diff --git a/CMakeLists.txt b/CMakeLists.txt index 588526c28614f753f6ff70e898f13f3fd65318db..eb2b1cceb48ec05e93bf20174e00169ca132a58a 100755 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -13,7 +13,6 @@ ENDIF () SET(TD_ACCOUNT FALSE) SET(TD_ADMIN FALSE) SET(TD_GRANT FALSE) -SET(TD_SYNC TRUE) SET(TD_MQTT TRUE) SET(TD_TSDB_PLUGINS FALSE) diff --git a/Jenkinsfile b/Jenkinsfile index dc7836c3daacaa457f721f9278687b99770fc394..edbe11d428d0381a81f64bbc823b5b51a5edf199 100644 --- a/Jenkinsfile +++ b/Jenkinsfile @@ -14,10 +14,12 @@ pipeline { sh ''' date cd ${WKC} + git reset --hard git checkout develop git pull git submodule update cd ${WK} + git reset --hard git checkout develop git pull export TZ=Asia/Harbin @@ -39,11 +41,13 @@ pipeline { steps { sh ''' cd ${WKC} + git reset --hard git checkout develop git pull git submodule update cd ${WK} + git reset --hard git checkout develop git pull export TZ=Asia/Harbin @@ -65,11 +69,13 @@ pipeline { steps { sh ''' cd ${WKC} + git reset --hard git checkout develop git pull git submodule update cd ${WK} + git reset --hard git checkout develop git pull export TZ=Asia/Harbin @@ -108,11 +114,13 @@ pipeline { steps { sh ''' cd ${WKC} + git reset --hard git checkout develop git pull git submodule update cd ${WK} + git reset --hard git checkout develop git pull export TZ=Asia/Harbin @@ -167,7 +175,47 @@ pipeline { } } - + stage('arm64_build'){ + agent{label 'arm64'} + steps{ + sh ''' + cd ${WK} + git fetch + git checkout develop + git pull + cd ${WKC} + git fetch + git checkout develop + git pull + git submodule update + cd ${WKC}/packaging + ./release.sh -v cluster -c aarch64 -n 2.0.0.0 -m 2.0.0.0 + + ''' + } + } + stage('arm32_build'){ + agent{label 'arm32'} + steps{ + catchError(buildResult: 'SUCCESS', stageResult: 'FAILURE') { + sh ''' + cd ${WK} + git fetch + git checkout develop + git pull + cd ${WKC} + git fetch + git checkout develop + git pull + git submodule update + cd ${WKC}/packaging + ./release.sh -v cluster -c aarch32 -n 2.0.0.0 -m 2.0.0.0 + + ''' + } + + } + } } } diff --git a/cmake/define.inc b/cmake/define.inc index 6e64c2709abee366856a61916c6905363ea2002d..782dc625bf5f82b09637648063649b944a7747e3 100755 --- a/cmake/define.inc +++ b/cmake/define.inc @@ -13,10 +13,6 @@ IF (TD_GRANT) ADD_DEFINITIONS(-D_GRANT) ENDIF () -IF (TD_SYNC) - ADD_DEFINITIONS(-D_SYNC) -ENDIF () - IF (TD_MQTT) ADD_DEFINITIONS(-D_MQTT) ENDIF () diff --git a/cmake/input.inc b/cmake/input.inc index 1ef2045f57e5c9df94b6fe8284930711e8624a8a..e8324887a051ae6ac27eff6b748c824d1f0a3fa5 100755 --- a/cmake/input.inc +++ b/cmake/input.inc @@ -47,11 +47,6 @@ IF (${MQTT} MATCHES "false") MESSAGE(STATUS "build without mqtt module") ENDIF () -IF (${SYNC} MATCHES "false") - SET(TD_SYNC FALSE) - MESSAGE(STATUS "build without sync module") -ENDIF () - IF (${RANDOM_FILE_FAIL} MATCHES "true") SET(TD_RANDOM_FILE_FAIL TRUE) MESSAGE(STATUS "build with random-file-fail enabled") diff --git a/cmake/install.inc b/cmake/install.inc index 4b2d4828f8c2884752dbc12b29ce21ca7e0495a6..8418612d4c54d7f7643e190da0af81064197b5d1 100755 --- a/cmake/install.inc +++ b/cmake/install.inc @@ -32,7 +32,7 @@ ELSEIF (TD_WINDOWS) #INSTALL(TARGETS taos RUNTIME DESTINATION driver) #INSTALL(TARGETS shell RUNTIME DESTINATION .) IF (TD_MVN_INSTALLED) - INSTALL(FILES ${LIBRARY_OUTPUT_PATH}/taos-jdbcdriver-2.0.13-dist.jar DESTINATION connector/jdbc) + INSTALL(FILES ${LIBRARY_OUTPUT_PATH}/taos-jdbcdriver-2.0.14-dist.jar DESTINATION connector/jdbc) ENDIF () ELSEIF (TD_DARWIN) SET(TD_MAKE_INSTALL_SH "${TD_COMMUNITY_DIR}/packaging/tools/make_install.sh") diff --git a/cmake/version.inc b/cmake/version.inc index 2f0ec81aea2dcc932845537bcb37f5ab64d9bcb8..948c7d2d0b7181cc29d91884166c844f39d8bf6b 100644 --- a/cmake/version.inc +++ b/cmake/version.inc @@ -4,7 +4,7 @@ PROJECT(TDengine) IF (DEFINED VERNUMBER) SET(TD_VER_NUMBER ${VERNUMBER}) ELSE () - SET(TD_VER_NUMBER "2.0.8.0") + SET(TD_VER_NUMBER "2.0.9.0") ENDIF () IF (DEFINED VERCOMPATIBLE) diff --git a/documentation20/webdocs/markdowndocs/Model-ch.md b/documentation20/webdocs/markdowndocs/Model-ch.md index d698e3daaf60c02a76da8470d11fc292b93df058..c0b64bde6e5b4ee8842931751cc64b726c991321 100644 --- a/documentation20/webdocs/markdowndocs/Model-ch.md +++ b/documentation20/webdocs/markdowndocs/Model-ch.md @@ -12,7 +12,7 @@ TDengine采用关系型数据模型,需要建库、建表。因此对于一个 CREATE DATABASE power KEEP 365 DAYS 10 BLOCKS 4; ``` 上述语句将创建一个名为power的库,这个库的数据将保留365天(超过365天将被自动删除),每10天一个数据文件,内存块数为4。详细的语法及参数请见TAOS SQL - + 创建库之后,需要使用SQL命令USE将当前库切换过来,例如: ```cmd @@ -20,7 +20,7 @@ USE power; ``` 就当前连接里操作的库换为power,否则对具体表操作前,需要使用“库名.表名”来指定库的名字。 - + **注意:** - 任何一张表或超级表是属于一个库的,在创建表之前,必须先创建库。 @@ -44,15 +44,17 @@ CREATE TABLE d1001 USING meters TAGS ("Beijing.Chaoyang", 2); ``` 其中d1001是表名,meters是超级表的表名,后面紧跟标签Location的具体标签值”Beijing.Chaoyang",标签groupId的具体标签值2。虽然在创建表时,需要指定标签值,但可以事后修改。详细细则请见 TAOS SQL。 +**注意:**目前 TDengine 没有从技术层面限制使用一个 database (dbA)的超级表作为模板建立另一个 database (dbB)的子表,后续会禁止这种用法,不建议使用这种方法建表。 + TDengine建议将数据采集点的全局唯一ID作为表名(比如设备序列号)。但对于有的场景,并没有唯一的ID,可以将多个ID组合成一个唯一的ID。不建议将具有唯一性的ID作为标签值。 - + **自动建表**:在某些特殊场景中,用户在写数据时并不确定某个数据采集点的表是否存在,此时可在写入数据时使用自动建表语法来创建不存在的表,若该表已存在则不会建立新表。比如: ```cmd INSERT INTO d1001 USING METERS TAGS ("Beijng.Chaoyang", 2) VALUES (now, 10.2, 219, 0.32); ``` 上述SQL语句将记录(now, 10.2, 219, 0.32) 插入进表d1001。如果表d1001还未创建,则使用超级表meters做模板自动创建,同时打上标签值“Beijing.Chaoyang", 2。 - + ## 多列模型 vs 单列模型 TDengine支持多列模型,只要物理量是一个数据采集点同时采集的(时间戳一致),这些量就可以作为不同列放在一张超级表里。但还有一种极限的设计,单列模型,每个采集的物理量都单独建表,因此每种类型的物理量都单独建立一超级表。比如电流、电压、相位,就建三张超级表。 diff --git a/documentation20/webdocs/markdowndocs/administrator-ch.md b/documentation20/webdocs/markdowndocs/administrator-ch.md index 36466d2b7ea29307b5c17ad21b0db47a98598fa4..f54c6b91a1f6b171676176615f5377dfdbe8ff95 100644 --- a/documentation20/webdocs/markdowndocs/administrator-ch.md +++ b/documentation20/webdocs/markdowndocs/administrator-ch.md @@ -95,6 +95,7 @@ TDengine系统后台服务由taosd提供,可以在配置文件taos.cfg里修 - logKeepDays:日志文件的最长保存时间。大于0时,日志文件会被重命名为taosdlog.xxx,其中xxx为日志文件最后修改的时间戳,单位为秒。默认值:0天。 - maxSQLLength:单条SQL语句允许最长限制。默认值:65380字节。 - telemetryReporting: 是否允许 TDengine 采集和上报基本使用信息,0表示不允许,1表示允许。 默认值:1。 +- stream: 是否启用连续查询(流计算功能),0表示不允许,1表示允许。 默认值:1。 **注意:**对于端口,TDengine会使用从serverPort起13个连续的TCP和UDP端口号,请务必在防火墙打开。因此如果是缺省配置,需要打开从6030都6042共13个端口,而且必须TCP和UDP都打开。 diff --git a/documentation20/webdocs/markdowndocs/cluster-ch.md b/documentation20/webdocs/markdowndocs/cluster-ch.md index 10c28c284c870d1a522a178beb5303901f079a5a..db479417c5a22acf26bc7a829083cd21fd7dffb1 100644 --- a/documentation20/webdocs/markdowndocs/cluster-ch.md +++ b/documentation20/webdocs/markdowndocs/cluster-ch.md @@ -37,7 +37,7 @@ fqdn h1.taosdata.com // 配置本数据节点的端口号,缺省是6030 serverPort 6030 -// 服务端节点数为偶数的时候,需要配置,请参考《Arbitrator的使用》的部分 +// 使用场景,请参考《Arbitrator的使用》的部分 arbitrator ha.taosdata.com:6042 ``` diff --git a/documentation20/webdocs/markdowndocs/connector-ch.md b/documentation20/webdocs/markdowndocs/connector-ch.md index c5a955f43fcc30e70d8aac3919433c2e767f76ba..69c560bbc48a49ac040c964baa0d0f20ef9ca879 100644 --- a/documentation20/webdocs/markdowndocs/connector-ch.md +++ b/documentation20/webdocs/markdowndocs/connector-ch.md @@ -142,7 +142,7 @@ C/C++的API类似于MySQL的C API。应用程序使用时,需要包含TDengine 获取最近一次API调用失败的原因,返回值为错误代码。 -**注意**:对于单个数据库连接,在同一时刻只能有一个线程使用该连接调用API,否则会有未定义的行为出现并可能导致客户端crash。客户端应用可以通过建立多个连接进行多线程的数据写入或查询处理。 +**注意**:对于每个数据库应用,2.0及以上版本 TDengine 推荐只建立一个连接。同时在应用中将该连接 (TAOS*) 结构体传递到不同的线程共享使用。基于 TAOS 结构体发出的查询、写入等操作具有多线程安全性。C 语言的连接器可以按照需求动态建立面向数据库的新连接(该过程对用户不可见),同时建议只有在程序最后退出的时候才调用 taos_close 关闭连接。 ### 异步查询API diff --git a/documentation20/webdocs/markdowndocs/faq-ch.md b/documentation20/webdocs/markdowndocs/faq-ch.md index 757b6d9929b68a90e7bc558b553233f24b09ba17..61f70c5962b4f83d720dfafd0e3e1759dad8dd25 100644 --- a/documentation20/webdocs/markdowndocs/faq-ch.md +++ b/documentation20/webdocs/markdowndocs/faq-ch.md @@ -36,16 +36,20 @@ 5. ping服务器FQDN,如果没有反应,请检查你的网络,DNS设置,或客户端所在计算机的系统hosts文件 -6. 检查防火墙设置,确认TCP/UDP 端口6030-6042 是打开的 +6. 检查防火墙设置(Ubuntu 使用 ufw status,CentOS 使用 firewall-cmd --list-port),确认TCP/UDP 端口6030-6042 是打开的 7. 对于Linux上的JDBC(ODBC, Python, Go等接口类似)连接, 确保*libtaos.so*在目录*/usr/local/taos/driver*里, 并且*/usr/local/taos/driver*在系统库函数搜索路径*LD_LIBRARY_PATH*里 8. 对于windows上的JDBC, ODBC, Python, Go等连接,确保*C:\TDengine\driver\taos.dll*在你的系统库函数搜索目录里 (建议*taos.dll*放在目录 *C:\Windows\System32*) -9. 如果仍不能排除连接故障,请使用命令行工具nc来分别判断指定端口的TCP和UDP连接是否通畅 - 检查UDP端口连接是否工作:`nc -vuz {hostIP} {port} ` - 检查服务器侧TCP端口连接是否工作:`nc -l {port}` - 检查客户端侧TCP端口连接是否工作:`nc {hostIP} {port}` +9. 如果仍不能排除连接故障 + + * Linux 系统请使用命令行工具nc来分别判断指定端口的TCP和UDP连接是否通畅 + 检查UDP端口连接是否工作:`nc -vuz {hostIP} {port} ` + 检查服务器侧TCP端口连接是否工作:`nc -l {port}` + 检查客户端侧TCP端口连接是否工作:`nc {hostIP} {port}` + + * Windows 系统请使用 PowerShell 命令 Net-TestConnection -ComputerName {fqdn} -Port {port} 检测服务段端口是否访问 10. 也可以使用taos程序内嵌的网络连通检测功能,来验证服务器和客户端之间指定的端口连接是否通畅(包括TCP和UDP):[TDengine 内嵌网络检测工具使用指南](https://www.taosdata.com/blog/2020/09/08/1816.html)。 @@ -101,7 +105,20 @@ Connection = DriverManager.getConnection(url, properties); 2.0.4 ``` -## 14. 怎么报告问题? +## 14. taos connect failed, reason: invalid timestamp + +常见原因是服务器和客户端时间没有校准,可以通过和时间服务器同步的方式(Linux 下使用 ntpdate 命令,Windows 在系统时间设置中选择自动同步)校准。 + + + +## 15. 表名显示不全 + +由于 taos shell 在终端中显示宽度有限,有可能比较长的表名显示不全,如果按照显示的不全的表名进行相关操作会发生 Table does not exist 错误。解决方法可以是通过修改 taos.cfg 文件中的设置项 maxBinaryDisplayWidth, 或者直接输入命令 set max_binary_display_width 100。或者在命令结尾使用 \G 参数来调整结果的显示方式。 + + + +## 16. 怎么报告问题? + 如果 FAQ 中的信息不能够帮到您,需要 TDengine 技术团队的技术支持与协助,请将以下两个目录中内容打包: 1. /var/log/taos 2. /etc/taos diff --git a/packaging/cfg/taos.cfg b/packaging/cfg/taos.cfg index ca88bca3c863dcfd4a95497fc572499dc93f31f9..8a68d02dfd232e557d0877e738aa8aa33a02467f 100644 --- a/packaging/cfg/taos.cfg +++ b/packaging/cfg/taos.cfg @@ -236,7 +236,7 @@ # httpDebugFlag 131 # debug flag for monitor -# monitorDebugFlag 131 +# monDebugFlag 131 # debug flag for query # qDebugflag 131 @@ -251,7 +251,7 @@ # cqDebugFlag 131 # enable/disable recording the SQL in taos client -# tscEnableRecordSql 0 +# enableRecordSql 0 # generate core file when service crash # enableCoreFile 1 @@ -260,4 +260,10 @@ # maxBinaryDisplayWidth 30 # enable/disable telemetry reporting -# telemetryReporting 1 \ No newline at end of file +# telemetryReporting 1 + +# enable/disable stream (continuous query) +# stream 1 + +# only 50% CPU resources will be used in query processing +# halfCoresForQuery 0 diff --git a/packaging/tools/install.sh b/packaging/tools/install.sh index ddf7114f08387c2277df47e9174948e479d432ba..d6dccf7045a02a3019d95d73685a3ca27bc5205f 100755 --- a/packaging/tools/install.sh +++ b/packaging/tools/install.sh @@ -172,6 +172,7 @@ function install_bin() { ${csudo} rm -f ${bin_link_dir}/taos || : ${csudo} rm -f ${bin_link_dir}/taosd || : ${csudo} rm -f ${bin_link_dir}/taosdemo || : + ${csudo} rm -f ${bin_link_dir}/taosdump || : ${csudo} rm -f ${bin_link_dir}/rmtaos || : ${csudo} rm -f ${bin_link_dir}/tarbitrator || : ${csudo} rm -f ${bin_link_dir}/set_core || : @@ -182,6 +183,7 @@ function install_bin() { [ -x ${install_main_dir}/bin/taos ] && ${csudo} ln -s ${install_main_dir}/bin/taos ${bin_link_dir}/taos || : [ -x ${install_main_dir}/bin/taosd ] && ${csudo} ln -s ${install_main_dir}/bin/taosd ${bin_link_dir}/taosd || : [ -x ${install_main_dir}/bin/taosdemo ] && ${csudo} ln -s ${install_main_dir}/bin/taosdemo ${bin_link_dir}/taosdemo || : + [ -x ${install_main_dir}/bin/taosdump ] && ${csudo} ln -s ${install_main_dir}/bin/taosdump ${bin_link_dir}/taosdump || : [ -x ${install_main_dir}/bin/remove.sh ] && ${csudo} ln -s ${install_main_dir}/bin/remove.sh ${bin_link_dir}/rmtaos || : [ -x ${install_main_dir}/bin/set_core.sh ] && ${csudo} ln -s ${install_main_dir}/bin/set_core.sh ${bin_link_dir}/set_core || : [ -x ${install_main_dir}/bin/tarbitrator ] && ${csudo} ln -s ${install_main_dir}/bin/tarbitrator ${bin_link_dir}/tarbitrator || : diff --git a/packaging/tools/install_client.sh b/packaging/tools/install_client.sh index 34a9bfaecb4d0366008ea9b78a60434297ea9f51..0467300953ca34d1f9b24ba9999ccd156a0d8d09 100755 --- a/packaging/tools/install_client.sh +++ b/packaging/tools/install_client.sh @@ -84,8 +84,9 @@ function install_main_path() { function install_bin() { # Remove links ${csudo} rm -f ${bin_link_dir}/taos || : - if [ "$osType" == "Darwin" ]; then + if [ "$osType" != "Darwin" ]; then ${csudo} rm -f ${bin_link_dir}/taosdemo || : + ${csudo} rm -f ${bin_link_dir}/taosdump || : fi ${csudo} rm -f ${bin_link_dir}/rmtaos || : ${csudo} rm -f ${bin_link_dir}/set_core || : @@ -94,8 +95,9 @@ function install_bin() { #Make link [ -x ${install_main_dir}/bin/taos ] && ${csudo} ln -s ${install_main_dir}/bin/taos ${bin_link_dir}/taos || : - if [ "$osType" == "Darwin" ]; then + if [ "$osType" != "Darwin" ]; then [ -x ${install_main_dir}/bin/taosdemo ] && ${csudo} ln -s ${install_main_dir}/bin/taosdemo ${bin_link_dir}/taosdemo || : + [ -x ${install_main_dir}/bin/taosdump ] && ${csudo} ln -s ${install_main_dir}/bin/taosdump ${bin_link_dir}/taosdump || : fi [ -x ${install_main_dir}/bin/remove_client.sh ] && ${csudo} ln -s ${install_main_dir}/bin/remove_client.sh ${bin_link_dir}/rmtaos || : [ -x ${install_main_dir}/bin/set_core.sh ] && ${csudo} ln -s ${install_main_dir}/bin/set_core.sh ${bin_link_dir}/set_core || : diff --git a/packaging/tools/install_client_power.sh b/packaging/tools/install_client_power.sh index 0108d1d44a186043aa5f83e01f44fb1540f6e0fd..26977e12f4b7ee780948399605f28603817f1aa2 100755 --- a/packaging/tools/install_client_power.sh +++ b/packaging/tools/install_client_power.sh @@ -84,8 +84,9 @@ function install_main_path() { function install_bin() { # Remove links ${csudo} rm -f ${bin_link_dir}/power || : - if [ "$osType" == "Darwin" ]; then + if [ "$osType" != "Darwin" ]; then ${csudo} rm -f ${bin_link_dir}/powerdemo || : + ${csudo} rm -f ${bin_link_dir}/powerdump || : fi ${csudo} rm -f ${bin_link_dir}/rmpower || : ${csudo} rm -f ${bin_link_dir}/set_core || : @@ -94,8 +95,9 @@ function install_bin() { #Make link [ -x ${install_main_dir}/bin/power ] && ${csudo} ln -s ${install_main_dir}/bin/power ${bin_link_dir}/power || : - if [ "$osType" == "Darwin" ]; then + if [ "$osType" != "Darwin" ]; then [ -x ${install_main_dir}/bin/powerdemo ] && ${csudo} ln -s ${install_main_dir}/bin/powerdemo ${bin_link_dir}/powerdemo || : + [ -x ${install_main_dir}/bin/powerdump ] && ${csudo} ln -s ${install_main_dir}/bin/powerdump ${bin_link_dir}/powerdump || : fi [ -x ${install_main_dir}/bin/remove_client_power.sh ] && ${csudo} ln -s ${install_main_dir}/bin/remove_client_power.sh ${bin_link_dir}/rmpower || : [ -x ${install_main_dir}/bin/set_core.sh ] && ${csudo} ln -s ${install_main_dir}/bin/set_core.sh ${bin_link_dir}/set_core || : diff --git a/packaging/tools/install_power.sh b/packaging/tools/install_power.sh index df6291f4ae54f6706f1a74e9a33f5b1bb361edcd..5929b52afca59f5e4364aa294fe7cefec5f6b828 100755 --- a/packaging/tools/install_power.sh +++ b/packaging/tools/install_power.sh @@ -172,6 +172,7 @@ function install_bin() { ${csudo} rm -f ${bin_link_dir}/power || : ${csudo} rm -f ${bin_link_dir}/powerd || : ${csudo} rm -f ${bin_link_dir}/powerdemo || : + ${csudo} rm -f ${bin_link_dir}/powerdump || : ${csudo} rm -f ${bin_link_dir}/rmpower || : ${csudo} rm -f ${bin_link_dir}/tarbitrator || : ${csudo} rm -f ${bin_link_dir}/set_core || : @@ -182,6 +183,7 @@ function install_bin() { [ -x ${install_main_dir}/bin/power ] && ${csudo} ln -s ${install_main_dir}/bin/power ${bin_link_dir}/power || : [ -x ${install_main_dir}/bin/powerd ] && ${csudo} ln -s ${install_main_dir}/bin/powerd ${bin_link_dir}/powerd || : [ -x ${install_main_dir}/bin/powerdemo ] && ${csudo} ln -s ${install_main_dir}/bin/powerdemo ${bin_link_dir}/powerdemo || : + [ -x ${install_main_dir}/bin/powerdump ] && ${csudo} ln -s ${install_main_dir}/bin/powerdump ${bin_link_dir}/powerdump || : [ -x ${install_main_dir}/bin/remove_power.sh ] && ${csudo} ln -s ${install_main_dir}/bin/remove_power.sh ${bin_link_dir}/rmpower || : [ -x ${install_main_dir}/bin/set_core.sh ] && ${csudo} ln -s ${install_main_dir}/bin/set_core.sh ${bin_link_dir}/set_core || : [ -x ${install_main_dir}/bin/tarbitrator ] && ${csudo} ln -s ${install_main_dir}/bin/tarbitrator ${bin_link_dir}/tarbitrator || : diff --git a/packaging/tools/post.sh b/packaging/tools/post.sh index 00705fad778c065eddbd7cb65a2f9c5583a9997b..52919976ee8a1b8c3ab7c38d9dad4eb7e98b22bd 100755 --- a/packaging/tools/post.sh +++ b/packaging/tools/post.sh @@ -92,6 +92,7 @@ function install_bin() { ${csudo} rm -f ${bin_link_dir}/taos || : ${csudo} rm -f ${bin_link_dir}/taosd || : ${csudo} rm -f ${bin_link_dir}/taosdemo || : + ${csudo} rm -f ${bin_link_dir}/taosdump || : ${csudo} rm -f ${bin_link_dir}/rmtaos || : ${csudo} rm -f ${bin_link_dir}/set_core || : @@ -101,6 +102,7 @@ function install_bin() { [ -x ${bin_dir}/taos ] && ${csudo} ln -s ${bin_dir}/taos ${bin_link_dir}/taos || : [ -x ${bin_dir}/taosd ] && ${csudo} ln -s ${bin_dir}/taosd ${bin_link_dir}/taosd || : [ -x ${bin_dir}/taosdemo ] && ${csudo} ln -s ${bin_dir}/taosdemo ${bin_link_dir}/taosdemo || : + [ -x ${bin_dir}/taosdump ] && ${csudo} ln -s ${bin_dir}/taosdump ${bin_link_dir}/taosdump || : [ -x ${bin_dir}/set_core.sh ] && ${csudo} ln -s ${bin_dir}/set_core.sh ${bin_link_dir}/set_core || : } diff --git a/packaging/tools/remove.sh b/packaging/tools/remove.sh index e9a4f48cf78469ac3b5d3a46cc058430d6da4d56..2f2660d44635c86df3b51d2b86e37b3399869a88 100755 --- a/packaging/tools/remove.sh +++ b/packaging/tools/remove.sh @@ -72,6 +72,7 @@ function clean_bin() { ${csudo} rm -f ${bin_link_dir}/taos || : ${csudo} rm -f ${bin_link_dir}/taosd || : ${csudo} rm -f ${bin_link_dir}/taosdemo || : + ${csudo} rm -f ${bin_link_dir}/taosdump || : ${csudo} rm -f ${bin_link_dir}/rmtaos || : ${csudo} rm -f ${bin_link_dir}/tarbitrator || : ${csudo} rm -f ${bin_link_dir}/set_core || : @@ -222,4 +223,4 @@ elif echo $osinfo | grep -qwi "centos" ; then fi echo -e "${GREEN}TDengine is removed successfully!${NC}" -echo \ No newline at end of file +echo diff --git a/packaging/tools/remove_client.sh b/packaging/tools/remove_client.sh index 2c28b7b6bf41b3798d533b8835a5ff311eecfdb4..7579162dc60e290754e71ed6a71c10cfaee5537b 100755 --- a/packaging/tools/remove_client.sh +++ b/packaging/tools/remove_client.sh @@ -38,6 +38,7 @@ function clean_bin() { # Remove link ${csudo} rm -f ${bin_link_dir}/taos || : ${csudo} rm -f ${bin_link_dir}/taosdemo || : + ${csudo} rm -f ${bin_link_dir}/taosdump || : ${csudo} rm -f ${bin_link_dir}/rmtaos || : ${csudo} rm -f ${bin_link_dir}/set_core || : } diff --git a/packaging/tools/remove_client_power.sh b/packaging/tools/remove_client_power.sh index 7a3c99e100c7b0e363bdf15bf19b8b523124c45b..580c46e2077d7f21e06d4d4a8f69dcd5b6bbf51d 100755 --- a/packaging/tools/remove_client_power.sh +++ b/packaging/tools/remove_client_power.sh @@ -38,6 +38,7 @@ function clean_bin() { # Remove link ${csudo} rm -f ${bin_link_dir}/power || : ${csudo} rm -f ${bin_link_dir}/powerdemo || : + ${csudo} rm -f ${bin_link_dir}/powerdump || : ${csudo} rm -f ${bin_link_dir}/rmpower || : ${csudo} rm -f ${bin_link_dir}/set_core || : } diff --git a/packaging/tools/remove_power.sh b/packaging/tools/remove_power.sh index d6d6c5dd7c7744363add2cafb23ea0178533ab8d..816869cf444d8001e0c0aae30840d2c40a9e6af4 100755 --- a/packaging/tools/remove_power.sh +++ b/packaging/tools/remove_power.sh @@ -72,6 +72,7 @@ function clean_bin() { ${csudo} rm -f ${bin_link_dir}/power || : ${csudo} rm -f ${bin_link_dir}/powerd || : ${csudo} rm -f ${bin_link_dir}/powerdemo || : + ${csudo} rm -f ${bin_link_dir}/powerdump || : ${csudo} rm -f ${bin_link_dir}/rmpower || : ${csudo} rm -f ${bin_link_dir}/tarbitrator || : ${csudo} rm -f ${bin_link_dir}/set_core || : @@ -223,4 +224,4 @@ fi #fi echo -e "${GREEN}PowerDB is removed successfully!${NC}" -echo \ No newline at end of file +echo diff --git a/snap/snapcraft.yaml b/snap/snapcraft.yaml index 059c0650c2bbba99694253a73b59d77cd16f06bf..b5d06a4adb2acc57e7220dfde72290fec0eec0fb 100644 --- a/snap/snapcraft.yaml +++ b/snap/snapcraft.yaml @@ -1,6 +1,6 @@ name: tdengine base: core18 -version: '2.0.8.0' +version: '2.0.9.0' icon: snap/gui/t-dengine.svg summary: an open-source big data platform designed and optimized for IoT. description: | @@ -72,7 +72,7 @@ parts: - usr/bin/taosd - usr/bin/taos - usr/bin/taosdemo - - usr/lib/libtaos.so.2.0.8.0 + - usr/lib/libtaos.so.2.0.9.0 - usr/lib/libtaos.so.1 - usr/lib/libtaos.so diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index a2600785c35a80d040b9f91137c630d7761cef7e..931a0a132e86b94cb78bf6df553ffdf8e9920abb 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -10,9 +10,7 @@ ADD_SUBDIRECTORY(client) ADD_SUBDIRECTORY(query) ADD_SUBDIRECTORY(kit) ADD_SUBDIRECTORY(plugins) -IF (TD_SYNC) - ADD_SUBDIRECTORY(sync) -ENDIF () +ADD_SUBDIRECTORY(sync) ADD_SUBDIRECTORY(balance) ADD_SUBDIRECTORY(mnode) ADD_SUBDIRECTORY(vnode) diff --git a/src/balance/inc/bnInt.h b/src/balance/inc/bnInt.h new file mode 100644 index 0000000000000000000000000000000000000000..e924776ff151a9e81fe1254cdd1f1def04e6fef3 --- /dev/null +++ b/src/balance/inc/bnInt.h @@ -0,0 +1,57 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_BALANCE_INT_H +#define TDENGINE_BALANCE_INT_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "mnodeInt.h" +#include "mnodeDef.h" +#include "mnodeDnode.h" + +typedef struct { + int32_t size; + int32_t maxSize; + SDnodeObj **list; +} SBnDnodes; + +typedef struct { + void * timer; + bool stop; + pthread_mutex_t mutex; + pthread_cond_t cond; + pthread_t thread; +} SBnThread; + +typedef struct { + pthread_mutex_t mutex; +} SBnMgmt; + +int32_t bnInit(); +void bnCleanUp(); +bool bnStart(); +void bnCheckStatus(); +void bnCheckModules(); + +extern SBnDnodes tsBnDnodes; +extern void *tsMnodeTmr; + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/balance/inc/bnScore.h b/src/balance/inc/bnScore.h new file mode 100644 index 0000000000000000000000000000000000000000..a28c4459dd444001674305e404bfb0d23ed2fc3b --- /dev/null +++ b/src/balance/inc/bnScore.h @@ -0,0 +1,34 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_BALANCE_SCORE_H +#define TDENGINE_BALANCE_SCORE_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "bnInt.h" + +void bnInitDnodes(); +void bnCleanupDnodes(); +void bnAccquireDnodes(); +void bnReleaseDnodes(); +float bnTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extraVnode); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/balance/inc/bnThread.h b/src/balance/inc/bnThread.h new file mode 100644 index 0000000000000000000000000000000000000000..8f54b66028c5ddc859861bb78e0d790f7d3e5507 --- /dev/null +++ b/src/balance/inc/bnThread.h @@ -0,0 +1,33 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_BALANCE_THREAD_H +#define TDENGINE_BALANCE_THREAD_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "bnInt.h" + +int32_t bnInitThread(); +void bnCleanupThread(); +void bnNotify(); +void bnStartTimer(int64_t mseconds); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/balance/src/balance.c b/src/balance/src/bnMain.c similarity index 50% rename from src/balance/src/balance.c rename to src/balance/src/bnMain.c index df78f4fe270d34d17e436e80322f6c81b68d0d2c..d80488fe9fa35df7af217b26cb93c85f9b11192a 100644 --- a/src/balance/src/balance.c +++ b/src/balance/src/bnMain.c @@ -15,17 +15,13 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "tutil.h" -#include "tbalance.h" +#include "tref.h" #include "tsync.h" -#include "ttimer.h" #include "tglobal.h" -#include "tdataformat.h" #include "dnode.h" -#include "mnode.h" -#include "mnodeDef.h" -#include "mnodeInt.h" -#include "mnodeDnode.h" +#include "bnInt.h" +#include "bnScore.h" +#include "bnThread.h" #include "mnodeDb.h" #include "mnodeMnode.h" #include "mnodeSdb.h" @@ -33,36 +29,20 @@ #include "mnodeUser.h" #include "mnodeVgroup.h" -/* - * once sdb work as mater, then tsAccessSquence reset to zero - * increase tsAccessSquence every balance interval - */ -extern void * tsMnodeTmr; -static void * tsBalanceTimer = NULL; -static int32_t tsBalanceDnodeListSize = 0; -static SDnodeObj ** tsBalanceDnodeList = NULL; -static int32_t tsBalanceDnodeListMallocSize = 16; -static pthread_mutex_t tsBalanceMutex; - -static void balanceStartTimer(int64_t mseconds); -static void balanceInitDnodeList(); -static void balanceCleanupDnodeList(); -static void balanceAccquireDnodeList(); -static void balanceReleaseDnodeList(); -static void balanceMonitorDnodeModule(); -static float balanceTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extraVnode); -static int32_t balanceGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); -static int32_t balanceRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn); - -static void balanceLock() { - pthread_mutex_lock(&tsBalanceMutex); +extern int64_t tsDnodeRid; +extern int64_t tsSdbRid; +static SBnMgmt tsBnMgmt; +static void bnMonitorDnodeModule(); + +static void bnLock() { + pthread_mutex_lock(&tsBnMgmt.mutex); } -static void balanceUnLock() { - pthread_mutex_unlock(&tsBalanceMutex); +static void bnUnLock() { + pthread_mutex_unlock(&tsBnMgmt.mutex); } -static bool balanceCheckFree(SDnodeObj *pDnode) { +static bool bnCheckFree(SDnodeObj *pDnode) { if (pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) { mError("dnode:%d, status:%s not available", pDnode->dnodeId, mnodeGetDnodeStatusStr(pDnode->status)); return false; @@ -86,7 +66,7 @@ static bool balanceCheckFree(SDnodeObj *pDnode) { return true; } -static void balanceDiscardVnode(SVgObj *pVgroup, SVnodeGid *pVnodeGid) { +static void bnDiscardVnode(SVgObj *pVgroup, SVnodeGid *pVnodeGid) { mDebug("vgId:%d, dnode:%d is dropping", pVgroup->vgId, pVnodeGid->dnodeId); SDnodeObj *pDnode = mnodeGetDnode(pVnodeGid->dnodeId); @@ -111,27 +91,26 @@ static void balanceDiscardVnode(SVgObj *pVgroup, SVnodeGid *pVnodeGid) { mnodeUpdateVgroup(pVgroup); } -static void balanceSwapVnodeGid(SVnodeGid *pVnodeGid1, SVnodeGid *pVnodeGid2) { +static void bnSwapVnodeGid(SVnodeGid *pVnodeGid1, SVnodeGid *pVnodeGid2) { // SVnodeGid tmp = *pVnodeGid1; // *pVnodeGid1 = *pVnodeGid2; // *pVnodeGid2 = tmp; } -int32_t balanceAllocVnodes(SVgObj *pVgroup) { +int32_t bnAllocVnodes(SVgObj *pVgroup) { static int32_t randIndex = 0; int32_t dnode = 0; int32_t vnodes = 0; - balanceLock(); - - balanceAccquireDnodeList(); + bnLock(); + bnAccquireDnodes(); mDebug("db:%s, try alloc %d vnodes to vgroup, dnodes total:%d, avail:%d", pVgroup->dbName, pVgroup->numOfVnodes, - mnodeGetDnodesNum(), tsBalanceDnodeListSize); + mnodeGetDnodesNum(), tsBnDnodes.size); for (int32_t i = 0; i < pVgroup->numOfVnodes; ++i) { - for (; dnode < tsBalanceDnodeListSize; ++dnode) { - SDnodeObj *pDnode = tsBalanceDnodeList[dnode]; - if (balanceCheckFree(pDnode)) { + for (; dnode < tsBnDnodes.size; ++dnode) { + SDnodeObj *pDnode = tsBnDnodes.list[dnode]; + if (bnCheckFree(pDnode)) { SVnodeGid *pVnodeGid = pVgroup->vnodeGid + i; pVnodeGid->dnodeId = pDnode->dnodeId; pVnodeGid->pDnode = pDnode; @@ -148,8 +127,8 @@ int32_t balanceAllocVnodes(SVgObj *pVgroup) { } if (vnodes != pVgroup->numOfVnodes) { - balanceReleaseDnodeList(); - balanceUnLock(); + bnReleaseDnodes(); + bnUnLock(); mDebug("db:%s, need vnodes:%d, but alloc:%d", pVgroup->dbName, pVgroup->numOfVnodes, vnodes); @@ -162,7 +141,6 @@ int32_t balanceAllocVnodes(SVgObj *pVgroup) { pDnode->openVnodes, pDnode->diskAvailable, pDnode->alternativeRole); mnodeDecDnodeRef(pDnode); } - sdbFreeIter(pIter); if (mnodeGetOnlineDnodesNum() == 0) { return TSDB_CODE_MND_NOT_READY; @@ -180,33 +158,33 @@ int32_t balanceAllocVnodes(SVgObj *pVgroup) { if (pVgroup->numOfVnodes == 1) { } else if (pVgroup->numOfVnodes == 2) { if (randIndex++ % 2 == 0) { - balanceSwapVnodeGid(pVgroup->vnodeGid, pVgroup->vnodeGid + 1); + bnSwapVnodeGid(pVgroup->vnodeGid, pVgroup->vnodeGid + 1); } } else { int32_t randVal = randIndex++ % 6; if (randVal == 1) { // 1, 0, 2 - balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1); + bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1); } else if (randVal == 2) { // 1, 2, 0 - balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1); - balanceSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2); + bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 1); + bnSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2); } else if (randVal == 3) { // 2, 1, 0 - balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2); + bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2); } else if (randVal == 4) { // 2, 0, 1 - balanceSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2); - balanceSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2); + bnSwapVnodeGid(pVgroup->vnodeGid + 0, pVgroup->vnodeGid + 2); + bnSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2); } if (randVal == 5) { // 0, 2, 1 - balanceSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2); + bnSwapVnodeGid(pVgroup->vnodeGid + 1, pVgroup->vnodeGid + 2); } else { } // 0, 1, 2 } - balanceReleaseDnodeList(); - balanceUnLock(); + bnReleaseDnodes(); + bnUnLock(); return TSDB_CODE_SUCCESS; } -static bool balanceCheckVgroupReady(SVgObj *pVgroup, SVnodeGid *pRmVnode) { +static bool bnCheckVgroupReady(SVgObj *pVgroup, SVnodeGid *pRmVnode) { if (pVgroup->lbTime + 5 * tsStatusInterval > tsAccessSquence) { return false; } @@ -233,7 +211,7 @@ static bool balanceCheckVgroupReady(SVgObj *pVgroup, SVnodeGid *pRmVnode) { * desc: remove one vnode from vgroup * all vnodes in vgroup should in ready state, except the balancing one **/ -static int32_t balanceRemoveVnode(SVgObj *pVgroup) { +static int32_t bnRemoveVnode(SVgObj *pVgroup) { if (pVgroup->numOfVnodes <= 1) return -1; SVnodeGid *pRmVnode = NULL; @@ -275,17 +253,17 @@ static int32_t balanceRemoveVnode(SVgObj *pVgroup) { pSelVnode = pRmVnode; } - if (!balanceCheckVgroupReady(pVgroup, pSelVnode)) { + if (!bnCheckVgroupReady(pVgroup, pSelVnode)) { mDebug("vgId:%d, is not ready", pVgroup->vgId); return -1; } else { mDebug("vgId:%d, is ready, discard dnode:%d", pVgroup->vgId, pSelVnode->dnodeId); - balanceDiscardVnode(pVgroup, pSelVnode); + bnDiscardVnode(pVgroup, pSelVnode); return TSDB_CODE_SUCCESS; } } -static bool balanceCheckDnodeInVgroup(SDnodeObj *pDnode, SVgObj *pVgroup) { +static bool bnCheckDnodeInVgroup(SDnodeObj *pDnode, SVgObj *pVgroup) { for (int32_t i = 0; i < pVgroup->numOfVnodes; ++i) { SVnodeGid *pGid = &pVgroup->vnodeGid[i]; if (pGid->dnodeId == 0) break; @@ -300,13 +278,13 @@ static bool balanceCheckDnodeInVgroup(SDnodeObj *pDnode, SVgObj *pVgroup) { /** * desc: add vnode to vgroup, find a new one if dest dnode is null **/ -static int32_t balanceAddVnode(SVgObj *pVgroup, SDnodeObj *pSrcDnode, SDnodeObj *pDestDnode) { +static int32_t bnAddVnode(SVgObj *pVgroup, SDnodeObj *pSrcDnode, SDnodeObj *pDestDnode) { if (pDestDnode == NULL) { - for (int32_t i = 0; i < tsBalanceDnodeListSize; ++i) { - SDnodeObj *pDnode = tsBalanceDnodeList[i]; + for (int32_t i = 0; i < tsBnDnodes.size; ++i) { + SDnodeObj *pDnode = tsBnDnodes.list[i]; if (pDnode == pSrcDnode) continue; - if (balanceCheckDnodeInVgroup(pDnode, pVgroup)) continue; - if (!balanceCheckFree(pDnode)) continue; + if (bnCheckDnodeInVgroup(pDnode, pVgroup)) continue; + if (!bnCheckFree(pDnode)) continue; pDestDnode = pDnode; mDebug("vgId:%d, add vnode to dnode:%d", pVgroup->vgId, pDnode->dnodeId); @@ -334,25 +312,25 @@ static int32_t balanceAddVnode(SVgObj *pVgroup, SDnodeObj *pSrcDnode, SDnodeObj return TSDB_CODE_SUCCESS; } -static bool balanceMonitorBalance() { - if (tsBalanceDnodeListSize < 2) return false; +static bool bnMonitorBalance() { + if (tsBnDnodes.size < 2) return false; - for (int32_t src = tsBalanceDnodeListSize - 1; src >= 0; --src) { - SDnodeObj *pDnode = tsBalanceDnodeList[src]; - mDebug("%d-dnode:%d, state:%s, score:%.1f, numOfCores:%d, openVnodes:%d", tsBalanceDnodeListSize - src - 1, + for (int32_t src = tsBnDnodes.size - 1; src >= 0; --src) { + SDnodeObj *pDnode = tsBnDnodes.list[src]; + mDebug("%d-dnode:%d, state:%s, score:%.1f, numOfCores:%d, openVnodes:%d", tsBnDnodes.size - src - 1, pDnode->dnodeId, mnodeGetDnodeStatusStr(pDnode->status), pDnode->score, pDnode->numOfCores, pDnode->openVnodes); } - float scoresDiff = tsBalanceDnodeList[tsBalanceDnodeListSize - 1]->score - tsBalanceDnodeList[0]->score; + float scoresDiff = tsBnDnodes.list[tsBnDnodes.size - 1]->score - tsBnDnodes.list[0]->score; if (scoresDiff < 0.01) { - mDebug("all dnodes:%d is already balanced, scoresDiff:%f", tsBalanceDnodeListSize, scoresDiff); + mDebug("all dnodes:%d is already balanced, scoresDiff:%f", tsBnDnodes.size, scoresDiff); return false; } - for (int32_t src = tsBalanceDnodeListSize - 1; src > 0; --src) { - SDnodeObj *pSrcDnode = tsBalanceDnodeList[src]; - float srcScore = balanceTryCalcDnodeScore(pSrcDnode, -1); + for (int32_t src = tsBnDnodes.size - 1; src > 0; --src) { + SDnodeObj *pSrcDnode = tsBnDnodes.list[src]; + float srcScore = bnTryCalcDnodeScore(pSrcDnode, -1); if (tsEnableBalance == 0 && pSrcDnode->status != TAOS_DN_STATUS_DROPPING) { continue; } @@ -363,29 +341,27 @@ static bool balanceMonitorBalance() { pIter = mnodeGetNextVgroup(pIter, &pVgroup); if (pVgroup == NULL) break; - if (balanceCheckDnodeInVgroup(pSrcDnode, pVgroup)) { + if (bnCheckDnodeInVgroup(pSrcDnode, pVgroup)) { for (int32_t dest = 0; dest < src; dest++) { - SDnodeObj *pDestDnode = tsBalanceDnodeList[dest]; - if (balanceCheckDnodeInVgroup(pDestDnode, pVgroup)) continue; + SDnodeObj *pDestDnode = tsBnDnodes.list[dest]; + if (bnCheckDnodeInVgroup(pDestDnode, pVgroup)) continue; - float destScore = balanceTryCalcDnodeScore(pDestDnode, 1); + float destScore = bnTryCalcDnodeScore(pDestDnode, 1); if (srcScore + 0.0001 < destScore) continue; - if (!balanceCheckFree(pDestDnode)) continue; + if (!bnCheckFree(pDestDnode)) continue; mDebug("vgId:%d, balance from dnode:%d to dnode:%d, srcScore:%.1f:%.1f, destScore:%.1f:%.1f", pVgroup->vgId, pSrcDnode->dnodeId, pDestDnode->dnodeId, pSrcDnode->score, srcScore, pDestDnode->score, destScore); - balanceAddVnode(pVgroup, pSrcDnode, pDestDnode); + bnAddVnode(pVgroup, pSrcDnode, pDestDnode); mnodeDecVgroupRef(pVgroup); - sdbFreeIter(pIter); + mnodeCancelGetNextVgroup(pIter); return true; } } mnodeDecVgroupRef(pVgroup); } - - sdbFreeIter(pIter); } return false; @@ -395,7 +371,7 @@ static bool balanceMonitorBalance() { // 1. reset balanceAccessSquence to zero // 2. reset state of dnodes to offline // 3. reset lastAccess of dnodes to zero -void balanceReset() { +void bnReset() { void * pIter = NULL; SDnodeObj *pDnode = NULL; while (1) { @@ -413,12 +389,10 @@ void balanceReset() { mnodeDecDnodeRef(pDnode); } - sdbFreeIter(pIter); - tsAccessSquence = 0; } -static int32_t balanceMonitorVgroups() { +static int32_t bnMonitorVgroups() { void * pIter = NULL; SVgObj *pVgroup = NULL; bool hasUpdatingVgroup = false; @@ -434,25 +408,24 @@ static int32_t balanceMonitorVgroups() { if (vgReplica > dbReplica) { mInfo("vgId:%d, replica:%d numOfVnodes:%d, try remove one vnode", pVgroup->vgId, dbReplica, vgReplica); hasUpdatingVgroup = true; - code = balanceRemoveVnode(pVgroup); + code = bnRemoveVnode(pVgroup); } else if (vgReplica < dbReplica) { mInfo("vgId:%d, replica:%d numOfVnodes:%d, try add one vnode", pVgroup->vgId, dbReplica, vgReplica); hasUpdatingVgroup = true; - code = balanceAddVnode(pVgroup, NULL, NULL); + code = bnAddVnode(pVgroup, NULL, NULL); } mnodeDecVgroupRef(pVgroup); if (code == TSDB_CODE_SUCCESS) { + mnodeCancelGetNextVgroup(pIter); break; } } - sdbFreeIter(pIter); - return hasUpdatingVgroup; } -static bool balanceMonitorDnodeDropping(SDnodeObj *pDnode) { +static bool bnMonitorDnodeDropping(SDnodeObj *pDnode) { mDebug("dnode:%d, in dropping state", pDnode->dnodeId); void * pIter = NULL; @@ -462,14 +435,15 @@ static bool balanceMonitorDnodeDropping(SDnodeObj *pDnode) { pIter = mnodeGetNextVgroup(pIter, &pVgroup); if (pVgroup == NULL) break; - hasThisDnode = balanceCheckDnodeInVgroup(pDnode, pVgroup); + hasThisDnode = bnCheckDnodeInVgroup(pDnode, pVgroup); mnodeDecVgroupRef(pVgroup); - if (hasThisDnode) break; + if (hasThisDnode) { + mnodeCancelGetNextVgroup(pIter); + break; + } } - sdbFreeIter(pIter); - if (!hasThisDnode) { mInfo("dnode:%d, dropped for all vnodes are moving to other dnodes", pDnode->dnodeId); mnodeDropDnode(pDnode, NULL); @@ -479,7 +453,7 @@ static bool balanceMonitorDnodeDropping(SDnodeObj *pDnode) { return false; } -static bool balanceMontiorDropping() { +static bool bnMontiorDropping() { void *pIter = NULL; SDnodeObj *pDnode = NULL; @@ -499,50 +473,46 @@ static bool balanceMontiorDropping() { pDnode->status = TAOS_DN_STATUS_DROPPING; mnodeUpdateDnode(pDnode); mnodeDecDnodeRef(pDnode); - sdbFreeIter(pIter); + mnodeCancelGetNextDnode(pIter); return true; } if (pDnode->status == TAOS_DN_STATUS_DROPPING) { - bool ret = balanceMonitorDnodeDropping(pDnode); + bool ret = bnMonitorDnodeDropping(pDnode); mnodeDecDnodeRef(pDnode); - sdbFreeIter(pIter); + mnodeCancelGetNextDnode(pIter); return ret; } } - sdbFreeIter(pIter); - return false; } -static bool balanceStart() { +bool bnStart() { if (!sdbIsMaster()) return false; - balanceLock(); - - balanceAccquireDnodeList(); + bnLock(); + bnAccquireDnodes(); - balanceMonitorDnodeModule(); + bnMonitorDnodeModule(); - bool updateSoon = balanceMontiorDropping(); + bool updateSoon = bnMontiorDropping(); if (!updateSoon) { - updateSoon = balanceMonitorVgroups(); + updateSoon = bnMonitorVgroups(); } if (!updateSoon) { - updateSoon = balanceMonitorBalance(); + updateSoon = bnMonitorBalance(); } - balanceReleaseDnodeList(); - - balanceUnLock(); + bnReleaseDnodes(); + bnUnLock(); return updateSoon; } -static void balanceSetVgroupOffline(SDnodeObj* pDnode) { +static void bnSetVgroupOffline(SDnodeObj* pDnode) { void *pIter = NULL; while (1) { SVgObj *pVgroup; @@ -556,14 +526,15 @@ static void balanceSetVgroupOffline(SDnodeObj* pDnode) { } mnodeDecVgroupRef(pVgroup); } - - sdbFreeIter(pIter); } -static void balanceCheckDnodeAccess() { +void bnCheckStatus() { void * pIter = NULL; SDnodeObj *pDnode = NULL; + void *dnodeSdb = taosAcquireRef(tsSdbRid, tsDnodeRid); + if (dnodeSdb == NULL) return; + while (1) { pIter = mnodeGetNextDnode(pIter, &pDnode); if (pDnode == NULL) break; @@ -573,85 +544,41 @@ static void balanceCheckDnodeAccess() { pDnode->offlineReason = TAOS_DN_OFF_STATUS_MSG_TIMEOUT; mInfo("dnode:%d, set to offline state, access seq:%d last seq:%d laststat:%d", pDnode->dnodeId, tsAccessSquence, pDnode->lastAccess, pDnode->status); - balanceSetVgroupOffline(pDnode); + bnSetVgroupOffline(pDnode); } } mnodeDecDnodeRef(pDnode); } - sdbFreeIter(pIter); + taosReleaseRef(tsSdbRid, tsDnodeRid); } -static void balanceProcessBalanceTimer(void *handle, void *tmrId) { - if (!sdbIsMaster()) return; - - tsBalanceTimer = NULL; - tsAccessSquence ++; - - balanceCheckDnodeAccess(); - bool updateSoon = false; - - if (handle == NULL) { - if (tsAccessSquence % tsBalanceInterval == 0) { - mDebug("balance function is scheduled by timer"); - updateSoon = balanceStart(); - } - } else { - int64_t mseconds = (int64_t)handle; - mDebug("balance function is scheduled by event for %" PRId64 " mseconds arrived", mseconds); - updateSoon = balanceStart(); - } - - if (updateSoon) { - balanceStartTimer(1000); - } else { - taosTmrReset(balanceProcessBalanceTimer, tsStatusInterval * 1000, NULL, tsMnodeTmr, &tsBalanceTimer); - } -} - -static void balanceStartTimer(int64_t mseconds) { - taosTmrReset(balanceProcessBalanceTimer, mseconds, (void *)mseconds, tsMnodeTmr, &tsBalanceTimer); -} - -void balanceSyncNotify() { +void bnCheckModules() { if (sdbIsMaster()) { - balanceLock(); - balanceAccquireDnodeList(); - balanceMonitorDnodeModule(); - balanceReleaseDnodeList(); - balanceUnLock(); + bnLock(); + bnAccquireDnodes(); + bnMonitorDnodeModule(); + bnReleaseDnodes(); + bnUnLock(); } } -void balanceAsyncNotify() { - balanceStartTimer(500); -} - -int32_t balanceInit() { - mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_SCORES, balanceGetScoresMeta); - mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_SCORES, balanceRetrieveScores); - - pthread_mutex_init(&tsBalanceMutex, NULL); - balanceInitDnodeList(); - balanceStartTimer(2000); - mDebug("balance start fp:%p initialized", balanceProcessBalanceTimer); - - balanceReset(); +int32_t bnInit() { + pthread_mutex_init(&tsBnMgmt.mutex, NULL); + bnInitDnodes(); + bnInitThread(); + bnReset(); return 0; } -void balanceCleanUp() { - if (tsBalanceTimer != NULL) { - taosTmrStopA(&tsBalanceTimer); - pthread_mutex_destroy(&tsBalanceMutex); - tsBalanceTimer = NULL; - mDebug("stop balance timer"); - } - balanceCleanupDnodeList(); +void bnCleanUp() { + bnCleanupThread(); + bnCleanupDnodes(); + pthread_mutex_destroy(&tsBnMgmt.mutex); } -int32_t balanceDropDnode(SDnodeObj *pDnode) { +int32_t bnDropDnode(SDnodeObj *pDnode) { int32_t totalFreeVnodes = 0; void * pIter = NULL; SDnodeObj *pTempDnode = NULL; @@ -660,15 +587,13 @@ int32_t balanceDropDnode(SDnodeObj *pDnode) { pIter = mnodeGetNextDnode(pIter, &pTempDnode); if (pTempDnode == NULL) break; - if (pTempDnode != pDnode && balanceCheckFree(pTempDnode)) { + if (pTempDnode != pDnode && bnCheckFree(pTempDnode)) { totalFreeVnodes += (TSDB_MAX_VNODES - pTempDnode->openVnodes); } mnodeDecDnodeRef(pTempDnode); } - sdbFreeIter(pIter); - if (pDnode->openVnodes > totalFreeVnodes) { mError("dnode:%d, openVnodes:%d totalFreeVnodes:%d no enough dnodes", pDnode->dnodeId, pDnode->openVnodes, totalFreeVnodes); return TSDB_CODE_MND_NO_ENOUGH_DNODES; @@ -677,296 +602,17 @@ int32_t balanceDropDnode(SDnodeObj *pDnode) { pDnode->status = TAOS_DN_STATUS_DROPPING; mnodeUpdateDnode(pDnode); - balanceStartTimer(1100); + bnStartTimer(1100); return TSDB_CODE_SUCCESS; } -static int32_t balanceCalcCpuScore(SDnodeObj *pDnode) { - if (pDnode->cpuAvgUsage < 80) - return 0; - else if (pDnode->cpuAvgUsage < 90) - return 10; - else - return 50; -} - -static int32_t balanceCalcMemoryScore(SDnodeObj *pDnode) { - if (pDnode->memoryAvgUsage < 80) - return 0; - else if (pDnode->memoryAvgUsage < 90) - return 10; - else - return 50; -} - -static int32_t balanceCalcDiskScore(SDnodeObj *pDnode) { - if (pDnode->diskAvgUsage < 80) - return 0; - else if (pDnode->diskAvgUsage < 90) - return 10; - else - return 50; -} - -static int32_t balanceCalcBandwidthScore(SDnodeObj *pDnode) { - if (pDnode->bandwidthUsage < 30) - return 0; - else if (pDnode->bandwidthUsage < 80) - return 10; - else - return 50; -} - -static float balanceCalcModuleScore(SDnodeObj *pDnode) { - if (pDnode->numOfCores <= 0) return 0; - if (pDnode->isMgmt) { - return (float)tsMnodeEqualVnodeNum / pDnode->numOfCores; - } - return 0; -} - -static float balanceCalcVnodeScore(SDnodeObj *pDnode, int32_t extra) { - if (pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) return 100000000; - if (pDnode->numOfCores <= 0) return 0; - return (float)(pDnode->openVnodes + extra) / pDnode->numOfCores; -} - -/** - * calc singe score, such as cpu/memory/disk/bandwitdh/vnode - * 1. get the score config - * 2. if the value is out of range, use border data - * 3. otherwise use interpolation method - **/ -void balanceCalcDnodeScore(SDnodeObj *pDnode) { - pDnode->score = balanceCalcCpuScore(pDnode) + balanceCalcMemoryScore(pDnode) + balanceCalcDiskScore(pDnode) + - balanceCalcBandwidthScore(pDnode) + balanceCalcModuleScore(pDnode) + - balanceCalcVnodeScore(pDnode, 0) + pDnode->customScore; -} - -float balanceTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extra) { - int32_t systemScore = balanceCalcCpuScore(pDnode) + balanceCalcMemoryScore(pDnode) + balanceCalcDiskScore(pDnode) + - balanceCalcBandwidthScore(pDnode); - float moduleScore = balanceCalcModuleScore(pDnode); - float vnodeScore = balanceCalcVnodeScore(pDnode, extra); - - float score = systemScore + moduleScore + vnodeScore + pDnode->customScore; - return score; -} - -static void balanceInitDnodeList() { - tsBalanceDnodeList = calloc(tsBalanceDnodeListMallocSize, sizeof(SDnodeObj *)); -} - -static void balanceCleanupDnodeList() { - if (tsBalanceDnodeList != NULL) { - free(tsBalanceDnodeList); - tsBalanceDnodeList = NULL; - } -} - -static void balanceCheckDnodeListSize(int32_t dnodesNum) { - if (tsBalanceDnodeListMallocSize <= dnodesNum) { - tsBalanceDnodeListMallocSize = dnodesNum * 2; - tsBalanceDnodeList = realloc(tsBalanceDnodeList, tsBalanceDnodeListMallocSize * sizeof(SDnodeObj *)); - } -} - -void balanceAccquireDnodeList() { - int32_t dnodesNum = mnodeGetDnodesNum(); - balanceCheckDnodeListSize(dnodesNum); - - void * pIter = NULL; - SDnodeObj *pDnode = NULL; - int32_t dnodeIndex = 0; - - while (1) { - if (dnodeIndex >= dnodesNum) break; - pIter = mnodeGetNextDnode(pIter, &pDnode); - if (pDnode == NULL) break; - if (pDnode->status == TAOS_DN_STATUS_OFFLINE) { - mnodeDecDnodeRef(pDnode); - continue; - } - - balanceCalcDnodeScore(pDnode); - - int32_t orderIndex = dnodeIndex; - for (; orderIndex > 0; --orderIndex) { - if (pDnode->score > tsBalanceDnodeList[orderIndex - 1]->score) { - break; - } - tsBalanceDnodeList[orderIndex] = tsBalanceDnodeList[orderIndex - 1]; - } - tsBalanceDnodeList[orderIndex] = pDnode; - dnodeIndex++; - } - - sdbFreeIter(pIter); - - tsBalanceDnodeListSize = dnodeIndex; -} - -void balanceReleaseDnodeList() { - for (int32_t i = 0; i < tsBalanceDnodeListSize; ++i) { - SDnodeObj *pDnode = tsBalanceDnodeList[i]; - if (pDnode != NULL) { - mnodeDecDnodeRef(pDnode); - } - } -} - -static int32_t balanceGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn) { - SUserObj *pUser = mnodeGetUserFromConn(pConn); - if (pUser == NULL) return 0; - - if (strcmp(pUser->pAcct->user, "root") != 0) { - mnodeDecUserRef(pUser); - return TSDB_CODE_MND_NO_RIGHTS; - } - - int32_t cols = 0; - SSchema *pSchema = pMeta->schema; - - pShow->bytes[cols] = 2; - pSchema[cols].type = TSDB_DATA_TYPE_SMALLINT; - strcpy(pSchema[cols].name, "id"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 4; - pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; - strcpy(pSchema[cols].name, "system scores"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 4; - pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; - strcpy(pSchema[cols].name, "custom scores"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 4; - pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; - strcpy(pSchema[cols].name, "module scores"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 4; - pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; - strcpy(pSchema[cols].name, "vnode scores"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 4; - pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; - strcpy(pSchema[cols].name, "total scores"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 4; - pSchema[cols].type = TSDB_DATA_TYPE_INT; - strcpy(pSchema[cols].name, "open vnodes"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 4; - pSchema[cols].type = TSDB_DATA_TYPE_INT; - strcpy(pSchema[cols].name, "cpu cores"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pShow->bytes[cols] = 18 + VARSTR_HEADER_SIZE; - pSchema[cols].type = TSDB_DATA_TYPE_BINARY; - strcpy(pSchema[cols].name, "balance state"); - pSchema[cols].bytes = htons(pShow->bytes[cols]); - cols++; - - pMeta->numOfColumns = htons(cols); - pShow->numOfColumns = cols; - - pShow->offset[0] = 0; - for (int32_t i = 1; i < cols; ++i) { - pShow->offset[i] = pShow->offset[i - 1] + pShow->bytes[i - 1]; - } - - pShow->numOfRows = mnodeGetDnodesNum(); - pShow->rowSize = pShow->offset[cols - 1] + pShow->bytes[cols - 1]; - pShow->pIter = NULL; - - mnodeDecUserRef(pUser); - - return 0; -} - -static int32_t balanceRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn) { - int32_t numOfRows = 0; - SDnodeObj *pDnode = NULL; - char * pWrite; - int32_t cols = 0; - - while (numOfRows < rows) { - pShow->pIter = mnodeGetNextDnode(pShow->pIter, &pDnode); - if (pDnode == NULL) break; - - int32_t systemScore = balanceCalcCpuScore(pDnode) + balanceCalcMemoryScore(pDnode) + balanceCalcDiskScore(pDnode) + - balanceCalcBandwidthScore(pDnode); - float moduleScore = balanceCalcModuleScore(pDnode); - float vnodeScore = balanceCalcVnodeScore(pDnode, 0); - - cols = 0; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(int16_t *)pWrite = pDnode->dnodeId; - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(float *)pWrite = systemScore; - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(float *)pWrite = pDnode->customScore; - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(float *)pWrite = (int32_t)moduleScore; - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(float *)pWrite = (int32_t)vnodeScore; - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(float *)pWrite = (int32_t)(vnodeScore + moduleScore + pDnode->customScore + systemScore); - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(int32_t *)pWrite = pDnode->openVnodes; - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - *(int32_t *)pWrite = pDnode->numOfCores; - cols++; - - pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; - STR_TO_VARSTR(pWrite, mnodeGetDnodeStatusStr(pDnode->status)); - cols++; - - numOfRows++; - mnodeDecDnodeRef(pDnode); - } - - mnodeVacuumResult(data, pShow->numOfColumns, numOfRows, rows, pShow); - pShow->numOfReads += numOfRows; - return numOfRows; -} - -static void balanceMonitorDnodeModule() { +static void bnMonitorDnodeModule() { int32_t numOfMnodes = mnodeGetMnodesNum(); if (numOfMnodes >= tsNumOfMnodes) return; - for (int32_t i = 0; i < tsBalanceDnodeListSize; ++i) { - SDnodeObj *pDnode = tsBalanceDnodeList[i]; + for (int32_t i = 0; i < tsBnDnodes.size; ++i) { + SDnodeObj *pDnode = tsBnDnodes.list[i]; if (pDnode == NULL) break; if (pDnode->isMgmt || pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) { @@ -990,7 +636,7 @@ static void balanceMonitorDnodeModule() { } } -int32_t balanceAlterDnode(struct SDnodeObj *pSrcDnode, int32_t vnodeId, int32_t dnodeId) { +int32_t bnAlterDnode(struct SDnodeObj *pSrcDnode, int32_t vnodeId, int32_t dnodeId) { if (!sdbIsMaster()) { mError("dnode:%d, failed to alter vgId:%d to dnode:%d, for self not master", pSrcDnode->dnodeId, vnodeId, dnodeId); return TSDB_CODE_MND_DNODE_NOT_EXIST; @@ -1014,29 +660,29 @@ int32_t balanceAlterDnode(struct SDnodeObj *pSrcDnode, int32_t vnodeId, int32_t return TSDB_CODE_MND_DNODE_NOT_EXIST; } - balanceLock(); - balanceAccquireDnodeList(); + bnLock(); + bnAccquireDnodes(); int32_t code = TSDB_CODE_SUCCESS; - if (!balanceCheckDnodeInVgroup(pSrcDnode, pVgroup)) { + if (!bnCheckDnodeInVgroup(pSrcDnode, pVgroup)) { mError("dnode:%d, failed to alter vgId:%d to dnode:%d, vgroup not in dnode:%d", pSrcDnode->dnodeId, vnodeId, dnodeId, pSrcDnode->dnodeId); code = TSDB_CODE_MND_VGROUP_NOT_IN_DNODE; - } else if (balanceCheckDnodeInVgroup(pDestDnode, pVgroup)) { + } else if (bnCheckDnodeInVgroup(pDestDnode, pVgroup)) { mError("dnode:%d, failed to alter vgId:%d to dnode:%d, vgroup already in dnode:%d", pSrcDnode->dnodeId, vnodeId, dnodeId, dnodeId); code = TSDB_CODE_MND_VGROUP_ALREADY_IN_DNODE; - } else if (!balanceCheckFree(pDestDnode)) { + } else if (!bnCheckFree(pDestDnode)) { mError("dnode:%d, failed to alter vgId:%d to dnode:%d, for dnode:%d not free", pSrcDnode->dnodeId, vnodeId, dnodeId, dnodeId); code = TSDB_CODE_MND_DNODE_NOT_FREE; } else { - code = balanceAddVnode(pVgroup, pSrcDnode, pDestDnode); + code = bnAddVnode(pVgroup, pSrcDnode, pDestDnode); mInfo("dnode:%d, alter vgId:%d to dnode:%d, result:%s", pSrcDnode->dnodeId, vnodeId, dnodeId, tstrerror(code)); } - balanceReleaseDnodeList(); - balanceUnLock(); + bnReleaseDnodes(); + bnUnLock(); mnodeDecVgroupRef(pVgroup); mnodeDecDnodeRef(pDestDnode); diff --git a/src/balance/src/bnScore.c b/src/balance/src/bnScore.c new file mode 100644 index 0000000000000000000000000000000000000000..e5ad7a211958dce9507c5fca22f57a6d99c82029 --- /dev/null +++ b/src/balance/src/bnScore.c @@ -0,0 +1,312 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "tglobal.h" +#include "mnodeShow.h" +#include "mnodeUser.h" +#include "bnScore.h" + +SBnDnodes tsBnDnodes; + +static int32_t bnGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); +static int32_t bnRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn); + +static int32_t bnCalcCpuScore(SDnodeObj *pDnode) { + if (pDnode->cpuAvgUsage < 80) + return 0; + else if (pDnode->cpuAvgUsage < 90) + return 10; + else + return 50; +} + +static int32_t bnCalcMemoryScore(SDnodeObj *pDnode) { + if (pDnode->memoryAvgUsage < 80) + return 0; + else if (pDnode->memoryAvgUsage < 90) + return 10; + else + return 50; +} + +static int32_t bnCalcDiskScore(SDnodeObj *pDnode) { + if (pDnode->diskAvgUsage < 80) + return 0; + else if (pDnode->diskAvgUsage < 90) + return 10; + else + return 50; +} + +static int32_t bnCalcBandScore(SDnodeObj *pDnode) { + if (pDnode->bandwidthUsage < 30) + return 0; + else if (pDnode->bandwidthUsage < 80) + return 10; + else + return 50; +} + +static float bnCalcModuleScore(SDnodeObj *pDnode) { + if (pDnode->numOfCores <= 0) return 0; + if (pDnode->isMgmt) { + return (float)tsMnodeEqualVnodeNum / pDnode->numOfCores; + } + return 0; +} + +static float bnCalcVnodeScore(SDnodeObj *pDnode, int32_t extra) { + if (pDnode->status == TAOS_DN_STATUS_DROPPING || pDnode->status == TAOS_DN_STATUS_OFFLINE) return 100000000; + if (pDnode->numOfCores <= 0) return 0; + return (float)(pDnode->openVnodes + extra) / pDnode->numOfCores; +} + +/** + * calc singe score, such as cpu/memory/disk/bandwitdh/vnode + * 1. get the score config + * 2. if the value is out of range, use border data + * 3. otherwise use interpolation method + **/ +static void bnCalcDnodeScore(SDnodeObj *pDnode) { + pDnode->score = bnCalcCpuScore(pDnode) + bnCalcMemoryScore(pDnode) + bnCalcDiskScore(pDnode) + + bnCalcBandScore(pDnode) + bnCalcModuleScore(pDnode) + bnCalcVnodeScore(pDnode, 0) + + pDnode->customScore; +} + +float bnTryCalcDnodeScore(SDnodeObj *pDnode, int32_t extra) { + int32_t systemScore = bnCalcCpuScore(pDnode) + bnCalcMemoryScore(pDnode) + bnCalcDiskScore(pDnode) + + bnCalcBandScore(pDnode); + float moduleScore = bnCalcModuleScore(pDnode); + float vnodeScore = bnCalcVnodeScore(pDnode, extra); + + float score = systemScore + moduleScore + vnodeScore + pDnode->customScore; + return score; +} + +void bnInitDnodes() { + mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_SCORES, bnGetScoresMeta); + mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_SCORES, bnRetrieveScores); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_SCORES, mnodeCancelGetNextDnode); + + memset(&tsBnDnodes, 0, sizeof(SBnDnodes)); + tsBnDnodes.maxSize = 16; + tsBnDnodes.list = calloc(tsBnDnodes.maxSize, sizeof(SDnodeObj *)); +} + +void bnCleanupDnodes() { + if (tsBnDnodes.list != NULL) { + free(tsBnDnodes.list); + tsBnDnodes.list = NULL; + } +} + +static void bnCheckDnodesSize(int32_t dnodesNum) { + if (tsBnDnodes.maxSize <= dnodesNum) { + tsBnDnodes.maxSize = dnodesNum * 2; + tsBnDnodes.list = realloc(tsBnDnodes.list, tsBnDnodes.maxSize * sizeof(SDnodeObj *)); + } +} + +void bnAccquireDnodes() { + int32_t dnodesNum = mnodeGetDnodesNum(); + bnCheckDnodesSize(dnodesNum); + + void * pIter = NULL; + SDnodeObj *pDnode = NULL; + int32_t dnodeIndex = 0; + + while (1) { + if (dnodeIndex >= dnodesNum) { + mnodeCancelGetNextDnode(pIter); + break; + } + + pIter = mnodeGetNextDnode(pIter, &pDnode); + if (pDnode == NULL) break; + if (pDnode->status == TAOS_DN_STATUS_OFFLINE) { + mnodeDecDnodeRef(pDnode); + continue; + } + + bnCalcDnodeScore(pDnode); + + int32_t orderIndex = dnodeIndex; + for (; orderIndex > 0; --orderIndex) { + if (pDnode->score > tsBnDnodes.list[orderIndex - 1]->score) { + break; + } + tsBnDnodes.list[orderIndex] = tsBnDnodes.list[orderIndex - 1]; + } + tsBnDnodes.list[orderIndex] = pDnode; + dnodeIndex++; + } + + tsBnDnodes.size = dnodeIndex; +} + +void bnReleaseDnodes() { + for (int32_t i = 0; i < tsBnDnodes.size; ++i) { + SDnodeObj *pDnode = tsBnDnodes.list[i]; + if (pDnode != NULL) { + mnodeDecDnodeRef(pDnode); + } + } +} + +static int32_t bnGetScoresMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn) { + SUserObj *pUser = mnodeGetUserFromConn(pConn); + if (pUser == NULL) return 0; + + if (strcmp(pUser->pAcct->user, "root") != 0) { + mnodeDecUserRef(pUser); + return TSDB_CODE_MND_NO_RIGHTS; + } + + int32_t cols = 0; + SSchema *pSchema = pMeta->schema; + + pShow->bytes[cols] = 2; + pSchema[cols].type = TSDB_DATA_TYPE_SMALLINT; + strcpy(pSchema[cols].name, "id"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 4; + pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; + strcpy(pSchema[cols].name, "system scores"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 4; + pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; + strcpy(pSchema[cols].name, "custom scores"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 4; + pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; + strcpy(pSchema[cols].name, "module scores"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 4; + pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; + strcpy(pSchema[cols].name, "vnode scores"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 4; + pSchema[cols].type = TSDB_DATA_TYPE_FLOAT; + strcpy(pSchema[cols].name, "total scores"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 4; + pSchema[cols].type = TSDB_DATA_TYPE_INT; + strcpy(pSchema[cols].name, "open vnodes"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 4; + pSchema[cols].type = TSDB_DATA_TYPE_INT; + strcpy(pSchema[cols].name, "cpu cores"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pShow->bytes[cols] = 18 + VARSTR_HEADER_SIZE; + pSchema[cols].type = TSDB_DATA_TYPE_BINARY; + strcpy(pSchema[cols].name, "balance state"); + pSchema[cols].bytes = htons(pShow->bytes[cols]); + cols++; + + pMeta->numOfColumns = htons(cols); + pShow->numOfColumns = cols; + + pShow->offset[0] = 0; + for (int32_t i = 1; i < cols; ++i) { + pShow->offset[i] = pShow->offset[i - 1] + pShow->bytes[i - 1]; + } + + pShow->numOfRows = mnodeGetDnodesNum(); + pShow->rowSize = pShow->offset[cols - 1] + pShow->bytes[cols - 1]; + pShow->pIter = NULL; + + mnodeDecUserRef(pUser); + + return 0; +} + +static int32_t bnRetrieveScores(SShowObj *pShow, char *data, int32_t rows, void *pConn) { + int32_t numOfRows = 0; + SDnodeObj *pDnode = NULL; + char * pWrite; + int32_t cols = 0; + + while (numOfRows < rows) { + pShow->pIter = mnodeGetNextDnode(pShow->pIter, &pDnode); + if (pDnode == NULL) break; + + int32_t systemScore = bnCalcCpuScore(pDnode) + bnCalcMemoryScore(pDnode) + bnCalcDiskScore(pDnode) + bnCalcBandScore(pDnode); + float moduleScore = bnCalcModuleScore(pDnode); + float vnodeScore = bnCalcVnodeScore(pDnode, 0); + + cols = 0; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(int16_t *)pWrite = pDnode->dnodeId; + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(float *)pWrite = systemScore; + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(float *)pWrite = pDnode->customScore; + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(float *)pWrite = (int32_t)moduleScore; + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(float *)pWrite = (int32_t)vnodeScore; + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(float *)pWrite = (int32_t)(vnodeScore + moduleScore + pDnode->customScore + systemScore); + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(int32_t *)pWrite = pDnode->openVnodes; + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + *(int32_t *)pWrite = pDnode->numOfCores; + cols++; + + pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows; + STR_TO_VARSTR(pWrite, mnodeGetDnodeStatusStr(pDnode->status)); + cols++; + + numOfRows++; + mnodeDecDnodeRef(pDnode); + } + + mnodeVacuumResult(data, pShow->numOfColumns, numOfRows, rows, pShow); + pShow->numOfReads += numOfRows; + return numOfRows; +} diff --git a/src/balance/src/bnThread.c b/src/balance/src/bnThread.c new file mode 100644 index 0000000000000000000000000000000000000000..bf046a9faea6e9e3d7b324e7786c594d16a9aeb7 --- /dev/null +++ b/src/balance/src/bnThread.c @@ -0,0 +1,132 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "ttimer.h" +#include "tglobal.h" +#include "mnodeSdb.h" +#include "bnThread.h" + +static SBnThread tsBnThread; + +static void *bnThreadFunc(void *arg) { + while (1) { + pthread_mutex_lock(&tsBnThread.mutex); + if (tsBnThread.stop) { + pthread_mutex_unlock(&tsBnThread.mutex); + break; + } + + pthread_cond_wait(&tsBnThread.cond, &tsBnThread.mutex); + bool updateSoon = bnStart(); + bnStartTimer(updateSoon ? 1000 : -1); + pthread_mutex_unlock(&(tsBnThread.mutex)); + } + + mDebug("balance thread is stopped"); + return NULL; +} + +int32_t bnInitThread() { + memset(&tsBnThread, 0, sizeof(SBnThread)); + tsBnThread.stop = false; + pthread_mutex_init(&tsBnThread.mutex, NULL); + pthread_cond_init(&tsBnThread.cond, NULL); + + pthread_attr_t thattr; + pthread_attr_init(&thattr); + pthread_attr_setdetachstate(&thattr, PTHREAD_CREATE_JOINABLE); + int32_t ret = pthread_create(&tsBnThread.thread, &thattr, bnThreadFunc, NULL); + pthread_attr_destroy(&thattr); + + if (ret != 0) { + mError("failed to create balance thread since %s", strerror(errno)); + return -1; + } + + bnStartTimer(2000); + mDebug("balance thread is created"); + return 0; +} + +void bnCleanupThread() { + mDebug("balance thread will be cleanup"); + + if (tsBnThread.timer != NULL) { + taosTmrStopA(&tsBnThread.timer); + tsBnThread.timer = NULL; + mDebug("stop balance timer"); + } + + pthread_mutex_lock(&tsBnThread.mutex); + tsBnThread.stop = true; + pthread_cond_signal(&tsBnThread.cond); + pthread_mutex_unlock(&(tsBnThread.mutex)); + pthread_join(tsBnThread.thread, NULL); + + pthread_cond_destroy(&tsBnThread.cond); + pthread_mutex_destroy(&tsBnThread.mutex); +} + +static void bnPostSignal() { + if (tsBnThread.stop) return; + + pthread_mutex_lock(&tsBnThread.mutex); + pthread_cond_signal(&tsBnThread.cond); + pthread_mutex_unlock(&(tsBnThread.mutex)); +} + +/* + * once sdb work as mater, then tsAccessSquence reset to zero + * increase tsAccessSquence every balance interval + */ + +static void bnProcessTimer(void *handle, void *tmrId) { + if (!sdbIsMaster()) return; + if (tsBnThread.stop) return; + + tsBnThread.timer = NULL; + tsAccessSquence++; + + bnCheckStatus(); + bnStartTimer(-1); + + if (handle == NULL) { + if (tsAccessSquence % tsBalanceInterval == 0) { + mDebug("balance function is scheduled by timer"); + bnPostSignal(); + } + } else { + int64_t mseconds = (int64_t)handle; + mDebug("balance function is scheduled by event for %" PRId64 " mseconds arrived", mseconds); + bnPostSignal(); + } +} + +void bnStartTimer(int64_t mseconds) { + if (tsBnThread.stop) return; + + bool updateSoon = (mseconds != -1); + if (updateSoon) { + taosTmrReset(bnProcessTimer, mseconds, (void *)mseconds, tsMnodeTmr, &tsBnThread.timer); + } else { + taosTmrReset(bnProcessTimer, tsStatusInterval * 1000, NULL, tsMnodeTmr, &tsBnThread.timer); + } +} + +void bnNotify() { + bnStartTimer(500); +} diff --git a/src/client/inc/tscLocalMerge.h b/src/client/inc/tscLocalMerge.h index 2c7c2f51d02ea1b9943004fc62f8e48f2034d0fe..43ba31f331fc5c92d4142a008db458b23395a473 100644 --- a/src/client/inc/tscLocalMerge.h +++ b/src/client/inc/tscLocalMerge.h @@ -56,7 +56,6 @@ typedef struct SLocalReducer { tFilePage * pTempBuffer; struct SQLFunctionCtx *pCtx; int32_t rowSize; // size of each intermediate result. - int32_t finalRowSize; // final result row size int32_t status; // denote it is in reduce process, in reduce process, it bool hasPrevRow; // cannot be released bool hasUnprocessedRow; diff --git a/src/client/inc/tscUtil.h b/src/client/inc/tscUtil.h index bde27d2932a5bacd09864c76ee81faa6adef04a7..2c8641da760fd2e1f340cc4627559ec1a226686c 100644 --- a/src/client/inc/tscUtil.h +++ b/src/client/inc/tscUtil.h @@ -92,7 +92,7 @@ typedef struct SVgroupTableInfo { } SVgroupTableInfo; static FORCE_INLINE SQueryInfo* tscGetQueryInfoDetail(SSqlCmd* pCmd, int32_t subClauseIndex) { - assert(pCmd != NULL && subClauseIndex >= 0 && subClauseIndex < TSDB_MAX_UNION_CLAUSE); + assert(pCmd != NULL && subClauseIndex >= 0); if (pCmd->pQueryInfo == NULL || subClauseIndex >= pCmd->numOfClause) { return NULL; @@ -278,7 +278,7 @@ bool hasMoreClauseToTry(SSqlObj* pSql); void tscTryQueryNextVnode(SSqlObj *pSql, __async_cb_func_t fp); void tscAsyncQuerySingleRowForNextVnode(void *param, TAOS_RES *tres, int numOfRows); void tscTryQueryNextClause(SSqlObj* pSql, __async_cb_func_t fp); -int tscSetMgmtEpSetFromCfg(const char *first, const char *second); +int tscSetMgmtEpSetFromCfg(const char *first, const char *second, SRpcCorEpSet *corEpSet); bool tscSetSqlOwner(SSqlObj* pSql); void tscClearSqlOwner(SSqlObj* pSql); diff --git a/src/client/inc/tsclient.h b/src/client/inc/tsclient.h index ff36cf0f5ac31ecf3baa4abae5697e08a1c8bc0b..a1b6174de0d21f02fdde1e367bcb45901707c6ba 100644 --- a/src/client/inc/tsclient.h +++ b/src/client/inc/tsclient.h @@ -246,11 +246,14 @@ typedef struct SQueryInfo { int16_t fillType; // final result fill type int16_t numOfTables; STableMetaInfo **pTableMetaInfo; - struct STSBuf * tsBuf; + struct STSBuf *tsBuf; int64_t * fillVal; // default value for fill char * msg; // pointer to the pCmd->payload to keep error message temporarily int64_t clauseLimit; // limit for current sub clause + int64_t prjOffset; // offset value in the original sql expression, only applied at client side + int64_t tableLimit; // table limit in case of super table projection query + global order + limit + int32_t udColumnId; // current user-defined constant output field column id, monotonically decreases from TSDB_UD_COLUMN_INDEX int16_t resColumnId; // result column id } SQueryInfo; @@ -282,7 +285,7 @@ typedef struct { int8_t dataSourceType; // load data from file or not int8_t submitSchema; // submit block is built with table schema - STagData tagData; + STagData *pTagData; // NOTE: pTagData->data is used as a variant length array SHashObj *pTableList; // referred table involved in sql SArray *pDataBlocks; // SArray submit data blocks after parsing sql } SSqlCmd; @@ -333,9 +336,10 @@ typedef struct STscObj { char superAuth : 1; uint32_t connId; uint64_t rid; // ref ID returned by taosAddRef - struct SSqlObj * pHb; + int64_t hbrid; struct SSqlObj * sqlList; struct SSqlStream *streamList; + SRpcCorEpSet *tscCorMgmtEpSet; void* pDnodeConn; pthread_mutex_t mutex; T_REF_DECLARE() @@ -373,7 +377,7 @@ typedef struct SSqlObj { struct SSqlObj **pSubs; struct SSqlObj *prev, *next; - struct SSqlObj **self; + int64_t self; } SSqlObj; typedef struct SSqlStream { @@ -507,7 +511,7 @@ static FORCE_INLINE void tscGetResultColumnChr(SSqlRes* pRes, SFieldInfo* pField } extern SCacheObj* tscMetaCache; -extern SCacheObj* tscObjCache; +extern int tscObjRef; extern void * tscTmr; extern void * tscQhandle; extern int tscKeepConn[]; @@ -515,7 +519,6 @@ extern int tsInsertHeadSize; extern int tscNumOfThreads; extern int tscRefId; -extern SRpcCorEpSet tscMgmtEpSet; extern int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo); diff --git a/src/client/src/tscFunctionImpl.c b/src/client/src/tscFunctionImpl.c index d39b833374b4891358053788a1afbf9246dfb0dd..7921399330f876a217b8f14f4871d7f47cebec21 100644 --- a/src/client/src/tscFunctionImpl.c +++ b/src/client/src/tscFunctionImpl.c @@ -54,8 +54,8 @@ #define DO_UPDATE_TAG_COLUMNS(ctx, ts) \ do { \ - for (int32_t i = 0; i < (ctx)->tagInfo.numOfTagCols; ++i) { \ - SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[i]; \ + for (int32_t _i = 0; _i < (ctx)->tagInfo.numOfTagCols; ++_i) { \ + SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[_i]; \ if (__ctx->functionId == TSDB_FUNC_TS_DUMMY) { \ __ctx->tag.i64Key = (ts); \ __ctx->tag.nType = TSDB_DATA_TYPE_BIGINT; \ @@ -64,13 +64,13 @@ } \ } while (0); -#define DO_UPDATE_TAG_COLUMNS_WITHOUT_TS(ctx) \ -do {\ -for (int32_t i = 0; i < (ctx)->tagInfo.numOfTagCols; ++i) { \ - SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[i]; \ - aAggs[TSDB_FUNC_TAG].xFunction(__ctx); \ - } \ -} while(0); +#define DO_UPDATE_TAG_COLUMNS_WITHOUT_TS(ctx) \ + do { \ + for (int32_t _i = 0; _i < (ctx)->tagInfo.numOfTagCols; ++_i) { \ + SQLFunctionCtx *__ctx = (ctx)->tagInfo.pTagCtxList[_i]; \ + aAggs[TSDB_FUNC_TAG].xFunction(__ctx); \ + } \ + } while (0); void noop1(SQLFunctionCtx *UNUSED_PARAM(pCtx)) {} void noop2(SQLFunctionCtx *UNUSED_PARAM(pCtx), int32_t UNUSED_PARAM(index)) {} @@ -305,7 +305,7 @@ int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionI } else if (functionId == TSDB_FUNC_FIRST || functionId == TSDB_FUNC_LAST) { *type = (int16_t)dataType; *bytes = (int16_t)dataBytes; - *interBytes = dataBytes; + *interBytes = (int16_t)(dataBytes + sizeof(SFirstLastInfo)); } else if (functionId == TSDB_FUNC_SPREAD) { *type = (int16_t)TSDB_DATA_TYPE_DOUBLE; *bytes = sizeof(double); @@ -426,8 +426,7 @@ static void count_function_f(SQLFunctionCtx *pCtx, int32_t index) { } SET_VAL(pCtx, 1, 1); - - *((int64_t *)pCtx->aOutputBuf) += 1; + *((int64_t *)pCtx->aOutputBuf) += pCtx->size; // do not need it actually SResultRowCellInfo *pInfo = GET_RES_INFO(pCtx); @@ -1170,8 +1169,8 @@ static int32_t minmax_merge_impl(SQLFunctionCtx *pCtx, int32_t bytes, char *outp if ((*(int32_t *)output < v) ^ isMin) { *(int32_t *)output = v; - for (int32_t i = 0; i < pCtx->tagInfo.numOfTagCols; ++i) { - SQLFunctionCtx *__ctx = pCtx->tagInfo.pTagCtxList[i]; + for (int32_t j = 0; j < pCtx->tagInfo.numOfTagCols; ++j) { + SQLFunctionCtx *__ctx = pCtx->tagInfo.pTagCtxList[j]; aAggs[TSDB_FUNC_TAG].xFunction(__ctx); } @@ -1680,16 +1679,35 @@ static void last_function_f(SQLFunctionCtx *pCtx, int32_t index) { if (pCtx->hasNull && isNull(pData, pCtx->inputType)) { return; } - - SET_VAL(pCtx, 1, 1); - memcpy(pCtx->aOutputBuf, pData, pCtx->inputBytes); - - TSKEY ts = pCtx->ptsList[index]; - DO_UPDATE_TAG_COLUMNS(pCtx, ts); - - SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - pResInfo->hasResult = DATA_SET_FLAG; - pResInfo->complete = true; // set query completed + + // the scan order is not the required order, ignore it + if (pCtx->order != pCtx->param[0].i64Key) { + return; + } + + if (pCtx->order == TSDB_ORDER_DESC) { + SET_VAL(pCtx, 1, 1); + memcpy(pCtx->aOutputBuf, pData, pCtx->inputBytes); + + TSKEY ts = pCtx->ptsList[index]; + DO_UPDATE_TAG_COLUMNS(pCtx, ts); + + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + pResInfo->hasResult = DATA_SET_FLAG; + pResInfo->complete = true; // set query completed + } else { // in case of ascending order check, all data needs to be checked + SResultRowCellInfo* pResInfo = GET_RES_INFO(pCtx); + TSKEY ts = pCtx->ptsList[index]; + + char* buf = GET_ROWCELL_INTERBUF(pResInfo); + if (pResInfo->hasResult != DATA_SET_FLAG || (*(TSKEY*)buf) < ts) { + pResInfo->hasResult = DATA_SET_FLAG; + memcpy(pCtx->aOutputBuf, pData, pCtx->inputBytes); + + *(TSKEY*)buf = ts; + DO_UPDATE_TAG_COLUMNS(pCtx, ts); + } + } } static void last_data_assign_impl(SQLFunctionCtx *pCtx, char *pData, int32_t index) { @@ -1712,7 +1730,7 @@ static void last_data_assign_impl(SQLFunctionCtx *pCtx, char *pData, int32_t ind static void last_dist_function(SQLFunctionCtx *pCtx) { /* - * 1. for scan data in asc order, no need to check data + * 1. for scan data is not the required order * 2. for data blocks that are not loaded, no need to check data */ if (pCtx->order != pCtx->param[0].i64Key) { @@ -2448,7 +2466,7 @@ static bool percentile_function_setup(SQLFunctionCtx *pCtx) { static void percentile_function(SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; - SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); SPercentileInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); // the first stage, only acquire the min/max value @@ -2549,12 +2567,14 @@ static void percentile_finalizer(SQLFunctionCtx *pCtx) { double v = pCtx->param[0].nType == TSDB_DATA_TYPE_INT ? pCtx->param[0].i64Key : pCtx->param[0].dKey; SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - tMemBucket * pMemBucket = ((SPercentileInfo *)GET_ROWCELL_INTERBUF(pResInfo))->pMemBucket; - - if (pMemBucket->total > 0) { // check for null - *(double *)pCtx->aOutputBuf = getPercentile(pMemBucket, v); - } else { + SPercentileInfo* ppInfo = (SPercentileInfo *) GET_ROWCELL_INTERBUF(pResInfo); + + tMemBucket * pMemBucket = ppInfo->pMemBucket; + if (pMemBucket == NULL || pMemBucket->total == 0) { // check for null + assert(ppInfo->numOfElems == 0); setNull(pCtx->aOutputBuf, pCtx->outputType, pCtx->outputBytes); + } else { + *(double *)pCtx->aOutputBuf = getPercentile(pMemBucket, v); } tMemBucketDestroy(pMemBucket); @@ -3624,94 +3644,158 @@ static bool twa_function_setup(SQLFunctionCtx *pCtx) { return false; } - SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); //->aOutputBuf + pCtx->outputBytes; - STwaInfo * pInfo = GET_ROWCELL_INTERBUF(pResInfo); - - pInfo->lastKey = INT64_MIN; - pInfo->type = pCtx->inputType; - + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + + STwaInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); + pInfo->lastKey = INT64_MIN; + pInfo->win = TSWINDOW_INITIALIZER; return true; } -static FORCE_INLINE void setTWALastVal(SQLFunctionCtx *pCtx, const char *data, int32_t i, STwaInfo *pInfo) { - switch (pCtx->inputType) { - case TSDB_DATA_TYPE_INT: - pInfo->iLastValue = GET_INT32_VAL(data + pCtx->inputBytes * i); +static int32_t twa_function_impl(SQLFunctionCtx* pCtx, int32_t tsIndex, int32_t index, int32_t size) { + int32_t notNullElems = 0; + TSKEY *primaryKey = pCtx->ptsList; + + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + + STwaInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); + + int32_t i = index; + int32_t step = GET_FORWARD_DIRECTION_FACTOR(pCtx->order); + + if (pCtx->start.key != INT64_MIN) { + assert((pCtx->start.key < primaryKey[tsIndex + i] && pCtx->order == TSDB_ORDER_ASC) || + (pCtx->start.key > primaryKey[tsIndex + i] && pCtx->order == TSDB_ORDER_DESC)); + + assert(pInfo->lastKey == INT64_MIN); + + pInfo->lastKey = primaryKey[tsIndex + i]; + GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index)); + + pInfo->dOutput += ((pInfo->lastValue + pCtx->start.val) / 2) * (pInfo->lastKey - pCtx->start.key); + + pInfo->hasResult = DATA_SET_FLAG; + pInfo->win.skey = pCtx->start.key; + notNullElems++; + i += step; + } else if (pInfo->lastKey == INT64_MIN) { + pInfo->lastKey = primaryKey[tsIndex + i]; + GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index)); + + pInfo->hasResult = DATA_SET_FLAG; + pInfo->win.skey = pInfo->lastKey; + notNullElems++; + i += step; + } + + // calculate the value of + switch(pCtx->inputType) { + case TSDB_DATA_TYPE_TINYINT: { + int8_t *val = (int8_t*) GET_INPUT_CHAR_INDEX(pCtx, 0); + for (; i < size && i >= 0; i += step) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + tsIndex]; + } break; - case TSDB_DATA_TYPE_TINYINT: - pInfo->iLastValue = GET_INT8_VAL(data + pCtx->inputBytes * i); + } + case TSDB_DATA_TYPE_SMALLINT: { + int16_t *val = (int16_t*) GET_INPUT_CHAR_INDEX(pCtx, 0); + for (; i < size && i >= 0; i += step) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + tsIndex]; + } break; - case TSDB_DATA_TYPE_SMALLINT: - pInfo->iLastValue = GET_INT16_VAL(data + pCtx->inputBytes * i); + } + case TSDB_DATA_TYPE_INT: { + int32_t *val = (int32_t*) GET_INPUT_CHAR_INDEX(pCtx, 0); + for (; i < size && i >= 0; i += step) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + tsIndex]; + } break; - case TSDB_DATA_TYPE_BIGINT: - pInfo->iLastValue = GET_INT64_VAL(data + pCtx->inputBytes * i); + } + case TSDB_DATA_TYPE_BIGINT: { + int64_t *val = (int64_t*) GET_INPUT_CHAR_INDEX(pCtx, 0); + for (; i < size && i >= 0; i += step) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey); + pInfo->lastValue = (double) val[i]; + pInfo->lastKey = primaryKey[i + tsIndex]; + } break; - case TSDB_DATA_TYPE_FLOAT: - pInfo->dLastValue = GET_FLOAT_VAL(data + pCtx->inputBytes * i); + } + case TSDB_DATA_TYPE_FLOAT: { + float *val = (float*) GET_INPUT_CHAR_INDEX(pCtx, 0); + for (; i < size && i >= 0; i += step) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + tsIndex]; + } break; - case TSDB_DATA_TYPE_DOUBLE: - pInfo->dLastValue = GET_DOUBLE_VAL(data + pCtx->inputBytes * i); + } + case TSDB_DATA_TYPE_DOUBLE: { + double *val = (double*) GET_INPUT_CHAR_INDEX(pCtx, 0); + for (; i < size && i >= 0; i += step) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + tsIndex] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + tsIndex]; + } break; - default: - assert(0); + } + default: assert(0); } + + // the last interpolated time window value + if (pCtx->end.key != INT64_MIN) { + pInfo->dOutput += ((pInfo->lastValue + pCtx->end.val) / 2) * (pCtx->end.key - pInfo->lastKey); + pInfo->lastValue = pCtx->end.val; + pInfo->lastKey = pCtx->end.key; + } + + pInfo->win.ekey = pInfo->lastKey; + return notNullElems; } static void twa_function(SQLFunctionCtx *pCtx) { void * data = GET_INPUT_CHAR(pCtx); - TSKEY *primaryKey = pCtx->ptsList; - - int32_t notNullElems = 0; - + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); STwaInfo * pInfo = GET_ROWCELL_INTERBUF(pResInfo); - int32_t i = 0; - // skip null value + int32_t step = GET_FORWARD_DIRECTION_FACTOR(pCtx->order); + int32_t i = (pCtx->order == TSDB_ORDER_ASC)? 0:(pCtx->size - 1); while (pCtx->hasNull && i < pCtx->size && isNull((char *)data + pCtx->inputBytes * i, pCtx->inputType)) { - i++; + i += step; } - - if (i >= pCtx->size) { - return; - } - - if (pInfo->lastKey == INT64_MIN) { - pInfo->lastKey = pCtx->nStartQueryTimestamp; - setTWALastVal(pCtx, data, i, pInfo); - - pInfo->hasResult = DATA_SET_FLAG; - } - - notNullElems++; - - if (pCtx->inputType == TSDB_DATA_TYPE_FLOAT || pCtx->inputType == TSDB_DATA_TYPE_DOUBLE) { - pInfo->dOutput += pInfo->dLastValue * (primaryKey[i] - pInfo->lastKey); - } else { - pInfo->iOutput += pInfo->iLastValue * (primaryKey[i] - pInfo->lastKey); - } - - pInfo->lastKey = primaryKey[i]; - setTWALastVal(pCtx, data, i, pInfo); - - for (++i; i < pCtx->size; i++) { - if (pCtx->hasNull && isNull((char *)data + pCtx->inputBytes * i, pCtx->inputType)) { - continue; - } - - notNullElems++; - if (pCtx->inputType == TSDB_DATA_TYPE_FLOAT || pCtx->inputType == TSDB_DATA_TYPE_DOUBLE) { - pInfo->dOutput += pInfo->dLastValue * (primaryKey[i] - pInfo->lastKey); - } else { - pInfo->iOutput += pInfo->iLastValue * (primaryKey[i] - pInfo->lastKey); - } - - pInfo->lastKey = primaryKey[i]; - setTWALastVal(pCtx, data, i, pInfo); - } - + + int32_t notNullElems = twa_function_impl(pCtx, pCtx->startOffset, i, pCtx->size); SET_VAL(pCtx, notNullElems, 1); if (notNullElems > 0) { @@ -3721,8 +3805,6 @@ static void twa_function(SQLFunctionCtx *pCtx) { if (pCtx->stableQuery) { memcpy(pCtx->aOutputBuf, pInfo, sizeof(STwaInfo)); } - - // pCtx->numOfIteratedElems += notNullElems; } static void twa_function_f(SQLFunctionCtx *pCtx, int32_t index) { @@ -3730,34 +3812,136 @@ static void twa_function_f(SQLFunctionCtx *pCtx, int32_t index) { if (pCtx->hasNull && isNull(pData, pCtx->inputType)) { return; } - - SET_VAL(pCtx, 1, 1); - + + int32_t notNullElems = 0; TSKEY *primaryKey = pCtx->ptsList; - + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + STwaInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); - - if (pInfo->lastKey == INT64_MIN) { - pInfo->lastKey = pCtx->nStartQueryTimestamp; - setTWALastVal(pCtx, pData, 0, pInfo); - + int32_t i = pCtx->startOffset; + int32_t size = pCtx->size; + + if (pCtx->start.key != INT64_MIN) { + assert(pInfo->lastKey == INT64_MIN); + + pInfo->lastKey = primaryKey[index]; + GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index)); + + pInfo->dOutput += ((pInfo->lastValue + pCtx->start.val) / 2) * (pInfo->lastKey - pCtx->start.key); + pInfo->hasResult = DATA_SET_FLAG; + pInfo->win.skey = pCtx->start.key; + notNullElems++; + i += 1; + } else if (pInfo->lastKey == INT64_MIN) { + pInfo->lastKey = primaryKey[index]; + GET_TYPED_DATA(pInfo->lastValue, double, pCtx->inputType, GET_INPUT_CHAR_INDEX(pCtx, index)); + + pInfo->hasResult = DATA_SET_FLAG; + pInfo->win.skey = pInfo->lastKey; + notNullElems++; + i += 1; } - - if (pCtx->inputType == TSDB_DATA_TYPE_FLOAT || pCtx->inputType == TSDB_DATA_TYPE_DOUBLE) { - pInfo->dOutput += pInfo->dLastValue * (primaryKey[index] - pInfo->lastKey); - } else { - pInfo->iOutput += pInfo->iLastValue * (primaryKey[index] - pInfo->lastKey); + + // calculate the value of + switch(pCtx->inputType) { + case TSDB_DATA_TYPE_TINYINT: { + int8_t *val = (int8_t*) GET_INPUT_CHAR_INDEX(pCtx, index); + for (; i < size; i++) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + index]; + } + break; + } + case TSDB_DATA_TYPE_SMALLINT: { + int16_t *val = (int16_t*) GET_INPUT_CHAR_INDEX(pCtx, index); + for (; i < size; i++) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + index]; + } + break; + } + case TSDB_DATA_TYPE_INT: { + int32_t *val = (int32_t*) GET_INPUT_CHAR_INDEX(pCtx, index); + for (; i < size; i++) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + index]; + } + break; + } + case TSDB_DATA_TYPE_BIGINT: { + int64_t *val = (int64_t*) GET_INPUT_CHAR_INDEX(pCtx, index); + for (; i < size; i++) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey); + pInfo->lastValue = (double) val[i]; + pInfo->lastKey = primaryKey[i + index]; + } + break; + } + case TSDB_DATA_TYPE_FLOAT: { + float *val = (float*) GET_INPUT_CHAR_INDEX(pCtx, index); + for (; i < size; i++) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + index]; + } + break; + } + case TSDB_DATA_TYPE_DOUBLE: { + double *val = (double*) GET_INPUT_CHAR_INDEX(pCtx, index); + for (; i < size; i++) { + if (pCtx->hasNull && isNull((const char*) &val[i], pCtx->inputType)) { + continue; + } + + pInfo->dOutput += ((val[i] + pInfo->lastValue) / 2) * (primaryKey[i + index] - pInfo->lastKey); + pInfo->lastValue = val[i]; + pInfo->lastKey = primaryKey[i + index]; + } + break; + } + default: assert(0); } - - // record the last key/value - pInfo->lastKey = primaryKey[index]; - setTWALastVal(pCtx, pData, 0, pInfo); - - // pCtx->numOfIteratedElems += 1; - pResInfo->hasResult = DATA_SET_FLAG; - + + // the last interpolated time window value + if (pCtx->end.key != INT64_MIN) { + pInfo->dOutput += ((pInfo->lastValue + pCtx->end.val) / 2) * (pCtx->end.key - pInfo->lastKey); + pInfo->lastValue = pCtx->end.val; + pInfo->lastKey = pCtx->end.key; + } + + pInfo->win.ekey = pInfo->lastKey; + + SET_VAL(pCtx, notNullElems, 1); + + if (notNullElems > 0) { + pResInfo->hasResult = DATA_SET_FLAG; + } + if (pCtx->stableQuery) { memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(STwaInfo)); } @@ -3778,16 +3962,10 @@ static void twa_func_merge(SQLFunctionCtx *pCtx) { } numOfNotNull++; - if (pCtx->inputType >= TSDB_DATA_TYPE_TINYINT && pCtx->inputType <= TSDB_DATA_TYPE_BIGINT) { - pBuf->iOutput += pInput->iOutput; - } else { - pBuf->dOutput += pInput->dOutput; - } - - pBuf->SKey = pInput->SKey; - pBuf->EKey = pInput->EKey; + pBuf->dOutput += pInput->dOutput; + + pBuf->win = pInput->win; pBuf->lastKey = pInput->lastKey; - pBuf->iLastValue = pInput->iLastValue; } SET_VAL(pCtx, numOfNotNull, 1); @@ -3814,21 +3992,17 @@ void twa_function_finalizer(SQLFunctionCtx *pCtx) { SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); STwaInfo *pInfo = (STwaInfo *)GET_ROWCELL_INTERBUF(pResInfo); - assert(pInfo->EKey >= pInfo->lastKey && pInfo->hasResult == pResInfo->hasResult); + assert(pInfo->win.ekey == pInfo->lastKey && pInfo->hasResult == pResInfo->hasResult); if (pInfo->hasResult != DATA_SET_FLAG) { setNull(pCtx->aOutputBuf, TSDB_DATA_TYPE_DOUBLE, sizeof(double)); return; } - if (pInfo->SKey == pInfo->EKey) { - *(double *)pCtx->aOutputBuf = 0; - } else if (pInfo->type >= TSDB_DATA_TYPE_TINYINT && pInfo->type <= TSDB_DATA_TYPE_BIGINT) { - pInfo->iOutput += pInfo->iLastValue * (pInfo->EKey - pInfo->lastKey); - *(double *)pCtx->aOutputBuf = pInfo->iOutput / (double)(pInfo->EKey - pInfo->SKey); + if (pInfo->win.ekey == pInfo->win.skey) { + *(double *)pCtx->aOutputBuf = pInfo->lastValue; } else { - pInfo->dOutput += pInfo->dLastValue * (pInfo->EKey - pInfo->lastKey); - *(double *)pCtx->aOutputBuf = pInfo->dOutput / (pInfo->EKey - pInfo->SKey); + *(double *)pCtx->aOutputBuf = pInfo->dOutput / (pInfo->win.ekey - pInfo->win.skey); } GET_RES_INFO(pCtx)->numOfRes = 1; @@ -4387,11 +4561,11 @@ static void sumrate_finalizer(SQLFunctionCtx *pCtx) { * */ int32_t functionCompatList[] = { - // count, sum, avg, min, max, stddev, percentile, apercentile, first, last - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - // last_row, top, bottom, spread, twa, leastsqr, ts, ts_dummy, tag_dummy, ts_z + // count, sum, avg, min, max, stddev, percentile, apercentile, first, last + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + // last_row,top, bottom, spread, twa, leastsqr, ts, ts_dummy, tag_dummy, ts_z 4, -1, -1, 1, 1, 1, 1, 1, 1, -1, - // tag, colprj, tagprj, arithmetic, diff, first_dist, last_dist, interp rate irate + // tag, colprj, tagprj, arithmetic, diff, first_dist, last_dist, interp rate irate 1, 1, 1, 1, -1, 1, 1, 5, 1, 1, // sum_rate, sum_irate, avg_rate, avg_irate 1, 1, 1, 1, diff --git a/src/client/src/tscLocal.c b/src/client/src/tscLocal.c index 538e652f3c6577098363565a4e76fa637b60709c..4c28adc261f6b83b3c1bbb68d3dd3d56e6007fc9 100644 --- a/src/client/src/tscLocal.c +++ b/src/client/src/tscLocal.c @@ -825,8 +825,11 @@ static int32_t tscProcessClientVer(SSqlObj *pSql) { static int32_t tscProcessServStatus(SSqlObj *pSql) { STscObj* pObj = pSql->pTscObj; - if (pObj->pHb != NULL) { - if (pObj->pHb->res.code == TSDB_CODE_RPC_NETWORK_UNAVAIL) { + SSqlObj* pHb = (SSqlObj*)taosAcquireRef(tscObjRef, pObj->hbrid); + if (pHb != NULL) { + int32_t code = pHb->res.code; + taosReleaseRef(tscObjRef, pObj->hbrid); + if (code == TSDB_CODE_RPC_NETWORK_UNAVAIL) { pSql->res.code = TSDB_CODE_RPC_NETWORK_UNAVAIL; return pSql->res.code; } diff --git a/src/client/src/tscLocalMerge.c b/src/client/src/tscLocalMerge.c index 3c7d46f914dd2632ac97cbcb728898ff9a657e82..7fc5b8debb82d18f3067286eea5571c748229072 100644 --- a/src/client/src/tscLocalMerge.c +++ b/src/client/src/tscLocalMerge.c @@ -198,6 +198,7 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd if (numOfFlush == 0 || numOfBuffer == 0) { tscLocalReducerEnvDestroy(pMemBuffer, pDesc, finalmodel, pFFModel, numOfBuffer); + pCmd->command = TSDB_SQL_RETRIEVE_EMPTY_RESULT; // no result, set the result empty tscDebug("%p retrieved no data", pSql); return; } @@ -330,22 +331,19 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd pReducer->nResultBufSize = pMemBuffer[0]->pageSize * 16; pReducer->pResultBuf = (tFilePage *)calloc(1, pReducer->nResultBufSize + sizeof(tFilePage)); - pReducer->finalRowSize = tscGetResRowLength(pQueryInfo->exprList); pReducer->resColModel = finalmodel; pReducer->resColModel->capacity = pReducer->nResultBufSize; - pReducer->finalModel = pFFModel; - assert(pReducer->finalRowSize > 0); - if (pReducer->finalRowSize > 0) { - pReducer->resColModel->capacity /= pReducer->finalRowSize; + if (finalmodel->rowSize > 0) { + pReducer->resColModel->capacity /= finalmodel->rowSize; } - assert(pReducer->finalRowSize <= pReducer->rowSize); + assert(finalmodel->rowSize > 0 && finalmodel->rowSize <= pReducer->rowSize); pReducer->pFinalRes = calloc(1, pReducer->rowSize * pReducer->resColModel->capacity); if (pReducer->pTempBuffer == NULL || pReducer->discardData == NULL || pReducer->pResultBuf == NULL || - /*pReducer->pBufForInterpo == NULL || */pReducer->pFinalRes == NULL || pReducer->prevRowOfInput == NULL) { + pReducer->pFinalRes == NULL || pReducer->prevRowOfInput == NULL) { tfree(pReducer->pTempBuffer); tfree(pReducer->discardData); tfree(pReducer->pResultBuf); @@ -723,10 +721,16 @@ int32_t tscLocalReducerEnvCreate(SSqlObj *pSql, tExtMemBuffer ***pMemBuffer, tOr // final result depends on the fields number memset(pSchema, 0, sizeof(SSchema) * size); + for (int32_t i = 0; i < size; ++i) { SSqlExpr *pExpr = tscSqlExprGet(pQueryInfo, i); - SSchema *p1 = tscGetTableColumnSchema(pTableMetaInfo->pTableMeta, pExpr->colInfo.colIndex); + SSchema p1 = {0}; + if (pExpr->colInfo.colIndex != TSDB_TBNAME_COLUMN_INDEX) { + p1 = *tscGetTableColumnSchema(pTableMetaInfo->pTableMeta, pExpr->colInfo.colIndex); + } else { + p1 = tGetTableNameColumnSchema(); + } int32_t inter = 0; int16_t type = -1; @@ -745,7 +749,8 @@ int32_t tscLocalReducerEnvCreate(SSqlObj *pSql, tExtMemBuffer ***pMemBuffer, tOr functionId = TSDB_FUNC_LAST; } - getResultDataInfo(p1->type, p1->bytes, functionId, 0, &type, &bytes, &inter, 0, false); + int32_t ret = getResultDataInfo(p1.type, p1.bytes, functionId, 0, &type, &bytes, &inter, 0, false); + assert(ret == TSDB_CODE_SUCCESS); } pSchema[i].type = (uint8_t)type; @@ -920,7 +925,7 @@ static void genFinalResWithoutFill(SSqlRes* pRes, SLocalReducer *pLocalReducer, savePrevRecordAndSetupFillInfo(pLocalReducer, pQueryInfo, pLocalReducer->pFillInfo); } - memcpy(pRes->data, pBeforeFillData->data, (size_t)(pRes->numOfRows * pLocalReducer->finalRowSize)); + memcpy(pRes->data, pBeforeFillData->data, (size_t)(pRes->numOfRows * pLocalReducer->finalModel->rowSize)); pRes->numOfClauseTotal += pRes->numOfRows; pBeforeFillData->num = 0; @@ -1256,7 +1261,7 @@ bool genFinalResults(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool noMoreCur tColModelCompact(pModel, pResBuf, pModel->capacity); if (tscIsSecondStageQuery(pQueryInfo)) { - pLocalReducer->finalRowSize = doArithmeticCalculate(pQueryInfo, pResBuf, pModel->rowSize, pLocalReducer->finalRowSize); + doArithmeticCalculate(pQueryInfo, pResBuf, pModel->rowSize, pLocalReducer->finalModel->rowSize); } #ifdef _DEBUG_VIEW @@ -1627,7 +1632,8 @@ void tscInitResObjForLocalQuery(SSqlObj *pObj, int32_t numOfRes, int32_t rowLen) } int32_t doArithmeticCalculate(SQueryInfo* pQueryInfo, tFilePage* pOutput, int32_t rowSize, int32_t finalRowSize) { - char* pbuf = calloc(1, pOutput->num * rowSize); + int32_t maxRowSize = MAX(rowSize, finalRowSize); + char* pbuf = calloc(1, (size_t)(pOutput->num * maxRowSize)); size_t size = tscNumOfFields(pQueryInfo); SArithmeticSupport arithSup = {0}; @@ -1654,17 +1660,16 @@ int32_t doArithmeticCalculate(SQueryInfo* pQueryInfo, tFilePage* pOutput, int32_ tExprTreeCalcTraverse(arithSup.pArithExpr->pExpr, (int32_t) pOutput->num, pbuf + pOutput->num*offset, &arithSup, TSDB_ORDER_ASC, getArithmeticInputSrc); } else { SSqlExpr* pExpr = pSup->pSqlExpr; - memcpy(pbuf + pOutput->num * offset, pExpr->offset * pOutput->num + pOutput->data, pExpr->resBytes * pOutput->num); + memcpy(pbuf + pOutput->num * offset, pExpr->offset * pOutput->num + pOutput->data, (size_t)(pExpr->resBytes * pOutput->num)); } offset += pSup->field.bytes; } - assert(finalRowSize <= rowSize); - memcpy(pOutput->data, pbuf, pOutput->num * offset); + memcpy(pOutput->data, pbuf, (size_t)(pOutput->num * offset)); tfree(pbuf); tfree(arithSup.data); return offset; -} \ No newline at end of file +} diff --git a/src/client/src/tscParseInsert.c b/src/client/src/tscParseInsert.c index a44a158f93095c752f6aa203eb35e58e1d7ba2a0..18e5b6f0744a06e5d28f2ce5cfa3df1244a0dc89 100644 --- a/src/client/src/tscParseInsert.c +++ b/src/client/src/tscParseInsert.c @@ -630,11 +630,17 @@ int32_t tscAllocateMemIfNeed(STableDataBlocks *pDataBlock, int32_t rowSize, int3 return TSDB_CODE_SUCCESS; } -static void tsSetBlockInfo(SSubmitBlk *pBlocks, const STableMeta *pTableMeta, int32_t numOfRows) { +static int32_t tsSetBlockInfo(SSubmitBlk *pBlocks, const STableMeta *pTableMeta, int32_t numOfRows) { pBlocks->tid = pTableMeta->id.tid; pBlocks->uid = pTableMeta->id.uid; pBlocks->sversion = pTableMeta->sversion; - pBlocks->numOfRows += numOfRows; + + if (pBlocks->numOfRows + numOfRows >= INT16_MAX) { + return TSDB_CODE_TSC_INVALID_SQL; + } else { + pBlocks->numOfRows += numOfRows; + return TSDB_CODE_SUCCESS; + } } // data block is disordered, sort it in ascending order @@ -722,7 +728,11 @@ static int32_t doParseInsertStatement(SSqlObj *pSql, void *pTableList, char **st } SSubmitBlk *pBlocks = (SSubmitBlk *)(dataBuf->pData); - tsSetBlockInfo(pBlocks, pTableMeta, numOfRows); + code = tsSetBlockInfo(pBlocks, pTableMeta, numOfRows); + if (code != TSDB_CODE_SUCCESS) { + tscInvalidSQLErrMsg(pCmd->payload, "too many rows in sql, total number of rows should be less than 32767", *str); + return code; + } dataBuf->vgId = pTableMeta->vgroupInfo.vgId; dataBuf->numOfTables = 1; @@ -790,9 +800,6 @@ static int32_t tscCheckIfCreateTable(char **sqlstr, SSqlObj *pSql) { sql += index; tscAllocPayload(pCmd, sizeof(STagData)); - STagData *pTag = &pCmd->tagData; - - memset(pTag, 0, sizeof(STagData)); //the source super table is moved to the secondary position of the pTableMetaInfo list if (pQueryInfo->numOfTables < 2) { @@ -805,7 +812,14 @@ static int32_t tscCheckIfCreateTable(char **sqlstr, SSqlObj *pSql) { return code; } + STagData *pTag = realloc(pCmd->pTagData, offsetof(STagData, data)); + if (pTag == NULL) { + return TSDB_CODE_TSC_OUT_OF_MEMORY; + } + memset(pTag, 0, offsetof(STagData, data)); tstrncpy(pTag->name, pSTableMeterMetaInfo->name, sizeof(pTag->name)); + pCmd->pTagData = pTag; + code = tscGetTableMeta(pSql, pSTableMeterMetaInfo); if (code != TSDB_CODE_SUCCESS) { return code; @@ -934,7 +948,13 @@ static int32_t tscCheckIfCreateTable(char **sqlstr, SSqlObj *pSql) { return TSDB_CODE_TSC_OUT_OF_MEMORY; } tdSortKVRowByColIdx(row); - pTag->dataLen = kvRowLen(row); + + pTag = (STagData*)realloc(pCmd->pTagData, offsetof(STagData, data) + kvRowLen(row)); + if (pTag == NULL) { + return TSDB_CODE_TSC_OUT_OF_MEMORY; + } + pCmd->pTagData = pTag; + pTag->dataLen = htonl(kvRowLen(row)); kvRowCpy(pTag->data, row); free(row); @@ -945,8 +965,6 @@ static int32_t tscCheckIfCreateTable(char **sqlstr, SSqlObj *pSql) { return tscSQLSyntaxErrMsg(pCmd->payload, ") expected", sToken.z); } - pTag->dataLen = htonl(pTag->dataLen); - if (tscValidateName(&tableToken) != TSDB_CODE_SUCCESS) { return tscInvalidSQLErrMsg(pCmd->payload, "invalid table name", *sqlstr); } @@ -1376,7 +1394,10 @@ static int doPackSendDataBlock(SSqlObj *pSql, int32_t numOfRows, STableDataBlock STableMeta *pTableMeta = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0)->pTableMeta; SSubmitBlk *pBlocks = (SSubmitBlk *)(pTableDataBlocks->pData); - tsSetBlockInfo(pBlocks, pTableMeta, numOfRows); + code = tsSetBlockInfo(pBlocks, pTableMeta, numOfRows); + if (code != TSDB_CODE_SUCCESS) { + return tscInvalidSQLErrMsg(pCmd->payload, "too many rows in sql, total number of rows should be less than 32767", NULL); + } if ((code = tscMergeTableDataBlocks(pSql, pCmd->pDataBlocks)) != TSDB_CODE_SUCCESS) { return code; diff --git a/src/client/src/tscProfile.c b/src/client/src/tscProfile.c index acc5acd786bfe00036e539a104da04af1485a263..18fc79c4748f84d4ec073e4bc678e3a067afb025 100644 --- a/src/client/src/tscProfile.c +++ b/src/client/src/tscProfile.c @@ -39,6 +39,7 @@ void tscInitConnCb(void *param, TAOS_RES *result, int code) { tscSlowQueryConnInitialized = true; tscSaveSlowQueryFp(sql, NULL); } + taos_free_result(result); } void tscAddIntoSqlList(SSqlObj *pSql) { @@ -69,6 +70,7 @@ void tscSaveSlowQueryFpCb(void *param, TAOS_RES *result, int code) { } else { tscDebug("success to save slow query, code:%d", code); } + taos_free_result(result); } void tscSaveSlowQueryFp(void *handle, void *tmrId) { diff --git a/src/client/src/tscSQLParser.c b/src/client/src/tscSQLParser.c index 743cb42eb3d0c85d7c77270dca85e225ff4b72c5..91b00e0109b0551b3bde52e326966ff396921019 100644 --- a/src/client/src/tscSQLParser.c +++ b/src/client/src/tscSQLParser.c @@ -2835,6 +2835,10 @@ static bool functionCompatibleCheck(SQueryInfo* pQueryInfo, bool joinQuery) { if (functionCompatList[functionId] != factor) { return false; + } else { + if (factor == -1) { // two functions with the same -1 flag + return false; + } } if (functionId == TSDB_FUNC_LAST_ROW && joinQuery) { @@ -4248,7 +4252,7 @@ static int32_t getTagQueryCondExpr(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SCondE tExprTreeDestroy(&p, NULL); taosArrayDestroy(colList); - if (taosArrayGetSize(pQueryInfo->tagCond.pCond) > 0 && !UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { + if (pQueryInfo->tagCond.pCond != NULL && taosArrayGetSize(pQueryInfo->tagCond.pCond) > 0 && !UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), "filter on tag not supported for normal table"); } } @@ -4256,6 +4260,7 @@ static int32_t getTagQueryCondExpr(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SCondE pCondExpr->pTagCond = NULL; return ret; } + int32_t parseWhereClause(SQueryInfo* pQueryInfo, tSQLExpr** pExpr, SSqlObj* pSql) { if (pExpr == NULL) { return TSDB_CODE_SUCCESS; @@ -5102,7 +5107,7 @@ int32_t validateDNodeConfig(tDCLSQL* pOptions) { const int tokenDebugFlagEnd = 20; const SDNodeDynConfOption cfgOptions[] = { {"resetLog", 8}, {"resetQueryCache", 15}, {"balance", 7}, {"monitor", 7}, - {"debugFlag", 9}, {"monitorDebugFlag", 16}, {"vDebugFlag", 10}, {"mDebugFlag", 10}, + {"debugFlag", 9}, {"monDebugFlag", 12}, {"vDebugFlag", 10}, {"mDebugFlag", 10}, {"cDebugFlag", 10}, {"httpDebugFlag", 13}, {"qDebugflag", 10}, {"sdbDebugFlag", 12}, {"uDebugFlag", 10}, {"tsdbDebugFlag", 13}, {"sDebugflag", 10}, {"rpcDebugFlag", 12}, {"dDebugFlag", 10}, {"mqttDebugFlag", 13}, {"wDebugFlag", 10}, {"tmrDebugFlag", 12}, @@ -5306,15 +5311,18 @@ int32_t parseLimitClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t clauseIn // keep original limitation value in globalLimit pQueryInfo->clauseLimit = pQueryInfo->limit.limit; - pQueryInfo->prjOffset = pQueryInfo->limit.offset; + pQueryInfo->prjOffset = pQueryInfo->limit.offset; + pQueryInfo->tableLimit = -1; if (tscOrderedProjectionQueryOnSTable(pQueryInfo, 0)) { /* - * the limitation/offset value should be removed during retrieve data from virtual node, - * since the global order are done in client side, so the limitation should also - * be done at the client side. + * the offset value should be removed during retrieve data from virtual node, since the + * global order are done in client side, so the offset is applied at the client side + * However, note that the maximum allowed number of result for each table should be less + * than or equal to the value of limit. */ if (pQueryInfo->limit.limit > 0) { + pQueryInfo->tableLimit = pQueryInfo->limit.limit + pQueryInfo->limit.offset; pQueryInfo->limit.limit = -1; } @@ -6648,7 +6656,7 @@ int32_t exprTreeFromSqlExpr(SSqlCmd* pCmd, tExprNode **pExpr, const tSQLExpr* pS return TSDB_CODE_SUCCESS; } else { - return TSDB_CODE_TSC_INVALID_SQL; + return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), "not support filter expression"); } } else { diff --git a/src/client/src/tscServer.c b/src/client/src/tscServer.c index 15859674ab9c891814b26fe054a478de8b398ca7..59bcdd691d6a8cfe2f2e65da0c87da9c232fee57 100644 --- a/src/client/src/tscServer.c +++ b/src/client/src/tscServer.c @@ -26,7 +26,7 @@ #include "ttimer.h" #include "tlockfree.h" -SRpcCorEpSet tscMgmtEpSet; +///SRpcCorEpSet tscMgmtEpSet; int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo) = {0}; @@ -73,10 +73,11 @@ static void tscSetDnodeEpSet(SSqlObj* pSql, SVgroupInfo* pVgroupInfo) { assert(hasFqdn); } -static void tscDumpMgmtEpSet(SRpcEpSet *epSet) { - taosCorBeginRead(&tscMgmtEpSet.version); - *epSet = tscMgmtEpSet.epSet; - taosCorEndRead(&tscMgmtEpSet.version); +static void tscDumpMgmtEpSet(SSqlObj *pSql) { + SRpcCorEpSet *pCorEpSet = pSql->pTscObj->tscCorMgmtEpSet; + taosCorBeginRead(&pCorEpSet->version); + pSql->epSet = pCorEpSet->epSet; + taosCorEndRead(&pCorEpSet->version); } static void tscEpSetHtons(SRpcEpSet *s) { for (int32_t i = 0; i < s->numOfEps; i++) { @@ -94,11 +95,12 @@ bool tscEpSetIsEqual(SRpcEpSet *s1, SRpcEpSet *s2) { } return true; } -void tscUpdateMgmtEpSet(SRpcEpSet *pEpSet) { +void tscUpdateMgmtEpSet(SSqlObj *pSql, SRpcEpSet *pEpSet) { // no need to update if equal - taosCorBeginWrite(&tscMgmtEpSet.version); - tscMgmtEpSet.epSet = *pEpSet; - taosCorEndWrite(&tscMgmtEpSet.version); + SRpcCorEpSet *pCorEpSet = pSql->pTscObj->tscCorMgmtEpSet; + taosCorBeginWrite(&pCorEpSet->version); + pCorEpSet->epSet = *pEpSet; + taosCorEndWrite(&pCorEpSet->version); } static void tscDumpEpSetFromVgroupInfo(SCorVgroupInfo *pVgroupInfo, SRpcEpSet *pEpSet) { if (pVgroupInfo == NULL) { return;} @@ -133,18 +135,6 @@ static void tscUpdateVgroupInfo(SSqlObj *pObj, SRpcEpSet *pEpSet) { taosCorEndWrite(&pVgroupInfo->version); } -void tscPrintMgmtEp() { - SRpcEpSet dump; - tscDumpMgmtEpSet(&dump); - if (dump.numOfEps <= 0) { - tscError("invalid mnode EP list:%d", dump.numOfEps); - } else { - for (int i = 0; i < dump.numOfEps; ++i) { - tscDebug("mnode index:%d %s:%d", i, dump.fqdn[i], dump.port[i]); - } - } -} - void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) { STscObj *pObj = (STscObj *)param; if (pObj == NULL) return; @@ -162,7 +152,7 @@ void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) { SRpcEpSet * epSet = &pRsp->epSet; if (epSet->numOfEps > 0) { tscEpSetHtons(epSet); - tscUpdateMgmtEpSet(epSet); + tscUpdateMgmtEpSet(pSql, epSet); } pSql->pTscObj->connId = htonl(pRsp->connId); @@ -175,10 +165,10 @@ void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) { if (pRsp->streamId) tscKillStream(pObj, htonl(pRsp->streamId)); } } else { - tscDebug("%p heartbeat failed, code:%s", pObj->pHb, tstrerror(code)); + tscDebug("%" PRId64 " heartbeat failed, code:%s", pObj->hbrid, tstrerror(code)); } - if (pObj->pHb != NULL) { + if (pObj->hbrid != 0) { int32_t waitingDuring = tsShellActivityTimer * 500; tscDebug("%p send heartbeat in %dms", pSql, waitingDuring); @@ -191,22 +181,21 @@ void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) { void tscProcessActivityTimer(void *handle, void *tmrId) { int64_t rid = (int64_t) handle; STscObj *pObj = taosAcquireRef(tscRefId, rid); - if (pObj == NULL) return; - - SSqlObj* pHB = pObj->pHb; + if (pObj == NULL) { + return; + } - void** p = taosCacheAcquireByKey(tscObjCache, &pHB, sizeof(TSDB_CACHE_PTR_TYPE)); - if (p == NULL) { - tscWarn("%p HB object has been released already", pHB); - taosReleaseRef(tscRefId, pObj->rid); + SSqlObj* pHB = taosAcquireRef(tscObjRef, pObj->hbrid); + if (pHB == NULL) { + taosReleaseRef(tscRefId, rid); return; } - assert(*pHB->self == pHB); + assert(pHB->self == pObj->hbrid); pHB->retry = 0; int32_t code = tscProcessSql(pHB); - taosCacheRelease(tscObjCache, (void**) &p, false); + taosReleaseRef(tscObjRef, pObj->hbrid); if (code != TSDB_CODE_SUCCESS) { tscError("%p failed to sent HB to server, reason:%s", pHB, tstrerror(code)); @@ -227,7 +216,7 @@ int tscSendMsgToServer(SSqlObj *pSql) { // set the mgmt ip list if (pSql->cmd.command >= TSDB_SQL_MGMT) { - tscDumpMgmtEpSet(&pSql->epSet); + tscDumpMgmtEpSet(pSql); } memcpy(pMsg, pSql->cmd.payload, pSql->cmd.payloadLen); @@ -236,7 +225,7 @@ int tscSendMsgToServer(SSqlObj *pSql) { .msgType = pSql->cmd.msgType, .pCont = pMsg, .contLen = pSql->cmd.payloadLen, - .ahandle = pSql, + .ahandle = (void*)pSql->self, .handle = NULL, .code = 0 }; @@ -247,26 +236,24 @@ int tscSendMsgToServer(SSqlObj *pSql) { void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcEpSet *pEpSet) { TSDB_CACHE_PTR_TYPE handle = (TSDB_CACHE_PTR_TYPE) rpcMsg->ahandle; - void** p = taosCacheAcquireByKey(tscObjCache, &handle, sizeof(TSDB_CACHE_PTR_TYPE)); - if (p == NULL) { + SSqlObj* pSql = (SSqlObj*)taosAcquireRef(tscObjRef, handle); + if (pSql == NULL) { rpcFreeCont(rpcMsg->pCont); return; } - - SSqlObj* pSql = *p; - assert(pSql != NULL); + assert(pSql->self == handle); STscObj *pObj = pSql->pTscObj; SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; - assert(*pSql->self == pSql); pSql->rpcRid = -1; if (pObj->signature != pObj) { tscDebug("%p DB connection is closed, cmd:%d pObj:%p signature:%p", pSql, pCmd->command, pObj, pObj->signature); - taosCacheRelease(tscObjCache, (void**) &p, true); + taosRemoveRef(tscObjRef, pSql->self); + taosReleaseRef(tscObjRef, pSql->self); rpcFreeCont(rpcMsg->pCont); return; } @@ -276,10 +263,8 @@ void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcEpSet *pEpSet) { tscDebug("%p sqlObj needs to be released or DB connection is closed, cmd:%d type:%d, pObj:%p signature:%p", pSql, pCmd->command, pQueryInfo->type, pObj, pObj->signature); - void** p1 = p; - taosCacheRelease(tscObjCache, (void**) &p1, false); - - taosCacheRelease(tscObjCache, (void**) &p, true); + taosRemoveRef(tscObjRef, pSql->self); + taosReleaseRef(tscObjRef, pSql->self); rpcFreeCont(rpcMsg->pCont); return; } @@ -289,7 +274,7 @@ void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcEpSet *pEpSet) { if (pCmd->command < TSDB_SQL_MGMT) { tscUpdateVgroupInfo(pSql, pEpSet); } else { - tscUpdateMgmtEpSet(pEpSet); + tscUpdateMgmtEpSet(pSql, pEpSet); } } } @@ -322,7 +307,7 @@ void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcEpSet *pEpSet) { // if there is an error occurring, proceed to the following error handling procedure. if (rpcMsg->code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) { - taosCacheRelease(tscObjCache, (void**) &p, false); + taosReleaseRef(tscObjRef, pSql->self); rpcFreeCont(rpcMsg->pCont); return; } @@ -390,11 +375,10 @@ void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcEpSet *pEpSet) { (*pSql->fp)(pSql->param, pSql, rpcMsg->code); } - void** p1 = p; - taosCacheRelease(tscObjCache, (void**) &p1, false); + taosReleaseRef(tscObjRef, pSql->self); if (shouldFree) { // in case of table-meta/vgrouplist query, automatically free it - taosCacheRelease(tscObjCache, (void **)&p, true); + taosRemoveRef(tscObjRef, pSql->self); tscDebug("%p sqlObj is automatically freed", pSql); } @@ -697,6 +681,7 @@ int tscBuildQueryMsg(SSqlObj *pSql, SSqlInfo *pInfo) { pQueryMsg->tagNameRelType = htons(pQueryInfo->tagCond.relType); pQueryMsg->numOfTags = htonl(numOfTags); pQueryMsg->queryType = htonl(pQueryInfo->type); + pQueryMsg->tableLimit = htobe64(pQueryInfo->tableLimit); size_t numOfOutput = tscSqlExprNumOfExprs(pQueryInfo); pQueryMsg->numOfOutput = htons((int16_t)numOfOutput); // this is the stage one output column number @@ -1565,11 +1550,11 @@ int tscBuildTableMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { char *pMsg = (char *)pInfoMsg + sizeof(STableInfoMsg); - size_t len = htonl(pCmd->tagData.dataLen); - if (pSql->cmd.autoCreated) { + if (pCmd->autoCreated && pCmd->pTagData != NULL) { + int len = htonl(pCmd->pTagData->dataLen); if (len > 0) { - len += sizeof(pCmd->tagData.name) + sizeof(pCmd->tagData.dataLen); - memcpy(pInfoMsg->tags, &pCmd->tagData, len); + len += sizeof(pCmd->pTagData->name) + sizeof(pCmd->pTagData->dataLen); + memcpy(pInfoMsg->tags, pCmd->pTagData, len); pMsg += len; } } @@ -2020,7 +2005,7 @@ int tscProcessShowRsp(SSqlObj *pSql) { // TODO multithread problem static void createHBObj(STscObj* pObj) { - if (pObj->pHb != NULL) { + if (pObj->hbrid != 0) { return; } @@ -2052,7 +2037,7 @@ static void createHBObj(STscObj* pObj) { registerSqlObj(pSql); tscDebug("%p HB is allocated, pObj:%p", pSql, pObj); - pObj->pHb = pSql; + pObj->hbrid = pSql->self; } int tscProcessConnectRsp(SSqlObj *pSql) { @@ -2070,7 +2055,7 @@ int tscProcessConnectRsp(SSqlObj *pSql) { if (pConnect->epSet.numOfEps > 0) { tscEpSetHtons(&pConnect->epSet); - tscUpdateMgmtEpSet(&pConnect->epSet); + tscUpdateMgmtEpSet(pSql, &pConnect->epSet); } strcpy(pObj->sversion, pConnect->serverVersion); @@ -2239,8 +2224,6 @@ static int32_t getTableMetaFromMgmt(SSqlObj *pSql, STableMetaInfo *pTableMetaInf pNew->signature = pNew; pNew->cmd.command = TSDB_SQL_META; - registerSqlObj(pNew); - tscAddSubqueryInfo(&pNew->cmd); SQueryInfo *pNewQueryInfo = tscGetQueryInfoDetailSafely(&pNew->cmd, 0); @@ -2248,8 +2231,7 @@ static int32_t getTableMetaFromMgmt(SSqlObj *pSql, STableMetaInfo *pTableMetaInf pNew->cmd.autoCreated = pSql->cmd.autoCreated; // create table if not exists if (TSDB_CODE_SUCCESS != tscAllocPayload(&pNew->cmd, TSDB_DEFAULT_PAYLOAD_SIZE + pSql->cmd.payloadLen)) { tscError("%p malloc failed for payload to get table meta", pSql); - free(pNew); - + tscFreeSqlObj(pNew); return TSDB_CODE_TSC_OUT_OF_MEMORY; } @@ -2257,12 +2239,25 @@ static int32_t getTableMetaFromMgmt(SSqlObj *pSql, STableMetaInfo *pTableMetaInf assert(pNew->cmd.numOfClause == 1 && pNewQueryInfo->numOfTables == 1); tstrncpy(pNewMeterMetaInfo->name, pTableMetaInfo->name, sizeof(pNewMeterMetaInfo->name)); - memcpy(&pNew->cmd.tagData, &pSql->cmd.tagData, sizeof(pSql->cmd.tagData)); + + if (pSql->cmd.pTagData != NULL) { + int size = offsetof(STagData, data) + htonl(pSql->cmd.pTagData->dataLen); + pNew->cmd.pTagData = calloc(1, size); + if (pNew->cmd.pTagData == NULL) { + tscError("%p malloc failed for new tag data to get table meta", pSql); + tscFreeSqlObj(pNew); + return TSDB_CODE_TSC_OUT_OF_MEMORY; + } + memcpy(pNew->cmd.pTagData, pSql->cmd.pTagData, size); + } + tscDebug("%p new pSqlObj:%p to get tableMeta, auto create:%d", pSql, pNew, pNew->cmd.autoCreated); pNew->fp = tscTableMetaCallBack; pNew->param = pSql; + registerSqlObj(pNew); + int32_t code = tscProcessSql(pNew); if (code == TSDB_CODE_SUCCESS) { code = TSDB_CODE_TSC_ACTION_IN_PROGRESS; // notify upper application that current process need to be terminated diff --git a/src/client/src/tscSql.c b/src/client/src/tscSql.c index de0177647234d0b63fa21a3b0fe2ec5763f40da3..fae5b5856fb8804ac9d8d9d67eb1592db16bdaff 100644 --- a/src/client/src/tscSql.c +++ b/src/client/src/tscSql.c @@ -58,6 +58,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa terrno = TSDB_CODE_TSC_INVALID_USER_LENGTH; return NULL; } + SRpcCorEpSet corMgmtEpSet; char secretEncrypt[32] = {0}; int secretEncryptLen = 0; @@ -82,11 +83,13 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa } secretEncryptLen = outlen; } - + if (ip) { - if (tscSetMgmtEpSetFromCfg(ip, NULL) < 0) return NULL; - if (port) tscMgmtEpSet.epSet.port[0] = port; - } + if (tscSetMgmtEpSetFromCfg(ip, NULL, &corMgmtEpSet) < 0) return NULL; + if (port) corMgmtEpSet.epSet.port[0] = port; + } else { + if (tscSetMgmtEpSetFromCfg(tsFirst, tsSecond, &corMgmtEpSet) < 0) return NULL; + } void *pDnodeConn = NULL; if (tscInitRpc(user, secretEncrypt, &pDnodeConn) != 0) { @@ -100,10 +103,20 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa rpcClose(pDnodeConn); return NULL; } + // set up tscObj's mgmtEpSet + pObj->tscCorMgmtEpSet = (SRpcCorEpSet *)malloc(sizeof(SRpcCorEpSet)); + if (NULL == pObj->tscCorMgmtEpSet) { + terrno = TSDB_CODE_TSC_OUT_OF_MEMORY; + rpcClose(pDnodeConn); + free(pObj->tscCorMgmtEpSet); + free(pObj); + } + memcpy(pObj->tscCorMgmtEpSet, &corMgmtEpSet, sizeof(SRpcCorEpSet)); pObj->signature = pObj; pObj->pDnodeConn = pDnodeConn; T_REF_INIT_VAL(pObj, 1); + tstrncpy(pObj->user, user, sizeof(pObj->user)); secretEncryptLen = MIN(secretEncryptLen, sizeof(pObj->pass)); @@ -115,6 +128,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa if (len >= TSDB_DB_NAME_LEN) { terrno = TSDB_CODE_TSC_INVALID_DB_LENGTH; rpcClose(pDnodeConn); + free(pObj->tscCorMgmtEpSet); free(pObj); return NULL; } @@ -132,6 +146,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa if (NULL == pSql) { terrno = TSDB_CODE_TSC_OUT_OF_MEMORY; rpcClose(pDnodeConn); + free(pObj->tscCorMgmtEpSet); free(pObj); return NULL; } @@ -149,6 +164,7 @@ static SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pa terrno = TSDB_CODE_TSC_OUT_OF_MEMORY; rpcClose(pDnodeConn); free(pSql); + free(pObj->tscCorMgmtEpSet); free(pObj); return NULL; } @@ -276,15 +292,18 @@ void taos_close(TAOS *taos) { pObj->signature = NULL; taosTmrStopA(&(pObj->pTimer)); - SSqlObj* pHb = pObj->pHb; - if (pHb != NULL && atomic_val_compare_exchange_ptr(&pObj->pHb, pHb, 0) == pHb) { - if (pHb->rpcRid > 0) { // wait for rsp from dnode - rpcCancelRequest(pHb->rpcRid); - pHb->rpcRid = -1; - } + if (pObj->hbrid > 0) { + SSqlObj* pHb = (SSqlObj*)taosAcquireRef(tscObjRef, pObj->hbrid); + if (pHb != NULL) { + if (pHb->rpcRid > 0) { // wait for rsp from dnode + rpcCancelRequest(pHb->rpcRid); + pHb->rpcRid = -1; + } - tscDebug("%p HB is freed", pHb); - taos_free_result(pHb); + tscDebug("%p HB is freed", pHb); + taos_free_result(pHb); + taosReleaseRef(tscObjRef, pHb->self); + } } int32_t ref = T_REF_DEC(pObj); @@ -606,8 +625,7 @@ void taos_free_result(TAOS_RES *res) { bool freeNow = tscKillQueryInDnode(pSql); if (freeNow) { tscDebug("%p free sqlObj in cache", pSql); - SSqlObj** p = pSql->self; - taosCacheRelease(tscObjCache, (void**) &p, true); + taosReleaseRef(tscObjRef, pSql->self); } } @@ -700,13 +718,7 @@ static void tscKillSTableQuery(SSqlObj *pSql) { continue; } - void** p = taosCacheAcquireByKey(tscObjCache, &pSub, sizeof(TSDB_CACHE_PTR_TYPE)); - if (p == NULL) { - continue; - } - - SSqlObj* pSubObj = (SSqlObj*) (*p); - assert(pSubObj->self == (SSqlObj**) p); + SSqlObj* pSubObj = pSub; pSubObj->res.code = TSDB_CODE_TSC_QUERY_CANCELLED; if (pSubObj->rpcRid > 0) { @@ -715,7 +727,7 @@ static void tscKillSTableQuery(SSqlObj *pSql) { } tscQueueAsyncRes(pSubObj); - taosCacheRelease(tscObjCache, (void**) &p, false); + taosReleaseRef(tscObjRef, pSubObj->self); } tscDebug("%p super table query cancelled", pSql); diff --git a/src/client/src/tscSub.c b/src/client/src/tscSub.c index 2c81bd7c7ce25535dfffe9c1fd6f8a0197fb900b..52b74f7502db139f18d8230eefb36d41bae4079f 100644 --- a/src/client/src/tscSub.c +++ b/src/client/src/tscSub.c @@ -157,7 +157,7 @@ static SSub* tscCreateSubscription(STscObj* pObj, const char* topic, const char* registerSqlObj(pSql); - code = tsParseSql(pSql, false); + code = tsParseSql(pSql, true); if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) { tsem_wait(&pSub->sem); code = pSql->res.code; @@ -168,7 +168,7 @@ static SSub* tscCreateSubscription(STscObj* pObj, const char* topic, const char* goto fail; } - if (pSql->cmd.command != TSDB_SQL_SELECT) { + if (pSql->cmd.command != TSDB_SQL_SELECT && pSql->cmd.command != TSDB_SQL_RETRIEVE_EMPTY_RESULT) { line = __LINE__; code = TSDB_CODE_TSC_INVALID_SQL; goto fail; @@ -179,10 +179,10 @@ static SSub* tscCreateSubscription(STscObj* pObj, const char* topic, const char* fail: tscError("tscCreateSubscription failed at line %d, reason: %s", line, tstrerror(code)); if (pSql != NULL) { - if (pSql->self != NULL) { - taos_free_result(pSql); + if (pSql->self != 0) { + taosReleaseRef(tscObjRef, pSql->self); } else { - tscFreeSqlObj(pSql); + tscFreeSqlObj(pSql); } pSql = NULL; @@ -401,9 +401,11 @@ TAOS_SUB *taos_subscribe(TAOS *taos, int restart, const char* topic, const char tscLoadSubscriptionProgress(pSub); } - if (!tscUpdateSubscription(pObj, pSub)) { - taos_unsubscribe(pSub, 1); - return NULL; + if (pSub->pSql->cmd.command == TSDB_SQL_SELECT) { + if (!tscUpdateSubscription(pObj, pSub)) { + taos_unsubscribe(pSub, 1); + return NULL; + } } pSub->interval = interval; @@ -417,10 +419,80 @@ TAOS_SUB *taos_subscribe(TAOS *taos, int restart, const char* topic, const char return pSub; } +SSqlObj* recreateSqlObj(SSub* pSub) { + SSqlObj* pSql = calloc(1, sizeof(SSqlObj)); + if (pSql == NULL) { + return NULL; + } + + pSql->signature = pSql; + pSql->pTscObj = pSub->taos; + + SSqlCmd* pCmd = &pSql->cmd; + SSqlRes* pRes = &pSql->res; + if (tsem_init(&pSql->rspSem, 0, 0) == -1) { + tscFreeSqlObj(pSql); + return NULL; + } + + pSql->param = pSub; + pSql->maxRetry = TSDB_MAX_REPLICA; + pSql->fp = asyncCallback; + pSql->fetchFp = asyncCallback; + pSql->sqlstr = strdup(pSub->pSql->sqlstr); + if (pSql->sqlstr == NULL) { + tscFreeSqlObj(pSql); + return NULL; + } + + pRes->qhandle = 0; + pRes->numOfRows = 1; + + int code = tscAllocPayload(pCmd, TSDB_DEFAULT_PAYLOAD_SIZE); + if (code != TSDB_CODE_SUCCESS) { + tscFreeSqlObj(pSql); + return NULL; + } + + registerSqlObj(pSql); + + code = tsParseSql(pSql, true); + if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) { + tsem_wait(&pSub->sem); + code = pSql->res.code; + } + + if (code != TSDB_CODE_SUCCESS) { + taosReleaseRef(tscObjRef, pSql->self); + return NULL; + } + + if (pSql->cmd.command != TSDB_SQL_SELECT) { + taosReleaseRef(tscObjRef, pSql->self); + return NULL; + } + + return pSql; +} + TAOS_RES *taos_consume(TAOS_SUB *tsub) { SSub *pSub = (SSub *)tsub; if (pSub == NULL) return NULL; + if (pSub->pSql->cmd.command == TSDB_SQL_RETRIEVE_EMPTY_RESULT) { + SSqlObj* pSql = recreateSqlObj(pSub); + if (pSql == NULL) { + return NULL; + } + if (pSub->pSql->self != 0) { + taosReleaseRef(tscObjRef, pSub->pSql->self); + } else { + tscFreeSqlObj(pSub->pSql); + } + pSub->pSql = pSql; + pSql->pSubscription = pSub; + } + tscSaveSubscriptionProgress(pSub); SSqlObj *pSql = pSub->pSql; @@ -442,7 +514,12 @@ TAOS_RES *taos_consume(TAOS_SUB *tsub) { size_t size = taosArrayGetSize(pSub->progress) * sizeof(STableIdInfo); size += sizeof(SQueryTableMsg) + 4096; - tscAllocPayload(&pSql->cmd, (int)size); + int code = tscAllocPayload(&pSql->cmd, (int)size); + if (code != TSDB_CODE_SUCCESS) { + tscError("failed to alloc payload"); + return NULL; + } + for (int retry = 0; retry < 3; retry++) { tscRemoveFromSqlList(pSql); @@ -507,10 +584,13 @@ void taos_unsubscribe(TAOS_SUB *tsub, int keepProgress) { } if (pSub->pSql != NULL) { - taos_free_result(pSub->pSql); + if (pSub->pSql->self != 0) { + taosReleaseRef(tscObjRef, pSub->pSql->self); + } else { + tscFreeSqlObj(pSub->pSql); + } } - tscFreeSqlObj(pSub->pSql); taosArrayDestroy(pSub->progress); tsem_destroy(&pSub->sem); memset(pSub, 0, sizeof(*pSub)); diff --git a/src/client/src/tscSubquery.c b/src/client/src/tscSubquery.c index 6ebbeeef411f8922662e7045027cf55410cbf89d..819a323db5e947df46158b9665c0f5d5c491f532 100644 --- a/src/client/src/tscSubquery.c +++ b/src/client/src/tscSubquery.c @@ -2198,6 +2198,9 @@ int32_t tscHandleInsertRetry(SSqlObj* pSql) { STableDataBlocks* pTableDataBlock = taosArrayGetP(pCmd->pDataBlocks, pSupporter->index); int32_t code = tscCopyDataBlockToPayload(pSql, pTableDataBlock); + // free the data block created from insert sql string + pCmd->pDataBlocks = tscDestroyBlockArrayList(pCmd->pDataBlocks); + if ((pRes->code = code)!= TSDB_CODE_SUCCESS) { tscQueueAsyncRes(pSql); return code; // here the pSql may have been released already. diff --git a/src/client/src/tscSystem.c b/src/client/src/tscSystem.c index 9e9a00550a4497ae11cbdffe48991c2fd7b742d6..03b6ac8404c1f3d55a1afce81342c497101a6589 100644 --- a/src/client/src/tscSystem.c +++ b/src/client/src/tscSystem.c @@ -15,7 +15,7 @@ #include "os.h" #include "taosmsg.h" -#include "tcache.h" +#include "tref.h" #include "trpc.h" #include "tsystem.h" #include "ttimer.h" @@ -31,7 +31,7 @@ // global, not configurable SCacheObj* tscMetaCache; -SCacheObj* tscObjCache; +int tscObjRef = -1; void * tscTmr; void * tscQhandle; void * tscCheckDiskUsageTmr; @@ -116,11 +116,6 @@ void taos_init_imp(void) { taosInitNote(tsNumOfLogLines / 10, 1, (char*)"tsc_note"); } - if (tscSetMgmtEpSetFromCfg(tsFirst, tsSecond) < 0) { - tscError("failed to init mnode EP list"); - return; - } - tscInitMsgsFp(); int queueSize = tsMaxConnections*2; @@ -144,7 +139,7 @@ void taos_init_imp(void) { int64_t refreshTime = 10; // 10 seconds by default if (tscMetaCache == NULL) { tscMetaCache = taosCacheInit(TSDB_DATA_TYPE_BINARY, refreshTime, false, tscFreeTableMetaHelper, "tableMeta"); - tscObjCache = taosCacheInit(TSDB_CACHE_PTR_KEY, refreshTime / 2, false, tscFreeRegisteredSqlObj, "sqlObj"); + tscObjRef = taosOpenRef(40960, tscFreeRegisteredSqlObj); } tscRefId = taosOpenRef(200, tscCloseTscObj); @@ -167,9 +162,9 @@ void taos_cleanup(void) { taosCacheCleanup(m); } - m = tscObjCache; - if (m != NULL && atomic_val_compare_exchange_ptr(&tscObjCache, m, 0) == m) { - taosCacheCleanup(m); + int refId = atomic_exchange_32(&tscObjRef, -1); + if (refId != -1) { + taosCloseRef(refId); } m = tscQhandle; diff --git a/src/client/src/tscUtil.c b/src/client/src/tscUtil.c index a75f6d829615fac6dda3b5d654374d06cf699e8a..a98132d319cf5403503c415d99c5fd0a72fc1941 100644 --- a/src/client/src/tscUtil.c +++ b/src/client/src/tscUtil.c @@ -408,7 +408,7 @@ void tscResetSqlCmdObj(SSqlCmd* pCmd, bool removeFromCache) { pCmd->pTableList = NULL; pCmd->pDataBlocks = tscDestroyBlockArrayList(pCmd->pDataBlocks); - + tscFreeQueryInfo(pCmd, removeFromCache); } @@ -447,20 +447,18 @@ static void tscFreeSubobj(SSqlObj* pSql) { void tscFreeRegisteredSqlObj(void *pSql) { assert(pSql != NULL); - SSqlObj** p = (SSqlObj**)pSql; - STscObj* pTscObj = (*p)->pTscObj; - - assert((*p)->self != 0 && (*p)->self == (p)); + SSqlObj* p = *(SSqlObj**)pSql; + STscObj* pTscObj = p->pTscObj; - SSqlObj* ptr = *p; - tscFreeSqlObj(*p); + assert(p->self != 0); + tscFreeSqlObj(p); int32_t ref = T_REF_DEC(pTscObj); assert(ref >= 0); - tscDebug("%p free sqlObj completed, tscObj:%p ref:%d", ptr, pTscObj, ref); + tscDebug("%p free sqlObj completed, tscObj:%p ref:%d", p, pTscObj, ref); if (ref == 0) { - tscDebug("%p all sqlObj freed, free tscObj:%p", ptr, pTscObj); + tscDebug("%p all sqlObj freed, free tscObj:%p", p, pTscObj); taosRemoveRef(tscRefId, pTscObj->rid); } } @@ -512,6 +510,8 @@ void tscFreeSqlObj(SSqlObj* pSql) { tscFreeSqlResult(pSql); tscResetSqlCmdObj(pCmd, false); + tfree(pCmd->pTagData); + memset(pCmd->payload, 0, (size_t)pCmd->allocSize); tfree(pCmd->payload); pCmd->allocSize = 0; @@ -838,7 +838,6 @@ int32_t tscMergeTableDataBlocks(SSqlObj* pSql, SArray* pTableDataBlockList) { // the length does not include the SSubmitBlk structure pBlocks->dataLen = htonl(finalLen); - dataBuf->numOfTables += 1; } @@ -865,7 +864,7 @@ void tscCloseTscObj(void *param) { rpcClose(pObj->pDnodeConn); pObj->pDnodeConn = NULL; } - + tfree(pObj->tscCorMgmtEpSet); pthread_mutex_destroy(&pObj->mutex); tscDebug("%p DB connection is closed, dnodeConn:%p", pObj, p); @@ -1563,19 +1562,6 @@ void tscGetSrcColumnInfo(SSrcColumnInfo* pColInfo, SQueryInfo* pQueryInfo) { } } -void tscSetFreeHeatBeat(STscObj* pObj) { - if (pObj == NULL || pObj->signature != pObj || pObj->pHb == NULL) { - return; - } - - SSqlObj* pHeatBeat = pObj->pHb; - assert(pHeatBeat == pHeatBeat->signature); - - // to denote the heart-beat timer close connection and free all allocated resources - SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pHeatBeat->cmd, 0); - pQueryInfo->type = TSDB_QUERY_TYPE_FREE_RESOURCE; -} - /* * the following four kinds of SqlObj should not be freed * 1. SqlObj for stream computing @@ -1594,7 +1580,7 @@ bool tscShouldBeFreed(SSqlObj* pSql) { } STscObj* pTscObj = pSql->pTscObj; - if (pSql->pStream != NULL || pTscObj->pHb == pSql || pSql->pSubscription != NULL) { + if (pSql->pStream != NULL || pTscObj->hbrid == pSql->self || pSql->pSubscription != NULL) { return false; } @@ -1886,13 +1872,10 @@ void tscResetForNextRetrieve(SSqlRes* pRes) { } void registerSqlObj(SSqlObj* pSql) { - int32_t DEFAULT_LIFE_TIME = 2 * 600 * 1000; // 1200 sec - int32_t ref = T_REF_INC(pSql->pTscObj); tscDebug("%p add to tscObj:%p, ref:%d", pSql, pSql->pTscObj, ref); - TSDB_CACHE_PTR_TYPE p = (TSDB_CACHE_PTR_TYPE) pSql; - pSql->self = taosCachePut(tscObjCache, &p, sizeof(TSDB_CACHE_PTR_TYPE), &p, sizeof(TSDB_CACHE_PTR_TYPE), DEFAULT_LIFE_TIME); + pSql->self = taosAddRef(tscObjRef, pSql); } SSqlObj* createSimpleSubObj(SSqlObj* pSql, void (*fp)(), void* param, int32_t cmd) { @@ -1909,7 +1892,17 @@ SSqlObj* createSimpleSubObj(SSqlObj* pSql, void (*fp)(), void* param, int32_t cm pCmd->command = cmd; pCmd->parseFinished = 1; pCmd->autoCreated = pSql->cmd.autoCreated; - memcpy(&pCmd->tagData, &pSql->cmd.tagData, sizeof(pCmd->tagData)); + + if (pSql->cmd.pTagData != NULL) { + int size = offsetof(STagData, data) + htonl(pSql->cmd.pTagData->dataLen); + pNew->cmd.pTagData = calloc(1, size); + if (pNew->cmd.pTagData == NULL) { + tscError("%p new subquery failed, unable to malloc tag data, tableIndex:%d", pSql, 0); + free(pNew); + return NULL; + } + memcpy(pNew->cmd.pTagData, pSql->cmd.pTagData, size); + } if (tscAddSubqueryInfo(pCmd) != TSDB_CODE_SUCCESS) { tscFreeSqlObj(pNew); @@ -2447,10 +2440,10 @@ char* strdup_throw(const char* str) { return p; } -int tscSetMgmtEpSetFromCfg(const char *first, const char *second) { +int tscSetMgmtEpSetFromCfg(const char *first, const char *second, SRpcCorEpSet *corMgmtEpSet) { + corMgmtEpSet->version = 0; // init mgmt ip set - tscMgmtEpSet.version = 0; - SRpcEpSet *mgmtEpSet = &(tscMgmtEpSet.epSet); + SRpcEpSet *mgmtEpSet = &(corMgmtEpSet->epSet); mgmtEpSet->numOfEps = 0; mgmtEpSet->inUse = 0; diff --git a/src/common/inc/tglobal.h b/src/common/inc/tglobal.h index efe3d7678a39f140b413192adfe9654b3661251f..5b88c9b0d02804781c221d0bc36f0e944f34b049 100644 --- a/src/common/inc/tglobal.h +++ b/src/common/inc/tglobal.h @@ -35,6 +35,7 @@ extern int32_t tsNumOfMnodes; extern int32_t tsEnableVnodeBak; extern int32_t tsEnableTelemetryReporting; extern char tsEmail[]; +extern char tsArbitrator[]; // common extern int tsRpcTimer; @@ -56,6 +57,7 @@ extern char tsTempDir[]; //query buffer management extern int32_t tsQueryBufferSize; // maximum allowed usage buffer for each data node during query processing +extern int32_t tsHalfCoresForQuery; // only 50% will be used in query processing // client extern int32_t tsTableMetaKeepTimer; @@ -125,6 +127,9 @@ extern char tsMonitorDbName[]; extern char tsInternalPass[]; extern int32_t tsMonitorInterval; +// stream +extern int32_t tsEnableStream; + // internal extern int32_t tsPrintAuth; extern int32_t tscEmbedded; @@ -176,7 +181,7 @@ extern int32_t tmrDebugFlag; extern int32_t sdbDebugFlag; extern int32_t httpDebugFlag; extern int32_t mqttDebugFlag; -extern int32_t monitorDebugFlag; +extern int32_t monDebugFlag; extern int32_t uDebugFlag; extern int32_t rpcDebugFlag; extern int32_t odbcDebugFlag; diff --git a/src/common/inc/tname.h b/src/common/inc/tname.h index 6c48ca72f3b5ea010c662faa09211afd40152615..9e0093ebfed67a2cb4ca924d053ccffd38e316e6 100644 --- a/src/common/inc/tname.h +++ b/src/common/inc/tname.h @@ -25,6 +25,8 @@ void extractTableName(const char *tableId, char *name); char* extractDBName(const char *tableId, char *name); +size_t tableIdPrefix(const char* name, char* prefix, int32_t len); + void extractTableNameFromToken(SStrToken *pToken, SStrToken* pTable); SSchema tGetTableNameColumnSchema(); diff --git a/src/common/src/tglobal.c b/src/common/src/tglobal.c index f8bb965d282f994c35d32fabeb4c3ddba2fb4c27..a912cdfd7f531a6d863323056b1d9f78857a9284 100644 --- a/src/common/src/tglobal.c +++ b/src/common/src/tglobal.c @@ -107,6 +107,9 @@ int64_t tsMaxRetentWindow = 24 * 3600L; // maximum time window tolerance // positive value (in MB) int32_t tsQueryBufferSize = -1; +// only 50% cpu will be used in query processing in dnode +int32_t tsHalfCoresForQuery = 0; + // db parameters int32_t tsCacheBlockSize = TSDB_DEFAULT_CACHE_BLOCK_SIZE; int32_t tsBlocksPerVnode = TSDB_DEFAULT_TOTAL_BLOCKS; @@ -161,6 +164,9 @@ char tsMonitorDbName[TSDB_DB_NAME_LEN] = "log"; char tsInternalPass[] = "secretkey"; int32_t tsMonitorInterval = 30; // seconds +// stream +int32_t tsEnableStream = 1; + // internal int32_t tsPrintAuth = 0; int32_t tscEmbedded = 0; @@ -200,13 +206,13 @@ int32_t tsNumOfLogLines = 10000000; int32_t mDebugFlag = 135; int32_t sdbDebugFlag = 135; int32_t dDebugFlag = 135; -int32_t vDebugFlag = 131; +int32_t vDebugFlag = 135; int32_t cDebugFlag = 131; int32_t jniDebugFlag = 131; int32_t odbcDebugFlag = 131; int32_t httpDebugFlag = 131; int32_t mqttDebugFlag = 131; -int32_t monitorDebugFlag = 131; +int32_t monDebugFlag = 131; int32_t qDebugFlag = 131; int32_t rpcDebugFlag = 131; int32_t uDebugFlag = 131; @@ -216,9 +222,9 @@ int32_t wDebugFlag = 135; int32_t tsdbDebugFlag = 131; int32_t cqDebugFlag = 135; -int32_t (*monitorStartSystemFp)() = NULL; -void (*monitorStopSystemFp)() = NULL; -void (*monitorExecuteSQLFp)(char *sql) = NULL; +int32_t (*monStartSystemFp)() = NULL; +void (*monStopSystemFp)() = NULL; +void (*monExecuteSQLFp)(char *sql) = NULL; char *qtypeStr[] = {"rpc", "fwd", "wal", "cq", "query"}; @@ -235,7 +241,7 @@ void taosSetAllDebugFlag() { odbcDebugFlag = debugFlag; httpDebugFlag = debugFlag; mqttDebugFlag = debugFlag; - monitorDebugFlag = debugFlag; + monDebugFlag = debugFlag; qDebugFlag = debugFlag; rpcDebugFlag = debugFlag; uDebugFlag = debugFlag; @@ -276,15 +282,15 @@ bool taosCfgDynamicOptions(char *msg) { if (strncasecmp(cfg->option, "monitor", olen) == 0) { if (1 == vint) { - if (monitorStartSystemFp) { - (*monitorStartSystemFp)(); + if (monStartSystemFp) { + (*monStartSystemFp)(); uInfo("monitor is enabled"); } else { uError("monitor can't be updated, for monitor not initialized"); } } else { - if (monitorStopSystemFp) { - (*monitorStopSystemFp)(); + if (monStopSystemFp) { + (*monStopSystemFp)(); uInfo("monitor is disabled"); } else { uError("monitor can't be updated, for monitor not initialized"); @@ -307,8 +313,8 @@ bool taosCfgDynamicOptions(char *msg) { } if (strncasecmp(option, "resetQueryCache", 15) == 0) { - if (monitorExecuteSQLFp) { - (*monitorExecuteSQLFp)("resetQueryCache"); + if (monExecuteSQLFp) { + (*monExecuteSQLFp)("resetQueryCache"); uInfo("resetquerycache is executed"); } else { uError("resetquerycache can't be executed, for monitor not started"); @@ -881,6 +887,16 @@ static void doInitGlobalConfig(void) { cfg.unitType = TAOS_CFG_UTYPE_BYTE; taosInitConfigOption(cfg); + cfg.option = "halfCoresForQuery"; + cfg.ptr = &tsHalfCoresForQuery; + cfg.valType = TAOS_CFG_VTYPE_INT32; + cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW; + cfg.minValue = 0; + cfg.maxValue = 1; + cfg.ptrLength = 1; + cfg.unitType = TAOS_CFG_UTYPE_NONE; + taosInitConfigOption(cfg); + // locale & charset cfg.option = "timezone"; cfg.ptr = tsTimezone; @@ -1015,6 +1031,16 @@ static void doInitGlobalConfig(void) { cfg.unitType = TAOS_CFG_UTYPE_NONE; taosInitConfigOption(cfg); + cfg.option = "stream"; + cfg.ptr = &tsEnableStream; + cfg.valType = TAOS_CFG_VTYPE_INT32; + cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW; + cfg.minValue = 0; + cfg.maxValue = 1; + cfg.ptrLength = 1; + cfg.unitType = TAOS_CFG_UTYPE_NONE; + taosInitConfigOption(cfg); + cfg.option = "httpEnableRecordSql"; cfg.ptr = &tsHttpEnableRecordSql; cfg.valType = TAOS_CFG_VTYPE_INT32; @@ -1227,8 +1253,8 @@ static void doInitGlobalConfig(void) { cfg.unitType = TAOS_CFG_UTYPE_NONE; taosInitConfigOption(cfg); - cfg.option = "monitorDebugFlag"; - cfg.ptr = &monitorDebugFlag; + cfg.option = "monDebugFlag"; + cfg.ptr = &monDebugFlag; cfg.valType = TAOS_CFG_VTYPE_INT32; cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_LOG; cfg.minValue = 0; @@ -1277,7 +1303,7 @@ static void doInitGlobalConfig(void) { cfg.unitType = TAOS_CFG_UTYPE_NONE; taosInitConfigOption(cfg); - cfg.option = "tscEnableRecordSql"; + cfg.option = "enableRecordSql"; cfg.ptr = &tsTscEnableRecordSql; cfg.valType = TAOS_CFG_VTYPE_INT32; cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG; diff --git a/src/common/src/tname.c b/src/common/src/tname.c index bea8c52ef21ea607bee5379ace879ef936ff60bb..5c351edf48e10ff3176f38a177142f0f394ee5bd 100644 --- a/src/common/src/tname.c +++ b/src/common/src/tname.c @@ -39,6 +39,13 @@ char* extractDBName(const char* tableId, char* name) { return strncpy(name, &tableId[offset1 + 1], len); } +size_t tableIdPrefix(const char* name, char* prefix, int32_t len) { + tstrncpy(prefix, name, len); + strcat(prefix, TS_PATH_DELIMITER); + + return strlen(prefix); +} + SSchema tGetTableNameColumnSchema() { SSchema s = {0}; s.bytes = TSDB_TABLE_NAME_LEN - 1 + VARSTR_HEADER_SIZE; @@ -198,4 +205,4 @@ SSchema tscGetTbnameColumnSchema() { strcpy(s.name, TSQL_TBNAME_L); return s; -} \ No newline at end of file +} diff --git a/src/connector/jdbc/CMakeLists.txt b/src/connector/jdbc/CMakeLists.txt index 701a39b20994c2201328c7fd85a7da9aa4ab44a5..0eb3eb21cec7d79fc5fec5f7b0400a0a5a39a137 100644 --- a/src/connector/jdbc/CMakeLists.txt +++ b/src/connector/jdbc/CMakeLists.txt @@ -8,7 +8,7 @@ IF (TD_MVN_INSTALLED) ADD_CUSTOM_COMMAND(OUTPUT ${JDBC_CMD_NAME} POST_BUILD COMMAND mvn -Dmaven.test.skip=true install -f ${CMAKE_CURRENT_SOURCE_DIR}/pom.xml - COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/target/taos-jdbcdriver-2.0.13-dist.jar ${LIBRARY_OUTPUT_PATH} + COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/target/taos-jdbcdriver-2.0.14-dist.jar ${LIBRARY_OUTPUT_PATH} COMMAND mvn -Dmaven.test.skip=true clean -f ${CMAKE_CURRENT_SOURCE_DIR}/pom.xml COMMENT "build jdbc driver") ADD_CUSTOM_TARGET(${JDBC_TARGET_NAME} ALL WORKING_DIRECTORY ${EXECUTABLE_OUTPUT_PATH} DEPENDS ${JDBC_CMD_NAME}) diff --git a/src/connector/jdbc/deploy-pom.xml b/src/connector/jdbc/deploy-pom.xml index 1dc2625e62769a38eaa709c7e5493dd926466924..4564bde81e450f9c44be52ef69918dc0b3ada26f 100755 --- a/src/connector/jdbc/deploy-pom.xml +++ b/src/connector/jdbc/deploy-pom.xml @@ -5,7 +5,7 @@ com.taosdata.jdbc taos-jdbcdriver - 2.0.13 + 2.0.14 jar JDBCDriver diff --git a/src/connector/jdbc/pom.xml b/src/connector/jdbc/pom.xml index 3d1f40243547081d5ada664e0edaf3e9184f3172..7e087ebd9b05709518d45b8096cbd2b051c4a214 100755 --- a/src/connector/jdbc/pom.xml +++ b/src/connector/jdbc/pom.xml @@ -3,7 +3,7 @@ 4.0.0 com.taosdata.jdbc taos-jdbcdriver - 2.0.13 + 2.0.14 jar JDBCDriver https://github.com/taosdata/TDengine/tree/master/src/connector/jdbc diff --git a/src/connector/python/linux/python2/setup.py b/src/connector/python/linux/python2/setup.py index b3daa98bdce79adc479ac4c529bb2d2f4fe36696..92a931b504fd76d12c1a4094a04fed4f859c0a3e 100644 --- a/src/connector/python/linux/python2/setup.py +++ b/src/connector/python/linux/python2/setup.py @@ -5,7 +5,7 @@ with open("README.md", "r") as fh: setuptools.setup( name="taos", - version="2.0.2", + version="2.0.3", author="Taosdata Inc.", author_email="support@taosdata.com", description="TDengine python client package", diff --git a/src/connector/python/linux/python2/taos/__init__.py b/src/connector/python/linux/python2/taos/__init__.py index 4894488bc8d4e594c11a8be61fbea278ca2588fa..62e0536b6fca63a0aca0b6be3673104c27c631c6 100644 --- a/src/connector/python/linux/python2/taos/__init__.py +++ b/src/connector/python/linux/python2/taos/__init__.py @@ -3,7 +3,7 @@ from .connection import TDengineConnection from .cursor import TDengineCursor # Globals -apilevel = '2.0' +apilevel = '2.0.3' threadsafety = 0 paramstyle = 'pyformat' diff --git a/src/connector/python/linux/python2/taos/cinterface.py b/src/connector/python/linux/python2/taos/cinterface.py index 6f0435722ffba7a3cf6e2fb22591b1d037397156..2b1b29ee31af56499d626fc56d4950e2e25f9978 100644 --- a/src/connector/python/linux/python2/taos/cinterface.py +++ b/src/connector/python/linux/python2/taos/cinterface.py @@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) if num_of_rows > 0: return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]] else: @@ -106,6 +107,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) res=[] if num_of_rows > 0: for i in range(abs(num_of_rows)): diff --git a/src/connector/python/linux/python2/taos/cursor.py b/src/connector/python/linux/python2/taos/cursor.py index eee5e3948894b710217038f6e52db24a96212d60..82a01be67118813f9c83c3c3870d91535e13c526 100644 --- a/src/connector/python/linux/python2/taos/cursor.py +++ b/src/connector/python/linux/python2/taos/cursor.py @@ -26,7 +26,7 @@ class TDengineCursor(object): """ def __init__(self, connection=None): - self._description = None + self._description = [] self._rowcount = -1 self._connection = None self._result = None @@ -234,7 +234,7 @@ class TDengineCursor(object): def _reset_result(self): """Reset the result to unused version. """ - self._description = None + self._description = [] self._rowcount = -1 if self._result is not None: CTaosInterface.freeResult(self._result) diff --git a/src/connector/python/linux/python3/setup.py b/src/connector/python/linux/python3/setup.py index f49ebe2b6d188bf8c25623fbdde4c610a6b66231..655a12ad135e4eae575061e6069724a12a2a6996 100644 --- a/src/connector/python/linux/python3/setup.py +++ b/src/connector/python/linux/python3/setup.py @@ -5,7 +5,7 @@ with open("README.md", "r") as fh: setuptools.setup( name="taos", - version="2.0.2", + version="2.0.3", author="Taosdata Inc.", author_email="support@taosdata.com", description="TDengine python client package", diff --git a/src/connector/python/linux/python3/taos/__init__.py b/src/connector/python/linux/python3/taos/__init__.py index 4894488bc8d4e594c11a8be61fbea278ca2588fa..8cf095ea68dda55a071403734d39e9198a71d8a1 100644 --- a/src/connector/python/linux/python3/taos/__init__.py +++ b/src/connector/python/linux/python3/taos/__init__.py @@ -3,7 +3,7 @@ from .connection import TDengineConnection from .cursor import TDengineCursor # Globals -apilevel = '2.0' +apilevel = '2.0.3' threadsafety = 0 paramstyle = 'pyformat' @@ -21,4 +21,4 @@ def connect(*args, **kwargs): @rtype: TDengineConnector """ - return TDengineConnection(*args, **kwargs) \ No newline at end of file + return TDengineConnection(*args, **kwargs) diff --git a/src/connector/python/linux/python3/taos/cinterface.py b/src/connector/python/linux/python3/taos/cinterface.py index bbf5a0c7146f1fbc22befe50691df460ca8191c2..fdebe349fe9597260f231adb52022a2ecc9e1063 100644 --- a/src/connector/python/linux/python3/taos/cinterface.py +++ b/src/connector/python/linux/python3/taos/cinterface.py @@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) if num_of_rows > 0: return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]] else: @@ -106,6 +107,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) res=[] if num_of_rows > 0: for i in range(abs(num_of_rows)): diff --git a/src/connector/python/linux/python3/taos/cursor.py b/src/connector/python/linux/python3/taos/cursor.py index d8184668c8a13daafad86a17906d2077afe62b5a..0ce20f0edaabe5cf0d2bb6efa3e031f51779f03d 100644 --- a/src/connector/python/linux/python3/taos/cursor.py +++ b/src/connector/python/linux/python3/taos/cursor.py @@ -27,7 +27,7 @@ class TDengineCursor(object): """ def __init__(self, connection=None): - self._description = None + self._description = [] self._rowcount = -1 self._connection = None self._result = None @@ -242,7 +242,7 @@ class TDengineCursor(object): def _reset_result(self): """Reset the result to unused version. """ - self._description = None + self._description = [] self._rowcount = -1 if self._result is not None: CTaosInterface.freeResult(self._result) diff --git a/src/connector/python/windows/python2/setup.py b/src/connector/python/windows/python2/setup.py index 34cb7a04d701395cedc60ef13e541ad0c8303ec5..5ddbe83011daa9d408e22dbeda7172466329921e 100644 --- a/src/connector/python/windows/python2/setup.py +++ b/src/connector/python/windows/python2/setup.py @@ -5,7 +5,7 @@ with open("README.md", "r") as fh: setuptools.setup( name="taos", - version="2.0.2", + version="2.0.3", author="Taosdata Inc.", author_email="support@taosdata.com", description="TDengine python client package", diff --git a/src/connector/python/windows/python2/taos/__init__.py b/src/connector/python/windows/python2/taos/__init__.py index 4894488bc8d4e594c11a8be61fbea278ca2588fa..62e0536b6fca63a0aca0b6be3673104c27c631c6 100644 --- a/src/connector/python/windows/python2/taos/__init__.py +++ b/src/connector/python/windows/python2/taos/__init__.py @@ -3,7 +3,7 @@ from .connection import TDengineConnection from .cursor import TDengineCursor # Globals -apilevel = '2.0' +apilevel = '2.0.3' threadsafety = 0 paramstyle = 'pyformat' diff --git a/src/connector/python/windows/python2/taos/cinterface.py b/src/connector/python/windows/python2/taos/cinterface.py index d4bf1a0350680c407dd6b957ea3f59c0c3c9612f..14f4f49be8ae719da16dbed7e79631db7ef9689e 100644 --- a/src/connector/python/windows/python2/taos/cinterface.py +++ b/src/connector/python/windows/python2/taos/cinterface.py @@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) if num_of_rows > 0: return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]] else: @@ -106,6 +107,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) res=[] if num_of_rows > 0: for i in range(abs(num_of_rows)): diff --git a/src/connector/python/windows/python2/taos/cursor.py b/src/connector/python/windows/python2/taos/cursor.py index 71651afee1c2232ea6dc7a402e15a27ff9a6d72d..f6fde2619b0a727c74904c2338c813b97a2e602c 100644 --- a/src/connector/python/windows/python2/taos/cursor.py +++ b/src/connector/python/windows/python2/taos/cursor.py @@ -27,7 +27,7 @@ class TDengineCursor(object): """ def __init__(self, connection=None): - self._description = None + self._description = [] self._rowcount = -1 self._connection = None self._result = None @@ -193,7 +193,7 @@ class TDengineCursor(object): def _reset_result(self): """Reset the result to unused version. """ - self._description = None + self._description = [] self._rowcount = -1 if self._result is not None: CTaosInterface.freeResult(self._result) diff --git a/src/connector/python/windows/python3/setup.py b/src/connector/python/windows/python3/setup.py index c09644d3301b3a488b9bfdb7b60ffbaf6dba320e..ffed304c855fda28287735d0daab27e17b0e976e 100644 --- a/src/connector/python/windows/python3/setup.py +++ b/src/connector/python/windows/python3/setup.py @@ -5,7 +5,7 @@ with open("README.md", "r") as fh: setuptools.setup( name="taos", - version="2.0.2", + version="2.0.3", author="Taosdata Inc.", author_email="support@taosdata.com", description="TDengine python client package", diff --git a/src/connector/python/windows/python3/taos/__init__.py b/src/connector/python/windows/python3/taos/__init__.py index 2fec471a1adb29f08cd0d6539e8f9682bf6bc768..c6dd929a6a5110300ecd2e1042d48a5ad6a4bf46 100644 --- a/src/connector/python/windows/python3/taos/__init__.py +++ b/src/connector/python/windows/python3/taos/__init__.py @@ -3,7 +3,7 @@ from .connection import TDengineConnection from .cursor import TDengineCursor # Globals -apilevel = '2.0' +apilevel = '2.0.3' threadsafety = 0 paramstyle = 'pyformat' diff --git a/src/connector/python/windows/python3/taos/cinterface.py b/src/connector/python/windows/python3/taos/cinterface.py index fb1b65fc15f2f1c0f07745bb2aecf7b928a7c555..42b820ca80feda4a269e5e80f288bd1f5e87adcd 100644 --- a/src/connector/python/windows/python3/taos/cinterface.py +++ b/src/connector/python/windows/python3/taos/cinterface.py @@ -81,6 +81,7 @@ def _crow_double_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) if num_of_rows > 0: return [ None if ele.value[0:1] == FieldType.C_BINARY_NULL else ele.value.decode('utf-8') for ele in (ctypes.cast(data, ctypes.POINTER(ctypes.c_char * nbytes)))[:abs(num_of_rows)]] else: @@ -108,6 +109,7 @@ def _crow_nchar_to_python(data, num_of_rows, nbytes=None, micro=False): def _crow_binary_to_python_block(data, num_of_rows, nbytes=None, micro=False): """Function to convert C binary row to python row """ + assert(nbytes is not None) res=[] if num_of_rows > 0: for i in range(abs(num_of_rows)): diff --git a/src/connector/python/windows/python3/taos/cursor.py b/src/connector/python/windows/python3/taos/cursor.py index b813bba35756065b2082d39391de3367cf9865fb..db66b99d3bfe5935a3e5b04c3692453092bcf24e 100644 --- a/src/connector/python/windows/python3/taos/cursor.py +++ b/src/connector/python/windows/python3/taos/cursor.py @@ -28,7 +28,7 @@ class TDengineCursor(object): """ def __init__(self, connection=None): - self._description = None + self._description = [] self._rowcount = -1 self._connection = None self._result = None @@ -194,7 +194,7 @@ class TDengineCursor(object): def _reset_result(self): """Reset the result to unused version. """ - self._description = None + self._description = [] self._rowcount = -1 if self._result is not None: CTaosInterface.freeResult(self._result) diff --git a/src/cq/src/cqMain.c b/src/cq/src/cqMain.c index 1be7552a892e25cf44317a99fe52ff57689ad338..efb8795962e769dd0b11b8d313177855c994f52d 100644 --- a/src/cq/src/cqMain.c +++ b/src/cq/src/cqMain.c @@ -40,15 +40,14 @@ typedef struct { int32_t vgId; + int32_t master; + int32_t num; // number of continuous streams char user[TSDB_USER_LEN]; char pass[TSDB_PASSWORD_LEN]; char db[TSDB_DB_NAME_LEN]; FCqWrite cqWrite; - void *ahandle; - int32_t num; // number of continuous streams struct SCqObj *pHead; void *dbConn; - int32_t master; void *tmrCtrl; pthread_mutex_t mutex; } SCqContext; @@ -70,6 +69,9 @@ static void cqProcessStreamRes(void *param, TAOS_RES *tres, TAOS_ROW row); static void cqCreateStream(SCqContext *pContext, SCqObj *pObj); void *cqOpen(void *ahandle, const SCqCfg *pCfg) { + if (tsEnableStream == 0) { + return NULL; + } SCqContext *pContext = calloc(sizeof(SCqContext), 1); if (pContext == NULL) { terrno = TAOS_SYSTEM_ERROR(errno); @@ -90,7 +92,6 @@ void *cqOpen(void *ahandle, const SCqCfg *pCfg) { tstrncpy(pContext->db, db, sizeof(pContext->db)); pContext->vgId = pCfg->vgId; pContext->cqWrite = pCfg->cqWrite; - pContext->ahandle = ahandle; tscEmbedded = 1; pthread_mutex_init(&pContext->mutex, NULL); @@ -101,6 +102,9 @@ void *cqOpen(void *ahandle, const SCqCfg *pCfg) { } void cqClose(void *handle) { + if (tsEnableStream == 0) { + return; + } SCqContext *pContext = handle; if (handle == NULL) return; @@ -131,6 +135,9 @@ void cqClose(void *handle) { } void cqStart(void *handle) { + if (tsEnableStream == 0) { + return; + } SCqContext *pContext = handle; if (pContext->dbConn || pContext->master) return; @@ -149,6 +156,9 @@ void cqStart(void *handle) { } void cqStop(void *handle) { + if (tsEnableStream == 0) { + return; + } SCqContext *pContext = handle; cInfo("vgId:%d, stop all CQs", pContext->vgId); if (pContext->dbConn == NULL || pContext->master == 0) return; @@ -176,6 +186,9 @@ void cqStop(void *handle) { } void *cqCreate(void *handle, uint64_t uid, int32_t tid, char *sqlStr, STSchema *pSchema) { + if (tsEnableStream == 0) { + return NULL; + } SCqContext *pContext = handle; SCqObj *pObj = calloc(sizeof(SCqObj), 1); @@ -205,6 +218,9 @@ void *cqCreate(void *handle, uint64_t uid, int32_t tid, char *sqlStr, STSchema * } void cqDrop(void *handle) { + if (tsEnableStream == 0) { + return; + } SCqObj *pObj = handle; SCqContext *pContext = pObj->pContext; @@ -241,8 +257,12 @@ static void doCreateStream(void *param, TAOS_RES *result, int32_t code) { SCqObj* pObj = (SCqObj*)param; SCqContext* pContext = pObj->pContext; SSqlObj* pSql = (SSqlObj*)result; - pContext->dbConn = pSql->pTscObj; + if (atomic_val_compare_exchange_ptr(&(pContext->dbConn), NULL, pSql->pTscObj) != NULL) { + taos_close(pSql->pTscObj); + } + pthread_mutex_lock(&pContext->mutex); cqCreateStream(pContext, pObj); + pthread_mutex_unlock(&pContext->mutex); } static void cqProcessCreateTimer(void *param, void *tmrId) { @@ -253,7 +273,9 @@ static void cqProcessCreateTimer(void *param, void *tmrId) { cDebug("vgId:%d, try connect to TDengine", pContext->vgId); taos_connect_a(NULL, pContext->user, pContext->pass, pContext->db, 0, doCreateStream, param, NULL); } else { + pthread_mutex_lock(&pContext->mutex); cqCreateStream(pContext, pObj); + pthread_mutex_unlock(&pContext->mutex); } } @@ -267,12 +289,14 @@ static void cqCreateStream(SCqContext *pContext, SCqObj *pObj) { } pObj->tmrId = 0; - pObj->pStream = taos_open_stream(pContext->dbConn, pObj->sqlStr, cqProcessStreamRes, 0, pObj, NULL); - if (pObj->pStream) { - pContext->num++; - cInfo("vgId:%d, id:%d CQ:%s is openned", pContext->vgId, pObj->tid, pObj->sqlStr); - } else { - cError("vgId:%d, id:%d CQ:%s, failed to open", pContext->vgId, pObj->tid, pObj->sqlStr); + if (pObj->pStream == NULL) { + pObj->pStream = taos_open_stream(pContext->dbConn, pObj->sqlStr, cqProcessStreamRes, 0, pObj, NULL); + if (pObj->pStream) { + pContext->num++; + cInfo("vgId:%d, id:%d CQ:%s is openned", pContext->vgId, pObj->tid, pObj->sqlStr); + } else { + cError("vgId:%d, id:%d CQ:%s, failed to open", pContext->vgId, pObj->tid, pObj->sqlStr); + } } } @@ -334,7 +358,7 @@ static void cqProcessStreamRes(void *param, TAOS_RES *tres, TAOS_ROW row) { pHead->version = 0; // write into vnode write queue - pContext->cqWrite(pContext->ahandle, pHead, TAOS_QTYPE_CQ, NULL); + pContext->cqWrite(pContext->vgId, pHead, TAOS_QTYPE_CQ, NULL); free(buffer); } diff --git a/src/cq/test/cqtest.c b/src/cq/test/cqtest.c index e1114fc024054f7a4af32ca1497a63f7da942147..41380f0d86ed31b2466596993b8ce1a1d435f8cf 100644 --- a/src/cq/test/cqtest.c +++ b/src/cq/test/cqtest.c @@ -24,7 +24,7 @@ int64_t ver = 0; void *pCq = NULL; -int writeToQueue(void *pVnode, void *data, int type, void *pMsg) { +int writeToQueue(int32_t vgId, void *data, int type, void *pMsg) { return 0; } diff --git a/src/dnode/CMakeLists.txt b/src/dnode/CMakeLists.txt index 5608cfd6d16826e399f9ea41dc026d2ab2459610..ebb25bf85879e5ab1a53bd8ae4ae9ad3579bd6eb 100644 --- a/src/dnode/CMakeLists.txt +++ b/src/dnode/CMakeLists.txt @@ -12,7 +12,7 @@ AUX_SOURCE_DIRECTORY(src SRC) IF (TD_LINUX) ADD_EXECUTABLE(taosd ${SRC}) - TARGET_LINK_LIBRARIES(taosd mnode monitor http tsdb twal vnode cJson lz4) + TARGET_LINK_LIBRARIES(taosd mnode monitor http tsdb twal vnode cJson lz4 balance sync) IF (TD_SOMODE_STATIC) TARGET_LINK_LIBRARIES(taosd taos_static) @@ -32,10 +32,6 @@ IF (TD_LINUX) TARGET_LINK_LIBRARIES(taosd mqtt) ENDIF () - IF (TD_SYNC) - TARGET_LINK_LIBRARIES(taosd balance sync) - ENDIF () - SET(PREPARE_ENV_CMD "prepare_env_cmd") SET(PREPARE_ENV_TARGET "prepare_env_target") ADD_CUSTOM_COMMAND(OUTPUT ${PREPARE_ENV_CMD} diff --git a/src/dnode/inc/dnodeCfg.h b/src/dnode/inc/dnodeCfg.h index 35d889646021a079abe18c01a26fa5891718aaa0..d74303f3252e1fa54e174876bab5758be541955e 100644 --- a/src/dnode/inc/dnodeCfg.h +++ b/src/dnode/inc/dnodeCfg.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitCfg(); void dnodeCleanupCfg(); diff --git a/src/dnode/inc/dnodeCheck.h b/src/dnode/inc/dnodeCheck.h index a4880b3c114342c8900337372bd1dfd07e7815b7..c94b9e93197f4bfc9a175ee5825fd2c1c25d7072 100644 --- a/src/dnode/inc/dnodeCheck.h +++ b/src/dnode/inc/dnodeCheck.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitCheck(); void dnodeCleanupCheck(); diff --git a/src/dnode/inc/dnodeEps.h b/src/dnode/inc/dnodeEps.h index 2a203498c1f270c7e456694a4e1e195cbd9022cd..a5840997b0f9623c8ed504fa1c2eb8209216ea29 100644 --- a/src/dnode/inc/dnodeEps.h +++ b/src/dnode/inc/dnodeEps.h @@ -19,8 +19,7 @@ #ifdef __cplusplus extern "C" { #endif - -#include "taosmsg.h" +#include "dnodeInt.h" int32_t dnodeInitEps(); void dnodeCleanupEps(); diff --git a/src/dnode/inc/dnodeInt.h b/src/dnode/inc/dnodeInt.h index f4cbee1d1333e3f1ffc213f9e2ecdeb5f4c32d57..7595f5fd02fde85e64cf47d283d3e167cfa751d1 100644 --- a/src/dnode/inc/dnodeInt.h +++ b/src/dnode/inc/dnodeInt.h @@ -19,8 +19,13 @@ #ifdef __cplusplus extern "C" { #endif - +#include "taoserror.h" +#include "taosmsg.h" #include "tlog.h" +#include "trpc.h" +#include "tglobal.h" +#include "dnode.h" +#include "vnode.h" extern int32_t dDebugFlag; diff --git a/src/dnode/inc/dnodeMInfos.h b/src/dnode/inc/dnodeMInfos.h index 9c3c85c47e2dbcc11c5b5a80fbf091bd93855149..f05e2b6f7b710f9ebe547e27bf524c13ffcfaea8 100644 --- a/src/dnode/inc/dnodeMInfos.h +++ b/src/dnode/inc/dnodeMInfos.h @@ -19,16 +19,19 @@ #ifdef __cplusplus extern "C" { #endif - -#include "taosmsg.h" +#include "dnodeInt.h" int32_t dnodeInitMInfos(); void dnodeCleanupMInfos(); -void dnodeUpdateMInfos(SMnodeInfos *minfos); -void dnodeUpdateEpSetForPeer(SRpcEpSet *epSet); -void dnodeGetMInfos(SMnodeInfos *minfos); +void dnodeUpdateMInfos(SMInfos *pMinfos); +void dnodeUpdateEpSetForPeer(SRpcEpSet *pEpSet); +void dnodeGetMInfos(SMInfos *pMinfos); bool dnodeIsMasterEp(char *ep); +void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell); +void dnodeGetEpSetForPeer(SRpcEpSet *epSet); +void dnodeGetEpSetForShell(SRpcEpSet *epSet); + #ifdef __cplusplus } #endif diff --git a/src/dnode/inc/dnodeMPeer.h b/src/dnode/inc/dnodeMPeer.h index 00221baa221a411d614c1fa1bb3dc4525de2ed71..b7e566d7e4952d6afbd88e64b36197de4f0f4401 100644 --- a/src/dnode/inc/dnodeMPeer.h +++ b/src/dnode/inc/dnodeMPeer.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitMPeer(); void dnodeCleanupMPeer(); diff --git a/src/dnode/inc/dnodeMRead.h b/src/dnode/inc/dnodeMRead.h index 8a8e71227ddf3cd7b7c77ef664ac3d4c935889db..279098d30e137bc6b57ef09ce93b8d12fd421082 100644 --- a/src/dnode/inc/dnodeMRead.h +++ b/src/dnode/inc/dnodeMRead.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitMRead(); void dnodeCleanupMRead(); diff --git a/src/dnode/inc/dnodeMWrite.h b/src/dnode/inc/dnodeMWrite.h index 6a3d41bc81a03465a2e097bc711c83a8d8af362e..8d4fcce3befbad35b47b4817c02662adc37addf8 100644 --- a/src/dnode/inc/dnodeMWrite.h +++ b/src/dnode/inc/dnodeMWrite.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitMWrite(); void dnodeCleanupMWrite(); diff --git a/src/dnode/inc/dnodeMain.h b/src/dnode/inc/dnodeMain.h index c1480407bde0bbcf7aada20f90084b96fc380a73..ca79d53afdc0b211f3903b9323f7e567d0045093 100644 --- a/src/dnode/inc/dnodeMain.h +++ b/src/dnode/inc/dnodeMain.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitSystem(); void dnodeCleanUpSystem(); diff --git a/src/dnode/inc/dnodeModule.h b/src/dnode/inc/dnodeModule.h index 8618de324446ca3f2504d15a6422bcca3a4b51b0..edcefbdd0c9d6ce7e9d5ae84f383f6fa15fc2417 100644 --- a/src/dnode/inc/dnodeModule.h +++ b/src/dnode/inc/dnodeModule.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitModules(); void dnodeStartModules(); diff --git a/src/dnode/inc/dnodePeer.h b/src/dnode/inc/dnodePeer.h index 0dcf48f2322be91de792148165b2a127fa983196..6d337ef6dc65d717c382ebaf59a50cf334686a64 100644 --- a/src/dnode/inc/dnodePeer.h +++ b/src/dnode/inc/dnodePeer.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitServer(); void dnodeCleanupServer(); diff --git a/src/dnode/inc/dnodeShell.h b/src/dnode/inc/dnodeShell.h index 300c86c5999553b2332b3224fc6c65ef019dc5af..3fa66d6a3bc24d253ede6076f9b6bdacdb14db7d 100644 --- a/src/dnode/inc/dnodeShell.h +++ b/src/dnode/inc/dnodeShell.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitShell(); void dnodeCleanupShell(); diff --git a/src/dnode/inc/dnodeStep.h b/src/dnode/inc/dnodeStep.h new file mode 100644 index 0000000000000000000000000000000000000000..e181e19c4633aa6c844d92db15cee1a9930a41ac --- /dev/null +++ b/src/dnode/inc/dnodeStep.h @@ -0,0 +1,33 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_DNODE_STEP_H +#define TDENGINE_DNODE_STEP_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "dnodeInt.h" + +int32_t dnodeStepInit(SStep *pSteps, int32_t stepSize); +void dnodeStepCleanup(SStep *pSteps, int32_t stepSize); +void dnodeReportStep(char *name, char *desc, int8_t finished); +void dnodeSendStartupStep(SRpcMsg *pMsg); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/dnode/inc/dnodeTelemetry.h b/src/dnode/inc/dnodeTelemetry.h index 6fb62556ae50ec3759c7dc2260d16f36d1daf025..e4fd5a0376e3e7fdca0bbe1be15fa1cd70fd7695 100644 --- a/src/dnode/inc/dnodeTelemetry.h +++ b/src/dnode/inc/dnodeTelemetry.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitTelemetry(); void dnodeCleanupTelemetry(); diff --git a/src/dnode/inc/dnodeVMgmt.h b/src/dnode/inc/dnodeVMgmt.h new file mode 100644 index 0000000000000000000000000000000000000000..821196defc9b827961ec4ae8140be82186192f62 --- /dev/null +++ b/src/dnode/inc/dnodeVMgmt.h @@ -0,0 +1,32 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_DNODE_VMGMT_H +#define TDENGINE_DNODE_VMGMT_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "dnodeInt.h" + +int32_t dnodeInitVMgmt(); +void dnodeCleanupVMgmt(); +void dnodeDispatchToVMgmtQueue(SRpcMsg *rpcMsg); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/dnode/inc/dnodeVRead.h b/src/dnode/inc/dnodeVRead.h index 5b17693146cb68ab66943593bc2f9cd6587cc225..30dfb1b3a47509c1c3c5568520de0f092adba2c4 100644 --- a/src/dnode/inc/dnodeVRead.h +++ b/src/dnode/inc/dnodeVRead.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitVRead(); void dnodeCleanupVRead(); diff --git a/src/dnode/inc/dnodeVWrite.h b/src/dnode/inc/dnodeVWrite.h index 759e9ca8a5599236d08228ddd87ed8d4f8c55dca..2ddff210f8620cfe0544cb5925f633cab8d496af 100644 --- a/src/dnode/inc/dnodeVWrite.h +++ b/src/dnode/inc/dnodeVWrite.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "dnodeInt.h" int32_t dnodeInitVWrite(); void dnodeCleanupVWrite(); diff --git a/src/dnode/inc/dnodeVnodes.h b/src/dnode/inc/dnodeVnodes.h new file mode 100644 index 0000000000000000000000000000000000000000..1785ed3d0636b3db564041fc8487f73113781e7b --- /dev/null +++ b/src/dnode/inc/dnodeVnodes.h @@ -0,0 +1,34 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_DNODE_VNODES_H +#define TDENGINE_DNODE_VNODES_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "dnodeInt.h" + +int32_t dnodeInitVnodes(); +void dnodeCleanupVnodes(); +int32_t dnodeInitTimer(); +void dnodeCleanupTimer(); +void dnodeSendStatusMsgToMnode(); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/dnode/src/dnodeCfg.c b/src/dnode/src/dnodeCfg.c index 16d109a13a7600c5414b1f1edd175a8f15aa4cf0..89249d773b7c67b8939a7a70f65882f89f8f5190 100644 --- a/src/dnode/src/dnodeCfg.c +++ b/src/dnode/src/dnodeCfg.c @@ -16,9 +16,6 @@ #define _DEFAULT_SOURCE #include "os.h" #include "cJSON.h" -#include "tglobal.h" -#include "dnode.h" -#include "dnodeInt.h" #include "dnodeCfg.h" static SDnodeCfg tsCfg = {0}; diff --git a/src/dnode/src/dnodeCheck.c b/src/dnode/src/dnodeCheck.c index a9ee4ac649c2f4b2734ce7f0dd59004b08e2fb67..be26bb967bf7e987e9e990b25fb046f2b8664841 100644 --- a/src/dnode/src/dnodeCheck.c +++ b/src/dnode/src/dnodeCheck.c @@ -15,8 +15,6 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "tglobal.h" -#include "dnodeInt.h" #include "dnodeCheck.h" typedef struct { diff --git a/src/dnode/src/dnodeEps.c b/src/dnode/src/dnodeEps.c index 83f294e05eb0e559638fef1259ad96ac6f162460..09151533e29ed8c00a90c17c76b2b24fc5f0eb23 100644 --- a/src/dnode/src/dnodeEps.c +++ b/src/dnode/src/dnodeEps.c @@ -16,10 +16,7 @@ #define _DEFAULT_SOURCE #include "os.h" #include "cJSON.h" -#include "tglobal.h" #include "hash.h" -#include "dnode.h" -#include "dnodeInt.h" #include "dnodeEps.h" static SDnodeEps *tsEps = NULL; @@ -236,7 +233,14 @@ PRASE_EPS_OVER: dnodeResetEps(eps); if (eps) free(eps); +#if 0 dnodeUpdateEp(dnodeGetDnodeId(), tsLocalEp, tsLocalFqdn, &tsServerPort); +#else + if (dnodeCheckEpChanged(dnodeGetDnodeId(), tsLocalEp)) { + dError("dnode:%d, localEp is changed to %s in dnodeEps.json and need reconfigured", dnodeGetDnodeId(), tsLocalEp); + return -1; + } +#endif terrno = 0; return 0; diff --git a/src/dnode/src/dnodeMInfos.c b/src/dnode/src/dnodeMInfos.c index cefe44aebe7f87803141ce3d75c45dca18463849..7c385a889d10fcedd819f6eb10cbdaaf4910739f 100644 --- a/src/dnode/src/dnodeMInfos.c +++ b/src/dnode/src/dnodeMInfos.c @@ -16,18 +16,15 @@ #define _DEFAULT_SOURCE #include "os.h" #include "cJSON.h" -#include "tglobal.h" #include "mnode.h" -#include "dnode.h" -#include "dnodeInt.h" #include "dnodeMInfos.h" -static SMnodeInfos tsMInfos; -static SRpcEpSet tsMEpSet; +static SMInfos tsMInfos; +static SRpcEpSet tsMEpSet; static pthread_mutex_t tsMInfosMutex; -static void dnodeResetMInfos(SMnodeInfos *minfos); -static void dnodePrintMInfos(SMnodeInfos *minfos); +static void dnodeResetMInfos(SMInfos *minfos); +static void dnodePrintMInfos(SMInfos *minfos); static int32_t dnodeReadMInfos(); static int32_t dnodeWriteMInfos(); @@ -44,14 +41,14 @@ int32_t dnodeInitMInfos() { void dnodeCleanupMInfos() { pthread_mutex_destroy(&tsMInfosMutex); } -void dnodeUpdateMInfos(SMnodeInfos *minfos) { - if (minfos->mnodeNum <= 0 || minfos->mnodeNum > 3) { - dError("invalid mnode infos, mnodeNum:%d", minfos->mnodeNum); +void dnodeUpdateMInfos(SMInfos *pMinfos) { + if (pMinfos->mnodeNum <= 0 || pMinfos->mnodeNum > 3) { + dError("invalid mnode infos, mnodeNum:%d", pMinfos->mnodeNum); return; } - for (int32_t i = 0; i < minfos->mnodeNum; ++i) { - SMnodeInfo *minfo = &minfos->mnodeInfos[i]; + for (int32_t i = 0; i < pMinfos->mnodeNum; ++i) { + SMInfo *minfo = &pMinfos->mnodeInfos[i]; minfo->mnodeId = htonl(minfo->mnodeId); if (minfo->mnodeId <= 0 || strlen(minfo->mnodeEp) <= 5) { dError("invalid mnode info:%d, mnodeId:%d mnodeEp:%s", i, minfo->mnodeId, minfo->mnodeEp); @@ -60,14 +57,14 @@ void dnodeUpdateMInfos(SMnodeInfos *minfos) { } pthread_mutex_lock(&tsMInfosMutex); - if (minfos->mnodeNum != tsMInfos.mnodeNum) { - dnodeResetMInfos(minfos); + if (pMinfos->mnodeNum != tsMInfos.mnodeNum) { + dnodeResetMInfos(pMinfos); dnodeWriteMInfos(); sdbUpdateAsync(); } else { - int32_t size = sizeof(SMnodeInfos); - if (memcmp(minfos, &tsMInfos, size) != 0) { - dnodeResetMInfos(minfos); + int32_t size = sizeof(SMInfos); + if (memcmp(pMinfos, &tsMInfos, size) != 0) { + dnodeResetMInfos(pMinfos); dnodeWriteMInfos(); sdbUpdateAsync(); } @@ -99,11 +96,11 @@ bool dnodeIsMasterEp(char *ep) { return isMaster; } -void dnodeGetMInfos(SMnodeInfos *minfos) { +void dnodeGetMInfos(SMInfos *pMinfos) { pthread_mutex_lock(&tsMInfosMutex); - memcpy(minfos, &tsMInfos, sizeof(SMnodeInfos)); + memcpy(pMinfos, &tsMInfos, sizeof(SMInfos)); for (int32_t i = 0; i < tsMInfos.mnodeNum; ++i) { - minfos->mnodeInfos[i].mnodeId = htonl(tsMInfos.mnodeInfos[i].mnodeId); + pMinfos->mnodeInfos[i].mnodeId = htonl(tsMInfos.mnodeInfos[i].mnodeId); } pthread_mutex_unlock(&tsMInfosMutex); } @@ -123,15 +120,15 @@ void dnodeGetEpSetForShell(SRpcEpSet *epSet) { pthread_mutex_unlock(&tsMInfosMutex); } -static void dnodePrintMInfos(SMnodeInfos *minfos) { - dInfo("print mnode infos, mnodeNum:%d inUse:%d", minfos->mnodeNum, minfos->inUse); - for (int32_t i = 0; i < minfos->mnodeNum; i++) { - dInfo("mnode index:%d, %s", minfos->mnodeInfos[i].mnodeId, minfos->mnodeInfos[i].mnodeEp); +static void dnodePrintMInfos(SMInfos *pMinfos) { + dInfo("print minfos, mnodeNum:%d inUse:%d", pMinfos->mnodeNum, pMinfos->inUse); + for (int32_t i = 0; i < pMinfos->mnodeNum; i++) { + dInfo("mnode index:%d, %s", pMinfos->mnodeInfos[i].mnodeId, pMinfos->mnodeInfos[i].mnodeEp); } } -static void dnodeResetMInfos(SMnodeInfos *minfos) { - if (minfos == NULL) { +static void dnodeResetMInfos(SMInfos *pMinfos) { + if (pMinfos == NULL) { tsMEpSet.numOfEps = 1; taosGetFqdnPortFromEp(tsFirst, tsMEpSet.fqdn[0], &tsMEpSet.port[0]); @@ -142,10 +139,10 @@ static void dnodeResetMInfos(SMnodeInfos *minfos) { return; } - if (minfos->mnodeNum == 0) return; + if (pMinfos->mnodeNum == 0) return; - int32_t size = sizeof(SMnodeInfos); - memcpy(&tsMInfos, minfos, size); + int32_t size = sizeof(SMInfos); + memcpy(&tsMInfos, pMinfos, size); tsMEpSet.inUse = tsMInfos.inUse; tsMEpSet.numOfEps = tsMInfos.mnodeNum; @@ -153,16 +150,17 @@ static void dnodeResetMInfos(SMnodeInfos *minfos) { taosGetFqdnPortFromEp(tsMInfos.mnodeInfos[i].mnodeEp, tsMEpSet.fqdn[i], &tsMEpSet.port[i]); } - dnodePrintMInfos(minfos); + dnodePrintMInfos(pMinfos); } static int32_t dnodeReadMInfos() { - int32_t len = 0; - int32_t maxLen = 2000; - char * content = calloc(1, maxLen + 1); - cJSON * root = NULL; - FILE * fp = NULL; - SMnodeInfos minfos = {0}; + int32_t len = 0; + int32_t maxLen = 2000; + char * content = calloc(1, maxLen + 1); + cJSON * root = NULL; + FILE * fp = NULL; + SMInfos minfos = {0}; + bool nodeChanged = false; char file[TSDB_FILENAME_LEN + 20] = {0}; sprintf(file, "%s/mnodeEpSet.json", tsDnodeDir); @@ -221,14 +219,19 @@ static int32_t dnodeReadMInfos() { dError("failed to read mnodeEpSet.json, nodeId not found"); goto PARSE_MINFOS_OVER; } - minfos.mnodeInfos[i].mnodeId = nodeId->valueint; cJSON *nodeEp = cJSON_GetObjectItem(nodeInfo, "nodeEp"); if (!nodeEp || nodeEp->type != cJSON_String || nodeEp->valuestring == NULL) { dError("failed to read mnodeEpSet.json, nodeName not found"); goto PARSE_MINFOS_OVER; } - strncpy(minfos.mnodeInfos[i].mnodeEp, nodeEp->valuestring, TSDB_EP_LEN); + + SMInfo *pMinfo = &minfos.mnodeInfos[i]; + pMinfo->mnodeId = nodeId->valueint; + tstrncpy(pMinfo->mnodeEp, nodeEp->valuestring, TSDB_EP_LEN); + + bool changed = dnodeCheckEpChanged(pMinfo->mnodeId, pMinfo->mnodeEp); + if (changed) nodeChanged = changed; } dInfo("read file %s successed", file); @@ -241,10 +244,15 @@ PARSE_MINFOS_OVER: terrno = 0; for (int32_t i = 0; i < minfos.mnodeNum; ++i) { - SMnodeInfo *mInfo = &minfos.mnodeInfos[i]; + SMInfo *mInfo = &minfos.mnodeInfos[i]; dnodeUpdateEp(mInfo->mnodeId, mInfo->mnodeEp, NULL, NULL); } dnodeResetMInfos(&minfos); + + if (nodeChanged) { + dnodeWriteMInfos(); + } + return 0; } @@ -286,3 +294,25 @@ static int32_t dnodeWriteMInfos() { dInfo("successed to write %s", file); return 0; } + +void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) { + SRpcConnInfo connInfo = {0}; + rpcGetConnInfo(rpcMsg->handle, &connInfo); + + SRpcEpSet epSet = {0}; + if (forShell) { + dnodeGetEpSetForShell(&epSet); + } else { + dnodeGetEpSetForPeer(&epSet); + } + + dDebug("msg:%s will be redirected, dnodeIp:%s user:%s, numOfEps:%d inUse:%d", taosMsg[rpcMsg->msgType], + taosIpStr(connInfo.clientIp), connInfo.user, epSet.numOfEps, epSet.inUse); + + for (int32_t i = 0; i < epSet.numOfEps; ++i) { + dDebug("mnode index:%d %s:%d", i, epSet.fqdn[i], epSet.port[i]); + epSet.port[i] = htons(epSet.port[i]); + } + + rpcSendRedirectRsp(rpcMsg->handle, &epSet); +} \ No newline at end of file diff --git a/src/dnode/src/dnodeMPeer.c b/src/dnode/src/dnodeMPeer.c index ee6dc5212e4fdb98d4bfb7c3a7678fb6bd81bc8c..0863666f762e792e495c972b63a754492f2a3a9f 100644 --- a/src/dnode/src/dnodeMPeer.c +++ b/src/dnode/src/dnodeMPeer.c @@ -15,16 +15,11 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" -#include "taosmsg.h" -#include "tutil.h" #include "tqueue.h" #include "twal.h" -#include "tglobal.h" #include "mnode.h" -#include "dnode.h" -#include "dnodeInt.h" -#include "dnodeMgmt.h" +#include "dnodeVMgmt.h" +#include "dnodeMInfos.h" #include "dnodeMWrite.h" typedef struct { diff --git a/src/dnode/src/dnodeMRead.c b/src/dnode/src/dnodeMRead.c index 65f3af7b3bf13efe0ffa80ca5752d66cc4a43e9e..0fc6400d99f8dbcbc57caa2ed1265e36e18f1a93 100644 --- a/src/dnode/src/dnodeMRead.c +++ b/src/dnode/src/dnodeMRead.c @@ -15,16 +15,11 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" -#include "taosmsg.h" -#include "tutil.h" #include "tqueue.h" #include "twal.h" -#include "tglobal.h" #include "mnode.h" -#include "dnode.h" -#include "dnodeInt.h" -#include "dnodeMgmt.h" +#include "dnodeVMgmt.h" +#include "dnodeMInfos.h" #include "dnodeMRead.h" typedef struct { diff --git a/src/dnode/src/dnodeMWrite.c b/src/dnode/src/dnodeMWrite.c index ef2d49ef4233eb94f3efd82f77a6918b2b671ffd..bc387e21714a61021fb8140a080937c1eed9d75c 100644 --- a/src/dnode/src/dnodeMWrite.c +++ b/src/dnode/src/dnodeMWrite.c @@ -15,17 +15,11 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" -#include "taosmsg.h" -#include "tutil.h" #include "ttimer.h" #include "tqueue.h" -#include "twal.h" -#include "tglobal.h" #include "mnode.h" -#include "dnode.h" -#include "dnodeInt.h" -#include "dnodeMgmt.h" +#include "dnodeVMgmt.h" +#include "dnodeMInfos.h" #include "dnodeMWrite.h" typedef struct { diff --git a/src/dnode/src/dnodeMain.c b/src/dnode/src/dnodeMain.c index 130be0af202a4882a100335713ba0f81af7a3c11..502c410ce3aabfcfaf8789e0e7854efde64ef926 100644 --- a/src/dnode/src/dnodeMain.c +++ b/src/dnode/src/dnodeMain.c @@ -16,14 +16,12 @@ #define _DEFAULT_SOURCE #include "os.h" #include "taos.h" -#include "tutil.h" #include "tconfig.h" -#include "tglobal.h" +#include "tfile.h" #include "twal.h" -#include "trpc.h" -#include "dnode.h" -#include "dnodeInt.h" -#include "dnodeMgmt.h" +// #include "tfs.h" +#include "tsync.h" +#include "dnodeStep.h" #include "dnodePeer.h" #include "dnodeModule.h" #include "dnodeEps.h" @@ -32,6 +30,8 @@ #include "dnodeCheck.h" #include "dnodeVRead.h" #include "dnodeVWrite.h" +#include "dnodeVMgmt.h" +#include "dnodeVnodes.h" #include "dnodeMRead.h" #include "dnodeMWrite.h" #include "dnodeMPeer.h" @@ -44,37 +44,32 @@ static int32_t dnodeInitStorage(); static void dnodeCleanupStorage(); static void dnodeSetRunStatus(SRunStatus status); static void dnodeCheckDataDirOpenned(char *dir); -static int32_t dnodeInitComponents(); -static void dnodeCleanupComponents(int32_t stepId); static int dnodeCreateDir(const char *dir); -typedef struct { - const char *const name; - int32_t (*init)(); - void (*cleanup)(); -} SDnodeComponent; - -static const SDnodeComponent tsDnodeComponents[] = { - {"rpc", rpcInit, rpcCleanup}, - {"storage", dnodeInitStorage, dnodeCleanupStorage}, - {"dnodecfg", dnodeInitCfg, dnodeCleanupCfg}, - {"dnodeeps", dnodeInitEps, dnodeCleanupEps}, - {"globalcfg" ,taosCheckGlobalCfg, NULL}, - {"mnodeinfos",dnodeInitMInfos, dnodeCleanupMInfos}, - {"wal", walInit, walCleanUp}, - {"check", dnodeInitCheck, dnodeCleanupCheck}, // NOTES: dnodeInitCheck must be behind the dnodeinitStorage component !!! - {"vread", dnodeInitVRead, dnodeCleanupVRead}, - {"vwrite", dnodeInitVWrite, dnodeCleanupVWrite}, - {"mread", dnodeInitMRead, dnodeCleanupMRead}, - {"mwrite", dnodeInitMWrite, dnodeCleanupMWrite}, - {"mpeer", dnodeInitMPeer, dnodeCleanupMPeer}, - {"client", dnodeInitClient, dnodeCleanupClient}, - {"server", dnodeInitServer, dnodeCleanupServer}, - {"mgmt", dnodeInitMgmt, dnodeCleanupMgmt}, - {"modules", dnodeInitModules, dnodeCleanupModules}, - {"mgmt-tmr", dnodeInitMgmtTimer, dnodeCleanupMgmtTimer}, - {"shell", dnodeInitShell, dnodeCleanupShell}, - {"telemetry", dnodeInitTelemetry, dnodeCleanupTelemetry}, +static SStep tsDnodeSteps[] = { + {"dnode-tfile", tfInit, tfCleanup}, + {"dnode-rpc", rpcInit, rpcCleanup}, + {"dnode-globalcfg", taosCheckGlobalCfg, NULL}, + {"dnode-storage", dnodeInitStorage, dnodeCleanupStorage}, + {"dnode-cfg", dnodeInitCfg, dnodeCleanupCfg}, + {"dnode-eps", dnodeInitEps, dnodeCleanupEps}, + {"dnode-minfos", dnodeInitMInfos, dnodeCleanupMInfos}, + {"dnode-wal", walInit, walCleanUp}, + {"dnode-sync", syncInit, syncCleanUp}, + {"dnode-check", dnodeInitCheck, dnodeCleanupCheck}, // NOTES: dnodeInitCheck must be behind the dnodeinitStorage component !!! + {"dnode-vread", dnodeInitVRead, dnodeCleanupVRead}, + {"dnode-vwrite", dnodeInitVWrite, dnodeCleanupVWrite}, + {"dnode-vmgmt", dnodeInitVMgmt, dnodeCleanupVMgmt}, + {"dnode-mread", dnodeInitMRead, dnodeCleanupMRead}, + {"dnode-mwrite", dnodeInitMWrite, dnodeCleanupMWrite}, + {"dnode-mpeer", dnodeInitMPeer, dnodeCleanupMPeer}, + {"dnode-client", dnodeInitClient, dnodeCleanupClient}, + {"dnode-server", dnodeInitServer, dnodeCleanupServer}, + {"dnode-vnodes", dnodeInitVnodes, dnodeCleanupVnodes}, + {"dnode-modules", dnodeInitModules, dnodeCleanupModules}, + {"dnode-tmr", dnodeInitTimer, dnodeCleanupTimer}, + {"dnode-shell", dnodeInitShell, dnodeCleanupShell}, + {"dnode-telemetry", dnodeInitTelemetry, dnodeCleanupTelemetry}, }; static int dnodeCreateDir(const char *dir) { @@ -85,24 +80,14 @@ static int dnodeCreateDir(const char *dir) { return 0; } -static void dnodeCleanupComponents(int32_t stepId) { - for (int32_t i = stepId; i >= 0; i--) { - if (tsDnodeComponents[i].cleanup) { - (*tsDnodeComponents[i].cleanup)(); - } - } +static void dnodeCleanupComponents() { + int32_t stepSize = sizeof(tsDnodeSteps) / sizeof(SStep); + dnodeStepCleanup(tsDnodeSteps, stepSize); } static int32_t dnodeInitComponents() { - int32_t code = 0; - for (int32_t i = 0; i < sizeof(tsDnodeComponents) / sizeof(tsDnodeComponents[0]); i++) { - if (tsDnodeComponents[i].init() != 0) { - dnodeCleanupComponents(i); - code = -1; - break; - } - } - return code; + int32_t stepSize = sizeof(tsDnodeSteps) / sizeof(SStep); + return dnodeStepInit(tsDnodeSteps, stepSize); } int32_t dnodeInitSystem() { @@ -149,7 +134,7 @@ int32_t dnodeInitSystem() { void dnodeCleanUpSystem() { if (dnodeGetRunStatus() != TSDB_RUN_STATUS_STOPPED) { dnodeSetRunStatus(TSDB_RUN_STATUS_STOPPED); - dnodeCleanupComponents(sizeof(tsDnodeComponents) / sizeof(tsDnodeComponents[0]) - 1); + dnodeCleanupComponents(); taos_cleanup(); taosCloseLog(); } diff --git a/src/dnode/src/dnodeMgmt.c b/src/dnode/src/dnodeMgmt.c deleted file mode 100644 index da1852e05eca11ae9909fe75ff1bb14eee6ae5fb..0000000000000000000000000000000000000000 --- a/src/dnode/src/dnodeMgmt.c +++ /dev/null @@ -1,572 +0,0 @@ -/* - * Copyright (c) 2019 TAOS Data, Inc. - * - * This program is free software: you can use, redistribute, and/or modify - * it under the terms of the GNU Affero General Public License, version 3 - * or later ("AGPL"), as published by the Free Software Foundation. - * - * This program is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. - * - * You should have received a copy of the GNU Affero General Public License - * along with this program. If not, see . - */ - -#define _DEFAULT_SOURCE -#include "os.h" -#include "cJSON.h" -#include "taoserror.h" -#include "taosmsg.h" -#include "ttimer.h" -#include "tsdb.h" -#include "twal.h" -#include "tqueue.h" -#include "tsync.h" -#include "ttimer.h" -#include "tbalance.h" -#include "tglobal.h" -#include "dnode.h" -#include "vnode.h" -#include "mnode.h" -#include "dnodeInt.h" -#include "dnodeMgmt.h" -#include "dnodeEps.h" -#include "dnodeCfg.h" -#include "dnodeMInfos.h" -#include "dnodeVRead.h" -#include "dnodeVWrite.h" -#include "dnodeModule.h" - -typedef struct { - pthread_t thread; - int32_t threadIndex; - int32_t failed; - int32_t opened; - int32_t vnodeNum; - int32_t * vnodeList; -} SOpenVnodeThread; - -typedef struct { - SRpcMsg rpcMsg; - char pCont[]; -} SMgmtMsg; - -void * tsDnodeTmr = NULL; -static void * tsStatusTimer = NULL; -static uint32_t tsRebootTime; -static taos_qset tsMgmtQset = NULL; -static taos_queue tsMgmtQueue = NULL; -static pthread_t tsQthread; - -static void dnodeProcessStatusRsp(SRpcMsg *pMsg); -static void dnodeSendStatusMsg(void *handle, void *tmrId); -static void *dnodeProcessMgmtQueue(void *param); - -static int32_t dnodeOpenVnodes(); -static void dnodeCloseVnodes(); -static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *pMsg); -static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *pMsg); -static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *pMsg); -static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg); -static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg); -static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg); -static int32_t (*dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *pMsg); - -int32_t dnodeInitMgmt() { - dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeProcessCreateVnodeMsg; - dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeProcessAlterVnodeMsg; - dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeProcessDropVnodeMsg; - dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeProcessAlterStreamMsg; - dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeProcessConfigDnodeMsg; - dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeProcessCreateMnodeMsg; - - dnodeAddClientRspHandle(TSDB_MSG_TYPE_DM_STATUS_RSP, dnodeProcessStatusRsp); - tsRebootTime = taosGetTimestampSec(); - - int32_t code = vnodeInitResources(); - if (code != TSDB_CODE_SUCCESS) { - dnodeCleanupMgmt(); - return -1; - } - - // create the queue and thread to handle the message - tsMgmtQset = taosOpenQset(); - if (tsMgmtQset == NULL) { - dError("failed to create the mgmt queue set"); - dnodeCleanupMgmt(); - return -1; - } - - tsMgmtQueue = taosOpenQueue(); - if (tsMgmtQueue == NULL) { - dError("failed to create the mgmt queue"); - dnodeCleanupMgmt(); - return -1; - } - - taosAddIntoQset(tsMgmtQset, tsMgmtQueue, NULL); - - pthread_attr_t thAttr; - pthread_attr_init(&thAttr); - pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); - - code = pthread_create(&tsQthread, &thAttr, dnodeProcessMgmtQueue, NULL); - pthread_attr_destroy(&thAttr); - if (code != 0) { - dError("failed to create thread to process mgmt queue, reason:%s", strerror(errno)); - dnodeCleanupMgmt(); - return -1; - } - - code = dnodeOpenVnodes(); - if (code != TSDB_CODE_SUCCESS) { - dnodeCleanupMgmt(); - return -1; - } - - dInfo("dnode mgmt is initialized"); - - return TSDB_CODE_SUCCESS; -} - -int32_t dnodeInitMgmtTimer() { - tsDnodeTmr = taosTmrInit(100, 200, 60000, "DND-DM"); - if (tsDnodeTmr == NULL) { - dError("failed to init dnode timer"); - dnodeCleanupMgmt(); - return -1; - } - - taosTmrReset(dnodeSendStatusMsg, 500, NULL, tsDnodeTmr, &tsStatusTimer); - dInfo("dnode mgmt timer is initialized"); - return TSDB_CODE_SUCCESS; -} - -void dnodeSendStatusMsgToMnode() { - if (tsDnodeTmr != NULL && tsStatusTimer != NULL) { - dInfo("force send status msg to mnode"); - taosTmrReset(dnodeSendStatusMsg, 3, NULL, tsDnodeTmr, &tsStatusTimer); - } -} - -void dnodeCleanupMgmtTimer() { - if (tsStatusTimer != NULL) { - taosTmrStopA(&tsStatusTimer); - tsStatusTimer = NULL; - } - - if (tsDnodeTmr != NULL) { - taosTmrCleanUp(tsDnodeTmr); - tsDnodeTmr = NULL; - } -} - -void dnodeCleanupMgmt() { - dnodeCleanupMgmtTimer(); - dnodeCloseVnodes(); - - if (tsMgmtQset) taosQsetThreadResume(tsMgmtQset); - if (tsQthread) pthread_join(tsQthread, NULL); - - if (tsMgmtQueue) taosCloseQueue(tsMgmtQueue); - if (tsMgmtQset) taosCloseQset(tsMgmtQset); - tsMgmtQset = NULL; - tsMgmtQueue = NULL; - - vnodeCleanupResources(); -} - -static int32_t dnodeWriteToMgmtQueue(SRpcMsg *pMsg) { - int32_t size = sizeof(SMgmtMsg) + pMsg->contLen; - SMgmtMsg *pMgmt = taosAllocateQitem(size); - if (pMgmt == NULL) { - return TSDB_CODE_DND_OUT_OF_MEMORY; - } - - pMgmt->rpcMsg = *pMsg; - pMgmt->rpcMsg.pCont = pMgmt->pCont; - memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen); - taosWriteQitem(tsMgmtQueue, TAOS_QTYPE_RPC, pMgmt); - - return TSDB_CODE_SUCCESS; -} - -void dnodeDispatchToMgmtQueue(SRpcMsg *pMsg) { - int32_t code = dnodeWriteToMgmtQueue(pMsg); - if (code != TSDB_CODE_SUCCESS) { - SRpcMsg rsp = {.handle = pMsg->handle, .code = code}; - rpcSendResponse(&rsp); - } - - rpcFreeCont(pMsg->pCont); -} - -static void *dnodeProcessMgmtQueue(void *param) { - SMgmtMsg *pMgmt; - SRpcMsg * pMsg; - SRpcMsg rsp = {0}; - int32_t qtype; - void * handle; - - while (1) { - if (taosReadQitemFromQset(tsMgmtQset, &qtype, (void **)&pMgmt, &handle) == 0) { - dDebug("qset:%p, dnode mgmt got no message from qset, exit", tsMgmtQset); - break; - } - - pMsg = &pMgmt->rpcMsg; - dDebug("msg:%p, ahandle:%p type:%s will be processed", pMgmt, pMsg->ahandle, taosMsg[pMsg->msgType]); - if (dnodeProcessMgmtMsgFp[pMsg->msgType]) { - rsp.code = (*dnodeProcessMgmtMsgFp[pMsg->msgType])(pMsg); - } else { - rsp.code = TSDB_CODE_DND_MSG_NOT_PROCESSED; - } - - rsp.handle = pMsg->handle; - rsp.pCont = NULL; - rpcSendResponse(&rsp); - - taosFreeQitem(pMsg); - } - - return NULL; -} - -static int32_t dnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) { - DIR *dir = opendir(tsVnodeDir); - if (dir == NULL) { - return TSDB_CODE_DND_NO_WRITE_ACCESS; - } - - *numOfVnodes = 0; - struct dirent *de = NULL; - while ((de = readdir(dir)) != NULL) { - if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue; - if (de->d_type & DT_DIR) { - if (strncmp("vnode", de->d_name, 5) != 0) continue; - int32_t vnode = atoi(de->d_name + 5); - if (vnode == 0) continue; - - (*numOfVnodes)++; - - if (*numOfVnodes >= TSDB_MAX_VNODES) { - dError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES); - continue; - } else { - vnodeList[*numOfVnodes - 1] = vnode; - } - } - } - closedir(dir); - - return TSDB_CODE_SUCCESS; -} - -static void *dnodeOpenVnode(void *param) { - SOpenVnodeThread *pThread = param; - char vnodeDir[TSDB_FILENAME_LEN * 3]; - - dDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum); - - for (int32_t v = 0; v < pThread->vnodeNum; ++v) { - int32_t vgId = pThread->vnodeList[v]; - snprintf(vnodeDir, TSDB_FILENAME_LEN * 3, "%s/vnode%d", tsVnodeDir, vgId); - if (vnodeOpen(vgId, vnodeDir) < 0) { - dError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex); - pThread->failed++; - } else { - dDebug("vgId:%d, is openned by thread:%d", vgId, pThread->threadIndex); - pThread->opened++; - } - } - - dDebug("thread:%d, total vnodes:%d, openned:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened, - pThread->failed); - return NULL; -} - -static int32_t dnodeOpenVnodes() { - int32_t vnodeList[TSDB_MAX_VNODES] = {0}; - int32_t numOfVnodes = 0; - int32_t status = dnodeGetVnodeList(vnodeList, &numOfVnodes); - - if (status != TSDB_CODE_SUCCESS) { - dInfo("get dnode list failed"); - return status; - } - - int32_t threadNum = tsNumOfCores; - int32_t vnodesPerThread = numOfVnodes / threadNum + 1; - SOpenVnodeThread *threads = calloc(threadNum, sizeof(SOpenVnodeThread)); - for (int32_t t = 0; t < threadNum; ++t) { - threads[t].threadIndex = t; - threads[t].vnodeList = calloc(vnodesPerThread, sizeof(int32_t)); - } - - for (int32_t v = 0; v < numOfVnodes; ++v) { - int32_t t = v % threadNum; - SOpenVnodeThread *pThread = &threads[t]; - pThread->vnodeList[pThread->vnodeNum++] = vnodeList[v]; - } - - dDebug("start %d threads to open %d vnodes", threadNum, numOfVnodes); - - for (int32_t t = 0; t < threadNum; ++t) { - SOpenVnodeThread *pThread = &threads[t]; - if (pThread->vnodeNum == 0) continue; - - pthread_attr_t thAttr; - pthread_attr_init(&thAttr); - pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); - if (pthread_create(&pThread->thread, &thAttr, dnodeOpenVnode, pThread) != 0) { - dError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno)); - } - - pthread_attr_destroy(&thAttr); - } - - int32_t openVnodes = 0; - int32_t failedVnodes = 0; - for (int32_t t = 0; t < threadNum; ++t) { - SOpenVnodeThread *pThread = &threads[t]; - if (pThread->vnodeNum > 0 && pThread->thread) { - pthread_join(pThread->thread, NULL); - } - openVnodes += pThread->opened; - failedVnodes += pThread->failed; - free(pThread->vnodeList); - } - - free(threads); - dInfo("there are total vnodes:%d, openned:%d failed:%d", numOfVnodes, openVnodes, failedVnodes); - - return TSDB_CODE_SUCCESS; -} - -static void dnodeCloseVnodes() { - int32_t vnodeList[TSDB_MAX_VNODES]= {0}; - int32_t numOfVnodes = 0; - int32_t status; - - status = vnodeGetVnodeList(vnodeList, &numOfVnodes); - - if (status != TSDB_CODE_SUCCESS) { - dInfo("get dnode list failed"); - return; - } - - for (int32_t i = 0; i < numOfVnodes; ++i) { - vnodeClose(vnodeList[i]); - } - - dInfo("total vnodes:%d are all closed", numOfVnodes); -} - -static void* dnodeParseVnodeMsg(SRpcMsg *rpcMsg) { - SCreateVnodeMsg *pCreate = rpcMsg->pCont; - pCreate->cfg.vgId = htonl(pCreate->cfg.vgId); - pCreate->cfg.cfgVersion = htonl(pCreate->cfg.cfgVersion); - pCreate->cfg.maxTables = htonl(pCreate->cfg.maxTables); - pCreate->cfg.cacheBlockSize = htonl(pCreate->cfg.cacheBlockSize); - pCreate->cfg.totalBlocks = htonl(pCreate->cfg.totalBlocks); - pCreate->cfg.daysPerFile = htonl(pCreate->cfg.daysPerFile); - pCreate->cfg.daysToKeep1 = htonl(pCreate->cfg.daysToKeep1); - pCreate->cfg.daysToKeep2 = htonl(pCreate->cfg.daysToKeep2); - pCreate->cfg.daysToKeep = htonl(pCreate->cfg.daysToKeep); - pCreate->cfg.minRowsPerFileBlock = htonl(pCreate->cfg.minRowsPerFileBlock); - pCreate->cfg.maxRowsPerFileBlock = htonl(pCreate->cfg.maxRowsPerFileBlock); - pCreate->cfg.fsyncPeriod = htonl(pCreate->cfg.fsyncPeriod); - pCreate->cfg.commitTime = htonl(pCreate->cfg.commitTime); - - for (int32_t j = 0; j < pCreate->cfg.replications; ++j) { - pCreate->nodes[j].nodeId = htonl(pCreate->nodes[j].nodeId); - } - - return pCreate; -} - -static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) { - SCreateVnodeMsg *pCreate = dnodeParseVnodeMsg(rpcMsg); - - void *pVnode = vnodeAcquire(pCreate->cfg.vgId); - if (pVnode != NULL) { - dDebug("vgId:%d, already exist, return success", pCreate->cfg.vgId); - vnodeRelease(pVnode); - return TSDB_CODE_SUCCESS; - } else { - dDebug("vgId:%d, create vnode msg is received", pCreate->cfg.vgId); - return vnodeCreate(pCreate); - } -} - -static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) { - SAlterVnodeMsg *pAlter = dnodeParseVnodeMsg(rpcMsg); - - void *pVnode = vnodeAcquire(pAlter->cfg.vgId); - if (pVnode != NULL) { - dDebug("vgId:%d, alter vnode msg is received", pAlter->cfg.vgId); - int32_t code = vnodeAlter(pVnode, pAlter); - vnodeRelease(pVnode); - return code; - } else { - dError("vgId:%d, vnode not exist, can't alter it", pAlter->cfg.vgId); - return TSDB_CODE_VND_INVALID_VGROUP_ID; - } -} - -static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) { - SDropVnodeMsg *pDrop = rpcMsg->pCont; - pDrop->vgId = htonl(pDrop->vgId); - - return vnodeDrop(pDrop->vgId); -} - -static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg) { -// SAlterStreamMsg *pStream = pCont; -// pStream->uid = htobe64(pStream->uid); -// pStream->stime = htobe64(pStream->stime); -// pStream->vnode = htonl(pStream->vnode); -// pStream->sid = htonl(pStream->sid); -// pStream->status = htonl(pStream->status); -// -// int32_t code = dnodeCreateStream(pStream); - - return 0; -} - -static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg) { - SCfgDnodeMsg *pCfg = pMsg->pCont; - return taosCfgDynamicOptions(pCfg->config); -} - -static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) { - SCreateMnodeMsg *pCfg = pMsg->pCont; - pCfg->dnodeId = htonl(pCfg->dnodeId); - if (pCfg->dnodeId != dnodeGetDnodeId()) { - dError("dnodeId:%d, in create mnode msg is not equal with saved dnodeId:%d", pCfg->dnodeId, dnodeGetDnodeId()); - return TSDB_CODE_MND_DNODE_ID_NOT_CONFIGURED; - } - - if (strcmp(pCfg->dnodeEp, tsLocalEp) != 0) { - dError("dnodeEp:%s, in create mnode msg is not equal with saved dnodeEp:%s", pCfg->dnodeEp, tsLocalEp); - return TSDB_CODE_MND_DNODE_EP_NOT_CONFIGURED; - } - - dDebug("dnodeId:%d, create mnode msg is received from mnodes, numOfMnodes:%d", pCfg->dnodeId, pCfg->mnodes.mnodeNum); - for (int i = 0; i < pCfg->mnodes.mnodeNum; ++i) { - pCfg->mnodes.mnodeInfos[i].mnodeId = htonl(pCfg->mnodes.mnodeInfos[i].mnodeId); - dDebug("mnode index:%d, mnode:%d:%s", i, pCfg->mnodes.mnodeInfos[i].mnodeId, pCfg->mnodes.mnodeInfos[i].mnodeEp); - } - - dnodeStartMnode(&pCfg->mnodes); - - return TSDB_CODE_SUCCESS; -} - -static void dnodeProcessStatusRsp(SRpcMsg *pMsg) { - if (pMsg->code != TSDB_CODE_SUCCESS) { - dError("status rsp is received, error:%s", tstrerror(pMsg->code)); - taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); - return; - } - - SStatusRsp *pStatusRsp = pMsg->pCont; - SMnodeInfos *minfos = &pStatusRsp->mnodes; - dnodeUpdateMInfos(minfos); - - SDnodeCfg *pCfg = &pStatusRsp->dnodeCfg; - pCfg->numOfVnodes = htonl(pCfg->numOfVnodes); - pCfg->moduleStatus = htonl(pCfg->moduleStatus); - pCfg->dnodeId = htonl(pCfg->dnodeId); - dnodeUpdateCfg(pCfg); - - vnodeSetAccess(pStatusRsp->vgAccess, pCfg->numOfVnodes); - - SDnodeEps *pEps = (SDnodeEps *)((char *)pStatusRsp->vgAccess + pCfg->numOfVnodes * sizeof(SVgroupAccess)); - dnodeUpdateEps(pEps); - - taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); -} - -static void dnodeSendStatusMsg(void *handle, void *tmrId) { - if (tsDnodeTmr == NULL) { - dError("dnode timer is already released"); - return; - } - - if (tsStatusTimer == NULL) { - taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); - dError("failed to start status timer"); - return; - } - - int32_t contLen = sizeof(SStatusMsg) + TSDB_MAX_VNODES * sizeof(SVnodeLoad); - SStatusMsg *pStatus = rpcMallocCont(contLen); - if (pStatus == NULL) { - taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); - dError("failed to malloc status message"); - return; - } - - dnodeGetCfg(&pStatus->dnodeId, pStatus->clusterId); - pStatus->dnodeId = htonl(dnodeGetDnodeId()); - pStatus->version = htonl(tsVersion); - pStatus->lastReboot = htonl(tsRebootTime); - pStatus->numOfCores = htons((uint16_t) tsNumOfCores); - pStatus->diskAvailable = tsAvailDataDirGB; - pStatus->alternativeRole = (uint8_t) tsAlternativeRole; - tstrncpy(pStatus->dnodeEp, tsLocalEp, TSDB_EP_LEN); - - // fill cluster cfg parameters - pStatus->clusterCfg.numOfMnodes = htonl(tsNumOfMnodes); - pStatus->clusterCfg.enableBalance = htonl(tsEnableBalance); - pStatus->clusterCfg.mnodeEqualVnodeNum = htonl(tsMnodeEqualVnodeNum); - pStatus->clusterCfg.offlineThreshold = htonl(tsOfflineThreshold); - pStatus->clusterCfg.statusInterval = htonl(tsStatusInterval); - pStatus->clusterCfg.maxtablesPerVnode = htonl(tsMaxTablePerVnode); - pStatus->clusterCfg.maxVgroupsPerDb = htonl(tsMaxVgroupsPerDb); - tstrncpy(pStatus->clusterCfg.arbitrator, tsArbitrator, TSDB_EP_LEN); - tstrncpy(pStatus->clusterCfg.timezone, tsTimezone, 64); - pStatus->clusterCfg.checkTime = 0; - char timestr[32] = "1970-01-01 00:00:00.00"; - (void)taosParseTime(timestr, &pStatus->clusterCfg.checkTime, strlen(timestr), TSDB_TIME_PRECISION_MILLI, 0); - tstrncpy(pStatus->clusterCfg.locale, tsLocale, TSDB_LOCALE_LEN); - tstrncpy(pStatus->clusterCfg.charset, tsCharset, TSDB_LOCALE_LEN); - - vnodeBuildStatusMsg(pStatus); - contLen = sizeof(SStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad); - pStatus->openVnodes = htons(pStatus->openVnodes); - - SRpcMsg rpcMsg = { - .pCont = pStatus, - .contLen = contLen, - .msgType = TSDB_MSG_TYPE_DM_STATUS - }; - - SRpcEpSet epSet; - dnodeGetEpSetForPeer(&epSet); - dnodeSendMsgToDnode(&epSet, &rpcMsg); -} - -void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) { - SRpcConnInfo connInfo = {0}; - rpcGetConnInfo(rpcMsg->handle, &connInfo); - - SRpcEpSet epSet = {0}; - if (forShell) { - dnodeGetEpSetForShell(&epSet); - } else { - dnodeGetEpSetForPeer(&epSet); - } - - dDebug("msg:%s will be redirected, dnodeIp:%s user:%s, numOfEps:%d inUse:%d", taosMsg[rpcMsg->msgType], - taosIpStr(connInfo.clientIp), connInfo.user, epSet.numOfEps, epSet.inUse); - - for (int i = 0; i < epSet.numOfEps; ++i) { - dDebug("mnode index:%d %s:%d", i, epSet.fqdn[i], epSet.port[i]); - epSet.port[i] = htons(epSet.port[i]); - } - - rpcSendRedirectRsp(rpcMsg->handle, &epSet); -} diff --git a/src/dnode/src/dnodeModule.c b/src/dnode/src/dnodeModule.c index bd9500ba51226d138fd3fe52f027d144d289681b..9eb52cbf5aee83852e3f97a89ecf82f06f41f09e 100644 --- a/src/dnode/src/dnodeModule.c +++ b/src/dnode/src/dnodeModule.c @@ -15,15 +15,10 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taosdef.h" -#include "taosmsg.h" -#include "tglobal.h" #include "mnode.h" #include "http.h" #include "tmqtt.h" #include "monitor.h" -#include "dnode.h" -#include "dnodeInt.h" #include "dnodeModule.h" typedef struct { @@ -78,10 +73,10 @@ static void dnodeAllocModules() { tsModule[TSDB_MOD_MONITOR].enable = (tsEnableMonitorModule == 1); tsModule[TSDB_MOD_MONITOR].name = "monitor"; - tsModule[TSDB_MOD_MONITOR].initFp = monitorInitSystem; - tsModule[TSDB_MOD_MONITOR].cleanUpFp = monitorCleanUpSystem; - tsModule[TSDB_MOD_MONITOR].startFp = monitorStartSystem; - tsModule[TSDB_MOD_MONITOR].stopFp = monitorStopSystem; + tsModule[TSDB_MOD_MONITOR].initFp = monInitSystem; + tsModule[TSDB_MOD_MONITOR].cleanUpFp = monCleanupSystem; + tsModule[TSDB_MOD_MONITOR].startFp = monStartSystem; + tsModule[TSDB_MOD_MONITOR].stopFp = monStopSystem; if (tsEnableMonitorModule) { dnodeSetModuleStatus(TSDB_MOD_MONITOR); } @@ -147,8 +142,8 @@ void dnodeProcessModuleStatus(uint32_t moduleStatus) { } } -bool dnodeStartMnode(SMnodeInfos *minfos) { - SMnodeInfos *mnodes = minfos; +bool dnodeStartMnode(SMInfos *pMinfos) { + SMInfos *pMnodes = pMinfos; if (tsModuleStatus & (1 << TSDB_MOD_MNODE)) { dDebug("mnode module is already started, module status:%d", tsModuleStatus); @@ -159,7 +154,7 @@ bool dnodeStartMnode(SMnodeInfos *minfos) { dInfo("start mnode module, module status:%d, new status:%d", tsModuleStatus, moduleStatus); dnodeProcessModuleStatus(moduleStatus); - sdbUpdateSync(mnodes); + sdbUpdateSync(pMnodes); return true; } diff --git a/src/dnode/src/dnodePeer.c b/src/dnode/src/dnodePeer.c index 4c44924cd0b79bae5b0c83f9ddad8ae30afcf292..bf74e14963e048c72cb57fdedfd788c92b90cbce 100644 --- a/src/dnode/src/dnodePeer.c +++ b/src/dnode/src/dnodePeer.c @@ -21,15 +21,12 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taosmsg.h" -#include "tglobal.h" #include "mnode.h" -#include "dnode.h" -#include "dnodeInt.h" -#include "dnodeMgmt.h" +#include "dnodeVMgmt.h" #include "dnodeVWrite.h" #include "dnodeMPeer.h" #include "dnodeMInfos.h" +#include "dnodeStep.h" static void (*dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *); static void dnodeProcessReqMsgFromDnode(SRpcMsg *pMsg, SRpcEpSet *); @@ -44,19 +41,19 @@ int32_t dnodeInitServer() { dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = dnodeDispatchToVWriteQueue; dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = dnodeDispatchToVWriteQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeDispatchToMgmtQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeDispatchToMgmtQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeDispatchToMgmtQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeDispatchToMgmtQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeDispatchToMgmtQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeDispatchToMgmtQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeDispatchToVMgmtQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeDispatchToVMgmtQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeDispatchToVMgmtQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeDispatchToVMgmtQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeDispatchToVMgmtQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeDispatchToVMgmtQueue; dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_CONFIG_TABLE] = dnodeDispatchToMPeerQueue; dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_CONFIG_VNODE] = dnodeDispatchToMPeerQueue; dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_AUTH] = dnodeDispatchToMPeerQueue; dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_GRANT] = dnodeDispatchToMPeerQueue; dnodeProcessReqMsgFp[TSDB_MSG_TYPE_DM_STATUS] = dnodeDispatchToMPeerQueue; - + SRpcInit rpcInit; memset(&rpcInit, 0, sizeof(rpcInit)); rpcInit.localPort = tsDnodeDnodePort; @@ -91,8 +88,9 @@ static void dnodeProcessReqMsgFromDnode(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { .pCont = NULL, .contLen = 0 }; - + if (pMsg->pCont == NULL) return; + if (pMsg->msgType == TSDB_MSG_TYPE_NETWORK_TEST) return dnodeSendStartupStep(pMsg); if (dnodeGetRunStatus() != TSDB_RUN_STATUS_RUNING) { rspMsg.code = TSDB_CODE_APP_NOT_READY; @@ -151,9 +149,9 @@ void dnodeCleanupClient() { } static void dnodeProcessRspFromDnode(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { - if (dnodeGetRunStatus() != TSDB_RUN_STATUS_RUNING) { + if (dnodeGetRunStatus() == TSDB_RUN_STATUS_STOPPED) { if (pMsg == NULL || pMsg->pCont == NULL) return; - dDebug("msg:%p is ignored since dnode not running", pMsg); + dDebug("msg:%p is ignored since dnode is stopping", pMsg); rpcFreeCont(pMsg->pCont); return; } @@ -182,6 +180,8 @@ void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg) { void dnodeSendMsgToMnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp) { SRpcEpSet epSet = {0}; dnodeGetEpSetForPeer(&epSet); + + assert(tsClientRpc != 0); rpcSendRecv(tsClientRpc, &epSet, rpcMsg, rpcRsp); } diff --git a/src/dnode/src/dnodeShell.c b/src/dnode/src/dnodeShell.c index 89f657f78986b8e57c0b5e1dedb841c451db00ba..d76af4e3dcf4027191e7cb92e0a9701e2c30c942 100644 --- a/src/dnode/src/dnodeShell.c +++ b/src/dnode/src/dnodeShell.c @@ -15,20 +15,14 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" -#include "taosdef.h" -#include "taosmsg.h" -#include "tglobal.h" -#include "tutil.h" #include "http.h" #include "mnode.h" -#include "dnode.h" -#include "dnodeInt.h" #include "dnodeVRead.h" #include "dnodeVWrite.h" #include "dnodeMRead.h" #include "dnodeMWrite.h" #include "dnodeShell.h" +#include "dnodeStep.h" static void (*dnodeProcessShellMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *); static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *); @@ -74,6 +68,8 @@ int32_t dnodeInitShell() { dnodeProcessShellMsgFp[TSDB_MSG_TYPE_CM_SHOW] = dnodeDispatchToMReadQueue; dnodeProcessShellMsgFp[TSDB_MSG_TYPE_CM_RETRIEVE] = dnodeDispatchToMReadQueue; + dnodeProcessShellMsgFp[TSDB_MSG_TYPE_NETWORK_TEST] = dnodeSendStartupStep; + int32_t numOfThreads = tsNumOfCores * tsNumOfThreadsPerCore; numOfThreads = (int32_t) ((1.0 - tsRatioOfQueryThreads) * numOfThreads / 2.0); if (numOfThreads < 1) { @@ -142,7 +138,23 @@ static void dnodeProcessMsgFromShell(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { } } +static int32_t dnodeAuthNettestUser(char *user, char *spi, char *encrypt, char *secret, char *ckey) { + if (strcmp(user, "nettestinternal") == 0) { + char pass[32] = {0}; + taosEncryptPass((uint8_t *)user, strlen(user), pass); + *spi = 0; + *encrypt = 0; + *ckey = 0; + memcpy(secret, pass, TSDB_KEY_LEN); + dTrace("nettest user is authorized"); + return 0; + } + + return -1; +} + static int dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char *secret, char *ckey) { + if (dnodeAuthNettestUser(user, spi, encrypt, secret, ckey) == 0) return 0; int code = mnodeRetriveAuth(user, spi, encrypt, secret, ckey); if (code != TSDB_CODE_APP_NOT_READY) return code; @@ -220,4 +232,4 @@ SStatisInfo dnodeGetStatisInfo() { } return info; -} +} \ No newline at end of file diff --git a/src/dnode/src/dnodeStep.c b/src/dnode/src/dnodeStep.c new file mode 100644 index 0000000000000000000000000000000000000000..0f535b9470ad53a62b294e82e118dd1e5e78175d --- /dev/null +++ b/src/dnode/src/dnodeStep.c @@ -0,0 +1,73 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "dnodeStep.h" + +static SStartupStep tsStartupStep; + +void dnodeReportStep(char *name, char *desc, int8_t finished) { + tstrncpy(tsStartupStep.name, name, sizeof(tsStartupStep.name)); + tstrncpy(tsStartupStep.desc, desc, sizeof(tsStartupStep.desc)); + tsStartupStep.finished = finished; +} + +void dnodeSendStartupStep(SRpcMsg *pMsg) { + dInfo("nettest msg is received, cont:%s", (char *)pMsg->pCont); + + SStartupStep *pStep = rpcMallocCont(sizeof(SStartupStep)); + memcpy(pStep, &tsStartupStep, sizeof(SStartupStep)); + + dDebug("startup msg is sent, step:%s desc:%s finished:%d", pStep->name, pStep->desc, pStep->finished); + + SRpcMsg rpcRsp = {.handle = pMsg->handle, .pCont = pStep, .contLen = sizeof(SStartupStep)}; + rpcSendResponse(&rpcRsp); + rpcFreeCont(pMsg->pCont); +} + +void taosStepCleanupImp(SStep *pSteps, int32_t stepId) { + for (int32_t step = stepId; step >= 0; step--) { + SStep *pStep = pSteps + step; + dDebug("step:%s will cleanup", pStep->name); + if (pStep->cleanupFp != NULL) { + (*pStep->cleanupFp)(); + } + } +} + +int32_t dnodeStepInit(SStep *pSteps, int32_t stepSize) { + for (int32_t step = 0; step < stepSize; step++) { + SStep *pStep = pSteps + step; + if (pStep->initFp == NULL) continue; + + dnodeReportStep(pStep->name, "Start initialization", 0); + + int32_t code = (*pStep->initFp)(); + if (code != 0) { + dDebug("step:%s will init", pStep->name); + taosStepCleanupImp(pSteps, step); + return code; + } + + dnodeReportStep(pStep->name, "Initialization complete", step + 1 >= stepSize); + } + + return 0; +} + +void dnodeStepCleanup(SStep *pSteps, int32_t stepSize) { + return taosStepCleanupImp(pSteps, stepSize - 1); +} \ No newline at end of file diff --git a/src/dnode/src/dnodeSystem.c b/src/dnode/src/dnodeSystem.c index 56316e9619344243ab2e5fa46e0acfc6e0331e8a..a135cda055434c46bd990541ecbf02e44efc14bf 100644 --- a/src/dnode/src/dnodeSystem.c +++ b/src/dnode/src/dnodeSystem.c @@ -16,9 +16,6 @@ #define _DEFAULT_SOURCE #include "os.h" #include "tgrant.h" -#include "tutil.h" -#include "tglobal.h" -#include "dnodeInt.h" #include "dnodeMain.h" static void signal_handler(int32_t signum, siginfo_t *sigInfo, void *context); diff --git a/src/dnode/src/dnodeTelemetry.c b/src/dnode/src/dnodeTelemetry.c index e973f9901f19b7aa6f4d4958d88e01ba258a82bb..85f0137d896ff65440117dcc6fc6e2a3c1a28038 100644 --- a/src/dnode/src/dnodeTelemetry.c +++ b/src/dnode/src/dnodeTelemetry.c @@ -15,9 +15,6 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" -#include "tglobal.h" -#include "tutil.h" #include "osTime.h" #include "tsocket.h" #include "tbuffer.h" @@ -32,8 +29,6 @@ #include "mnodeTable.h" #include "mnodeSdb.h" #include "mnodeAcct.h" -#include "dnode.h" -#include "dnodeInt.h" #include "dnodeTelemetry.h" static tsem_t tsExitSem; @@ -313,4 +308,4 @@ void dnodeCleanupTelemetry() { pthread_join(tsTelemetryThread, NULL); tsem_destroy(&tsExitSem); } -} +} \ No newline at end of file diff --git a/src/dnode/src/dnodeVMgmt.c b/src/dnode/src/dnodeVMgmt.c new file mode 100644 index 0000000000000000000000000000000000000000..87302026ecd4ba5e2192be3e7a50c58b5c906f6b --- /dev/null +++ b/src/dnode/src/dnodeVMgmt.c @@ -0,0 +1,239 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "tqueue.h" +#include "dnodeVMgmt.h" + +typedef struct { + SRpcMsg rpcMsg; + char pCont[]; +} SMgmtMsg; + +static taos_qset tsMgmtQset = NULL; +static taos_queue tsMgmtQueue = NULL; +static pthread_t tsQthread; + +static void * dnodeProcessMgmtQueue(void *param); +static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *pMsg); +static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *pMsg); +static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *pMsg); +static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg); +static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg); +static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg); +static int32_t (*dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *pMsg); + +int32_t dnodeInitVMgmt() { + dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = dnodeProcessCreateVnodeMsg; + dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = dnodeProcessAlterVnodeMsg; + dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = dnodeProcessDropVnodeMsg; + dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = dnodeProcessAlterStreamMsg; + dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CONFIG_DNODE] = dnodeProcessConfigDnodeMsg; + dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeProcessCreateMnodeMsg; + + int32_t code = vnodeInitMgmt(); + if (code != TSDB_CODE_SUCCESS) return -1; + + tsMgmtQset = taosOpenQset(); + if (tsMgmtQset == NULL) { + dError("failed to create the vmgmt queue set"); + return -1; + } + + tsMgmtQueue = taosOpenQueue(); + if (tsMgmtQueue == NULL) { + dError("failed to create the vmgmt queue"); + return -1; + } + + taosAddIntoQset(tsMgmtQset, tsMgmtQueue, NULL); + + pthread_attr_t thAttr; + pthread_attr_init(&thAttr); + pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); + + code = pthread_create(&tsQthread, &thAttr, dnodeProcessMgmtQueue, NULL); + pthread_attr_destroy(&thAttr); + if (code != 0) { + dError("failed to create thread to process vmgmt queue, reason:%s", strerror(errno)); + return -1; + } + + dInfo("dnode vmgmt is initialized"); + return TSDB_CODE_SUCCESS; +} + +void dnodeCleanupVMgmt() { + if (tsMgmtQset) taosQsetThreadResume(tsMgmtQset); + if (tsQthread) pthread_join(tsQthread, NULL); + + if (tsMgmtQueue) taosCloseQueue(tsMgmtQueue); + if (tsMgmtQset) taosCloseQset(tsMgmtQset); + + tsMgmtQset = NULL; + tsMgmtQueue = NULL; + + vnodeCleanupMgmt(); +} + +static int32_t dnodeWriteToMgmtQueue(SRpcMsg *pMsg) { + int32_t size = sizeof(SMgmtMsg) + pMsg->contLen; + SMgmtMsg *pMgmt = taosAllocateQitem(size); + if (pMgmt == NULL) return TSDB_CODE_DND_OUT_OF_MEMORY; + + pMgmt->rpcMsg = *pMsg; + pMgmt->rpcMsg.pCont = pMgmt->pCont; + memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen); + taosWriteQitem(tsMgmtQueue, TAOS_QTYPE_RPC, pMgmt); + + return TSDB_CODE_SUCCESS; +} + +void dnodeDispatchToVMgmtQueue(SRpcMsg *pMsg) { + int32_t code = dnodeWriteToMgmtQueue(pMsg); + if (code != TSDB_CODE_SUCCESS) { + SRpcMsg rsp = {.handle = pMsg->handle, .code = code}; + rpcSendResponse(&rsp); + } + + rpcFreeCont(pMsg->pCont); +} + +static void *dnodeProcessMgmtQueue(void *param) { + SMgmtMsg *pMgmt; + SRpcMsg * pMsg; + SRpcMsg rsp = {0}; + int32_t qtype; + void * handle; + + while (1) { + if (taosReadQitemFromQset(tsMgmtQset, &qtype, (void **)&pMgmt, &handle) == 0) { + dDebug("qset:%p, dnode mgmt got no message from qset, exit", tsMgmtQset); + break; + } + + pMsg = &pMgmt->rpcMsg; + dDebug("msg:%p, ahandle:%p type:%s will be processed", pMgmt, pMsg->ahandle, taosMsg[pMsg->msgType]); + if (dnodeProcessMgmtMsgFp[pMsg->msgType]) { + rsp.code = (*dnodeProcessMgmtMsgFp[pMsg->msgType])(pMsg); + } else { + rsp.code = TSDB_CODE_DND_MSG_NOT_PROCESSED; + } + + dDebug("msg:%p, is processed, code:0x%x", pMgmt, rsp.code); + if (rsp.code != TSDB_CODE_DND_ACTION_IN_PROGRESS) { + rsp.handle = pMsg->handle; + rsp.pCont = NULL; + rpcSendResponse(&rsp); + } + + taosFreeQitem(pMsg); + } + + return NULL; +} + +static SCreateVnodeMsg* dnodeParseVnodeMsg(SRpcMsg *rpcMsg) { + SCreateVnodeMsg *pCreate = rpcMsg->pCont; + pCreate->cfg.vgId = htonl(pCreate->cfg.vgId); + pCreate->cfg.cfgVersion = htonl(pCreate->cfg.cfgVersion); + pCreate->cfg.maxTables = htonl(pCreate->cfg.maxTables); + pCreate->cfg.cacheBlockSize = htonl(pCreate->cfg.cacheBlockSize); + pCreate->cfg.totalBlocks = htonl(pCreate->cfg.totalBlocks); + pCreate->cfg.daysPerFile = htonl(pCreate->cfg.daysPerFile); + pCreate->cfg.daysToKeep1 = htonl(pCreate->cfg.daysToKeep1); + pCreate->cfg.daysToKeep2 = htonl(pCreate->cfg.daysToKeep2); + pCreate->cfg.daysToKeep = htonl(pCreate->cfg.daysToKeep); + pCreate->cfg.minRowsPerFileBlock = htonl(pCreate->cfg.minRowsPerFileBlock); + pCreate->cfg.maxRowsPerFileBlock = htonl(pCreate->cfg.maxRowsPerFileBlock); + pCreate->cfg.fsyncPeriod = htonl(pCreate->cfg.fsyncPeriod); + pCreate->cfg.commitTime = htonl(pCreate->cfg.commitTime); + + for (int32_t j = 0; j < pCreate->cfg.replications; ++j) { + pCreate->nodes[j].nodeId = htonl(pCreate->nodes[j].nodeId); + } + + return pCreate; +} + +static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) { + SCreateVnodeMsg *pCreate = dnodeParseVnodeMsg(rpcMsg); + + void *pVnode = vnodeAcquire(pCreate->cfg.vgId); + if (pVnode != NULL) { + dDebug("vgId:%d, already exist, return success", pCreate->cfg.vgId); + vnodeRelease(pVnode); + return TSDB_CODE_SUCCESS; + } else { + dDebug("vgId:%d, create vnode msg is received", pCreate->cfg.vgId); + return vnodeCreate(pCreate); + } +} + +static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) { + SAlterVnodeMsg *pAlter = dnodeParseVnodeMsg(rpcMsg); + + void *pVnode = vnodeAcquire(pAlter->cfg.vgId); + if (pVnode != NULL) { + dDebug("vgId:%d, alter vnode msg is received", pAlter->cfg.vgId); + int32_t code = vnodeAlter(pVnode, pAlter); + vnodeRelease(pVnode); + return code; + } else { + dError("vgId:%d, vnode not exist, can't alter it", pAlter->cfg.vgId); + return TSDB_CODE_VND_INVALID_VGROUP_ID; + } +} + +static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) { + SDropVnodeMsg *pDrop = rpcMsg->pCont; + pDrop->vgId = htonl(pDrop->vgId); + + return vnodeDrop(pDrop->vgId); +} + +static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg) { + return 0; +} + +static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg) { + SCfgDnodeMsg *pCfg = pMsg->pCont; + return taosCfgDynamicOptions(pCfg->config); +} + +static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) { + SCreateMnodeMsg *pCfg = pMsg->pCont; + pCfg->dnodeId = htonl(pCfg->dnodeId); + if (pCfg->dnodeId != dnodeGetDnodeId()) { + dDebug("dnodeId:%d, in create mnode msg is not equal with saved dnodeId:%d", pCfg->dnodeId, dnodeGetDnodeId()); + return TSDB_CODE_MND_DNODE_ID_NOT_CONFIGURED; + } + + if (strcmp(pCfg->dnodeEp, tsLocalEp) != 0) { + dDebug("dnodeEp:%s, in create mnode msg is not equal with saved dnodeEp:%s", pCfg->dnodeEp, tsLocalEp); + return TSDB_CODE_MND_DNODE_EP_NOT_CONFIGURED; + } + + dDebug("dnodeId:%d, create mnode msg is received from mnodes, numOfMnodes:%d", pCfg->dnodeId, pCfg->mnodes.mnodeNum); + for (int i = 0; i < pCfg->mnodes.mnodeNum; ++i) { + pCfg->mnodes.mnodeInfos[i].mnodeId = htonl(pCfg->mnodes.mnodeInfos[i].mnodeId); + dDebug("mnode index:%d, mnode:%d:%s", i, pCfg->mnodes.mnodeInfos[i].mnodeId, pCfg->mnodes.mnodeInfos[i].mnodeEp); + } + + dnodeStartMnode(&pCfg->mnodes); + + return TSDB_CODE_SUCCESS; +} \ No newline at end of file diff --git a/src/dnode/src/dnodeVRead.c b/src/dnode/src/dnodeVRead.c index b42a627a3a78031be3191d7be5fb1202723dcc57..07496b142a210904e9842c0fe7a0ad613e41ba7a 100644 --- a/src/dnode/src/dnodeVRead.c +++ b/src/dnode/src/dnodeVRead.c @@ -15,12 +15,8 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" -#include "taosmsg.h" -#include "tglobal.h" #include "tqueue.h" -#include "vnode.h" -#include "dnodeInt.h" +#include "dnodeVRead.h" typedef struct { pthread_t thread; // thread diff --git a/src/dnode/src/dnodeVWrite.c b/src/dnode/src/dnodeVWrite.c index 6d4b50ee542b565669f0a4782142a3919c19ebe8..a5ae8ac83063c599ad4c215bd0a7fa4468810580 100644 --- a/src/dnode/src/dnodeVWrite.c +++ b/src/dnode/src/dnodeVWrite.c @@ -15,13 +15,8 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" -#include "taosmsg.h" -#include "tglobal.h" #include "tqueue.h" -#include "twal.h" -#include "vnode.h" -#include "dnodeInt.h" +#include "dnodeVWrite.h" typedef struct { taos_qall qall; diff --git a/src/dnode/src/dnodeVnodes.c b/src/dnode/src/dnodeVnodes.c new file mode 100644 index 0000000000000000000000000000000000000000..ba7f7625fa7f71d4c2db5b5570c6d126899d9deb --- /dev/null +++ b/src/dnode/src/dnodeVnodes.c @@ -0,0 +1,289 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "ttimer.h" +#include "dnodeEps.h" +#include "dnodeCfg.h" +#include "dnodeMInfos.h" +#include "dnodeVnodes.h" + +typedef struct { + pthread_t thread; + int32_t threadIndex; + int32_t failed; + int32_t opened; + int32_t vnodeNum; + int32_t * vnodeList; +} SOpenVnodeThread; + +void * tsDnodeTmr = NULL; +static void * tsStatusTimer = NULL; +static uint32_t tsRebootTime = 0; + +static void dnodeSendStatusMsg(void *handle, void *tmrId); +static void dnodeProcessStatusRsp(SRpcMsg *pMsg); + +int32_t dnodeInitTimer() { + tsDnodeTmr = taosTmrInit(100, 200, 60000, "DND-DM"); + if (tsDnodeTmr == NULL) { + dError("failed to init dnode timer"); + return -1; + } + + dnodeAddClientRspHandle(TSDB_MSG_TYPE_DM_STATUS_RSP, dnodeProcessStatusRsp); + + tsRebootTime = taosGetTimestampSec(); + taosTmrReset(dnodeSendStatusMsg, 500, NULL, tsDnodeTmr, &tsStatusTimer); + + dInfo("dnode timer is initialized"); + return TSDB_CODE_SUCCESS; +} + +void dnodeCleanupTimer() { + if (tsStatusTimer != NULL) { + taosTmrStopA(&tsStatusTimer); + tsStatusTimer = NULL; + } + + if (tsDnodeTmr != NULL) { + taosTmrCleanUp(tsDnodeTmr); + tsDnodeTmr = NULL; + } +} + +static int32_t dnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) { + DIR *dir = opendir(tsVnodeDir); + if (dir == NULL) return TSDB_CODE_DND_NO_WRITE_ACCESS; + + *numOfVnodes = 0; + struct dirent *de = NULL; + while ((de = readdir(dir)) != NULL) { + if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue; + if (de->d_type & DT_DIR) { + if (strncmp("vnode", de->d_name, 5) != 0) continue; + int32_t vnode = atoi(de->d_name + 5); + if (vnode == 0) continue; + + (*numOfVnodes)++; + + if (*numOfVnodes >= TSDB_MAX_VNODES) { + dError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES); + continue; + } else { + vnodeList[*numOfVnodes - 1] = vnode; + } + } + } + closedir(dir); + + return TSDB_CODE_SUCCESS; +} + +static void *dnodeOpenVnode(void *param) { + SOpenVnodeThread *pThread = param; + + dDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum); + + for (int32_t v = 0; v < pThread->vnodeNum; ++v) { + int32_t vgId = pThread->vnodeList[v]; + if (vnodeOpen(vgId) < 0) { + dError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex); + pThread->failed++; + } else { + dDebug("vgId:%d, is openned by thread:%d", vgId, pThread->threadIndex); + pThread->opened++; + } + } + + dDebug("thread:%d, total vnodes:%d, openned:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened, + pThread->failed); + return NULL; +} + +int32_t dnodeInitVnodes() { + int32_t vnodeList[TSDB_MAX_VNODES] = {0}; + int32_t numOfVnodes = 0; + int32_t status = dnodeGetVnodeList(vnodeList, &numOfVnodes); + + if (status != TSDB_CODE_SUCCESS) { + dInfo("get dnode list failed"); + return status; + } + + int32_t threadNum = tsNumOfCores; + int32_t vnodesPerThread = numOfVnodes / threadNum + 1; + SOpenVnodeThread *threads = calloc(threadNum, sizeof(SOpenVnodeThread)); + for (int32_t t = 0; t < threadNum; ++t) { + threads[t].threadIndex = t; + threads[t].vnodeList = calloc(vnodesPerThread, sizeof(int32_t)); + } + + for (int32_t v = 0; v < numOfVnodes; ++v) { + int32_t t = v % threadNum; + SOpenVnodeThread *pThread = &threads[t]; + pThread->vnodeList[pThread->vnodeNum++] = vnodeList[v]; + } + + dDebug("start %d threads to open %d vnodes", threadNum, numOfVnodes); + + for (int32_t t = 0; t < threadNum; ++t) { + SOpenVnodeThread *pThread = &threads[t]; + if (pThread->vnodeNum == 0) continue; + + pthread_attr_t thAttr; + pthread_attr_init(&thAttr); + pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); + if (pthread_create(&pThread->thread, &thAttr, dnodeOpenVnode, pThread) != 0) { + dError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno)); + } + + pthread_attr_destroy(&thAttr); + } + + int32_t openVnodes = 0; + int32_t failedVnodes = 0; + for (int32_t t = 0; t < threadNum; ++t) { + SOpenVnodeThread *pThread = &threads[t]; + if (pThread->vnodeNum > 0 && pThread->thread) { + pthread_join(pThread->thread, NULL); + } + openVnodes += pThread->opened; + failedVnodes += pThread->failed; + free(pThread->vnodeList); + } + + free(threads); + dInfo("there are total vnodes:%d, openned:%d", numOfVnodes, openVnodes); + + if (failedVnodes != 0) { + dError("there are total vnodes:%d, failed:%d", numOfVnodes, failedVnodes); + return -1; + } + + return TSDB_CODE_SUCCESS; +} + +void dnodeCleanupVnodes() { + int32_t vnodeList[TSDB_MAX_VNODES]= {0}; + int32_t numOfVnodes = 0; + int32_t status; + + status = vnodeGetVnodeList(vnodeList, &numOfVnodes); + + if (status != TSDB_CODE_SUCCESS) { + dInfo("get dnode list failed"); + return; + } + + for (int32_t i = 0; i < numOfVnodes; ++i) { + vnodeClose(vnodeList[i]); + } + + dInfo("total vnodes:%d are all closed", numOfVnodes); +} + +static void dnodeProcessStatusRsp(SRpcMsg *pMsg) { + if (pMsg->code != TSDB_CODE_SUCCESS) { + dError("status rsp is received, error:%s", tstrerror(pMsg->code)); + taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); + return; + } + + SStatusRsp *pStatusRsp = pMsg->pCont; + SMInfos *minfos = &pStatusRsp->mnodes; + dnodeUpdateMInfos(minfos); + + SDnodeCfg *pCfg = &pStatusRsp->dnodeCfg; + pCfg->numOfVnodes = htonl(pCfg->numOfVnodes); + pCfg->moduleStatus = htonl(pCfg->moduleStatus); + pCfg->dnodeId = htonl(pCfg->dnodeId); + dnodeUpdateCfg(pCfg); + + vnodeSetAccess(pStatusRsp->vgAccess, pCfg->numOfVnodes); + + SDnodeEps *pEps = (SDnodeEps *)((char *)pStatusRsp->vgAccess + pCfg->numOfVnodes * sizeof(SVgroupAccess)); + dnodeUpdateEps(pEps); + + taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); +} + +static void dnodeSendStatusMsg(void *handle, void *tmrId) { + if (tsDnodeTmr == NULL) { + dError("dnode timer is already released"); + return; + } + + if (tsStatusTimer == NULL) { + taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); + dError("failed to start status timer"); + return; + } + + int32_t contLen = sizeof(SStatusMsg) + TSDB_MAX_VNODES * sizeof(SVnodeLoad); + SStatusMsg *pStatus = rpcMallocCont(contLen); + if (pStatus == NULL) { + taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); + dError("failed to malloc status message"); + return; + } + + dnodeGetCfg(&pStatus->dnodeId, pStatus->clusterId); + pStatus->dnodeId = htonl(dnodeGetDnodeId()); + pStatus->version = htonl(tsVersion); + pStatus->lastReboot = htonl(tsRebootTime); + pStatus->numOfCores = htons((uint16_t) tsNumOfCores); + pStatus->diskAvailable = tsAvailDataDirGB; + pStatus->alternativeRole = (uint8_t) tsAlternativeRole; + tstrncpy(pStatus->dnodeEp, tsLocalEp, TSDB_EP_LEN); + + // fill cluster cfg parameters + pStatus->clusterCfg.numOfMnodes = htonl(tsNumOfMnodes); + pStatus->clusterCfg.enableBalance = htonl(tsEnableBalance); + pStatus->clusterCfg.mnodeEqualVnodeNum = htonl(tsMnodeEqualVnodeNum); + pStatus->clusterCfg.offlineThreshold = htonl(tsOfflineThreshold); + pStatus->clusterCfg.statusInterval = htonl(tsStatusInterval); + pStatus->clusterCfg.maxtablesPerVnode = htonl(tsMaxTablePerVnode); + pStatus->clusterCfg.maxVgroupsPerDb = htonl(tsMaxVgroupsPerDb); + tstrncpy(pStatus->clusterCfg.arbitrator, tsArbitrator, TSDB_EP_LEN); + tstrncpy(pStatus->clusterCfg.timezone, tsTimezone, 64); + pStatus->clusterCfg.checkTime = 0; + char timestr[32] = "1970-01-01 00:00:00.00"; + (void)taosParseTime(timestr, &pStatus->clusterCfg.checkTime, strlen(timestr), TSDB_TIME_PRECISION_MILLI, 0); + tstrncpy(pStatus->clusterCfg.locale, tsLocale, TSDB_LOCALE_LEN); + tstrncpy(pStatus->clusterCfg.charset, tsCharset, TSDB_LOCALE_LEN); + + vnodeBuildStatusMsg(pStatus); + contLen = sizeof(SStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad); + pStatus->openVnodes = htons(pStatus->openVnodes); + + SRpcMsg rpcMsg = { + .pCont = pStatus, + .contLen = contLen, + .msgType = TSDB_MSG_TYPE_DM_STATUS + }; + + SRpcEpSet epSet; + dnodeGetEpSetForPeer(&epSet); + dnodeSendMsgToDnode(&epSet, &rpcMsg); +} + +void dnodeSendStatusMsgToMnode() { + if (tsDnodeTmr != NULL && tsStatusTimer != NULL) { + dInfo("force send status msg to mnode"); + taosTmrReset(dnodeSendStatusMsg, 3, NULL, tsDnodeTmr, &tsStatusTimer); + } +} \ No newline at end of file diff --git a/src/inc/dnode.h b/src/inc/dnode.h index eef4490800a4191c2dee55c450cf99b8381bb64d..9dd95e32d759166490230bc06beac6b652b4fe77 100644 --- a/src/inc/dnode.h +++ b/src/inc/dnode.h @@ -45,7 +45,7 @@ void dnodeGetEpSetForShell(SRpcEpSet *epSet); int32_t dnodeGetDnodeId(); void dnodeUpdateEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port); bool dnodeCheckEpChanged(int32_t dnodeId, char *epstr); -bool dnodeStartMnode(SMnodeInfos *minfos); +bool dnodeStartMnode(SMInfos *pMinfos); void dnodeAddClientRspHandle(uint8_t msgType, void (*fp)(SRpcMsg *rpcMsg)); void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg); @@ -71,8 +71,18 @@ void dnodeDelayReprocessMWriteMsg(void *pMsg); void dnodeSendStatusMsgToMnode(); +typedef struct { + char *name; + int32_t (*initFp)(); + void (*cleanupFp)(); +} SStep; + +int32_t dnodeStepInit(SStep *pSteps, int32_t stepSize); +void dnodeStepCleanup(SStep *pSteps, int32_t stepSize); +void dnodeReportStep(char *name, char *desc, int8_t finished); + #ifdef __cplusplus } #endif -#endif +#endif \ No newline at end of file diff --git a/src/inc/monitor.h b/src/inc/monitor.h index b1229cca6b1fc18ac9bef4d6ea506c9a0ef59627..1aefb0b84887270c7997c381402096006fe6eaee 100644 --- a/src/inc/monitor.h +++ b/src/inc/monitor.h @@ -47,13 +47,13 @@ typedef struct { int8_t accessState; } SAcctMonitorObj; -int32_t monitorInitSystem(); -int32_t monitorStartSystem(); -void monitorStopSystem(); -void monitorCleanUpSystem(); -void monitorSaveAcctLog(SAcctMonitorObj *pMonObj); -void monitorSaveLog(int32_t level, const char *const format, ...); -void monitorExecuteSQL(char *sql); +int32_t monInitSystem(); +int32_t monStartSystem(); +void monStopSystem(); +void monCleanupSystem(); +void monSaveAcctLog(SAcctMonitorObj *pMonObj); +void monSaveLog(int32_t level, const char *const format, ...); +void monExecuteSQL(char *sql); #ifdef __cplusplus } diff --git a/src/inc/taosdef.h b/src/inc/taosdef.h index 20c7af6a21d9a385d00fb3e52fe2a6ffd3d0729e..0cc06be1dbcafbaba1ab00b041aff7af9db6e445 100644 --- a/src/inc/taosdef.h +++ b/src/inc/taosdef.h @@ -257,7 +257,7 @@ void tsDataSwap(void *pLeft, void *pRight, int32_t type, int32_t size, void* buf #define TSDB_MAX_SAVED_SQL_LEN TSDB_MAX_COLUMNS * 64 #define TSDB_MAX_SQL_LEN TSDB_PAYLOAD_SIZE #define TSDB_MAX_SQL_SHOW_LEN 512 -#define TSDB_MAX_ALLOWED_SQL_LEN (8*1024*1024U) // sql length should be less than 8mb +#define TSDB_MAX_ALLOWED_SQL_LEN (1*1024*1024U) // sql length should be less than 1mb #define TSDB_APPNAME_LEN TSDB_UNI_LEN @@ -286,6 +286,9 @@ void tsDataSwap(void *pLeft, void *pRight, int32_t type, int32_t size, void* buf #define TSDB_SHOW_SQL_LEN 512 #define TSDB_SLOW_QUERY_SQL_LEN 512 +#define TSDB_STEP_NAME_LEN 32 +#define TSDB_STEP_DESC_LEN 128 + #define TSDB_MQTT_HOSTNAME_LEN 64 #define TSDB_MQTT_PORT_LEN 8 #define TSDB_MQTT_USER_LEN 24 @@ -428,11 +431,6 @@ void tsDataSwap(void *pLeft, void *pRight, int32_t type, int32_t size, void* buf #define TSDB_PORT_SYNC 10 #define TSDB_PORT_HTTP 11 #define TSDB_PORT_ARBITRATOR 12 -#define TSDB_PORT_DNODESHELL 0 -#define TSDB_PORT_DNODEDNODE 5 -#define TSDB_PORT_SYNC 10 -#define TSDB_PORT_HTTP 11 -#define TSDB_PORT_ARBITRATOR 12 #define TSDB_MAX_WAL_SIZE (1024*1024) @@ -444,6 +442,10 @@ typedef enum { TAOS_QTYPE_QUERY = 4 } EQType; +#define TSDB_MAX_TIERS 3 +#define TSDB_MAX_DISKS_PER_TIER 16 +#define TSDB_MAX_DISKS (TSDB_MAX_TIERS * TSDB_MAX_DISKS_PER_TIER) + typedef enum { TSDB_SUPER_TABLE = 0, // super table TSDB_CHILD_TABLE = 1, // table created from super table diff --git a/src/inc/taoserror.h b/src/inc/taoserror.h index 77ec5350ba0205ebff94e8b740eb90b458cf12e6..e0d7e018434f15d96fe559440ba6b07f027cd930 100644 --- a/src/inc/taoserror.h +++ b/src/inc/taoserror.h @@ -64,7 +64,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_RPC_INVALID_TRAN_ID, 0, 0x000F, "Invalid tr TAOS_DEFINE_ERROR(TSDB_CODE_RPC_INVALID_SESSION_ID, 0, 0x0010, "Invalid session id") TAOS_DEFINE_ERROR(TSDB_CODE_RPC_INVALID_MSG_TYPE, 0, 0x0011, "Invalid message type") TAOS_DEFINE_ERROR(TSDB_CODE_RPC_INVALID_RESPONSE_TYPE, 0, 0x0012, "Invalid response type") -TAOS_DEFINE_ERROR(TSDB_CODE_RPC_INVALID_TIME_STAMP, 0, 0x0013, "Invalid timestamp") +TAOS_DEFINE_ERROR(TSDB_CODE_RPC_INVALID_TIME_STAMP, 0, 0x0013, "Client and server's time is not synchronized") TAOS_DEFINE_ERROR(TSDB_CODE_APP_NOT_READY, 0, 0x0014, "Database not ready") TAOS_DEFINE_ERROR(TSDB_CODE_RPC_FQDN_ERROR, 0, 0x0015, "Unable to resolve FQDN") @@ -184,11 +184,15 @@ TAOS_DEFINE_ERROR(TSDB_CODE_MND_TOO_MANY_DATABASES, 0, 0x0385, "Too many d TAOS_DEFINE_ERROR(TSDB_CODE_MND_DB_IN_DROPPING, 0, 0x0386, "Database not available") TAOS_DEFINE_ERROR(TSDB_CODE_MND_VGROUP_NOT_READY, 0, 0x0387, "Database unsynced") +TAOS_DEFINE_ERROR(TSDB_CODE_MND_INVALID_DB_OPTION_DAYS, 0, 0x0390, "Invalid database option: days out of range") +TAOS_DEFINE_ERROR(TSDB_CODE_MND_INVALID_DB_OPTION_KEEP, 0, 0x0391, "Invalid database option: keep >= keep2 >= keep1 >= days") + // dnode TAOS_DEFINE_ERROR(TSDB_CODE_DND_MSG_NOT_PROCESSED, 0, 0x0400, "Message not processed") TAOS_DEFINE_ERROR(TSDB_CODE_DND_OUT_OF_MEMORY, 0, 0x0401, "Dnode out of memory") TAOS_DEFINE_ERROR(TSDB_CODE_DND_NO_WRITE_ACCESS, 0, 0x0402, "No permission for disk files in dnode") TAOS_DEFINE_ERROR(TSDB_CODE_DND_INVALID_MSG_LEN, 0, 0x0403, "Invalid message length") +TAOS_DEFINE_ERROR(TSDB_CODE_DND_ACTION_IN_PROGRESS, 0, 0x0404, "Action in progress") // vnode TAOS_DEFINE_ERROR(TSDB_CODE_VND_ACTION_IN_PROGRESS, 0, 0x0500, "Action in progress") @@ -367,6 +371,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_HTTP_OP_TAG_VALUE_TOO_LONG, 0, 0x11A4, "tag value TAOS_DEFINE_ERROR(TSDB_CODE_HTTP_OP_VALUE_NULL, 0, 0x11A5, "value not find") TAOS_DEFINE_ERROR(TSDB_CODE_HTTP_OP_VALUE_TYPE, 0, 0x11A6, "value type should be boolean, number or string") +// odbc TAOS_DEFINE_ERROR(TSDB_CODE_ODBC_OOM, 0, 0x2100, "out of memory") TAOS_DEFINE_ERROR(TSDB_CODE_ODBC_CONV_CHAR_NOT_NUM, 0, 0x2101, "convertion not a valid literal input") TAOS_DEFINE_ERROR(TSDB_CODE_ODBC_CONV_UNDEF, 0, 0x2102, "convertion undefined") @@ -390,6 +395,17 @@ TAOS_DEFINE_ERROR(TSDB_CODE_ODBC_CONV_SRC_BAD_SEQ, 0, 0x2113, "src bad se TAOS_DEFINE_ERROR(TSDB_CODE_ODBC_CONV_SRC_INCOMPLETE, 0, 0x2114, "src incomplete") TAOS_DEFINE_ERROR(TSDB_CODE_ODBC_CONV_SRC_GENERAL, 0, 0x2115, "src general") +// tfs +TAOS_DEFINE_ERROR(TSDB_CODE_FS_OUT_OF_MEMORY, 0, 0x2200, "tfs out of memory") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_INVLD_CFG, 0, 0x2201, "tfs invalid mount config") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_TOO_MANY_MOUNT, 0, 0x2202, "tfs too many mount") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_DUP_PRIMARY, 0, 0x2203, "tfs duplicate primary mount") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_NO_PRIMARY_DISK, 0, 0x2204, "tfs no primary mount") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_NO_MOUNT_AT_TIER, 0, 0x2205, "tfs no mount at tier") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_FILE_ALREADY_EXISTS, 0, 0x2206, "tfs file already exists") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_INVLD_LEVEL, 0, 0x2207, "tfs invalid level") +TAOS_DEFINE_ERROR(TSDB_CODE_FS_NO_VALID_DISK, 0, 0x2208, "tfs no valid disk") + #ifdef TAOS_ERROR_C }; diff --git a/src/inc/taosmsg.h b/src/inc/taosmsg.h index b77db69c46038419703944764cbb2837674d1c21..ea9d608d92a3fb84ec447ce0ea00ca8ce388f8f4 100644 --- a/src/inc/taosmsg.h +++ b/src/inc/taosmsg.h @@ -105,10 +105,7 @@ TAOS_DEFINE_MESSAGE_TYPE( TSDB_MSG_TYPE_DM_AUTH, "auth" ) TAOS_DEFINE_MESSAGE_TYPE( TSDB_MSG_TYPE_DUMMY12, "dummy12" ) TAOS_DEFINE_MESSAGE_TYPE( TSDB_MSG_TYPE_DUMMY13, "dummy13" ) TAOS_DEFINE_MESSAGE_TYPE( TSDB_MSG_TYPE_DUMMY14, "dummy14" ) - - -TAOS_DEFINE_MESSAGE_TYPE( TSDB_MSG_TYPE_NETWORK_TEST, "network-test" ) - +TAOS_DEFINE_MESSAGE_TYPE( TSDB_MSG_TYPE_NETWORK_TEST, "nettest" ) #ifndef TAOS_MESSAGE_C TSDB_MSG_TYPE_MAX // 105 @@ -476,19 +473,21 @@ typedef struct { int16_t numOfGroupCols; // num of group by columns int16_t orderByIdx; int16_t orderType; // used in group by xx order by xxx + int64_t tableLimit; // limit the number of rows for each table, used in order by + limit in stable projection query. + int16_t prjOrder; // global order in super table projection query. int64_t limit; int64_t offset; uint32_t queryType; // denote another query process int16_t numOfOutput; // final output columns numbers int16_t tagNameRelType; // relation of tag criteria and tbname criteria - int16_t fillType; // interpolate type - uint64_t fillVal; // default value array list + int16_t fillType; // interpolate type + uint64_t fillVal; // default value array list int32_t secondStageOutput; - int32_t tsOffset; // offset value in current msg body, NOTE: ts list is compressed - int32_t tsLen; // total length of ts comp block - int32_t tsNumOfBlocks; // ts comp block numbers - int32_t tsOrder; // ts comp block order - int32_t numOfTags; // number of tags columns involved + int32_t tsOffset; // offset value in current msg body, NOTE: ts list is compressed + int32_t tsLen; // total length of ts comp block + int32_t tsNumOfBlocks; // ts comp block numbers + int32_t tsOrder; // ts comp block order + int32_t numOfTags; // number of tags columns involved SColumnInfo colList[]; } SQueryTableMsg; @@ -589,13 +588,13 @@ typedef struct { typedef struct { int32_t mnodeId; char mnodeEp[TSDB_EP_LEN]; -} SMnodeInfo; +} SMInfo; typedef struct { - int8_t inUse; - int8_t mnodeNum; - SMnodeInfo mnodeInfos[TSDB_MAX_REPLICA]; -} SMnodeInfos; + int8_t inUse; + int8_t mnodeNum; + SMInfo mnodeInfos[TSDB_MAX_REPLICA]; +} SMInfos; typedef struct { int32_t numOfMnodes; // tsNumOfMnodes @@ -630,7 +629,7 @@ typedef struct { } SStatusMsg; typedef struct { - SMnodeInfos mnodes; + SMInfos mnodes; SDnodeCfg dnodeCfg; SVgroupAccess vgAccess[]; } SStatusRsp; @@ -759,7 +758,7 @@ typedef struct { typedef struct { int32_t dnodeId; char dnodeEp[TSDB_EP_LEN]; // end point, hostname:port - SMnodeInfos mnodes; + SMInfos mnodes; } SCreateMnodeMsg; typedef struct { @@ -838,6 +837,14 @@ typedef struct { char ckey[TSDB_KEY_LEN]; } SAuthMsg, SAuthRsp; +typedef struct { + int8_t finished; + int8_t reserved1[7]; + char name[TSDB_STEP_NAME_LEN]; + char desc[TSDB_STEP_DESC_LEN]; + char reserved2[64]; +} SStartupStep; + #pragma pack(pop) #ifdef __cplusplus diff --git a/src/inc/tbalance.h b/src/inc/tbn.h similarity index 71% rename from src/inc/tbalance.h rename to src/inc/tbn.h index f0da4a37476157e235993ef8c2c32baedabc2a1e..b9f4e3c608a1ae3df3a4ea0dca32c7bf9d5820a9 100644 --- a/src/inc/tbalance.h +++ b/src/inc/tbn.h @@ -23,14 +23,14 @@ extern "C" { struct SVgObj; struct SDnodeObj; -int32_t balanceInit(); -void balanceCleanUp(); -void balanceAsyncNotify(); -void balanceSyncNotify(); -void balanceReset(); -int32_t balanceAllocVnodes(struct SVgObj *pVgroup); -int32_t balanceAlterDnode(struct SDnodeObj *pDnode, int32_t vnodeId, int32_t dnodeId); -int32_t balanceDropDnode(struct SDnodeObj *pDnode); +int32_t bnInit(); +void bnCleanUp(); +void bnNotify(); +void bnCheckModules(); +void bnReset(); +int32_t bnAllocVnodes(struct SVgObj *pVgroup); +int32_t bnAlterDnode(struct SDnodeObj *pDnode, int32_t vnodeId, int32_t dnodeId); +int32_t bnDropDnode(struct SDnodeObj *pDnode); #ifdef __cplusplus } diff --git a/src/inc/tcq.h b/src/inc/tcq.h index 7a0727f1b8dd7cdea1a815c174b9f85004eedf87..afa744a9c4856521075fb9427d0ed1bb412aa7d5 100644 --- a/src/inc/tcq.h +++ b/src/inc/tcq.h @@ -21,7 +21,7 @@ extern "C" { #include "tdataformat.h" -typedef int32_t (*FCqWrite)(void *ahandle, void *pHead, int32_t qtype, void *pMsg); +typedef int32_t (*FCqWrite)(int32_t vgId, void *pHead, int32_t qtype, void *pMsg); typedef struct { int32_t vgId; diff --git a/src/inc/tsdb.h b/src/inc/tsdb.h index 58859f42bc80daa3317d789950c1625c1533cf5f..1769bd6566bbc27d75bf6b830add8ef79356f111 100644 --- a/src/inc/tsdb.h +++ b/src/inc/tsdb.h @@ -321,7 +321,7 @@ void tsdbCleanupQueryHandle(TsdbQueryHandleT queryHandle); */ void tsdbReportStat(void *repo, int64_t *totalPoints, int64_t *totalStorage, int64_t *compStorage); -int tsdbInitCommitQueue(int nthreads); +int tsdbInitCommitQueue(); void tsdbDestroyCommitQueue(); int tsdbSyncCommit(TSDB_REPO_T *repo); void tsdbIncCommitRef(int vgId); diff --git a/src/inc/tsync.h b/src/inc/tsync.h index 398e1bf97c739406d1f8f568c998d891191b71bf..1303195ef149a98b6ddd7ffe55615d27eaeba04d 100644 --- a/src/inc/tsync.h +++ b/src/inc/tsync.h @@ -85,6 +85,9 @@ typedef void (*FNotifyFlowCtrl)(int32_t vgId, int32_t level); // when data file is synced successfully, notity app typedef int32_t (*FNotifyFileSynced)(int32_t vgId, uint64_t fversion); +// get file version +typedef int32_t (*FGetVersion)(int32_t vgId, uint64_t *fver, uint64_t *vver); + typedef struct { int32_t vgId; // vgroup ID uint64_t version; // initial version @@ -97,6 +100,7 @@ typedef struct { FNotifyRole notifyRole; FNotifyFlowCtrl notifyFlowCtrl; FNotifyFileSynced notifyFileSynced; + FGetVersion getVersion; } SSyncInfo; typedef void *tsync_h; diff --git a/src/inc/twal.h b/src/inc/twal.h index 8dd3a8a91209e840abeb9560f94a52ce362492a9..1645de77aacf020048dc21cd47907163089f269f 100644 --- a/src/inc/twal.h +++ b/src/inc/twal.h @@ -51,9 +51,8 @@ typedef struct { typedef void * twalh; // WAL HANDLE typedef int32_t FWalWrite(void *ahandle, void *pHead, int32_t qtype, void *pMsg); -int32_t walInit(); -void walCleanUp(); - +int32_t walInit(); +void walCleanUp(); twalh walOpen(char *path, SWalCfg *pCfg); int32_t walAlter(twalh pWal, SWalCfg *pCfg); void walStop(twalh); diff --git a/src/inc/vnode.h b/src/inc/vnode.h index 5f643295d6e3d70999de55da47470ad0f0c0df93..33bb2c278c4693347b1641349de8cd269229e77a 100644 --- a/src/inc/vnode.h +++ b/src/inc/vnode.h @@ -19,18 +19,9 @@ #ifdef __cplusplus extern "C" { #endif - #include "trpc.h" #include "twal.h" -typedef enum _VN_STATUS { - TAOS_VN_STATUS_INIT = 0, - TAOS_VN_STATUS_READY = 1, - TAOS_VN_STATUS_CLOSING = 2, - TAOS_VN_STATUS_UPDATING = 3, - TAOS_VN_STATUS_RESET = 4, -} EVnodeStatus; - typedef struct { int32_t len; void * rsp; @@ -60,29 +51,35 @@ typedef struct { SWalHead pHead[]; } SVWriteMsg; +// vnodeStatus extern char *vnodeStatus[]; +// vnodeMain int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg); int32_t vnodeDrop(int32_t vgId); -int32_t vnodeOpen(int32_t vgId, char *rootDir); +int32_t vnodeOpen(int32_t vgId); int32_t vnodeAlter(void *pVnode, SCreateVnodeMsg *pVnodeCfg); int32_t vnodeClose(int32_t vgId); -void* vnodeAcquire(int32_t vgId); // add refcount -void vnodeRelease(void *pVnode); // dec refCount +// vnodeMgmt +int32_t vnodeInitMgmt(); +void vnodeCleanupMgmt(); +void* vnodeAcquire(int32_t vgId); +void vnodeRelease(void *pVnode); void* vnodeGetWal(void *pVnode); +int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes); +void vnodeBuildStatusMsg(void *pStatus); +void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes); +// vnodeWrite int32_t vnodeWriteToWQueue(void *pVnode, void *pHead, int32_t qtype, void *pRpcMsg); void vnodeFreeFromWQueue(void *pVnode, SVWriteMsg *pWrite); int32_t vnodeProcessWrite(void *pVnode, void *pHead, int32_t qtype, void *pRspRet); -int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes); -void vnodeBuildStatusMsg(void *pStatus); -void vnodeConfirmForward(void *pVnode, uint64_t version, int32_t code); -void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes); -int32_t vnodeInitResources(); -void vnodeCleanupResources(); +// vnodeSync +void vnodeConfirmForward(void *pVnode, uint64_t version, int32_t code); +// vnodeRead int32_t vnodeWriteToRQueue(void *pVnode, void *pCont, int32_t contLen, int8_t qtype, void *rparam); void vnodeFreeFromRQueue(void *pVnode, SVReadMsg *pRead); int32_t vnodeProcessRead(void *pVnode, SVReadMsg *pRead); @@ -91,4 +88,4 @@ int32_t vnodeProcessRead(void *pVnode, SVReadMsg *pRead); } #endif -#endif +#endif \ No newline at end of file diff --git a/src/kit/shell/inc/shell.h b/src/kit/shell/inc/shell.h index d65c943e28d7d5a63aba4fc1839a4ba9cf744746..2c6e4a308ca04e825ea3b876a1b278ccb6feb279 100644 --- a/src/kit/shell/inc/shell.h +++ b/src/kit/shell/inc/shell.h @@ -51,7 +51,6 @@ typedef struct SShellArguments { char* commands; int abort; int port; - int endPort; int pktLen; char* netTestRole; } SShellArguments; diff --git a/src/kit/shell/src/shellDarwin.c b/src/kit/shell/src/shellDarwin.c index 995b56f341b36b84e413aec154ae1abcc832741d..ddf7b21bef354817f67ccda4106abf1bc648ad6b 100644 --- a/src/kit/shell/src/shellDarwin.c +++ b/src/kit/shell/src/shellDarwin.c @@ -407,7 +407,11 @@ void get_history_path(char *history) { sprintf(history, "%s/%s", getpwuid(getuid void clearScreen(int ecmd_pos, int cursor_pos) { struct winsize w; - ioctl(0, TIOCGWINSZ, &w); + if (ioctl(0, TIOCGWINSZ, &w) < 0 || w.ws_col == 0 || w.ws_row == 0) { + //fprintf(stderr, "No stream device, and use default value(col 120, row 30)\n"); + w.ws_col = 120; + w.ws_row = 30; + } int cursor_x = cursor_pos / w.ws_col; int cursor_y = cursor_pos % w.ws_col; @@ -425,8 +429,9 @@ void clearScreen(int ecmd_pos, int cursor_pos) { void showOnScreen(Command *cmd) { struct winsize w; if (ioctl(0, TIOCGWINSZ, &w) < 0 || w.ws_col == 0 || w.ws_row == 0) { - fprintf(stderr, "No stream device\n"); - exit(EXIT_FAILURE); + //fprintf(stderr, "No stream device\n"); + w.ws_col = 120; + w.ws_row = 30; } wchar_t wc; diff --git a/src/kit/shell/src/shellEngine.c b/src/kit/shell/src/shellEngine.c index 22f01ac142c72603b9fa0595ce8c1fc89c626d7e..af8beb7987b2cf5003a59e623c6e76710490992b 100644 --- a/src/kit/shell/src/shellEngine.c +++ b/src/kit/shell/src/shellEngine.c @@ -32,14 +32,14 @@ /**************** Global variables ****************/ #ifdef _TD_POWER_ char CLIENT_VERSION[] = "Welcome to the PowerDB shell from %s, Client Version:%s\n" - "Copyright (c) 2017 by PowerDB, Inc. All rights reserved.\n\n"; + "Copyright (c) 2020 by PowerDB, Inc. All rights reserved.\n\n"; char PROMPT_HEADER[] = "power> "; char CONTINUE_PROMPT[] = " -> "; int prompt_size = 7; #else char CLIENT_VERSION[] = "Welcome to the TDengine shell from %s, Client Version:%s\n" - "Copyright (c) 2017 by TAOS Data, Inc. All rights reserved.\n\n"; + "Copyright (c) 2020 by TAOS Data, Inc. All rights reserved.\n\n"; char PROMPT_HEADER[] = "taos> "; char CONTINUE_PROMPT[] = " -> "; diff --git a/src/kit/shell/src/shellLinux.c b/src/kit/shell/src/shellLinux.c index 04f5824d8da547557a4dc7fce599d721c1885611..f896253fb4ea4fedca07ce82d0c99ef162146df5 100644 --- a/src/kit/shell/src/shellLinux.c +++ b/src/kit/shell/src/shellLinux.c @@ -46,8 +46,7 @@ static struct argp_option options[] = { {"thread", 'T', "THREADNUM", 0, "Number of threads when using multi-thread to import data."}, {"database", 'd', "DATABASE", 0, "Database to use when connecting to the server."}, {"timezone", 't', "TIMEZONE", 0, "Time zone of the shell, default is local."}, - {"netrole", 'n', "NETROLE", 0, "Net role when network connectivity test, default is NULL, options: client|clients|server."}, - {"endport", 'e', "ENDPORT", 0, "Net test end port, default is 6042."}, + {"netrole", 'n', "NETROLE", 0, "Net role when network connectivity test, default is NULL, options: client|server|rpc|startup."}, {"pktlen", 'l', "PKTLEN", 0, "Packet length used for net test, default is 1000 bytes."}, {0}}; @@ -130,20 +129,9 @@ static error_t parse_opt(int key, char *arg, struct argp_state *state) { case 'd': arguments->database = arg; break; - case 'n': arguments->netTestRole = arg; break; - - case 'e': - if (arg) { - arguments->endPort = atoi(arg); - } else { - fprintf(stderr, "Invalid end port\n"); - return -1; - } - break; - case 'l': if (arg) { arguments->pktLen = atoi(arg); @@ -152,7 +140,6 @@ static error_t parse_opt(int key, char *arg, struct argp_state *state) { return -1; } break; - case OPT_ABORT: arguments->abort = 1; break; @@ -413,7 +400,11 @@ void get_history_path(char *history) { snprintf(history, TSDB_FILENAME_LEN, "%s/ void clearScreen(int ecmd_pos, int cursor_pos) { struct winsize w; - ioctl(0, TIOCGWINSZ, &w); + if (ioctl(0, TIOCGWINSZ, &w) < 0 || w.ws_col == 0 || w.ws_row == 0) { + //fprintf(stderr, "No stream device, and use default value(col 120, row 30)\n"); + w.ws_col = 120; + w.ws_row = 30; + } int cursor_x = cursor_pos / w.ws_col; int cursor_y = cursor_pos % w.ws_col; @@ -431,8 +422,9 @@ void clearScreen(int ecmd_pos, int cursor_pos) { void showOnScreen(Command *cmd) { struct winsize w; if (ioctl(0, TIOCGWINSZ, &w) < 0 || w.ws_col == 0 || w.ws_row == 0) { - fprintf(stderr, "No stream device\n"); - exit(EXIT_FAILURE); + //fprintf(stderr, "No stream device\n"); + w.ws_col = 120; + w.ws_row = 30; } wchar_t wc; diff --git a/src/kit/shell/src/shellMain.c b/src/kit/shell/src/shellMain.c index 2083ad3e9b7d5a101f35af4fe20dcd38957d5be2..a2ce78d36f7b624985592c4cfc60ce0fe741d39e 100644 --- a/src/kit/shell/src/shellMain.c +++ b/src/kit/shell/src/shellMain.c @@ -61,8 +61,7 @@ SShellArguments args = { .file = "\0", .dir = "\0", .threadNum = 5, - .commands = NULL, - .endPort = 6042, + .commands = NULL, .pktLen = 1000, .netTestRole = NULL }; @@ -81,9 +80,7 @@ int main(int argc, char* argv[]) { if (args.netTestRole && args.netTestRole[0] != 0) { taos_init(); - CmdArguments cmdArgs; - memcpy(&cmdArgs, &args, sizeof(SShellArguments)); - taosNetTest(&cmdArgs); + taosNetTest(args.netTestRole, args.host, args.port, args.pktLen); exit(0); } diff --git a/src/mnode/inc/mnodeAcct.h b/src/mnode/inc/mnodeAcct.h index 744a62f948dae7174eb782fa788c2276c99efd01..595dcca413f41eafa83e791394fd295e2b2ab525 100644 --- a/src/mnode/inc/mnodeAcct.h +++ b/src/mnode/inc/mnodeAcct.h @@ -27,6 +27,7 @@ void mnodeCleanupAccts(); void mnodeGetStatOfAllAcct(SAcctInfo* pAcctInfo); void * mnodeGetAcct(char *acctName); void * mnodeGetNextAcct(void *pIter, SAcctObj **pAcct); +void mnodeCancelGetNextAcct(void *pIter); void mnodeIncAcctRef(SAcctObj *pAcct); void mnodeDecAcctRef(SAcctObj *pAcct); void mnodeAddDbToAcct(SAcctObj *pAcct, SDbObj *pDb); diff --git a/src/mnode/inc/mnodeDb.h b/src/mnode/inc/mnodeDb.h index 7cbd08ed92bc42e165469d60e3fb25e724415981..9354b923d74de055d3905ebe6581d3f721630b08 100644 --- a/src/mnode/inc/mnodeDb.h +++ b/src/mnode/inc/mnodeDb.h @@ -34,6 +34,7 @@ int64_t mnodeGetDbNum(); SDbObj *mnodeGetDb(char *db); SDbObj *mnodeGetDbByTableId(char *db); void * mnodeGetNextDb(void *pIter, SDbObj **pDb); +void mnodeCancelGetNextDb(void *pIter); void mnodeIncDbRef(SDbObj *pDb); void mnodeDecDbRef(SDbObj *pDb); bool mnodeCheckIsMonitorDB(char *db, char *monitordb); diff --git a/src/mnode/inc/mnodeDnode.h b/src/mnode/inc/mnodeDnode.h index b6ddb7a9bf27252bb9dc97d8c6645245354a6428..8bc29ef9efabf73470896b49b86aad276a29a748 100644 --- a/src/mnode/inc/mnodeDnode.h +++ b/src/mnode/inc/mnodeDnode.h @@ -65,6 +65,7 @@ int32_t mnodeGetDnodesNum(); int32_t mnodeGetOnlinDnodesCpuCoreNum(); int32_t mnodeGetOnlineDnodesNum(); void * mnodeGetNextDnode(void *pIter, SDnodeObj **pDnode); +void mnodeCancelGetNextDnode(void *pIter); void mnodeIncDnodeRef(SDnodeObj *pDnode); void mnodeDecDnodeRef(SDnodeObj *pDnode); void * mnodeGetDnode(int32_t dnodeId); diff --git a/src/mnode/inc/mnodeInt.h b/src/mnode/inc/mnodeInt.h index 44626fd1672d5cf7e216e3fb256d434a4c27ca27..7a791d76e6796cfed22657f1bc7ffdb26890ea40 100644 --- a/src/mnode/inc/mnodeInt.h +++ b/src/mnode/inc/mnodeInt.h @@ -41,9 +41,9 @@ extern int32_t sdbDebugFlag; #define sdbDebug(...) { if (sdbDebugFlag & DEBUG_DEBUG) { taosPrintLog("SDB ", sdbDebugFlag, __VA_ARGS__); }} #define sdbTrace(...) { if (sdbDebugFlag & DEBUG_TRACE) { taosPrintLog("SDB ", sdbDebugFlag, __VA_ARGS__); }} -#define mLError(...) { monitorSaveLog(2, __VA_ARGS__); mError(__VA_ARGS__) } -#define mLWarn(...) { monitorSaveLog(1, __VA_ARGS__); mWarn(__VA_ARGS__) } -#define mLInfo(...) { monitorSaveLog(0, __VA_ARGS__); mInfo(__VA_ARGS__) } +#define mLError(...) { monSaveLog(2, __VA_ARGS__); mError(__VA_ARGS__) } +#define mLWarn(...) { monSaveLog(1, __VA_ARGS__); mWarn(__VA_ARGS__) } +#define mLInfo(...) { monSaveLog(0, __VA_ARGS__); mInfo(__VA_ARGS__) } #ifdef __cplusplus } diff --git a/src/mnode/inc/mnodeMnode.h b/src/mnode/inc/mnodeMnode.h index a28a03ea4028a243a27e79688de54627aaeb7064..93f2fa11ea95f22ca4addf12496b79cdaddbf290 100644 --- a/src/mnode/inc/mnodeMnode.h +++ b/src/mnode/inc/mnodeMnode.h @@ -38,6 +38,7 @@ void mnodeDropMnodeLocal(int32_t dnodeId); void * mnodeGetMnode(int32_t mnodeId); int32_t mnodeGetMnodesNum(); void * mnodeGetNextMnode(void *pIter, struct SMnodeObj **pMnode); +void mnodeCancelGetNextMnode(void *pIter); void mnodeIncMnodeRef(struct SMnodeObj *pMnode); void mnodeDecMnodeRef(struct SMnodeObj *pMnode); @@ -47,7 +48,7 @@ void mnodeGetMnodeEpSetForShell(SRpcEpSet *epSet); char* mnodeGetMnodeMasterEp(); void mnodeGetMnodeInfos(void *mnodes); -void mnodeUpdateMnodeEpSet(); +void mnodeUpdateMnodeEpSet(SMInfos *pMnodes); #ifdef __cplusplus } diff --git a/src/mnode/inc/mnodeSdb.h b/src/mnode/inc/mnodeSdb.h index 29d8cf1207c0cb82773780b5aa212bfa08791acd..31ea2da640ef3b20b84ca769a22637332d6ffe34 100644 --- a/src/mnode/inc/mnodeSdb.h +++ b/src/mnode/inc/mnodeSdb.h @@ -79,22 +79,26 @@ typedef struct { int32_t (*fpRestored)(); } SSdbTableDesc; +int32_t sdbInitRef(); +void sdbCleanUpRef(); int32_t sdbInit(); void sdbCleanUp(); -void * sdbOpenTable(SSdbTableDesc *desc); -void sdbCloseTable(void *handle); +int64_t sdbOpenTable(SSdbTableDesc *desc); +void sdbCloseTable(int64_t rid); +void* sdbGetTableByRid(int64_t rid); bool sdbIsMaster(); bool sdbIsServing(); void sdbUpdateMnodeRoles(); +int32_t sdbGetReplicaNum(); int32_t sdbInsertRow(SSdbRow *pRow); int32_t sdbDeleteRow(SSdbRow *pRow); int32_t sdbUpdateRow(SSdbRow *pRow); int32_t sdbInsertRowToQueue(SSdbRow *pRow); -void *sdbGetRow(void *pTable, void *key); -void *sdbFetchRow(void *pTable, void *pIter, void **ppRow); -void sdbFreeIter(void *pIter); +void * sdbGetRow(void *pTable, void *key); +void * sdbFetchRow(void *pTable, void *pIter, void **ppRow); +void sdbFreeIter(void *pTable, void *pIter); void sdbIncRef(void *pTable, void *pRow); void sdbDecRef(void *pTable, void *pRow); int64_t sdbGetNumOfRows(void *pTable); diff --git a/src/mnode/inc/mnodeShow.h b/src/mnode/inc/mnodeShow.h index da66e7167822d0077a8b5cad99489868eab7485d..f985fe792dec7b09dd91bc6c037ffed8da771fb7 100644 --- a/src/mnode/inc/mnodeShow.h +++ b/src/mnode/inc/mnodeShow.h @@ -26,8 +26,10 @@ void mnodeCleanUpShow(); typedef int32_t (*SShowMetaFp)(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); typedef int32_t (*SShowRetrieveFp)(SShowObj *pShow, char *data, int32_t rows, void *pConn); +typedef void (*SShowFreeIterFp)(void *pIter); void mnodeAddShowMetaHandle(uint8_t showType, SShowMetaFp fp); void mnodeAddShowRetrieveHandle(uint8_t showType, SShowRetrieveFp fp); +void mnodeAddShowFreeIterHandle(uint8_t msgType, SShowFreeIterFp fp); void mnodeVacuumResult(char *data, int32_t numOfCols, int32_t rows, int32_t capacity, SShowObj *pShow); #ifdef __cplusplus diff --git a/src/mnode/inc/mnodeTable.h b/src/mnode/inc/mnodeTable.h index 7c0077aa609843aa99e9f5114573aea8f7446fc1..bf04f26a90b6264bc226db0d64d4b808bb1f99b2 100644 --- a/src/mnode/inc/mnodeTable.h +++ b/src/mnode/inc/mnodeTable.h @@ -31,6 +31,8 @@ void mnodeIncTableRef(void *pTable); void mnodeDecTableRef(void *pTable); void * mnodeGetNextChildTable(void *pIter, SCTableObj **pTable); void * mnodeGetNextSuperTable(void *pIter, SSTableObj **pTable); +void mnodeCancelGetNextChildTable(void *pIter); +void mnodeCancelGetNextSuperTable(void *pIter); void mnodeDropAllChildTables(SDbObj *pDropDb); void mnodeDropAllSuperTables(SDbObj *pDropDb); void mnodeDropAllChildTablesInVgroups(SVgObj *pVgroup); diff --git a/src/mnode/inc/mnodeUser.h b/src/mnode/inc/mnodeUser.h index 073460f9d3621c7ec1f1eb57fd7da15791753b9e..156bc7aaebefb9accc59aa76979b7086c5cce9ae 100644 --- a/src/mnode/inc/mnodeUser.h +++ b/src/mnode/inc/mnodeUser.h @@ -25,6 +25,7 @@ int32_t mnodeInitUsers(); void mnodeCleanupUsers(); SUserObj *mnodeGetUser(char *name); void * mnodeGetNextUser(void *pIter, SUserObj **pUser); +void mnodeCancelGetNextUser(void *pIter); void mnodeIncUserRef(SUserObj *pUser); void mnodeDecUserRef(SUserObj *pUser); SUserObj *mnodeGetUserFromConn(void *pConn); diff --git a/src/mnode/inc/mnodeVgroup.h b/src/mnode/inc/mnodeVgroup.h index 0e6d9dfde463dc50810654921b2006fda212e496..ee9ec7ae9305c3d436b6e5bec0cd4e94ecc1adf4 100644 --- a/src/mnode/inc/mnodeVgroup.h +++ b/src/mnode/inc/mnodeVgroup.h @@ -34,6 +34,7 @@ void mnodeDropAllDnodeVgroups(SDnodeObj *pDropDnode); //void mnodeUpdateAllDbVgroups(SDbObj *pAlterDb); void * mnodeGetNextVgroup(void *pIter, SVgObj **pVgroup); +void mnodeCancelGetNextVgroup(void *pIter); void mnodeUpdateVgroup(SVgObj *pVgroup); void mnodeUpdateVgroupStatus(SVgObj *pVgroup, SDnodeObj *pDnode, SVnodeLoad *pVload); void mnodeCheckUnCreatedVgroup(SDnodeObj *pDnode, SVnodeLoad *pVloads, int32_t openVnodes); diff --git a/src/mnode/src/mnodeAcct.c b/src/mnode/src/mnodeAcct.c index 9fff2f02291d6b23163c0e58d3d2bfe545da6eb5..6fba05674fadbe967a673bbc0cc6e02ad7f2b41a 100644 --- a/src/mnode/src/mnodeAcct.c +++ b/src/mnode/src/mnodeAcct.c @@ -26,6 +26,7 @@ #include "mnodeUser.h" #include "mnodeVgroup.h" +int64_t tsAcctRid = -1; void * tsAcctSdb = NULL; static int32_t tsAcctUpdateSize; static int32_t mnodeCreateRootAcct(); @@ -114,7 +115,8 @@ int32_t mnodeInitAccts() { .fpRestored = mnodeAcctActionRestored }; - tsAcctSdb = sdbOpenTable(&desc); + tsAcctRid = sdbOpenTable(&desc); + tsAcctSdb = sdbGetTableByRid(tsAcctRid); if (tsAcctSdb == NULL) { mError("table:%s, failed to create hash", desc.name); return -1; @@ -126,7 +128,7 @@ int32_t mnodeInitAccts() { void mnodeCleanupAccts() { acctCleanUp(); - sdbCloseTable(tsAcctSdb); + sdbCloseTable(tsAcctRid); tsAcctSdb = NULL; } @@ -144,7 +146,6 @@ void mnodeGetStatOfAllAcct(SAcctInfo* pAcctInfo) { pAcctInfo->numOfTimeSeries += pAcct->acctInfo.numOfTimeSeries; mnodeDecAcctRef(pAcct); } - sdbFreeIter(pIter); SVgObj *pVgroup = NULL; pIter = NULL; @@ -158,7 +159,6 @@ void mnodeGetStatOfAllAcct(SAcctInfo* pAcctInfo) { pAcctInfo->totalPoints += pVgroup->pointsWritten; mnodeDecVgroupRef(pVgroup); } - sdbFreeIter(pIter); } void *mnodeGetAcct(char *name) { @@ -169,6 +169,10 @@ void *mnodeGetNextAcct(void *pIter, SAcctObj **pAcct) { return sdbFetchRow(tsAcctSdb, pIter, (void **)pAcct); } +void mnodeCancelGetNextAcct(void *pIter) { + sdbFreeIter(tsAcctSdb, pIter); +} + void mnodeIncAcctRef(SAcctObj *pAcct) { sdbIncRef(tsAcctSdb, pAcct); } diff --git a/src/mnode/src/mnodeCluster.c b/src/mnode/src/mnodeCluster.c index 5be67e4ad9b1b1746598331997e428f4e78607cf..56229daffa87d3300cc597d9353b5a720751a0f9 100644 --- a/src/mnode/src/mnodeCluster.c +++ b/src/mnode/src/mnodeCluster.c @@ -24,6 +24,7 @@ #include "mnodeShow.h" #include "tglobal.h" +int64_t tsClusterRid = -1; static void * tsClusterSdb = NULL; static int32_t tsClusterUpdateSize; static char tsClusterId[TSDB_CLUSTER_ID_LEN]; @@ -31,6 +32,7 @@ static int32_t mnodeCreateCluster(); static int32_t mnodeGetClusterMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); static int32_t mnodeRetrieveClusters(SShowObj *pShow, char *data, int32_t rows, void *pConn); +static void mnodeCancelGetNextCluster(void *pIter); static int32_t mnodeClusterActionDestroy(SSdbRow *pRow) { tfree(pRow->pObj); @@ -100,21 +102,23 @@ int32_t mnodeInitCluster() { .fpRestored = mnodeClusterActionRestored }; - tsClusterSdb = sdbOpenTable(&desc); + tsClusterRid = sdbOpenTable(&desc); + tsClusterSdb = sdbGetTableByRid(tsClusterRid); if (tsClusterSdb == NULL) { - mError("table:%s, failed to create hash", desc.name); + mError("table:%s, rid:%" PRId64 ", failed to create hash", desc.name, tsClusterRid); return -1; } mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_CLUSTER, mnodeGetClusterMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_CLUSTER, mnodeRetrieveClusters); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_CLUSTER, mnodeCancelGetNextCluster); mDebug("table:%s, hash is created", desc.name); return TSDB_CODE_SUCCESS; } void mnodeCleanupCluster() { - sdbCloseTable(tsClusterSdb); + sdbCloseTable(tsClusterRid); tsClusterSdb = NULL; } @@ -122,6 +126,10 @@ void *mnodeGetNextCluster(void *pIter, SClusterObj **pCluster) { return sdbFetchRow(tsClusterSdb, pIter, (void **)pCluster); } +void mnodeCancelGetNextCluster(void *pIter) { + sdbFreeIter(tsClusterSdb, pIter); +} + void mnodeIncClusterRef(SClusterObj *pCluster) { sdbIncRef(tsClusterSdb, pCluster); } @@ -167,7 +175,7 @@ void mnodeUpdateClusterId() { } mnodeDecClusterRef(pCluster); - sdbFreeIter(pIter); + mnodeCancelGetNextCluster(pIter); } static int32_t mnodeGetClusterMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn) { diff --git a/src/mnode/src/mnodeDb.c b/src/mnode/src/mnodeDb.c index d121208447c498280fb6f397e2181e2993ad1dbe..25dbb10536c5d0d784b5cbd5624dfca322321c5d 100644 --- a/src/mnode/src/mnodeDb.c +++ b/src/mnode/src/mnodeDb.c @@ -20,7 +20,7 @@ #include "tgrant.h" #include "tglobal.h" #include "tname.h" -#include "tbalance.h" +#include "tbn.h" #include "tdataformat.h" #include "mnode.h" #include "mnodeDef.h" @@ -38,6 +38,7 @@ #include "mnodeVgroup.h" #define VG_LIST_SIZE 8 +int64_t tsDbRid = -1; static void * tsDbSdb = NULL; static int32_t tsDbUpdateSize; @@ -160,7 +161,8 @@ int32_t mnodeInitDbs() { .fpRestored = mnodeDbActionRestored }; - tsDbSdb = sdbOpenTable(&desc); + tsDbRid = sdbOpenTable(&desc); + tsDbSdb = sdbGetTableByRid(tsDbRid); if (tsDbSdb == NULL) { mError("failed to init db data"); return -1; @@ -171,6 +173,7 @@ int32_t mnodeInitDbs() { mnodeAddWriteMsgHandle(TSDB_MSG_TYPE_CM_DROP_DB, mnodeProcessDropDbMsg); mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_DB, mnodeGetDbMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_DB, mnodeRetrieveDbs); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_DB, mnodeCancelGetNextDb); mDebug("table:dbs table is created"); return 0; @@ -180,6 +183,10 @@ void *mnodeGetNextDb(void *pIter, SDbObj **pDb) { return sdbFetchRow(tsDbSdb, pIter, (void **)pDb); } +void mnodeCancelGetNextDb(void *pIter) { + sdbFreeIter(tsDbSdb, pIter); +} + SDbObj *mnodeGetDb(char *db) { return (SDbObj *)sdbGetRow(tsDbSdb, db); } @@ -229,30 +236,28 @@ static int32_t mnodeCheckDbCfg(SDbCfg *pCfg) { if (pCfg->daysPerFile < TSDB_MIN_DAYS_PER_FILE || pCfg->daysPerFile > TSDB_MAX_DAYS_PER_FILE) { mError("invalid db option daysPerFile:%d valid range: [%d, %d]", pCfg->daysPerFile, TSDB_MIN_DAYS_PER_FILE, TSDB_MAX_DAYS_PER_FILE); - return TSDB_CODE_MND_INVALID_DB_OPTION; + return TSDB_CODE_MND_INVALID_DB_OPTION_DAYS; } if (pCfg->daysToKeep < TSDB_MIN_KEEP || pCfg->daysToKeep > TSDB_MAX_KEEP) { mError("invalid db option daysToKeep:%d valid range: [%d, %d]", pCfg->daysToKeep, TSDB_MIN_KEEP, TSDB_MAX_KEEP); - return TSDB_CODE_MND_INVALID_DB_OPTION; + return TSDB_CODE_MND_INVALID_DB_OPTION_KEEP; } if (pCfg->daysToKeep < pCfg->daysPerFile) { mError("invalid db option daysToKeep:%d should larger than daysPerFile:%d", pCfg->daysToKeep, pCfg->daysPerFile); - return TSDB_CODE_MND_INVALID_DB_OPTION; + return TSDB_CODE_MND_INVALID_DB_OPTION_KEEP; } -#if 0 if (pCfg->daysToKeep2 < TSDB_MIN_KEEP || pCfg->daysToKeep2 > pCfg->daysToKeep) { - mError("invalid db option daysToKeep2:%d valid range: [%d, %d]", pCfg->daysToKeep, TSDB_MIN_KEEP, pCfg->daysToKeep); - return TSDB_CODE_MND_INVALID_DB_OPTION; + mError("invalid db option daysToKeep2:%d valid range: [%d, %d]", pCfg->daysToKeep2, TSDB_MIN_KEEP, pCfg->daysToKeep); + return TSDB_CODE_MND_INVALID_DB_OPTION_KEEP; } if (pCfg->daysToKeep1 < TSDB_MIN_KEEP || pCfg->daysToKeep1 > pCfg->daysToKeep2) { mError("invalid db option daysToKeep1:%d valid range: [%d, %d]", pCfg->daysToKeep1, TSDB_MIN_KEEP, pCfg->daysToKeep2); - return TSDB_CODE_MND_INVALID_DB_OPTION; + return TSDB_CODE_MND_INVALID_DB_OPTION_KEEP; } -#endif if (pCfg->maxRowsPerFileBlock < TSDB_MIN_MAX_ROW_FBLOCK || pCfg->maxRowsPerFileBlock > TSDB_MAX_MAX_ROW_FBLOCK) { mError("invalid db option maxRowsPerFileBlock:%d valid range: [%d, %d]", pCfg->maxRowsPerFileBlock, @@ -312,13 +317,6 @@ static int32_t mnodeCheckDbCfg(SDbCfg *pCfg) { return TSDB_CODE_MND_INVALID_DB_OPTION; } -#ifndef _SYNC - if (pCfg->replications != 1) { - mError("invalid db option replications:%d can only be 1 in this version", pCfg->replications); - return TSDB_CODE_MND_INVALID_DB_OPTION; - } -#endif - if (pCfg->update < TSDB_MIN_DB_UPDATE || pCfg->update > TSDB_MAX_DB_UPDATE) { mError("invalid db option update:%d valid range: [%d, %d]", pCfg->update, TSDB_MIN_DB_UPDATE, TSDB_MAX_DB_UPDATE); return TSDB_CODE_MND_INVALID_DB_OPTION; @@ -491,7 +489,7 @@ void mnodeRemoveVgroupFromDb(SVgObj *pVgroup) { } void mnodeCleanupDbs() { - sdbCloseTable(tsDbSdb); + sdbCloseTable(tsDbRid); tsDbSdb = NULL; } @@ -986,8 +984,8 @@ static int32_t mnodeAlterDbCb(SMnodeMsg *pMsg, int32_t code) { SDbObj *pDb = pMsg->pDb; void *pIter = NULL; - while (1) { - SVgObj *pVgroup = NULL; + SVgObj *pVgroup = NULL; + while (1) { pIter = mnodeGetNextVgroup(pIter, &pVgroup); if (pVgroup == NULL) break; if (pVgroup->pDb == pDb) { @@ -995,12 +993,11 @@ static int32_t mnodeAlterDbCb(SMnodeMsg *pMsg, int32_t code) { } mnodeDecVgroupRef(pVgroup); } - sdbFreeIter(pIter); mDebug("db:%s, all vgroups is altered", pDb->name); mLInfo("db:%s, is alterd by %s", pDb->name, mnodeGetUserFromMsg(pMsg)); - balanceAsyncNotify(); + bnNotify(); return TSDB_CODE_SUCCESS; } @@ -1146,7 +1143,5 @@ void mnodeDropAllDbs(SAcctObj *pAcct) { mnodeDecDbRef(pDb); } - sdbFreeIter(pIter); - mInfo("acct:%s, all dbs:%d is dropped from sdb", pAcct->user, numOfDbs); } diff --git a/src/mnode/src/mnodeDnode.c b/src/mnode/src/mnodeDnode.c index f76533c76032907e627eea7c1d16bb99009e0ed3..f297dd51ddd9d265900d57e5b4693015143ede30 100644 --- a/src/mnode/src/mnodeDnode.c +++ b/src/mnode/src/mnodeDnode.c @@ -16,12 +16,12 @@ #define _DEFAULT_SOURCE #include "os.h" #include "tgrant.h" -#include "tbalance.h" +#include "tbn.h" #include "tglobal.h" #include "tconfig.h" #include "tutil.h" #include "tsocket.h" -#include "tbalance.h" +#include "tbn.h" #include "tsync.h" #include "tdataformat.h" #include "mnode.h" @@ -39,6 +39,7 @@ #include "mnodeCluster.h" int32_t tsAccessSquence = 0; +int64_t tsDnodeRid = -1; static void * tsDnodeSdb = NULL; static int32_t tsDnodeUpdateSize = 0; extern void * tsMnodeSdb; @@ -110,11 +111,8 @@ static int32_t mnodeDnodeActionInsert(SSdbRow *pRow) { static int32_t mnodeDnodeActionDelete(SSdbRow *pRow) { SDnodeObj *pDnode = pRow->pObj; -#ifndef _SYNC - mnodeDropAllDnodeVgroups(pDnode); -#endif mnodeDropMnodeLocal(pDnode->dnodeId); - balanceAsyncNotify(); + bnNotify(); mnodeUpdateDnodeEps(); mDebug("dnode:%d, all vgroups is dropped from sdb", pDnode->dnodeId); @@ -187,7 +185,8 @@ int32_t mnodeInitDnodes() { .fpRestored = mnodeDnodeActionRestored }; - tsDnodeSdb = sdbOpenTable(&desc); + tsDnodeRid = sdbOpenTable(&desc); + tsDnodeSdb = sdbGetTableByRid(tsDnodeRid); if (tsDnodeSdb == NULL) { mError("failed to init dnodes data"); return -1; @@ -206,13 +205,14 @@ int32_t mnodeInitDnodes() { mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_VNODES, mnodeRetrieveVnodes); mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_DNODE, mnodeGetDnodeMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_DNODE, mnodeRetrieveDnodes); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_DNODE, mnodeCancelGetNextDnode); mDebug("table:dnodes table is created"); return 0; } void mnodeCleanupDnodes() { - sdbCloseTable(tsDnodeSdb); + sdbCloseTable(tsDnodeRid); pthread_mutex_destroy(&tsDnodeEpsMutex); free(tsDnodeEps); tsDnodeEps = NULL; @@ -223,6 +223,10 @@ void *mnodeGetNextDnode(void *pIter, SDnodeObj **pDnode) { return sdbFetchRow(tsDnodeSdb, pIter, (void **)pDnode); } +void mnodeCancelGetNextDnode(void *pIter) { + sdbFreeIter(tsDnodeSdb, pIter); +} + int32_t mnodeGetDnodesNum() { return sdbGetNumOfRows(tsDnodeSdb); } @@ -241,8 +245,6 @@ int32_t mnodeGetOnlinDnodesCpuCoreNum() { mnodeDecDnodeRef(pDnode); } - sdbFreeIter(pIter); - if (cpuCores < 2) cpuCores = 2; return cpuCores; } @@ -259,8 +261,6 @@ int32_t mnodeGetOnlineDnodesNum() { mnodeDecDnodeRef(pDnode); } - sdbFreeIter(pIter); - return onlineDnodes; } @@ -276,13 +276,12 @@ void *mnodeGetDnodeByEp(char *ep) { pIter = mnodeGetNextDnode(pIter, &pDnode); if (pDnode == NULL) break; if (strcmp(ep, pDnode->dnodeEp) == 0) { - sdbFreeIter(pIter); + mnodeCancelGetNextDnode(pIter); return pDnode; } mnodeDecDnodeRef(pDnode); } - sdbFreeIter(pIter); return NULL; } @@ -345,7 +344,7 @@ static int32_t mnodeProcessCfgDnodeMsg(SMnodeMsg *pMsg) { return TSDB_CODE_MND_INVALID_DNODE_CFG_OPTION; } - int32_t code = balanceAlterDnode(pDnode, vnodeId, dnodeId); + int32_t code = bnAlterDnode(pDnode, vnodeId, dnodeId); mnodeDecDnodeRef(pDnode); return code; } else { @@ -464,7 +463,10 @@ static void mnodeUpdateDnodeEps() { while (1) { pIter = mnodeGetNextDnode(pIter, &pDnode); if (pDnode == NULL) break; - if (dnodesNum >= totalDnodes) break; + if (dnodesNum >= totalDnodes) { + mnodeCancelGetNextDnode(pIter); + break; + } SDnodeEp *pEp = &tsDnodeEps->dnodeEps[dnodesNum]; dnodesNum++; @@ -474,7 +476,6 @@ static void mnodeUpdateDnodeEps() { mnodeDecDnodeRef(pDnode); } - sdbFreeIter(pIter); pthread_mutex_unlock(&tsDnodeEpsMutex); } @@ -587,8 +588,8 @@ static int32_t mnodeProcessDnodeStatusMsg(SMnodeMsg *pMsg) { mInfo("dnode:%d, from offline to online", pDnode->dnodeId); pDnode->status = TAOS_DN_STATUS_READY; pDnode->offlineReason = TAOS_DN_OFF_ONLINE; - balanceSyncNotify(); - balanceAsyncNotify(); + bnCheckModules(); + bnNotify(); } if (openVnodes != pDnode->openVnodes) { @@ -701,11 +702,7 @@ static int32_t mnodeDropDnodeByEp(char *ep, SMnodeMsg *pMsg) { mInfo("dnode:%d, start to drop it", pDnode->dnodeId); -#ifndef _SYNC - int32_t code = mnodeDropDnode(pDnode, pMsg); -#else - int32_t code = balanceDropDnode(pDnode); -#endif + int32_t code = bnDropDnode(pDnode); mnodeDecDnodeRef(pDnode); return code; } @@ -1100,7 +1097,7 @@ static int32_t mnodeGetVnodeMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pC pDnode = mnodeGetDnodeByEp(pShow->payload); } else { void *pIter = mnodeGetNextDnode(NULL, (SDnodeObj **)&pDnode); - sdbFreeIter(pIter); + mnodeCancelGetNextDnode(pIter); } if (pDnode != NULL) { @@ -1148,7 +1145,6 @@ static int32_t mnodeRetrieveVnodes(SShowObj *pShow, char *data, int32_t rows, vo mnodeDecVgroupRef(pVgroup); } - sdbFreeIter(pIter); } else { numOfRows = 0; } @@ -1176,60 +1172,3 @@ static char* mnodeGetDnodeAlternativeRoleStr(int32_t alternativeRole) { default:return "any"; } } - -#ifndef _SYNC - -int32_t balanceInit() { return TSDB_CODE_SUCCESS; } -void balanceCleanUp() {} -void balanceAsyncNotify() {} -void balanceSyncNotify() {} -void balanceReset() {} -int32_t balanceAlterDnode(struct SDnodeObj *pDnode, int32_t vnodeId, int32_t dnodeId) { return TSDB_CODE_SYN_NOT_ENABLED; } - -char* syncRole[] = { - "offline", - "unsynced", - "syncing", - "slave", - "master" -}; - -int32_t balanceAllocVnodes(SVgObj *pVgroup) { - void * pIter = NULL; - SDnodeObj *pDnode = NULL; - SDnodeObj *pSelDnode = NULL; - float vnodeUsage = 1000.0; - - while (1) { - pIter = mnodeGetNextDnode(pIter, &pDnode); - if (pDnode == NULL) break; - - if (pDnode->numOfCores > 0 && pDnode->openVnodes < TSDB_MAX_VNODES) { - float openVnodes = pDnode->openVnodes; - if (pDnode->isMgmt) openVnodes += tsMnodeEqualVnodeNum; - - float usage = openVnodes / pDnode->numOfCores; - if (usage <= vnodeUsage) { - pSelDnode = pDnode; - vnodeUsage = usage; - } - } - mnodeDecDnodeRef(pDnode); - } - - sdbFreeIter(pIter); - - if (pSelDnode == NULL) { - mError("failed to alloc vnode to vgroup"); - return TSDB_CODE_MND_NO_ENOUGH_DNODES; - } - - pVgroup->vnodeGid[0].dnodeId = pSelDnode->dnodeId; - pVgroup->vnodeGid[0].pDnode = pSelDnode; - - mDebug("dnode:%d, alloc one vnode to vgroup, openVnodes:%d", pSelDnode->dnodeId, pSelDnode->openVnodes); - return TSDB_CODE_SUCCESS; -} - -#endif - diff --git a/src/mnode/src/mnodeMain.c b/src/mnode/src/mnodeMain.c index 1f5ad42bdead75064c60f8626ce4916e7b08fd9f..7b520c6022bd3b38450657d4a0fa68fc3cfcc581 100644 --- a/src/mnode/src/mnodeMain.c +++ b/src/mnode/src/mnodeMain.c @@ -17,7 +17,7 @@ #include "os.h" #include "taosdef.h" #include "tsched.h" -#include "tbalance.h" +#include "tbn.h" #include "tgrant.h" #include "ttimer.h" #include "tglobal.h" @@ -37,16 +37,11 @@ #include "mnodeShow.h" #include "mnodeProfile.h" -typedef struct { - const char *const name; - int (*init)(); - void (*cleanup)(); -} SMnodeComponent; - void *tsMnodeTmr = NULL; static bool tsMgmtIsRunning = false; -static const SMnodeComponent tsMnodeComponents[] = { +static SStep tsMnodeSteps[] = { + {"sdbref", sdbInitRef, sdbCleanUpRef}, {"profile", mnodeInitProfile, mnodeCleanupProfile}, {"cluster", mnodeInitCluster, mnodeCleanupCluster}, {"accts", mnodeInitAccts, mnodeCleanupAccts}, @@ -57,7 +52,7 @@ static const SMnodeComponent tsMnodeComponents[] = { {"tables", mnodeInitTables, mnodeCleanupTables}, {"mnodes", mnodeInitMnodes, mnodeCleanupMnodes}, {"sdb", sdbInit, sdbCleanUp}, - {"balance", balanceInit, balanceCleanUp}, + {"balance", bnInit, bnCleanUp}, {"grant", grantInit, grantCleanUp}, {"show", mnodeInitShow, mnodeCleanUpShow} }; @@ -66,22 +61,14 @@ static void mnodeInitTimer(); static void mnodeCleanupTimer(); static bool mnodeNeedStart() ; -static void mnodeCleanupComponents(int32_t stepId) { - for (int32_t i = stepId; i >= 0; i--) { - tsMnodeComponents[i].cleanup(); - } +static void mnodeCleanupComponents() { + int32_t stepSize = sizeof(tsMnodeSteps) / sizeof(SStep); + dnodeStepCleanup(tsMnodeSteps, stepSize); } static int32_t mnodeInitComponents() { - int32_t code = 0; - for (int32_t i = 0; i < sizeof(tsMnodeComponents) / sizeof(tsMnodeComponents[0]); i++) { - if (tsMnodeComponents[i].init() != 0) { - mnodeCleanupComponents(i); - code = -1; - break; - } - } - return code; + int32_t stepSize = sizeof(tsMnodeSteps) / sizeof(SStep); + return dnodeStepInit(tsMnodeSteps, stepSize); } int32_t mnodeStartSystem() { @@ -131,7 +118,7 @@ void mnodeCleanupSystem() { dnodeFreeMReadQueue(); dnodeFreeMPeerQueue(); mnodeCleanupTimer(); - mnodeCleanupComponents(sizeof(tsMnodeComponents) / sizeof(tsMnodeComponents[0]) - 1); + mnodeCleanupComponents(); mInfo("mnode is cleaned up"); } diff --git a/src/mnode/src/mnodeMnode.c b/src/mnode/src/mnodeMnode.c index 205bfda4b9571b488ccb335ab98c677f94166aae..68acae7dec0e5505e72733a638abcc142eeeb407 100644 --- a/src/mnode/src/mnodeMnode.c +++ b/src/mnode/src/mnodeMnode.c @@ -19,7 +19,7 @@ #include "tglobal.h" #include "trpc.h" #include "tsync.h" -#include "tbalance.h" +#include "tbn.h" #include "tutil.h" #include "tsocket.h" #include "tdataformat.h" @@ -34,13 +34,14 @@ #include "mnodeUser.h" #include "mnodeVgroup.h" -static void * tsMnodeSdb = NULL; -static int32_t tsMnodeUpdateSize = 0; -static SRpcEpSet tsMnodeEpSetForShell; -static SRpcEpSet tsMnodeEpSetForPeer; -static SMnodeInfos tsMnodeInfos; -static int32_t mnodeGetMnodeMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); -static int32_t mnodeRetrieveMnodes(SShowObj *pShow, char *data, int32_t rows, void *pConn); +int64_t tsMnodeRid = -1; +static void * tsMnodeSdb = NULL; +static int32_t tsMnodeUpdateSize = 0; +static SRpcEpSet tsMEpForShell; +static SRpcEpSet tsMEpForPeer; +static SMInfos tsMInfos; +static int32_t mnodeGetMnodeMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); +static int32_t mnodeRetrieveMnodes(SShowObj *pShow, char *data, int32_t rows, void *pConn); #if defined(LINUX) static pthread_rwlock_t tsMnodeLock; @@ -123,10 +124,10 @@ static int32_t mnodeMnodeActionRestored() { pMnode->role = TAOS_SYNC_ROLE_MASTER; mnodeDecMnodeRef(pMnode); } - sdbFreeIter(pIter); + mnodeCancelGetNextMnode(pIter); } - mnodeUpdateMnodeEpSet(); + mnodeUpdateMnodeEpSet(NULL); return TSDB_CODE_SUCCESS; } @@ -153,7 +154,8 @@ int32_t mnodeInitMnodes() { .fpRestored = mnodeMnodeActionRestored }; - tsMnodeSdb = sdbOpenTable(&desc); + tsMnodeRid = sdbOpenTable(&desc); + tsMnodeSdb = sdbGetTableByRid(tsMnodeRid); if (tsMnodeSdb == NULL) { mError("failed to init mnodes data"); return -1; @@ -161,13 +163,14 @@ int32_t mnodeInitMnodes() { mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_MNODE, mnodeGetMnodeMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_MNODE, mnodeRetrieveMnodes); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_MNODE, mnodeCancelGetNextMnode); mDebug("table:mnodes table is created"); return TSDB_CODE_SUCCESS; } void mnodeCleanupMnodes() { - sdbCloseTable(tsMnodeSdb); + sdbCloseTable(tsMnodeRid); tsMnodeSdb = NULL; mnodeMnodeDestroyLock(); } @@ -192,90 +195,134 @@ void *mnodeGetNextMnode(void *pIter, SMnodeObj **pMnode) { return sdbFetchRow(tsMnodeSdb, pIter, (void **)pMnode); } -void mnodeUpdateMnodeEpSet() { - mInfo("update mnodes epSet, numOfEps:%d ", mnodeGetMnodesNum()); - - mnodeMnodeWrLock(); - - memset(&tsMnodeEpSetForShell, 0, sizeof(SRpcEpSet)); - memset(&tsMnodeEpSetForPeer, 0, sizeof(SRpcEpSet)); - memset(&tsMnodeInfos, 0, sizeof(SMnodeInfos)); - - int32_t index = 0; - void * pIter = NULL; - while (1) { - SMnodeObj *pMnode = NULL; - pIter = mnodeGetNextMnode(pIter, &pMnode); - if (pMnode == NULL) break; - - SDnodeObj *pDnode = mnodeGetDnode(pMnode->mnodeId); - if (pDnode != NULL) { - strcpy(tsMnodeEpSetForShell.fqdn[index], pDnode->dnodeFqdn); - tsMnodeEpSetForShell.port[index] = htons(pDnode->dnodePort); - mDebug("mnode:%d, for shell fqdn:%s %d", pDnode->dnodeId, tsMnodeEpSetForShell.fqdn[index], htons(tsMnodeEpSetForShell.port[index])); - - strcpy(tsMnodeEpSetForPeer.fqdn[index], pDnode->dnodeFqdn); - tsMnodeEpSetForPeer.port[index] = htons(pDnode->dnodePort + TSDB_PORT_DNODEDNODE); - mDebug("mnode:%d, for peer fqdn:%s %d", pDnode->dnodeId, tsMnodeEpSetForPeer.fqdn[index], htons(tsMnodeEpSetForPeer.port[index])); +void mnodeCancelGetNextMnode(void *pIter) { + sdbFreeIter(tsMnodeSdb, pIter); +} - tsMnodeInfos.mnodeInfos[index].mnodeId = htonl(pMnode->mnodeId); - strcpy(tsMnodeInfos.mnodeInfos[index].mnodeEp, pDnode->dnodeEp); +void mnodeUpdateMnodeEpSet(SMInfos *pMinfos) { + bool set = false; + SMInfos mInfos = {0}; + mInfo("vgId:1, update mnodes epSet, numOfMnodes:%d pMinfos:%p", mnodeGetMnodesNum(), pMinfos); - if (pMnode->role == TAOS_SYNC_ROLE_MASTER) { - tsMnodeEpSetForShell.inUse = index; - tsMnodeEpSetForPeer.inUse = index; - tsMnodeInfos.inUse = index; + if (pMinfos != NULL) { + set = true; + mInfos = *pMinfos; + } + else { + int32_t index = 0; + void * pIter = NULL; + while (1) { + SMnodeObj *pMnode = NULL; + pIter = mnodeGetNextMnode(pIter, &pMnode); + if (pMnode == NULL) break; + + SDnodeObj *pDnode = mnodeGetDnode(pMnode->mnodeId); + if (pDnode != NULL) { + set = true; + mInfos.mnodeInfos[index].mnodeId = pMnode->mnodeId; + strcpy(mInfos.mnodeInfos[index].mnodeEp, pDnode->dnodeEp); + if (pMnode->role == TAOS_SYNC_ROLE_MASTER) mInfos.inUse = index; + index++; + } else { + set = false; } - mInfo("mnode:%d, ep:%s %s", pDnode->dnodeId, pDnode->dnodeEp, pMnode->role == TAOS_SYNC_ROLE_MASTER ? "master" : ""); - index++; + mnodeDecDnodeRef(pDnode); + mnodeDecMnodeRef(pMnode); } - mnodeDecDnodeRef(pDnode); - mnodeDecMnodeRef(pMnode); + mInfos.mnodeNum = index; + if (mInfos.mnodeNum < sdbGetReplicaNum()) { + set = false; + mDebug("vgId:1, mnodes info not synced, current:%d syncCfgNum:%d", mInfos.mnodeNum, sdbGetReplicaNum()); + } } - tsMnodeInfos.mnodeNum = index; - tsMnodeEpSetForShell.numOfEps = index; - tsMnodeEpSetForPeer.numOfEps = index; + mnodeMnodeWrLock(); - sdbFreeIter(pIter); + if (set) { + memset(&tsMEpForShell, 0, sizeof(SRpcEpSet)); + memset(&tsMEpForPeer, 0, sizeof(SRpcEpSet)); + memcpy(&tsMInfos, &mInfos, sizeof(SMInfos)); + tsMEpForShell.inUse = tsMInfos.inUse; + tsMEpForPeer.inUse = tsMInfos.inUse; + tsMEpForShell.numOfEps = tsMInfos.mnodeNum; + tsMEpForPeer.numOfEps = tsMInfos.mnodeNum; + + mInfo("vgId:1, mnodes epSet is set, num:%d inUse:%d", tsMInfos.mnodeNum, tsMInfos.inUse); + for (int index = 0; index < mInfos.mnodeNum; ++index) { + SMInfo *pInfo = &tsMInfos.mnodeInfos[index]; + taosGetFqdnPortFromEp(pInfo->mnodeEp, tsMEpForShell.fqdn[index], &tsMEpForShell.port[index]); + taosGetFqdnPortFromEp(pInfo->mnodeEp, tsMEpForPeer.fqdn[index], &tsMEpForPeer.port[index]); + tsMEpForPeer.port[index] = tsMEpForPeer.port[index] + TSDB_PORT_DNODEDNODE; + + mInfo("vgId:1, mnode:%d, fqdn:%s shell:%u peer:%u", pInfo->mnodeId, tsMEpForShell.fqdn[index], + tsMEpForShell.port[index], tsMEpForPeer.port[index]); + + tsMEpForShell.port[index] = htons(tsMEpForShell.port[index]); + tsMEpForPeer.port[index] = htons(tsMEpForPeer.port[index]); + pInfo->mnodeId = htonl(pInfo->mnodeId); + } + } else { + mInfo("vgId:1, mnodes epSet not set, num:%d inUse:%d", tsMInfos.mnodeNum, tsMInfos.inUse); + for (int index = 0; index < tsMInfos.mnodeNum; ++index) { + mInfo("vgId:1, index:%d, ep:%s:%u", index, tsMEpForShell.fqdn[index], htons(tsMEpForShell.port[index])); + } + } mnodeMnodeUnLock(); } void mnodeGetMnodeEpSetForPeer(SRpcEpSet *epSet) { mnodeMnodeRdLock(); - *epSet = tsMnodeEpSetForPeer; + *epSet = tsMEpForPeer; mnodeMnodeUnLock(); + + mTrace("vgId:1, mnodes epSet for peer is returned, num:%d inUse:%d", tsMEpForPeer.numOfEps, tsMEpForPeer.inUse); + for (int32_t i = 0; i < epSet->numOfEps; ++i) { + if (strcmp(epSet->fqdn[i], tsLocalFqdn) == 0 && htons(epSet->port[i]) == tsServerPort + TSDB_PORT_DNODEDNODE) { + epSet->inUse = (i + 1) % epSet->numOfEps; + mTrace("vgId:1, mnode:%d, for peer ep:%s:%u, set inUse to %d", i, epSet->fqdn[i], htons(epSet->port[i]), epSet->inUse); + } else { + mTrace("vgId:1, mpeer:%d, for peer ep:%s:%u", i, epSet->fqdn[i], htons(epSet->port[i])); + } + } } void mnodeGetMnodeEpSetForShell(SRpcEpSet *epSet) { mnodeMnodeRdLock(); - *epSet = tsMnodeEpSetForShell; + *epSet = tsMEpForShell; mnodeMnodeUnLock(); + + mTrace("vgId:1, mnodes epSet for shell is returned, num:%d inUse:%d", tsMEpForShell.numOfEps, tsMEpForShell.inUse); + for (int32_t i = 0; i < epSet->numOfEps; ++i) { + if (strcmp(epSet->fqdn[i], tsLocalFqdn) == 0 && htons(epSet->port[i]) == tsServerPort) { + epSet->inUse = (i + 1) % epSet->numOfEps; + mTrace("vgId:1, mnode:%d, for shell ep:%s:%u, set inUse to %d", i, epSet->fqdn[i], htons(epSet->port[i]), epSet->inUse); + } else { + mTrace("vgId:1, mnode:%d, for shell ep:%s:%u", i, epSet->fqdn[i], htons(epSet->port[i])); + } + } } char* mnodeGetMnodeMasterEp() { - return tsMnodeInfos.mnodeInfos[tsMnodeInfos.inUse].mnodeEp; + return tsMInfos.mnodeInfos[tsMInfos.inUse].mnodeEp; } -void mnodeGetMnodeInfos(void *mnodeInfos) { +void mnodeGetMnodeInfos(void *pMinfos) { mnodeMnodeRdLock(); - *(SMnodeInfos *)mnodeInfos = tsMnodeInfos; + *(SMInfos *)pMinfos = tsMInfos; mnodeMnodeUnLock(); } static int32_t mnodeSendCreateMnodeMsg(int32_t dnodeId, char *dnodeEp) { - mDebug("dnode:%d, send create mnode msg to dnode %s", dnodeId, dnodeEp); - SCreateMnodeMsg *pCreate = rpcMallocCont(sizeof(SCreateMnodeMsg)); if (pCreate == NULL) { return TSDB_CODE_MND_OUT_OF_MEMORY; } else { pCreate->dnodeId = htonl(dnodeId); tstrncpy(pCreate->dnodeEp, dnodeEp, sizeof(pCreate->dnodeEp)); - pCreate->mnodes = tsMnodeInfos; + mnodeGetMnodeInfos(&pCreate->mnodes); bool found = false; for (int i = 0; i < pCreate->mnodes.mnodeNum; ++i) { if (pCreate->mnodes.mnodeInfos[i].mnodeId == htonl(dnodeId)) { @@ -289,6 +336,11 @@ static int32_t mnodeSendCreateMnodeMsg(int32_t dnodeId, char *dnodeEp) { } } + mDebug("dnode:%d, send create mnode msg to dnode %s, numOfMnodes:%d", dnodeId, dnodeEp, pCreate->mnodes.mnodeNum); + for (int32_t i = 0; i < pCreate->mnodes.mnodeNum; ++i) { + mDebug("index:%d, mnodeId:%d ep:%s", i, pCreate->mnodes.mnodeInfos[i].mnodeId, pCreate->mnodes.mnodeInfos[i].mnodeEp); + } + SRpcMsg rpcMsg = {0}; rpcMsg.pCont = pCreate; rpcMsg.contLen = sizeof(SCreateMnodeMsg); @@ -313,7 +365,7 @@ static int32_t mnodeCreateMnodeCb(SMnodeMsg *pMsg, int32_t code) { mError("failed to create mnode, reason:%s", tstrerror(code)); } else { mDebug("mnode is created successfully"); - mnodeUpdateMnodeEpSet(); + mnodeUpdateMnodeEpSet(NULL); sdbUpdateAsync(); } @@ -357,7 +409,7 @@ void mnodeDropMnodeLocal(int32_t dnodeId) { mnodeDecMnodeRef(pMnode); } - mnodeUpdateMnodeEpSet(); + mnodeUpdateMnodeEpSet(NULL); sdbUpdateAsync(); } @@ -377,7 +429,7 @@ int32_t mnodeDropMnode(int32_t dnodeId) { sdbDecRef(tsMnodeSdb, pMnode); - mnodeUpdateMnodeEpSet(); + mnodeUpdateMnodeEpSet(NULL); sdbUpdateAsync(); return code; diff --git a/src/mnode/src/mnodePeer.c b/src/mnode/src/mnodePeer.c index f13ef7539800919e5f31d43b55def590b801e86d..cfb7b7781b04c0145f66a58c6d3d2eaefa0913f3 100644 --- a/src/mnode/src/mnodePeer.c +++ b/src/mnode/src/mnodePeer.c @@ -20,7 +20,7 @@ #include "tsystem.h" #include "tutil.h" #include "tgrant.h" -#include "tbalance.h" +#include "tbn.h" #include "tglobal.h" #include "mnode.h" #include "dnode.h" @@ -58,16 +58,8 @@ int32_t mnodeProcessPeerReq(SMnodeMsg *pMsg) { rpcRsp->rsp = epSet; rpcRsp->len = sizeof(SRpcEpSet); - mDebug("msg:%p, ahandle:%p type:%s in mpeer queue will be redirected, numOfEps:%d inUse:%d", pMsg, - pMsg->rpcMsg.ahandle, taosMsg[pMsg->rpcMsg.msgType], epSet->numOfEps, epSet->inUse); - for (int32_t i = 0; i < epSet->numOfEps; ++i) { - if (strcmp(epSet->fqdn[i], tsLocalFqdn) == 0 && htons(epSet->port[i]) == tsServerPort + TSDB_PORT_DNODEDNODE) { - epSet->inUse = (i + 1) % epSet->numOfEps; - mDebug("mpeer:%d ep:%s:%u, set inUse to %d", i, epSet->fqdn[i], htons(epSet->port[i]), epSet->inUse); - } else { - mDebug("mpeer:%d ep:%s:%u", i, epSet->fqdn[i], htons(epSet->port[i])); - } - } + mDebug("msg:%p, ahandle:%p type:%s in mpeer queue is redirected, numOfEps:%d inUse:%d", pMsg, pMsg->rpcMsg.ahandle, + taosMsg[pMsg->rpcMsg.msgType], epSet->numOfEps, epSet->inUse); return TSDB_CODE_RPC_REDIRECT; } diff --git a/src/mnode/src/mnodeProfile.c b/src/mnode/src/mnodeProfile.c index 12ac64854c6834ca0cd1509f9bfedd7b65a0900c..36b6ff7a59fc1f1c73e64c40ff3a5a1cef65ece6 100644 --- a/src/mnode/src/mnodeProfile.c +++ b/src/mnode/src/mnodeProfile.c @@ -34,7 +34,6 @@ #define QUERY_ID_SIZE 20 #define QUERY_STREAM_SAVE_SIZE 20 -extern void *tsMnodeTmr; static SCacheObj *tsMnodeConnCache = NULL; static int32_t tsConnIndex = 0; @@ -42,6 +41,7 @@ static int32_t mnodeGetQueryMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pC static int32_t mnodeRetrieveQueries(SShowObj *pShow, char *data, int32_t rows, void *pConn); static int32_t mnodeGetConnsMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); static int32_t mnodeRetrieveConns(SShowObj *pShow, char *data, int32_t rows, void *pConn); +static void mnodeCancelGetNextConn(void *pIter); static int32_t mnodeGetStreamMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); static int32_t mnodeRetrieveStreams(SShowObj *pShow, char *data, int32_t rows, void *pConn); static void mnodeFreeConn(void *data); @@ -52,10 +52,13 @@ static int32_t mnodeProcessKillConnectionMsg(SMnodeMsg *pMsg); int32_t mnodeInitProfile() { mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_QUERIES, mnodeGetQueryMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_QUERIES, mnodeRetrieveQueries); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_QUERIES, mnodeCancelGetNextConn); mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_CONNS, mnodeGetConnsMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_CONNS, mnodeRetrieveConns); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_CONNS, mnodeCancelGetNextConn); mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_STREAMS, mnodeGetStreamMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_STREAMS, mnodeRetrieveStreams); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_STREAMS, mnodeCancelGetNextConn); mnodeAddWriteMsgHandle(TSDB_MSG_TYPE_CM_KILL_QUERY, mnodeProcessKillQueryMsg); mnodeAddWriteMsgHandle(TSDB_MSG_TYPE_CM_KILL_STREAM, mnodeProcessKillStreamMsg); @@ -137,21 +140,15 @@ static void mnodeFreeConn(void *data) { mDebug("connId:%d, is destroyed", pConn->connId); } -static void *mnodeGetNextConn(SHashMutableIterator *pIter, SConnObj **pConn) { +static void *mnodeGetNextConn(void *pIter, SConnObj **pConn) { *pConn = NULL; - if (pIter == NULL) { - pIter = taosHashCreateIter(tsMnodeConnCache->pHashTable); - } - - if (!taosHashIterNext(pIter)) { - taosHashDestroyIter(pIter); - return NULL; - } + pIter = taosHashIterate(tsMnodeConnCache->pHashTable, pIter); + if (pIter == NULL) return NULL; - SCacheDataNode **pNode = taosHashIterGet(pIter); + SCacheDataNode **pNode = pIter; if (pNode == NULL || *pNode == NULL) { - taosHashDestroyIter(pIter); + taosHashCancelIterate(tsMnodeConnCache->pHashTable, pIter); return NULL; } @@ -159,6 +156,10 @@ static void *mnodeGetNextConn(SHashMutableIterator *pIter, SConnObj **pConn) { return pIter; } +static void mnodeCancelGetNextConn(void *pIter) { + taosHashCancelIterate(tsMnodeConnCache->pHashTable, pIter); +} + static int32_t mnodeGetConnsMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn) { SUserObj *pUser = mnodeGetUserFromConn(pConn); if (pUser == NULL) return 0; diff --git a/src/mnode/src/mnodeRead.c b/src/mnode/src/mnodeRead.c index ea7ce783e8f0a018d3648257206cb6abcaf28e26..c2a70bc01d2194fff5b8095d06e40535b940dc25 100644 --- a/src/mnode/src/mnodeRead.c +++ b/src/mnode/src/mnodeRead.c @@ -17,7 +17,7 @@ #include "os.h" #include "taosdef.h" #include "tsched.h" -#include "tbalance.h" +#include "tbn.h" #include "tgrant.h" #include "ttimer.h" #include "tglobal.h" @@ -51,21 +51,12 @@ int32_t mnodeProcessRead(SMnodeMsg *pMsg) { SMnodeRsp *rpcRsp = &pMsg->rpcRsp; SRpcEpSet *epSet = rpcMallocCont(sizeof(SRpcEpSet)); mnodeGetMnodeEpSetForShell(epSet); - - mDebug("msg:%p, app:%p type:%s in mread queue will be redirected, numOfEps:%d inUse:%d", pMsg, pMsg->rpcMsg.ahandle, - taosMsg[pMsg->rpcMsg.msgType], epSet->numOfEps, epSet->inUse); - for (int32_t i = 0; i < epSet->numOfEps; ++i) { - if (strcmp(epSet->fqdn[i], tsLocalFqdn) == 0 && htons(epSet->port[i]) == tsServerPort) { - epSet->inUse = (i + 1) % epSet->numOfEps; - mDebug("mnode index:%d ep:%s:%u, set inUse to %d", i, epSet->fqdn[i], htons(epSet->port[i]), epSet->inUse); - } else { - mDebug("mnode index:%d ep:%s:%u", i, epSet->fqdn[i], htons(epSet->port[i])); - } - } - rpcRsp->rsp = epSet; rpcRsp->len = sizeof(SRpcEpSet); + mDebug("msg:%p, app:%p type:%s in mread queue is redirected, numOfEps:%d inUse:%d", pMsg, pMsg->rpcMsg.ahandle, + taosMsg[pMsg->rpcMsg.msgType], epSet->numOfEps, epSet->inUse); + return TSDB_CODE_RPC_REDIRECT; } diff --git a/src/mnode/src/mnodeSdb.c b/src/mnode/src/mnodeSdb.c index e5fda26687224066b10ee4fe24ed05a34a5f1648..6cc4e097350783a93300c46adbd5f68faf43e143 100644 --- a/src/mnode/src/mnodeSdb.c +++ b/src/mnode/src/mnodeSdb.c @@ -18,7 +18,8 @@ #include "taoserror.h" #include "hash.h" #include "tutil.h" -#include "tbalance.h" +#include "tref.h" +#include "tbn.h" #include "tqueue.h" #include "twal.h" #include "tsync.h" @@ -98,6 +99,7 @@ typedef struct { SSdbWorker *worker; } SSdbWorkerPool; +int32_t tsSdbRid; extern void * tsMnodeTmr; static void * tsSdbTmr; static SSdbMgmt tsSdbMgmt = {0}; @@ -118,6 +120,7 @@ static void sdbFreeQueue(); static int32_t sdbInsertHash(SSdbTable *pTable, SSdbRow *pRow); static int32_t sdbUpdateHash(SSdbTable *pTable, SSdbRow *pRow); static int32_t sdbDeleteHash(SSdbTable *pTable, SSdbRow *pRow); +static void sdbCloseTableObj(void *handle); int32_t sdbGetId(void *pTable) { return ((SSdbTable *)pTable)->autoIndex; @@ -221,11 +224,13 @@ void sdbUpdateMnodeRoles() { sdbInfo("vgId:1, mnode:%d, role:%s", pMnode->mnodeId, syncRole[pMnode->role]); if (pMnode->mnodeId == dnodeGetDnodeId()) tsSdbMgmt.role = pMnode->role; mnodeDecMnodeRef(pMnode); + } else { + sdbDebug("vgId:1, mnode:%d not found", roles.nodeId[i]); } } mnodeUpdateClusterId(); - mnodeUpdateMnodeEpSet(); + mnodeUpdateMnodeEpSet(NULL); } static uint32_t sdbGetFileInfo(int32_t vgId, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fversion) { @@ -241,13 +246,23 @@ static void sdbNotifyRole(int32_t vgId, int8_t role) { sdbInfo("vgId:1, mnode role changed from %s to %s", syncRole[tsSdbMgmt.role], syncRole[role]); if (role == TAOS_SYNC_ROLE_MASTER && tsSdbMgmt.role != TAOS_SYNC_ROLE_MASTER) { - balanceReset(); + bnReset(); } tsSdbMgmt.role = role; sdbUpdateMnodeRoles(); } +static int32_t sdbNotifyFileSynced(int32_t vgId, uint64_t fversion) { return 0; } + +static void sdbNotifyFlowCtrl(int32_t vgId, int32_t level) {} + +static int32_t sdbGetSyncVersion(int32_t vgId, uint64_t *fver, uint64_t *vver) { + *fver = 0; + *vver = 0; + return 0; +} + // failed to forward, need revert insert static void sdbHandleFailedConfirm(SSdbRow *pRow) { SWalHead *pHead = pRow->pHead; @@ -295,18 +310,20 @@ void sdbUpdateAsync() { } void sdbUpdateSync(void *pMnodes) { - SMnodeInfos *mnodes = pMnodes; + SMInfos *pMinfos = pMnodes; if (!mnodeIsRunning()) { mDebug("vgId:1, mnode not start yet, update sync config later"); return; } - mDebug("vgId:1, update sync config in sync module, mnodes:%p", pMnodes); + mDebug("vgId:1, update sync config, pMnodes:%p", pMnodes); SSyncCfg syncCfg = {0}; int32_t index = 0; - if (mnodes == NULL) { + if (pMinfos == NULL) { + mDebug("vgId:1, mInfos not input, use mInfos in sdb, numOfMnodes:%d", syncCfg.replica); + void *pIter = NULL; while (1) { SMnodeObj *pMnode = NULL; @@ -325,18 +342,18 @@ void sdbUpdateSync(void *pMnodes) { mnodeDecDnodeRef(pDnode); mnodeDecMnodeRef(pMnode); } - sdbFreeIter(pIter); syncCfg.replica = index; - mDebug("vgId:1, mnodes info not input, use infos in sdb, numOfMnodes:%d", syncCfg.replica); } else { - for (index = 0; index < mnodes->mnodeNum; ++index) { - SMnodeInfo *node = &mnodes->mnodeInfos[index]; + mDebug("vgId:1, mInfos input, numOfMnodes:%d", pMinfos->mnodeNum); + + for (index = 0; index < pMinfos->mnodeNum; ++index) { + SMInfo *node = &pMinfos->mnodeInfos[index]; syncCfg.nodeInfo[index].nodeId = node->mnodeId; taosGetFqdnPortFromEp(node->mnodeEp, syncCfg.nodeInfo[index].nodeFqdn, &syncCfg.nodeInfo[index].nodePort); syncCfg.nodeInfo[index].nodePort += TSDB_PORT_SYNC; } syncCfg.replica = index; - mDebug("vgId:1, mnodes info input, numOfMnodes:%d", syncCfg.replica); + mnodeUpdateMnodeEpSet(pMnodes); } syncCfg.quorum = (syncCfg.replica == 1) ? 1 : 2; @@ -370,11 +387,14 @@ void sdbUpdateSync(void *pMnodes) { syncInfo.version = sdbGetVersion(); syncInfo.syncCfg = syncCfg; sprintf(syncInfo.path, "%s", tsMnodeDir); - syncInfo.getWalInfo = sdbGetWalInfo; syncInfo.getFileInfo = sdbGetFileInfo; + syncInfo.getWalInfo = sdbGetWalInfo; syncInfo.writeToCache = sdbWriteFwdToQueue; syncInfo.confirmForward = sdbConfirmForward; syncInfo.notifyRole = sdbNotifyRole; + syncInfo.notifyFileSynced = sdbNotifyFileSynced; + syncInfo.notifyFlowCtrl = sdbNotifyFlowCtrl; + syncInfo.getVersion = sdbGetSyncVersion; tsSdbMgmt.cfg = syncCfg; if (tsSdbMgmt.sync) { @@ -386,6 +406,17 @@ void sdbUpdateSync(void *pMnodes) { sdbUpdateMnodeRoles(); } +int32_t sdbInitRef() { + tsSdbRid = taosOpenRef(10, sdbCloseTableObj); + if (tsSdbRid <= 0) { + sdbError("failed to init sdb ref"); + return -1; + } + return 0; +} + +void sdbCleanUpRef() { taosCloseRef(tsSdbRid); } + int32_t sdbInit() { pthread_mutex_init(&tsSdbMgmt.mutex, NULL); @@ -424,7 +455,7 @@ void sdbCleanUp() { walClose(tsSdbMgmt.wal); tsSdbMgmt.wal = NULL; } - + pthread_mutex_destroy(&tsSdbMgmt.mutex); } @@ -507,7 +538,7 @@ static int32_t sdbInsertHash(SSdbTable *pTable, SSdbRow *pRow) { atomic_add_fetch_32(&pTable->autoIndex, 1); } - sdbDebug("vgId:1, sdb:%s, insert key:%s to hash, rowSize:%d rows:%" PRId64 ", msg:%p", pTable->name, + sdbTrace("vgId:1, sdb:%s, insert key:%s to hash, rowSize:%d rows:%" PRId64 ", msg:%p", pTable->name, sdbGetRowStr(pTable, pRow->pObj), pRow->rowSize, pTable->numOfRows, pRow->pMsg); int32_t code = (*pTable->fpInsert)(pRow); @@ -543,7 +574,7 @@ static int32_t sdbDeleteHash(SSdbTable *pTable, SSdbRow *pRow) { atomic_sub_fetch_32(&pTable->numOfRows, 1); - sdbDebug("vgId:1, sdb:%s, delete key:%s from hash, numOfRows:%" PRId64 ", msg:%p", pTable->name, + sdbTrace("vgId:1, sdb:%s, delete key:%s from hash, numOfRows:%" PRId64 ", msg:%p", pTable->name, sdbGetRowStr(pTable, pRow->pObj), pTable->numOfRows, pRow->pMsg); sdbDecRef(pTable, pRow->pObj); @@ -552,7 +583,7 @@ static int32_t sdbDeleteHash(SSdbTable *pTable, SSdbRow *pRow) { } static int32_t sdbUpdateHash(SSdbTable *pTable, SSdbRow *pRow) { - sdbDebug("vgId:1, sdb:%s, update key:%s in hash, numOfRows:%" PRId64 ", msg:%p", pTable->name, + sdbTrace("vgId:1, sdb:%s, update key:%s in hash, numOfRows:%" PRId64 ", msg:%p", pTable->name, sdbGetRowStr(pTable, pRow->pObj), pTable->numOfRows, pRow->pMsg); (*pTable->fpUpdate)(pRow); @@ -650,7 +681,7 @@ static int32_t sdbProcessWrite(void *wparam, void *hparam, int32_t qtype, void * return syncCode; } - sdbDebug("vgId:1, sdb:%s, record from wal/fwd is disposed, action:%s key:%s hver:%" PRIu64, pTable->name, + sdbTrace("vgId:1, sdb:%s, record from %s is disposed, action:%s key:%s hver:%" PRIu64, pTable->name, qtypeStr[qtype], actStr[action], sdbGetKeyStr(pTable, pHead->cont), pHead->version); // even it is WAL/FWD, it shall be called to update version in sync @@ -775,24 +806,17 @@ int32_t sdbUpdateRow(SSdbRow *pRow) { } } -void *sdbFetchRow(void *tparam, void *pNode, void **ppRow) { +void *sdbFetchRow(void *tparam, void *pIter, void **ppRow) { SSdbTable *pTable = tparam; *ppRow = NULL; if (pTable == NULL) return NULL; - SHashMutableIterator *pIter = pNode; - if (pIter == NULL) { - pIter = taosHashCreateIter(pTable->iHandle); - } - - if (!taosHashIterNext(pIter)) { - taosHashDestroyIter(pIter); - return NULL; - } + pIter = taosHashIterate(pTable->iHandle, pIter); + if (pIter == NULL) return NULL; - void **ppMetaRow = taosHashIterGet(pIter); + void **ppMetaRow = pIter; if (ppMetaRow == NULL) { - taosHashDestroyIter(pIter); + taosHashCancelIterate(pTable->iHandle, pIter); return NULL; } @@ -802,16 +826,17 @@ void *sdbFetchRow(void *tparam, void *pNode, void **ppRow) { return pIter; } -void sdbFreeIter(void *pIter) { - if (pIter != NULL) { - taosHashDestroyIter(pIter); - } +void sdbFreeIter(void *tparam, void *pIter) { + SSdbTable *pTable = tparam; + if (pTable == NULL || pIter == NULL) return; + + taosHashCancelIterate(pTable->iHandle, pIter); } -void *sdbOpenTable(SSdbTableDesc *pDesc) { +int64_t sdbOpenTable(SSdbTableDesc *pDesc) { SSdbTable *pTable = (SSdbTable *)calloc(1, sizeof(SSdbTable)); - if (pTable == NULL) return NULL; + if (pTable == NULL) return -1; pthread_mutex_init(&pTable->mutex, NULL); tstrncpy(pTable->name, pDesc->name, SDB_TABLE_LEN); @@ -836,19 +861,31 @@ void *sdbOpenTable(SSdbTableDesc *pDesc) { tsSdbMgmt.numOfTables++; tsSdbMgmt.tableList[pTable->id] = pTable; - return pTable; + + return taosAddRef(tsSdbRid, pTable); +} + +void sdbCloseTable(int64_t rid) { + taosRemoveRef(tsSdbRid, rid); +} + +void *sdbGetTableByRid(int64_t rid) { + void *handle = taosAcquireRef(tsSdbRid, rid); + taosReleaseRef(tsSdbRid, rid); + return handle; } -void sdbCloseTable(void *handle) { +static void sdbCloseTableObj(void *handle) { SSdbTable *pTable = (SSdbTable *)handle; if (pTable == NULL) return; tsSdbMgmt.numOfTables--; tsSdbMgmt.tableList[pTable->id] = NULL; - SHashMutableIterator *pIter = taosHashCreateIter(pTable->iHandle); - while (taosHashIterNext(pIter)) { - void **ppRow = taosHashIterGet(pIter); + void *pIter = taosHashIterate(pTable->iHandle, NULL); + while (pIter) { + void **ppRow = pIter; + pIter = taosHashIterate(pTable->iHandle, pIter); if (ppRow == NULL) continue; SSdbRow row = { @@ -859,8 +896,9 @@ void sdbCloseTable(void *handle) { (*pTable->fpDestroy)(&row); } - taosHashDestroyIter(pIter); + taosHashCancelIterate(pTable->iHandle, pIter); taosHashCleanup(pTable->iHandle); + pTable->iHandle = NULL; pthread_mutex_destroy(&pTable->mutex); sdbDebug("vgId:1, sdb:%s, is closed, numOfTables:%d", pTable->name, tsSdbMgmt.numOfTables); @@ -1070,3 +1108,7 @@ static void *sdbWorkerFp(void *pWorker) { return NULL; } + +int32_t sdbGetReplicaNum() { + return tsSdbMgmt.cfg.replica; +} \ No newline at end of file diff --git a/src/mnode/src/mnodeShow.c b/src/mnode/src/mnodeShow.c index 0d53fa9617b500d61457889e2228e5e756a244ae..2da46d5b4bd01d47794b0e77011101ce7dc137cc 100644 --- a/src/mnode/src/mnodeShow.c +++ b/src/mnode/src/mnodeShow.c @@ -52,11 +52,11 @@ static bool mnodeCheckShowFinished(SShowObj *pShow); static void *mnodePutShowObj(SShowObj *pShow); static void mnodeReleaseShowObj(SShowObj *pShow, bool forceRemove); -extern void *tsMnodeTmr; static void *tsMnodeShowCache = NULL; static int32_t tsShowObjIndex = 0; static SShowMetaFp tsMnodeShowMetaFp[TSDB_MGMT_TABLE_MAX] = {0}; static SShowRetrieveFp tsMnodeShowRetrieveFp[TSDB_MGMT_TABLE_MAX] = {0}; +static SShowFreeIterFp tsMnodeShowFreeIterFp[TSDB_MGMT_TABLE_MAX] = {0}; int32_t mnodeInitShow() { mnodeAddReadMsgHandle(TSDB_MSG_TYPE_CM_SHOW, mnodeProcessShowMsg); @@ -85,6 +85,10 @@ void mnodeAddShowRetrieveHandle(uint8_t msgType, SShowRetrieveFp fp) { tsMnodeShowRetrieveFp[msgType] = fp; } +void mnodeAddShowFreeIterHandle(uint8_t msgType, SShowFreeIterFp fp) { + tsMnodeShowFreeIterFp[msgType] = fp; +} + static char *mnodeGetShowType(int32_t showType) { switch (showType) { case TSDB_MGMT_TABLE_ACCT: return "show accounts"; @@ -412,7 +416,9 @@ static void* mnodePutShowObj(SShowObj *pShow) { static void mnodeFreeShowObj(void *data) { SShowObj *pShow = *(SShowObj **)data; - sdbFreeIter(pShow->pIter); + if (tsMnodeShowFreeIterFp[pShow->type] != NULL && pShow->pIter != NULL) { + (*tsMnodeShowFreeIterFp[pShow->type])(pShow->pIter); + } mDebug("%p, show is destroyed, data:%p index:%d", pShow, data, pShow->index); tfree(pShow); diff --git a/src/mnode/src/mnodeTable.c b/src/mnode/src/mnodeTable.c index d0a5402986fcc9d24d3c8dce4e07e87af9a2061a..ff81c37de78b99af19a70b8cdb6882957b029972 100644 --- a/src/mnode/src/mnodeTable.c +++ b/src/mnode/src/mnodeTable.c @@ -49,7 +49,9 @@ #define CREATE_CTABLE_RETRY_TIMES 10 #define CREATE_CTABLE_RETRY_SEC 14 +int64_t tsCTableRid = -1; static void * tsChildTableSdb; +int64_t tsSTableRid = -1; static void * tsSuperTableSdb; static int32_t tsChildTableUpdateSize; static int32_t tsSuperTableUpdateSize; @@ -342,8 +344,7 @@ static int32_t mnodeChildTableActionRestored() { mnodeDecTableRef(pTable); } - sdbFreeIter(pIter); - + mnodeCancelGetNextChildTable(pIter); return 0; } @@ -351,7 +352,7 @@ static int32_t mnodeInitChildTables() { SCTableObj tObj; tsChildTableUpdateSize = (int8_t *)tObj.updateEnd - (int8_t *)&tObj.info.type; - SSdbTableDesc tableDesc = { + SSdbTableDesc desc = { .id = SDB_TABLE_CTABLE, .name = "ctables", .hashSessions = TSDB_DEFAULT_CTABLES_HASH_SIZE, @@ -367,7 +368,8 @@ static int32_t mnodeInitChildTables() { .fpRestored = mnodeChildTableActionRestored }; - tsChildTableSdb = sdbOpenTable(&tableDesc); + tsCTableRid = sdbOpenTable(&desc); + tsChildTableSdb = sdbGetTableByRid(tsCTableRid); if (tsChildTableSdb == NULL) { mError("failed to init child table data"); return -1; @@ -378,7 +380,7 @@ static int32_t mnodeInitChildTables() { } static void mnodeCleanupChildTables() { - sdbCloseTable(tsChildTableSdb); + sdbCloseTable(tsCTableRid); tsChildTableSdb = NULL; } @@ -544,7 +546,7 @@ static int32_t mnodeInitSuperTables() { SSTableObj tObj; tsSuperTableUpdateSize = (int8_t *)tObj.updateEnd - (int8_t *)&tObj.info.type; - SSdbTableDesc tableDesc = { + SSdbTableDesc desc = { .id = SDB_TABLE_STABLE, .name = "stables", .hashSessions = TSDB_DEFAULT_STABLES_HASH_SIZE, @@ -560,7 +562,8 @@ static int32_t mnodeInitSuperTables() { .fpRestored = mnodeSuperTableActionRestored }; - tsSuperTableSdb = sdbOpenTable(&tableDesc); + tsSTableRid = sdbOpenTable(&desc); + tsSuperTableSdb = sdbGetTableByRid(tsSTableRid); if (tsSuperTableSdb == NULL) { mError("failed to init stables data"); return -1; @@ -571,7 +574,7 @@ static int32_t mnodeInitSuperTables() { } static void mnodeCleanupSuperTables() { - sdbCloseTable(tsSuperTableSdb); + sdbCloseTable(tsSTableRid); tsSuperTableSdb = NULL; } @@ -602,10 +605,13 @@ int32_t mnodeInitTables() { mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_TABLE, mnodeGetShowTableMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_TABLE, mnodeRetrieveShowTables); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_TABLE, mnodeCancelGetNextChildTable); mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_METRIC, mnodeGetShowSuperTableMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_METRIC, mnodeRetrieveShowSuperTables); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_METRIC, mnodeCancelGetNextSuperTable); mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_STREAMTABLES, mnodeGetStreamTableMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_STREAMTABLES, mnodeRetrieveStreamTables); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_STREAMTABLES, mnodeCancelGetNextChildTable); return TSDB_CODE_SUCCESS; } @@ -626,14 +632,12 @@ static void *mnodeGetSuperTableByUid(uint64_t uid) { pIter = mnodeGetNextSuperTable(pIter, &pStable); if (pStable == NULL) break; if (pStable->uid == uid) { - sdbFreeIter(pIter); + mnodeCancelGetNextSuperTable(pIter); return pStable; } mnodeDecTableRef(pStable); } - sdbFreeIter(pIter); - return NULL; } @@ -655,10 +659,18 @@ void *mnodeGetNextChildTable(void *pIter, SCTableObj **pTable) { return sdbFetchRow(tsChildTableSdb, pIter, (void **)pTable); } +void mnodeCancelGetNextChildTable(void *pIter) { + sdbFreeIter(tsChildTableSdb, pIter); +} + void *mnodeGetNextSuperTable(void *pIter, SSTableObj **pTable) { return sdbFetchRow(tsSuperTableSdb, pIter, (void **)pTable); } +void mnodeCancelGetNextSuperTable(void *pIter) { + sdbFreeIter(tsSuperTableSdb, pIter); +} + void mnodeIncTableRef(void *p1) { STableObj *pTable = (STableObj *)p1; if (pTable->type == TSDB_SUPER_TABLE) { @@ -914,10 +926,10 @@ static int32_t mnodeProcessDropSuperTableMsg(SMnodeMsg *pMsg) { SSTableObj *pStable = (SSTableObj *)pMsg->pTable; if (pStable->vgHash != NULL /*pStable->numOfTables != 0*/) { - SHashMutableIterator *pIter = taosHashCreateIter(pStable->vgHash); - while (taosHashIterNext(pIter)) { - int32_t *pVgId = taosHashIterGet(pIter); + int32_t *pVgId = taosHashIterate(pStable->vgHash, NULL); + while (pVgId) { SVgObj *pVgroup = mnodeGetVgroup(*pVgId); + pVgId = taosHashIterate(pStable->vgHash, pVgId); if (pVgroup == NULL) break; SDropSTableMsg *pDrop = rpcMallocCont(sizeof(SDropSTableMsg)); @@ -933,7 +945,8 @@ static int32_t mnodeProcessDropSuperTableMsg(SMnodeMsg *pMsg) { dnodeSendMsgToDnode(&epSet, &rpcMsg); mnodeDecVgroupRef(pVgroup); } - taosHashDestroyIter(pIter); + + taosHashCancelIterate(pStable->vgHash, pVgId); mnodeDropAllChildTablesInStable(pStable); } @@ -1430,8 +1443,6 @@ void mnodeDropAllSuperTables(SDbObj *pDropDb) { mnodeDecTableRef(pTable); } - sdbFreeIter(pIter); - mInfo("db:%s, all super tables:%d is dropped from sdb", pDropDb->name, numOfTables); } @@ -1523,11 +1534,11 @@ static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) { } else { SVgroupsMsg *pVgroupMsg = (SVgroupsMsg *)msg; - SHashMutableIterator *pIter = taosHashCreateIter(pTable->vgHash); - int32_t vgSize = 0; - while (taosHashIterNext(pIter)) { - int32_t *pVgId = taosHashIterGet(pIter); - SVgObj * pVgroup = mnodeGetVgroup(*pVgId); + int32_t *pVgId = taosHashIterate(pTable->vgHash, NULL); + int32_t vgSize = 0; + while (pVgId) { + SVgObj *pVgroup = mnodeGetVgroup(*pVgId); + pVgId = taosHashIterate(pTable->vgHash, pVgId); if (pVgroup == NULL) continue; pVgroupMsg->vgroups[vgSize].vgId = htonl(pVgroup->vgId); @@ -1547,7 +1558,7 @@ static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) { mnodeDecVgroupRef(pVgroup); } - taosHashDestroyIter(pIter); + taosHashCancelIterate(pTable->vgHash, pVgId); mnodeDecTableRef(pTable); pVgroupMsg->numOfVgroups = htonl(vgSize); @@ -1723,6 +1734,16 @@ static int32_t mnodeDoCreateChildTable(SMnodeMsg *pMsg, int32_t tid) { if (pTable->info.type == TSDB_CHILD_TABLE) { STagData *pTagData = (STagData *)pCreate->schema; // it is a tag key + + char prefix[64] = {0}; + size_t prefixLen = tableIdPrefix(pMsg->pDb->name, prefix, 64); + if (0 != strncasecmp(prefix, pTagData->name, prefixLen)) { + mError("msg:%p, app:%p table:%s, corresponding super table:%s not in this db", pMsg, pMsg->rpcMsg.ahandle, + pCreate->tableId, pTagData->name); + mnodeDestroyChildTable(pTable); + return TSDB_CODE_TDB_INVALID_CREATE_TB_MSG; + } + if (pMsg->pSTable == NULL) pMsg->pSTable = mnodeGetSuperTable(pTagData->name); if (pMsg->pSTable == NULL) { mError("msg:%p, app:%p table:%s, corresponding super table:%s does not exist", pMsg, pMsg->rpcMsg.ahandle, @@ -2230,8 +2251,6 @@ void mnodeDropAllChildTablesInVgroups(SVgObj *pVgroup) { mnodeDecTableRef(pTable); } - sdbFreeIter(pIter); - mInfo("vgId:%d, all child tables is dropped from sdb", pVgroup->vgId); } @@ -2263,8 +2282,6 @@ void mnodeDropAllChildTables(SDbObj *pDropDb) { mnodeDecTableRef(pTable); } - sdbFreeIter(pIter); - mInfo("db:%s, all child tables:%d is dropped from sdb", pDropDb->name, numOfTables); } @@ -2293,8 +2310,6 @@ static void mnodeDropAllChildTablesInStable(SSTableObj *pStable) { mnodeDecTableRef(pTable); } - sdbFreeIter(pIter); - mInfo("stable:%s, all child tables:%d is dropped from sdb", pStable->info.tableId, numOfTables); } @@ -2624,9 +2639,7 @@ static int32_t mnodeRetrieveShowTables(SShowObj *pShow, char *data, int32_t rows SPatternCompareInfo info = PATTERN_COMPARE_INFO_INITIALIZER; char prefix[64] = {0}; - tstrncpy(prefix, pDb->name, 64); - strcat(prefix, TS_PATH_DELIMITER); - int32_t prefixLen = strlen(prefix); + int32_t prefixLen = tableIdPrefix(pDb->name, prefix, 64); char* pattern = NULL; if (pShow->payloadLen > 0) { diff --git a/src/mnode/src/mnodeUser.c b/src/mnode/src/mnodeUser.c index dc76d92eb8c2c01396705b33cc9ea55ccc4275df..fb26086d0433d7760c9cc7ed906abb570a4fcfee 100644 --- a/src/mnode/src/mnodeUser.c +++ b/src/mnode/src/mnodeUser.c @@ -33,6 +33,7 @@ #include "mnodeWrite.h" #include "mnodePeer.h" +int64_t tsUserRid = -1; static void * tsUserSdb = NULL; static int32_t tsUserUpdateSize = 0; static int32_t mnodeGetUserMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); @@ -123,7 +124,6 @@ static void mnodePrintUserAuth() { } fflush(fp); - sdbFreeIter(pIter); fclose(fp); } @@ -166,7 +166,8 @@ int32_t mnodeInitUsers() { .fpRestored = mnodeUserActionRestored }; - tsUserSdb = sdbOpenTable(&desc); + tsUserRid = sdbOpenTable(&desc); + tsUserSdb = sdbGetTableByRid(tsUserRid); if (tsUserSdb == NULL) { mError("table:%s, failed to create hash", desc.name); return -1; @@ -177,6 +178,8 @@ int32_t mnodeInitUsers() { mnodeAddWriteMsgHandle(TSDB_MSG_TYPE_CM_DROP_USER, mnodeProcessDropUserMsg); mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_USER, mnodeGetUserMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_USER, mnodeRetrieveUsers); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_USER, mnodeCancelGetNextUser); + mnodeAddPeerMsgHandle(TSDB_MSG_TYPE_DM_AUTH, mnodeProcessAuthMsg); mDebug("table:%s, hash is created", desc.name); @@ -184,7 +187,7 @@ int32_t mnodeInitUsers() { } void mnodeCleanupUsers() { - sdbCloseTable(tsUserSdb); + sdbCloseTable(tsUserRid); tsUserSdb = NULL; } @@ -196,6 +199,10 @@ void *mnodeGetNextUser(void *pIter, SUserObj **pUser) { return sdbFetchRow(tsUserSdb, pIter, (void **)pUser); } +void mnodeCancelGetNextUser(void *pIter) { + sdbFreeIter(tsUserSdb, pIter); +} + void mnodeIncUserRef(SUserObj *pUser) { return sdbIncRef(tsUserSdb, pUser); } @@ -574,8 +581,6 @@ void mnodeDropAllUsers(SAcctObj *pAcct) { mnodeDecUserRef(pUser); } - sdbFreeIter(pIter); - mDebug("acct:%s, all users:%d is dropped from sdb", pAcct->user, numOfUsers); } diff --git a/src/mnode/src/mnodeVgroup.c b/src/mnode/src/mnodeVgroup.c index f9a49e5ec20d611415023535f3ba2e8fce311054..d3020de6bd32dfcee97a068181d1e957e7a1a62a 100644 --- a/src/mnode/src/mnodeVgroup.c +++ b/src/mnode/src/mnodeVgroup.c @@ -20,7 +20,7 @@ #include "tsocket.h" #include "tidpool.h" #include "tsync.h" -#include "tbalance.h" +#include "tbn.h" #include "tglobal.h" #include "tdataformat.h" #include "dnode.h" @@ -51,6 +51,7 @@ char* vgroupStatus[] = { "updating" }; +int64_t tsVgroupRid = -1; static void *tsVgroupSdb = NULL; static int32_t tsVgUpdateSize = 0; @@ -222,7 +223,8 @@ int32_t mnodeInitVgroups() { .fpRestored = mnodeVgroupActionRestored, }; - tsVgroupSdb = sdbOpenTable(&desc); + tsVgroupRid = sdbOpenTable(&desc); + tsVgroupSdb = sdbGetTableByRid(tsVgroupRid); if (tsVgroupSdb == NULL) { mError("failed to init vgroups data"); return -1; @@ -230,6 +232,7 @@ int32_t mnodeInitVgroups() { mnodeAddShowMetaHandle(TSDB_MGMT_TABLE_VGROUP, mnodeGetVgroupMeta); mnodeAddShowRetrieveHandle(TSDB_MGMT_TABLE_VGROUP, mnodeRetrieveVgroups); + mnodeAddShowFreeIterHandle(TSDB_MGMT_TABLE_VGROUP, mnodeCancelGetNextVgroup); mnodeAddPeerRspHandle(TSDB_MSG_TYPE_MD_CREATE_VNODE_RSP, mnodeProcessCreateVnodeRsp); mnodeAddPeerRspHandle(TSDB_MSG_TYPE_MD_ALTER_VNODE_RSP, mnodeProcessAlterVnodeRsp); mnodeAddPeerRspHandle(TSDB_MSG_TYPE_MD_DROP_VNODE_RSP, mnodeProcessDropVnodeRsp); @@ -304,7 +307,7 @@ void mnodeCheckUnCreatedVgroup(SDnodeObj *pDnode, SVnodeLoad *pVloads, int32_t o mnodeDecVgroupRef(pVgroup); } - sdbFreeIter(pIter); + mnodeCancelGetNextVgroup(pIter); } void mnodeUpdateVgroupStatus(SVgObj *pVgroup, SDnodeObj *pDnode, SVnodeLoad *pVload) { @@ -491,6 +494,10 @@ void *mnodeGetNextVgroup(void *pIter, SVgObj **pVgroup) { return sdbFetchRow(tsVgroupSdb, pIter, (void **)pVgroup); } +void mnodeCancelGetNextVgroup(void *pIter) { + sdbFreeIter(tsVgroupSdb, pIter); +} + static int32_t mnodeCreateVgroupFp(SMnodeMsg *pMsg) { SVgObj *pVgroup = pMsg->pVgroup; SDbObj *pDb = pMsg->pDb; @@ -556,7 +563,7 @@ int32_t mnodeCreateVgroup(SMnodeMsg *pMsg) { pVgroup->numOfVnodes = pDb->cfg.replications; pVgroup->createdTime = taosGetTimestampMs(); pVgroup->accessState = TSDB_VN_ALL_ACCCESS; - int32_t code = balanceAllocVnodes(pVgroup); + int32_t code = bnAllocVnodes(pVgroup); if (code != TSDB_CODE_SUCCESS) { mError("db:%s, no enough dnode to alloc %d vnodes to vgroup, reason:%s", pDb->name, pVgroup->numOfVnodes, tstrerror(code)); @@ -605,7 +612,7 @@ void mnodeDropVgroup(SVgObj *pVgroup, void *ahandle) { } void mnodeCleanupVgroups() { - sdbCloseTable(tsVgroupSdb); + sdbCloseTable(tsVgroupRid); tsVgroupSdb = NULL; } @@ -1095,8 +1102,6 @@ void mnodeDropAllDnodeVgroups(SDnodeObj *pDropDnode) { mnodeDecVgroupRef(pVgroup); } - sdbFreeIter(pIter); - mInfo("dnode:%d, all vgroups:%d is dropped from sdb", pDropDnode->dnodeId, numOfVgroups); } @@ -1118,8 +1123,6 @@ void mnodeUpdateAllDbVgroups(SDbObj *pAlterDb) { mnodeDecVgroupRef(pVgroup); } - sdbFreeIter(pIter); - mInfo("db:%s, all vgroups is updated in sdb", pAlterDb->name); } #endif @@ -1147,8 +1150,6 @@ void mnodeDropAllDbVgroups(SDbObj *pDropDb) { mnodeDecVgroupRef(pVgroup); } - sdbFreeIter(pIter); - mInfo("db:%s, all vgroups:%d is dropped from sdb", pDropDb->name, numOfVgroups); } @@ -1170,7 +1171,5 @@ void mnodeSendDropAllDbVgroupsMsg(SDbObj *pDropDb) { numOfVgroups++; } - sdbFreeIter(pIter); - mInfo("db:%s, all vgroups:%d drop msg is sent to dnode", pDropDb->name, numOfVgroups); } diff --git a/src/mnode/src/mnodeWrite.c b/src/mnode/src/mnodeWrite.c index 8893316ffc2e9ac3291a1411d7ea5054e198ff9b..53981238a76ee3976689f77315eabab118816e81 100644 --- a/src/mnode/src/mnodeWrite.c +++ b/src/mnode/src/mnodeWrite.c @@ -17,7 +17,7 @@ #include "os.h" #include "taosdef.h" #include "tsched.h" -#include "tbalance.h" +#include "tbn.h" #include "tgrant.h" #include "tglobal.h" #include "trpc.h" @@ -54,18 +54,8 @@ int32_t mnodeProcessWrite(SMnodeMsg *pMsg) { rpcRsp->rsp = epSet; rpcRsp->len = sizeof(SRpcEpSet); - mDebug("msg:%p, app:%p type:%s in write queue, will be redirected, numOfEps:%d inUse:%d", pMsg, pMsg->rpcMsg.ahandle, - taosMsg[pMsg->rpcMsg.msgType], epSet->numOfEps, epSet->inUse); - for (int32_t i = 0; i < epSet->numOfEps; ++i) { - if (strcmp(epSet->fqdn[i], tsLocalFqdn) == 0 && htons(epSet->port[i]) == tsServerPort) { - epSet->inUse = (i + 1) % epSet->numOfEps; - mDebug("msg:%p, app:%p mnode index:%d ep:%s:%d, set inUse to %d", pMsg, pMsg->rpcMsg.ahandle, i, epSet->fqdn[i], - htons(epSet->port[i]), epSet->inUse); - } else { - mDebug("msg:%p, app:%p mnode index:%d ep:%s:%d", pMsg, pMsg->rpcMsg.ahandle, i, epSet->fqdn[i], - htons(epSet->port[i])); - } - } + mDebug("msg:%p, app:%p type:%s in write queue, is redirected, numOfEps:%d inUse:%d", pMsg, + pMsg->rpcMsg.ahandle, taosMsg[pMsg->rpcMsg.msgType], epSet->numOfEps, epSet->inUse); return TSDB_CODE_RPC_REDIRECT; } diff --git a/src/os/inc/osFile.h b/src/os/inc/osFile.h index 62e44d8eb0b70fb1526693895847637daa72247a..c9b3b9cd76d675169746a2478bfbd775d6da190a 100644 --- a/src/os/inc/osFile.h +++ b/src/os/inc/osFile.h @@ -20,17 +20,6 @@ extern "C" { #endif -#define tread(fd, buf, count) read(fd, buf, count) -#define twrite(fd, buf, count) write(fd, buf, count) -#define tlseek(fd, offset, whence) lseek(fd, offset, whence) -#define tclose(fd) \ - { \ - if (FD_VALID(fd)) { \ - close(fd); \ - fd = FD_INITIALIZER; \ - } \ - } - int64_t taosReadImp(int32_t fd, void *buf, int64_t count); int64_t taosWriteImp(int32_t fd, void *buf, int64_t count); int64_t taosLSeekImp(int32_t fd, int64_t offset, int32_t whence); @@ -39,7 +28,13 @@ int32_t taosRenameFile(char *fullPath, char *suffix, char delimiter, char **dstP #define taosRead(fd, buf, count) taosReadImp(fd, buf, count) #define taosWrite(fd, buf, count) taosWriteImp(fd, buf, count) #define taosLSeek(fd, offset, whence) taosLSeekImp(fd, offset, whence) -#define taosClose(x) tclose(x) +#define taosClose(fd) \ + { \ + if (FD_VALID(fd)) { \ + close(fd); \ + fd = FD_INITIALIZER; \ + } \ + } // TAOS_OS_FUNC_FILE_SENDIFLE int64_t taosSendFile(int32_t dfd, int32_t sfd, int64_t *offset, int64_t size); diff --git a/src/os/src/detail/osFile.c b/src/os/src/detail/osFile.c index 23fc88b8e1f5d4c0e2fe6b2140278b1e32d35abb..2e6886aa21f8c083e1134122ac334a20edf54874 100644 --- a/src/os/src/detail/osFile.c +++ b/src/os/src/detail/osFile.c @@ -116,7 +116,7 @@ int64_t taosWriteImp(int32_t fd, void *buf, int64_t n) { } int64_t taosLSeekImp(int32_t fd, int64_t offset, int32_t whence) { - return (int64_t)tlseek(fd, (long)offset, whence); + return (int64_t)lseek(fd, (long)offset, whence); } #ifndef TAOS_OS_FUNC_FILE_SENDIFLE diff --git a/src/plugins/monitor/src/monitorMain.c b/src/plugins/monitor/src/monMain.c similarity index 70% rename from src/plugins/monitor/src/monitorMain.c rename to src/plugins/monitor/src/monMain.c index 24998b54cd1a0114697c411fa145465f52e1e48a..9443b1ce12deba75a070f25ae4f67f7887870f7d 100644 --- a/src/plugins/monitor/src/monitorMain.c +++ b/src/plugins/monitor/src/monMain.c @@ -27,12 +27,12 @@ #include "monitor.h" #include "taoserror.h" -#define mnFatal(...) { if (monitorDebugFlag & DEBUG_FATAL) { taosPrintLog("MON FATAL ", 255, __VA_ARGS__); }} -#define mnError(...) { if (monitorDebugFlag & DEBUG_ERROR) { taosPrintLog("MON ERROR ", 255, __VA_ARGS__); }} -#define mnWarn(...) { if (monitorDebugFlag & DEBUG_WARN) { taosPrintLog("MON WARN ", 255, __VA_ARGS__); }} -#define mnInfo(...) { if (monitorDebugFlag & DEBUG_INFO) { taosPrintLog("MON ", 255, __VA_ARGS__); }} -#define mnDebug(...) { if (monitorDebugFlag & DEBUG_DEBUG) { taosPrintLog("MON ", monitorDebugFlag, __VA_ARGS__); }} -#define mnTrace(...) { if (monitorDebugFlag & DEBUG_TRACE) { taosPrintLog("MON ", monitorDebugFlag, __VA_ARGS__); }} +#define monFatal(...) { if (monDebugFlag & DEBUG_FATAL) { taosPrintLog("MON FATAL ", 255, __VA_ARGS__); }} +#define monError(...) { if (monDebugFlag & DEBUG_ERROR) { taosPrintLog("MON ERROR ", 255, __VA_ARGS__); }} +#define monWarn(...) { if (monDebugFlag & DEBUG_WARN) { taosPrintLog("MON WARN ", 255, __VA_ARGS__); }} +#define monInfo(...) { if (monDebugFlag & DEBUG_INFO) { taosPrintLog("MON ", 255, __VA_ARGS__); }} +#define monDebug(...) { if (monDebugFlag & DEBUG_DEBUG) { taosPrintLog("MON ", monDebugFlag, __VA_ARGS__); }} +#define monTrace(...) { if (monDebugFlag & DEBUG_TRACE) { taosPrintLog("MON ", monDebugFlag, __VA_ARGS__); }} #define SQL_LENGTH 1030 #define LOG_LEN_STR 100 @@ -48,12 +48,12 @@ typedef enum { MON_CMD_CREATE_TB_ACCT_ROOT, MON_CMD_CREATE_TB_SLOWQUERY, MON_CMD_MAX -} EMonitorCommand; +} EMonCmd; typedef enum { MON_STATE_NOT_INIT, MON_STATE_INITED -} EMonitorState; +} EMonState; typedef struct { pthread_t thread; @@ -64,17 +64,17 @@ typedef struct { int8_t start; // enable/disable by mnode int8_t quiting; // taosd is quiting char sql[SQL_LENGTH + 1]; -} SMonitorConn; +} SMonConn; -static SMonitorConn tsMonitor = {0}; -static void monitorSaveSystemInfo(); -static void *monitorThreadFunc(void *param); -static void monitorBuildMonitorSql(char *sql, int32_t cmd); -extern int32_t (*monitorStartSystemFp)(); -extern void (*monitorStopSystemFp)(); -extern void (*monitorExecuteSQLFp)(char *sql); +static SMonConn tsMonitor = {0}; +static void monSaveSystemInfo(); +static void *monThreadFunc(void *param); +static void monBuildMonitorSql(char *sql, int32_t cmd); +extern int32_t (*monStartSystemFp)(); +extern void (*monStopSystemFp)(); +extern void (*monExecuteSQLFp)(char *sql); -int32_t monitorInitSystem() { +int32_t monInitSystem() { if (tsMonitor.ep[0] == 0) { strcpy(tsMonitor.ep, tsLocalEp); } @@ -90,29 +90,29 @@ int32_t monitorInitSystem() { pthread_attr_init(&thAttr); pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); - if (pthread_create(&tsMonitor.thread, &thAttr, monitorThreadFunc, NULL)) { - mnError("failed to create thread to for monitor module, reason:%s", strerror(errno)); + if (pthread_create(&tsMonitor.thread, &thAttr, monThreadFunc, NULL)) { + monError("failed to create thread to for monitor module, reason:%s", strerror(errno)); return -1; } pthread_attr_destroy(&thAttr); - mnDebug("monitor thread is launched"); + monDebug("monitor thread is launched"); - monitorStartSystemFp = monitorStartSystem; - monitorStopSystemFp = monitorStopSystem; + monStartSystemFp = monStartSystem; + monStopSystemFp = monStopSystem; return 0; } -int32_t monitorStartSystem() { +int32_t monStartSystem() { taos_init(); tsMonitor.start = 1; - monitorExecuteSQLFp = monitorExecuteSQL; - mnInfo("monitor module start"); + monExecuteSQLFp = monExecuteSQL; + monInfo("monitor module start"); return 0; } -static void *monitorThreadFunc(void *param) { - mnDebug("starting to initialize monitor module ..."); +static void *monThreadFunc(void *param) { + monDebug("starting to initialize monitor module ..."); while (1) { static int32_t accessTimes = 0; @@ -121,7 +121,7 @@ static void *monitorThreadFunc(void *param) { if (tsMonitor.quiting) { tsMonitor.state = MON_STATE_NOT_INIT; - mnInfo("monitor thread will quit, for taosd is quiting"); + monInfo("monitor thread will quit, for taosd is quiting"); break; } else { taosGetDisk(); @@ -132,7 +132,7 @@ static void *monitorThreadFunc(void *param) { } if (dnodeGetDnodeId() <= 0) { - mnDebug("dnode not initialized, waiting for 3000 ms to start monitor module"); + monDebug("dnode not initialized, waiting for 3000 ms to start monitor module"); continue; } @@ -140,10 +140,10 @@ static void *monitorThreadFunc(void *param) { tsMonitor.state = MON_STATE_NOT_INIT; tsMonitor.conn = taos_connect(NULL, "monitor", tsInternalPass, "", 0); if (tsMonitor.conn == NULL) { - mnError("failed to connect to database, reason:%s", tstrerror(terrno)); + monError("failed to connect to database, reason:%s", tstrerror(terrno)); continue; } else { - mnDebug("connect to database success"); + monDebug("connect to database success"); } } @@ -151,16 +151,16 @@ static void *monitorThreadFunc(void *param) { int code = 0; for (; tsMonitor.cmdIndex < MON_CMD_MAX; ++tsMonitor.cmdIndex) { - monitorBuildMonitorSql(tsMonitor.sql, tsMonitor.cmdIndex); + monBuildMonitorSql(tsMonitor.sql, tsMonitor.cmdIndex); void *res = taos_query(tsMonitor.conn, tsMonitor.sql); code = taos_errno(res); taos_free_result(res); if (code != 0) { - mnError("failed to exec sql:%s, reason:%s", tsMonitor.sql, tstrerror(code)); + monError("failed to exec sql:%s, reason:%s", tsMonitor.sql, tstrerror(code)); break; } else { - mnDebug("successfully to exec sql:%s", tsMonitor.sql); + monDebug("successfully to exec sql:%s", tsMonitor.sql); } } @@ -171,16 +171,16 @@ static void *monitorThreadFunc(void *param) { if (tsMonitor.state == MON_STATE_INITED) { if (accessTimes % tsMonitorInterval == 0) { - monitorSaveSystemInfo(); + monSaveSystemInfo(); } } } - mnInfo("monitor thread is stopped"); + monInfo("monitor thread is stopped"); return NULL; } -static void monitorBuildMonitorSql(char *sql, int32_t cmd) { +static void monBuildMonitorSql(char *sql, int32_t cmd) { memset(sql, 0, SQL_LENGTH); if (cmd == MON_CMD_CREATE_DB) { @@ -236,47 +236,47 @@ static void monitorBuildMonitorSql(char *sql, int32_t cmd) { sql[SQL_LENGTH] = 0; } -void monitorStopSystem() { +void monStopSystem() { tsMonitor.start = 0; tsMonitor.state = MON_STATE_NOT_INIT; - monitorExecuteSQLFp = NULL; - mnInfo("monitor module stopped"); + monExecuteSQLFp = NULL; + monInfo("monitor module stopped"); } -void monitorCleanUpSystem() { +void monCleanupSystem() { tsMonitor.quiting = 1; - monitorStopSystem(); + monStopSystem(); pthread_join(tsMonitor.thread, NULL); if (tsMonitor.conn != NULL) { taos_close(tsMonitor.conn); tsMonitor.conn = NULL; } - mnInfo("monitor module is cleaned up"); + monInfo("monitor module is cleaned up"); } // unit is MB -static int32_t monitorBuildMemorySql(char *sql) { +static int32_t monBuildMemorySql(char *sql) { float sysMemoryUsedMB = 0; bool suc = taosGetSysMemory(&sysMemoryUsedMB); if (!suc) { - mnDebug("failed to get sys memory info"); + monDebug("failed to get sys memory info"); } float procMemoryUsedMB = 0; suc = taosGetProcMemory(&procMemoryUsedMB); if (!suc) { - mnDebug("failed to get proc memory info"); + monDebug("failed to get proc memory info"); } return sprintf(sql, ", %f, %f, %d", procMemoryUsedMB, sysMemoryUsedMB, tsTotalMemoryMB); } // unit is % -static int32_t monitorBuildCpuSql(char *sql) { +static int32_t monBuildCpuSql(char *sql) { float sysCpuUsage = 0, procCpuUsage = 0; bool suc = taosGetCpuUsage(&sysCpuUsage, &procCpuUsage); if (!suc) { - mnDebug("failed to get cpu usage"); + monDebug("failed to get cpu usage"); } if (sysCpuUsage <= procCpuUsage) { @@ -287,72 +287,72 @@ static int32_t monitorBuildCpuSql(char *sql) { } // unit is GB -static int32_t monitorBuildDiskSql(char *sql) { +static int32_t monBuildDiskSql(char *sql) { return sprintf(sql, ", %f, %d", (tsTotalDataDirGB - tsAvailDataDirGB), (int32_t)tsTotalDataDirGB); } // unit is Kb -static int32_t monitorBuildBandSql(char *sql) { +static int32_t monBuildBandSql(char *sql) { float bandSpeedKb = 0; bool suc = taosGetBandSpeed(&bandSpeedKb); if (!suc) { - mnDebug("failed to get bandwidth speed"); + monDebug("failed to get bandwidth speed"); } return sprintf(sql, ", %f", bandSpeedKb); } -static int32_t monitorBuildReqSql(char *sql) { +static int32_t monBuildReqSql(char *sql) { SStatisInfo info = dnodeGetStatisInfo(); return sprintf(sql, ", %d, %d, %d)", info.httpReqNum, info.queryReqNum, info.submitReqNum); } -static int32_t monitorBuildIoSql(char *sql) { +static int32_t monBuildIoSql(char *sql) { float readKB = 0, writeKB = 0; bool suc = taosGetProcIO(&readKB, &writeKB); if (!suc) { - mnDebug("failed to get io info"); + monDebug("failed to get io info"); } return sprintf(sql, ", %f, %f", readKB, writeKB); } -static void monitorSaveSystemInfo() { +static void monSaveSystemInfo() { int64_t ts = taosGetTimestampUs(); char * sql = tsMonitor.sql; int32_t pos = snprintf(sql, SQL_LENGTH, "insert into %s.dn%d values(%" PRId64, tsMonitorDbName, dnodeGetDnodeId(), ts); - pos += monitorBuildCpuSql(sql + pos); - pos += monitorBuildMemorySql(sql + pos); - pos += monitorBuildDiskSql(sql + pos); - pos += monitorBuildBandSql(sql + pos); - pos += monitorBuildIoSql(sql + pos); - pos += monitorBuildReqSql(sql + pos); + pos += monBuildCpuSql(sql + pos); + pos += monBuildMemorySql(sql + pos); + pos += monBuildDiskSql(sql + pos); + pos += monBuildBandSql(sql + pos); + pos += monBuildIoSql(sql + pos); + pos += monBuildReqSql(sql + pos); void *res = taos_query(tsMonitor.conn, tsMonitor.sql); int code = taos_errno(res); taos_free_result(res); if (code != 0) { - mnError("failed to save system info, reason:%s, sql:%s", tstrerror(code), tsMonitor.sql); + monError("failed to save system info, reason:%s, sql:%s", tstrerror(code), tsMonitor.sql); } else { - mnDebug("successfully to save system info, sql:%s", tsMonitor.sql); + monDebug("successfully to save system info, sql:%s", tsMonitor.sql); } } -static void montiorExecSqlCb(void *param, TAOS_RES *result, int32_t code) { +static void monExecSqlCb(void *param, TAOS_RES *result, int32_t code) { int32_t c = taos_errno(result); if (c != TSDB_CODE_SUCCESS) { - mnError("save %s failed, reason:%s", (char *)param, tstrerror(c)); + monError("save %s failed, reason:%s", (char *)param, tstrerror(c)); } else { int32_t rows = taos_affected_rows(result); - mnDebug("save %s succ, rows:%d", (char *)param, rows); + monDebug("save %s succ, rows:%d", (char *)param, rows); } taos_free_result(result); } -void monitorSaveAcctLog(SAcctMonitorObj *pMon) { +void monSaveAcctLog(SAcctMonitorObj *pMon) { if (tsMonitor.state != MON_STATE_INITED) return; char sql[1024] = {0}; @@ -382,11 +382,11 @@ void monitorSaveAcctLog(SAcctMonitorObj *pMon) { pMon->totalConns, pMon->maxConns, pMon->accessState); - mnDebug("save account info, sql:%s", sql); - taos_query_a(tsMonitor.conn, sql, montiorExecSqlCb, "account info"); + monDebug("save account info, sql:%s", sql); + taos_query_a(tsMonitor.conn, sql, monExecSqlCb, "account info"); } -void monitorSaveLog(int32_t level, const char *const format, ...) { +void monSaveLog(int32_t level, const char *const format, ...) { if (tsMonitor.state != MON_STATE_INITED) return; va_list argpointer; @@ -403,13 +403,13 @@ void monitorSaveLog(int32_t level, const char *const format, ...) { len += sprintf(sql + len, "', '%s')", tsLocalEp); sql[len++] = 0; - mnDebug("save log, sql: %s", sql); - taos_query_a(tsMonitor.conn, sql, montiorExecSqlCb, "log"); + monDebug("save log, sql: %s", sql); + taos_query_a(tsMonitor.conn, sql, monExecSqlCb, "log"); } -void monitorExecuteSQL(char *sql) { +void monExecuteSQL(char *sql) { if (tsMonitor.state != MON_STATE_INITED) return; - mnDebug("execute sql:%s", sql); - taos_query_a(tsMonitor.conn, sql, montiorExecSqlCb, "sql"); + monDebug("execute sql:%s", sql); + taos_query_a(tsMonitor.conn, sql, monExecSqlCb, "sql"); } diff --git a/src/query/inc/qExecutor.h b/src/query/inc/qExecutor.h index 895b414a56287bb6e35bcc0c50adcc338edbee1b..b73f7ce3f5cad809c12fcf17e5c57a79811605fd 100644 --- a/src/query/inc/qExecutor.h +++ b/src/query/inc/qExecutor.h @@ -63,9 +63,11 @@ typedef struct SSqlGroupbyExpr { typedef struct SResultRow { int32_t pageId; // pageId & rowId is the position of current result in disk-based output buffer - int32_t rowId:15; - bool closed:1; // this result status: closed or opened - uint16_t numOfRows; // number of rows of current time window + int32_t rowId:29; // row index in buffer page + bool startInterp; // the time window start timestamp has done the interpolation already. + bool endInterp; // the time window end timestamp has done the interpolation already. + bool closed; // this result status: closed or opened + uint32_t numOfRows; // number of rows of current time window SResultRowCellInfo* pCellInfo; // For each result column, there is a resultInfo union {STimeWindow win; char* key;}; // start key of current time window } SResultRow; @@ -81,16 +83,15 @@ typedef struct SResultRec { int32_t threshold; // result size threshold in rows. } SResultRec; -typedef struct SWindowResInfo { - SResultRow** pResult; // result list - int16_t type:8; // data type for hash key - int32_t size:24; // number of result set - int32_t threshold; // threshold to halt query and return the generated results. - int32_t capacity; // max capacity - int32_t curIndex; // current start active index - int64_t startTime; // start time of the first time window for sliding query - int64_t prevSKey; // previous (not completed) sliding window start key -} SWindowResInfo; +typedef struct SResultRowInfo { + SResultRow** pResult; // result list + int16_t type:8; // data type for hash key + int32_t size:24; // number of result set + int32_t capacity; // max capacity + int32_t curIndex; // current start active index + int64_t startTime; // start time of the first time window for sliding query + int64_t prevSKey; // previous (not completed) sliding window start key +} SResultRowInfo; typedef struct SColumnFilterElem { int16_t bytes; // column length @@ -113,7 +114,7 @@ typedef struct STableQueryInfo { STimeWindow win; STSCursor cur; void* pTable; // for retrieve the page id list - SWindowResInfo windowResInfo; + SResultRowInfo windowResInfo; } STableQueryInfo; typedef struct SQueryCostInfo { @@ -177,7 +178,7 @@ typedef struct SQueryRuntimeEnv { uint16_t* offset; uint16_t scanFlag; // denotes reversed scan of data or not SFillInfo* pFillInfo; - SWindowResInfo windowResInfo; + SResultRowInfo windowResInfo; STSBuf* pTSBuf; STSCursor cur; SQueryCostInfo summary; @@ -187,6 +188,8 @@ typedef struct SQueryRuntimeEnv { bool topBotQuery; // false bool groupbyNormalCol; // denote if this is a groupby normal column query bool hasTagResults; // if there are tag values in final result or not + bool timeWindowInterpo;// if the time window start/end required interpolation + bool queryWindowIdentical; // all query time windows are identical for all tables in one group int32_t interBufSize; // intermediate buffer sizse int32_t prevGroupId; // previous executed group id SDiskbasedResultBuf* pResultBuf; // query result buffer based on blocked-wised disk file @@ -195,6 +198,8 @@ typedef struct SQueryRuntimeEnv { SResultRowPool* pool; // window result object pool int32_t* rowCellInfoOffset;// offset value for each row result cell info + char** prevRow; + char** nextRow; } SQueryRuntimeEnv; enum { @@ -212,7 +217,8 @@ typedef struct SQInfo { STableGroupInfo tableGroupInfo; // table list SArray STableGroupInfo tableqinfoGroupInfo; // this is a group array list, including SArray structure SQueryRuntimeEnv runtimeEnv; - SArray* arrTableIdInfo; +// SArray* arrTableIdInfo; + SHashObj* arrTableIdInfo; int32_t groupIndex; /* diff --git a/src/query/inc/qUtil.h b/src/query/inc/qUtil.h index bc23ffadd40e1b5285f442aa87b8de934d5a9ea4..fb71c8a5fe0a7948c09c4bb4d83ed77b5d291389 100644 --- a/src/query/inc/qUtil.h +++ b/src/query/inc/qUtil.h @@ -27,22 +27,22 @@ int32_t getOutputInterResultBufSize(SQuery* pQuery); void clearResultRow(SQueryRuntimeEnv* pRuntimeEnv, SResultRow* pRow, int16_t type); -void copyResultRow(SQueryRuntimeEnv* pRuntimeEnv, SResultRow* dst, const SResultRow* src); +void copyResultRow(SQueryRuntimeEnv* pRuntimeEnv, SResultRow* dst, const SResultRow* src, int16_t type); SResultRowCellInfo* getResultCell(SQueryRuntimeEnv* pRuntimeEnv, const SResultRow* pRow, int32_t index); -int32_t initWindowResInfo(SWindowResInfo* pWindowResInfo, int32_t size, int32_t threshold, int16_t type); +int32_t initWindowResInfo(SResultRowInfo* pWindowResInfo, int32_t size, int16_t type); -void cleanupTimeWindowInfo(SWindowResInfo* pWindowResInfo); -void resetTimeWindowInfo(SQueryRuntimeEnv* pRuntimeEnv, SWindowResInfo* pWindowResInfo); +void cleanupTimeWindowInfo(SResultRowInfo* pWindowResInfo); +void resetTimeWindowInfo(SQueryRuntimeEnv* pRuntimeEnv, SResultRowInfo* pWindowResInfo); void clearFirstNWindowRes(SQueryRuntimeEnv *pRuntimeEnv, int32_t num); void clearClosedTimeWindow(SQueryRuntimeEnv* pRuntimeEnv); -int32_t numOfClosedTimeWindow(SWindowResInfo* pWindowResInfo); -void closeTimeWindow(SWindowResInfo* pWindowResInfo, int32_t slot); -void closeAllTimeWindow(SWindowResInfo* pWindowResInfo); -void removeRedundantWindow(SWindowResInfo *pWindowResInfo, TSKEY lastKey, int32_t order); +int32_t numOfClosedTimeWindow(SResultRowInfo* pWindowResInfo); +void closeTimeWindow(SResultRowInfo* pWindowResInfo, int32_t slot); +void closeAllTimeWindow(SResultRowInfo* pWindowResInfo); +void removeRedundantWindow(SResultRowInfo *pWindowResInfo, TSKEY lastKey, int32_t order); -static FORCE_INLINE SResultRow *getResultRow(SWindowResInfo *pWindowResInfo, int32_t slot) { +static FORCE_INLINE SResultRow *getResultRow(SResultRowInfo *pWindowResInfo, int32_t slot) { assert(pWindowResInfo != NULL && slot >= 0 && slot < pWindowResInfo->size); return pWindowResInfo->pResult[slot]; } @@ -50,7 +50,7 @@ static FORCE_INLINE SResultRow *getResultRow(SWindowResInfo *pWindowResInfo, int #define curTimeWindowIndex(_winres) ((_winres)->curIndex) #define GET_ROW_PARAM_FOR_MULTIOUTPUT(_q, tbq, sq) (((tbq) && (!sq))? (_q)->pExpr1[1].base.arg->argValue.i64:1) -bool isWindowResClosed(SWindowResInfo *pWindowResInfo, int32_t slot); +bool isWindowResClosed(SResultRowInfo *pWindowResInfo, int32_t slot); int32_t initResultRow(SResultRow *pResultRow); diff --git a/src/query/inc/tsqlfunction.h b/src/query/inc/tsqlfunction.h index 32cbb56c62514001ea5314e7e5dce8549daddb97..5a923db52c6b9a7897de2b2ac72e59bf570a3a56 100644 --- a/src/query/inc/tsqlfunction.h +++ b/src/query/inc/tsqlfunction.h @@ -152,6 +152,11 @@ typedef struct SResultRowCellInfo { uint32_t numOfRes; // num of output result in current buffer } SResultRowCellInfo; +typedef struct SPoint1 { + int64_t key; + double val; +} SPoint1; + #define GET_ROWCELL_INTERBUF(_c) ((void*) ((char*)(_c) + sizeof(SResultRowCellInfo))) struct SQLFunctionCtx; @@ -194,6 +199,8 @@ typedef struct SQLFunctionCtx { SResultRowCellInfo *resultInfo; SExtTagsInfo tagInfo; + SPoint1 start; + SPoint1 end; } SQLFunctionCtx; typedef struct SQLAggFuncElem { @@ -243,21 +250,11 @@ enum { }; typedef struct STwaInfo { - TSKEY lastKey; - int8_t hasResult; // flag to denote has value - int16_t type; // source data type - TSKEY SKey; - TSKEY EKey; - - union { - double dOutput; - int64_t iOutput; - }; - - union { - double dLastValue; - int64_t iLastValue; - }; + TSKEY lastKey; + int8_t hasResult; // flag to denote has value + double dOutput; + double lastValue; + STimeWindow win; } STwaInfo; /* global sql function array */ @@ -276,8 +273,6 @@ bool topbot_datablock_filter(SQLFunctionCtx *pCtx, int32_t functionId, const cha (_r)->initialized = false; \ } while (0) -//void setResultInfoBuf(SResultRowCellInfo *pResInfo, char* buf); - static FORCE_INLINE void initResultInfo(SResultRowCellInfo *pResInfo, uint32_t bufLen) { pResInfo->initialized = true; // the this struct has been initialized flag diff --git a/src/query/src/qAst.c b/src/query/src/qAst.c index e813688d8408bd124149352b337f1c0bef71d2af..a65f4a6dc95a586896372a648dc19ad2bb88b154 100644 --- a/src/query/src/qAst.c +++ b/src/query/src/qAst.c @@ -370,6 +370,66 @@ void tExprTreeTraverse(tExprNode *pExpr, SSkipList *pSkipList, SArray *result, S #endif } +static void reverseCopy(char* dest, const char* src, int16_t type, int32_t numOfRows) { + switch(type) { + case TSDB_DATA_TYPE_TINYINT: { + int8_t* p = (int8_t*) dest; + int8_t* pSrc = (int8_t*) src; + + for(int32_t i = 0; i < numOfRows; ++i) { + p[i] = pSrc[numOfRows - i - 1]; + } + break; + } + case TSDB_DATA_TYPE_SMALLINT: { + int16_t* p = (int16_t*) dest; + int16_t* pSrc = (int16_t*) src; + + for(int32_t i = 0; i < numOfRows; ++i) { + p[i] = pSrc[numOfRows - i - 1]; + } + break; + } + case TSDB_DATA_TYPE_INT: { + int32_t* p = (int32_t*) dest; + int32_t* pSrc = (int32_t*) src; + + for(int32_t i = 0; i < numOfRows; ++i) { + p[i] = pSrc[numOfRows - i - 1]; + } + break; + } + case TSDB_DATA_TYPE_BIGINT: { + int64_t* p = (int64_t*) dest; + int64_t* pSrc = (int64_t*) src; + + for(int32_t i = 0; i < numOfRows; ++i) { + p[i] = pSrc[numOfRows - i - 1]; + } + break; + } + case TSDB_DATA_TYPE_FLOAT: { + float* p = (float*) dest; + float* pSrc = (float*) src; + + for(int32_t i = 0; i < numOfRows; ++i) { + p[i] = pSrc[numOfRows - i - 1]; + } + break; + } + case TSDB_DATA_TYPE_DOUBLE: { + double* p = (double*) dest; + double* pSrc = (double*) src; + + for(int32_t i = 0; i < numOfRows; ++i) { + p[i] = pSrc[numOfRows - i - 1]; + } + break; + } + default: assert(0); + } +} + void tExprTreeCalcTraverse(tExprNode *pExprs, int32_t numOfRows, char *pOutput, void *param, int32_t order, char *(*getSourceDataBlock)(void *, const char*, int32_t)) { if (pExprs == NULL) { @@ -387,6 +447,8 @@ void tExprTreeCalcTraverse(tExprNode *pExprs, int32_t numOfRows, char *pOutput, /* the right output has result from the right child syntax tree */ char *pRightOutput = malloc(sizeof(int64_t) * numOfRows); + char *pdata = malloc(sizeof(int64_t) * numOfRows); + if (pRight->nodeType == TSQL_NODE_EXPR) { tExprTreeCalcTraverse(pRight, numOfRows, pRightOutput, param, order, getSourceDataBlock); } @@ -398,52 +460,75 @@ void tExprTreeCalcTraverse(tExprNode *pExprs, int32_t numOfRows, char *pOutput, * the type of returned value of one expression is always double float precious */ _bi_consumer_fn_t fp = tGetBiConsumerFn(TSDB_DATA_TYPE_DOUBLE, TSDB_DATA_TYPE_DOUBLE, pExprs->_node.optr); - fp(pLeftOutput, pRightOutput, numOfRows, numOfRows, pOutput, order); + fp(pLeftOutput, pRightOutput, numOfRows, numOfRows, pOutput, TSDB_ORDER_ASC); } else if (pRight->nodeType == TSQL_NODE_COL) { // exprLeft + columnRight _bi_consumer_fn_t fp = tGetBiConsumerFn(TSDB_DATA_TYPE_DOUBLE, pRight->pSchema->type, pExprs->_node.optr); + // set input buffer char *pInputData = getSourceDataBlock(param, pRight->pSchema->name, pRight->pSchema->colId); - fp(pLeftOutput, pInputData, numOfRows, numOfRows, pOutput, order); + if (order == TSDB_ORDER_DESC) { + reverseCopy(pdata, pInputData, pRight->pSchema->type, numOfRows); + fp(pLeftOutput, pdata, numOfRows, numOfRows, pOutput, TSDB_ORDER_ASC); + } else { + fp(pLeftOutput, pInputData, numOfRows, numOfRows, pOutput, TSDB_ORDER_ASC); + } } else if (pRight->nodeType == TSQL_NODE_VALUE) { // exprLeft + 12 _bi_consumer_fn_t fp = tGetBiConsumerFn(TSDB_DATA_TYPE_DOUBLE, pRight->pVal->nType, pExprs->_node.optr); - fp(pLeftOutput, &pRight->pVal->i64Key, numOfRows, 1, pOutput, order); + fp(pLeftOutput, &pRight->pVal->i64Key, numOfRows, 1, pOutput, TSDB_ORDER_ASC); } } else if (pLeft->nodeType == TSQL_NODE_COL) { // column data specified on left-hand-side char *pLeftInputData = getSourceDataBlock(param, pLeft->pSchema->name, pLeft->pSchema->colId); if (pRight->nodeType == TSQL_NODE_EXPR) { // columnLeft + expr2 _bi_consumer_fn_t fp = tGetBiConsumerFn(pLeft->pSchema->type, TSDB_DATA_TYPE_DOUBLE, pExprs->_node.optr); - fp(pLeftInputData, pRightOutput, numOfRows, numOfRows, pOutput, order); + + if (order == TSDB_ORDER_DESC) { + reverseCopy(pdata, pLeftInputData, pLeft->pSchema->type, numOfRows); + fp(pdata, pRightOutput, numOfRows, numOfRows, pOutput, TSDB_ORDER_ASC); + } else { + fp(pLeftInputData, pRightOutput, numOfRows, numOfRows, pOutput, TSDB_ORDER_ASC); + } } else if (pRight->nodeType == TSQL_NODE_COL) { // columnLeft + columnRight // column data specified on right-hand-side char *pRightInputData = getSourceDataBlock(param, pRight->pSchema->name, pRight->pSchema->colId); - _bi_consumer_fn_t fp = tGetBiConsumerFn(pLeft->pSchema->type, pRight->pSchema->type, pExprs->_node.optr); - fp(pLeftInputData, pRightInputData, numOfRows, numOfRows, pOutput, order); + // both columns are descending order, do not reverse the source data + fp(pLeftInputData, pRightInputData, numOfRows, numOfRows, pOutput, order); } else if (pRight->nodeType == TSQL_NODE_VALUE) { // columnLeft + 12 _bi_consumer_fn_t fp = tGetBiConsumerFn(pLeft->pSchema->type, pRight->pVal->nType, pExprs->_node.optr); - fp(pLeftInputData, &pRight->pVal->i64Key, numOfRows, 1, pOutput, order); + + if (order == TSDB_ORDER_DESC) { + reverseCopy(pdata, pLeftInputData, pLeft->pSchema->type, numOfRows); + fp(pdata, &pRight->pVal->i64Key, numOfRows, 1, pOutput, TSDB_ORDER_ASC); + } else { + fp(pLeftInputData, &pRight->pVal->i64Key, numOfRows, 1, pOutput, TSDB_ORDER_ASC); + } } } else { // column data specified on left-hand-side if (pRight->nodeType == TSQL_NODE_EXPR) { // 12 + expr2 _bi_consumer_fn_t fp = tGetBiConsumerFn(pLeft->pVal->nType, TSDB_DATA_TYPE_DOUBLE, pExprs->_node.optr); - fp(&pLeft->pVal->i64Key, pRightOutput, 1, numOfRows, pOutput, order); + fp(&pLeft->pVal->i64Key, pRightOutput, 1, numOfRows, pOutput, TSDB_ORDER_ASC); } else if (pRight->nodeType == TSQL_NODE_COL) { // 12 + columnRight // column data specified on right-hand-side char *pRightInputData = getSourceDataBlock(param, pRight->pSchema->name, pRight->pSchema->colId); - _bi_consumer_fn_t fp = tGetBiConsumerFn(pLeft->pVal->nType, pRight->pSchema->type, pExprs->_node.optr); - fp(&pLeft->pVal->i64Key, pRightInputData, 1, numOfRows, pOutput, order); + + if (order == TSDB_ORDER_DESC) { + reverseCopy(pdata, pRightInputData, pRight->pSchema->type, numOfRows); + fp(&pLeft->pVal->i64Key, pdata, numOfRows, 1, pOutput, TSDB_ORDER_ASC); + } else { + fp(&pLeft->pVal->i64Key, pRightInputData, 1, numOfRows, pOutput, TSDB_ORDER_ASC); + } } else if (pRight->nodeType == TSQL_NODE_VALUE) { // 12 + 12 _bi_consumer_fn_t fp = tGetBiConsumerFn(pLeft->pVal->nType, pRight->pVal->nType, pExprs->_node.optr); - fp(&pLeft->pVal->i64Key, &pRight->pVal->i64Key, 1, 1, pOutput, order); + fp(&pLeft->pVal->i64Key, &pRight->pVal->i64Key, 1, 1, pOutput, TSDB_ORDER_ASC); } } diff --git a/src/query/src/qExecutor.c b/src/query/src/qExecutor.c index fce25ae119bd842b59a5645bbe465c5de0c85f8a..a45c0ac6ef666db3ce17165525b8eeacf1fff6d6 100644 --- a/src/query/src/qExecutor.c +++ b/src/query/src/qExecutor.c @@ -27,6 +27,7 @@ #include "query.h" #include "queryLog.h" #include "tlosertree.h" +#include "ttype.h" #define MAX_ROWS_PER_RESBUF_PAGE ((1u<<12) - 1) @@ -50,8 +51,8 @@ #define SDATA_BLOCK_INITIALIZER (SDataBlockInfo) {{0}, 0} #define TIME_WINDOW_COPY(_dst, _src) do {\ - _dst.skey = _src.skey;\ - _dst.ekey = _src.ekey;\ + (_dst).skey = (_src).skey;\ + (_dst).ekey = (_src).ekey;\ } while (0); enum { @@ -194,6 +195,10 @@ static int32_t setAdditionalInfo(SQInfo *pQInfo, void *pTable, STableQueryInfo * static int32_t flushFromResultBuf(SQueryRuntimeEnv* pRuntimeEnv, SGroupResInfo* pGroupResInfo); static int32_t checkForQueryBuf(size_t numOfTables); static void releaseQueryBuf(size_t numOfTables); +static int32_t binarySearchForKey(char *pValue, int num, TSKEY key, int order); +static void doRowwiseTimeWindowInterpolation(SQueryRuntimeEnv* pRuntimeEnv, SArray* pDataBlock, TSKEY prevTs, int32_t prevRowIndex, TSKEY curTs, int32_t curRowIndex, TSKEY windowKey, int32_t type); +static STsdbQueryCond createTsdbQueryCond(SQuery* pQuery, STimeWindow* win); +static STableIdInfo createTableIdInfo(SQuery* pQuery); bool doFilterData(SQuery *pQuery, int32_t elemPos) { for (int32_t k = 0; k < pQuery->numOfFilterCols; ++k) { @@ -400,6 +405,17 @@ static bool isTopBottomQuery(SQuery *pQuery) { return false; } +static bool timeWindowInterpoRequired(SQuery *pQuery) { + for(int32_t i = 0; i < pQuery->numOfOutput; ++i) { + int32_t functionId = pQuery->pExpr1[i].base.functionId; + if (functionId == TSDB_FUNC_TWA) { + return true; + } + } + + return false; +} + static bool hasTagValOutput(SQuery* pQuery) { SExprInfo *pExprInfo = &pQuery->pExpr1[0]; if (pQuery->numOfOutput == 1 && pExprInfo->base.functionId == TSDB_FUNC_TS_COMP) { @@ -445,7 +461,7 @@ static bool hasNullValue(SColIndex* pColIndex, SDataStatis *pStatis, SDataStatis return true; } -static SResultRow *doPrepareResultRowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowResInfo, char *pData, +static SResultRow *doPrepareResultRowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SResultRowInfo *pWindowResInfo, char *pData, int16_t bytes, bool masterscan, uint64_t uid) { SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, pData, bytes, uid); int32_t *p1 = @@ -457,6 +473,7 @@ static SResultRow *doPrepareResultRowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SWin return NULL; } + // TODO refactor // more than the capacity, reallocate the resources if (pWindowResInfo->size >= pWindowResInfo->capacity) { int64_t newCapacity = 0; @@ -501,7 +518,7 @@ static SResultRow *doPrepareResultRowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SWin } // get the correct time window according to the handled timestamp -static STimeWindow getActiveTimeWindow(SWindowResInfo *pWindowResInfo, int64_t ts, SQuery *pQuery) { +static STimeWindow getActiveTimeWindow(SResultRowInfo *pWindowResInfo, int64_t ts, SQuery *pQuery) { STimeWindow w = {0}; if (pWindowResInfo->curIndex == -1) { // the first window, from the previous stored value @@ -594,15 +611,18 @@ static int32_t addNewWindowResultBuf(SResultRow *pWindowRes, SDiskbasedResultBuf return 0; } -static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowResInfo, SDataBlockInfo* pBockInfo, - STimeWindow *win, bool masterscan, bool* newWind) { +static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SResultRowInfo *pWindowResInfo, SDataBlockInfo* pBockInfo, + STimeWindow *win, bool masterscan, bool* newWind, SResultRow** pResult) { assert(win->skey <= win->ekey); SDiskbasedResultBuf *pResultBuf = pRuntimeEnv->pResultBuf; - SResultRow *pResultRow = doPrepareResultRowFromKey(pRuntimeEnv, pWindowResInfo, (char *)&win->skey, TSDB_KEYSIZE, masterscan, pBockInfo->uid); + // todo refactor + int64_t uid = getResultInfoUId(pRuntimeEnv); + SResultRow *pResultRow = doPrepareResultRowFromKey(pRuntimeEnv, pWindowResInfo, (char *)&win->skey, TSDB_KEYSIZE, masterscan, uid); if (pResultRow == NULL) { *newWind = false; + // no master scan, no result generated means error occurs return masterscan? -1:0; } @@ -618,15 +638,40 @@ static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowRes // set time window for current result pResultRow->win = (*win); + *pResult = pResultRow; setResultRowOutputBufInitCtx(pRuntimeEnv, pResultRow); + return TSDB_CODE_SUCCESS; } -static bool getTimeWindowResStatus(SWindowResInfo *pWindowResInfo, int32_t slot) { +static bool getResultRowStatus(SResultRowInfo *pWindowResInfo, int32_t slot) { assert(slot >= 0 && slot < pWindowResInfo->size); return pWindowResInfo->pResult[slot]->closed; } +typedef enum SResultTsInterpType { + RESULT_ROW_START_INTERP = 1, + RESULT_ROW_END_INTERP = 2, +} SResultTsInterpType; + +static void setResultRowInterpo(SResultRow* pResult, SResultTsInterpType type) { + assert(pResult != NULL && (type == RESULT_ROW_START_INTERP || type == RESULT_ROW_END_INTERP)); + if (type == RESULT_ROW_START_INTERP) { + pResult->startInterp = true; + } else { + pResult->endInterp = true; + } +} + +static bool resultRowInterpolated(SResultRow* pResult, SResultTsInterpType type) { + assert(pResult != NULL && (type == RESULT_ROW_START_INTERP || type == RESULT_ROW_END_INTERP)); + if (type == RESULT_ROW_START_INTERP) { + return pResult->startInterp == true; + } else { + return pResult->endInterp == true; + } +} + static FORCE_INLINE int32_t getForwardStepsInBlock(int32_t numOfRows, __block_search_fn_t searchFn, TSKEY ekey, int16_t pos, int16_t order, int64_t *pData) { int32_t forwardStep = 0; @@ -658,7 +703,7 @@ static FORCE_INLINE int32_t getForwardStepsInBlock(int32_t numOfRows, __block_se /** * NOTE: the query status only set for the first scan of master scan. */ -static int32_t doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKey, SWindowResInfo *pWindowResInfo) { +static int32_t doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKey, SResultRowInfo *pWindowResInfo) { SQuery *pQuery = pRuntimeEnv->pQuery; if (pRuntimeEnv->scanFlag != MASTER_SCAN) { return pWindowResInfo->size; @@ -716,7 +761,7 @@ static int32_t doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKe pWindowResInfo->prevSKey = pWindowResInfo->pResult[pWindowResInfo->curIndex]->win.skey; // the number of completed slots are larger than the threshold, return current generated results to client. - if (numOfClosed > pWindowResInfo->threshold) { + if (numOfClosed > pQuery->rec.threshold) { qDebug("QInfo:%p total result window:%d closed:%d, reached the output threshold %d, return", GET_QINFO_ADDR(pRuntimeEnv), pWindowResInfo->size, numOfClosed, pQuery->rec.threshold); @@ -947,7 +992,7 @@ static char *getDataBlock(SQueryRuntimeEnv *pRuntimeEnv, SArithmeticSupport *sas if (functionId == TSDB_FUNC_ARITHM) { sas->pArithExpr = &pQuery->pExpr1[col]; - sas->offset = (QUERY_IS_ASC_QUERY(pQuery)) ? pQuery->pos : pQuery->pos - (size - 1); + sas->offset = (QUERY_IS_ASC_QUERY(pQuery))? pQuery->pos : pQuery->pos - (size - 1); sas->colList = pQuery->colList; sas->numOfCols = pQuery->numOfCols; sas->data = calloc(pQuery->numOfCols, POINTER_BYTES); @@ -990,6 +1035,117 @@ static char *getDataBlock(SQueryRuntimeEnv *pRuntimeEnv, SArithmeticSupport *sas return dataBlock; } +static void setNotInterpoWindowKey(SQLFunctionCtx* pCtx, int32_t numOfOutput, int32_t type) { + if (type == RESULT_ROW_START_INTERP) { + for (int32_t k = 0; k < numOfOutput; ++k) { + pCtx[k].start.key = INT64_MIN; + } + } else { + for (int32_t k = 0; k < numOfOutput; ++k) { + pCtx[k].end.key = INT64_MIN; + } + } +} + +//static double getTSWindowInterpoVal(SColumnInfoData* pColInfo, int16_t srcColIndex, int16_t rowIndex, TSKEY key, char** prevRow, TSKEY* tsCols, int32_t step) { +// TSKEY start = tsCols[rowIndex]; +// TSKEY prevTs = (rowIndex == 0)? *(TSKEY *) prevRow[0] : tsCols[rowIndex - step]; +// +// double v1 = 0, v2 = 0, v = 0; +// char *prevVal = (rowIndex == 0)? prevRow[srcColIndex] : ((char*)pColInfo->pData) + (rowIndex - step) * pColInfo->info.bytes; +// +// GET_TYPED_DATA(v1, double, pColInfo->info.type, (char *)prevVal); +// GET_TYPED_DATA(v2, double, pColInfo->info.type, (char *)pColInfo->pData + rowIndex * pColInfo->info.bytes); +// +// SPoint point1 = (SPoint){.key = prevTs, .val = &v1}; +// SPoint point2 = (SPoint){.key = start, .val = &v2}; +// SPoint point = (SPoint){.key = key, .val = &v}; +// taosGetLinearInterpolationVal(TSDB_DATA_TYPE_DOUBLE, &point1, &point2, &point); +// +// return v; +//} + +// window start key interpolation +static bool setTimeWindowInterpolationStartTs(SQueryRuntimeEnv* pRuntimeEnv, int32_t pos, int32_t numOfRows, SArray* pDataBlock, TSKEY* tsCols, STimeWindow* win) { + SQuery* pQuery = pRuntimeEnv->pQuery; + + TSKEY curTs = tsCols[pos]; + TSKEY lastTs = *(TSKEY *) pRuntimeEnv->prevRow[0]; + + // lastTs == INT64_MIN and pos == 0 means this is the first time window, interpolation is not needed. + // start exactly from this point, no need to do interpolation + TSKEY key = QUERY_IS_ASC_QUERY(pQuery)? win->skey:win->ekey; + if (key == curTs) { + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_START_INTERP); + return true; + } + + if (lastTs == INT64_MIN && ((pos == 0 && QUERY_IS_ASC_QUERY(pQuery)) || (pos == (numOfRows - 1) && !QUERY_IS_ASC_QUERY(pQuery)))) { + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_START_INTERP); + return true; + } + + int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order); + TSKEY prevTs = ((pos == 0 && QUERY_IS_ASC_QUERY(pQuery)) || (pos == (numOfRows - 1) && !QUERY_IS_ASC_QUERY(pQuery)))? + lastTs:tsCols[pos - step]; + + doRowwiseTimeWindowInterpolation(pRuntimeEnv, pDataBlock, prevTs, pos - step, curTs, pos, key, RESULT_ROW_START_INTERP); + return true; +} + +static bool setTimeWindowInterpolationEndTs(SQueryRuntimeEnv* pRuntimeEnv, int32_t endRowIndex, SArray* pDataBlock, TSKEY* tsCols, TSKEY blockEkey, STimeWindow* win) { + SQuery* pQuery = pRuntimeEnv->pQuery; + TSKEY actualEndKey = tsCols[endRowIndex]; + + TSKEY key = QUERY_IS_ASC_QUERY(pQuery)? win->ekey:win->skey; + + // not ended in current data block, do not invoke interpolation + if ((key > blockEkey && QUERY_IS_ASC_QUERY(pQuery)) || (key < blockEkey && !QUERY_IS_ASC_QUERY(pQuery))) { + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_END_INTERP); + return false; + } + + // there is actual end point of current time window, no interpolation need + if (key == actualEndKey) { + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_END_INTERP); + return true; + } + + int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order); + int32_t nextRowIndex = endRowIndex + step; + assert(nextRowIndex >= 0); + + TSKEY nextKey = tsCols[nextRowIndex]; + doRowwiseTimeWindowInterpolation(pRuntimeEnv, pDataBlock, actualEndKey, endRowIndex, nextKey, nextRowIndex, key, RESULT_ROW_END_INTERP); + return true; +} + +static void saveDataBlockLastRow(SQueryRuntimeEnv* pRuntimeEnv, SDataBlockInfo* pDataBlockInfo, SArray* pDataBlock) { + if (pDataBlock == NULL) { + return; + } + + SQuery* pQuery = pRuntimeEnv->pQuery; + int32_t rowIndex = QUERY_IS_ASC_QUERY(pQuery)? pDataBlockInfo->rows-1:0; + for (int32_t k = 0; k < pQuery->numOfCols; ++k) { + SColumnInfoData *pColInfo = taosArrayGet(pDataBlock, k); + memcpy(pRuntimeEnv->prevRow[k], ((char*)pColInfo->pData) + (pColInfo->info.bytes * rowIndex), pColInfo->info.bytes); + } +} + +static TSKEY getStartTsKey(SQuery* pQuery, SDataBlockInfo* pDataBlockInfo, TSKEY* tsCols, int32_t step) { + TSKEY ts = TSKEY_INITIAL_VAL; + + if (tsCols == NULL) { + ts = QUERY_IS_ASC_QUERY(pQuery) ? pDataBlockInfo->window.skey : pDataBlockInfo->window.ekey; + } else { + int32_t offset = GET_COL_DATA_POS(pQuery, 0, step); + ts = tsCols[offset]; + } + + return ts; +} + /** * todo set the last value for pQueryTableInfo as in rowwiseapplyfunctions * @param pRuntimeEnv @@ -1000,16 +1156,15 @@ static char *getDataBlock(SQueryRuntimeEnv *pRuntimeEnv, SArithmeticSupport *sas * @return the incremental number of output value, so it maybe 0 for fixed number of query, * such as count/min/max etc. */ -static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pStatis, - SDataBlockInfo *pDataBlockInfo, SWindowResInfo *pWindowResInfo, - __block_search_fn_t searchFn, SArray *pDataBlock) { +static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pStatis, SDataBlockInfo *pDataBlockInfo, + SResultRowInfo *pWindowResInfo, __block_search_fn_t searchFn, SArray *pDataBlock) { SQLFunctionCtx *pCtx = pRuntimeEnv->pCtx; - bool masterScan = IS_MASTER_SCAN(pRuntimeEnv); + bool masterScan = IS_MASTER_SCAN(pRuntimeEnv); SQuery *pQuery = pRuntimeEnv->pQuery; TSKEY *tsCols = NULL; if (pDataBlock != NULL) { - SColumnInfoData* pColInfo = taosArrayGet(pDataBlock, 0); + SColumnInfoData *pColInfo = taosArrayGet(pDataBlock, 0); tsCols = (TSKEY *)(pColInfo->pData); } @@ -1018,7 +1173,7 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_OUT_OF_MEMORY); } - SQInfo* pQInfo = GET_QINFO_ADDR(pRuntimeEnv); + SQInfo *pQInfo = GET_QINFO_ADDR(pRuntimeEnv); for (int32_t k = 0; k < pQuery->numOfOutput; ++k) { char *dataBlock = getDataBlock(pRuntimeEnv, &sasArray[k], k, pDataBlockInfo->rows, pDataBlock); setExecParams(pQuery, &pCtx[k], dataBlock, tsCols, pDataBlockInfo, pStatis, &sasArray[k], k, pQInfo->vgId); @@ -1026,18 +1181,15 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order); if (QUERY_IS_INTERVAL_QUERY(pQuery)) { - TSKEY ts = TSKEY_INITIAL_VAL; - - if (tsCols == NULL) { - ts = QUERY_IS_ASC_QUERY(pQuery)? pDataBlockInfo->window.skey:pDataBlockInfo->window.ekey; - } else { - int32_t offset = GET_COL_DATA_POS(pQuery, 0, step); - ts = tsCols[offset]; - } + int32_t prevIndex = curTimeWindowIndex(pWindowResInfo); - bool hasTimeWindow = false; + TSKEY ts = getStartTsKey(pQuery, pDataBlockInfo, tsCols, step); STimeWindow win = getActiveTimeWindow(pWindowResInfo, ts, pQuery); - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { + + bool hasTimeWindow = false; + SResultRow* pResult = NULL; + int32_t ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow, &pResult); + if (ret != TSDB_CODE_SUCCESS) { tfree(sasArray); return; } @@ -1045,11 +1197,59 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * int32_t forwardStep = 0; int32_t startPos = pQuery->pos; + // in case of repeat scan/reverse scan, no new time window added. if (hasTimeWindow) { TSKEY ekey = reviseWindowEkey(pQuery, &win); forwardStep = getNumOfRowsInTimeWindow(pQuery, pDataBlockInfo, tsCols, pQuery->pos, ekey, searchFn, true); - bool pStatus = getTimeWindowResStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); + // prev time window not interpolation yet. + int32_t curIndex = curTimeWindowIndex(pWindowResInfo); + if (prevIndex != -1 && prevIndex < curIndex && pRuntimeEnv->timeWindowInterpo) { + for(int32_t j = prevIndex; j < curIndex; ++j) { + SResultRow *pRes = pWindowResInfo->pResult[j]; + + STimeWindow w = pRes->win; + ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &w, masterScan, &hasTimeWindow, &pResult); + assert(ret == TSDB_CODE_SUCCESS && !resultRowInterpolated(pResult, RESULT_ROW_END_INTERP)); + + int32_t p = QUERY_IS_ASC_QUERY(pQuery)? 0:pDataBlockInfo->rows-1; + doRowwiseTimeWindowInterpolation(pRuntimeEnv, pDataBlock, *(TSKEY*) pRuntimeEnv->prevRow[0], -1, tsCols[0], p, w.ekey, RESULT_ROW_END_INTERP); + setResultRowInterpo(pResult, RESULT_ROW_END_INTERP); + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_START_INTERP); + + bool closed = getResultRowStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); + doBlockwiseApplyFunctions(pRuntimeEnv, closed, &w, startPos, 0, tsCols, pDataBlockInfo->rows); + } + + // restore current time window + ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow, &pResult); + assert (ret == TSDB_CODE_SUCCESS); // null data, too many state code + } + + // window start key interpolation + if (pRuntimeEnv->timeWindowInterpo) { + bool done = resultRowInterpolated(pResult, RESULT_ROW_START_INTERP); + if (!done) { + int32_t startRowIndex = pQuery->pos; + bool interp = setTimeWindowInterpolationStartTs(pRuntimeEnv, startRowIndex, pDataBlockInfo->rows, pDataBlock, tsCols, &win); + if (interp) { + setResultRowInterpo(pResult, RESULT_ROW_START_INTERP); + } + } + + done = resultRowInterpolated(pResult, RESULT_ROW_END_INTERP); + if (!done) { + int32_t endRowIndex = pQuery->pos + (forwardStep - 1) * step; + + TSKEY endKey = QUERY_IS_ASC_QUERY(pQuery)? pDataBlockInfo->window.ekey:pDataBlockInfo->window.skey; + bool interp = setTimeWindowInterpolationEndTs(pRuntimeEnv, endRowIndex, pDataBlock, tsCols, endKey, &win); + if (interp) { + setResultRowInterpo(pResult, RESULT_ROW_END_INTERP); + } + } + } + + bool pStatus = getResultRowStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); doBlockwiseApplyFunctions(pRuntimeEnv, pStatus, &win, startPos, forwardStep, tsCols, pDataBlockInfo->rows); } @@ -1065,7 +1265,8 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * // null data, failed to allocate more memory buffer hasTimeWindow = false; - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &nextWin, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { + if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &nextWin, masterScan, &hasTimeWindow, &pResult) != + TSDB_CODE_SUCCESS) { break; } @@ -1076,7 +1277,29 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * TSKEY ekey = reviseWindowEkey(pQuery, &nextWin); forwardStep = getNumOfRowsInTimeWindow(pQuery, pDataBlockInfo, tsCols, startPos, ekey, searchFn, true); - bool closed = getTimeWindowResStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); + // window start(end) key interpolation + if (pRuntimeEnv->timeWindowInterpo) { + bool done = resultRowInterpolated(pResult, RESULT_ROW_START_INTERP); + if (!done) { + int32_t startRowIndex = startPos; + bool interp = setTimeWindowInterpolationStartTs(pRuntimeEnv, startRowIndex, pDataBlockInfo->rows, pDataBlock, tsCols, &nextWin); + if (interp) { + setResultRowInterpo(pResult, RESULT_ROW_START_INTERP); + } + } + + done = resultRowInterpolated(pResult, RESULT_ROW_END_INTERP); + if (!done) { + int32_t endRowIndex = startPos + (forwardStep - 1)*step; + TSKEY endKey = QUERY_IS_ASC_QUERY(pQuery)? pDataBlockInfo->window.ekey:pDataBlockInfo->window.skey; + bool interp = setTimeWindowInterpolationEndTs(pRuntimeEnv, endRowIndex, pDataBlock, tsCols, endKey, &nextWin); + if (interp) { + setResultRowInterpo(pResult, RESULT_ROW_END_INTERP); + } + } + } + + bool closed = getResultRowStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); doBlockwiseApplyFunctions(pRuntimeEnv, closed, &nextWin, startPos, forwardStep, tsCols, pDataBlockInfo->rows); } @@ -1090,12 +1313,17 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * for (int32_t k = 0; k < pQuery->numOfOutput; ++k) { int32_t functionId = pQuery->pExpr1[k].base.functionId; if (functionNeedToExecute(pRuntimeEnv, &pCtx[k], functionId)) { + pCtx[k].nStartQueryTimestamp = pDataBlockInfo->window.skey; aAggs[functionId].xFunction(&pCtx[k]); } } } - for(int32_t i = 0; i < pQuery->numOfOutput; ++i) { + if (pRuntimeEnv->timeWindowInterpo) { + saveDataBlockLastRow(pRuntimeEnv, pDataBlockInfo, pDataBlock); + } + + for (int32_t i = 0; i < pQuery->numOfOutput; ++i) { if (pQuery->pExpr1[i].base.functionId != TSDB_FUNC_ARITHM) { continue; } @@ -1144,8 +1372,12 @@ static int32_t setGroupResultOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, char *pDat } if (type == TSDB_DATA_TYPE_BINARY || type == TSDB_DATA_TYPE_NCHAR) { - pResultRow->key = malloc(varDataTLen(pData)); - varDataCopy(pResultRow->key, pData); + if (pResultRow->key == NULL) { + pResultRow->key = malloc(varDataTLen(pData)); + varDataCopy(pResultRow->key, pData); + } else { + assert(memcmp(pResultRow->key, pData, varDataTLen(pData)) == 0); + } } else { pResultRow->win.skey = v; pResultRow->win.ekey = v; @@ -1270,8 +1502,84 @@ static bool functionNeedToExecute(SQueryRuntimeEnv *pRuntimeEnv, SQLFunctionCtx return true; } +void doRowwiseTimeWindowInterpolation(SQueryRuntimeEnv* pRuntimeEnv, SArray* pDataBlock, TSKEY prevTs, int32_t prevRowIndex, TSKEY curTs, int32_t curRowIndex, TSKEY windowKey, int32_t type) { + SQuery* pQuery = pRuntimeEnv->pQuery; + for (int32_t k = 0; k < pQuery->numOfOutput; ++k) { + int32_t functionId = pQuery->pExpr1[k].base.functionId; + if (functionId != TSDB_FUNC_TWA) { + pRuntimeEnv->pCtx[k].start.key = INT64_MIN; + continue; + } + + SColIndex* pColIndex = &pQuery->pExpr1[k].base.colInfo; + int16_t index = pColIndex->colIndex; + SColumnInfoData* pColInfo = taosArrayGet(pDataBlock, index); + + assert(pColInfo->info.colId == pColIndex->colId && curTs != windowKey); + double v1 = 0, v2 = 0, v = 0; + + if (prevRowIndex == -1) { + GET_TYPED_DATA(v1, double, pColInfo->info.type, (char *)pRuntimeEnv->prevRow[k]); + } else { + GET_TYPED_DATA(v1, double, pColInfo->info.type, (char *)pColInfo->pData + prevRowIndex * pColInfo->info.bytes); + } + + GET_TYPED_DATA(v2, double, pColInfo->info.type, (char *)pColInfo->pData + curRowIndex * pColInfo->info.bytes); + + SPoint point1 = (SPoint){.key = prevTs, .val = &v1}; + SPoint point2 = (SPoint){.key = curTs, .val = &v2}; + SPoint point = (SPoint){.key = windowKey, .val = &v}; + taosGetLinearInterpolationVal(TSDB_DATA_TYPE_DOUBLE, &point1, &point2, &point); + + if (type == RESULT_ROW_START_INTERP) { + pRuntimeEnv->pCtx[k].start.key = point.key; + pRuntimeEnv->pCtx[k].start.val = v; + } else { + pRuntimeEnv->pCtx[k].end.key = point.key; + pRuntimeEnv->pCtx[k].end.val = v; + } + } +} + +static void setTimeWindowSKeyInterp(SQueryRuntimeEnv* pRuntimeEnv, SArray* pDataBlock, TSKEY prevTs, int32_t prevRowIndex, TSKEY ts, int32_t offset, SResultRow* pResult, STimeWindow* win) { + SQuery* pQuery = pRuntimeEnv->pQuery; + + bool done = resultRowInterpolated(pResult, RESULT_ROW_START_INTERP); + if (!done) { + TSKEY key = QUERY_IS_ASC_QUERY(pQuery)? win->skey:win->ekey; + if (key == ts) { + setResultRowInterpo(pResult, RESULT_ROW_START_INTERP); + } else if (prevTs != INT64_MIN && ((QUERY_IS_ASC_QUERY(pQuery) && prevTs < key) || (!QUERY_IS_ASC_QUERY(pQuery) && prevTs > key))) { + doRowwiseTimeWindowInterpolation(pRuntimeEnv, pDataBlock, prevTs, prevRowIndex, ts, offset, key, RESULT_ROW_START_INTERP); + setResultRowInterpo(pResult, RESULT_ROW_START_INTERP); + } else { + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_START_INTERP); + } + + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_END_INTERP); + for (int32_t k = 0; k < pQuery->numOfOutput; ++k) { + pRuntimeEnv->pCtx[k].size = 1; + } + } else { + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_START_INTERP); + } +} + +static void setTimeWindowEKeyInterp(SQueryRuntimeEnv* pRuntimeEnv, SArray* pDataBlock, TSKEY prevTs, int32_t prevRowIndex, TSKEY ts, int32_t offset, SResultRow* pResult, STimeWindow* win) { + SQuery* pQuery = pRuntimeEnv->pQuery; + + TSKEY key = QUERY_IS_ASC_QUERY(pQuery)? win->ekey:win->skey; + doRowwiseTimeWindowInterpolation(pRuntimeEnv, pDataBlock, prevTs, prevRowIndex, ts, offset, key, RESULT_ROW_END_INTERP); + setResultRowInterpo(pResult, RESULT_ROW_END_INTERP); + + setNotInterpoWindowKey(pRuntimeEnv->pCtx, pQuery->numOfOutput, RESULT_ROW_START_INTERP); + for (int32_t i = 0; i < pQuery->numOfOutput; ++i) { + pRuntimeEnv->pCtx[i].size = 0; + } +} + static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pStatis, SDataBlockInfo *pDataBlockInfo, - SWindowResInfo *pWindowResInfo, SArray *pDataBlock) { + SResultRowInfo *pWindowResInfo, SArray *pDataBlock) { SQLFunctionCtx *pCtx = pRuntimeEnv->pCtx; bool masterScan = IS_MASTER_SCAN(pRuntimeEnv); @@ -1300,6 +1608,7 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS for (int32_t k = 0; k < pQuery->numOfOutput; ++k) { char *dataBlock = getDataBlock(pRuntimeEnv, &sasArray[k], k, pDataBlockInfo->rows, pDataBlock); setExecParams(pQuery, &pCtx[k], dataBlock, tsCols, pDataBlockInfo, pStatis, &sasArray[k], k, pQInfo->vgId); + pCtx[k].size = 1; } // set the input column data @@ -1318,20 +1627,21 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS pQuery->order.order, pRuntimeEnv->pTSBuf->cur.order); } - int32_t j = 0; int32_t offset = -1; + TSKEY prevTs = *(TSKEY*) pRuntimeEnv->prevRow[0]; + int32_t prevRowIndex = -1; - for (j = 0; j < pDataBlockInfo->rows; ++j) { + for (int32_t j = 0; j < pDataBlockInfo->rows; ++j) { offset = GET_COL_DATA_POS(pQuery, j, step); if (pRuntimeEnv->pTSBuf != NULL) { - int32_t r = doTSJoinFilter(pRuntimeEnv, offset); - if (r == TS_JOIN_TAG_NOT_EQUALS) { + int32_t ret = doTSJoinFilter(pRuntimeEnv, offset); + if (ret == TS_JOIN_TAG_NOT_EQUALS) { break; - } else if (r == TS_JOIN_TS_NOT_EQUALS) { + } else if (ret == TS_JOIN_TS_NOT_EQUALS) { continue; } else { - assert(r == TS_JOIN_TS_EQUAL); + assert(ret == TS_JOIN_TS_EQUAL); } } @@ -1341,11 +1651,14 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS // interval window query, decide the time window according to the primary timestamp if (QUERY_IS_INTERVAL_QUERY(pQuery)) { - int64_t ts = tsCols[offset]; + int32_t prevWindowIndex = curTimeWindowIndex(pWindowResInfo); + int64_t ts = tsCols[offset]; + STimeWindow win = getActiveTimeWindow(pWindowResInfo, ts, pQuery); - bool hasTimeWindow = false; - int32_t ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow); + bool hasTimeWindow = false; + SResultRow* pResult = NULL; + int32_t ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow, &pResult); if (ret != TSDB_CODE_SUCCESS) { // null data, too many state code continue; } @@ -1354,7 +1667,35 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS continue; } - bool closed = getTimeWindowResStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); + // window start key interpolation + if (pRuntimeEnv->timeWindowInterpo) { + // check for the time window end time interpolation + int32_t curIndex = curTimeWindowIndex(pWindowResInfo); + if (prevWindowIndex != -1 && prevWindowIndex < curIndex) { + for (int32_t k = prevWindowIndex; k < curIndex; ++k) { + SResultRow *pRes = pWindowResInfo->pResult[k]; + + ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &pRes->win, masterScan, &hasTimeWindow, &pResult); + assert(ret == TSDB_CODE_SUCCESS && !resultRowInterpolated(pResult, RESULT_ROW_END_INTERP)); + + setTimeWindowEKeyInterp(pRuntimeEnv, pDataBlock, prevTs, prevRowIndex, ts, offset, pResult, &pRes->win); + + bool closed = getResultRowStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); + doRowwiseApplyFunctions(pRuntimeEnv, closed, &pRes->win, offset); + } + + // restore current time window + ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow, + &pResult); + if (ret != TSDB_CODE_SUCCESS) { // null data, too many state code + continue; + } + } + + setTimeWindowSKeyInterp(pRuntimeEnv, pDataBlock, prevTs, prevRowIndex, ts, offset, pResult, &win); + } + + bool closed = getResultRowStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); doRowwiseApplyFunctions(pRuntimeEnv, closed, &win, offset); STimeWindow nextWin = win; @@ -1373,12 +1714,13 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS // null data, failed to allocate more memory buffer hasTimeWindow = false; - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &nextWin, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { + if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &nextWin, masterScan, &hasTimeWindow, &pResult) != TSDB_CODE_SUCCESS) { break; } if (hasTimeWindow) { - closed = getTimeWindowResStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); + setTimeWindowSKeyInterp(pRuntimeEnv, pDataBlock, prevTs, prevRowIndex, ts, offset, pResult, &nextWin); + closed = getResultRowStatus(pWindowResInfo, curTimeWindowIndex(pWindowResInfo)); doRowwiseApplyFunctions(pRuntimeEnv, closed, &nextWin, offset); } } @@ -1403,6 +1745,9 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS } } + prevTs = tsCols[offset]; + prevRowIndex = offset; + if (pRuntimeEnv->pTSBuf != NULL) { // if timestamp filter list is empty, quit current query if (!tsBufNextPos(pRuntimeEnv->pTSBuf)) { @@ -1440,7 +1785,7 @@ static int32_t tableApplyFunctionsOnBlock(SQueryRuntimeEnv *pRuntimeEnv, SDataBl SQuery *pQuery = pRuntimeEnv->pQuery; STableQueryInfo* pTableQInfo = pQuery->current; - SWindowResInfo* pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo* pWindowResInfo = &pRuntimeEnv->windowResInfo; if (pQuery->numOfFilterCols > 0 || pRuntimeEnv->pTSBuf != NULL || pRuntimeEnv->groupbyNormalCol) { rowwiseApplyFunctions(pRuntimeEnv, pStatis, pDataBlockInfo, pWindowResInfo, pDataBlock); @@ -1528,10 +1873,10 @@ void setExecParams(SQuery *pQuery, SQLFunctionCtx *pCtx, void* inputData, TSKEY * top/bottom values emerge, so does diff function */ if (functionId == TSDB_FUNC_TWA) { - SResultRowCellInfo* pInfo = GET_RES_INFO(pCtx); - STwaInfo *pTWAInfo = (STwaInfo*) GET_ROWCELL_INTERBUF(pInfo); - pTWAInfo->SKey = pQuery->window.skey; - pTWAInfo->EKey = pQuery->window.ekey; + pCtx->param[1].i64Key = pQuery->window.skey; + pCtx->param[1].nType = TSDB_DATA_TYPE_BIGINT; + pCtx->param[2].i64Key = pQuery->window.ekey; + pCtx->param[2].nType = TSDB_DATA_TYPE_BIGINT; } } else if (functionId == TSDB_FUNC_ARITHM) { @@ -1677,6 +2022,8 @@ static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int16_t order pCtx->functionId = pSqlFuncMsg->functionId; pCtx->stableQuery = pRuntimeEnv->stableQuery; pCtx->interBufBytes = pQuery->pExpr1[i].interBytes; + pCtx->start.key = INT64_MIN; + pCtx->end.key = INT64_MIN; pCtx->numOfParams = pSqlFuncMsg->numOfParams; for (int32_t j = 0; j < pCtx->numOfParams; ++j) { @@ -1711,6 +2058,8 @@ static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int16_t order } + *(int64_t*) pRuntimeEnv->prevRow[0] = INT64_MIN; + // if it is group by normal column, do not set output buffer, the output buffer is pResult // fixed output query/multi-output query for normal table if (!pRuntimeEnv->groupbyNormalCol && !pRuntimeEnv->stableQuery && !QUERY_IS_INTERVAL_QUERY(pRuntimeEnv->pQuery)) { @@ -1781,6 +2130,7 @@ static void teardownQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv) { tfree(pRuntimeEnv->offset); tfree(pRuntimeEnv->keyBuf); tfree(pRuntimeEnv->rowCellInfoOffset); + tfree(pRuntimeEnv->prevRow); taosHashCleanup(pRuntimeEnv->pResultRowHashTable); pRuntimeEnv->pResultRowHashTable = NULL; @@ -2259,7 +2609,7 @@ static bool overlapWithTimeWindow(SQuery* pQuery, SDataBlockInfo* pBlockInfo) { return false; } -int32_t loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo * pWindowResInfo, void* pQueryHandle, SDataBlockInfo* pBlockInfo, SDataStatis **pStatis, SArray** pDataBlock, uint32_t* status) { +int32_t loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, SResultRowInfo * pWindowResInfo, void* pQueryHandle, SDataBlockInfo* pBlockInfo, SDataStatis **pStatis, SArray** pDataBlock, uint32_t* status) { SQuery *pQuery = pRuntimeEnv->pQuery; *status = BLK_DATA_NO_NEEDED; @@ -2279,12 +2629,14 @@ int32_t loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo * pW // the filter result may be incorrect. So in case of interval query, we need to set the correct time output buffer if (QUERY_IS_INTERVAL_QUERY(pQuery)) { bool hasTimeWindow = false; + SResultRow* pResult = NULL; + bool masterScan = IS_MASTER_SCAN(pRuntimeEnv); TSKEY k = QUERY_IS_ASC_QUERY(pQuery)? pBlockInfo->window.skey:pBlockInfo->window.ekey; STimeWindow win = getActiveTimeWindow(pWindowResInfo, k, pQuery); - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pBlockInfo, &win, masterScan, &hasTimeWindow) != + if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pBlockInfo, &win, masterScan, &hasTimeWindow, &pResult) != TSDB_CODE_SUCCESS) { // todo handle error in set result for timewindow } @@ -2435,6 +2787,7 @@ static void ensureOutputBufferSimple(SQueryRuntimeEnv* pRuntimeEnv, int32_t capa pQuery->rec.capacity = capacity; } +// TODO merge with enuserOutputBufferSimple static void ensureOutputBuffer(SQueryRuntimeEnv* pRuntimeEnv, SDataBlockInfo* pBlockInfo) { // in case of prj/diff query, ensure the output buffer is sufficient to accommodate the results of current block SQuery* pQuery = pRuntimeEnv->pQuery; @@ -2479,7 +2832,7 @@ static void doSetInitialTimewindow(SQueryRuntimeEnv* pRuntimeEnv, SDataBlockInfo if (QUERY_IS_INTERVAL_QUERY(pQuery) && pRuntimeEnv->windowResInfo.prevSKey == TSKEY_INITIAL_VAL) { STimeWindow w = TSWINDOW_INITIALIZER; - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; if (QUERY_IS_ASC_QUERY(pQuery)) { getAlignQueryTimeWindow(pQuery, pBlockInfo->window.skey, pBlockInfo->window.skey, pQuery->window.ekey, &w); @@ -2843,14 +3196,14 @@ int32_t tableResultComparFn(const void *pLeft, const void *pRight, void *param) return -1; } - SWindowResInfo *pWindowResInfo1 = &supporter->pTableQueryInfo[left]->windowResInfo; + SResultRowInfo *pWindowResInfo1 = &supporter->pTableQueryInfo[left]->windowResInfo; SResultRow * pWindowRes1 = getResultRow(pWindowResInfo1, leftPos); tFilePage *page1 = getResBufPage(pRuntimeEnv->pResultBuf, pWindowRes1->pageId); char *b1 = getPosInResultPage(pRuntimeEnv, PRIMARYKEY_TIMESTAMP_COL_INDEX, pWindowRes1, page1); TSKEY leftTimestamp = GET_INT64_VAL(b1); - SWindowResInfo *pWindowResInfo2 = &supporter->pTableQueryInfo[right]->windowResInfo; + SResultRowInfo *pWindowResInfo2 = &supporter->pTableQueryInfo[right]->windowResInfo; SResultRow * pWindowRes2 = getResultRow(pWindowResInfo2, rightPos); tFilePage *page2 = getResBufPage(pRuntimeEnv->pResultBuf, pWindowRes2->pageId); @@ -3090,7 +3443,7 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) { int32_t pos = pTree->pNode[0].index; - SWindowResInfo *pWindowResInfo = &pTableList[pos]->windowResInfo; + SResultRowInfo *pWindowResInfo = &pTableList[pos]->windowResInfo; SResultRow *pWindowRes = getResultRow(pWindowResInfo, cs.position[pos]); tFilePage *page = getResBufPage(pRuntimeEnv->pResultBuf, pWindowRes->pageId); @@ -3238,17 +3591,9 @@ static void updateTableQueryInfoForReverseScan(SQuery *pQuery, STableQueryInfo * // order has changed already int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order); - - // TODO validate the assertion -// if (!QUERY_IS_ASC_QUERY(pQuery)) { -// assert(pTableQueryInfo->win.ekey >= pTableQueryInfo->lastKey + step); -// } else { -// assert(pTableQueryInfo->win.ekey <= pTableQueryInfo->lastKey + step); -// } - if (pTableQueryInfo->lastKey == pTableQueryInfo->win.skey) { // do nothing, no results - } else { + } else {// NOTE: even win.skey != lastKey, the results may not generated. pTableQueryInfo->win.ekey = pTableQueryInfo->lastKey + step; } @@ -3262,11 +3607,11 @@ static void updateTableQueryInfoForReverseScan(SQuery *pQuery, STableQueryInfo * pTableQueryInfo->windowResInfo.curIndex = pTableQueryInfo->windowResInfo.size - 1; } -static void disableFuncInReverseScanImpl(SQueryRuntimeEnv* pRuntimeEnv, SWindowResInfo *pWindowResInfo, int32_t order) { +static void disableFuncInReverseScanImpl(SQueryRuntimeEnv* pRuntimeEnv, SResultRowInfo *pWindowResInfo, int32_t order) { SQuery* pQuery = pRuntimeEnv->pQuery; for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - bool closed = getTimeWindowResStatus(pWindowResInfo, i); + bool closed = getResultRowStatus(pWindowResInfo, i); if (!closed) { continue; } @@ -3294,7 +3639,7 @@ void disableFuncInReverseScan(SQInfo *pQInfo) { int32_t order = pQuery->order.order; // group by normal columns and interval query on normal table - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; if (pRuntimeEnv->groupbyNormalCol || QUERY_IS_INTERVAL_QUERY(pQuery)) { disableFuncInReverseScanImpl(pRuntimeEnv, pWindowResInfo, order); } else { // for simple result of table query, @@ -3485,7 +3830,7 @@ bool needScanDataBlocksAgain(SQueryRuntimeEnv *pRuntimeEnv) { bool toContinue = false; if (pRuntimeEnv->groupbyNormalCol || QUERY_IS_INTERVAL_QUERY(pQuery)) { // for each group result, call the finalize function for each column - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; for (int32_t i = 0; i < pWindowResInfo->size; ++i) { SResultRow *pResult = getResultRow(pWindowResInfo, i); @@ -3569,14 +3914,7 @@ static void setEnvBeforeReverseScan(SQueryRuntimeEnv *pRuntimeEnv, SQueryStatusI } SET_REVERSE_SCAN_FLAG(pRuntimeEnv); - - STsdbQueryCond cond = { - .order = pQuery->order.order, - .colList = pQuery->colList, - .numOfCols = pQuery->numOfCols, - }; - - TIME_WINDOW_COPY(cond.twindow, pQuery->window); + STsdbQueryCond cond = createTsdbQueryCond(pQuery, &pQuery->window); setQueryStatus(pQuery, QUERY_NOT_COMPLETED); switchCtxOrder(pRuntimeEnv); @@ -3645,6 +3983,8 @@ void scanOneTableDataBlocks(SQueryRuntimeEnv *pRuntimeEnv, TSKEY start) { // do nothing if no data blocks are found qualified during scan if (qstatus.lastKey != pTableQueryInfo->lastKey) { qstatus.curWindow.ekey = pTableQueryInfo->lastKey - step; + } else { // the lastkey does not increase, which means no data checked yet + qDebug("QInfo:%p no results generated in this scan", pQInfo); } qstatus.lastKey = pTableQueryInfo->lastKey; @@ -3659,18 +3999,11 @@ void scanOneTableDataBlocks(SQueryRuntimeEnv *pRuntimeEnv, TSKEY start) { break; } - STsdbQueryCond cond = { - .order = pQuery->order.order, - .colList = pQuery->colList, - .numOfCols = pQuery->numOfCols, - }; - - TIME_WINDOW_COPY(cond.twindow, qstatus.curWindow); - if (pRuntimeEnv->pSecQueryHandle != NULL) { tsdbCleanupQueryHandle(pRuntimeEnv->pSecQueryHandle); } + STsdbQueryCond cond = createTsdbQueryCond(pQuery, &qstatus.curWindow); restoreTimeWindow(&pQInfo->tableGroupInfo, &cond); pRuntimeEnv->pSecQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &pQInfo->tableGroupInfo, pQInfo, &pQInfo->memRef); if (pRuntimeEnv->pSecQueryHandle == NULL) { @@ -3708,7 +4041,7 @@ void finalizeQueryResult(SQueryRuntimeEnv *pRuntimeEnv) { if (pRuntimeEnv->groupbyNormalCol || QUERY_IS_INTERVAL_QUERY(pQuery)) { // for each group result, call the finalize function for each column - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; if (pRuntimeEnv->groupbyNormalCol) { closeAllTimeWindow(pWindowResInfo); } @@ -3764,9 +4097,8 @@ static STableQueryInfo *createTableQueryInfo(SQueryRuntimeEnv *pRuntimeEnv, void // set more initial size of interval/groupby query if (QUERY_IS_INTERVAL_QUERY(pQuery) || pRuntimeEnv->groupbyNormalCol) { - int32_t initialSize = 16; - int32_t initialThreshold = 100; - int32_t code = initWindowResInfo(&pTableQueryInfo->windowResInfo, initialSize, initialThreshold, TSDB_DATA_TYPE_INT); + int32_t initialSize = 128; + int32_t code = initWindowResInfo(&pTableQueryInfo->windowResInfo, initialSize, TSDB_DATA_TYPE_INT); if (code != TSDB_CODE_SUCCESS) { return NULL; } @@ -3793,7 +4125,7 @@ void destroyTableQueryInfoImpl(STableQueryInfo *pTableQueryInfo) { void setExecutionContext(SQInfo *pQInfo, int32_t groupIndex, TSKEY nextKey) { SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv; STableQueryInfo *pTableQueryInfo = pRuntimeEnv->pQuery->current; - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; // lastKey needs to be updated pTableQueryInfo->lastKey = nextKey; @@ -3961,7 +4293,7 @@ void setIntervalQueryRange(SQInfo *pQInfo, TSKEY key) { * operations involve. */ STimeWindow w = TSWINDOW_INITIALIZER; - SWindowResInfo *pWindowResInfo = &pTableQueryInfo->windowResInfo; + SResultRowInfo *pWindowResInfo = &pTableQueryInfo->windowResInfo; TSKEY sk = MIN(win.skey, win.ekey); TSKEY ek = MAX(win.skey, win.ekey); @@ -4005,7 +4337,7 @@ bool needPrimaryTimestampCol(SQuery *pQuery, SDataBlockInfo *pDataBlockInfo) { return loadPrimaryTS; } -static int32_t doCopyToSData(SQInfo *pQInfo, SWindowResInfo *pResultInfo, int32_t orderType) { +static int32_t doCopyToSData(SQInfo *pQInfo, SResultRowInfo *pResultInfo, int32_t orderType) { SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv; SQuery * pQuery = pRuntimeEnv->pQuery; @@ -4082,7 +4414,7 @@ static int32_t doCopyToSData(SQInfo *pQInfo, SWindowResInfo *pResultInfo, int32_ * @param pQInfo * @param result */ -void copyFromWindowResToSData(SQInfo *pQInfo, SWindowResInfo *pResultInfo) { +void copyFromWindowResToSData(SQInfo *pQInfo, SResultRowInfo *pResultInfo) { SQuery *pQuery = pQInfo->runtimeEnv.pQuery; int32_t orderType = (pQuery->pGroupbyExpr != NULL) ? pQuery->pGroupbyExpr->orderType : TSDB_ORDER_ASC; @@ -4121,7 +4453,7 @@ static void stableApplyFunctionsOnBlock(SQueryRuntimeEnv *pRuntimeEnv, SDataBloc SQuery * pQuery = pRuntimeEnv->pQuery; STableQueryInfo* pTableQueryInfo = pQuery->current; - SWindowResInfo * pWindowResInfo = &pTableQueryInfo->windowResInfo; + SResultRowInfo * pWindowResInfo = &pTableQueryInfo->windowResInfo; pQuery->pos = QUERY_IS_ASC_QUERY(pQuery)? 0 : pDataBlockInfo->rows - 1; if (pQuery->numOfFilterCols > 0 || pRuntimeEnv->pTSBuf != NULL || pRuntimeEnv->groupbyNormalCol) { @@ -4195,16 +4527,19 @@ static void doCopyQueryResultToMsg(SQInfo *pQInfo, int32_t numOfRows, char *data } } - int32_t numOfTables = (int32_t)taosArrayGetSize(pQInfo->arrTableIdInfo); + int32_t numOfTables = (int32_t) taosHashGetSize(pQInfo->arrTableIdInfo); *(int32_t*)data = htonl(numOfTables); data += sizeof(int32_t); - for(int32_t i = 0; i < numOfTables; i++) { - STableIdInfo* pSrc = taosArrayGet(pQInfo->arrTableIdInfo, i); + + STableIdInfo* item = taosHashIterate(pQInfo->arrTableIdInfo, NULL); + while(item) { STableIdInfo* pDst = (STableIdInfo*)data; - pDst->uid = htobe64(pSrc->uid); - pDst->tid = htonl(pSrc->tid); - pDst->key = htobe64(pSrc->key); + pDst->uid = htobe64(item->uid); + pDst->tid = htonl(item->tid); + pDst->key = htobe64(item->key); + data += sizeof(STableIdInfo); + item = taosHashIterate(pQInfo->arrTableIdInfo, item); } // Check if query is completed or not for stable query or normal table query respectively. @@ -4366,7 +4701,7 @@ void skipBlocks(SQueryRuntimeEnv *pRuntimeEnv) { static TSKEY doSkipIntervalProcess(SQueryRuntimeEnv* pRuntimeEnv, STimeWindow* win, SDataBlockInfo* pBlockInfo, STableQueryInfo* pTableQueryInfo) { SQuery *pQuery = pRuntimeEnv->pQuery; - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; assert(pQuery->limit.offset == 0); STimeWindow tw = *win; @@ -4416,7 +4751,23 @@ static TSKEY doSkipIntervalProcess(SQueryRuntimeEnv* pRuntimeEnv, STimeWindow* w static bool skipTimeInterval(SQueryRuntimeEnv *pRuntimeEnv, TSKEY* start) { SQuery *pQuery = pRuntimeEnv->pQuery; - *start = pQuery->current->lastKey; + + // get the first unclosed time window + bool assign = false; + for(int32_t i = 0; i < pRuntimeEnv->windowResInfo.size; ++i) { + if (pRuntimeEnv->windowResInfo.pResult[i]->closed) { + continue; + } + + assign = true; + *start = pRuntimeEnv->windowResInfo.pResult[i]->win.skey; + } + + if (!assign) { + *start = pQuery->current->lastKey; + } + + assert(*start <= pQuery->current->lastKey); // if queried with value filter, do NOT forward query start position if (pQuery->limit.offset <= 0 || pQuery->numOfFilterCols > 0 || pRuntimeEnv->pTSBuf != NULL || pRuntimeEnv->pFillInfo != NULL) { @@ -4432,7 +4783,7 @@ static bool skipTimeInterval(SQueryRuntimeEnv *pRuntimeEnv, TSKEY* start) { STimeWindow w = TSWINDOW_INITIALIZER; - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; STableQueryInfo *pTableQueryInfo = pQuery->current; SDataBlockInfo blockInfo = SDATA_BLOCK_INITIALIZER; @@ -4531,13 +4882,7 @@ static int32_t setupQueryHandle(void* tsdb, SQInfo* pQInfo, bool isSTableQuery) return TSDB_CODE_SUCCESS; } - STsdbQueryCond cond = { - .order = pQuery->order.order, - .colList = pQuery->colList, - .numOfCols = pQuery->numOfCols, - }; - - TIME_WINDOW_COPY(cond.twindow, pQuery->window); + STsdbQueryCond cond = createTsdbQueryCond(pQuery, &pQuery->window); if (!isSTableQuery && (pQInfo->tableqinfoGroupInfo.numOfTables == 1) @@ -4617,6 +4962,7 @@ int32_t doInitQInfo(SQInfo *pQInfo, STSBuf *pTsBuf, void *tsdb, int32_t vgId, bo pRuntimeEnv->topBotQuery = isTopBottomQuery(pQuery); pRuntimeEnv->hasTagResults = hasTagValOutput(pQuery); + pRuntimeEnv->timeWindowInterpo = timeWindowInterpoRequired(pQuery); setScanLimitationByResultBuffer(pQuery); @@ -4653,20 +4999,13 @@ int32_t doInitQInfo(SQInfo *pQInfo, STSBuf *pTsBuf, void *tsdb, int32_t vgId, bo if (!QUERY_IS_INTERVAL_QUERY(pQuery)) { int16_t type = TSDB_DATA_TYPE_NULL; - int32_t threshold = 0; - if (pRuntimeEnv->groupbyNormalCol) { // group by columns not tags; type = getGroupbyColumnType(pQuery, pQuery->pGroupbyExpr); - threshold = 4000; } else { type = TSDB_DATA_TYPE_INT; // group id - threshold = (int32_t)(GET_NUM_OF_TABLEGROUP(pQInfo)); - if (threshold < 8) { - threshold = 8; - } } - code = initWindowResInfo(&pRuntimeEnv->windowResInfo, 8, threshold, type); + code = initWindowResInfo(&pRuntimeEnv->windowResInfo, 8, type); if (code != TSDB_CODE_SUCCESS) { return code; } @@ -4686,7 +5025,7 @@ int32_t doInitQInfo(SQInfo *pQInfo, STSBuf *pTsBuf, void *tsdb, int32_t vgId, bo type = TSDB_DATA_TYPE_TIMESTAMP; } - code = initWindowResInfo(&pRuntimeEnv->windowResInfo, numOfResultRows, 1024, type); + code = initWindowResInfo(&pRuntimeEnv->windowResInfo, numOfResultRows, type); if (code != TSDB_CODE_SUCCESS) { return code; } @@ -4744,6 +5083,20 @@ static FORCE_INLINE void setEnvForEachBlock(SQInfo* pQInfo, STableQueryInfo* pTa } } +static void doTableQueryInfoTimeWindowCheck(SQuery* pQuery, STableQueryInfo* pTableQueryInfo) { + if (QUERY_IS_ASC_QUERY(pQuery)) { + assert( + (pTableQueryInfo->win.skey <= pTableQueryInfo->win.ekey) && + (pTableQueryInfo->lastKey >= pTableQueryInfo->win.skey) && + (pTableQueryInfo->win.skey >= pQuery->window.skey && pTableQueryInfo->win.ekey <= pQuery->window.ekey)); + } else { + assert( + (pTableQueryInfo->win.skey >= pTableQueryInfo->win.ekey) && + (pTableQueryInfo->lastKey <= pTableQueryInfo->win.skey) && + (pTableQueryInfo->win.skey <= pQuery->window.skey && pTableQueryInfo->win.ekey >= pQuery->window.ekey)); + } +} + static int64_t scanMultiTableDataBlocks(SQInfo *pQInfo) { SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv; SQuery* pQuery = pRuntimeEnv->pQuery; @@ -4770,17 +5123,7 @@ static int64_t scanMultiTableDataBlocks(SQInfo *pQInfo) { } pQuery->current = *pTableQueryInfo; - if (QUERY_IS_ASC_QUERY(pQuery)) { - assert( - ((*pTableQueryInfo)->win.skey <= (*pTableQueryInfo)->win.ekey) && - ((*pTableQueryInfo)->lastKey >= (*pTableQueryInfo)->win.skey) && - ((*pTableQueryInfo)->win.skey >= pQuery->window.skey && (*pTableQueryInfo)->win.ekey <= pQuery->window.ekey)); - } else { - assert( - ((*pTableQueryInfo)->win.skey >= (*pTableQueryInfo)->win.ekey) && - ((*pTableQueryInfo)->lastKey <= (*pTableQueryInfo)->win.skey) && - ((*pTableQueryInfo)->win.skey <= pQuery->window.skey && (*pTableQueryInfo)->win.ekey >= pQuery->window.ekey)); - } + doTableQueryInfoTimeWindowCheck(pQuery, *pTableQueryInfo); if (!pRuntimeEnv->groupbyNormalCol) { setEnvForEachBlock(pQInfo, *pTableQueryInfo, &blockInfo); @@ -4929,6 +5272,41 @@ static bool multiTableMultioutputHelper(SQInfo *pQInfo, int32_t index) { return true; } +STsdbQueryCond createTsdbQueryCond(SQuery* pQuery, STimeWindow* win) { + STsdbQueryCond cond = { + .colList = pQuery->colList, + .order = pQuery->order.order, + .numOfCols = pQuery->numOfCols, + }; + + TIME_WINDOW_COPY(cond.twindow, *win); + return cond; +} + +static STableIdInfo createTableIdInfo(SQuery* pQuery) { + assert(pQuery != NULL && pQuery->current != NULL); + + STableIdInfo tidInfo; + STableId* id = TSDB_TABLEID(pQuery->current->pTable); + + tidInfo.uid = id->uid; + tidInfo.tid = id->tid; + tidInfo.key = pQuery->current->lastKey; + + return tidInfo; +} + +static void updateTableIdInfo(SQuery* pQuery, SHashObj* pTableIdInfo) { + STableIdInfo tidInfo = createTableIdInfo(pQuery); + STableIdInfo* idinfo = taosHashGet(pTableIdInfo, &tidInfo.tid, sizeof(tidInfo.tid)); + if (idinfo != NULL) { + assert(idinfo->tid == tidInfo.tid && idinfo->uid == tidInfo.uid); + idinfo->key = tidInfo.key; + } else { + taosHashPut(pTableIdInfo, &tidInfo.tid, sizeof(tidInfo.tid), &tidInfo, sizeof(STableIdInfo)); + } +} + /** * super table query handler * 1. super table projection query, group-by on normal columns query, ts-comp query @@ -4948,18 +5326,11 @@ static void sequentialTableProcess(SQInfo *pQInfo) { assert(pQuery->limit.offset == 0 && pQuery->limit.limit != 0); while (pQInfo->groupIndex < numOfGroups) { - SArray* group = taosArrayGetP(pQInfo->tableGroupInfo.pGroupList, pQInfo->groupIndex); - - qDebug("QInfo:%p last_row query on group:%d, total group:%" PRIzu ", current group:%p", pQInfo, pQInfo->groupIndex, - numOfGroups, group); + SArray *group = taosArrayGetP(pQInfo->tableGroupInfo.pGroupList, pQInfo->groupIndex); - STsdbQueryCond cond = { - .colList = pQuery->colList, - .order = pQuery->order.order, - .numOfCols = pQuery->numOfCols, - }; - - TIME_WINDOW_COPY(cond.twindow, pQuery->window); + qDebug("QInfo:%p point interpolation query on group:%d, total group:%" PRIzu ", current group:%p", pQInfo, + pQInfo->groupIndex, numOfGroups, group); + STsdbQueryCond cond = createTsdbQueryCond(pQuery, &pQuery->window); SArray *g1 = taosArrayInit(1, POINTER_BYTES); SArray *tx = taosArrayClone(group); @@ -4983,14 +5354,14 @@ static void sequentialTableProcess(SQInfo *pQInfo) { initCtxOutputBuf(pRuntimeEnv); - SArray* s = tsdbGetQueriedTableList(pRuntimeEnv->pQueryHandle); + SArray *s = tsdbGetQueriedTableList(pRuntimeEnv->pQueryHandle); assert(taosArrayGetSize(s) >= 1); setTagVal(pRuntimeEnv, taosArrayGetP(s, 0), pQInfo->tsdb); taosArrayDestroy(s); // here we simply set the first table as current table - SArray* first = GET_TABLEGROUP(pQInfo, pQInfo->groupIndex); + SArray *first = GET_TABLEGROUP(pQInfo, pQInfo->groupIndex); pQuery->current = taosArrayGetP(first, 0); scanOneTableDataBlocks(pRuntimeEnv, pQuery->current->lastKey); @@ -5012,19 +5383,14 @@ static void sequentialTableProcess(SQInfo *pQInfo) { break; } } - } else if (pRuntimeEnv->groupbyNormalCol) { // group-by on normal columns query + } else if (pRuntimeEnv->groupbyNormalCol) { // group-by on normal columns query while (pQInfo->groupIndex < numOfGroups) { - SArray* group = taosArrayGetP(pQInfo->tableGroupInfo.pGroupList, pQInfo->groupIndex); - - qDebug("QInfo:%p group by normal columns group:%d, total group:%" PRIzu "", pQInfo, pQInfo->groupIndex, numOfGroups); + SArray *group = taosArrayGetP(pQInfo->tableGroupInfo.pGroupList, pQInfo->groupIndex); - STsdbQueryCond cond = { - .colList = pQuery->colList, - .order = pQuery->order.order, - .numOfCols = pQuery->numOfCols, - }; + qDebug("QInfo:%p group by normal columns group:%d, total group:%" PRIzu "", pQInfo, pQInfo->groupIndex, + numOfGroups); - TIME_WINDOW_COPY(cond.twindow, pQuery->window); + STsdbQueryCond cond = createTsdbQueryCond(pQuery, &pQuery->window); SArray *g1 = taosArrayInit(1, POINTER_BYTES); SArray *tx = taosArrayClone(group); @@ -5047,7 +5413,7 @@ static void sequentialTableProcess(SQInfo *pQInfo) { longjmp(pRuntimeEnv->env, terrno); } - SArray* s = tsdbGetQueriedTableList(pRuntimeEnv->pQueryHandle); + SArray *s = tsdbGetQueriedTableList(pRuntimeEnv->pQueryHandle); assert(taosArrayGetSize(s) >= 1); setTagVal(pRuntimeEnv, taosArrayGetP(s, 0), pQInfo->tsdb); @@ -5056,26 +5422,26 @@ static void sequentialTableProcess(SQInfo *pQInfo) { scanMultiTableDataBlocks(pQInfo); pQInfo->groupIndex += 1; - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; + SResultRowInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; - // no results generated for current group, continue to try the next group + // no results generated for current group, continue to try the next group taosArrayDestroy(s); if (pWindowResInfo->size <= 0) { continue; } for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - pWindowResInfo->pResult[i]->closed = true; // enable return all results for group by normal columns + pWindowResInfo->pResult[i]->closed = true; // enable return all results for group by normal columns SResultRow *pResult = pWindowResInfo->pResult[i]; for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { - SResultRowCellInfo* pCell = getResultCell(pRuntimeEnv, pResult, j); + SResultRowCellInfo *pCell = getResultCell(pRuntimeEnv, pResult, j); pResult->numOfRows = (uint16_t)(MAX(pResult->numOfRows, pCell->numOfRes)); } } qDebug("QInfo:%p generated groupby columns results %d rows for group %d completed", pQInfo, pWindowResInfo->size, - pQInfo->groupIndex); + pQInfo->groupIndex); int32_t currentGroupIndex = pQInfo->groupIndex; pQuery->rec.rows = 0; @@ -5084,16 +5450,114 @@ static void sequentialTableProcess(SQInfo *pQInfo) { ensureOutputBufferSimple(pRuntimeEnv, pWindowResInfo->size); copyFromWindowResToSData(pQInfo, pWindowResInfo); - pQInfo->groupIndex = currentGroupIndex; //restore the group index + pQInfo->groupIndex = currentGroupIndex; // restore the group index assert(pQuery->rec.rows == pWindowResInfo->size); clearClosedTimeWindow(pRuntimeEnv); break; } + } else if (pRuntimeEnv->queryWindowIdentical && pRuntimeEnv->pTSBuf == NULL && !isTSCompQuery(pQuery)) { + //super table projection query with identical query time range for all tables. + SDataBlockInfo blockInfo = SDATA_BLOCK_INITIALIZER; + resetDefaultResInfoOutputBuf(pRuntimeEnv); + + SArray *group = GET_TABLEGROUP(pQInfo, 0); + assert(taosArrayGetSize(group) == pQInfo->tableqinfoGroupInfo.numOfTables && + 1 == taosArrayGetSize(pQInfo->tableqinfoGroupInfo.pGroupList)); + + void *pQueryHandle = pRuntimeEnv->pQueryHandle; + if (pQueryHandle == NULL) { + STsdbQueryCond con = createTsdbQueryCond(pQuery, &pQuery->window); + pRuntimeEnv->pQueryHandle = tsdbQueryTables(pQInfo->tsdb, &con, &pQInfo->tableGroupInfo, pQInfo, &pQInfo->memRef); + pQueryHandle = pRuntimeEnv->pQueryHandle; + } + + // skip blocks without load the actual data block from file if no filter condition present + // skipBlocks(&pQInfo->runtimeEnv); + // if (pQuery->limit.offset > 0 && pQuery->numOfFilterCols == 0) { + // setQueryStatus(pQuery, QUERY_COMPLETED); + // return; + // } + + bool hasMoreBlock = true; + int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order); + SQueryCostInfo *summary = &pRuntimeEnv->summary; + while ((hasMoreBlock = tsdbNextDataBlock(pQueryHandle)) == true) { + summary->totalBlocks += 1; + + if (IS_QUERY_KILLED(pQInfo)) { + longjmp(pRuntimeEnv->env, TSDB_CODE_TSC_QUERY_CANCELLED); + } + + tsdbRetrieveDataBlockInfo(pQueryHandle, &blockInfo); + STableQueryInfo **pTableQueryInfo = + (STableQueryInfo **)taosHashGet(pQInfo->tableqinfoGroupInfo.map, &blockInfo.tid, sizeof(blockInfo.tid)); + if (pTableQueryInfo == NULL) { + break; + } + + pQuery->current = *pTableQueryInfo; + doTableQueryInfoTimeWindowCheck(pQuery, *pTableQueryInfo); + + if (pRuntimeEnv->hasTagResults) { + setTagVal(pRuntimeEnv, pQuery->current->pTable, pQInfo->tsdb); + } + + uint32_t status = 0; + SDataStatis *pStatis = NULL; + SArray *pDataBlock = NULL; + + int32_t ret = loadDataBlockOnDemand(pRuntimeEnv, &pQuery->current->windowResInfo, pQueryHandle, &blockInfo, + &pStatis, &pDataBlock, &status); + if (ret != TSDB_CODE_SUCCESS) { + break; + } + + if(status == BLK_DATA_DISCARD) { + pQuery->current->lastKey = + QUERY_IS_ASC_QUERY(pQuery) ? blockInfo.window.ekey + step : blockInfo.window.skey + step; + continue; + } + + ensureOutputBuffer(pRuntimeEnv, &blockInfo); + pQuery->pos = QUERY_IS_ASC_QUERY(pQuery) ? 0 : blockInfo.rows - 1; + int32_t numOfRes = tableApplyFunctionsOnBlock(pRuntimeEnv, &blockInfo, pStatis, binarySearchForKey, pDataBlock); + + summary->totalRows += blockInfo.rows; + qDebug("QInfo:%p check data block, brange:%" PRId64 "-%" PRId64 ", numOfRows:%d, numOfRes:%d, lastKey:%" PRId64, + GET_QINFO_ADDR(pRuntimeEnv), blockInfo.window.skey, blockInfo.window.ekey, blockInfo.rows, numOfRes, + pQuery->current->lastKey); + + pQuery->rec.rows = getNumOfResult(pRuntimeEnv); + + // the flag may be set by tableApplyFunctionsOnBlock, clear it here + CLEAR_QUERY_STATUS(pQuery, QUERY_COMPLETED); + + updateTableIdInfo(pQuery, pQInfo->arrTableIdInfo); + skipResults(pRuntimeEnv); + + // the limitation of output result is reached, set the query completed + if (limitResults(pRuntimeEnv)) { + setQueryStatus(pQuery, QUERY_COMPLETED); + SET_STABLE_QUERY_OVER(pQInfo); + break; + } + + // while the output buffer is full or limit/offset is applied, query may be paused here + if (Q_STATUS_EQUAL(pQuery->status, QUERY_RESBUF_FULL|QUERY_COMPLETED)) { + break; + } + } + + if (!hasMoreBlock) { + setQueryStatus(pQuery, QUERY_COMPLETED); + SET_STABLE_QUERY_OVER(pQInfo); + } } else { /* - * 1. super table projection query, 2. ts-comp query - * if the subgroup index is larger than 0, results generated by group by tbname,k is existed. + * the following two cases handled here. + * 1. ts-comp query, and 2. the super table projection query with different query time range for each table. + * If the subgroup index is larger than 0, results generated by group by tbname,k is existed. * we need to return it to client in the first place. */ if (pQInfo->groupIndex > 0) { @@ -5156,14 +5620,7 @@ static void sequentialTableProcess(SQInfo *pQInfo) { * to ensure that, we can reset the query range once query on a meter is completed. */ pQInfo->tableIndex++; - - STableIdInfo tidInfo = {0}; - - STableId* id = TSDB_TABLEID(pQuery->current->pTable); - tidInfo.uid = id->uid; - tidInfo.tid = id->tid; - tidInfo.key = pQuery->current->lastKey; - taosArrayPush(pQInfo->arrTableIdInfo, &tidInfo); + updateTableIdInfo(pQuery, pQInfo->arrTableIdInfo); // if the buffer is full or group by each table, we need to jump out of the loop if (Q_STATUS_EQUAL(pQuery->status, QUERY_RESBUF_FULL)) { @@ -5190,31 +5647,31 @@ static void sequentialTableProcess(SQInfo *pQInfo) { if (pQInfo->tableIndex >= pQInfo->tableqinfoGroupInfo.numOfTables) { setQueryStatus(pQuery, QUERY_COMPLETED); } - } - /* - * 1. super table projection query, group-by on normal columns query, ts-comp query - * 2. point interpolation query, last row query - * - * group-by on normal columns query and last_row query do NOT invoke the finalizer here, - * since the finalize stage will be done at the client side. - * - * projection query, point interpolation query do not need the finalizer. - * - * Only the ts-comp query requires the finalizer function to be executed here. - */ - if (isTSCompQuery(pQuery)) { - finalizeQueryResult(pRuntimeEnv); - } + /* + * 1. super table projection query, group-by on normal columns query, ts-comp query + * 2. point interpolation query, last row query + * + * group-by on normal columns query and last_row query do NOT invoke the finalizer here, + * since the finalize stage will be done at the client side. + * + * projection query, point interpolation query do not need the finalizer. + * + * Only the ts-comp query requires the finalizer function to be executed here. + */ + if (isTSCompQuery(pQuery)) { + finalizeQueryResult(pRuntimeEnv); + } - if (pRuntimeEnv->pTSBuf != NULL) { - pRuntimeEnv->cur = pRuntimeEnv->pTSBuf->cur; - } + if (pRuntimeEnv->pTSBuf != NULL) { + pRuntimeEnv->cur = pRuntimeEnv->pTSBuf->cur; + } - qDebug( - "QInfo %p numOfTables:%" PRIu64 ", index:%d, numOfGroups:%" PRIzu ", %" PRId64 " points returned, total:%" PRId64 ", offset:%" PRId64, - pQInfo, (uint64_t)pQInfo->tableqinfoGroupInfo.numOfTables, pQInfo->tableIndex, numOfGroups, pQuery->rec.rows, pQuery->rec.total, - pQuery->limit.offset); + qDebug("QInfo %p numOfTables:%" PRIu64 ", index:%d, numOfGroups:%" PRIzu ", %" PRId64 + " points returned, total:%" PRId64 ", offset:%" PRId64, + pQInfo, (uint64_t)pQInfo->tableqinfoGroupInfo.numOfTables, pQInfo->tableIndex, numOfGroups, pQuery->rec.rows, + pQuery->rec.total, pQuery->limit.offset); + } } static void doSaveContext(SQInfo *pQInfo) { @@ -5229,13 +5686,7 @@ static void doSaveContext(SQInfo *pQInfo) { SWITCH_ORDER(pRuntimeEnv->pTSBuf->cur.order); } - STsdbQueryCond cond = { - .order = pQuery->order.order, - .colList = pQuery->colList, - .numOfCols = pQuery->numOfCols, - }; - - TIME_WINDOW_COPY(cond.twindow, pQuery->window); + STsdbQueryCond cond = createTsdbQueryCond(pQuery, &pQuery->window); // clean unused handle if (pRuntimeEnv->pSecQueryHandle != NULL) { @@ -5386,7 +5837,7 @@ static void doSecondaryArithmeticProcess(SQuery* pQuery) { tFilePage **data = calloc(pQuery->numOfExpr2, POINTER_BYTES); for (int32_t i = 0; i < pQuery->numOfExpr2; ++i) { int32_t bytes = pQuery->pExpr2[i].bytes; - data[i] = (tFilePage *)malloc(bytes * pQuery->rec.rows + sizeof(tFilePage)); + data[i] = (tFilePage *)malloc((size_t)(bytes * pQuery->rec.rows) + sizeof(tFilePage)); } arithSup.offset = 0; @@ -5408,7 +5859,7 @@ static void doSecondaryArithmeticProcess(SQuery* pQuery) { for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { if (pSqlFunc->functionId == pQuery->pExpr1[j].base.functionId && pSqlFunc->colInfo.colId == pQuery->pExpr1[j].base.colInfo.colId) { - memcpy(data[i]->data, pQuery->sdata[j]->data, pQuery->pExpr1[j].bytes * pQuery->rec.rows); + memcpy(data[i]->data, pQuery->sdata[j]->data, (size_t)(pQuery->pExpr1[j].bytes * pQuery->rec.rows)); break; } } @@ -5420,7 +5871,7 @@ static void doSecondaryArithmeticProcess(SQuery* pQuery) { } for (int32_t i = 0; i < pQuery->numOfExpr2; ++i) { - memcpy(pQuery->sdata[i]->data, data[i]->data, pQuery->pExpr2[i].bytes * pQuery->rec.rows); + memcpy(pQuery->sdata[i]->data, data[i]->data, (size_t)(pQuery->pExpr2[i].bytes * pQuery->rec.rows)); } for (int32_t i = 0; i < pQuery->numOfExpr2; ++i) { @@ -5508,13 +5959,8 @@ static void tableMultiOutputProcess(SQInfo *pQInfo, STableQueryInfo* pTableInfo) qDebug("QInfo:%p query paused due to output limitation, next qrange:%" PRId64 "-%" PRId64, pQInfo, pQuery->current->lastKey, pQuery->window.ekey); } else if (Q_STATUS_EQUAL(pQuery->status, QUERY_COMPLETED)) { - STableIdInfo tidInfo; - STableId* id = TSDB_TABLEID(pQuery->current->pTable); - - tidInfo.uid = id->uid; - tidInfo.tid = id->tid; - tidInfo.key = pQuery->current->lastKey; - taosArrayPush(pQInfo->arrTableIdInfo, &tidInfo); + STableIdInfo tidInfo = createTableIdInfo(pQuery); + taosHashPut(pQInfo->arrTableIdInfo, &tidInfo.tid, sizeof(tidInfo.tid), &tidInfo, sizeof(STableIdInfo)); } if (!isTSCompQuery(pQuery)) { @@ -5836,11 +6282,9 @@ static int32_t convertQueryMsg(SQueryTableMsg *pQueryMsg, SArray **pTableIdList, pQueryMsg->interval.interval = htobe64(pQueryMsg->interval.interval); pQueryMsg->interval.sliding = htobe64(pQueryMsg->interval.sliding); pQueryMsg->interval.offset = htobe64(pQueryMsg->interval.offset); - pQueryMsg->interval.intervalUnit = pQueryMsg->interval.intervalUnit; - pQueryMsg->interval.slidingUnit = pQueryMsg->interval.slidingUnit; - pQueryMsg->interval.offsetUnit = pQueryMsg->interval.offsetUnit; pQueryMsg->limit = htobe64(pQueryMsg->limit); pQueryMsg->offset = htobe64(pQueryMsg->offset); + pQueryMsg->tableLimit = htobe64(pQueryMsg->tableLimit); pQueryMsg->order = htons(pQueryMsg->order); pQueryMsg->orderColId = htons(pQueryMsg->orderColId); @@ -6447,9 +6891,11 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou goto _cleanup; } + int32_t srcSize = 0; for (int16_t i = 0; i < numOfCols; ++i) { pQuery->colList[i] = pQueryMsg->colList[i]; pQuery->colList[i].filters = tscFilterInfoClone(pQueryMsg->colList[i].filters, pQuery->colList[i].numOfFilters); + srcSize += pQuery->colList[i].bytes; } // calculate the result row size @@ -6510,14 +6956,20 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_NO_LOCK); } - int tableIndex = 0; - pQInfo->runtimeEnv.interBufSize = getOutputInterResultBufSize(pQuery); pQInfo->runtimeEnv.summary.tableInfoSize += (pTableGroupInfo->numOfTables * sizeof(STableQueryInfo)); pQInfo->runtimeEnv.pResultRowHashTable = taosHashInit(pTableGroupInfo->numOfTables, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_NO_LOCK); pQInfo->runtimeEnv.keyBuf = malloc(TSDB_MAX_BYTES_PER_ROW); pQInfo->runtimeEnv.pool = initResultRowPool(getWindowResultSize(&pQInfo->runtimeEnv)); + pQInfo->runtimeEnv.prevRow = malloc(POINTER_BYTES * pQuery->numOfCols + srcSize); + + char* start = POINTER_BYTES * pQuery->numOfCols + (char*) pQInfo->runtimeEnv.prevRow; + pQInfo->runtimeEnv.prevRow[0] = start; + + for(int32_t i = 1; i < pQuery->numOfCols; ++i) { + pQInfo->runtimeEnv.prevRow[i] = pQInfo->runtimeEnv.prevRow[i - 1] + pQuery->colList[i-1].bytes; + } pQInfo->pBuf = calloc(pTableGroupInfo->numOfTables, sizeof(STableQueryInfo)); if (pQInfo->pBuf == NULL) { @@ -6525,7 +6977,7 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou } // NOTE: pTableCheckInfo need to update the query time range and the lastKey info - pQInfo->arrTableIdInfo = taosArrayInit(tableIndex, sizeof(STableIdInfo)); + pQInfo->arrTableIdInfo = taosHashInit(pTableGroupInfo->numOfTables, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), false, HASH_NO_LOCK); pQInfo->dataReady = QUERY_RESULT_NOT_READY; pQInfo->rspContext = NULL; pthread_mutex_init(&pQInfo->lock, NULL); @@ -6535,10 +6987,10 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou pQuery->window = pQueryMsg->window; changeExecuteScanOrder(pQInfo, pQueryMsg, stableQuery); + pQInfo->runtimeEnv.queryWindowIdentical = true; STimeWindow window = pQuery->window; int32_t index = 0; - for(int32_t i = 0; i < numOfGroups; ++i) { SArray* pa = taosArrayGetP(pQInfo->tableGroupInfo.pGroupList, i); @@ -6553,9 +7005,12 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou for(int32_t j = 0; j < s; ++j) { STableKeyInfo* info = taosArrayGet(pa, j); - void* buf = (char*) pQInfo->pBuf + index * sizeof(STableQueryInfo); - window.skey = info->lastKey; + if (info->lastKey != pQuery->window.skey) { + pQInfo->runtimeEnv.queryWindowIdentical = false; + } + + void* buf = (char*) pQInfo->pBuf + index * sizeof(STableQueryInfo); STableQueryInfo* item = createTableQueryInfo(&pQInfo->runtimeEnv, info->pTable, window, buf); if (item == NULL) { goto _cleanup; @@ -6769,7 +7224,7 @@ static void freeQInfo(SQInfo *pQInfo) { tfree(pQInfo->pBuf); tsdbDestroyTableGroup(&pQInfo->tableGroupInfo); - taosArrayDestroy(pQInfo->arrTableIdInfo); + taosHashCleanup(pQInfo->arrTableIdInfo); pQInfo->signature = 0; @@ -7149,7 +7604,7 @@ int32_t qDumpRetrieveResult(qinfo_t qinfo, SRetrieveTableRsp **pRsp, int32_t *co size_t size = getResultSize(pQInfo, &pQuery->rec.rows); size += sizeof(int32_t); - size += sizeof(STableIdInfo) * taosArrayGetSize(pQInfo->arrTableIdInfo); + size += sizeof(STableIdInfo) * taosHashGetSize(pQInfo->arrTableIdInfo); *contLen = (int32_t)(size + sizeof(SRetrieveTableRsp)); diff --git a/src/query/src/qParserImpl.c b/src/query/src/qParserImpl.c index 5cd9d3c77a5211fa1c0058978799447f38867180..7d71d9f7f15ad88d0bc1565047ae7686320d2832 100644 --- a/src/query/src/qParserImpl.c +++ b/src/query/src/qParserImpl.c @@ -405,14 +405,29 @@ void tSQLSetColumnType(TAOS_FIELD *pField, SStrToken *type) { if (type->type == 0) { pField->bytes = 0; } else { - pField->bytes = (int16_t)(-(int32_t)type->type * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE); + int32_t bytes = -(int32_t)(type->type); + if (bytes > (TSDB_MAX_NCHAR_LEN - VARSTR_HEADER_SIZE) / TSDB_NCHAR_SIZE) { + // we have to postpone reporting the error because it cannot be done here + // as pField->bytes is int16_t, use 'TSDB_MAX_NCHAR_LEN + 1' to avoid overflow + bytes = TSDB_MAX_NCHAR_LEN + 1; + } else { + bytes = bytes * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE; + } + pField->bytes = (int16_t)bytes; } } else if (i == TSDB_DATA_TYPE_BINARY) { /* for binary, the TOKENTYPE is the length of binary */ if (type->type == 0) { pField->bytes = 0; } else { - pField->bytes = (int16_t) (-(int32_t) type->type + VARSTR_HEADER_SIZE); + int32_t bytes = -(int32_t)(type->type); + if (bytes > TSDB_MAX_BINARY_LEN - VARSTR_HEADER_SIZE) { + // refer comment for NCHAR above + bytes = TSDB_MAX_BINARY_LEN + 1; + } else { + bytes += VARSTR_HEADER_SIZE; + } + pField->bytes = (int16_t)bytes; } } break; diff --git a/src/query/src/qResultbuf.c b/src/query/src/qResultbuf.c index 55a7aea53a8d53ef66bf33eaa73a47d8cf0d77b6..bc7243830d836287cde808076cfb24de80fa574d 100644 --- a/src/query/src/qResultbuf.c +++ b/src/query/src/qResultbuf.c @@ -423,9 +423,8 @@ void destroyResultBuf(SDiskbasedResultBuf* pResultBuf) { unlink(pResultBuf->path); tfree(pResultBuf->path); - SHashMutableIterator* iter = taosHashCreateIter(pResultBuf->groupSet); - while(taosHashIterNext(iter)) { - SArray** p = (SArray**) taosHashIterGet(iter); + SArray** p = taosHashIterate(pResultBuf->groupSet, NULL); + while(p) { size_t n = taosArrayGetSize(*p); for(int32_t i = 0; i < n; ++i) { SPageInfo* pi = taosArrayGetP(*p, i); @@ -434,10 +433,9 @@ void destroyResultBuf(SDiskbasedResultBuf* pResultBuf) { } taosArrayDestroy(*p); + p = taosHashIterate(pResultBuf->groupSet, p); } - taosHashDestroyIter(iter); - tdListFree(pResultBuf->lruList); taosArrayDestroy(pResultBuf->emptyDummyIdList); taosHashCleanup(pResultBuf->groupSet); diff --git a/src/query/src/qUtil.c b/src/query/src/qUtil.c index 2b4cdef9b0b4a069ef4d3148b9ea67361ae88f7b..6c845b012faade8fbb693d3406013b8264b0c4a7 100644 --- a/src/query/src/qUtil.c +++ b/src/query/src/qUtil.c @@ -43,51 +43,48 @@ int32_t getOutputInterResultBufSize(SQuery* pQuery) { return size; } -int32_t initWindowResInfo(SWindowResInfo *pWindowResInfo, int32_t size, int32_t threshold, int16_t type) { - pWindowResInfo->capacity = size; - pWindowResInfo->threshold = threshold; - - pWindowResInfo->type = type; - pWindowResInfo->curIndex = -1; - pWindowResInfo->size = 0; - pWindowResInfo->prevSKey = TSKEY_INITIAL_VAL; +int32_t initWindowResInfo(SResultRowInfo *pResultRowInfo, int32_t size, int16_t type) { + pResultRowInfo->capacity = size; + + pResultRowInfo->type = type; + pResultRowInfo->curIndex = -1; + pResultRowInfo->size = 0; + pResultRowInfo->prevSKey = TSKEY_INITIAL_VAL; - pWindowResInfo->pResult = calloc(pWindowResInfo->capacity, POINTER_BYTES); - if (pWindowResInfo->pResult == NULL) { + pResultRowInfo->pResult = calloc(pResultRowInfo->capacity, POINTER_BYTES); + if (pResultRowInfo->pResult == NULL) { return TSDB_CODE_QRY_OUT_OF_MEMORY; } return TSDB_CODE_SUCCESS; } -void cleanupTimeWindowInfo(SWindowResInfo *pWindowResInfo) { - if (pWindowResInfo == NULL) { +void cleanupTimeWindowInfo(SResultRowInfo *pResultRowInfo) { + if (pResultRowInfo == NULL) { return; } - if (pWindowResInfo->capacity == 0) { - assert(pWindowResInfo->pResult == NULL); + if (pResultRowInfo->capacity == 0) { + assert(pResultRowInfo->pResult == NULL); return; } - if (pWindowResInfo->type == TSDB_DATA_TYPE_BINARY || pWindowResInfo->type == TSDB_DATA_TYPE_NCHAR) { - for(int32_t i = 0; i < pWindowResInfo->size; ++i) { - tfree(pWindowResInfo->pResult[i]->key); + if (pResultRowInfo->type == TSDB_DATA_TYPE_BINARY || pResultRowInfo->type == TSDB_DATA_TYPE_NCHAR) { + for(int32_t i = 0; i < pResultRowInfo->size; ++i) { + tfree(pResultRowInfo->pResult[i]->key); } } - tfree(pWindowResInfo->pResult); + tfree(pResultRowInfo->pResult); } -void resetTimeWindowInfo(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowResInfo) { - if (pWindowResInfo == NULL || pWindowResInfo->capacity == 0) { +void resetTimeWindowInfo(SQueryRuntimeEnv *pRuntimeEnv, SResultRowInfo *pResultRowInfo) { + if (pResultRowInfo == NULL || pResultRowInfo->capacity == 0) { return; } -// assert(pWindowResInfo->size == 1); - - for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - SResultRow *pWindowRes = pWindowResInfo->pResult[i]; - clearResultRow(pRuntimeEnv, pWindowRes, pWindowResInfo->type); + for (int32_t i = 0; i < pResultRowInfo->size; ++i) { + SResultRow *pWindowRes = pResultRowInfo->pResult[i]; + clearResultRow(pRuntimeEnv, pWindowRes, pResultRowInfo->type); int32_t groupIndex = 0; int64_t uid = 0; @@ -96,30 +93,30 @@ void resetTimeWindowInfo(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowR taosHashRemove(pRuntimeEnv->pResultRowHashTable, (const char *)pRuntimeEnv->keyBuf, GET_RES_WINDOW_KEY_LEN(sizeof(groupIndex))); } - pWindowResInfo->curIndex = -1; - pWindowResInfo->size = 0; + pResultRowInfo->curIndex = -1; + pResultRowInfo->size = 0; - pWindowResInfo->startTime = TSKEY_INITIAL_VAL; - pWindowResInfo->prevSKey = TSKEY_INITIAL_VAL; + pResultRowInfo->startTime = TSKEY_INITIAL_VAL; + pResultRowInfo->prevSKey = TSKEY_INITIAL_VAL; } void clearFirstNWindowRes(SQueryRuntimeEnv *pRuntimeEnv, int32_t num) { - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; - if (pWindowResInfo == NULL || pWindowResInfo->capacity == 0 || pWindowResInfo->size == 0 || num == 0) { + SResultRowInfo *pResultRowInfo = &pRuntimeEnv->windowResInfo; + if (pResultRowInfo == NULL || pResultRowInfo->capacity == 0 || pResultRowInfo->size == 0 || num == 0) { return; } - int32_t numOfClosed = numOfClosedTimeWindow(pWindowResInfo); + int32_t numOfClosed = numOfClosedTimeWindow(pResultRowInfo); assert(num >= 0 && num <= numOfClosed); - int16_t type = pWindowResInfo->type; + int16_t type = pResultRowInfo->type; int64_t uid = getResultInfoUId(pRuntimeEnv); char *key = NULL; int16_t bytes = -1; for (int32_t i = 0; i < num; ++i) { - SResultRow *pResult = pWindowResInfo->pResult[i]; + SResultRow *pResult = pResultRowInfo->pResult[i]; if (pResult->closed) { // remove the window slot from hash table getResultRowKeyInfo(pResult, type, &key, &bytes); SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, key, bytes, uid); @@ -129,23 +126,23 @@ void clearFirstNWindowRes(SQueryRuntimeEnv *pRuntimeEnv, int32_t num) { } } - int32_t remain = pWindowResInfo->size - num; + int32_t remain = pResultRowInfo->size - num; // clear all the closed windows from the window list for (int32_t k = 0; k < remain; ++k) { - copyResultRow(pRuntimeEnv, pWindowResInfo->pResult[k], pWindowResInfo->pResult[num + k]); + copyResultRow(pRuntimeEnv, pResultRowInfo->pResult[k], pResultRowInfo->pResult[num + k], type); } // move the unclosed window in the front of the window list - for (int32_t k = remain; k < pWindowResInfo->size; ++k) { - SResultRow *pWindowRes = pWindowResInfo->pResult[k]; - clearResultRow(pRuntimeEnv, pWindowRes, pWindowResInfo->type); + for (int32_t k = remain; k < pResultRowInfo->size; ++k) { + SResultRow *pWindowRes = pResultRowInfo->pResult[k]; + clearResultRow(pRuntimeEnv, pWindowRes, pResultRowInfo->type); } - pWindowResInfo->size = remain; + pResultRowInfo->size = remain; - for (int32_t k = 0; k < pWindowResInfo->size; ++k) { - SResultRow *pResult = pWindowResInfo->pResult[k]; + for (int32_t k = 0; k < pResultRowInfo->size; ++k) { + SResultRow *pResult = pResultRowInfo->pResult[k]; getResultRowKeyInfo(pResult, type, &key, &bytes); SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, key, bytes, uid); @@ -153,43 +150,43 @@ void clearFirstNWindowRes(SQueryRuntimeEnv *pRuntimeEnv, int32_t num) { assert(p != NULL); int32_t v = (*p - num); - assert(v >= 0 && v <= pWindowResInfo->size); + assert(v >= 0 && v <= pResultRowInfo->size); SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, key, bytes, uid); taosHashPut(pRuntimeEnv->pResultRowHashTable, pRuntimeEnv->keyBuf, GET_RES_WINDOW_KEY_LEN(bytes), (char *)&v, sizeof(int32_t)); } - pWindowResInfo->curIndex = -1; + pResultRowInfo->curIndex = -1; } void clearClosedTimeWindow(SQueryRuntimeEnv *pRuntimeEnv) { - SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; - if (pWindowResInfo == NULL || pWindowResInfo->capacity == 0 || pWindowResInfo->size == 0) { + SResultRowInfo *pResultRowInfo = &pRuntimeEnv->windowResInfo; + if (pResultRowInfo == NULL || pResultRowInfo->capacity == 0 || pResultRowInfo->size == 0) { return; } - int32_t numOfClosed = numOfClosedTimeWindow(pWindowResInfo); + int32_t numOfClosed = numOfClosedTimeWindow(pResultRowInfo); clearFirstNWindowRes(pRuntimeEnv, numOfClosed); } -int32_t numOfClosedTimeWindow(SWindowResInfo *pWindowResInfo) { +int32_t numOfClosedTimeWindow(SResultRowInfo *pResultRowInfo) { int32_t i = 0; - while (i < pWindowResInfo->size && pWindowResInfo->pResult[i]->closed) { + while (i < pResultRowInfo->size && pResultRowInfo->pResult[i]->closed) { ++i; } return i; } -void closeAllTimeWindow(SWindowResInfo *pWindowResInfo) { - assert(pWindowResInfo->size >= 0 && pWindowResInfo->capacity >= pWindowResInfo->size); +void closeAllTimeWindow(SResultRowInfo *pResultRowInfo) { + assert(pResultRowInfo->size >= 0 && pResultRowInfo->capacity >= pResultRowInfo->size); - for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - if (pWindowResInfo->pResult[i]->closed) { + for (int32_t i = 0; i < pResultRowInfo->size; ++i) { + if (pResultRowInfo->pResult[i]->closed) { continue; } - pWindowResInfo->pResult[i]->closed = true; + pResultRowInfo->pResult[i]->closed = true; } } @@ -198,41 +195,41 @@ void closeAllTimeWindow(SWindowResInfo *pWindowResInfo) { * the last qualified time stamp in case of sliding query, which the sliding time is not equalled to the interval time. * NOTE: remove redundant, only when the result set order equals to traverse order */ -void removeRedundantWindow(SWindowResInfo *pWindowResInfo, TSKEY lastKey, int32_t order) { - assert(pWindowResInfo->size >= 0 && pWindowResInfo->capacity >= pWindowResInfo->size); - if (pWindowResInfo->size <= 1) { +void removeRedundantWindow(SResultRowInfo *pResultRowInfo, TSKEY lastKey, int32_t order) { + assert(pResultRowInfo->size >= 0 && pResultRowInfo->capacity >= pResultRowInfo->size); + if (pResultRowInfo->size <= 1) { return; } // get the result order - int32_t resultOrder = (pWindowResInfo->pResult[0]->win.skey < pWindowResInfo->pResult[1]->win.skey)? 1:-1; + int32_t resultOrder = (pResultRowInfo->pResult[0]->win.skey < pResultRowInfo->pResult[1]->win.skey)? 1:-1; if (order != resultOrder) { return; } int32_t i = 0; if (order == QUERY_ASC_FORWARD_STEP) { - TSKEY ekey = pWindowResInfo->pResult[i]->win.ekey; - while (i < pWindowResInfo->size && (ekey < lastKey)) { + TSKEY ekey = pResultRowInfo->pResult[i]->win.ekey; + while (i < pResultRowInfo->size && (ekey < lastKey)) { ++i; } } else if (order == QUERY_DESC_FORWARD_STEP) { - while (i < pWindowResInfo->size && (pWindowResInfo->pResult[i]->win.skey > lastKey)) { + while (i < pResultRowInfo->size && (pResultRowInfo->pResult[i]->win.skey > lastKey)) { ++i; } } - if (i < pWindowResInfo->size) { - pWindowResInfo->size = (i + 1); + if (i < pResultRowInfo->size) { + pResultRowInfo->size = (i + 1); } } -bool isWindowResClosed(SWindowResInfo *pWindowResInfo, int32_t slot) { - return (getResultRow(pWindowResInfo, slot)->closed == true); +bool isWindowResClosed(SResultRowInfo *pResultRowInfo, int32_t slot) { + return (getResultRow(pResultRowInfo, slot)->closed == true); } -void closeTimeWindow(SWindowResInfo *pWindowResInfo, int32_t slot) { - getResultRow(pWindowResInfo, slot)->closed = true; +void closeTimeWindow(SResultRowInfo *pResultRowInfo, int32_t slot) { + getResultRow(pResultRowInfo, slot)->closed = true; } void clearResultRow(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *pWindowRes, int16_t type) { @@ -272,9 +269,15 @@ void clearResultRow(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *pWindowRes, int16 * since the attribute of "Pos" is bound to each window result when the window result is created in the * disk-based result buffer. */ -void copyResultRow(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *dst, const SResultRow *src) { +void copyResultRow(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *dst, const SResultRow *src, int16_t type) { dst->numOfRows = src->numOfRows; - dst->win = src->win; + + if (type == TSDB_DATA_TYPE_BINARY || type == TSDB_DATA_TYPE_NCHAR) { + dst->key = realloc(dst->key, varDataTLen(src->key)); + varDataCopy(dst->key, src->key); + } else { + dst->win = src->win; + } dst->closed = src->closed; int32_t nOutputCols = pRuntimeEnv->pQuery->numOfOutput; @@ -389,11 +392,10 @@ uint64_t getResultInfoUId(SQueryRuntimeEnv* pRuntimeEnv) { } SQuery* pQuery = pRuntimeEnv->pQuery; - if ((pQuery->checkBuffer == 1 && pQuery->interval.interval == 0) || isPointInterpoQuery(pQuery) || - pRuntimeEnv->groupbyNormalCol) { + if (pQuery->interval.interval == 0 || isPointInterpoQuery(pQuery) || pRuntimeEnv->groupbyNormalCol) { return 0; } - STableId* id = TSDB_TABLEID(pRuntimeEnv->pQuery->current); + STableId* id = TSDB_TABLEID(pRuntimeEnv->pQuery->current->pTable); return id->uid; } \ No newline at end of file diff --git a/src/rpc/src/rpcMain.c b/src/rpc/src/rpcMain.c index acceaf9d7a63f3378afbc8d8f485f14f77b50af6..587c079fe6201ac804aca35ab15d54bff968688e 100644 --- a/src/rpc/src/rpcMain.c +++ b/src/rpc/src/rpcMain.c @@ -630,8 +630,16 @@ static void rpcReleaseConn(SRpcConn *pConn) { } else { // if there is an outgoing message, free it if (pConn->outType && pConn->pReqMsg) { - if (pConn->pContext) pConn->pContext->pConn = NULL; - taosRemoveRef(tsRpcRefId, pConn->pContext->rid); + SRpcReqContext *pContext = pConn->pContext; + if (pContext->pRsp) { + // for synchronous API, post semaphore to unblock app + pContext->pRsp->code = TSDB_CODE_RPC_APP_ERROR; + pContext->pRsp->pCont = NULL; + pContext->pRsp->contLen = 0; + tsem_post(pContext->pSem); + } + pContext->pConn = NULL; + taosRemoveRef(tsRpcRefId, pContext->rid); } } @@ -1078,13 +1086,6 @@ static void *rpcProcessMsgFromPeer(SRecvInfo *pRecv) { tDebug("%s %p %p, %s is sent with error code:0x%x", pRpc->label, pConn, (void *)pHead->ahandle, taosMsg[pHead->msgType+1], code); } } else { // msg is passed to app only parsing is ok - - if (pHead->msgType == TSDB_MSG_TYPE_NETWORK_TEST) { - rpcSendQuickRsp(pConn, TSDB_CODE_SUCCESS); - rpcFreeMsg(pRecv->msg); - return pConn; - } - rpcProcessIncomingMsg(pConn, pHead, pContext); } } diff --git a/src/rpc/src/rpcTcp.c b/src/rpc/src/rpcTcp.c index 7b8cf3cda2f98137ac68bacec257d5b7d14a2a4c..2850046d050cba4adceeac615c0672d36dfed7d9 100644 --- a/src/rpc/src/rpcTcp.c +++ b/src/rpc/src/rpcTcp.c @@ -371,10 +371,13 @@ void taosCloseTcpConnection(void *chandle) { int taosSendTcpData(uint32_t ip, uint16_t port, void *data, int len, void *chandle) { SFdObj *pFdObj = chandle; - if (pFdObj == NULL || pFdObj->signature != pFdObj) return -1; + SThreadObj *pThreadObj = pFdObj->pThreadObj; + + int ret = taosWriteMsg(pFdObj->fd, data, len); + tTrace("%s %p TCP data is sent, FD:%p fd:%d bytes:%d", pThreadObj->label, pFdObj->thandle, pFdObj, pFdObj->fd, ret); - return taosWriteMsg(pFdObj->fd, data, len); + return ret; } static void taosReportBrokenLink(SFdObj *pFdObj) { @@ -409,7 +412,7 @@ static int taosReadTcpData(SFdObj *pFdObj, SRecvInfo *pInfo) { headLen = taosReadMsg(pFdObj->fd, &rpcHead, sizeof(SRpcHead)); if (headLen != sizeof(SRpcHead)) { - tDebug("%s %p read error, headLen:%d", pThreadObj->label, pFdObj->thandle, headLen); + tDebug("%s %p read error, FD:%p headLen:%d", pThreadObj->label, pFdObj->thandle, pFdObj, headLen); return -1; } @@ -420,7 +423,7 @@ static int taosReadTcpData(SFdObj *pFdObj, SRecvInfo *pInfo) { tError("%s %p TCP malloc(size:%d) fail", pThreadObj->label, pFdObj->thandle, msgLen); return -1; } else { - tTrace("TCP malloc mem:%p size:%d", buffer, size); + tTrace("%s %p read data, FD:%p fd:%d TCP malloc mem:%p", pThreadObj->label, pFdObj->thandle, pFdObj, pFdObj->fd, buffer); } msg = buffer + tsRpcOverhead; @@ -583,8 +586,8 @@ static void taosFreeFdObj(SFdObj *pFdObj) { pthread_mutex_unlock(&pThreadObj->mutex); - tDebug("%s %p TCP connection is closed, FD:%p numOfFds:%d", - pThreadObj->label, pFdObj->thandle, pFdObj, pThreadObj->numOfFds); + tDebug("%s %p TCP connection is closed, FD:%p fd:%d numOfFds:%d", + pThreadObj->label, pFdObj->thandle, pFdObj, pFdObj->fd, pThreadObj->numOfFds); tfree(pFdObj); } diff --git a/src/sync/inc/syncInt.h b/src/sync/inc/syncInt.h index 05b7adc5f44b5853cd177443d062167c77681dce..6d0c52284f7f2429ccfa20389ccd66b2447610b9 100644 --- a/src/sync/inc/syncInt.h +++ b/src/sync/inc/syncInt.h @@ -139,15 +139,14 @@ typedef struct SsyncPeer { char id[TSDB_EP_LEN + 32]; // peer vgId + end point uint64_t version; uint64_t sversion; // track the peer version in retrieve process + uint64_t lastFileVer; // track the file version while retrieve + uint64_t lastWalVer; // track the wal version while retrieve int32_t syncFd; int32_t peerFd; // forward FD int32_t numOfRetrieves; // number of retrieves tried int32_t fileChanged; // a flag to indicate file is changed during retrieving process void * timer; void * pConn; - int32_t notifyFd; - int32_t watchNum; - int32_t *watchFd; int32_t refCount; // reference count struct SSyncNode *pSyncNode; } SSyncPeer; @@ -172,6 +171,7 @@ typedef struct SSyncNode { FNotifyRole notifyRole; FNotifyFlowCtrl notifyFlowCtrl; FNotifyFileSynced notifyFileSynced; + FGetVersion getVersion; pthread_mutex_t mutex; } SSyncNode; diff --git a/src/sync/src/syncMain.c b/src/sync/src/syncMain.c index d2d6d2d7fae03e2341c58e73ff4984046655d682..c86265d55622fc563b781aae851ee5af8f35b752 100644 --- a/src/sync/src/syncMain.c +++ b/src/sync/src/syncMain.c @@ -196,6 +196,7 @@ int64_t syncStart(const SSyncInfo *pInfo) { pNode->confirmForward = pInfo->confirmForward; pNode->notifyFlowCtrl = pInfo->notifyFlowCtrl; pNode->notifyFileSynced = pInfo->notifyFileSynced; + pNode->getVersion = pInfo->getVersion; pNode->selfIndex = -1; pNode->vgId = pInfo->vgId; @@ -225,7 +226,7 @@ int64_t syncStart(const SSyncInfo *pInfo) { } if (pNode->selfIndex < 0) { - sInfo("vgId:%d, this node is not configured", pNode->vgId); + sError("vgId:%d, this node is not configured", pNode->vgId); terrno = TSDB_CODE_SYN_INVALID_CONFIG; syncStop(pNode->rid); return -1; @@ -497,7 +498,6 @@ int32_t syncDecPeerRef(SSyncPeer *pPeer) { taosReleaseRef(tsSyncRefId, pPeer->pSyncNode->rid); sDebug("%s, resource is freed", pPeer->id); - tfree(pPeer->watchFd); tfree(pPeer); return 0; } @@ -540,7 +540,7 @@ static SSyncPeer *syncAddPeer(SSyncNode *pNode, const SNodeInfo *pInfo) { pPeer->ip = ip; pPeer->port = pInfo->nodePort; pPeer->fqdn[sizeof(pPeer->fqdn) - 1] = 0; - snprintf(pPeer->id, sizeof(pPeer->id), "vgId:%d, peer:%s:%u", pNode->vgId, pPeer->fqdn, pPeer->port); + snprintf(pPeer->id, sizeof(pPeer->id), "vgId:%d, nodeId:%d", pNode->vgId, pPeer->nodeId); pPeer->peerFd = -1; pPeer->syncFd = -1; @@ -548,7 +548,7 @@ static SSyncPeer *syncAddPeer(SSyncNode *pNode, const SNodeInfo *pInfo) { pPeer->pSyncNode = pNode; pPeer->refCount = 1; - sInfo("%s, it is configured", pPeer->id); + sInfo("%s, it is configured, ep:%s:%u", pPeer->id, pPeer->fqdn, pPeer->port); int32_t ret = strcmp(pPeer->fqdn, tsNodeFqdn); if (pPeer->nodeId == 0 || (ret > 0) || (ret == 0 && pPeer->port > tsSyncPort)) { int32_t checkMs = 100 + (pNode->vgId * 10) % 100; @@ -1134,7 +1134,7 @@ static void syncProcessIncommingConnection(int32_t connFd, uint32_t sourceIp) { pPeer = (i < pNode->replica) ? pNode->peerInfo[i] : NULL; if (pPeer == NULL) { - sError("vgId:%d, peer:%s not configured", pNode->vgId, firstPkt.fqdn); + sError("vgId:%d, peer:%s:%u not configured", pNode->vgId, firstPkt.fqdn, firstPkt.port); taosCloseSocket(connFd); // syncSendVpeerCfgMsg(sync); } else { @@ -1143,8 +1143,7 @@ static void syncProcessIncommingConnection(int32_t connFd, uint32_t sourceIp) { pPeer->syncFd = connFd; syncCreateRestoreDataThread(pPeer); } else { - sDebug("%s, TCP connection is already up(pfd:%d), close one, new pfd:%d sfd:%d", pPeer->id, pPeer->peerFd, connFd, - pPeer->syncFd); + sDebug("%s, TCP connection is up, pfd:%d sfd:%d, old pfd:%d", pPeer->id, connFd, pPeer->syncFd, pPeer->peerFd); syncClosePeerConn(pPeer); pPeer->peerFd = connFd; pPeer->pConn = taosAllocateTcpConn(tsTcpPool, pPeer, connFd); diff --git a/src/sync/src/syncRestore.c b/src/sync/src/syncRestore.c index ed6b63c92deaa1f077311599ff3e88469bd77593..d156c93865a3f4757cd9c9af0aff7b6789f59822 100644 --- a/src/sync/src/syncRestore.c +++ b/src/sync/src/syncRestore.c @@ -52,12 +52,12 @@ static void syncRemoveExtraFile(SSyncPeer *pPeer, int32_t sindex, int32_t eindex static int32_t syncRestoreFile(SSyncPeer *pPeer, uint64_t *fversion) { SSyncNode *pNode = pPeer->pSyncNode; - SFileInfo minfo; memset(&minfo, 0, sizeof(minfo)); /* = {0}; */ // master file info - SFileInfo sinfo; memset(&sinfo, 0, sizeof(sinfo)); /* = {0}; */ // slave file info - SFileAck fileAck; + SFileInfo minfo; memset(&minfo, 0, sizeof(SFileInfo)); /* = {0}; */ + SFileInfo sinfo; memset(&sinfo, 0, sizeof(SFileInfo)); /* = {0}; */ + SFileAck fileAck = {0}; int32_t code = -1; char name[TSDB_FILENAME_LEN * 2] = {0}; - uint32_t pindex = 0; // index in last restore + uint32_t pindex = 0; // index in last restore bool fileChanged = false; *fversion = 0; @@ -134,7 +134,7 @@ static int32_t syncRestoreFile(SSyncPeer *pPeer, uint64_t *fversion) { // data file is changed, code shall be set to 1 *fversion = minfo.fversion; code = 1; - sDebug("%s, file changed while restore file", pPeer->id); + sDebug("%s, file changed after restore file, fver:%" PRIu64, pPeer->id, *fversion); } if (code < 0) { @@ -160,7 +160,7 @@ static int32_t syncRestoreWal(SSyncPeer *pPeer) { } if (pHead->len == 0) { - sDebug("%s, wal is synced over", pPeer->id); + sDebug("%s, wal is synced over, last wver:%" PRIu64, pPeer->id, lastVer); code = 0; break; } // wal sync over diff --git a/src/sync/src/syncRetrieve.c b/src/sync/src/syncRetrieve.c index 060badba9dc188c2741d280d600c709ee2740099..36b197dd46484c7ce676391aea5dc4efaf00b9d8 100644 --- a/src/sync/src/syncRetrieve.c +++ b/src/sync/src/syncRetrieve.c @@ -16,6 +16,7 @@ #define _DEFAULT_SOURCE #include #include "os.h" +#include "taoserror.h" #include "tlog.h" #include "tutil.h" #include "tglobal.h" @@ -25,110 +26,102 @@ #include "tsync.h" #include "syncInt.h" -static int32_t syncAddIntoWatchList(SSyncPeer *pPeer, char *name) { - sDebug("%s, start to monitor:%s", pPeer->id, name); - - if (pPeer->notifyFd <= 0) { - pPeer->watchNum = 0; - pPeer->notifyFd = inotify_init1(IN_NONBLOCK); - if (pPeer->notifyFd < 0) { - sError("%s, failed to init inotify since %s", pPeer->id, strerror(errno)); - return -1; - } +static int32_t syncGetWalVersion(SSyncNode *pNode, SSyncPeer *pPeer) { + uint64_t fver, wver; + int32_t code = (*pNode->getVersion)(pNode->vgId, &fver, &wver); + if (code != 0) { + sDebug("%s, vnode is commiting while retrieve, last wver:%" PRIu64, pPeer->id, pPeer->lastWalVer); + return -1; + } - if (pPeer->watchFd == NULL) pPeer->watchFd = malloc(sizeof(int32_t) * tsMaxWatchFiles); - if (pPeer->watchFd == NULL) { - sError("%s, failed to allocate watchFd", pPeer->id); - return -1; - } + pPeer->lastWalVer = wver; + return code; +} - memset(pPeer->watchFd, -1, sizeof(int32_t) * tsMaxWatchFiles); +static bool syncIsWalModified(SSyncNode *pNode, SSyncPeer *pPeer) { + uint64_t fver, wver; + int32_t code = (*pNode->getVersion)(pNode->vgId, &fver, &wver); + if (code != 0) { + sDebug("%s, vnode is commiting while retrieve, last wver:%" PRIu64, pPeer->id, pPeer->lastWalVer); + return true; } - int32_t *wd = pPeer->watchFd + pPeer->watchNum; - - if (*wd >= 0) { - if (inotify_rm_watch(pPeer->notifyFd, *wd) < 0) { - sError("%s, failed to remove wd:%d since %s", pPeer->id, *wd, strerror(errno)); - return -1; - } + if (wver != pPeer->lastWalVer) { + sDebug("%s, wal is modified while retrieve, wver:%" PRIu64 ", last:%" PRIu64, pPeer->id, wver, pPeer->lastWalVer); + return true; } - *wd = inotify_add_watch(pPeer->notifyFd, name, IN_MODIFY | IN_DELETE); - if (*wd == -1) { - sError("%s, failed to add %s since %s", pPeer->id, name, strerror(errno)); + return false; +} + +static int32_t syncGetFileVersion(SSyncNode *pNode, SSyncPeer *pPeer) { + uint64_t fver, wver; + int32_t code = (*pNode->getVersion)(pNode->vgId, &fver, &wver); + if (code != 0) { + sDebug("%s, vnode is commiting while retrieve, last fver:%" PRIu64, pPeer->id, pPeer->lastFileVer); return -1; - } else { - sDebug("%s, monitor %s, wd:%d watchNum:%d", pPeer->id, name, *wd, pPeer->watchNum); } - pPeer->watchNum = (pPeer->watchNum + 1) % tsMaxWatchFiles; - - return 0; + pPeer->lastFileVer = fver; + return code; } -static int32_t syncAreFilesModified(SSyncPeer *pPeer) { - if (pPeer->notifyFd <= 0) return 0; - - char buf[2048]; - int32_t len = read(pPeer->notifyFd, buf, sizeof(buf)); - if (len < 0 && errno != EAGAIN) { - sError("%s, failed to read notify FD since %s", pPeer->id, strerror(errno)); - return -1; +static bool syncAreFilesModified(SSyncNode *pNode, SSyncPeer *pPeer) { + uint64_t fver, wver; + int32_t code = (*pNode->getVersion)(pNode->vgId, &fver, &wver); + if (code != 0) { + sDebug("%s, vnode is commiting while retrieve, last fver:%" PRIu64, pPeer->id, pPeer->lastFileVer); + pPeer->fileChanged = 1; + return true; } - int32_t code = 0; - if (len > 0) { - const struct inotify_event *event; - char *ptr; - for (ptr = buf; ptr < buf + len; ptr += sizeof(struct inotify_event) + event->len) { - event = (const struct inotify_event *)ptr; - if ((event->mask & IN_MODIFY) || (event->mask & IN_DELETE)) { - sDebug("%s, processed file is changed", pPeer->id); - pPeer->fileChanged = 1; - code = 1; - break; - } - } + if (fver != pPeer->lastFileVer) { + sDebug("%s, files are modified while retrieve, fver:%" PRIu64 ", last:%" PRIu64, pPeer->id, fver, pPeer->lastFileVer); + pPeer->fileChanged = 1; + return true; } - return code; + pPeer->fileChanged = 0; + return false; } static int32_t syncRetrieveFile(SSyncPeer *pPeer) { SSyncNode *pNode = pPeer->pSyncNode; - SFileInfo fileInfo; - SFileAck fileAck; + SFileInfo fileInfo; memset(&fileInfo, 0, sizeof(SFileInfo)); + SFileAck fileAck = {0}; int32_t code = -1; char name[TSDB_FILENAME_LEN * 2] = {0}; - memset(&fileInfo, 0, sizeof(fileInfo)); - memset(&fileAck, 0, sizeof(fileAck)); + if (syncGetFileVersion(pNode, pPeer) < 0) return -1; while (1) { // retrieve file info fileInfo.name[0] = 0; + fileInfo.size = 0; fileInfo.magic = (*pNode->getFileInfo)(pNode->vgId, fileInfo.name, &fileInfo.index, TAOS_SYNC_MAX_INDEX, &fileInfo.size, &fileInfo.fversion); // fileInfo.size = htonl(size); + sDebug("%s, file:%s info is sent, size:%" PRId64, pPeer->id, fileInfo.name, fileInfo.size); // send the file info int32_t ret = taosWriteMsg(pPeer->syncFd, &(fileInfo), sizeof(fileInfo)); if (ret < 0) { + code = -1; sError("%s, failed to write file:%s info while retrieve file since %s", pPeer->id, fileInfo.name, strerror(errno)); break; } // if no file anymore, break if (fileInfo.magic == 0 || fileInfo.name[0] == 0) { - sDebug("%s, no more files to sync", pPeer->id); code = 0; + sDebug("%s, no more files to sync", pPeer->id); break; } // wait for the ack from peer ret = taosReadMsg(pPeer->syncFd, &fileAck, sizeof(fileAck)); if (ret < 0) { + code = -1; sError("%s, failed to read file:%s ack while retrieve file since %s", pPeer->id, fileInfo.name, strerror(errno)); break; } @@ -136,15 +129,6 @@ static int32_t syncRetrieveFile(SSyncPeer *pPeer) { // set the peer sync version pPeer->sversion = fileInfo.fversion; - // get the full path to file - snprintf(name, sizeof(name), "%s/%s", pNode->path, fileInfo.name); - - // add the file into watch list - if (syncAddIntoWatchList(pPeer, name) < 0) { - sError("%s, failed to watch file:%s while retrieve file since %s", pPeer->id, fileInfo.name, strerror(errno)); - break; - } - // if sync is not required, continue if (fileAck.sync == 0) { fileInfo.index++; @@ -152,9 +136,13 @@ static int32_t syncRetrieveFile(SSyncPeer *pPeer) { continue; } + // get the full path to file + snprintf(name, sizeof(name), "%s/%s", pNode->path, fileInfo.name); + // send the file to peer int32_t sfd = open(name, O_RDONLY); if (sfd < 0) { + code = -1; sError("%s, failed to open file:%s while retrieve file since %s", pPeer->id, fileInfo.name, strerror(errno)); break; } @@ -162,138 +150,112 @@ static int32_t syncRetrieveFile(SSyncPeer *pPeer) { ret = taosSendFile(pPeer->syncFd, sfd, NULL, fileInfo.size); close(sfd); if (ret < 0) { + code = -1; sError("%s, failed to send file:%s while retrieve file since %s", pPeer->id, fileInfo.name, strerror(errno)); break; } - sDebug("%s, %s is sent, size:%" PRId64, pPeer->id, name, fileInfo.size); + sDebug("%s, file:%s is sent, size:%" PRId64, pPeer->id, fileInfo.name, fileInfo.size); fileInfo.index++; // check if processed files are modified - if (syncAreFilesModified(pPeer) != 0) { - sInfo("%s, file:%s are modified while retrieve file since %s", pPeer->id, fileInfo.name, strerror(errno)); + if (syncAreFilesModified(pNode, pPeer)) { + code = -1; break; } } - if (code < 0) { - sError("%s, failed to retrieve file", pPeer->id); + if (code != TSDB_CODE_SUCCESS) { + sError("%s, failed to retrieve file, code:0x%x", pPeer->id, code); } return code; } -/* if only a partial record is read out, set the IN_MODIFY flag in event, - so upper layer will reload the file to get a complete record */ -static int32_t syncReadOneWalRecord(int32_t sfd, SWalHead *pHead, uint32_t *pEvent) { - int32_t ret; +// if only a partial record is read out, upper layer will reload the file to get a complete record +static int32_t syncReadOneWalRecord(int32_t sfd, SWalHead *pHead) { + int32_t ret = read(sfd, pHead, sizeof(SWalHead)); + if (ret < 0) { + sError("sfd:%d, failed to read wal head since %s, ret:%d", sfd, strerror(errno), ret); + return -1; + } - ret = read(sfd, pHead, sizeof(SWalHead)); - if (ret < 0) return -1; - if (ret == 0) return 0; + if (ret == 0) { + sTrace("sfd:%d, read to the end of file, ret:%d", sfd, ret); + return 0; + } if (ret != sizeof(SWalHead)) { // file is not at end yet, it shall be reloaded - *pEvent = *pEvent | IN_MODIFY; + sDebug("sfd:%d, a partial wal head is read out, ret:%d", sfd, ret); return 0; } assert(pHead->len <= TSDB_MAX_WAL_SIZE); ret = read(sfd, pHead->cont, pHead->len); - if (ret < 0) return -1; + if (ret < 0) { + sError("sfd:%d, failed to read wal content since %s, ret:%d", sfd, strerror(errno), ret); + return -1; + } if (ret != pHead->len) { // file is not at end yet, it shall be reloaded - *pEvent = *pEvent | IN_MODIFY; + sDebug("sfd:%d, a partial wal conetnt is read out, ret:%d", sfd, ret); return 0; } return sizeof(SWalHead) + pHead->len; } -static int32_t syncMonitorLastWal(SSyncPeer *pPeer, char *name) { - pPeer->watchNum = 0; - taosClose(pPeer->notifyFd); - pPeer->notifyFd = inotify_init1(IN_NONBLOCK); - if (pPeer->notifyFd < 0) { - sError("%s, failed to init inotify since %s", pPeer->id, strerror(errno)); - return -1; - } - - if (pPeer->watchFd == NULL) pPeer->watchFd = malloc(sizeof(int32_t) * tsMaxWatchFiles); - if (pPeer->watchFd == NULL) { - sError("%s, failed to allocate watchFd", pPeer->id); - return -1; - } - - memset(pPeer->watchFd, -1, sizeof(int32_t) * tsMaxWatchFiles); - int32_t *wd = pPeer->watchFd; - - *wd = inotify_add_watch(pPeer->notifyFd, name, IN_MODIFY | IN_CLOSE_WRITE); - if (*wd == -1) { - sError("%s, failed to watch last wal since %s", pPeer->id, strerror(errno)); +static int32_t syncRetrieveLastWal(SSyncPeer *pPeer, char *name, uint64_t fversion, int64_t offset) { + int32_t sfd = open(name, O_RDONLY); + if (sfd < 0) { + sError("%s, failed to open wal:%s for retrieve since:%s", pPeer->id, name, tstrerror(errno)); return -1; } - return 0; -} - -static int32_t syncCheckLastWalChanges(SSyncPeer *pPeer, uint32_t *pEvent) { - char buf[2048]; - int32_t len = read(pPeer->notifyFd, buf, sizeof(buf)); - if (len < 0 && errno != EAGAIN) { - sError("%s, failed to read notify FD since %s", pPeer->id, strerror(errno)); + int32_t code = taosLSeek(sfd, offset, SEEK_SET); + if (code < 0) { + sError("%s, failed to seek %" PRId64 " in wal:%s for retrieve since:%s", pPeer->id, offset, name, tstrerror(errno)); + close(sfd); return -1; } - if (len == 0) return 0; - - struct inotify_event *event; - for (char *ptr = buf; ptr < buf + len; ptr += sizeof(struct inotify_event) + event->len) { - event = (struct inotify_event *)ptr; - if (event->mask & IN_MODIFY) *pEvent = *pEvent | IN_MODIFY; - if (event->mask & IN_CLOSE_WRITE) *pEvent = *pEvent | IN_CLOSE_WRITE; - } - - if (pEvent != 0) sDebug("%s, last wal event:0x%x", pPeer->id, *pEvent); + sDebug("%s, retrieve last wal:%s, offset:%" PRId64 " fver:%" PRIu64, pPeer->id, name, offset, fversion); - return 0; -} - -static int32_t syncRetrieveLastWal(SSyncPeer *pPeer, char *name, uint64_t fversion, int64_t offset, uint32_t *pEvent) { SWalHead *pHead = malloc(SYNC_MAX_SIZE); - int32_t code = -1; int32_t bytes = 0; - int32_t sfd; - - sfd = open(name, O_RDONLY); - if (sfd < 0) { - free(pHead); - return -1; - } - - (void)lseek(sfd, offset, SEEK_SET); - sDebug("%s, retrieve last wal, offset:%" PRId64 " fver:%" PRIu64, pPeer->id, offset, fversion); while (1) { - int32_t wsize = syncReadOneWalRecord(sfd, pHead, pEvent); - if (wsize < 0) break; - if (wsize == 0) { - code = 0; + code = syncReadOneWalRecord(sfd, pHead); + if (code < 0) { + sError("%s, failed to read one record from wal:%s", pPeer->id, name); + break; + } + + if (code == 0) { + code = bytes; + sDebug("%s, read to the end of wal, bytes:%d", pPeer->id, bytes); break; } sTrace("%s, last wal is forwarded, hver:%" PRIu64, pPeer->id, pHead->version); + + int32_t wsize = code; int32_t ret = taosWriteMsg(pPeer->syncFd, pHead, wsize); - if (ret != wsize) break; - pPeer->sversion = pHead->version; + if (ret != wsize) { + code = -1; + sError("%s, failed to forward wal since %s, hver:%" PRIu64, pPeer->id, strerror(errno), pHead->version); + break; + } + pPeer->sversion = pHead->version; bytes += wsize; if (pHead->version >= fversion && fversion > 0) { code = 0; - bytes = 0; + sDebug("%s, retrieve wal finished, hver:%" PRIu64 " fver:%" PRIu64, pPeer->id, pHead->version, fversion); break; } } @@ -301,92 +263,62 @@ static int32_t syncRetrieveLastWal(SSyncPeer *pPeer, char *name, uint64_t fversi free(pHead); close(sfd); - if (code == 0) return bytes; - return -1; + return code; } static int32_t syncProcessLastWal(SSyncPeer *pPeer, char *wname, int64_t index) { SSyncNode *pNode = pPeer->pSyncNode; - int32_t code = -1; - char fname[TSDB_FILENAME_LEN * 2]; // full path to wal file + int32_t once = 0; // last WAL has once ever been processed + int64_t offset = 0; + uint64_t fversion = 0; + char fname[TSDB_FILENAME_LEN * 2] = {0}; // full path to wal file - if (syncAreFilesModified(pPeer) != 0) return -1; + // get full path to wal file + snprintf(fname, sizeof(fname), "%s/%s", pNode->path, wname); + sDebug("%s, start to retrieve last wal:%s", pPeer->id, fname); while (1) { - int32_t once = 0; // last WAL has once ever been processed - int64_t offset = 0; - uint64_t fversion = 0; - uint32_t event = 0; - - // get full path to wal file - snprintf(fname, sizeof(fname), "%s/%s", pNode->path, wname); - sDebug("%s, start to retrieve last wal:%s", pPeer->id, fname); - - // monitor last wal - if (syncMonitorLastWal(pPeer, fname) < 0) break; - - while (1) { - int32_t bytes = syncRetrieveLastWal(pPeer, fname, fversion, offset, &event); - if (bytes < 0) break; - - // check file changes - if (syncCheckLastWalChanges(pPeer, &event) < 0) break; + if (syncAreFilesModified(pNode, pPeer)) return -1; + if (syncGetWalVersion(pNode, pPeer) < 0) return -1; - // if file is not updated or updated once, set the fversion and sstatus - if (((event & IN_MODIFY) == 0) || once) { - if (fversion == 0) { - pPeer->sstatus = TAOS_SYNC_STATUS_CACHE; // start to forward pkt - sDebug("%s, fversion is 0 then set sstatus:%s", pPeer->id, syncStatus[pPeer->sstatus]); - fversion = nodeVersion; // must read data to fversion - } - } + int32_t bytes = syncRetrieveLastWal(pPeer, fname, fversion, offset); + if (bytes < 0) { + sDebug("%s, failed to retrieve last wal", pPeer->id); + return bytes; + } - // if all data up to fversion is read out, it is over - if (pPeer->sversion >= fversion && fversion > 0) { - code = 0; - sDebug("%s, data up to fver:%" PRIu64 " has been read out, bytes:%d", pPeer->id, fversion, bytes); - break; - } + // check file changes + bool walModified = syncIsWalModified(pNode, pPeer); - // if all data are read out, and no update - if ((bytes == 0) && ((event & IN_MODIFY) == 0)) { - // wal file is closed, break - if (event & IN_CLOSE_WRITE) { - code = 0; - sDebug("%s, current wal is closed", pPeer->id); - break; - } - - // wal not closed, it means some data not flushed to disk, wait for a while - usleep(10000); + // if file is not updated or updated once, set the fversion and sstatus + if (!walModified || once) { + if (fversion == 0) { + pPeer->sstatus = TAOS_SYNC_STATUS_CACHE; // start to forward pkt + fversion = nodeVersion; // must read data to fversion + sDebug("%s, set sstatus:%s and fver:%" PRIu64, pPeer->id, syncStatus[pPeer->sstatus], fversion); } - - // if bytes>0, file is updated, or fversion is not reached but file still open, read again - once = 1; - offset += bytes; - sDebug("%s, retrieve last wal, bytes:%d", pPeer->id, bytes); - event = event & (~IN_MODIFY); // clear IN_MODIFY flag } - if (code < 0) break; - if (pPeer->sversion >= fversion && fversion > 0) break; + // if all data up to fversion is read out, it is over + if (pPeer->sversion >= fversion && fversion > 0) { + sDebug("%s, data up to fver:%" PRIu64 " has been read out, bytes:%d sver:%" PRIu64, pPeer->id, fversion, bytes, + pPeer->sversion); + return 0; + } - index++; - wname[0] = 0; - code = (*pNode->getWalInfo)(pNode->vgId, wname, &index); - if (code < 0) break; - if (wname[0] == 0) { - code = 0; - break; + // if all data are read out, and no update + if (bytes == 0 && !walModified) { + // wal not closed, it means some data not flushed to disk, wait for a while + usleep(10000); } - // current last wal is closed, there is a new one - sDebug("%s, last wal is closed, try new one", pPeer->id); + // if bytes > 0, file is updated, or fversion is not reached but file still open, read again + once = 1; + offset += bytes; + sDebug("%s, continue retrieve last wal, bytes:%d offset:%" PRId64, pPeer->id, bytes, offset); } - taosClose(pPeer->notifyFd); - - return code; + return -1; } static int32_t syncRetrieveWal(SSyncPeer *pPeer) { @@ -394,7 +326,6 @@ static int32_t syncRetrieveWal(SSyncPeer *pPeer) { char fname[TSDB_FILENAME_LEN * 3]; char wname[TSDB_FILENAME_LEN * 2]; int32_t size; - struct stat fstat; int32_t code = -1; int64_t index = 0; @@ -402,9 +333,14 @@ static int32_t syncRetrieveWal(SSyncPeer *pPeer) { // retrieve wal info wname[0] = 0; code = (*pNode->getWalInfo)(pNode->vgId, wname, &index); - if (code < 0) break; // error + if (code < 0) { + sError("%s, failed to get wal info since:%s, code:0x%x", pPeer->id, strerror(errno), code); + break; + } + if (wname[0] == 0) { // no wal file - sDebug("%s, no wal file", pPeer->id); + code = 0; + sDebug("%s, no wal file anymore", pPeer->id); break; } @@ -416,22 +352,35 @@ static int32_t syncRetrieveWal(SSyncPeer *pPeer) { // get the full path to wal file snprintf(fname, sizeof(fname), "%s/%s", pNode->path, wname); - // send wal file, - // inotify is not required, old wal file won't be modified, even remove is ok - if (stat(fname, &fstat) < 0) break; - size = fstat.st_size; + // send wal file, old wal file won't be modified, even remove is ok + struct stat fstat; + if (stat(fname, &fstat) < 0) { + code = -1; + sDebug("%s, failed to stat wal:%s for retrieve since %s, code:0x%x", pPeer->id, fname, strerror(errno), code); + break; + } + size = fstat.st_size; sDebug("%s, retrieve wal:%s size:%d", pPeer->id, fname, size); + int32_t sfd = open(fname, O_RDONLY); - if (sfd < 0) break; + if (sfd < 0) { + code = -1; + sError("%s, failed to open wal:%s for retrieve since %s, code:0x%x", pPeer->id, fname, strerror(errno), code); + break; + } code = taosSendFile(pPeer->syncFd, sfd, NULL, size); close(sfd); - if (code < 0) break; - - index++; + if (code < 0) { + sError("%s, failed to send wal:%s for retrieve since %s, code:0x%x", pPeer->id, fname, strerror(errno), code); + break; + } - if (syncAreFilesModified(pPeer) != 0) break; + if (syncAreFilesModified(pNode, pPeer)) { + code = -1; + break; + } } if (code == 0) { @@ -440,9 +389,9 @@ static int32_t syncRetrieveWal(SSyncPeer *pPeer) { SWalHead walHead; memset(&walHead, 0, sizeof(walHead)); - code = taosWriteMsg(pPeer->syncFd, &walHead, sizeof(walHead)); + taosWriteMsg(pPeer->syncFd, &walHead, sizeof(walHead)); } else { - sError("%s, failed to send wal since %s", pPeer->id, strerror(errno)); + sError("%s, failed to send wal since %s, code:0x%x", pPeer->id, strerror(errno), code); } return code; @@ -481,18 +430,19 @@ static int32_t syncRetrieveDataStepByStep(SSyncPeer *pPeer) { pPeer->sversion = 0; pPeer->sstatus = TAOS_SYNC_STATUS_FILE; - sInfo("%s, start to retrieve file, set sstatus:%s", pPeer->id, syncStatus[pPeer->sstatus]); - if (syncRetrieveFile(pPeer) < 0) { - sError("%s, failed to retrieve file", pPeer->id); + sInfo("%s, start to retrieve files, set sstatus:%s", pPeer->id, syncStatus[pPeer->sstatus]); + if (syncRetrieveFile(pPeer) != 0) { + sError("%s, failed to retrieve files", pPeer->id); return -1; } // if no files are synced, there must be wal to sync, sversion must be larger than one if (pPeer->sversion == 0) pPeer->sversion = 1; - sInfo("%s, start to retrieve wal", pPeer->id); - if (syncRetrieveWal(pPeer) < 0) { - sError("%s, failed to retrieve wal", pPeer->id); + sInfo("%s, start to retrieve wals", pPeer->id); + int32_t code = syncRetrieveWal(pPeer); + if (code != 0) { + sError("%s, failed to retrieve wals, code:0x%x", pPeer->id, code); return -1; } @@ -506,7 +456,6 @@ void *syncRetrieveData(void *param) { if (pNode->notifyFlowCtrl) (*pNode->notifyFlowCtrl)(pNode->vgId, pPeer->numOfRetrieves); - pPeer->fileChanged = 0; pPeer->syncFd = taosOpenTcpClientSocket(pPeer->ip, pPeer->port, 0); if (pPeer->syncFd < 0) { sError("%s, failed to open socket to sync", pPeer->id); @@ -529,7 +478,6 @@ void *syncRetrieveData(void *param) { } pPeer->fileChanged = 0; - taosClose(pPeer->notifyFd); taosClose(pPeer->syncFd); syncDecPeerRef(pPeer); diff --git a/src/tsdb/src/tsdbCommitQueue.c b/src/tsdb/src/tsdbCommitQueue.c index c86b8f32b7ff6bfb30e7b734c7b38504200e44e6..75a2cbcb8deb869922295e623144569e90a0cecf 100644 --- a/src/tsdb/src/tsdbCommitQueue.c +++ b/src/tsdb/src/tsdbCommitQueue.c @@ -14,6 +14,7 @@ */ #include "os.h" +#include "tglobal.h" #include "tlist.h" #include "tref.h" #include "tsdbMain.h" @@ -36,7 +37,8 @@ static void *tsdbLoopCommit(void *arg); SCommitQueue tsCommitQueue = {0}; -int tsdbInitCommitQueue(int nthreads) { +int tsdbInitCommitQueue() { + int nthreads = tsNumOfCommitThreads; SCommitQueue *pQueue = &tsCommitQueue; if (nthreads < 1) nthreads = 1; diff --git a/src/tsdb/src/tsdbFile.c b/src/tsdb/src/tsdbFile.c index 03c50d42f7324c1cbe7889138bced9d9d986c424..5d8933d14142e4c569808fd932efbed63a9ed8e9 100644 --- a/src/tsdb/src/tsdbFile.c +++ b/src/tsdb/src/tsdbFile.c @@ -248,8 +248,13 @@ SFileGroup *tsdbCreateFGroupIfNeed(STsdbRepo *pRepo, char *dataDir, int fid) { if (pGroup == NULL) { // if not exists, create one pFGroup->fileId = fid; for (int type = 0; type < TSDB_FILE_TYPE_MAX; type++) { - if (tsdbCreateFile(&pFGroup->files[type], pRepo, fid, type) < 0) - goto _err; + if (tsdbCreateFile(&pFGroup->files[type], pRepo, fid, type) < 0) { + for (int i = type; i >= 0; i--) { + remove(pFGroup->files[i].fname); + } + + return NULL; + } } pthread_rwlock_wrlock(&pFileH->fhlock); @@ -261,10 +266,6 @@ SFileGroup *tsdbCreateFGroupIfNeed(STsdbRepo *pRepo, char *dataDir, int fid) { } return pGroup; - -_err: - for (int type = 0; type < TSDB_FILE_TYPE_MAX; type++) tsdbDestroyFile(&pGroup->files[type]); - return NULL; } void tsdbInitFileGroupIter(STsdbFileH *pFileH, SFileGroupIter *pIter, int direction) { diff --git a/src/tsdb/src/tsdbMemTable.c b/src/tsdb/src/tsdbMemTable.c index 71944c87c6d68d530e07a13a05bf5ac89ee3754d..999e2deb413eb13cc7119ee9042a260366c5f9a3 100644 --- a/src/tsdb/src/tsdbMemTable.c +++ b/src/tsdb/src/tsdbMemTable.c @@ -786,9 +786,9 @@ static int tsdbCheckTableSchema(STsdbRepo *pRepo, SSubmitBlk *pBlock, STable *pT if (tsdbGetTableSchemaImpl(pTable, false, false, pBlock->sversion) == NULL) { tsdbError("vgId:%d invalid submit schema version %d to table %s tid %d from client", REPO_ID(pRepo), pBlock->sversion, TABLE_CHAR_NAME(pTable), TABLE_TID(pTable)); + terrno = TSDB_CODE_TDB_IVD_TB_SCHEMA_VERSION; + return -1; } - terrno = TSDB_CODE_TDB_IVD_TB_SCHEMA_VERSION; - return -1; } return 0; @@ -871,10 +871,10 @@ static void tsdbFreeRows(STsdbRepo *pRepo, void **rows, int rowCounter) { listNEles(pRepo->mem->bufBlockList), pBufBlock->offset, pBufBlock->remain); if (pBufBlock->offset == 0) { // return the block to buffer pool - tsdbLockRepo(pRepo); + if (tsdbLockRepo(pRepo) < 0) return; SListNode *pNode = tdListPopTail(pRepo->mem->bufBlockList); tdListPrependNode(pBufPool->bufBlockList, pNode); - tsdbUnlockRepo(pRepo); + if (tsdbUnlockRepo(pRepo) < 0) return; } } else { ASSERT(listNEles(pRepo->mem->extraBuffList) > 0); diff --git a/src/tsdb/src/tsdbRWHelper.c b/src/tsdb/src/tsdbRWHelper.c index 5b65b2185a873b9ea241b0b0e4147a4e9d522e48..8e57066d2703b40c647e5f7df9065abe3a77ff20 100644 --- a/src/tsdb/src/tsdbRWHelper.c +++ b/src/tsdb/src/tsdbRWHelper.c @@ -917,6 +917,8 @@ static int tsdbInsertSuperBlock(SRWHelper *pHelper, SCompBlock *pCompBlock, int ASSERT(pHelper->pCompInfo->blocks[0].keyLast < pHelper->pCompInfo->blocks[1].keyFirst); } + ASSERT((blkIdx == pIdx->numOfBlocks -1) || (!pCompBlock->last)); + tsdbDebug("vgId:%d tid:%d a super block is inserted at index %d", REPO_ID(pHelper->pRepo), pHelper->tableInfo.tid, blkIdx); @@ -1042,6 +1044,8 @@ static int tsdbUpdateSuperBlock(SRWHelper *pHelper, SCompBlock *pCompBlock, int pIdx->maxKey = blockAtIdx(pHelper, pIdx->numOfBlocks - 1)->keyLast; pIdx->hasLast = (uint32_t)blockAtIdx(pHelper, pIdx->numOfBlocks - 1)->last; + ASSERT((blkIdx == pIdx->numOfBlocks-1) || (!pCompBlock->last)); + tsdbDebug("vgId:%d tid:%d a super block is updated at index %d", REPO_ID(pHelper->pRepo), pHelper->tableInfo.tid, blkIdx); @@ -1622,11 +1626,7 @@ static int tsdbProcessMergeCommit(SRWHelper *pHelper, SCommitIter *pCommitIter, pCfg->update); if (pDataCols->numOfRows == 0) break; - if (tblkIdx == pIdx->numOfBlocks - 1) { - if (tsdbWriteBlockToProperFile(pHelper, pDataCols, &compBlock) < 0) return -1; - } else { - if (tsdbWriteBlockToFile(pHelper, helperDataF(pHelper), pDataCols, &compBlock, false, true) < 0) return -1; - } + if (tsdbWriteBlockToFile(pHelper, helperDataF(pHelper), pDataCols, &compBlock, false, true) < 0) return -1; if (round == 0) { if (oBlock.last && pHelper->hasOldLastBlock) pHelper->hasOldLastBlock = false; diff --git a/src/util/inc/hash.h b/src/util/inc/hash.h index 42bc136584d618e62826c307b698746ca387ff41..b0319d3e139f72bdaff7993e2e46d61b82c4d0a5 100644 --- a/src/util/inc/hash.h +++ b/src/util/inc/hash.h @@ -31,16 +31,18 @@ extern "C" { typedef void (*_hash_free_fn_t)(void *param); typedef struct SHashNode { -// char *key; struct SHashNode *next; uint32_t hashVal; // the hash value of key uint32_t keyLen; // length of the key -// char *data; + size_t dataLen; // length of data + int8_t count; // reference count + int8_t removed; // flag to indicate removed + char data[]; } SHashNode; -#define GET_HASH_NODE_KEY(_n) ((char*)(_n) + sizeof(SHashNode)) -#define GET_HASH_NODE_DATA(_n) ((char*)(_n) + sizeof(SHashNode) + (_n)->keyLen) - +#define GET_HASH_NODE_KEY(_n) ((char*)(_n) + sizeof(SHashNode) + (_n)->dataLen) +#define GET_HASH_NODE_DATA(_n) ((char*)(_n) + sizeof(SHashNode)) +#define GET_HASH_PNODE(_n) ((char*)(_n) - sizeof(SHashNode)); typedef enum SHashLockTypeE { HASH_NO_LOCK = 0, HASH_ENTRY_LOCK = 1, @@ -65,15 +67,6 @@ typedef struct SHashObj { SArray *pMemBlock; // memory block allocated for SHashEntry } SHashObj; -typedef struct SHashMutableIterator { - SHashObj *pHashObj; - int32_t entryIndex; - SHashNode *pCur; - SHashNode *pNext; // current node can be deleted for mutable iterator, so keep the next one before return current - size_t numOfChecked; // already check number of elements in hash table - size_t numOfEntries; // number of entries while the iterator is created -} SHashMutableIterator; - /** * init the hash table * @@ -142,33 +135,9 @@ int32_t taosHashCondTraverse(SHashObj *pHashObj, bool (*fp)(void *, void *), voi */ void taosHashCleanup(SHashObj *pHashObj); -/** - * - * @param pHashObj - * @return - */ -SHashMutableIterator* taosHashCreateIter(SHashObj *pHashObj); - -/** - * - * @param iter - * @return - */ -bool taosHashIterNext(SHashMutableIterator *iter); - -/** - * - * @param iter - * @return - */ -void *taosHashIterGet(SHashMutableIterator *iter); - -/** - * - * @param iter - * @return - */ -void* taosHashDestroyIter(SHashMutableIterator* iter); +/* +void *SHashMutableIterator* taosHashCreateIter(SHashObj *pHashObj, void *); +*/ /** * @@ -179,6 +148,9 @@ int32_t taosHashGetMaxOverflowLinkLength(const SHashObj *pHashObj); size_t taosHashGetMemSize(const SHashObj *pHashObj); +void *taosHashIterate(SHashObj *pHashObj, void *p); +void taosHashCancelIterate(SHashObj *pHashObj, void *p); + #ifdef __cplusplus } #endif diff --git a/src/util/inc/tfile.h b/src/util/inc/tfile.h index 10b7c1df35caaa675c9334106785b91af3fdeb05..066040170e44c24539e29b5de5acd438e8b9b9d0 100644 --- a/src/util/inc/tfile.h +++ b/src/util/inc/tfile.h @@ -20,23 +20,26 @@ extern "C" { #endif -#include - // init taos file module -int32_t tfinit(); +int32_t tfInit(); // clean up taos file module -void tfcleanup(); +void tfCleanup(); // the same syntax as UNIX standard open/close/read/write // but FD is int64_t and will never be reused -int64_t tfopen(const char *pathname, int32_t flags); -int64_t tfclose(int64_t tfd); -int64_t tfwrite(int64_t tfd, void *buf, int64_t count); -int64_t tfread(int64_t tfd, void *buf, int64_t count); +int64_t tfOpen(const char *pathname, int32_t flags); +int64_t tfOpenM(const char *pathname, int32_t flags, mode_t mode); +int64_t tfClose(int64_t tfd); +int64_t tfWrite(int64_t tfd, void *buf, int64_t count); +int64_t tfRead(int64_t tfd, void *buf, int64_t count); +int32_t tfFsync(int64_t tfd); +bool tfValid(int64_t tfd); +int64_t tfLseek(int64_t tfd, int64_t offset, int32_t whence); +int32_t tfFtruncate(int64_t tfd, int64_t length); #ifdef __cplusplus } #endif -#endif // TDENGINE_TREF_H +#endif // TDENGINE_TFILE_H diff --git a/src/util/inc/tnettest.h b/src/util/inc/tnettest.h index 426df5cbb28b9c0fcada049c7242730359d2a3fc..b7585bd7155421d1f22e5f989dc7d1ae6f8be491 100644 --- a/src/util/inc/tnettest.h +++ b/src/util/inc/tnettest.h @@ -20,27 +20,7 @@ extern "C" { #endif -typedef struct CmdArguments { - char* host; - char* password; - char* user; - char* auth; - char* database; - char* timezone; - bool is_raw_time; - bool is_use_passwd; - char file[TSDB_FILENAME_LEN]; - char dir[TSDB_FILENAME_LEN]; - int threadNum; - char* commands; - int abort; - int port; - int endPort; - int pktLen; - char* netTestRole; -} CmdArguments; - -void taosNetTest(CmdArguments* args); +void taosNetTest(char *role, char *host, int port, int pkgLen); #ifdef __cplusplus } diff --git a/src/util/src/hash.c b/src/util/src/hash.c index 03a73424971ba7e4b77d508b12f71c729889a51a..0e3e0d3e2437f4595179c880a37642e047eb4331 100644 --- a/src/util/src/hash.c +++ b/src/util/src/hash.c @@ -76,7 +76,7 @@ static FORCE_INLINE int32_t taosHashCapacity(int32_t length) { static FORCE_INLINE SHashNode *doSearchInEntryList(SHashEntry *pe, const void *key, size_t keyLen, uint32_t hashVal) { SHashNode *pNode = pe->next; while (pNode) { - if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0)) { + if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0) && pNode->removed == 0) { assert(pNode->hashVal == hashVal); break; } @@ -114,15 +114,25 @@ static SHashNode *doCreateHashNode(const void *key, size_t keyLen, const void *p * @param dsize size of actual data * @return hash node */ -static FORCE_INLINE SHashNode *doUpdateHashNode(SHashEntry* pe, SHashNode* prev, SHashNode *pNode, SHashNode *pNewNode) { +static FORCE_INLINE SHashNode *doUpdateHashNode(SHashObj *pHashObj, SHashEntry* pe, SHashNode* prev, SHashNode *pNode, SHashNode *pNewNode) { assert(pNode->keyLen == pNewNode->keyLen); + + pNode->count--; if (prev != NULL) { prev->next = pNewNode; } else { pe->next = pNewNode; } - pNewNode->next = pNode->next; + if (pNode->count <= 0) { + pNewNode->next = pNode->next; + DO_FREE_HASH_NODE(pNode); + } else { + pNewNode->next = pNode; + pe->num++; + atomic_add_fetch_64(&pHashObj->size, 1); + } + return pNewNode; } @@ -139,11 +149,11 @@ static void pushfrontNodeInEntryList(SHashEntry *pEntry, SHashNode *pNode); * @param pIter * @return */ -static SHashNode *getNextHashNode(SHashMutableIterator *pIter); SHashObj *taosHashInit(size_t capacity, _hash_fn_t fn, bool update, SHashLockTypeE type) { - if (capacity == 0 || fn == NULL) { - return NULL; + assert(fn != NULL); + if (capacity == 0) { + capacity = 4; } SHashObj *pHashObj = (SHashObj *)calloc(1, sizeof(SHashObj)); @@ -213,7 +223,7 @@ int32_t taosHashPut(SHashObj *pHashObj, const void *key, size_t keyLen, void *da SHashNode* prev = NULL; while (pNode) { - if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0)) { + if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0) && pNode->removed == 0) { assert(pNode->hashVal == hashVal); break; } @@ -244,8 +254,7 @@ int32_t taosHashPut(SHashObj *pHashObj, const void *key, size_t keyLen, void *da } else { // not support the update operation, return error if (pHashObj->enableUpdate) { - doUpdateHashNode(pe, prev, pNode, pNewNode); - DO_FREE_HASH_NODE(pNode); + doUpdateHashNode(pHashObj, pe, prev, pNode, pNewNode); } else { DO_FREE_HASH_NODE(pNewNode); } @@ -335,22 +344,10 @@ int32_t taosHashRemoveWithData(SHashObj *pHashObj, const void *key, size_t keyLe int32_t slot = HASH_INDEX(hashVal, pHashObj->capacity); SHashEntry *pe = pHashObj->hashList[slot]; - // no data, return directly - if (pe->num == 0) { - __rd_unlock(&pHashObj->lock, pHashObj->type); - return -1; - } - if (pHashObj->type == HASH_ENTRY_LOCK) { taosWLockLatch(&pe->latch); } - if (pe->num == 0) { - assert(pe->next == NULL); - } else { - assert(pe->next != NULL); - } - // double check after locked if (pe->num == 0) { assert(pe->next == NULL); @@ -360,37 +357,37 @@ int32_t taosHashRemoveWithData(SHashObj *pHashObj, const void *key, size_t keyLe return -1; } + int code = -1; SHashNode *pNode = pe->next; - SHashNode *pRes = NULL; + SHashNode *prevNode = NULL; - // remove it - if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0)) { - pe->num -= 1; - pRes = pNode; - pe->next = pNode->next; - } else { - while (pNode->next != NULL) { - if (((pNode->next)->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY((pNode->next)), key, keyLen) == 0)) { - assert((pNode->next)->hashVal == hashVal); - break; - } + while (pNode) { + if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0) && pNode->removed == 0) + break; - pNode = pNode->next; - } + prevNode = pNode; + pNode = pNode->next; + } + if (pNode) { + code = 0; // it is found - if (pNode->next != NULL) { - pe->num -= 1; - pRes = pNode->next; - pNode->next = pNode->next->next; - } - } + pNode->count--; + pNode->removed = 1; + if (pNode->count <= 0) { + if (prevNode) { + prevNode->next = pNode->next; + } else { + pe->next = pNode->next; + } + + if (data) memcpy(data, GET_HASH_NODE_DATA(pNode), dsize); - if (pe->num == 0) { - assert(pe->next == NULL); - } else { - assert(pe->next != NULL); - } + pe->num--; + atomic_sub_fetch_64(&pHashObj->size, 1); + FREE_HASH_NODE(pHashObj, pNode); + } + } if (pHashObj->type == HASH_ENTRY_LOCK) { taosWUnLockLatch(&pe->latch); @@ -398,17 +395,7 @@ int32_t taosHashRemoveWithData(SHashObj *pHashObj, const void *key, size_t keyLe __rd_unlock(&pHashObj->lock, pHashObj->type); - if (data != NULL && pRes != NULL) { - memcpy(data, GET_HASH_NODE_DATA(pRes), dsize); - } - - if (pRes != NULL) { - atomic_sub_fetch_64(&pHashObj->size, 1); - FREE_HASH_NODE(pHashObj, pRes); - return 0; - } else { - return -1; - } + return code; } int32_t taosHashCondTraverse(SHashObj *pHashObj, bool (*fp)(void *, void *), void *param) { @@ -531,98 +518,6 @@ void taosHashCleanup(SHashObj *pHashObj) { free(pHashObj); } -SHashMutableIterator *taosHashCreateIter(SHashObj *pHashObj) { - SHashMutableIterator *pIter = calloc(1, sizeof(SHashMutableIterator)); - if (pIter == NULL) { - return NULL; - } - - pIter->pHashObj = pHashObj; - - // keep it in local variable, in case the resize operation expand the size - pIter->numOfEntries = pHashObj->capacity; - return pIter; -} - -bool taosHashIterNext(SHashMutableIterator *pIter) { - if (pIter == NULL) { - return false; - } - - size_t size = taosHashGetSize(pIter->pHashObj); - if (size == 0) { - return false; - } - - // check the first one - if (pIter->numOfChecked == 0) { - assert(pIter->pCur == NULL && pIter->pNext == NULL); - - while (1) { - SHashEntry *pEntry = pIter->pHashObj->hashList[pIter->entryIndex]; - if (pEntry->num == 0) { - assert(pEntry->next == NULL); - - pIter->entryIndex++; - continue; - } - - if (pIter->pHashObj->type == HASH_ENTRY_LOCK) { - taosRLockLatch(&pEntry->latch); - } - - pIter->pCur = pEntry->next; - - if (pIter->pCur->next) { - pIter->pNext = pIter->pCur->next; - - if (pIter->pHashObj->type == HASH_ENTRY_LOCK) { - taosRUnLockLatch(&pEntry->latch); - } - } else { - if (pIter->pHashObj->type == HASH_ENTRY_LOCK) { - taosRUnLockLatch(&pEntry->latch); - } - - pIter->pNext = getNextHashNode(pIter); - } - - break; - } - - pIter->numOfChecked++; - return true; - } else { - assert(pIter->pCur != NULL); - if (pIter->pNext) { - pIter->pCur = pIter->pNext; - } else { // no more data in the hash list - return false; - } - - pIter->numOfChecked++; - - if (pIter->pCur->next) { - pIter->pNext = pIter->pCur->next; - } else { - pIter->pNext = getNextHashNode(pIter); - } - - return true; - } -} - -void *taosHashIterGet(SHashMutableIterator *iter) { return (iter == NULL) ? NULL : GET_HASH_NODE_DATA(iter->pCur); } - -void *taosHashDestroyIter(SHashMutableIterator *iter) { - if (iter == NULL) { - return NULL; - } - - free(iter); - return NULL; -} - // for profile only int32_t taosHashGetMaxOverflowLinkLength(const SHashObj *pHashObj) { if (pHashObj == NULL || pHashObj->size == 0) { @@ -759,6 +654,8 @@ SHashNode *doCreateHashNode(const void *key, size_t keyLen, const void *pData, s pNewNode->keyLen = (uint32_t)keyLen; pNewNode->hashVal = hashVal; + pNewNode->dataLen = dsize; + pNewNode->count = 1; memcpy(GET_HASH_NODE_DATA(pNewNode), pData, dsize); memcpy(GET_HASH_NODE_KEY(pNewNode), key, keyLen); @@ -775,39 +672,136 @@ void pushfrontNodeInEntryList(SHashEntry *pEntry, SHashNode *pNode) { pEntry->num += 1; } -SHashNode *getNextHashNode(SHashMutableIterator *pIter) { - assert(pIter != NULL); +size_t taosHashGetMemSize(const SHashObj *pHashObj) { + if (pHashObj == NULL) { + return 0; + } + + return (pHashObj->capacity * (sizeof(SHashEntry) + POINTER_BYTES)) + sizeof(SHashNode) * taosHashGetSize(pHashObj) + sizeof(SHashObj); +} + +// release the pNode, return next pNode, and lock the current entry +static void *taosHashReleaseNode(SHashObj *pHashObj, void *p, int *slot) { - pIter->entryIndex++; - SHashNode *p = NULL; + SHashNode *pOld = (SHashNode *)GET_HASH_PNODE(p); + SHashNode *prevNode = NULL; - while (pIter->entryIndex < pIter->numOfEntries) { - SHashEntry *pEntry = pIter->pHashObj->hashList[pIter->entryIndex]; - if (pEntry->num == 0) { - pIter->entryIndex++; - continue; + *slot = HASH_INDEX(pOld->hashVal, pHashObj->capacity); + SHashEntry *pe = pHashObj->hashList[*slot]; + + // lock entry + if (pHashObj->type == HASH_ENTRY_LOCK) { + taosWLockLatch(&pe->latch); + } + + SHashNode *pNode = pe->next; + + while (pNode) { + if (pNode == pOld) + break; + + prevNode = pNode; + pNode = pNode->next; + } + + if (pNode) { + pNode = pNode->next; + while (pNode) { + if (pNode->removed == 0) break; + pNode = pNode->next; } - if (pIter->pHashObj->type == HASH_ENTRY_LOCK) { - taosRLockLatch(&pEntry->latch); + pOld->count--; + if (pOld->count <=0) { + if (prevNode) { + prevNode->next = pOld->next; + } else { + pe->next = pOld->next; + } + + pe->num--; + atomic_sub_fetch_64(&pHashObj->size, 1); + FREE_HASH_NODE(pHashObj, pOld); + } + } else { + uError("pNode:%p data:%p is not there!!!", pNode, p); + } + + return pNode; +} + +void *taosHashIterate(SHashObj *pHashObj, void *p) { + if (pHashObj == NULL) return NULL; + + int slot = 0; + char *data = NULL; + + // only add the read lock to disable the resize process + __rd_lock(&pHashObj->lock, pHashObj->type); + + SHashNode *pNode = NULL; + if (p) { + pNode = taosHashReleaseNode(pHashObj, p, &slot); + if (pNode == NULL) { + SHashEntry *pe = pHashObj->hashList[slot]; + if (pHashObj->type == HASH_ENTRY_LOCK) { + taosWUnLockLatch(&pe->latch); + } + + slot = slot + 1; } + } + + if (pNode == NULL) { + for (; slot < pHashObj->capacity; ++slot) { + SHashEntry *pe = pHashObj->hashList[slot]; + + // lock entry + if (pHashObj->type == HASH_ENTRY_LOCK) { + taosWLockLatch(&pe->latch); + } - p = pEntry->next; + pNode = pe->next; + while (pNode) { + if (pNode->removed == 0) break; + pNode = pNode->next; + } - if (pIter->pHashObj->type == HASH_ENTRY_LOCK) { - taosRUnLockLatch(&pEntry->latch); + if (pNode) break; + + if (pHashObj->type == HASH_ENTRY_LOCK) { + taosWUnLockLatch(&pe->latch); + } } + } - return p; + if (pNode) { + SHashEntry *pe = pHashObj->hashList[slot]; + pNode->count++; + data = GET_HASH_NODE_DATA(pNode); + if (pHashObj->type == HASH_ENTRY_LOCK) { + taosWUnLockLatch(&pe->latch); + } } - return NULL; + __rd_unlock(&pHashObj->lock, pHashObj->type); + return data; + } -size_t taosHashGetMemSize(const SHashObj *pHashObj) { - if (pHashObj == NULL) { - return 0; - } +void taosHashCancelIterate(SHashObj *pHashObj, void *p) { + if (pHashObj == NULL || p == NULL) return; - return (pHashObj->capacity * (sizeof(SHashEntry) + POINTER_BYTES)) + sizeof(SHashNode) * taosHashGetSize(pHashObj) + sizeof(SHashObj); + // only add the read lock to disable the resize process + __rd_lock(&pHashObj->lock, pHashObj->type); + + int slot; + taosHashReleaseNode(pHashObj, p, &slot); + + SHashEntry *pe = pHashObj->hashList[slot]; + if (pHashObj->type == HASH_ENTRY_LOCK) { + taosWUnLockLatch(&pe->latch); + } + + __rd_unlock(&pHashObj->lock, pHashObj->type); } diff --git a/src/util/src/tconfig.c b/src/util/src/tconfig.c index e89dea5a244acb5823c49d4b7c9aefb4c254db4c..173de294cf333e20b9649a5634905421786a1b30 100644 --- a/src/util/src/tconfig.c +++ b/src/util/src/tconfig.c @@ -19,9 +19,7 @@ #include "taoserror.h" #include "tconfig.h" #include "tglobal.h" -#include "tkey.h" #include "tulog.h" -#include "tsocket.h" #include "tsystem.h" #include "tutil.h" diff --git a/src/util/src/tfile.c b/src/util/src/tfile.c index 27ba30fe8179ecd5e213d4cf78862fd4982b3ba4..64fea2843b6d54488c77f9da6aa4965749855229 100644 --- a/src/util/src/tfile.c +++ b/src/util/src/tfile.c @@ -13,6 +13,7 @@ * along with this program. If not, see . */ +#define _DEFAULT_SOURCE #include "os.h" #include "taoserror.h" #include "tulog.h" @@ -21,40 +22,52 @@ static int32_t tsFileRsetId = -1; -static void taosCloseFile(void *p) { +static void tfCloseFile(void *p) { close((int32_t)(uintptr_t)p); } -int32_t tfinit() { - tsFileRsetId = taosOpenRef(2000, taosCloseFile); - return tsFileRsetId; +int32_t tfInit() { + tsFileRsetId = taosOpenRef(2000, tfCloseFile); + if (tsFileRsetId > 0) { + return 0; + } else { + return -1; + } } -void tfcleanup() { +void tfCleanup() { if (tsFileRsetId >= 0) taosCloseRef(tsFileRsetId); tsFileRsetId = -1; } -int64_t tfopen(const char *pathname, int32_t flags) { - int32_t fd = open(pathname, flags); - +static int64_t tfOpenImp(int32_t fd) { if (fd < 0) { terrno = TAOS_SYSTEM_ERROR(errno); return -1; - } + } - void *p = (void *)(int64_t)fd; + void * p = (void *)(int64_t)fd; int64_t rid = taosAddRef(tsFileRsetId, p); if (rid < 0) close(fd); return rid; } -int64_t tfclose(int64_t tfd) { +int64_t tfOpen(const char *pathname, int32_t flags) { + int32_t fd = open(pathname, flags); + return tfOpenImp(fd); +} + +int64_t tfOpenM(const char *pathname, int32_t flags, mode_t mode) { + int32_t fd = open(pathname, flags, mode); + return tfOpenImp(fd); +} + +int64_t tfClose(int64_t tfd) { return taosRemoveRef(tsFileRsetId, tfd); } -int64_t tfwrite(int64_t tfd, void *buf, int64_t count) { +int64_t tfWrite(int64_t tfd, void *buf, int64_t count) { void *p = taosAcquireRef(tsFileRsetId, tfd); if (p == NULL) return -1; @@ -67,7 +80,7 @@ int64_t tfwrite(int64_t tfd, void *buf, int64_t count) { return ret; } -int64_t tfread(int64_t tfd, void *buf, int64_t count) { +int64_t tfRead(int64_t tfd, void *buf, int64_t count) { void *p = taosAcquireRef(tsFileRsetId, tfd); if (p == NULL) return -1; @@ -79,3 +92,44 @@ int64_t tfread(int64_t tfd, void *buf, int64_t count) { taosReleaseRef(tsFileRsetId, tfd); return ret; } + +int32_t tfFsync(int64_t tfd) { + void *p = taosAcquireRef(tsFileRsetId, tfd); + if (p == NULL) return -1; + + int32_t fd = (int32_t)(uintptr_t)p; + int32_t code = fsync(fd); + + taosReleaseRef(tsFileRsetId, tfd); + return code; +} + +bool tfValid(int64_t tfd) { + void *p = taosAcquireRef(tsFileRsetId, tfd); + if (p == NULL) return false; + + taosReleaseRef(tsFileRsetId, tfd); + return true; +} + +int64_t tfLseek(int64_t tfd, int64_t offset, int32_t whence) { + void *p = taosAcquireRef(tsFileRsetId, tfd); + if (p == NULL) return -1; + + int32_t fd = (int32_t)(uintptr_t)p; + int64_t ret = taosLSeek(fd, offset, whence); + + taosReleaseRef(tsFileRsetId, tfd); + return ret; +} + +int32_t tfFtruncate(int64_t tfd, int64_t length) { + void *p = taosAcquireRef(tsFileRsetId, tfd); + if (p == NULL) return -1; + + int32_t fd = (int32_t)(uintptr_t)p; + int32_t code = taosFtruncate(fd, length); + + taosReleaseRef(tsFileRsetId, tfd); + return code; +} diff --git a/src/util/src/tkvstore.c b/src/util/src/tkvstore.c index 31641ac9a74d3fc914dcc05443da9384ba20c339..2b1d13c78b3d52f511e0c01e5472a2cd19c5255b 100644 --- a/src/util/src/tkvstore.c +++ b/src/util/src/tkvstore.c @@ -529,7 +529,7 @@ static int tdRestoreKVStore(SKVStore *pStore) { void * buf = NULL; int64_t maxBufSize = 0; SKVRecord rInfo = {0}; - SHashMutableIterator *pIter = NULL; + SKVRecord *pRecord = NULL; ASSERT(TD_KVSTORE_HEADER_SIZE == lseek(pStore->fd, 0, SEEK_CUR)); ASSERT(pStore->info.size == TD_KVSTORE_HEADER_SIZE); @@ -582,16 +582,8 @@ static int tdRestoreKVStore(SKVStore *pStore) { goto _err; } - pIter = taosHashCreateIter(pStore->map); - if (pIter == NULL) { - uError("failed to create hash iter while opening KV store %s", pStore->fname); - terrno = TSDB_CODE_COM_OUT_OF_MEMORY; - goto _err; - } - - while (taosHashIterNext(pIter)) { - SKVRecord *pRecord = taosHashIterGet(pIter); - + pRecord = taosHashIterate(pStore->map, NULL); + while (pRecord) { if (lseek(pStore->fd, (off_t)(pRecord->offset + sizeof(SKVRecord)), SEEK_SET) < 0) { uError("failed to lseek file %s since %s, offset %" PRId64, pStore->fname, strerror(errno), pRecord->offset); terrno = TAOS_SYSTEM_ERROR(errno); @@ -613,16 +605,17 @@ static int tdRestoreKVStore(SKVStore *pStore) { goto _err; } } + + pRecord = taosHashIterate(pStore->map, pRecord); } if (pStore->aFunc) (*pStore->aFunc)(pStore->appH); - taosHashDestroyIter(pIter); tfree(buf); return 0; _err: - taosHashDestroyIter(pIter); + taosHashCancelIterate(pStore->map, pRecord); tfree(buf); return -1; } diff --git a/src/util/src/tnettest.c b/src/util/src/tnettest.c index 6fd526598365f831addd1bacb0b7f748d9552fdb..c269d9a1ffe07c162123c44c397aba0da9593068 100644 --- a/src/util/src/tnettest.c +++ b/src/util/src/tnettest.c @@ -13,50 +13,42 @@ * along with this program. If not, see . */ +#define _DEFAULT_SOURCE #include "os.h" #include "taosdef.h" #include "taosmsg.h" #include "taoserror.h" #include "tulog.h" -#include "tconfig.h" #include "tglobal.h" #include "tsocket.h" #include "trpc.h" #include "rpcHead.h" -#include "tutil.h" -#include "tnettest.h" -#define MAX_PKG_LEN (64*1000) -#define BUFFER_SIZE (MAX_PKG_LEN + 1024) +#define MAX_PKG_LEN (64 * 1000) +#define BUFFER_SIZE (MAX_PKG_LEN + 1024) + +extern int32_t tsRpcMaxUdpSize; typedef struct { + char * hostFqdn; uint32_t hostIp; - uint16_t port; - uint16_t pktLen; -} info_s; - -extern int tsRpcMaxUdpSize; - -static char g_user[TSDB_USER_LEN+1] = {0}; -static char g_pass[TSDB_PASSWORD_LEN+1] = {0}; -static char g_serverFqdn[TSDB_FQDN_LEN] = {0}; -static uint16_t g_startPort = 0; -static uint16_t g_endPort = 6042; -static uint32_t g_pktLen = 0; - - -static void *bindUdpPort(void *sarg) { - info_s *pinfo = (info_s *)sarg; - int port = pinfo->port; - SOCKET serverSocket; + int32_t port; + int32_t pktLen; +} STestInfo; + +static void *taosNetBindUdpPort(void *sarg) { + STestInfo *pinfo = (STestInfo *)sarg; + int32_t port = pinfo->port; + SOCKET serverSocket; + char buffer[BUFFER_SIZE]; + int32_t iDataNum; + socklen_t sin_size; struct sockaddr_in server_addr; struct sockaddr_in clientAddr; - char buffer[BUFFER_SIZE]; - int iDataNum; - + if ((serverSocket = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP)) < 0) { - perror("socket"); + uError("failed to create udp socket since %s", strerror(errno)); return NULL; } @@ -66,28 +58,25 @@ static void *bindUdpPort(void *sarg) { server_addr.sin_addr.s_addr = htonl(INADDR_ANY); if (bind(serverSocket, (struct sockaddr *)&server_addr, sizeof(server_addr)) < 0) { - perror("connect"); + uError("failed to bind udp port:%d since %s", port, strerror(errno)); return NULL; } - socklen_t sin_size; + uInfo("udp server at port:%d is listening", port); while (1) { memset(buffer, 0, BUFFER_SIZE); - sin_size = sizeof(*(struct sockaddr *)&server_addr); - iDataNum = recvfrom(serverSocket, buffer, BUFFER_SIZE, 0, (struct sockaddr *)&clientAddr, &sin_size); if (iDataNum < 0) { - perror("recvfrom null"); + uDebug("failed to perform recvfrom func at %d since %s", port, strerror(errno)); continue; } - if (iDataNum > 0) { - printf("recv Client: %s pkg from UDP port: %d, pkg len: %d\n", taosInetNtoa(clientAddr.sin_addr), port, iDataNum); - //printf("Read msg from udp:%s ... %s\n", buffer, buffer+iDataNum-16); - sendto(serverSocket, buffer, iDataNum, 0, (struct sockaddr *)&clientAddr, (int)sin_size); + if (iDataNum > 0) { + uInfo("UDP: recv:%d bytes from %s:%d", iDataNum, taosInetNtoa(clientAddr.sin_addr), port); + sendto(serverSocket, buffer, iDataNum, 0, (struct sockaddr *)&clientAddr, (int32_t)sin_size); } } @@ -95,20 +84,20 @@ static void *bindUdpPort(void *sarg) { return NULL; } -static void *bindTcpPort(void *sarg) { - info_s *pinfo = (info_s *)sarg; - int port = pinfo->port; - SOCKET serverSocket; - +static void *taosNetBindTcpPort(void *sarg) { struct sockaddr_in server_addr; struct sockaddr_in clientAddr; - int addr_len = sizeof(clientAddr); - SOCKET client; - char buffer[BUFFER_SIZE]; - int iDataNum = 0; + + STestInfo *pinfo = sarg; + int32_t port = pinfo->port; + SOCKET serverSocket; + int32_t addr_len = sizeof(clientAddr); + SOCKET client; + char buffer[BUFFER_SIZE]; + int32_t iDataNum = 0; if ((serverSocket = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP)) < 0) { - printf("socket() fail: %s", strerror(errno)); + uError("failed to create tcp socket since %s", strerror(errno)); return NULL; } @@ -118,28 +107,30 @@ static void *bindTcpPort(void *sarg) { server_addr.sin_addr.s_addr = htonl(INADDR_ANY); if (bind(serverSocket, (struct sockaddr *)&server_addr, sizeof(server_addr)) < 0) { - printf("port:%d bind() fail: %s", port, strerror(errno)); + uError("failed to bind tcp port:%d since %s", port, strerror(errno)); return NULL; } if (listen(serverSocket, 5) < 0) { - printf("listen() fail: %s", strerror(errno)); + uError("failed to listen tcp port:%d since %s", port, strerror(errno)); return NULL; } - //printf("Bind port: %d success\n", port); + uInfo("tcp server at port:%d is listening", port); + while (1) { client = accept(serverSocket, (struct sockaddr *)&clientAddr, (socklen_t *)&addr_len); if (client < 0) { - printf("accept() fail: %s", strerror(errno)); + uDebug("failed to accept from tcp port:%d since %s", port, strerror(errno)); continue; } iDataNum = 0; memset(buffer, 0, BUFFER_SIZE); - int nleft, nread; - char *ptr = buffer; + int32_t nleft, nread; + char * ptr = buffer; nleft = pinfo->pktLen; + while (nleft > 0) { nread = recv(client, ptr, BUFFER_SIZE, 0); @@ -149,7 +140,7 @@ static void *bindTcpPort(void *sarg) { if (errno == EINTR) { continue; } else { - printf("recv Client: %s pkg from TCP port: %d fail:%s.\n", taosInetNtoa(clientAddr.sin_addr), port, strerror(errno)); + uError("failed to perform recv func at %d since %s", port, strerror(errno)); taosCloseSocket(serverSocket); return NULL; } @@ -157,11 +148,11 @@ static void *bindTcpPort(void *sarg) { nleft -= nread; ptr += nread; iDataNum += nread; - } + } } - - printf("recv Client: %s pkg from TCP port: %d, pkg len: %d\n", taosInetNtoa(clientAddr.sin_addr), port, iDataNum); + if (iDataNum > 0) { + uInfo("TCP: recv:%d bytes from %s:%d", iDataNum, taosInetNtoa(clientAddr.sin_addr), port); send(client, buffer, iDataNum, 0); } } @@ -170,39 +161,38 @@ static void *bindTcpPort(void *sarg) { return NULL; } -static int checkTcpPort(info_s *info) { +static int32_t taosNetCheckTcpPort(STestInfo *info) { + SOCKET clientSocket; + char sendbuf[BUFFER_SIZE]; + char recvbuf[BUFFER_SIZE]; + int32_t iDataNum = 0; + struct sockaddr_in serverAddr; - SOCKET clientSocket; - char sendbuf[BUFFER_SIZE]; - char recvbuf[BUFFER_SIZE]; - int iDataNum = 0; if ((clientSocket = socket(AF_INET, SOCK_STREAM, 0)) < 0) { - printf("socket() fail: %s\n", strerror(errno)); + uError("failed to create tcp client socket since %s", strerror(errno)); return -1; } // set send and recv overtime - struct timeval timeout; - timeout.tv_sec = 2; //s - timeout.tv_usec = 0; //us - if (setsockopt(clientSocket, SOL_SOCKET,SO_SNDTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { - perror("setsockopt send timer failed:"); + struct timeval timeout; + timeout.tv_sec = 2; // s + timeout.tv_usec = 0; // us + if (setsockopt(clientSocket, SOL_SOCKET, SO_SNDTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { + uError("failed to setsockopt send timer since %s", strerror(errno)); } - if (setsockopt(clientSocket, SOL_SOCKET,SO_RCVTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { - perror("setsockopt recv timer failed:"); + if (setsockopt(clientSocket, SOL_SOCKET, SO_RCVTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { + uError("failed to setsockopt recv timer since %s", strerror(errno)); } serverAddr.sin_family = AF_INET; serverAddr.sin_port = htons(info->port); - serverAddr.sin_addr.s_addr = info->hostIp; - //printf("=================================\n"); if (connect(clientSocket, (struct sockaddr *)&serverAddr, sizeof(serverAddr)) < 0) { - printf("connect() fail: %s\t", strerror(errno)); + uError("failed to connect port:%d since %s", info->port, strerror(errno)); return -1; } - //printf("Connect to: %s:%d...success\n", host, port); + memset(sendbuf, 0, BUFFER_SIZE); memset(recvbuf, 0, BUFFER_SIZE); @@ -214,9 +204,10 @@ static int checkTcpPort(info_s *info) { send(clientSocket, sendbuf, info->pktLen, 0); memset(recvbuf, 0, BUFFER_SIZE); - int nleft, nread; - char *ptr = recvbuf; + int32_t nleft, nread; + char * ptr = recvbuf; nleft = info->pktLen; + while (nleft > 0) { nread = recv(clientSocket, ptr, BUFFER_SIZE, 0);; @@ -226,7 +217,7 @@ static int checkTcpPort(info_s *info) { if (errno == EINTR) { continue; } else { - printf("recv ack pkg from TCP port: %d fail:%s.\n", info->port, strerror(errno)); + uError("faild to recv pkg from TCP port:%d since %s", info->port, strerror(errno)); taosCloseSocket(clientSocket); return -1; } @@ -234,45 +225,46 @@ static int checkTcpPort(info_s *info) { nleft -= nread; ptr += nread; iDataNum += nread; - } + } } if (iDataNum < info->pktLen) { - printf("recv ack pkg len: %d, less than req pkg len: %d from tcp port: %d\n", iDataNum, info->pktLen, info->port); + uError("TCP: received ack:%d bytes, less than send:%d bytes from port:%d", iDataNum, info->pktLen, info->port); return -1; } - //printf("Read ack pkg len:%d from tcp port: %d, buffer: %s %s\n", info->pktLen, port, recvbuf, recvbuf+iDataNum-8); taosCloseSocket(clientSocket); return 0; } -static int checkUdpPort(info_s *info) { +static int32_t taosNetCheckUdpPort(STestInfo *info) { + SOCKET clientSocket; + char sendbuf[BUFFER_SIZE]; + char recvbuf[BUFFER_SIZE]; + int32_t iDataNum = 0; + struct sockaddr_in serverAddr; - SOCKET clientSocket; - char sendbuf[BUFFER_SIZE]; - char recvbuf[BUFFER_SIZE]; - int iDataNum = 0; + if ((clientSocket = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP)) < 0) { - perror("socket"); + uError("failed to create udp client socket since %s", strerror(errno)); return -1; } - // set overtime + // set overtime struct timeval timeout; - timeout.tv_sec = 2; //s - timeout.tv_usec = 0; //us - if (setsockopt(clientSocket, SOL_SOCKET,SO_SNDTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { - perror("setsockopt send timer failed:"); + timeout.tv_sec = 2; // s + timeout.tv_usec = 0; // us + if (setsockopt(clientSocket, SOL_SOCKET, SO_SNDTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { + uError("failed to setsockopt send timer since %s", strerror(errno)); } - if (setsockopt(clientSocket, SOL_SOCKET,SO_RCVTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { - perror("setsockopt recv timer failed:"); + if (setsockopt(clientSocket, SOL_SOCKET, SO_RCVTIMEO, (char *)&timeout, sizeof(struct timeval)) == -1) { + uError("failed to setsockopt recv timer since %s", strerror(errno)); } - + serverAddr.sin_family = AF_INET; serverAddr.sin_port = htons(info->port); serverAddr.sin_addr.s_addr = info->hostIp; - + memset(sendbuf, 0, BUFFER_SIZE); memset(recvbuf, 0, BUFFER_SIZE); @@ -283,69 +275,66 @@ static int checkUdpPort(info_s *info) { socklen_t sin_size = sizeof(*(struct sockaddr *)&serverAddr); - int code = sendto(clientSocket, sendbuf, info->pktLen, 0, (struct sockaddr *)&serverAddr, (int)sin_size); + int32_t code = sendto(clientSocket, sendbuf, info->pktLen, 0, (struct sockaddr *)&serverAddr, (int32_t)sin_size); if (code < 0) { - perror("sendto"); + uError("failed to perform sendto func since %s", strerror(errno)); return -1; } iDataNum = recvfrom(clientSocket, recvbuf, BUFFER_SIZE, 0, (struct sockaddr *)&serverAddr, &sin_size); if (iDataNum < info->pktLen) { - printf("Read ack pkg len: %d, less than req pkg len: %d from udp port: %d\t\t", iDataNum, info->pktLen, info->port); + uError("UDP: received ack:%d bytes, less than send:%d bytes from port:%d", iDataNum, info->pktLen, info->port); return -1; } - - //printf("Read ack pkg len:%d from udp port: %d, buffer: %s %s\n", info->pktLen, port, recvbuf, recvbuf+iDataNum-8); + taosCloseSocket(clientSocket); return 0; } -static void checkPort(uint32_t hostIp, uint16_t startPort, uint16_t maxPort, uint16_t pktLen) { - int ret; - info_s info; - memset(&info, 0, sizeof(info_s)); - info.hostIp = hostIp; - info.pktLen = pktLen; +static void taosNetCheckPort(uint32_t hostIp, int32_t startPort, int32_t endPort, int32_t pktLen) { + int32_t ret; + STestInfo info; - for (uint16_t port = startPort; port <= maxPort; port++) { - //printf("test: %s:%d\n", info.host, port); - printf("\n"); + memset(&info, 0, sizeof(STestInfo)); + info.hostIp = hostIp; + info.pktLen = pktLen; + for (int32_t port = startPort; port <= endPort; port++) { info.port = port; - ret = checkTcpPort(&info); + ret = taosNetCheckTcpPort(&info); if (ret != 0) { - printf("tcp port:%d test fail.\t\n", port); + uError("failed to test tcp port:%d", port); } else { - printf("tcp port:%d test ok.\t\t", port); + uInfo("successed to test tcp port:%d", port); } - - ret = checkUdpPort(&info); + + ret = taosNetCheckUdpPort(&info); if (ret != 0) { - printf("udp port:%d test fail.\t\n", port); + uError("failed to test udp port:%d", port); } else { - printf("udp port:%d test ok.\t\t", port); + uInfo("successed to test udp port:%d", port); } } - - printf("\n"); - return ; + return; } -void* tnetInitRpc(char* secretEncrypt, char spi) { +void *taosNetInitRpc(char *secretEncrypt, char spi) { SRpcInit rpcInit; - void* pRpcConn = NULL; + void * pRpcConn = NULL; + + char user[] = "nettestinternal"; + char pass[] = "nettestinternal"; + taosEncryptPass((uint8_t *)pass, strlen(pass), secretEncrypt); - taosEncryptPass((uint8_t *)g_pass, strlen(g_pass), secretEncrypt); - memset(&rpcInit, 0, sizeof(rpcInit)); rpcInit.localPort = 0; - rpcInit.label = "NET-TEST"; + rpcInit.label = "NT"; rpcInit.numOfThreads = 1; // every DB connection has only one thread rpcInit.cfp = NULL; rpcInit.sessions = 16; rpcInit.connType = TAOS_CONN_CLIENT; - rpcInit.user = g_user; + rpcInit.user = user; rpcInit.idleTime = 2000; rpcInit.ckey = "key"; rpcInit.spi = spi; @@ -355,17 +344,18 @@ void* tnetInitRpc(char* secretEncrypt, char spi) { return pRpcConn; } -static int rpcCheckPortImpl(const char* serverFqdn, uint16_t port, uint16_t pktLen, char spi) { +static int32_t taosNetCheckRpc(const char* serverFqdn, uint16_t port, uint16_t pktLen, char spi, SStartupStep *pStep) { SRpcEpSet epSet; SRpcMsg reqMsg; SRpcMsg rspMsg; - void* pRpcConn; + void * pRpcConn; char secretEncrypt[32] = {0}; - pRpcConn = tnetInitRpc(secretEncrypt, spi); + pRpcConn = taosNetInitRpc(secretEncrypt, spi); if (NULL == pRpcConn) { - return -1; + uError("failed to init client rpc"); + return TSDB_CODE_RPC_NETWORK_UNAVAIL; } memset(&epSet, 0, sizeof(SRpcEpSet)); @@ -373,205 +363,171 @@ static int rpcCheckPortImpl(const char* serverFqdn, uint16_t port, uint16_t pktL epSet.numOfEps = 1; epSet.port[0] = port; strcpy(epSet.fqdn[0], serverFqdn); - + reqMsg.msgType = TSDB_MSG_TYPE_NETWORK_TEST; reqMsg.pCont = rpcMallocCont(pktLen); reqMsg.contLen = pktLen; reqMsg.code = 0; reqMsg.handle = NULL; // rpc handle returned to app - reqMsg.ahandle = NULL; // app handle set by client + reqMsg.ahandle = NULL; // app handle set by client + strcpy(reqMsg.pCont, "nettest"); rpcSendRecv(pRpcConn, &epSet, &reqMsg, &rspMsg); - // handle response if ((rspMsg.code != 0) || (rspMsg.msgType != TSDB_MSG_TYPE_NETWORK_TEST + 1)) { - //printf("code:%d[%s]\n", rspMsg.code, tstrerror(rspMsg.code)); - return -1; + uDebug("ret code 0x%x %s", rspMsg.code, tstrerror(rspMsg.code)); + return rspMsg.code; } - - rpcFreeCont(rspMsg.pCont); + int32_t code = 0; + if (pStep != NULL && rspMsg.pCont != NULL && rspMsg.contLen > 0 && rspMsg.contLen <= sizeof(SStartupStep)) { + memcpy(pStep, rspMsg.pCont, rspMsg.contLen); + code = 1; + } + + rpcFreeCont(rspMsg.pCont); rpcClose(pRpcConn); + return code; +} - return 0; +static int32_t taosNetParseStartup(SStartupStep *pCont) { + SStartupStep *pStep = pCont; + uInfo("step:%s desc:%s", pStep->name, pStep->desc); + + if (pStep->finished) { + uInfo("check startup finished"); + } + + return pStep->finished ? 0 : 1; } -static void rpcCheckPort(uint32_t hostIp) { - int ret; - char spi; +static void taosNetTestStartup(char *host, int32_t port) { + uInfo("check startup, host:%s port:%d\n", host, port); - for (uint16_t port = g_startPort; port <= g_endPort; port++) { - //printf("test: %s:%d\n", info.host, port); - printf("\n"); + SStartupStep *pStep = malloc(sizeof(SStartupStep)); + while (1) { + int32_t code = taosNetCheckRpc(host, port + TSDB_PORT_DNODEDNODE, 20, 0, pStep); + if (code > 0) { + code = taosNetParseStartup(pStep); + } - //================ check tcp port ================ - int32_t pktLen; - if (g_pktLen <= tsRpcMaxUdpSize) { - pktLen = tsRpcMaxUdpSize + 1000; + if (code > 0) { + uDebug("continue check startup step"); } else { - pktLen = g_pktLen; + if (code < 0) { + uError("failed to check startup step, code:0x%x %s", code, tstrerror(code)); + } + break; } + } - spi = 1; - ret = rpcCheckPortImpl(g_serverFqdn, port, pktLen, spi); - if (ret != 0) { - spi = 0; - ret = rpcCheckPortImpl(g_serverFqdn, port, pktLen, spi); - if (ret != 0) { - printf("TCP port:%d test fail.\t\t", port); - } else { - //printf("tcp port:%d test ok.\t\t", port); - printf("TCP port:\033[32m%d test OK\033[0m\t\t", port); - } + free(pStep); +} + +static void taosNetTestRpc(char *host, int32_t startPort, int32_t pkgLen) { + int32_t endPort = startPort + 9; + char spi = 0; + + uInfo("check rpc, host:%s startPort:%d endPort:%d pkgLen:%d\n", host, startPort, endPort, pkgLen); + + for (uint16_t port = startPort; port <= endPort; port++) { + int32_t sendpkgLen; + if (pkgLen <= tsRpcMaxUdpSize) { + sendpkgLen = tsRpcMaxUdpSize + 1000; } else { - //printf("tcp port:%d test ok.\t\t", port); - printf("TCP port:\033[32m%d test OK\033[0m\t\t", port); + sendpkgLen = pkgLen; } - //================ check udp port ================ - if (g_pktLen >= tsRpcMaxUdpSize) { - pktLen = tsRpcMaxUdpSize - 1000; + int32_t ret = taosNetCheckRpc(host, port, sendpkgLen, spi, NULL); + if (ret < 0) { + uError("failed to test tcp port:%d", port); } else { - pktLen = g_pktLen; + uInfo("successed to test tcp port:%d", port); } - - spi = 0; - ret = rpcCheckPortImpl(g_serverFqdn, port, pktLen, spi); - if (ret != 0) { - spi = 1; - ret = rpcCheckPortImpl(g_serverFqdn, port, pktLen, spi); - if (ret != 0) { - printf("udp port:%d test fail.\t\n", port); - } else { - //printf("udp port:%d test ok.\t\n", port); - printf("UDP port:\033[32m%d test OK\033[0m\t\n", port); - } + + if (pkgLen >= tsRpcMaxUdpSize) { + sendpkgLen = tsRpcMaxUdpSize - 1000; + } else { + sendpkgLen = pkgLen; + } + + ret = taosNetCheckRpc(host, port, pkgLen, spi, NULL); + if (ret < 0) { + uError("failed to test udp port:%d", port); } else { - //printf("udp port:%d test ok.\t\n", port); - printf("UDP port:\033[32m%d test OK\033[0m\t\n", port); + uInfo("successed to test udp port:%d", port); } } - - printf("\n"); - return ; } -static void taosNetTestClient(int flag) { - uint32_t serverIp = taosGetIpFromFqdn(g_serverFqdn); +static void taosNetTestClient(char *host, int32_t startPort, int32_t pkgLen) { + int32_t endPort = startPort + 11; + uInfo("work as client, host:%s startPort:%d endPort:%d pkgLen:%d\n", host, startPort, endPort, pkgLen); + + uint32_t serverIp = taosGetIpFromFqdn(host); if (serverIp == 0xFFFFFFFF) { - printf("Failed to resolve FQDN:%s", g_serverFqdn); + uError("failed to resolve fqdn:%s", host); exit(-1); } - if (0 == flag) { - checkPort(serverIp, g_startPort, g_endPort, g_pktLen); - } else { - rpcCheckPort(serverIp); - } - - return; + uInfo("server ip:%s is resolved from host:%s", taosIpStr(serverIp), host); + taosNetCheckPort(serverIp, startPort, endPort, pkgLen); } -static void taosNetTestServer(uint16_t startPort, uint16_t endPort, int pktLen) { +static void taosNetTestServer(char *host, int32_t startPort, int32_t pkgLen) { + int32_t endPort = startPort + 11; + uInfo("work as server, host:%s startPort:%d endPort:%d pkgLen:%d\n", host, startPort, endPort, pkgLen); - int port = startPort; - int num = endPort - startPort + 1; + int32_t port = startPort; + int32_t num = endPort - startPort + 1; + if (num < 0) num = 1; - if (num < 0) { - num = 1; - } - pthread_t *pids = malloc(2 * num * sizeof(pthread_t)); - info_s * tinfos = malloc(num * sizeof(info_s)); - info_s * uinfos = malloc(num * sizeof(info_s)); + STestInfo *tinfos = malloc(num * sizeof(STestInfo)); + STestInfo *uinfos = malloc(num * sizeof(STestInfo)); - for (size_t i = 0; i < num; i++) { - info_s *tcpInfo = tinfos + i; - tcpInfo->port = (uint16_t)(port + i); - tcpInfo->pktLen = pktLen; + for (int32_t i = 0; i < num; i++) { + STestInfo *tcpInfo = tinfos + i; + tcpInfo->port = port + i; + tcpInfo->pktLen = pkgLen; - if (pthread_create(pids + i, NULL, bindTcpPort, tcpInfo) != 0) - { - printf("create thread fail, port:%d.\n", port); + if (pthread_create(pids + i, NULL, taosNetBindTcpPort, tcpInfo) != 0) { + uInfo("failed to create tcp test thread, %s:%d", tcpInfo->hostFqdn, tcpInfo->port); exit(-1); } - info_s *udpInfo = uinfos + i; + STestInfo *udpInfo = uinfos + i; udpInfo->port = (uint16_t)(port + i); - if (pthread_create(pids + num + i, NULL, bindUdpPort, udpInfo) != 0) - { - printf("create thread fail, port:%d.\n", port); + if (pthread_create(pids + num + i, NULL, taosNetBindUdpPort, udpInfo) != 0) { + uInfo("failed to create udp test thread, %s:%d", tcpInfo->hostFqdn, tcpInfo->port); exit(-1); } } - - for (int i = 0; i < num; i++) { + + for (int32_t i = 0; i < num; i++) { pthread_join(pids[i], NULL); pthread_join(pids[(num + i)], NULL); } } - -void taosNetTest(CmdArguments *args) { - if (0 == args->pktLen) { - g_pktLen = 1000; +void taosNetTest(char *role, char *host, int32_t port, int32_t pkgLen) { + tscEmbedded = 1; + if (host == NULL) host = tsLocalFqdn; + if (port == 0) port = tsServerPort; + if (pkgLen <= 10) pkgLen = 1000; + if (pkgLen > MAX_PKG_LEN) pkgLen = MAX_PKG_LEN; + + if (0 == strcmp("client", role)) { + taosNetTestClient(host, port, pkgLen); + } else if (0 == strcmp("server", role)) { + taosNetTestServer(host, port, pkgLen); + } else if (0 == strcmp("rpc", role)) { + taosNetTestRpc(host, port, pkgLen); + } else if (0 == strcmp("startup", role)) { + taosNetTestStartup(host, port); } else { - g_pktLen = args->pktLen; + taosNetTestStartup(host, port); } - - if (args->port && args->endPort) { - if (args->port > args->endPort) { - printf("endPort[%d] must not lesss port[%d]\n", args->endPort, args->port); - exit(-1); - } - } - - if (args->host && args->host[0] != 0) { - if (strlen(args->host) >= TSDB_EP_LEN) { - printf("host invalid: %s\n", args->host); - exit(-1); - } - taosGetFqdnPortFromEp(args->host, g_serverFqdn, &g_startPort); - } else { - tstrncpy(g_serverFqdn, "127.0.0.1", TSDB_IPv4ADDR_LEN); - g_startPort = tsServerPort; - } - - if (args->port) { - g_startPort = args->port; - } - - if (args->endPort) { - g_endPort = args->endPort; - } - - if (g_startPort > g_endPort) { - printf("endPort[%d] must not lesss port[%d]\n", g_endPort, g_startPort); - exit(-1); - } - - - if (args->is_use_passwd) { - if (args->password == NULL) args->password = getpass("Enter password: "); - } else { - args->password = TSDB_DEFAULT_PASS; - } - tstrncpy(g_pass, args->password, TSDB_PASSWORD_LEN); - - if (args->user == NULL) { - args->user = TSDB_DEFAULT_USER; - } - tstrncpy(g_user, args->user, TSDB_USER_LEN); - - if (0 == strcmp("client", args->netTestRole)) { - printf("host: %s\tstart port: %d\tend port: %d\tpacket len: %d\n", g_serverFqdn, g_startPort, g_endPort, g_pktLen); - taosNetTestClient(0); - } else if (0 == strcmp("clients", args->netTestRole)) { - printf("host: %s\tstart port: %d\tend port: %d\tpacket len: %d\n", g_serverFqdn, g_startPort, g_endPort, g_pktLen); - taosNetTestClient(1); - } else if (0 == strcmp("server", args->netTestRole)) { - taosNetTestServer(g_startPort, g_endPort, g_pktLen); - } + tscEmbedded = 0; } - diff --git a/src/util/src/tqueue.c b/src/util/src/tqueue.c index d3c17e7626c4566df32c1e4257bda1a4e68a200a..d72bc5f4121ec0cd71c1c1dad5d9db5421c66a37 100644 --- a/src/util/src/tqueue.c +++ b/src/util/src/tqueue.c @@ -79,15 +79,12 @@ void taosCloseQueue(taos_queue param) { if (queue->qset) taosRemoveFromQset(qset, queue); - pthread_mutex_lock(&queue->mutex); - while (pNode) { pTemp = pNode; pNode = pNode->next; free (pTemp); } - pthread_mutex_unlock(&queue->mutex); pthread_mutex_destroy(&queue->mutex); free(queue); @@ -361,15 +358,15 @@ int taosReadQitemFromQset(taos_qset param, int *type, void **pitem, void **phand if (queue->head) { pNode = queue->head; *pitem = pNode->item; - *type = pNode->type; - *phandle = queue->ahandle; + if (type) *type = pNode->type; + if (phandle) *phandle = queue->ahandle; queue->head = pNode->next; if (queue->head == NULL) queue->tail = NULL; queue->numOfItems--; atomic_sub_fetch_32(&qset->numOfItems, 1); code = 1; - uTrace("item:%p is read out from queue:%p, type:%d items:%d", *pitem, queue, *type, queue->numOfItems); + uTrace("item:%p is read out from queue:%p, type:%d items:%d", *pitem, queue, pNode->type, queue->numOfItems); } pthread_mutex_unlock(&queue->mutex); diff --git a/src/util/src/tref.c b/src/util/src/tref.c index 4c1a87c96070534e25149357b766a8539e362680..1f83abcb847147da1d9ad0d17d26081176df0c93 100644 --- a/src/util/src/tref.c +++ b/src/util/src/tref.c @@ -329,7 +329,7 @@ void *taosIterateRef(int rsetId, int64_t rid) { pNode->count++; // acquire it newP = pNode->p; taosUnlockList(pSet->lockedBy+hash); - uTrace("rsetId:%d p:%p rid:%" PRId64 " is returned", rsetId, newP, rid); + uTrace("rsetId:%d p:%p rid:%" PRId64 " is returned", rsetId, newP, rid); } else { uTrace("rsetId:%d the list is over", rsetId); } @@ -423,24 +423,25 @@ static int taosDecRefCount(int rsetId, int64_t rid, int remove) { if (pNode->next) { pNode->next->prev = pNode->prev; } - - (*pSet->fp)(pNode->p); - - uTrace("rsetId:%d p:%p rid:%" PRId64 " is removed, count:%d, free mem: %p", rsetId, pNode->p, rid, pSet->count, pNode); - free(pNode); released = 1; } else { - uTrace("rsetId:%d p:%p rid:%" PRId64 " is released, count:%d", rsetId, pNode->p, rid, pNode->count); + uTrace("rsetId:%d p:%p rid:%" PRId64 " is released", rsetId, pNode->p, rid); } } else { - uTrace("rsetId:%d rid:%" PRId64 " is not there, failed to release/remove", rsetId, rid); + uTrace("rsetId:%d rid:%" PRId64 " is not there, failed to release/remove", rsetId, rid); terrno = TSDB_CODE_REF_NOT_EXIST; code = -1; } taosUnlockList(pSet->lockedBy+hash); - if (released) taosDecRsetCount(pSet); + if (released) { + uTrace("rsetId:%d p:%p rid:%" PRId64 " is removed, count:%d, free mem: %p", rsetId, pNode->p, rid, pSet->count, pNode); + (*pSet->fp)(pNode->p); + free(pNode); + + taosDecRsetCount(pSet); + } return code; } diff --git a/src/util/src/tskiplist.c b/src/util/src/tskiplist.c index a36f7f0261ddb34cbfa82a5bc2485c6e9c45f03a..20bec169546d6514def2453257346d5f94826187 100644 --- a/src/util/src/tskiplist.c +++ b/src/util/src/tskiplist.c @@ -291,7 +291,10 @@ bool tSkipListIterNext(SSkipListIterator *iter) { iter->next = SL_NODE_GET_FORWARD_POINTER(iter->cur, 0); iter->step++; } else { - if (iter->cur == pSkipList->pHead) return false; + if (iter->cur == pSkipList->pHead) { + tSkipListUnlock(pSkipList); + return false; + } iter->cur = SL_NODE_GET_BACKWARD_POINTER(iter->cur, 0); // a new node is inserted into between iter->cur and iter->next, ignore it diff --git a/src/vnode/inc/vnodeCfg.h b/src/vnode/inc/vnodeCfg.h index c5887fef5d53ca32c215a5f2da9d182386ebc4f0..ba148c07c1e4f9451232c706fb79c0af0cad9746 100644 --- a/src/vnode/inc/vnodeCfg.h +++ b/src/vnode/inc/vnodeCfg.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "vnodeInt.h" int32_t vnodeReadCfg(SVnodeObj *pVnode); int32_t vnodeWriteCfg(SCreateVnodeMsg *pVnodeCfg); diff --git a/src/vnode/inc/vnodeInt.h b/src/vnode/inc/vnodeInt.h index 7fc9b100eff50bd6a68a1dd5f1f48acbb290c530..401c217b9a972f0929f38b76b88aac508f106723 100644 --- a/src/vnode/inc/vnodeInt.h +++ b/src/vnode/inc/vnodeInt.h @@ -19,11 +19,11 @@ #ifdef __cplusplus extern "C" { #endif - #include "tlog.h" #include "tsync.h" -#include "twal.h" #include "tcq.h" +#include "tsdb.h" +#include "vnode.h" extern int32_t vDebugFlag; @@ -35,38 +35,37 @@ extern int32_t vDebugFlag; #define vTrace(...) { if (vDebugFlag & DEBUG_TRACE) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }} typedef struct { - int32_t vgId; // global vnode group ID - int32_t refCount; // reference count - int32_t queuedWMsg; - int32_t queuedRMsg; - int32_t flowctrlLevel; - int8_t status; - int8_t role; - int8_t accessState; - int8_t isFull; - uint64_t version; // current version - uint64_t fversion; // version on saved data file - void *wqueue; - void *rqueue; - void *wal; - void *tsdb; - int64_t sync; - void *events; - void *cq; // continuous query - int32_t cfgVersion; - STsdbCfg tsdbCfg; - SSyncCfg syncCfg; - SWalCfg walCfg; - void *qMgmt; - char *rootDir; - tsem_t sem; - int8_t dropped; - char db[TSDB_ACCT_LEN + TSDB_DB_NAME_LEN]; + int32_t vgId; // global vnode group ID + int32_t refCount; // reference count + int32_t queuedWMsg; + int32_t queuedRMsg; + int32_t flowctrlLevel; + int8_t status; + int8_t role; + int8_t accessState; + int8_t isFull; + int8_t isCommiting; + uint64_t version; // current version + uint64_t fversion; // version on saved data file + void * wqueue; + void * rqueue; + void * wal; + void * tsdb; + int64_t sync; + void * events; + void * cq; // continuous query + int32_t cfgVersion; + STsdbCfg tsdbCfg; + SSyncCfg syncCfg; + SWalCfg walCfg; + void * qMgmt; + char * rootDir; + tsem_t sem; + int8_t dropped; + char db[TSDB_ACCT_LEN + TSDB_DB_NAME_LEN]; + pthread_mutex_t statusMutex; } SVnodeObj; -void vnodeInitWriteFp(void); -void vnodeInitReadFp(void); - #ifdef __cplusplus } #endif diff --git a/src/vnode/inc/vnodeMain.h b/src/vnode/inc/vnodeMain.h new file mode 100644 index 0000000000000000000000000000000000000000..e1ddcdc36aa1fbf434b138f8d6fef5966e1fbc3e --- /dev/null +++ b/src/vnode/inc/vnodeMain.h @@ -0,0 +1,38 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_VNODE_MAIN_H +#define TDENGINE_VNODE_MAIN_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "vnodeInt.h" + +int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg); +int32_t vnodeDrop(int32_t vgId); +int32_t vnodeOpen(int32_t vgId); +int32_t vnodeAlter(void *pVnode, SCreateVnodeMsg *pVnodeCfg); +int32_t vnodeClose(int32_t vgId); + +int32_t vnodeReset(SVnodeObj *pVnode); +void vnodeCleanUp(SVnodeObj *pVnode); +void vnodeDestroy(SVnodeObj *pVnode); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/vnode/inc/vnodeMgmt.h b/src/vnode/inc/vnodeMgmt.h new file mode 100644 index 0000000000000000000000000000000000000000..5a7e7456195f2f8a21f2c99b4d50761c83729acd --- /dev/null +++ b/src/vnode/inc/vnodeMgmt.h @@ -0,0 +1,42 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_VNODE_MGMT_H +#define TDENGINE_VNODE_MGMT_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "vnodeInt.h" + +int32_t vnodeInitMgmt(); +void vnodeCleanupMgmt(); + +void* vnodeAcquire(int32_t vgId); +void vnodeRelease(void *pVnode); +void* vnodeGetWal(void *pVnode); + +int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes); +void vnodeBuildStatusMsg(void *pStatus); +void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes); + +void vnodeAddIntoHash(SVnodeObj* pVnode); +void vnodeRemoveFromHash(SVnodeObj * pVnode); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/vnode/inc/vnodeRead.h b/src/vnode/inc/vnodeRead.h new file mode 100644 index 0000000000000000000000000000000000000000..f2953d79f4d07c3dac821e9a086d86c53647d9c7 --- /dev/null +++ b/src/vnode/inc/vnodeRead.h @@ -0,0 +1,35 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_VNODE_READ_H +#define TDENGINE_VNODE_READ_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "vnodeInt.h" + +int32_t vnodeInitRead(void); +void vnodeCleanupRead(void); + +int32_t vnodeWriteToRQueue(void *pVnode, void *pCont, int32_t contLen, int8_t qtype, void *rparam); +void vnodeFreeFromRQueue(void *pVnode, SVReadMsg *pRead); +int32_t vnodeProcessRead(void *pVnode, SVReadMsg *pRead); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/vnode/inc/vnodeStatus.h b/src/vnode/inc/vnodeStatus.h new file mode 100644 index 0000000000000000000000000000000000000000..00ac47df65fba91b9d7ef4b92ab5210de3652330 --- /dev/null +++ b/src/vnode/inc/vnodeStatus.h @@ -0,0 +1,47 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_VNODE_STATUS_H +#define TDENGINE_VNODE_STATUS_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "vnodeInt.h" + +typedef enum _VN_STATUS { + TAOS_VN_STATUS_INIT = 0, + TAOS_VN_STATUS_READY = 1, + TAOS_VN_STATUS_CLOSING = 2, + TAOS_VN_STATUS_UPDATING = 3, + TAOS_VN_STATUS_RESET = 4, +} EVnodeStatus; + +bool vnodeSetInitStatus(SVnodeObj* pVnode); +bool vnodeSetReadyStatus(SVnodeObj* pVnode); +bool vnodeSetClosingStatus(SVnodeObj* pVnode); +bool vnodeSetUpdatingStatus(SVnodeObj* pVnode); +bool vnodeSetResetStatus(SVnodeObj* pVnode); + +bool vnodeInInitStatus(SVnodeObj* pVnode); +bool vnodeInReadyStatus(SVnodeObj* pVnode); +bool vnodeInClosingStatus(SVnodeObj* pVnode); +bool vnodeInResetStatus(SVnodeObj* pVnode); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/vnode/inc/vnodeSync.h b/src/vnode/inc/vnodeSync.h new file mode 100644 index 0000000000000000000000000000000000000000..ae02ca17cb50ed35480916d624004e33c74cc611 --- /dev/null +++ b/src/vnode/inc/vnodeSync.h @@ -0,0 +1,39 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_VNODE_SYNC_H +#define TDENGINE_VNODE_SYNC_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "vnodeInt.h" + +uint32_t vnodeGetFileInfo(int32_t vgId, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fver); +int32_t vnodeGetWalInfo(int32_t vgId, char *fileName, int64_t *fileId); +void vnodeNotifyRole(int32_t vgId, int8_t role); +void vnodeCtrlFlow(int32_t vgId, int32_t level); +int32_t vnodeNotifyFileSynced(int32_t vgId, uint64_t fversion); +void vnodeConfirmForard(int32_t vgId, void *wparam, int32_t code); +int32_t vnodeWriteToCache(int32_t vgId, void *wparam, int32_t qtype, void *rparam); +int32_t vnodeGetVersion(int32_t vgId, uint64_t *fver, uint64_t *wver); + +void vnodeConfirmForward(void *pVnode, uint64_t version, int32_t code); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/vnode/inc/vnodeVersion.h b/src/vnode/inc/vnodeVersion.h index 1d086cb21fdab0247038c7f3d32d89b38c19d871..913e3915ab3b911746440be9eab2c7f2e05dce3d 100644 --- a/src/vnode/inc/vnodeVersion.h +++ b/src/vnode/inc/vnodeVersion.h @@ -19,6 +19,7 @@ #ifdef __cplusplus extern "C" { #endif +#include "vnodeInt.h" int32_t vnodeReadVersion(SVnodeObj *pVnode); int32_t vnodeSaveVersion(SVnodeObj *pVnode); diff --git a/src/vnode/inc/vnodeWorker.h b/src/vnode/inc/vnodeWorker.h new file mode 100644 index 0000000000000000000000000000000000000000..01d9d42900ef7e0c56bd396f5c698a43c4b501f8 --- /dev/null +++ b/src/vnode/inc/vnodeWorker.h @@ -0,0 +1,33 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_VNODE_WORKER_H +#define TDENGINE_VNODE_WORKER_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "vnodeInt.h" + +int32_t vnodeInitMWorker(); +void vnodeCleanupMWorker(); +int32_t vnodeCleanupInMWorker(SVnodeObj *pVnode); +int32_t vnodeDestroyInMWorker(SVnodeObj *pVnode); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/vnode/inc/vnodeWrite.h b/src/vnode/inc/vnodeWrite.h new file mode 100644 index 0000000000000000000000000000000000000000..8b3f0fdb58c8a510bcfc6da3aa36adb85297efca --- /dev/null +++ b/src/vnode/inc/vnodeWrite.h @@ -0,0 +1,35 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_VNODE_WRITE_H +#define TDENGINE_VNODE_WRITE_H + +#ifdef __cplusplus +extern "C" { +#endif +#include "vnodeInt.h" + +int32_t vnodeInitWrite(void); +void vnodeCleanupWrite(void); + +int32_t vnodeWriteToWQueue(void *pVnode, void *pHead, int32_t qtype, void *pRpcMsg); +void vnodeFreeFromWQueue(void *pVnode, SVWriteMsg *pWrite); +int32_t vnodeProcessWrite(void *pVnode, void *pHead, int32_t qtype, void *pRspRet); + +#ifdef __cplusplus +} +#endif + +#endif \ No newline at end of file diff --git a/src/vnode/src/vnodeCfg.c b/src/vnode/src/vnodeCfg.c index 2d56157328714a5ae4b8ac05abe3ac4468e361cb..e0881db00072d6482c1eb557f5b35743902e39b2 100644 --- a/src/vnode/src/vnodeCfg.c +++ b/src/vnode/src/vnodeCfg.c @@ -15,13 +15,9 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taosmsg.h" -#include "taoserror.h" #include "cJSON.h" #include "tglobal.h" -#include "tsdb.h" #include "dnode.h" -#include "vnodeInt.h" #include "vnodeCfg.h" static void vnodeLoadCfg(SVnodeObj *pVnode, SCreateVnodeMsg* vnodeMsg) { @@ -242,9 +238,8 @@ int32_t vnodeReadCfg(SVnodeObj *pVnode) { } tstrncpy(node->nodeEp, nodeEp->valuestring, TSDB_EP_LEN); - if (!nodeChanged) { - nodeChanged = dnodeCheckEpChanged(node->nodeId, node->nodeEp); - } + bool changed = dnodeCheckEpChanged(node->nodeId, node->nodeEp); + if (changed) nodeChanged = changed; } ret = TSDB_CODE_SUCCESS; diff --git a/src/vnode/src/vnodeMain.c b/src/vnode/src/vnodeMain.c index cd6d2ea7c0a4842207140d2e4cab1d30722998ad..e286a972dc97f94613614fdefea273cae7459f26 100644 --- a/src/vnode/src/vnodeMain.c +++ b/src/vnode/src/vnodeMain.c @@ -18,76 +18,18 @@ #include "taoserror.h" #include "taosmsg.h" #include "tglobal.h" -#include "trpc.h" -#include "tsdb.h" -#include "tutil.h" -#include "vnode.h" -#include "vnodeInt.h" +// #include "tfs.h" #include "query.h" #include "dnode.h" #include "vnodeCfg.h" +#include "vnodeStatus.h" +#include "vnodeSync.h" #include "vnodeVersion.h" +#include "vnodeMgmt.h" +#include "vnodeWorker.h" +#include "vnodeMain.h" -static SHashObj*tsVnodesHash; -static void vnodeCleanUp(SVnodeObj *pVnode); -static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno); -static uint32_t vnodeGetFileInfo(int32_t vgId, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fversion); -static int32_t vnodeGetWalInfo(int32_t vgId, char *fileName, int64_t *fileId); -static void vnodeNotifyRole(int32_t vgId, int8_t role); -static void vnodeCtrlFlow(int32_t vgId, int32_t level); -static int32_t vnodeNotifyFileSynced(int32_t vgId, uint64_t fversion); -static void vnodeConfirmForard(int32_t vgId, void *wparam, int32_t code); -static int32_t vnodeWriteToCache(int32_t vgId, void *wparam, int32_t qtype, void *rparam); - -#ifndef _SYNC -int64_t syncStart(const SSyncInfo *info) { return NULL; } -int32_t syncForwardToPeer(int64_t rid, void *pHead, void *mhandle, int32_t qtype) { return 0; } -void syncStop(int64_t rid) {} -int32_t syncReconfig(int64_t rid, const SSyncCfg *cfg) { return 0; } -int32_t syncGetNodesRole(int64_t rid, SNodesRole *cfg) { return 0; } -void syncConfirmForward(int64_t rid, uint64_t version, int32_t code) {} -#endif - -char* vnodeStatus[] = { - "init", - "ready", - "closing", - "updating", - "reset" -}; - -int32_t vnodeInitResources() { - int32_t code = syncInit(); - if (code != 0) return code; - - vnodeInitWriteFp(); - vnodeInitReadFp(); - - tsVnodesHash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK); - if (tsVnodesHash == NULL) { - vError("failed to init vnode list"); - return TSDB_CODE_VND_OUT_OF_MEMORY; - } - - if (tsdbInitCommitQueue(tsNumOfCommitThreads) < 0) { - vError("failed to init vnode commit queue"); - return terrno; - } - - return TSDB_CODE_SUCCESS; -} - -void vnodeCleanupResources() { - tsdbDestroyCommitQueue(); - - if (tsVnodesHash != NULL) { - vDebug("vnode list is cleanup"); - taosHashCleanup(tsVnodesHash); - tsVnodesHash = NULL; - } - - syncCleanUp(); -} +static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno); int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) { int32_t code; @@ -154,7 +96,7 @@ int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) { vInfo("vgId:%d, vnode dir is created, walLevel:%d fsyncPeriod:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.walLevel, pVnodeCfg->cfg.fsyncPeriod); - code = vnodeOpen(pVnodeCfg->cfg.vgId, rootDir); + code = vnodeOpen(pVnodeCfg->cfg.vgId); return code; } @@ -169,79 +111,89 @@ int32_t vnodeDrop(int32_t vgId) { vInfo("vgId:%d, vnode will be dropped, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode); pVnode->dropped = 1; + // remove from hash, so new messages wont be consumed + vnodeRemoveFromHash(pVnode); vnodeRelease(pVnode); - vnodeCleanUp(pVnode); + vnodeCleanupInMWorker(pVnode); return TSDB_CODE_SUCCESS; } -int32_t vnodeAlter(void *vparam, SCreateVnodeMsg *pVnodeCfg) { - SVnodeObj *pVnode = vparam; - - // vnode in non-ready state and still needs to return success instead of TSDB_CODE_VND_INVALID_STATUS - // cfgVersion can be corrected by status msg - if (atomic_val_compare_exchange_8(&pVnode->status, TAOS_VN_STATUS_READY, TAOS_VN_STATUS_UPDATING) != TAOS_VN_STATUS_READY) { - vDebug("vgId:%d, vnode is not ready, do alter operation later", pVnode->vgId); - return TSDB_CODE_SUCCESS; - } - +static int32_t vnodeAlterImp(SVnodeObj *pVnode, SCreateVnodeMsg *pVnodeCfg) { int32_t code = vnodeWriteCfg(pVnodeCfg); if (code != TSDB_CODE_SUCCESS) { - pVnode->status = TAOS_VN_STATUS_READY; return code; } code = vnodeReadCfg(pVnode); if (code != TSDB_CODE_SUCCESS) { - pVnode->status = TAOS_VN_STATUS_READY; return code; } code = walAlter(pVnode->wal, &pVnode->walCfg); if (code != TSDB_CODE_SUCCESS) { - pVnode->status = TAOS_VN_STATUS_READY; return code; } code = syncReconfig(pVnode->sync, &pVnode->syncCfg); if (code != TSDB_CODE_SUCCESS) { - pVnode->status = TAOS_VN_STATUS_READY; return code; - } + } if (pVnode->tsdb) { code = tsdbConfigRepo(pVnode->tsdb, &pVnode->tsdbCfg); if (code != TSDB_CODE_SUCCESS) { - pVnode->status = TAOS_VN_STATUS_READY; - return code; + return code; } } - pVnode->status = TAOS_VN_STATUS_READY; - vDebug("vgId:%d, vnode is altered", pVnode->vgId); + return 0; +} + +int32_t vnodeAlter(void *vparam, SCreateVnodeMsg *pVnodeCfg) { + SVnodeObj *pVnode = vparam; - return TSDB_CODE_SUCCESS; + // vnode in non-ready state and still needs to return success instead of TSDB_CODE_VND_INVALID_STATUS + // cfgVersion can be corrected by status msg + if (!vnodeSetUpdatingStatus(pVnode)) { + vDebug("vgId:%d, vnode is not ready, do alter operation later", pVnode->vgId); + return TSDB_CODE_SUCCESS; + } + + int32_t code = vnodeAlterImp(pVnode, pVnodeCfg); + vnodeSetReadyStatus(pVnode); + + if (code != 0) { + vError("vgId:%d, failed to alter vnode, code:0x%x", pVnode->vgId, code); + } else { + vDebug("vgId:%d, vnode is altered", pVnode->vgId); + } + + return code; } -int32_t vnodeOpen(int32_t vnode, char *rootDir) { - char temp[TSDB_FILENAME_LEN]; +int32_t vnodeOpen(int32_t vgId) { + char temp[TSDB_FILENAME_LEN * 3]; + char rootDir[TSDB_FILENAME_LEN * 2]; + snprintf(rootDir, TSDB_FILENAME_LEN * 2, "%s/vnode%d", tsVnodeDir, vgId); SVnodeObj *pVnode = calloc(sizeof(SVnodeObj), 1); if (pVnode == NULL) { - vError("vgId:%d, failed to open vnode since no enough memory", vnode); + vError("vgId:%d, failed to open vnode since no enough memory", vgId); return TAOS_SYSTEM_ERROR(errno); } atomic_add_fetch_32(&pVnode->refCount, 1); - pVnode->vgId = vnode; - pVnode->status = TAOS_VN_STATUS_INIT; + pVnode->vgId = vgId; pVnode->fversion = 0; pVnode->version = 0; pVnode->tsdbCfg.tsdbId = pVnode->vgId; pVnode->rootDir = strdup(rootDir); pVnode->accessState = TSDB_VN_ALL_ACCCESS; tsem_init(&pVnode->sem, 0, 0); + pthread_mutex_init(&pVnode->statusMutex, NULL); + vnodeSetInitStatus(pVnode); int32_t code = vnodeReadCfg(pVnode); if (code != TSDB_CODE_SUCCESS) { @@ -266,16 +218,18 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) { return terrno; } - SCqCfg cqCfg = {0}; - sprintf(cqCfg.user, "_root"); - strcpy(cqCfg.pass, tsInternalPass); - strcpy(cqCfg.db, pVnode->db); - cqCfg.vgId = vnode; - cqCfg.cqWrite = vnodeWriteToWQueue; - pVnode->cq = cqOpen(pVnode, &cqCfg); - if (pVnode->cq == NULL) { - vnodeCleanUp(pVnode); - return terrno; + if (tsEnableStream) { + SCqCfg cqCfg = {0}; + sprintf(cqCfg.user, "_root"); + strcpy(cqCfg.pass, tsInternalPass); + strcpy(cqCfg.db, pVnode->db); + cqCfg.vgId = vgId; + cqCfg.cqWrite = vnodeWriteToCache; + pVnode->cq = cqOpen(pVnode, &cqCfg); + if (pVnode->cq == NULL) { + vnodeCleanUp(pVnode); + return terrno; + } } STsdbAppH appH = {0}; @@ -338,13 +292,13 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) { vDebug("vgId:%d, vnode is opened in %s, pVnode:%p", pVnode->vgId, rootDir, pVnode); tsdbIncCommitRef(pVnode->vgId); - taosHashPut(tsVnodesHash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnodeObj *)); + vnodeAddIntoHash(pVnode); SSyncInfo syncInfo; syncInfo.vgId = pVnode->vgId; syncInfo.version = pVnode->version; syncInfo.syncCfg = pVnode->syncCfg; - sprintf(syncInfo.path, "%s", rootDir); + tstrncpy(syncInfo.path, rootDir, TSDB_FILENAME_LEN); syncInfo.getWalInfo = vnodeGetWalInfo; syncInfo.getFileInfo = vnodeGetFileInfo; syncInfo.writeToCache = vnodeWriteToCache; @@ -352,21 +306,18 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) { syncInfo.notifyRole = vnodeNotifyRole; syncInfo.notifyFlowCtrl = vnodeCtrlFlow; syncInfo.notifyFileSynced = vnodeNotifyFileSynced; + syncInfo.getVersion = vnodeGetVersion; pVnode->sync = syncStart(&syncInfo); -#ifndef _SYNC - pVnode->role = TAOS_SYNC_ROLE_MASTER; -#else if (pVnode->sync <= 0) { - vError("vgId:%d, failed to open sync module, replica:%d reason:%s", pVnode->vgId, pVnode->syncCfg.replica, + vError("vgId:%d, failed to open sync, replica:%d reason:%s", pVnode->vgId, pVnode->syncCfg.replica, tstrerror(terrno)); - vnodeRelease(pVnode); + vnodeRemoveFromHash(pVnode); vnodeCleanUp(pVnode); return terrno; } -#endif - pVnode->status = TAOS_VN_STATUS_READY; + vnodeSetReadyStatus(pVnode); return TSDB_CODE_SUCCESS; } @@ -375,30 +326,16 @@ int32_t vnodeClose(int32_t vgId) { if (pVnode == NULL) return 0; vDebug("vgId:%d, vnode will be closed, pVnode:%p", pVnode->vgId, pVnode); + vnodeRemoveFromHash(pVnode); vnodeRelease(pVnode); vnodeCleanUp(pVnode); return 0; } -void vnodeRelease(void *vparam) { - if (vparam == NULL) return; - SVnodeObj *pVnode = vparam; - int32_t code = 0; - int32_t vgId = pVnode->vgId; - - int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1); - vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", vgId, refCount, pVnode); - assert(refCount >= 0); - - if (refCount > 0) { - if (pVnode->status == TAOS_VN_STATUS_RESET && refCount <= 3) { - tsem_post(&pVnode->sem); - } - return; - } - - vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", vgId, refCount, pVnode); +void vnodeDestroy(SVnodeObj *pVnode) { + int32_t code = 0; + int32_t vgId = pVnode->vgId; if (pVnode->qMgmt) { qCleanupQueryMgmt(pVnode->qMgmt); @@ -461,122 +398,16 @@ void vnodeRelease(void *vparam) { } tsem_destroy(&pVnode->sem); + pthread_mutex_destroy(&pVnode->statusMutex); free(pVnode); tsdbDecCommitRef(vgId); - - int32_t count = taosHashGetSize(tsVnodesHash); - vDebug("vgId:%d, vnode is destroyed, vnodes:%d", vgId, count); -} - -static void vnodeIncRef(void *ptNode) { - assert(ptNode != NULL); - - SVnodeObj **ppVnode = (SVnodeObj **)ptNode; - assert(ppVnode); - assert(*ppVnode); - - SVnodeObj *pVnode = *ppVnode; - atomic_add_fetch_32(&pVnode->refCount, 1); - vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode); } -void *vnodeAcquire(int32_t vgId) { - SVnodeObj **ppVnode = taosHashGetCB(tsVnodesHash, &vgId, sizeof(int32_t), vnodeIncRef, NULL, sizeof(void *)); - - if (ppVnode == NULL || *ppVnode == NULL) { - terrno = TSDB_CODE_VND_INVALID_VGROUP_ID; - vDebug("vgId:%d, not exist", vgId); - return NULL; - } - - return *ppVnode; -} - -void *vnodeGetWal(void *pVnode) { - return ((SVnodeObj *)pVnode)->wal; -} - -static void vnodeBuildVloadMsg(SVnodeObj *pVnode, SStatusMsg *pStatus) { - int64_t totalStorage = 0; - int64_t compStorage = 0; - int64_t pointsWritten = 0; - - if (pVnode->status != TAOS_VN_STATUS_READY) return; - if (pStatus->openVnodes >= TSDB_MAX_VNODES) return; - - if (pVnode->tsdb) { - tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage); - } - - SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++]; - pLoad->vgId = htonl(pVnode->vgId); - pLoad->cfgVersion = htonl(pVnode->cfgVersion); - pLoad->totalStorage = htobe64(totalStorage); - pLoad->compStorage = htobe64(compStorage); - pLoad->pointsWritten = htobe64(pointsWritten); - pLoad->status = pVnode->status; - pLoad->role = pVnode->role; - pLoad->replica = pVnode->syncCfg.replica; -} - -int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) { - SHashMutableIterator *pIter = taosHashCreateIter(tsVnodesHash); - while (taosHashIterNext(pIter)) { - SVnodeObj **pVnode = taosHashIterGet(pIter); - if (pVnode == NULL) continue; - if (*pVnode == NULL) continue; - - (*numOfVnodes)++; - if (*numOfVnodes >= TSDB_MAX_VNODES) { - vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES); - continue; - } else { - vnodeList[*numOfVnodes - 1] = (*pVnode)->vgId; - } - } - - taosHashDestroyIter(pIter); - return TSDB_CODE_SUCCESS; -} - -void vnodeBuildStatusMsg(void *param) { - SStatusMsg *pStatus = param; - SHashMutableIterator *pIter = taosHashCreateIter(tsVnodesHash); - - while (taosHashIterNext(pIter)) { - SVnodeObj **pVnode = taosHashIterGet(pIter); - if (pVnode == NULL) continue; - if (*pVnode == NULL) continue; - - vnodeBuildVloadMsg(*pVnode, pStatus); - } - - taosHashDestroyIter(pIter); -} - -void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes) { - for (int32_t i = 0; i < numOfVnodes; ++i) { - pAccess[i].vgId = htonl(pAccess[i].vgId); - SVnodeObj *pVnode = vnodeAcquire(pAccess[i].vgId); - if (pVnode != NULL) { - pVnode->accessState = pAccess[i].accessState; - if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) { - vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState); - } - vnodeRelease(pVnode); - } - } -} - -static void vnodeCleanUp(SVnodeObj *pVnode) { - // remove from hash, so new messages wont be consumed - taosHashRemove(tsVnodesHash, &pVnode->vgId, sizeof(int32_t)); - - if (pVnode->status != TAOS_VN_STATUS_INIT) { +void vnodeCleanUp(SVnodeObj *pVnode) { + if (!vnodeInInitStatus(pVnode)) { // it may be in updateing or reset state, then it shall wait int32_t i = 0; - while (atomic_val_compare_exchange_8(&pVnode->status, TAOS_VN_STATUS_READY, TAOS_VN_STATUS_CLOSING) != - TAOS_VN_STATUS_READY) { + while (!vnodeSetClosingStatus(pVnode)) { if (++i % 1000 == 0) { sched_yield(); } @@ -597,21 +428,22 @@ static void vnodeCleanUp(SVnodeObj *pVnode) { vnodeRelease(pVnode); } -// TODO: this is a simple implement static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) { SVnodeObj *pVnode = arg; if (eno != TSDB_CODE_SUCCESS) { vError("vgId:%d, failed to commit since %s, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, tstrerror(eno), pVnode->fversion, pVnode->version); + pVnode->isCommiting = 0; pVnode->isFull = 1; return 0; } if (status == TSDB_STATUS_COMMIT_START) { + pVnode->isCommiting = 1; pVnode->fversion = pVnode->version; vDebug("vgId:%d, start commit, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, pVnode->fversion, pVnode->version); - if (pVnode->status != TAOS_VN_STATUS_INIT) { + if (!vnodeInInitStatus(pVnode)) { return walRenew(pVnode->wal); } return 0; @@ -619,8 +451,9 @@ static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) { if (status == TSDB_STATUS_COMMIT_OVER) { vDebug("vgId:%d, commit over, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, pVnode->fversion, pVnode->version); + pVnode->isCommiting = 0; pVnode->isFull = 0; - if (pVnode->status != TAOS_VN_STATUS_INIT) { + if (!vnodeInInitStatus(pVnode)) { walRemoveOneOldFile(pVnode->wal); } return vnodeSaveVersion(pVnode); @@ -629,73 +462,12 @@ static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) { return 0; } -static uint32_t vnodeGetFileInfo(int32_t vgId, char *name, uint32_t *index, uint32_t eindex, int64_t *size, - uint64_t *fversion) { - SVnodeObj *pVnode = vnodeAcquire(vgId); - if (pVnode == NULL) { - vError("vgId:%d, vnode not found while get file info", vgId); - return 0; - } - - *fversion = pVnode->fversion; - uint32_t ret = tsdbGetFileInfo(pVnode->tsdb, name, index, eindex, size); - - vnodeRelease(pVnode); - return ret; -} - -static int32_t vnodeGetWalInfo(int32_t vgId, char *fileName, int64_t *fileId) { - SVnodeObj *pVnode = vnodeAcquire(vgId); - if (pVnode == NULL) { - vError("vgId:%d, vnode not found while get wal info", vgId); - return -1; - } - - int32_t code = walGetWalFile(pVnode->wal, fileName, fileId); - - vnodeRelease(pVnode); - return code; -} - -static void vnodeNotifyRole(int32_t vgId, int8_t role) { - SVnodeObj *pVnode = vnodeAcquire(vgId); - if (pVnode == NULL) { - vTrace("vgId:%d, vnode not found while notify role", vgId); - return; - } - - vInfo("vgId:%d, sync role changed from %s to %s", pVnode->vgId, syncRole[pVnode->role], syncRole[role]); - pVnode->role = role; - dnodeSendStatusMsgToMnode(); - - if (pVnode->role == TAOS_SYNC_ROLE_MASTER) { - cqStart(pVnode->cq); - } else { - cqStop(pVnode->cq); - } - - vnodeRelease(pVnode); -} - -static void vnodeCtrlFlow(int32_t vgId, int32_t level) { - SVnodeObj *pVnode = vnodeAcquire(vgId); - if (pVnode == NULL) { - vTrace("vgId:%d, vnode not found while flow ctrl", vgId); - return; - } - - pVnode->flowctrlLevel = level; - vDebug("vgId:%d, set flowctrl level:%d", pVnode->vgId, level); - - vnodeRelease(pVnode); -} - -static int32_t vnodeResetTsdb(SVnodeObj *pVnode) { +int32_t vnodeReset(SVnodeObj *pVnode) { char rootDir[128] = "\0"; sprintf(rootDir, "%s/tsdb", pVnode->rootDir); - if (pVnode->status != TAOS_VN_STATUS_CLOSING && pVnode->status != TAOS_VN_STATUS_INIT) { - pVnode->status = TAOS_VN_STATUS_RESET; + if (!vnodeSetResetStatus(pVnode)) { + return -1; } void *tsdb = pVnode->tsdb; @@ -718,50 +490,8 @@ static int32_t vnodeResetTsdb(SVnodeObj *pVnode) { appH.cqDropFunc = cqDrop; pVnode->tsdb = tsdbOpenRepo(rootDir, &appH); - pVnode->status = TAOS_VN_STATUS_READY; + vnodeSetReadyStatus(pVnode); vnodeRelease(pVnode); return 0; } - -static int32_t vnodeNotifyFileSynced(int32_t vgId, uint64_t fversion) { - SVnodeObj *pVnode = vnodeAcquire(vgId); - if (pVnode == NULL) { - vError("vgId:%d, vnode not found while notify file synced", vgId); - return 0; - } - - pVnode->fversion = fversion; - pVnode->version = fversion; - vnodeSaveVersion(pVnode); - - vDebug("vgId:%d, data file is synced, fver:%" PRIu64 " vver:%" PRIu64, vgId, fversion, fversion); - int32_t code = vnodeResetTsdb(pVnode); - - vnodeRelease(pVnode); - return code; -} - -void vnodeConfirmForard(int32_t vgId, void *wparam, int32_t code) { - void *pVnode = vnodeAcquire(vgId); - if (pVnode == NULL) { - vError("vgId:%d, vnode not found while confirm forward", vgId); - return; - } - - dnodeSendRpcVWriteRsp(pVnode, wparam, code); - vnodeRelease(pVnode); -} - -static int32_t vnodeWriteToCache(int32_t vgId, void *wparam, int32_t qtype, void *rparam) { - SVnodeObj *pVnode = vnodeAcquire(vgId); - if (pVnode == NULL) { - vError("vgId:%d, vnode not found while write to cache", vgId); - return TSDB_CODE_VND_INVALID_VGROUP_ID; - } - - int32_t code = vnodeWriteToWQueue(pVnode, wparam, qtype, rparam); - - vnodeRelease(pVnode); - return code; -} diff --git a/src/vnode/src/vnodeMgmt.c b/src/vnode/src/vnodeMgmt.c new file mode 100644 index 0000000000000000000000000000000000000000..cf42690d7d0208f50b781e20a4c7c8885f8c5fba --- /dev/null +++ b/src/vnode/src/vnodeMgmt.c @@ -0,0 +1,192 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "dnode.h" +#include "vnodeStatus.h" +#include "vnodeWorker.h" +#include "vnodeRead.h" +#include "vnodeWrite.h" +#include "vnodeMain.h" + +static SHashObj *tsVnodesHash = NULL; + +static int32_t vnodeInitHash(void); +static void vnodeCleanupHash(void); +static void vnodeIncRef(void *ptNode); + +static SStep tsVnodeSteps[] = { + {"vnode-worker", vnodeInitMWorker, vnodeCleanupMWorker}, + {"vnode-write", vnodeInitWrite, vnodeCleanupWrite}, + {"vnode-read", vnodeInitRead, vnodeCleanupRead}, + {"vnode-hash", vnodeInitHash, vnodeCleanupHash}, + {"tsdb-queue", tsdbInitCommitQueue, tsdbDestroyCommitQueue} +}; + +int32_t vnodeInitMgmt() { + int32_t stepSize = sizeof(tsVnodeSteps) / sizeof(SStep); + return dnodeStepInit(tsVnodeSteps, stepSize); +} + +void vnodeCleanupMgmt() { + int32_t stepSize = sizeof(tsVnodeSteps) / sizeof(SStep); + dnodeStepCleanup(tsVnodeSteps, stepSize); +} + +static int32_t vnodeInitHash() { + tsVnodesHash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK); + if (tsVnodesHash == NULL) { + vError("failed to init vnode mgmt"); + return -1; + } + + return 0; +} + +static void vnodeCleanupHash() { + if (tsVnodesHash != NULL) { + vDebug("vnode mgmt is cleanup"); + taosHashCleanup(tsVnodesHash); + tsVnodesHash = NULL; + } +} + +void *vnodeGetWal(void *pVnode) { + return ((SVnodeObj *)pVnode)->wal; +} + +void vnodeAddIntoHash(SVnodeObj *pVnode) { + taosHashPut(tsVnodesHash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnodeObj *)); +} + +void vnodeRemoveFromHash(SVnodeObj *pVnode) { + taosHashRemove(tsVnodesHash, &pVnode->vgId, sizeof(int32_t)); +} + +static void vnodeIncRef(void *ptNode) { + assert(ptNode != NULL); + + SVnodeObj **ppVnode = (SVnodeObj **)ptNode; + assert(ppVnode); + assert(*ppVnode); + + SVnodeObj *pVnode = *ppVnode; + atomic_add_fetch_32(&pVnode->refCount, 1); + vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode); +} + +void *vnodeAcquire(int32_t vgId) { + SVnodeObj **ppVnode = taosHashGetCB(tsVnodesHash, &vgId, sizeof(int32_t), vnodeIncRef, NULL, sizeof(void *)); + + if (ppVnode == NULL || *ppVnode == NULL) { + terrno = TSDB_CODE_VND_INVALID_VGROUP_ID; + vDebug("vgId:%d, not exist", vgId); + return NULL; + } + + return *ppVnode; +} + +void vnodeRelease(void *vparam) { + SVnodeObj *pVnode = vparam; + if (vparam == NULL) return; + + int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1); + vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", pVnode->vgId, refCount, pVnode); + assert(refCount >= 0); + + if (refCount > 0) { + if (vnodeInResetStatus(pVnode) && refCount <= 3) { + tsem_post(&pVnode->sem); + } + } else { + vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", pVnode->vgId, refCount, pVnode); + vnodeDestroyInMWorker(pVnode); + int32_t count = taosHashGetSize(tsVnodesHash); + vDebug("vgId:%d, vnode is destroyed, vnodes:%d", pVnode->vgId, count); + } +} + +static void vnodeBuildVloadMsg(SVnodeObj *pVnode, SStatusMsg *pStatus) { + int64_t totalStorage = 0; + int64_t compStorage = 0; + int64_t pointsWritten = 0; + + if (!vnodeInReadyStatus(pVnode)) return; + if (pStatus->openVnodes >= TSDB_MAX_VNODES) return; + + if (pVnode->tsdb) { + tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage); + } + + SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++]; + pLoad->vgId = htonl(pVnode->vgId); + pLoad->cfgVersion = htonl(pVnode->cfgVersion); + pLoad->totalStorage = htobe64(totalStorage); + pLoad->compStorage = htobe64(compStorage); + pLoad->pointsWritten = htobe64(pointsWritten); + pLoad->status = pVnode->status; + pLoad->role = pVnode->role; + pLoad->replica = pVnode->syncCfg.replica; +} + +int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) { + void *pIter = taosHashIterate(tsVnodesHash, NULL); + while (pIter) { + SVnodeObj **pVnode = pIter; + if (*pVnode) { + + (*numOfVnodes)++; + if (*numOfVnodes >= TSDB_MAX_VNODES) { + vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES); + continue; + } else { + vnodeList[*numOfVnodes - 1] = (*pVnode)->vgId; + } + + } + + pIter = taosHashIterate(tsVnodesHash, pIter); + } + return TSDB_CODE_SUCCESS; +} + +void vnodeBuildStatusMsg(void *param) { + SStatusMsg *pStatus = param; + + void *pIter = taosHashIterate(tsVnodesHash, NULL); + while (pIter) { + SVnodeObj **pVnode = pIter; + if (*pVnode) { + vnodeBuildVloadMsg(*pVnode, pStatus); + } + pIter = taosHashIterate(tsVnodesHash, pIter); + } +} + +void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes) { + for (int32_t i = 0; i < numOfVnodes; ++i) { + pAccess[i].vgId = htonl(pAccess[i].vgId); + SVnodeObj *pVnode = vnodeAcquire(pAccess[i].vgId); + if (pVnode != NULL) { + pVnode->accessState = pAccess[i].accessState; + if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) { + vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState); + } + vnodeRelease(pVnode); + } + } +} \ No newline at end of file diff --git a/src/vnode/src/vnodeRead.c b/src/vnode/src/vnodeRead.c index 5ef79cfbf01fb49ebbfd4603b8a9163b3ae9bc32..34921a93b344657889ae868d06fe22864495969e 100644 --- a/src/vnode/src/vnodeRead.c +++ b/src/vnode/src/vnodeRead.c @@ -16,26 +16,26 @@ #define _DEFAULT_SOURCE #define _NON_BLOCKING_RETRIEVE 0 #include "os.h" -#include "tglobal.h" -#include "taoserror.h" #include "taosmsg.h" -#include "query.h" -#include "trpc.h" -#include "tsdb.h" -#include "vnode.h" -#include "vnodeInt.h" #include "tqueue.h" +#include "tglobal.h" +#include "query.h" +#include "vnodeStatus.h" static int32_t (*vnodeProcessReadMsgFp[TSDB_MSG_TYPE_MAX])(SVnodeObj *pVnode, SVReadMsg *pRead); static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead); static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead); static int32_t vnodeNotifyCurrentQhandle(void* handle, void* qhandle, int32_t vgId); -void vnodeInitReadFp(void) { +int32_t vnodeInitRead(void) { vnodeProcessReadMsgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessQueryMsg; vnodeProcessReadMsgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessFetchMsg; + + return 0; } +void vnodeCleanupRead() {} + // // After the fetch request enters the vnode queue, if the vnode cannot provide services, the process function are // still required, or there will be a deadlock, so we don’t do any check here, but put the check codes before the @@ -54,7 +54,7 @@ int32_t vnodeProcessRead(void *vparam, SVReadMsg *pRead) { } static int32_t vnodeCheckRead(SVnodeObj *pVnode) { - if (pVnode->status != TAOS_VN_STATUS_READY) { + if (!vnodeInReadyStatus(pVnode)) { vDebug("vgId:%d, vnode status is %s, refCount:%d pVnode:%p", pVnode->vgId, vnodeStatus[pVnode->status], pVnode->refCount, pVnode); return TSDB_CODE_APP_NOT_READY; @@ -275,41 +275,40 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { vDebug("vgId:%d, QInfo:%p, dnode continues to exec query", pVnode->vgId, *qhandle); + // In the retrieve blocking model, only 50% CPU will be used in query processing + if (tsHalfCoresForQuery) { + qTableQuery(*qhandle); // do execute query + qReleaseQInfo(pVnode->qMgmt, (void **)&qhandle, false); + } else { + bool freehandle = false; + bool buildRes = qTableQuery(*qhandle); // do execute query -#if _NON_BLOCKING_RETRIEVE - bool freehandle = false; - bool buildRes = qTableQuery(*qhandle); // do execute query - - // build query rsp, the retrieve request has reached here already - if (buildRes) { - // update the connection info according to the retrieve connection - pRead->rpcHandle = qGetResultRetrieveMsg(*qhandle); - assert(pRead->rpcHandle != NULL); - - vDebug("vgId:%d, QInfo:%p, start to build retrieval rsp after query paused, %p", pVnode->vgId, *qhandle, - pRead->rpcHandle); + // build query rsp, the retrieve request has reached here already + if (buildRes) { + // update the connection info according to the retrieve connection + pRead->rpcHandle = qGetResultRetrieveMsg(*qhandle); + assert(pRead->rpcHandle != NULL); - // set the real rsp error code - pRead->code = vnodeDumpQueryResult(&pRead->rspRet, pVnode, qhandle, &freehandle, pRead->rpcHandle); + vDebug("vgId:%d, QInfo:%p, start to build retrieval rsp after query paused, %p", pVnode->vgId, *qhandle, + pRead->rpcHandle); - // NOTE: set return code to be TSDB_CODE_QRY_HAS_RSP to notify dnode to return msg to client - code = TSDB_CODE_QRY_HAS_RSP; - } else { - void* h1 = qGetResultRetrieveMsg(*qhandle); - assert(h1 == NULL); + // set the real rsp error code + pRead->code = vnodeDumpQueryResult(&pRead->rspRet, pVnode, qhandle, &freehandle, pRead->rpcHandle); - freehandle = qQueryCompleted(*qhandle); - } + // NOTE: set return code to be TSDB_CODE_QRY_HAS_RSP to notify dnode to return msg to client + code = TSDB_CODE_QRY_HAS_RSP; + } else { + void *h1 = qGetResultRetrieveMsg(*qhandle); + assert(h1 == NULL); + freehandle = qQueryCompleted(*qhandle); + } - // NOTE: if the qhandle is not put into vread queue or query is completed, free the qhandle. - // If the building of result is not required, simply free it. Otherwise, mandatorily free the qhandle - if (freehandle || (!buildRes)) { - qReleaseQInfo(pVnode->qMgmt, (void **)&qhandle, freehandle); + // NOTE: if the qhandle is not put into vread queue or query is completed, free the qhandle. + // If the building of result is not required, simply free it. Otherwise, mandatorily free the qhandle + if (freehandle || (!buildRes)) { + qReleaseQInfo(pVnode->qMgmt, (void **)&qhandle, freehandle); + } } -#else - qTableQuery(*qhandle); // do execute query - qReleaseQInfo(pVnode->qMgmt, (void **)&qhandle, false); -#endif } return code; @@ -375,14 +374,16 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) { freeHandle = true; } else { // result is not ready, return immediately assert(buildRes == true); -#if _NON_BLOCKING_RETRIEVE - if (!buildRes) { - assert(pRead->rpcHandle != NULL); - qReleaseQInfo(pVnode->qMgmt, (void **)&handle, false); - return TSDB_CODE_QRY_NOT_READY; + // Only effects in the non-blocking model + if (!tsHalfCoresForQuery) { + if (!buildRes) { + assert(pRead->rpcHandle != NULL); + + qReleaseQInfo(pVnode->qMgmt, (void **)&handle, false); + return TSDB_CODE_QRY_NOT_READY; + } } -#endif // ahandle is the sqlObj pointer code = vnodeDumpQueryResult(pRet, pVnode, handle, &freeHandle, pRead->rpcHandle); diff --git a/src/vnode/src/vnodeStatus.c b/src/vnode/src/vnodeStatus.c new file mode 100644 index 0000000000000000000000000000000000000000..d09a6a86631837db6799aca5b4df87ae22e07853 --- /dev/null +++ b/src/vnode/src/vnodeStatus.c @@ -0,0 +1,142 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "vnodeStatus.h" + +char* vnodeStatus[] = { + "init", + "ready", + "closing", + "updating", + "reset" +}; + +bool vnodeSetInitStatus(SVnodeObj* pVnode) { + pthread_mutex_lock(&pVnode->statusMutex); + pVnode->status = TAOS_VN_STATUS_INIT; + pthread_mutex_unlock(&pVnode->statusMutex); + return true; +} + +bool vnodeSetReadyStatus(SVnodeObj* pVnode) { + bool set = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_READY || + pVnode->status == TAOS_VN_STATUS_UPDATING || pVnode->status == TAOS_VN_STATUS_RESET) { + pVnode->status = TAOS_VN_STATUS_READY; + set = true; + } else { + vDebug("vgId:%d, cannot set status:ready, old:%s", pVnode->vgId, vnodeStatus[pVnode->status]); + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return set; +} + +bool vnodeSetClosingStatus(SVnodeObj* pVnode) { + bool set = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status == TAOS_VN_STATUS_READY) { + pVnode->status = TAOS_VN_STATUS_CLOSING; + set = true; + } else { + vTrace("vgId:%d, cannot set status:closing, old:%s", pVnode->vgId, vnodeStatus[pVnode->status]); + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return set; +} + +bool vnodeSetUpdatingStatus(SVnodeObj* pVnode) { + bool set = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status == TAOS_VN_STATUS_READY) { + pVnode->status = TAOS_VN_STATUS_UPDATING; + set = true; + } else { + vDebug("vgId:%d, cannot set status:updating, old:%s", pVnode->vgId, vnodeStatus[pVnode->status]); + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return set; +} + +bool vnodeSetResetStatus(SVnodeObj* pVnode) { + bool set = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status != TAOS_VN_STATUS_CLOSING && pVnode->status != TAOS_VN_STATUS_INIT) { + pVnode->status = TAOS_VN_STATUS_RESET; + set = true; + } else { + vDebug("vgId:%d, cannot set status:reset, old:%s", pVnode->vgId, vnodeStatus[pVnode->status]); + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return set; +} + +bool vnodeInInitStatus(SVnodeObj* pVnode) { + bool in = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status == TAOS_VN_STATUS_INIT) { + in = true; + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return in; +} + +bool vnodeInReadyStatus(SVnodeObj* pVnode) { + bool in = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status == TAOS_VN_STATUS_READY) { + in = true; + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return in; +} + +bool vnodeInClosingStatus(SVnodeObj* pVnode) { + bool in = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status == TAOS_VN_STATUS_CLOSING) { + in = true; + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return in; +} + +bool vnodeInResetStatus(SVnodeObj* pVnode) { + bool in = false; + pthread_mutex_lock(&pVnode->statusMutex); + + if (pVnode->status == TAOS_VN_STATUS_RESET) { + in = true; + } + + pthread_mutex_unlock(&pVnode->statusMutex); + return in; +} diff --git a/src/vnode/src/vnodeSync.c b/src/vnode/src/vnodeSync.c new file mode 100644 index 0000000000000000000000000000000000000000..c67132c41f2a925c5d3224f8b59f0e94b5e8f1c3 --- /dev/null +++ b/src/vnode/src/vnodeSync.c @@ -0,0 +1,151 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "taosmsg.h" +#include "query.h" +#include "dnode.h" +#include "vnodeVersion.h" +#include "vnodeMain.h" + +uint32_t vnodeGetFileInfo(int32_t vgId, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fver) { + SVnodeObj *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vError("vgId:%d, vnode not found while get file info", vgId); + return 0; + } + + *fver = pVnode->fversion; + uint32_t ret = tsdbGetFileInfo(pVnode->tsdb, name, index, eindex, size); + + vnodeRelease(pVnode); + return ret; +} + +int32_t vnodeGetWalInfo(int32_t vgId, char *fileName, int64_t *fileId) { + SVnodeObj *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vError("vgId:%d, vnode not found while get wal info", vgId); + return -1; + } + + int32_t code = walGetWalFile(pVnode->wal, fileName, fileId); + + vnodeRelease(pVnode); + return code; +} + +void vnodeNotifyRole(int32_t vgId, int8_t role) { + SVnodeObj *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vTrace("vgId:%d, vnode not found while notify role", vgId); + return; + } + + vInfo("vgId:%d, sync role changed from %s to %s", pVnode->vgId, syncRole[pVnode->role], syncRole[role]); + pVnode->role = role; + dnodeSendStatusMsgToMnode(); + + if (pVnode->role == TAOS_SYNC_ROLE_MASTER) { + cqStart(pVnode->cq); + } else { + cqStop(pVnode->cq); + } + + vnodeRelease(pVnode); +} + +void vnodeCtrlFlow(int32_t vgId, int32_t level) { + SVnodeObj *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vTrace("vgId:%d, vnode not found while flow ctrl", vgId); + return; + } + + if (pVnode->flowctrlLevel != level) { + vDebug("vgId:%d, set flowctrl level from %d to %d", pVnode->vgId, pVnode->flowctrlLevel, level); + pVnode->flowctrlLevel = level; + } + + vnodeRelease(pVnode); +} + +int32_t vnodeNotifyFileSynced(int32_t vgId, uint64_t fversion) { + SVnodeObj *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vError("vgId:%d, vnode not found while notify file synced", vgId); + return 0; + } + + pVnode->fversion = fversion; + pVnode->version = fversion; + vnodeSaveVersion(pVnode); + + vDebug("vgId:%d, data file is synced, fver:%" PRIu64 " vver:%" PRIu64, vgId, fversion, fversion); + int32_t code = vnodeReset(pVnode); + + vnodeRelease(pVnode); + return code; +} + +void vnodeConfirmForard(int32_t vgId, void *wparam, int32_t code) { + void *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vError("vgId:%d, vnode not found while confirm forward", vgId); + return; + } + + dnodeSendRpcVWriteRsp(pVnode, wparam, code); + vnodeRelease(pVnode); +} + +int32_t vnodeWriteToCache(int32_t vgId, void *wparam, int32_t qtype, void *rparam) { + SVnodeObj *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vError("vgId:%d, vnode not found while write to cache", vgId); + return TSDB_CODE_VND_INVALID_VGROUP_ID; + } + + int32_t code = vnodeWriteToWQueue(pVnode, wparam, qtype, rparam); + + vnodeRelease(pVnode); + return code; +} + +int32_t vnodeGetVersion(int32_t vgId, uint64_t *fver, uint64_t *wver) { + SVnodeObj *pVnode = vnodeAcquire(vgId); + if (pVnode == NULL) { + vError("vgId:%d, vnode not found while write to cache", vgId); + return -1; + } + + int32_t code = 0; + if (pVnode->isCommiting) { + vDebug("vgId:%d, vnode is commiting while get version", vgId); + code = -1; + } else { + *fver = pVnode->fversion; + *wver = pVnode->version; + } + + vnodeRelease(pVnode); + return code; +} + +void vnodeConfirmForward(void *vparam, uint64_t version, int32_t code) { + SVnodeObj *pVnode = vparam; + syncConfirmForward(pVnode->sync, version, code); +} diff --git a/src/vnode/src/vnodeVersion.c b/src/vnode/src/vnodeVersion.c index 8f6360b4f98eac8f394f5078ed2b025cea4192b0..fb3b3ebd9e9b9b4e6ac4d3f62de3b7599a13cae9 100644 --- a/src/vnode/src/vnodeVersion.c +++ b/src/vnode/src/vnodeVersion.c @@ -15,11 +15,8 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taoserror.h" #include "cJSON.h" #include "tglobal.h" -#include "tsdb.h" -#include "vnodeInt.h" #include "vnodeVersion.h" int32_t vnodeReadVersion(SVnodeObj *pVnode) { diff --git a/src/vnode/src/vnodeWorker.c b/src/vnode/src/vnodeWorker.c new file mode 100644 index 0000000000000000000000000000000000000000..d6053cf18e2569a197c4bdf96ff6394ede716e42 --- /dev/null +++ b/src/vnode/src/vnodeWorker.c @@ -0,0 +1,204 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "taoserror.h" +#include "taosmsg.h" +#include "tutil.h" +#include "tqueue.h" +#include "tglobal.h" +#include "vnodeWorker.h" +#include "vnodeMain.h" + +typedef enum { + VNODE_WORKER_ACTION_CLEANUP, + VNODE_WORKER_ACTION_DESTROUY +} EVMWorkerAction; + +typedef struct { + int32_t vgId; + int32_t code; + void * rpcHandle; + SVnodeObj *pVnode; + EVMWorkerAction action; +} SVMWorkerMsg; + +typedef struct { + pthread_t thread; + int32_t workerId; +} SVMWorker; + +typedef struct { + int32_t curNum; + int32_t maxNum; + SVMWorker *worker; +} SVMWorkerPool; + +static SVMWorkerPool tsVMWorkerPool; +static taos_qset tsVMWorkerQset; +static taos_queue tsVMWorkerQueue; + +static void *vnodeMWorkerFunc(void *param); + +static int32_t vnodeStartMWorker() { + tsVMWorkerQueue = taosOpenQueue(); + if (tsVMWorkerQueue == NULL) return TSDB_CODE_DND_OUT_OF_MEMORY; + + taosAddIntoQset(tsVMWorkerQset, tsVMWorkerQueue, NULL); + + for (int32_t i = tsVMWorkerPool.curNum; i < tsVMWorkerPool.maxNum; ++i) { + SVMWorker *pWorker = tsVMWorkerPool.worker + i; + pWorker->workerId = i; + + pthread_attr_t thAttr; + pthread_attr_init(&thAttr); + pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); + + if (pthread_create(&pWorker->thread, &thAttr, vnodeMWorkerFunc, pWorker) != 0) { + vError("failed to create thread to process vmworker queue, reason:%s", strerror(errno)); + } + + pthread_attr_destroy(&thAttr); + + tsVMWorkerPool.curNum = i + 1; + vDebug("vmworker:%d is launched, total:%d", pWorker->workerId, tsVMWorkerPool.maxNum); + } + + vDebug("vmworker queue:%p is allocated", tsVMWorkerQueue); + return TSDB_CODE_SUCCESS; +} + +int32_t vnodeInitMWorker() { + tsVMWorkerQset = taosOpenQset(); + + tsVMWorkerPool.maxNum = 1; + tsVMWorkerPool.curNum = 0; + tsVMWorkerPool.worker = calloc(sizeof(SVMWorker), tsVMWorkerPool.maxNum); + + if (tsVMWorkerPool.worker == NULL) return -1; + for (int32_t i = 0; i < tsVMWorkerPool.maxNum; ++i) { + SVMWorker *pWorker = tsVMWorkerPool.worker + i; + pWorker->workerId = i; + vDebug("vmworker:%d is created", i); + } + + vDebug("vmworker is initialized, num:%d qset:%p", tsVMWorkerPool.maxNum, tsVMWorkerQset); + + return vnodeStartMWorker(); +} + +static void vnodeStopMWorker() { + vDebug("vmworker queue:%p is freed", tsVMWorkerQueue); + taosCloseQueue(tsVMWorkerQueue); + tsVMWorkerQueue = NULL; +} + +void vnodeCleanupMWorker() { + for (int32_t i = 0; i < tsVMWorkerPool.maxNum; ++i) { + SVMWorker *pWorker = tsVMWorkerPool.worker + i; + if (pWorker->thread) { + taosQsetThreadResume(tsVMWorkerQset); + } + vDebug("vmworker:%d is closed", i); + } + + for (int32_t i = 0; i < tsVMWorkerPool.maxNum; ++i) { + SVMWorker *pWorker = tsVMWorkerPool.worker + i; + vDebug("vmworker:%d start to join", i); + if (pWorker->thread) { + pthread_join(pWorker->thread, NULL); + } + vDebug("vmworker:%d join success", i); + } + + vDebug("vmworker is closed, qset:%p", tsVMWorkerQset); + + taosCloseQset(tsVMWorkerQset); + tsVMWorkerQset = NULL; + tfree(tsVMWorkerPool.worker); + + vnodeStopMWorker(); +} + +static int32_t vnodeWriteIntoMWorker(SVnodeObj *pVnode, EVMWorkerAction action, void *rpcHandle) { + SVMWorkerMsg *pMsg = taosAllocateQitem(sizeof(SVMWorkerMsg)); + if (pMsg == NULL) return TSDB_CODE_VND_OUT_OF_MEMORY; + + pMsg->vgId = pVnode->vgId; + pMsg->pVnode = pVnode; + pMsg->rpcHandle = rpcHandle; + pMsg->action = action; + + int32_t code = taosWriteQitem(tsVMWorkerQueue, TAOS_QTYPE_RPC, pMsg); + if (code == 0) code = TSDB_CODE_DND_ACTION_IN_PROGRESS; + + return code; +} + +int32_t vnodeCleanupInMWorker(SVnodeObj *pVnode) { + vTrace("vgId:%d, will cleanup in vmworker", pVnode->vgId); + return vnodeWriteIntoMWorker(pVnode, VNODE_WORKER_ACTION_CLEANUP, NULL); +} + +int32_t vnodeDestroyInMWorker(SVnodeObj *pVnode) { + vTrace("vgId:%d, will destroy in vmworker", pVnode->vgId); + return vnodeWriteIntoMWorker(pVnode, VNODE_WORKER_ACTION_DESTROUY, NULL); +} + +static void vnodeFreeMWorkerMsg(SVMWorkerMsg *pMsg) { + vTrace("vgId:%d, disposed in vmworker", pMsg->vgId); + taosFreeQitem(pMsg); +} + +static void vnodeSendVMWorkerRpcRsp(SVMWorkerMsg *pMsg) { + if (pMsg->rpcHandle != NULL) { + SRpcMsg rpcRsp = {.handle = pMsg->rpcHandle, .code = pMsg->code}; + rpcSendResponse(&rpcRsp); + } + + vnodeFreeMWorkerMsg(pMsg); +} + +static void vnodeProcessMWorkerMsg(SVMWorkerMsg *pMsg) { + pMsg->code = 0; + + switch (pMsg->action) { + case VNODE_WORKER_ACTION_CLEANUP: + vnodeCleanUp(pMsg->pVnode); + break; + case VNODE_WORKER_ACTION_DESTROUY: + vnodeDestroy(pMsg->pVnode); + break; + default: + break; + } +} + +static void *vnodeMWorkerFunc(void *param) { + while (1) { + SVMWorkerMsg *pMsg = NULL; + if (taosReadQitemFromQset(tsVMWorkerQset, NULL, (void **)&pMsg, NULL) == 0) { + vDebug("qset:%p, vmworker got no message from qset, exiting", tsVMWorkerQset); + break; + } + + vTrace("vgId:%d, action:%d will be processed in vmworker queue", pMsg->vgId, pMsg->action); + vnodeProcessMWorkerMsg(pMsg); + vnodeSendVMWorkerRpcRsp(pMsg); + } + + return NULL; +} diff --git a/src/vnode/src/vnodeWrite.c b/src/vnode/src/vnodeWrite.c index 57bd407cd1f47ab875c57326f361f80103ecc2b6..a826a4903fe6a498db0769a1a1d883791bbb3802 100644 --- a/src/vnode/src/vnodeWrite.c +++ b/src/vnode/src/vnodeWrite.c @@ -19,17 +19,9 @@ #include "taoserror.h" #include "tglobal.h" #include "tqueue.h" -#include "trpc.h" -#include "tsdb.h" -#include "twal.h" -#include "tsync.h" #include "ttimer.h" -#include "tdataformat.h" -#include "vnode.h" -#include "vnodeInt.h" -#include "syncInt.h" -#include "tcq.h" #include "dnode.h" +#include "vnodeStatus.h" #define MAX_QUEUED_MSG_NUM 10000 @@ -43,15 +35,19 @@ static int32_t vnodeProcessDropStableMsg(SVnodeObj *pVnode, void *pCont, SRspRet static int32_t vnodeProcessUpdateTagValMsg(SVnodeObj *pVnode, void *pCont, SRspRet *); static int32_t vnodePerformFlowCtrl(SVWriteMsg *pWrite); -void vnodeInitWriteFp(void) { +int32_t vnodeInitWrite(void) { vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessSubmitMsg; vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessCreateTableMsg; vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessDropTableMsg; vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessAlterTableMsg; vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessDropStableMsg; vnodeProcessWriteMsgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessUpdateTagValMsg; + + return 0; } +void vnodeCleanupWrite() {} + int32_t vnodeProcessWrite(void *vparam, void *wparam, int32_t qtype, void *rparam) { int32_t code = 0; SVnodeObj *pVnode = vparam; @@ -68,7 +64,7 @@ int32_t vnodeProcessWrite(void *vparam, void *wparam, int32_t qtype, void *rpara taosMsg[pHead->msgType], qtypeStr[qtype], pHead->version, pVnode->version); if (pHead->version == 0) { // from client or CQ - if (pVnode->status != TAOS_VN_STATUS_READY) { + if (!vnodeInReadyStatus(pVnode)) { vDebug("vgId:%d, msg:%s not processed since vstatus:%d, qtype:%s hver:%" PRIu64, pVnode->vgId, taosMsg[pHead->msgType], pVnode->status, qtypeStr[qtype], pHead->version); return TSDB_CODE_APP_NOT_READY; // it may be in deleting or closing state @@ -118,7 +114,7 @@ static int32_t vnodeCheckWrite(void *vparam) { return TSDB_CODE_APP_NOT_READY; } - if (pVnode->status == TAOS_VN_STATUS_CLOSING) { + if (vnodeInClosingStatus(pVnode)) { vDebug("vgId:%d, vnode status is %s, refCount:%d pVnode:%p", pVnode->vgId, vnodeStatus[pVnode->status], pVnode->refCount, pVnode); return TSDB_CODE_APP_NOT_READY; @@ -132,11 +128,6 @@ static int32_t vnodeCheckWrite(void *vparam) { return TSDB_CODE_SUCCESS; } -void vnodeConfirmForward(void *vparam, uint64_t version, int32_t code) { - SVnodeObj *pVnode = vparam; - syncConfirmForward(pVnode->sync, version, code); -} - static int32_t vnodeProcessSubmitMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) { int32_t code = TSDB_CODE_SUCCESS; @@ -282,13 +273,15 @@ static void vnodeFlowCtrlMsgToWQueue(void *param, void *tmrId) { pWrite->processedCount++; if (pWrite->processedCount > 100) { - vError("vgId:%d, msg:%p, failed to process since %s", pVnode->vgId, pWrite, tstrerror(code)); + vError("vgId:%d, msg:%p, failed to process since %s, retry:%d", pVnode->vgId, pWrite, tstrerror(code), + pWrite->processedCount); pWrite->processedCount = 1; dnodeSendRpcVWriteRsp(pWrite->pVnode, pWrite, code); } else { code = vnodePerformFlowCtrl(pWrite); if (code == 0) { - vTrace("vgId:%d, write into vwqueue after flowctrl", pVnode->vgId); + vDebug("vgId:%d, msg:%p, write into vwqueue after flowctrl, retry:%d", pVnode->vgId, pWrite, + pWrite->processedCount); pWrite->processedCount = 0; taosWriteQitem(pVnode->wqueue, pWrite->qtype, pWrite); } @@ -310,7 +303,7 @@ static int32_t vnodePerformFlowCtrl(SVWriteMsg *pWrite) { void *unUsed = NULL; taosTmrReset(vnodeFlowCtrlMsgToWQueue, 100, pWrite, tsDnodeTmr, &unUsed); - vTrace("vgId:%d, msg:%p, app:%p, perform flowctrl, count:%d", pVnode->vgId, pWrite, pWrite->rpcMsg.ahandle, + vTrace("vgId:%d, msg:%p, app:%p, perform flowctrl, retry:%d", pVnode->vgId, pWrite, pWrite->rpcMsg.ahandle, pWrite->processedCount); return TSDB_CODE_VND_ACTION_IN_PROGRESS; } diff --git a/src/wal/inc/walInt.h b/src/wal/inc/walInt.h index b0edabfbd8b0f9e0c71d478994c796f4de755c0a..06748d885f7b0fb798abc1873a1734abbf9f05ab 100644 --- a/src/wal/inc/walInt.h +++ b/src/wal/inc/walInt.h @@ -44,8 +44,8 @@ typedef struct { uint64_t version; int64_t fileId; int64_t rid; + int64_t tfd; int32_t vgId; - int32_t fd; int32_t keep; int32_t level; int32_t fsyncPeriod; diff --git a/src/wal/src/walMgmt.c b/src/wal/src/walMgmt.c index 36c190be3e0f72f96b9075b7fb90db39ee3c5412..72ea239817778db036e1eb05fc67fdcd128cbb6b 100644 --- a/src/wal/src/walMgmt.c +++ b/src/wal/src/walMgmt.c @@ -17,6 +17,7 @@ #include "os.h" #include "taoserror.h" #include "tref.h" +#include "tfile.h" #include "twal.h" #include "walInt.h" @@ -61,7 +62,7 @@ void *walOpen(char *path, SWalCfg *pCfg) { } pWal->vgId = pCfg->vgId; - pWal->fd = -1; + pWal->tfd = -1; pWal->fileId = -1; pWal->level = pCfg->walLevel; pWal->keep = pCfg->keep; @@ -124,7 +125,7 @@ void walClose(void *handle) { SWal *pWal = handle; pthread_mutex_lock(&pWal->mutex); - taosClose(pWal->fd); + tfClose(pWal->tfd); pthread_mutex_unlock(&pWal->mutex); taosRemoveRef(tsWal.refId, pWal->rid); } @@ -143,7 +144,7 @@ static void walFreeObj(void *wal) { SWal *pWal = wal; wDebug("vgId:%d, wal:%p is freed", pWal->vgId, pWal); - taosClose(pWal->fd); + tfClose(pWal->tfd); pthread_mutex_destroy(&pWal->mutex); tfree(pWal); } @@ -172,7 +173,7 @@ static void walFsyncAll() { while (pWal) { if (walNeedFsync(pWal)) { wTrace("vgId:%d, do fsync, level:%d seq:%d rseq:%d", pWal->vgId, pWal->level, pWal->fsyncSeq, tsWal.seq); - int32_t code = fsync(pWal->fd); + int32_t code = tfFsync(pWal->tfd); if (code != 0) { wError("vgId:%d, file:%s, failed to fsync since %s", pWal->vgId, pWal->name, strerror(code)); } diff --git a/src/wal/src/walWrite.c b/src/wal/src/walWrite.c index 36b3dba165cbf399360e95d090278a681b10646d..10e1b4dd614fb881ea8819a16ca2451b20f4393d 100644 --- a/src/wal/src/walWrite.c +++ b/src/wal/src/walWrite.c @@ -18,6 +18,7 @@ #include "os.h" #include "taoserror.h" #include "tchecksum.h" +#include "tfile.h" #include "twal.h" #include "walInt.h" @@ -36,8 +37,8 @@ int32_t walRenew(void *handle) { pthread_mutex_lock(&pWal->mutex); - if (pWal->fd >= 0) { - tclose(pWal->fd); + if (tfValid(pWal->tfd)) { + tfClose(pWal->tfd); wDebug("vgId:%d, file:%s, it is closed", pWal->vgId, pWal->name); } @@ -49,9 +50,9 @@ int32_t walRenew(void *handle) { } snprintf(pWal->name, sizeof(pWal->name), "%s/%s%" PRId64, pWal->path, WAL_PREFIX, pWal->fileId); - pWal->fd = open(pWal->name, O_WRONLY | O_CREAT, S_IRWXU | S_IRWXG | S_IRWXO); + pWal->tfd = tfOpenM(pWal->name, O_WRONLY | O_CREAT, S_IRWXU | S_IRWXG | S_IRWXO); - if (pWal->fd < 0) { + if (!tfValid(pWal->tfd)) { code = TAOS_SYSTEM_ERROR(errno); wError("vgId:%d, file:%s, failed to open since %s", pWal->vgId, pWal->name, strerror(errno)); } else { @@ -67,7 +68,7 @@ void walRemoveOneOldFile(void *handle) { SWal *pWal = handle; if (pWal == NULL) return; if (pWal->keep == TAOS_WAL_KEEP) return; - if (pWal->fd <= 0) return; + if (!tfValid(pWal->tfd)) return; pthread_mutex_lock(&pWal->mutex); @@ -113,7 +114,7 @@ int32_t walWrite(void *handle, SWalHead *pHead) { int32_t code = 0; // no wal - if (pWal->fd <= 0) return 0; + if (!tfValid(pWal->tfd)) return 0; if (pWal->level == TAOS_WAL_NOLOG) return 0; if (pHead->version <= pWal->version) return 0; @@ -123,12 +124,12 @@ int32_t walWrite(void *handle, SWalHead *pHead) { pthread_mutex_lock(&pWal->mutex); - if (taosWrite(pWal->fd, pHead, contLen) != contLen) { + if (tfWrite(pWal->tfd, pHead, contLen) != contLen) { code = TAOS_SYSTEM_ERROR(errno); wError("vgId:%d, file:%s, failed to write since %s", pWal->vgId, pWal->name, strerror(errno)); } else { - wTrace("vgId:%d, write wal, fileId:%" PRId64 " fd:%d hver:%" PRId64 " wver:%" PRIu64 " len:%d", pWal->vgId, - pWal->fileId, pWal->fd, pHead->version, pWal->version, pHead->len); + wTrace("vgId:%d, write wal, fileId:%" PRId64 " tfd:%" PRId64 " hver:%" PRId64 " wver:%" PRIu64 " len:%d", pWal->vgId, + pWal->fileId, pWal->tfd, pHead->version, pWal->version, pHead->len); pWal->version = pHead->version; } @@ -141,11 +142,11 @@ int32_t walWrite(void *handle, SWalHead *pHead) { void walFsync(void *handle, bool forceFsync) { SWal *pWal = handle; - if (pWal == NULL || pWal->fd < 0) return; + if (pWal == NULL || !tfValid(pWal->tfd)) return; if (forceFsync || (pWal->level == TAOS_WAL_FSYNC && pWal->fsyncPeriod == 0)) { wTrace("vgId:%d, fileId:%" PRId64 ", do fsync", pWal->vgId, pWal->fileId); - if (fsync(pWal->fd) < 0) { + if (tfFsync(pWal->tfd) < 0) { wError("vgId:%d, fileId:%" PRId64 ", fsync failed since %s", pWal->vgId, pWal->fileId, strerror(errno)); } } @@ -186,8 +187,8 @@ int32_t walRestore(void *handle, void *pVnode, FWalWrite writeFp) { // open the existing WAL file in append mode pWal->fileId = 0; snprintf(pWal->name, sizeof(pWal->name), "%s/%s%" PRId64, pWal->path, WAL_PREFIX, pWal->fileId); - pWal->fd = open(pWal->name, O_WRONLY | O_CREAT | O_APPEND, S_IRWXU | S_IRWXG | S_IRWXO); - if (pWal->fd < 0) { + pWal->tfd = tfOpenM(pWal->name, O_WRONLY | O_CREAT | O_APPEND, S_IRWXU | S_IRWXG | S_IRWXO); + if (!tfValid(pWal->tfd)) { wError("vgId:%d, file:%s, failed to open since %s", pWal->vgId, pWal->name, strerror(errno)); return TAOS_SYSTEM_ERROR(errno); } @@ -211,28 +212,28 @@ int32_t walGetWalFile(void *handle, char *fileName, int64_t *fileId) { code = (*fileId == pWal->fileId) ? 0 : 1; } - wTrace("vgId:%d, get wal file, code:%d curId:%" PRId64 " outId:%" PRId64, pWal->vgId, code, pWal->fileId, *fileId); + wDebug("vgId:%d, get wal file, code:%d curId:%" PRId64 " outId:%" PRId64, pWal->vgId, code, pWal->fileId, *fileId); pthread_mutex_unlock(&(pWal->mutex)); return code; } -static void walFtruncate(SWal *pWal, int32_t fd, int64_t offset) { - taosFtruncate(fd, offset); - fsync(fd); +static void walFtruncate(SWal *pWal, int64_t tfd, int64_t offset) { + tfFtruncate(tfd, offset); + tfFsync(tfd); } -static int32_t walSkipCorruptedRecord(SWal *pWal, SWalHead *pHead, int32_t fd, int64_t *offset) { +static int32_t walSkipCorruptedRecord(SWal *pWal, SWalHead *pHead, int64_t tfd, int64_t *offset) { int64_t pos = *offset; while (1) { pos++; - if (lseek(fd, pos, SEEK_SET) < 0) { + if (tfLseek(tfd, pos, SEEK_SET) < 0) { wError("vgId:%d, failed to seek from corrupted wal file since %s", pWal->vgId, strerror(errno)); return TSDB_CODE_WAL_FILE_CORRUPTED; } - if (taosRead(fd, pHead, sizeof(SWalHead)) <= 0) { + if (tfRead(tfd, pHead, sizeof(SWalHead)) <= 0) { wError("vgId:%d, read to end of corrupted wal file, offset:%" PRId64, pWal->vgId, pos); return TSDB_CODE_WAL_FILE_CORRUPTED; } @@ -259,8 +260,8 @@ static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, ch return TAOS_SYSTEM_ERROR(errno); } - int32_t fd = open(name, O_RDWR); - if (fd < 0) { + int64_t tfd = tfOpen(name, O_RDWR); + if (!tfValid(tfd)) { wError("vgId:%d, file:%s, failed to open for restore since %s", pWal->vgId, name, strerror(errno)); tfree(buffer); return TAOS_SYSTEM_ERROR(errno); @@ -273,7 +274,7 @@ static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, ch SWalHead *pHead = buffer; while (1) { - int32_t ret = taosRead(fd, pHead, sizeof(SWalHead)); + int32_t ret = tfRead(tfd, pHead, sizeof(SWalHead)); if (ret == 0) break; if (ret < 0) { @@ -284,16 +285,16 @@ static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, ch if (ret < sizeof(SWalHead)) { wError("vgId:%d, file:%s, failed to read wal head, ret is %d", pWal->vgId, name, ret); - walFtruncate(pWal, fd, offset); + walFtruncate(pWal, tfd, offset); break; } if (!taosCheckChecksumWhole((uint8_t *)pHead, sizeof(SWalHead))) { wError("vgId:%d, file:%s, wal head cksum is messed up, hver:%" PRIu64 " len:%d offset:%" PRId64, pWal->vgId, name, pHead->version, pHead->len, offset); - code = walSkipCorruptedRecord(pWal, pHead, fd, &offset); + code = walSkipCorruptedRecord(pWal, pHead, tfd, &offset); if (code != TSDB_CODE_SUCCESS) { - walFtruncate(pWal, fd, offset); + walFtruncate(pWal, tfd, offset); break; } } @@ -310,7 +311,7 @@ static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, ch pHead = buffer; } - ret = taosRead(fd, pHead->cont, pHead->len); + ret = tfRead(tfd, pHead->cont, pHead->len); if (ret < 0) { wError("vgId:%d, file:%s, failed to read wal body since %s", pWal->vgId, name, strerror(errno)); code = TAOS_SYSTEM_ERROR(errno); @@ -325,14 +326,14 @@ static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, ch offset = offset + sizeof(SWalHead) + pHead->len; - wTrace("vgId:%d, restore wal, fileId:%" PRId64 " hver:%" PRIu64 " wver:%" PRIu64 " len:%d", pWal->vgId, + wDebug("vgId:%d, restore wal, fileId:%" PRId64 " hver:%" PRIu64 " wver:%" PRIu64 " len:%d", pWal->vgId, fileId, pHead->version, pWal->version, pHead->len); pWal->version = pHead->version; (*writeFp)(pVnode, pHead, TAOS_QTYPE_WAL, NULL); } - tclose(fd); + tfClose(tfd); tfree(buffer); return code; diff --git a/tests/examples/JDBC/JDBCDemo/pom.xml b/tests/examples/JDBC/JDBCDemo/pom.xml index 121a3b5cd63579b47fb166c7365090f06d05c0da..98f908b77e8d3bfa03eebf8560205b0424720f1f 100644 --- a/tests/examples/JDBC/JDBCDemo/pom.xml +++ b/tests/examples/JDBC/JDBCDemo/pom.xml @@ -56,7 +56,7 @@ com.taosdata.jdbc taos-jdbcdriver - 2.0.8 + 2.0.12 log4j diff --git a/tests/examples/JDBC/JDBCDemo/readme.md b/tests/examples/JDBC/JDBCDemo/readme.md index 9b8790adaddb20246232392dd323ec502102fa18..e348e458fe938c4f2381c448f3c15e60af27040e 100644 --- a/tests/examples/JDBC/JDBCDemo/readme.md +++ b/tests/examples/JDBC/JDBCDemo/readme.md @@ -6,10 +6,24 @@ TDengine's JDBC demo project is organized in a Maven way so that users can easil Make sure you have already installed a tdengine client on your current develop environment. Download the tdengine package on our website: ``https://www.taosdata.com/cn/all-downloads/`` and install the client. +## How to run jdbcChecker +
mvn clean compile exec:java -Dexec.mainClass="com.taosdata.example.JdbcChecker" -Dexec.args="-host localhost"
+ +## How to run jdbcTaosDemo +run command: +
 mvn clean compile exec:java -Dexec.mainClass="com.taosdata.example.jdbcTaosdemo.JdbcTaosdemo"
+and run with your customed args +
mvn clean compile exec:java -Dexec.mainClass="com.taosdata.example.jdbcTaosdemo.JdbcTaosdemo" -Dexec.args="-host localhost"
+ ## Compile the Demo Code and Run It + To compile the demo project, go to the source directory ``TDengine/tests/examples/JDBC/JDBCDemo`` and execute -
mvn clean package assembly:single
+ +
+mvn clean package assembly:single
+
+ The ``pom.xml`` is configured to package all the dependencies into one executable jar file. To run it, go to ``examples/JDBC/JDBCDemo/target`` and execute -
java -jar jdbcChecker-SNAPSHOT-jar-with-dependencies.jar -host localhost
\ No newline at end of file +
java -jar jdbcChecker-SNAPSHOT-jar-with-dependencies.jar -host localhost
diff --git a/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/domain/JdbcTaosdemoConfig.java b/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/domain/JdbcTaosdemoConfig.java index 82613037dbccd3be1f2c8a85a2f25e7a25ffad01..36745a93941cc690f37d06d9a3662605723bbd2c 100644 --- a/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/domain/JdbcTaosdemoConfig.java +++ b/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/domain/JdbcTaosdemoConfig.java @@ -14,9 +14,9 @@ public final class JdbcTaosdemoConfig { //Destination database. Default is 'test' private String dbName = "test"; //keep - private int keep = 3650; + private int keep = 36500; //days - private int days = 10; + private int days = 120; //Super table Name. Default is 'meters' private String stbName = "meters"; diff --git a/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/task/InsertTableTask.java b/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/task/InsertTableTask.java index a35628bb58c6630d92bd2b6aebb09f9912e57536..644de52dd3e75a77c2d635a6f5328f186259096c 100644 --- a/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/task/InsertTableTask.java +++ b/tests/examples/JDBC/JDBCDemo/src/main/java/com/taosdata/example/jdbcTaosdemo/task/InsertTableTask.java @@ -41,7 +41,7 @@ public class InsertTableTask implements Runnable { long ts = start.toEpochMilli() + (j * timeGap); // insert data into echo table for (int i = startTbIndex; i < startTbIndex + tableNumber; i++) { - String sql = SqlSpeller.insertOneRowSQL(config.getDbName(), config.getTbPrefix(), i + 1, ts); + String sql = SqlSpeller.insertBatchSizeRowsSQL(config.getDbName(), config.getTbPrefix(), i + 1, ts, config.getNumberOfRecordsPerRequest()); logger.info(Thread.currentThread().getName() + ">>> " + sql); Statement statement = connection.createStatement(); statement.execute(sql); diff --git a/tests/examples/c/apitest.c b/tests/examples/c/apitest.c index be60a88ad70721bd6281d9ca8f1d73263a788532..e08d667d6be42bef3dc4e058fab10ac68f0809e0 100644 --- a/tests/examples/c/apitest.c +++ b/tests/examples/c/apitest.c @@ -79,11 +79,11 @@ static int print_result(TAOS_RES* res, int blockFetch) { if (blockFetch) { int rows = 0; while ((rows = taos_fetch_block(res, &row))) { - for (int i = 0; i < rows; i++) { - char temp[256]; - taos_print_row(temp, row + i, fields, num_fields); - puts(temp); - } + //for (int i = 0; i < rows; i++) { + // char temp[256]; + // taos_print_row(temp, row + i, fields, num_fields); + // puts(temp); + //} nRows += rows; } } else { @@ -498,4 +498,4 @@ int main(int argc, char *argv[]) { taos_close(taos); taos_cleanup(); -} \ No newline at end of file +} diff --git a/tests/examples/c/subscribe.c b/tests/examples/c/subscribe.c index 8368af18f7831d9c2cbb7402104b1c59042eda72..cdd8ddaf7f6c4d2e5088ef36cc00ad77a0c8ebc9 100644 --- a/tests/examples/c/subscribe.c +++ b/tests/examples/c/subscribe.c @@ -19,10 +19,10 @@ void print_result(TAOS_RES* res, int blockFetch) { if (blockFetch) { nRows = taos_fetch_block(res, &row); - for (int i = 0; i < nRows; i++) { - taos_print_row(buf, row + i, fields, num_fields); - puts(buf); - } + //for (int i = 0; i < nRows; i++) { + // taos_print_row(buf, row + i, fields, num_fields); + // puts(buf); + //} } else { while ((row = taos_fetch_row(res))) { taos_print_row(buf, row, fields, num_fields); diff --git a/tests/pytest/client/alterDatabase.py b/tests/pytest/client/alterDatabase.py index fa397d16c57625d53f6f1b2772afa99ade314e1a..8191312cc060f9c20632911c956065f777d47079 100644 --- a/tests/pytest/client/alterDatabase.py +++ b/tests/pytest/client/alterDatabase.py @@ -32,9 +32,9 @@ class TDTestCase: tdSql.query("show databases") tdSql.checkData(0, 14, 2) - tdSql.execute("alter database db keep 365") + tdSql.execute("alter database db keep 365,365,365") tdSql.query("show databases") - tdSql.checkData(0, 7, "3650,3650,365") + tdSql.checkData(0, 7, "365,365,365") tdSql.execute("alter database db quorum 2") tdSql.query("show databases") diff --git a/tests/pytest/client/noConnectionErrorTest.py b/tests/pytest/client/noConnectionErrorTest.py new file mode 100644 index 0000000000000000000000000000000000000000..2c13016cf19506a6cb76bef1a294e5ec1a9c2b84 --- /dev/null +++ b/tests/pytest/client/noConnectionErrorTest.py @@ -0,0 +1,48 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +from util.log import * +from util.cases import * +from util.sql import * +from util.dnodes import tdDnodes + + +class TDTestCase: + def init(self, conn, logSql): + tdLog.debug("start to execute %s" % __file__) + tdSql.init(conn.cursor(), logSql) + + def run(self): + tdSql.prepare() + + tdDnodes.stop(1) + sql = "use db" + + try: + tdSql.execute(sql) + except Exception as e: + expectError = 'Unable to establish connection' + if expectError in str(e): + pass + else: + caller = inspect.getframeinfo(inspect.stack()[1][1]) + tdLog.exit("%s(%d) failed: sql:%s, expect error not occured" % (caller.filename, caller.lineno, sql)) + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) diff --git a/tests/pytest/crash_gen/crash_gen_main.py b/tests/pytest/crash_gen/crash_gen_main.py index 8d68457ec8d07c9c00f8b8fd0a11e2f25284ce4a..e2ce4b26fa2fa8cc9c601218c1bf19d838b57c46 100755 --- a/tests/pytest/crash_gen/crash_gen_main.py +++ b/tests/pytest/crash_gen/crash_gen_main.py @@ -352,6 +352,12 @@ class ThreadCoordinator: self._execStats.registerFailure("Broken DB Connection") # continue # don't do that, need to tap all threads at # end, and maybe signal them to stop + if isinstance(err, CrashGenError): # our own transition failure + Logging.info("State transition error") + traceback.print_stack() + transitionFailed = True + self._te = None # Not running any more + self._execStats.registerFailure("State transition error") else: raise # return transitionFailed # Why did we have this??!! @@ -388,12 +394,20 @@ class ThreadCoordinator: self._syncAtBarrier() # For now just cross the barrier Progress.emit(Progress.END_THREAD_STEP) except threading.BrokenBarrierError as err: - Logging.info("Main loop aborted, caused by worker thread(s) time-out") self._execStats.registerFailure("Aborted due to worker thread timeout") - print("\n\nWorker Thread time-out detected, TAOS related threads are:") + Logging.error("\n") + Logging.error("Main loop aborted, caused by worker thread(s) time-out of {} seconds".format( + ThreadCoordinator.WORKER_THREAD_TIMEOUT)) + Logging.error("TAOS related threads blocked at (stack frames top-to-bottom):") ts = ThreadStacks() ts.print(filterInternal=True) workerTimeout = True + + # Enable below for deadlock debugging, using gdb to attach to process + # while True: + # Logging.error("Deadlock detected") + # time.sleep(60.0) + break # At this point, all threads should be pass the overall "barrier" and before the per-thread "gate" @@ -701,7 +715,7 @@ class AnyState: # task.logDebug("Task success found") sCnt += 1 if (sCnt >= 2): - raise RuntimeError( + raise CrashGenError( "Unexpected more than 1 success with task: {}".format(cls)) def assertIfExistThenSuccess(self, tasks, cls): @@ -714,7 +728,7 @@ class AnyState: if task.isSuccess(): sCnt += 1 if (exists and sCnt <= 0): - raise RuntimeError("Unexpected zero success for task type: {}, from tasks: {}" + raise CrashGenError("Unexpected zero success for task type: {}, from tasks: {}" .format(cls, tasks)) def assertNoTask(self, tasks, cls): @@ -727,7 +741,7 @@ class AnyState: for task in tasks: if isinstance(task, cls): if task.isSuccess(): - raise RuntimeError( + raise CrashGenError( "Unexpected successful task: {}".format(cls)) def hasSuccess(self, tasks, cls): @@ -926,8 +940,9 @@ class StateMechine: Logging.debug("[STT] DB_ONLY found, between {} and {}".format(ts, time.time())) return StateDbOnly() + # For sure we have tables, which means we must have the super table. # TODO: are we sure? sTable = self._db.getFixedSuperTable() - if sTable.hasRegTables(dbc, dbName): # no regular tables + if sTable.hasRegTables(dbc): # no regular tables Logging.debug("[STT] SUPER_TABLE_ONLY found, between {} and {}".format(ts, time.time())) return StateSuperTableOnly() else: # has actual tables @@ -1050,9 +1065,8 @@ class Database: def getFixedSuperTableName(cls): return "fs_table" - @classmethod - def getFixedSuperTable(cls) -> TdSuperTable: - return TdSuperTable(cls.getFixedSuperTableName()) + def getFixedSuperTable(self) -> TdSuperTable: + return TdSuperTable(self.getFixedSuperTableName(), self.getName()) # We aim to create a starting time tick, such that, whenever we run our test here once # We should be able to safely create 100,000 records, which will not have any repeated time stamp @@ -1107,6 +1121,11 @@ class Database: # print("Float obtained: {}".format(ret)) return ret + ALL_COLORS = ['red', 'white', 'blue', 'green', 'purple'] + + def getNextColor(self): + return random.choice(self.ALL_COLORS) + class TaskExecutor(): class BoundedList: @@ -1240,7 +1259,7 @@ class Task(): if errno in [ 0x05, # TSDB_CODE_RPC_NOT_READY 0x0B, # Unable to establish connection, more details in TD-1648 - 0x200, # invalid SQL, TODO: re-examine with TD-934 + # 0x200, # invalid SQL, TODO: re-examine with TD-934 0x20F, # query terminated, possibly due to vnoding being dropped, see TD-1776 0x213, # "Disconnected from service", result of "kill connection ???" 0x217, # "db not selected", client side defined error code @@ -1569,8 +1588,8 @@ class TaskCreateSuperTable(StateTransitionTask): sTable = self._db.getFixedSuperTable() # type: TdSuperTable # wt.execSql("use db") # should always be in place - sTable.create(wt.getDbConn(), self._db.getName(), - {'ts':'timestamp', 'speed':'int'}, {'b':'binary(200)', 'f':'float'}, + sTable.create(wt.getDbConn(), + {'ts':'TIMESTAMP', 'speed':'INT', 'color':'BINARY(16)'}, {'b':'BINARY(200)', 'f':'FLOAT'}, dropIfExists = True ) # self.execWtSql(wt,"create table db.{} (ts timestamp, speed int) tags (b binary(200), f float) ".format(tblName)) @@ -1579,30 +1598,33 @@ class TaskCreateSuperTable(StateTransitionTask): class TdSuperTable: - def __init__(self, stName): + def __init__(self, stName, dbName): self._stName = stName + self._dbName = dbName def getName(self): return self._stName - def drop(self, dbc, dbName, skipCheck = False): - if self.exists(dbc, dbName) : # if myself exists + def drop(self, dbc, skipCheck = False): + dbName = self._dbName + if self.exists(dbc) : # if myself exists fullTableName = dbName + '.' + self._stName dbc.execute("DROP TABLE {}".format(fullTableName)) else: if not skipCheck: raise CrashGenError("Cannot drop non-existant super table: {}".format(self._stName)) - def exists(self, dbc, dbName): - dbc.execute("USE " + dbName) + def exists(self, dbc): + dbc.execute("USE " + self._dbName) return dbc.existsSuperTable(self._stName) # TODO: odd semantic, create() method is usually static? - def create(self, dbc, dbName, cols: dict, tags: dict, + def create(self, dbc, cols: dict, tags: dict, dropIfExists = False ): - '''Creating a super table''' + + dbName = self._dbName dbc.execute("USE " + dbName) fullTableName = dbName + '.' + self._stName if dbc.existsSuperTable(self._stName): @@ -1623,7 +1645,8 @@ class TdSuperTable: ) dbc.execute(sql) - def getRegTables(self, dbc: DbConn, dbName: str): + def getRegTables(self, dbc: DbConn): + dbName = self._dbName try: dbc.query("select TBNAME from {}.{}".format(dbName, self._stName)) # TODO: analyze result set later except taos.error.ProgrammingError as err: @@ -1634,10 +1657,11 @@ class TdSuperTable: qr = dbc.getQueryResult() return [v[0] for v in qr] # list transformation, ref: https://stackoverflow.com/questions/643823/python-list-transformation - def hasRegTables(self, dbc: DbConn, dbName: str): - return dbc.query("SELECT * FROM {}.{}".format(dbName, self._stName)) > 0 + def hasRegTables(self, dbc: DbConn): + return dbc.query("SELECT * FROM {}.{}".format(self._dbName, self._stName)) > 0 - def ensureTable(self, task: Task, dbc: DbConn, dbName: str, regTableName: str): + def ensureTable(self, task: Task, dbc: DbConn, regTableName: str): + dbName = self._dbName sql = "select tbname from {}.{} where tbname in ('{}')".format(dbName, self._stName, regTableName) if dbc.query(sql) >= 1 : # reg table exists already return @@ -1650,15 +1674,15 @@ class TdSuperTable: # print("(" + fullTableName[-3:] + ")", end="", flush=True) try: sql = "CREATE TABLE {} USING {}.{} tags ({})".format( - fullTableName, dbName, self._stName, self._getTagStrForSql(dbc, dbName) + fullTableName, dbName, self._stName, self._getTagStrForSql(dbc) ) dbc.execute(sql) finally: if task is not None: task.unlockTable(fullTableName) # no matter what - def _getTagStrForSql(self, dbc, dbName: str) : - tags = self._getTags(dbc, dbName) + def _getTagStrForSql(self, dbc) : + tags = self._getTags(dbc) tagStrs = [] for tagName in tags: tagType = tags[tagName] @@ -1672,36 +1696,86 @@ class TdSuperTable: raise RuntimeError("Unexpected tag type: {}".format(tagType)) return ", ".join(tagStrs) - def _getTags(self, dbc, dbName) -> dict: - dbc.query("DESCRIBE {}.{}".format(dbName, self._stName)) + def _getTags(self, dbc) -> dict: + dbc.query("DESCRIBE {}.{}".format(self._dbName, self._stName)) stCols = dbc.getQueryResult() # print(stCols) ret = {row[0]:row[1] for row in stCols if row[3]=='TAG'} # name:type # print("Tags retrieved: {}".format(ret)) return ret - def addTag(self, dbc, dbName, tagName, tagType): - if tagName in self._getTags(dbc, dbName): # already + def addTag(self, dbc, tagName, tagType): + if tagName in self._getTags(dbc): # already return # sTable.addTag("extraTag", "int") - sql = "alter table {}.{} add tag {} {}".format(dbName, self._stName, tagName, tagType) + sql = "alter table {}.{} add tag {} {}".format( + self._dbName, self._stName, tagName, tagType) dbc.execute(sql) - def dropTag(self, dbc, dbName, tagName): - if not tagName in self._getTags(dbc, dbName): # don't have this tag + def dropTag(self, dbc, tagName): + if not tagName in self._getTags(dbc): # don't have this tag return - sql = "alter table {}.{} drop tag {}".format(dbName, self._stName, tagName) + sql = "alter table {}.{} drop tag {}".format(self._dbName, self._stName, tagName) dbc.execute(sql) - def changeTag(self, dbc, dbName, oldTag, newTag): - tags = self._getTags(dbc, dbName) + def changeTag(self, dbc, oldTag, newTag): + tags = self._getTags(dbc) if not oldTag in tags: # don't have this tag return if newTag in tags: # already have this tag return - sql = "alter table {}.{} change tag {} {}".format(dbName, self._stName, oldTag, newTag) + sql = "alter table {}.{} change tag {} {}".format(self._dbName, self._stName, oldTag, newTag) dbc.execute(sql) + def generateQueries(self, dbc: DbConn) -> List[SqlQuery]: + ''' Generate queries to test/exercise this super table ''' + ret = [] # type: List[SqlQuery] + + for rTbName in self.getRegTables(dbc): # regular tables + + filterExpr = Dice.choice([ # TODO: add various kind of WHERE conditions + None + ]) + + # Run the query against the regular table first + doAggr = (Dice.throw(2) == 0) # 1 in 2 chance + if not doAggr: # don't do aggregate query, just simple one + ret.append(SqlQuery( # reg table + "select {} from {}.{}".format('*', self._dbName, rTbName))) + ret.append(SqlQuery( # super table + "select {} from {}.{}".format('*', self._dbName, self.getName()))) + else: # Aggregate query + aggExpr = Dice.choice([ + 'count(*)', + 'avg(speed)', + # 'twa(speed)', # TODO: this one REQUIRES a where statement, not reasonable + 'sum(speed)', + 'stddev(speed)', + # SELECTOR functions + 'min(speed)', + 'max(speed)', + 'first(speed)', + 'last(speed)', + 'top(speed, 50)', # TODO: not supported? + 'bottom(speed, 50)', # TODO: not supported? + 'apercentile(speed, 10)', # TODO: TD-1316 + 'last_row(speed)', + # Transformation Functions + # 'diff(speed)', # TODO: no supported?! + 'spread(speed)' + ]) # TODO: add more from 'top' + + + if aggExpr not in ['stddev(speed)']: #TODO: STDDEV not valid for super tables?! + sql = "select {} from {}.{}".format(aggExpr, self._dbName, self.getName()) + if Dice.throw(3) == 0: # 1 in X chance + sql = sql + ' GROUP BY color' + Progress.emit(Progress.QUERY_GROUP_BY) + # Logging.info("Executing GROUP-BY query: " + sql) + ret.append(SqlQuery(sql)) + + return ret + class TaskReadData(StateTransitionTask): @classmethod def getEndState(cls): @@ -1716,10 +1790,8 @@ class TaskReadData(StateTransitionTask): # return True # always # return gSvcMgr.isActive() # only if it's running TODO: race condition here - def _executeInternal(self, te: TaskExecutor, wt: WorkerThread): - sTable = self._db.getFixedSuperTable() - - # 1 in 5 chance, simulate a broken connection, only if service stable (not restarting) + def _reconnectIfNeeded(self, wt): + # 1 in 20 chance, simulate a broken connection, only if service stable (not restarting) if random.randrange(20)==0: # and self._canRestartService(): # TODO: break connection in all situations # Logging.info("Attempting to reconnect to server") # TODO: change to DEBUG Progress.emit(Progress.SERVICE_RECONNECT_START) @@ -1744,43 +1816,36 @@ class TaskReadData(StateTransitionTask): return # TODO: fix server restart status race condtion + def _executeInternal(self, te: TaskExecutor, wt: WorkerThread): + self._reconnectIfNeeded(wt) + dbc = wt.getDbConn() - dbName = self._db.getName() - for rTbName in sTable.getRegTables(dbc, dbName): # regular tables - aggExpr = Dice.choice([ - '*', - 'count(*)', - 'avg(speed)', - # 'twa(speed)', # TODO: this one REQUIRES a where statement, not reasonable - 'sum(speed)', - 'stddev(speed)', - # SELECTOR functions - 'min(speed)', - 'max(speed)', - 'first(speed)', - 'last(speed)', - 'top(speed, 50)', # TODO: not supported? - 'bottom(speed, 50)', # TODO: not supported? - 'apercentile(speed, 10)', # TODO: TD-1316 - 'last_row(speed)', - # Transformation Functions - # 'diff(speed)', # TODO: no supported?! - 'spread(speed)' - ]) # TODO: add more from 'top' - filterExpr = Dice.choice([ # TODO: add various kind of WHERE conditions - None - ]) + sTable = self._db.getFixedSuperTable() + + for q in sTable.generateQueries(dbc): # regular tables try: - # Run the query against the regular table first - dbc.execute("select {} from {}.{}".format(aggExpr, dbName, rTbName)) - # Then run it against the super table - if aggExpr not in ['stddev(speed)']: #TODO: STDDEV not valid for super tables?! - dbc.execute("select {} from {}.{}".format(aggExpr, dbName, sTable.getName())) + sql = q.getSql() + # if 'GROUP BY' in sql: + # Logging.info("Executing GROUP-BY query: " + sql) + dbc.execute(sql) except taos.error.ProgrammingError as err: errno2 = Helper.convertErrno(err.errno) Logging.debug("[=] Read Failure: errno=0x{:X}, msg: {}, SQL: {}".format(errno2, err, dbc.getLastSql())) raise +class SqlQuery: + @classmethod + def buildRandom(cls, db: Database): + '''Build a random query against a certain database''' + + dbName = db.getName() + + def __init__(self, sql:str = None): + self._sql = sql + + def getSql(self): + return self._sql + class TaskDropSuperTable(StateTransitionTask): @classmethod def getEndState(cls): @@ -1837,19 +1902,18 @@ class TaskAlterTags(StateTransitionTask): # tblName = self._dbManager.getFixedSuperTableName() dbc = wt.getDbConn() sTable = self._db.getFixedSuperTable() - dbName = self._db.getName() dice = Dice.throw(4) if dice == 0: - sTable.addTag(dbc, dbName, "extraTag", "int") + sTable.addTag(dbc, "extraTag", "int") # sql = "alter table db.{} add tag extraTag int".format(tblName) elif dice == 1: - sTable.dropTag(dbc, dbName, "extraTag") + sTable.dropTag(dbc, "extraTag") # sql = "alter table db.{} drop tag extraTag".format(tblName) elif dice == 2: - sTable.dropTag(dbc, dbName, "newTag") + sTable.dropTag(dbc, "newTag") # sql = "alter table db.{} drop tag newTag".format(tblName) else: # dice == 3 - sTable.changeTag(dbc, dbName, "extraTag", "newTag") + sTable.changeTag(dbc, "extraTag", "newTag") # sql = "alter table db.{} change tag extraTag newTag".format(tblName) class TaskRestartService(StateTransitionTask): @@ -1920,15 +1984,17 @@ class TaskAddData(StateTransitionTask): for j in range(numRecords): # number of records per table nextInt = db.getNextInt() nextTick = db.getNextTick() - sql += "('{}', {});".format(nextTick, nextInt) + nextColor = db.getNextColor() + sql += "('{}', {}, '{}');".format(nextTick, nextInt, nextColor) dbc.execute(sql) - def _addData(self, db, dbc, regTableName, te: TaskExecutor): # implied: NOT in batches + def _addData(self, db: Database, dbc, regTableName, te: TaskExecutor): # implied: NOT in batches numRecords = self.LARGE_NUMBER_OF_RECORDS if gConfig.larger_data else self.SMALL_NUMBER_OF_RECORDS for j in range(numRecords): # number of records per table nextInt = db.getNextInt() nextTick = db.getNextTick() + nextColor = db.getNextColor() if gConfig.record_ops: self.prepToRecordOps() self.fAddLogReady.write("Ready to write {} to {}\n".format(nextInt, regTableName)) @@ -1942,11 +2008,11 @@ class TaskAddData(StateTransitionTask): # print("_w" + str(nextInt % 100), end="", flush=True) # Trace what was written try: - sql = "insert into {} values ('{}', {});".format( # removed: tags ('{}', {}) + sql = "insert into {} values ('{}', {}, '{}');".format( # removed: tags ('{}', {}) fullTableName, # ds.getFixedSuperTableName(), # ds.getNextBinary(), ds.getNextFloat(), - nextTick, nextInt) + nextTick, nextInt, nextColor) dbc.execute(sql) except: # Any exception at all if gConfig.verify_data: @@ -1964,10 +2030,10 @@ class TaskAddData(StateTransitionTask): .format(nextInt, readBack), 0x999) except taos.error.ProgrammingError as err: errno = Helper.convertErrno(err.errno) - if errno in [0x991, 0x992] : # not a single result + if errno in [CrashGenError.INVALID_EMPTY_RESULT, CrashGenError.INVALID_MULTIPLE_RESULT] : # not a single result raise taos.error.ProgrammingError( "Failed to read back same data for tick: {}, wrote: {}, read: {}" - .format(nextTick, nextInt, "Empty Result" if errno==0x991 else "Multiple Result"), + .format(nextTick, nextInt, "Empty Result" if errno == CrashGenError.INVALID_EMPTY_RESULT else "Multiple Result"), errno) elif errno in [0x218, 0x362]: # table doesn't exist # do nothing @@ -2000,11 +2066,12 @@ class TaskAddData(StateTransitionTask): else: self.activeTable.add(i) # marking it active + dbName = db.getName() sTable = db.getFixedSuperTable() regTableName = self.getRegTableName(i) # "db.reg_table_{}".format(i) - fullTableName = db.getName() + '.' + regTableName + fullTableName = dbName + '.' + regTableName # self._lockTable(fullTableName) # "create table" below. Stop it if the table is "locked" - sTable.ensureTable(self, wt.getDbConn(), db.getName(), regTableName) # Ensure the table exists + sTable.ensureTable(self, wt.getDbConn(), regTableName) # Ensure the table exists # self._unlockTable(fullTableName) if Dice.throw(1) == 0: # 1 in 2 chance @@ -2024,7 +2091,7 @@ class ThreadStacks: # stack info for all threads self._allStacks[th.native_id] = stack def print(self, filteredEndName = None, filterInternal = False): - for thNid, stack in self._allStacks.items(): # for each thread + for thNid, stack in self._allStacks.items(): # for each thread, stack frames top to bottom lastFrame = stack[-1] if filteredEndName: # we need to filter out stacks that match this name if lastFrame.name == filteredEndName : # end did not match @@ -2036,9 +2103,9 @@ class ThreadStacks: # stack info for all threads '__init__']: # the thread that extracted the stack continue # ignore # Now print - print("\n<----- Thread Info for LWP/ID: {} (Execution stopped at Bottom Frame) <-----".format(thNid)) + print("\n<----- Thread Info for LWP/ID: {} (most recent call last) <-----".format(thNid)) stackFrame = 0 - for frame in stack: + for frame in stack: # was using: reversed(stack) # print(frame) print("[{sf}] File {filename}, line {lineno}, in {name}".format( sf=stackFrame, filename=frame.filename, lineno=frame.lineno, name=frame.name)) diff --git a/tests/pytest/crash_gen/db.py b/tests/pytest/crash_gen/db.py index 2a4b362f82c4516195becf78ef9771b7c62c3c41..855e18be55fd5ae0d88865644a2ebfc836b7ccf3 100644 --- a/tests/pytest/crash_gen/db.py +++ b/tests/pytest/crash_gen/db.py @@ -78,7 +78,7 @@ class DbConn: if nRows != 1: raise taos.error.ProgrammingError( "Unexpected result for query: {}, rows = {}".format(sql, nRows), - (0x991 if nRows==0 else 0x992) + (CrashGenError.INVALID_EMPTY_RESULT if nRows==0 else CrashGenError.INVALID_MULTIPLE_RESULT) ) if self.getResultRows() != 1 or self.getResultCols() != 1: raise RuntimeError("Unexpected result set for query: {}".format(sql)) @@ -349,7 +349,8 @@ class DbConnNative(DbConn): def execute(self, sql): if (not self.isOpen): - raise RuntimeError("Cannot execute database commands until connection is open") + raise CrashGenError( + "Cannot exec SQL unless db connection is open", CrashGenError.DB_CONNECTION_NOT_OPEN) Logging.debug("[SQL] Executing SQL: {}".format(sql)) self._lastSql = sql nRows = self._tdSql.execute(sql) @@ -360,8 +361,8 @@ class DbConnNative(DbConn): def query(self, sql): # return rows affected if (not self.isOpen): - raise RuntimeError( - "Cannot query database until connection is open") + raise CrashGenError( + "Cannot query database until connection is open, restarting?", CrashGenError.DB_CONNECTION_NOT_OPEN) Logging.debug("[SQL] Executing SQL: {}".format(sql)) self._lastSql = sql nRows = self._tdSql.query(sql) diff --git a/tests/pytest/crash_gen/misc.py b/tests/pytest/crash_gen/misc.py index 2d2ce99d95582809a8940a8d777bc166b633747c..a374ed943b41b52c2ccbee0825bca3080ed43ea9 100644 --- a/tests/pytest/crash_gen/misc.py +++ b/tests/pytest/crash_gen/misc.py @@ -3,14 +3,20 @@ import random import logging import os +import taos -class CrashGenError(Exception): - def __init__(self, msg=None, errno=None): - self.msg = msg - self.errno = errno - def __str__(self): - return self.msg +class CrashGenError(taos.error.ProgrammingError): + INVALID_EMPTY_RESULT = 0x991 + INVALID_MULTIPLE_RESULT = 0x992 + DB_CONNECTION_NOT_OPEN = 0x993 + # def __init__(self, msg=None, errno=None): + # self.msg = msg + # self.errno = errno + + # def __str__(self): + # return self.msg + pass class LoggingFilter(logging.Filter): @@ -168,6 +174,7 @@ class Progress: SERVICE_RECONNECT_FAILURE = 6 SERVICE_START_NAP = 7 CREATE_TABLE_ATTEMPT = 8 + QUERY_GROUP_BY = 9 tokens = { STEP_BOUNDARY: '.', @@ -178,7 +185,8 @@ class Progress: SERVICE_RECONNECT_SUCCESS: '.r>', SERVICE_RECONNECT_FAILURE: '.xr>', SERVICE_START_NAP: '_zz', - CREATE_TABLE_ATTEMPT: '_c', + CREATE_TABLE_ATTEMPT: 'c', + QUERY_GROUP_BY: 'g', } @classmethod diff --git a/tests/pytest/crash_gen/service_manager.py b/tests/pytest/crash_gen/service_manager.py index d249abc4396462b6dbacbfcbb1f6619e48161c3b..ae6f8d5d3a5a4d6a8bf745a48ec2368aa5e832ad 100644 --- a/tests/pytest/crash_gen/service_manager.py +++ b/tests/pytest/crash_gen/service_manager.py @@ -51,10 +51,12 @@ class TdeInstance(): def prepareGcovEnv(cls, env): # Ref: https://gcc.gnu.org/onlinedocs/gcc/Cross-profiling.html bPath = cls._getBuildPath() # build PATH - numSegments = len(bPath.split('/')) - 1 # "/x/TDengine/build" should yield 3 - numSegments = numSegments - 1 # DEBUG only - env['GCOV_PREFIX'] = bPath + '/svc_gcov' + numSegments = len(bPath.split('/')) # "/x/TDengine/build" should yield 3 + # numSegments += 2 # cover "/src" after build + # numSegments = numSegments - 1 # DEBUG only + env['GCOV_PREFIX'] = bPath + '/src_s' # Server side source env['GCOV_PREFIX_STRIP'] = str(numSegments) # Strip every element, plus, ENV needs strings + # VERY VERY important note: GCOV data collection NOT effective upon SIG_KILL Logging.info("Preparing GCOV environement to strip {} elements and use path: {}".format( numSegments, env['GCOV_PREFIX'] )) @@ -258,14 +260,15 @@ class TdeSubProcess: TdeInstance.prepareGcovEnv(myEnv) # print(myEnv) - # print(myEnv.items()) + # print("Starting TDengine with env: ", myEnv.items()) # print("Starting TDengine via Shell: {}".format(cmdLineStr)) useShell = True self.subProcess = subprocess.Popen( - ' '.join(cmdLine) if useShell else cmdLine, - shell=useShell, - # svcCmdSingle, shell=True, # capture core dump? + # ' '.join(cmdLine) if useShell else cmdLine, + # shell=useShell, + ' '.join(cmdLine), + shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, # bufsize=1, # not supported in binary mode @@ -273,7 +276,8 @@ class TdeSubProcess: env=myEnv ) # had text=True, which interferred with reading EOF - STOP_SIGNAL = signal.SIGKILL # What signal to use (in kill) to stop a taosd process? + STOP_SIGNAL = signal.SIGKILL # signal.SIGKILL/SIGINT # What signal to use (in kill) to stop a taosd process? + SIG_KILL_RETCODE = 137 # ref: https://stackoverflow.com/questions/43268156/process-finished-with-exit-code-137-in-pycharm def stop(self): """ @@ -320,8 +324,12 @@ class TdeSubProcess: retCode = self.subProcess.returncode # should always be there # May throw subprocess.TimeoutExpired exception above, therefore # The process is guranteed to have ended by now - self.subProcess = None - if retCode != 0: # != (- signal.SIGINT): + self.subProcess = None + if retCode == self.SIG_KILL_RETCODE: + Logging.info("TSP.stop(): sub proc KILLED, as expected") + elif retCode == (- self.STOP_SIGNAL): + Logging.info("TSP.stop(), sub process STOPPED, as expected") + elif retCode != 0: # != (- signal.SIGINT): Logging.error("TSP.stop(): Failed to stop sub proc properly w/ SIG {}, retCode={}".format( self.STOP_SIGNAL, retCode)) else: diff --git a/tests/pytest/fulltest.sh b/tests/pytest/fulltest.sh index 0e3b482e3da2c6618edd64c5dd004e8812078d99..dc2c0099b4a1477755543b16312708beb98ade18 100755 --- a/tests/pytest/fulltest.sh +++ b/tests/pytest/fulltest.sh @@ -19,6 +19,7 @@ python3 ./test.py -f insert/randomNullCommit.py python3 insert/retentionpolicy.py python3 ./test.py -f insert/alterTableAndInsert.py python3 ./test.py -f insert/insertIntoTwoTables.py +python3 ./test.py -f insert/before_1970.py python3 ./test.py -f table/alter_wal0.py python3 ./test.py -f table/column_name.py @@ -158,7 +159,11 @@ python3 ./test.py -f query/bug1471.py python3 ./test.py -f query/bug1874.py python3 ./test.py -f query/bug1875.py python3 ./test.py -f query/bug1876.py -python3 ./test.py -f query/bug2218.py +python3 ./test.py -f query/bug2218.py +python3 ./test.py -f query/bug2117.py +python3 ./test.py -f query/bug2143.py +python3 ./test.py -f query/sliding.py +python3 ./test.py -f query/unionAllTest.py #stream python3 ./test.py -f stream/metric_1.py @@ -175,6 +180,7 @@ python3 ./test.py -f alter/alter_table_crash.py python3 ./test.py -f client/client.py python3 ./test.py -f client/version.py python3 ./test.py -f client/alterDatabase.py +python3 ./test.py -f client/noConnectionErrorTest.py # Misc python3 testCompress.py @@ -209,6 +215,7 @@ python3 test.py -f query/queryFillTest.py python3 test.py -f tools/taosdemoTest.py python3 test.py -f tools/taosdumpTest.py python3 test.py -f tools/lowaTest.py +python3 test.py -f tools/taosdemoTest2.py # subscribe python3 test.py -f subscribe/singlemeter.py @@ -228,6 +235,7 @@ python3 ./test.py -f update/merge_commit_data2.py python3 ./test.py -f update/merge_commit_data2_update0.py python3 ./test.py -f update/merge_commit_last-0.py python3 ./test.py -f update/merge_commit_last.py +python3 ./test.py -f update/bug_td2279.py # wal python3 ./test.py -f wal/addOldWalTest.py \ No newline at end of file diff --git a/tests/pytest/hivemq-extension-test.py b/tests/pytest/hivemq-extension-test.py new file mode 100644 index 0000000000000000000000000000000000000000..3d0b1ef83fe5841826ad0ab80384b5320662e104 --- /dev/null +++ b/tests/pytest/hivemq-extension-test.py @@ -0,0 +1,249 @@ +#!/usr/bin/python3 +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### +# install pip +# pip install src/connector/python/linux/python2/ +import sys +import os +import os.path +import time +import glob +import getopt +import subprocess +from shutil import which +from multipledispatch import dispatch + + +@dispatch(str, str) +def v_print(msg: str, arg: str): + if verbose: + print(msg % arg) + + +@dispatch(str, int) +def v_print(msg: str, arg: int): + if verbose: + print(msg % int(arg)) + + +@dispatch(str, int, int) +def v_print(msg: str, arg1: int, arg2: int): + if verbose: + print(msg % (int(arg1), int(arg2))) + + +@dispatch(str, int, int, int) +def v_print(msg: str, arg1: int, arg2: int, arg3: int): + if verbose: + print(msg % (int(arg1), int(arg2), int(arg3))) + + +@dispatch(str, int, int, int, int) +def v_print(msg: str, arg1: int, arg2: int, arg3: int, arg4: int): + if verbose: + print(msg % (int(arg1), int(arg2), int(arg3), int(arg4))) + + +def isHiveMQInstalled(): + v_print("%s", "Check if HiveMQ installed") + defaultHiveMQPath = "/opt/hivemq*" + hiveMQDir = glob.glob(defaultHiveMQPath) + if (len(hiveMQDir) == 0): + v_print("%s", "ERROR: HiveMQ NOT found") + return False + else: + v_print("HiveMQ installed at %s", hiveMQDir[0]) + return True + + +def isMosquittoInstalled(): + v_print("%s", "Check if mosquitto installed") + if not which('mosquitto_pub'): + v_print("%s", "ERROR: mosquitto is NOT installed") + return False + else: + return True + + +def installExtension(): + currentDir = os.getcwd() + extDir = 'src/connector/hivemq-tdengine-extension' + os.chdir('../..') + os.system('git submodule update --init -- %s' % extDir) + os.chdir(extDir) + v_print("%s", "build extension..") + os.system('mvn clean package') + + tdExtensionZip = 'target/hivemq-tdengine-extension*.zip' + tdExtensionZipDir = glob.glob(tdExtensionZip) + + defaultHiveMQPath = "/opt/hivemq*" + hiveMQDir = glob.glob(defaultHiveMQPath) + extPath = hiveMQDir[0] + '/extensions' + + tdExtDir = glob.glob(extPath + '/hivemq-tdengine-extension') + if len(tdExtDir): + v_print("%s", "delete exist extension..") + os.system('rm -rf %s' % tdExtDir[0]) + + v_print("%s", "unzip extension..") + os.system('unzip %s -d %s' % (tdExtensionZipDir[0], extPath)) + + os.chdir(currentDir) + + +def stopProgram(prog: str): + psCmd = "ps ax|grep -w %s| grep -v grep | awk '{print $1}'" % prog + + processID = subprocess.check_output( + psCmd, shell=True).decode("utf-8") + + while(processID): + killCmd = "kill -TERM %s > /dev/null 2>&1" % processID + os.system(killCmd) + time.sleep(1) + processID = subprocess.check_output( + psCmd, shell=True).decode("utf-8") + pass + + +def stopHiveMQ(): + stopProgram("hivemq.jar") + v_print("%s", "ERROR: HiveMQ is NOT running") + + +def checkProgramRunning(prog: str): + psCmd = "ps ax|grep -w %s| grep -v grep | awk '{print $1}'" % prog + + processID = subprocess.check_output( + psCmd, shell=True).decode("utf-8") + + if not processID: + v_print("ERROR: %s is NOT running", prog) + return False + else: + return True + + +def runHiveMQ(): + defaultHiveMQPath = "/opt/hivemq*" + hiveMQDir = glob.glob(defaultHiveMQPath) + runPath = hiveMQDir[0] + '/bin/run.sh > /dev/null &' + os.system(runPath) + time.sleep(10) + + if not checkProgramRunning("hivemq.jar"): + return False + else: + v_print("%s", "hivemq is running") + return True + + +def getBuildPath(): + selfPath = os.path.dirname(os.path.realpath(__file__)) + + binPath = '' + + if ("community" in selfPath): + projPath = selfPath[:selfPath.find("community")] + else: + projPath = selfPath[:selfPath.find("tests")] + + for root, dirs, files in os.walk(projPath): + if ("taosd" in files): + rootRealPath = os.path.dirname(os.path.realpath(root)) + if ("packaging" not in rootRealPath): + binPath = root[:len(root) - len("/build/bin")] + break + return binPath + + +def runTDengine(): + stopProgram("taosd") + + buildPath = getBuildPath() + + if (buildPath == ""): + v_print("%s", "ERROR: taosd NOT found!") + sys.exit(1) + else: + v_print("%s", "taosd found in %s" % buildPath) + + binPath = buildPath + "/build/bin/taosd" + + os.system('%s > /dev/null &' % binPath) + time.sleep(10) + if not checkProgramRunning("taosd"): + return False + else: + v_print("%s", "TDengine is running") + return True + + + +def reCreateDatabase(): + buildPath = getBuildPath() + binPath = buildPath + "/build/bin/taos" + + os.system('%s -s "DROP DATABASE IF EXISTS hivemq"' % binPath) + os.system('%s -s "CREATE DATABASE IF NOT EXISTS hivemq"' % binPath) + + +def sendMqttMsg(topic: str, payload: str): + testStr = 'mosquitto_pub -t %s -m "%s"' % (topic, payload) + os.system(testStr) + time.sleep(3) + + +def checkTDengineData(topic: str, payload: str): + buildPath = getBuildPath() + binPath = buildPath + "/build/bin/taos" + + output = subprocess.check_output( + '%s -s "select * from hivemq.mqtt_payload"' % + binPath, shell=True).decode('utf-8') + if (topic in output) and (payload in output): + v_print("%s", output) + return True + else: + v_print("%s", "ERROR: mqtt topic or payload NOT found") + return False + + +if __name__ == "__main__": + verbose = True + testTopic = 'test' + testPayload = 'hello world' + + if not isHiveMQInstalled(): + sys.exit(1) + + if not isMosquittoInstalled(): + sys.exit(1) + + stopHiveMQ() + + installExtension() + + if not runTDengine(): + sys.exit(1) + + reCreateDatabase() + + if not runHiveMQ(): + sys.exit(1) + + sendMqttMsg(testTopic, testPayload) + + if not checkTDengineData(testTopic, testPayload): + sys.exit(1) + + sys.exit(0) diff --git a/tests/pytest/insert/insertDemo.py b/tests/pytest/insert/insertDemo.py new file mode 100644 index 0000000000000000000000000000000000000000..d18206e7a46b960499263b2960a7e3a42ff806ca --- /dev/null +++ b/tests/pytest/insert/insertDemo.py @@ -0,0 +1,47 @@ +import taos +import datetime +import random +import multiprocessing + +def taos_excute(table, connect_host): + conn = taos.connect(host=connect_host, user="root", password="taosdata", config="/etc/taos", database='test') + cursor = conn.cursor() + for i in range(1000000): + pk = random.randint(100001, 300000) + time_now = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S.%f")[:-3] + col1 = random.randint(1, 10000) + col2 = random.randint(1, 10000) + col3 = random.randint(1, 10000) + col4 = random.randint(1, 10000) + col5 = random.randint(1, 10000) + col6 = random.randint(1, 10000) + sql = f"INSERT INTO {table}_{pk} USING {table} TAGS ({pk}) VALUES ('{time_now}', {col1}, {col2}, {col3}, {col4}, {col5}, {col6})" + cursor.execute(sql) + cursor.close() + conn.close() + +def taos_init(table, connect_host, pk): + conn = taos.connect(host=connect_host, user="root", password="taosdata", config="/etc/taos", database='test') + cursor = conn.cursor() + sql = f"CREATE TABLE {table}_{pk} USING {table} TAGS ({pk})" + cursor.execute(sql) + cursor.close() + conn.close() + +print("init time:", datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")) + +connect_list = ["node1", "node2", "node3", "node4", "node5"] +pool = multiprocessing.Pool(processes=108) + +for pk in range(100001, 300000): + pool.apply_async(func=taos_init, args=("test", connect_list[pk % 5], pk, )) + +print("start time:", datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")) + +for i in range(10000): + pool.apply_async(func=taos_excute, args=("test", connect_list[i % 5],)) + +pool.close() +pool.join() + +print("end time:", datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")) \ No newline at end of file diff --git a/tests/pytest/insert/insertFromCSVOurofOrder.py b/tests/pytest/insert/insertFromCSVOurofOrder.py new file mode 100644 index 0000000000000000000000000000000000000000..d4de85b7e93e78ad12962c54fbd1014615dc8e3b --- /dev/null +++ b/tests/pytest/insert/insertFromCSVOurofOrder.py @@ -0,0 +1,71 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +import taos +from util.log import tdLog +from util.cases import tdCases +from util.sql import tdSql +import time +import datetime +import csv +import random +import pandas as pd + + +class TDTestCase: + def init(self, conn, logSql): + tdLog.debug("start to execute %s" % __file__) + tdSql.init(conn.cursor(), logSql) + + self.ts = 1500074556514 + + def writeCSV(self): + with open('test3.csv','w', encoding='utf-8', newline='') as csvFile: + writer = csv.writer(csvFile, dialect='excel') + for i in range(1000000): + newTimestamp = self.ts + random.randint(10000000, 10000000000) + random.randint(1000, 10000000) + random.randint(1, 1000) + d = datetime.datetime.fromtimestamp(newTimestamp / 1000) + dt = str(d.strftime("%Y-%m-%d %H:%M:%S.%f")) + writer.writerow(["'%s'" % dt, random.randint(1, 100), random.uniform(1, 100), random.randint(1, 100), random.randint(1, 100)]) + + def removCSVHeader(self): + data = pd.read_csv("ordered.csv") + data = data.drop([0]) + data.to_csv("ordered.csv", header = False, index = False) + + def run(self): + tdSql.prepare() + + tdSql.execute("create table t1(ts timestamp, c1 int, c2 float, c3 int, c4 int)") + startTime = time.time() + tdSql.execute("insert into t1 file 'outoforder.csv'") + duration = time.time() - startTime + print("Out of Order - Insert time: %d" % duration) + tdSql.query("select count(*) from t1") + rows = tdSql.getData(0, 0) + + tdSql.execute("create table t2(ts timestamp, c1 int, c2 float, c3 int, c4 int)") + startTime = time.time() + tdSql.execute("insert into t2 file 'ordered.csv'") + duration = time.time() - startTime + print("Ordered - Insert time: %d" % duration) + tdSql.query("select count(*) from t2") + tdSql.checkData(0,0, rows) + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) \ No newline at end of file diff --git a/tests/pytest/insert/restfulInsert.py b/tests/pytest/insert/restfulInsert.py index c39d1cbe48f6d08d243d39808c21741806a0ee54..da797f788fcc3cc1ae06843b70da9aaf7230e97c 100644 --- a/tests/pytest/insert/restfulInsert.py +++ b/tests/pytest/insert/restfulInsert.py @@ -18,10 +18,10 @@ import time import argparse class RestfulInsert: - def __init__(self, host, dbname, threads, tables, records, batchSize, tbNamePerfix, outOfOrder): + def __init__(self, host, startTimestamp, dbname, threads, tables, records, batchSize, tbNamePerfix, outOfOrder,tablePerbatch): self.header = {'Authorization': 'Basic cm9vdDp0YW9zZGF0YQ=='} self.url = "http://%s:6041/rest/sql" % host - self.ts = 1500000000000 + self.ts = startTimestamp self.dbname = dbname self.numOfThreads = threads self.numOfTables = tables @@ -29,28 +29,71 @@ class RestfulInsert: self.batchSize = batchSize self.tableNamePerfix = tbNamePerfix self.outOfOrder = outOfOrder + self.tablePerbatch = tablePerbatch def createTable(self, threadID): - tablesPerThread = int (self.numOfTables / self.numOfThreads) - print("create table %d to %d" % (tablesPerThread * threadID, tablesPerThread * (threadID + 1) - 1)) - for i in range(tablesPerThread): + tablesPerThread = int (self.numOfTables / self.numOfThreads) + loop = tablesPerThread if threadID != self.numOfThreads - 1 else self.numOfTables - tablesPerThread * threadID + print("create table %d to %d" % (tablesPerThread * threadID, tablesPerThread * threadID + loop - 1)) + for i in range(loop): tableID = threadID * tablesPerThread + if tableID + i >= self.numOfTables : break name = 'beijing' if tableID % 2 == 0 else 'shanghai' - data = "create table %s.%s%d using %s.meters tags(%d, '%s')" % (self.dbname, self.tableNamePerfix, tableID + i, self.dbname, tableID + i, name) - requests.post(self.url, data, headers = self.header) + data = "create table if not exists %s.%s%d using %s.meters tags(%d, '%s')" % (self.dbname, self.tableNamePerfix, tableID + i, self.dbname, tableID + i, name) + response = requests.post(self.url, data, headers = self.header) + if response.status_code != 200: + print(response.content) + + def insertData(self, threadID): print("thread %d started" % threadID) - tablesPerThread = int (self.numOfTables / self.numOfThreads) - for i in range(tablesPerThread): - tableID = i + threadID * tablesPerThread - start = self.ts - for j in range(int(self.recordsPerTable / self.batchSize)): - data = "insert into %s.%s%d values" % (self.dbname, self.tableNamePerfix, tableID) - values = [] - for k in range(self.batchSize): - data += "(%d, %d, %d, %d)" % (start + j * self.batchSize + k, random.randint(1, 100), random.randint(1, 100), random.randint(1, 100)) - requests.post(self.url, data, headers = self.header) + tablesPerThread = int (self.numOfTables / self.numOfThreads) + loop = int(self.recordsPerTable / self.batchSize) + if self.tablePerbatch == 1 : + for i in range(tablesPerThread+1): + tableID = i + threadID * tablesPerThread + if tableID >= self.numOfTables: return + start = self.ts + start1=time.time() + for k in range(loop): + data = "insert into %s.%s%d values" % (self.dbname, self.tableNamePerfix, tableID) + values = [] + bloop = self.batchSize if k != loop - 1 else self.recordsPerTable - self.batchSize * k + for l in range(bloop): + values.append("(%d, %d, %d, %d)" % (start + k * self.batchSize + l, random.randint(1, 100), random.randint(1, 100), random.randint(1, 100))) + if len(data) > 1048576 : + print ('batch size is larger than 1M') + exit(-1) + if self.outOfOrder : + random.shuffle(values) + data+=''.join(values) + response = requests.post(self.url, data, headers = self.header) + if response.status_code != 200: + print(response.content) + else: + for i in range(0,tablesPerThread+self.tablePerbatch,self.tablePerbatch): + for k in range(loop): + data = "insert into " + for j in range(self.tablePerbatch): + tableID = i + threadID * tablesPerThread+j + if tableID >= self.numOfTables: return + start = self.ts + data += "%s.%s%d values" % (self.dbname, self.tableNamePerfix, tableID) + values = [] + bloop = self.batchSize if k != loop - 1 else self.recordsPerTable - self.batchSize * k + for l in range(bloop): + values.append("(%d, %d, %d, %d)" % (start + k * self.batchSize + l, random.randint(1, 100), random.randint(1, 100), random.randint(1, 100))) + if self.outOfOrder : + random.shuffle(values) + data+=''.join(values) + if len(data) > 1024*1024 : + print ('batch size is larger than 1M') + exit(-1) + response = requests.post(self.url, data, headers = self.header) + if response.status_code != 200: + print(response.content) + def insertUnlimitedData(self, threadID): print("thread %d started" % threadID) @@ -76,15 +119,15 @@ class RestfulInsert: else: random.shuffle(values) for k in range(len(values)): - data += values[k] - requests.post(self.url, data, headers = self.header) + data += values[k] + response = requests.post(self.url, data, headers = self.header) + if response.status_code != 200: + print(response.content) - def run(self): - data = "drop database if exists %s" % self.dbname + def run(self): + data = "create database if not exists %s" % self.dbname requests.post(self.url, data, headers = self.header) - data = "create database %s" % self.dbname - requests.post(self.url, data, headers = self.header) - data = "create table %s.meters(ts timestamp, f1 int, f2 int, f3 int) tags(id int, loc nchar(20))" % self.dbname + data = "create table if not exists %s.meters(ts timestamp, f1 int, f2 int, f3 int) tags(id int, loc nchar(20))" % self.dbname requests.post(self.url, data, headers = self.header) threads = [] @@ -110,7 +153,7 @@ class RestfulInsert: for i in range(self.numOfThreads): threads[i].join() - print("inserting %d records takes %d seconds" % (self.numOfTables * self.recordsPerTable, (time.time() - startTime))) + print("inserting %s records takes %d seconds" % (self.numOfTables * self.recordsPerTable, (time.time() - startTime))) parser = argparse.ArgumentParser() parser.add_argument( @@ -120,6 +163,13 @@ parser.add_argument( default='127.0.0.1', type=str, help='host name to be connected (default: 127.0.0.1)') +parser.add_argument( + '-S', + '--start-timestamp', + action='store', + default=1500000000000, + type=int, + help='insert data from timestamp (default: 1500000000000)') parser.add_argument( '-d', '--db-name', @@ -138,14 +188,14 @@ parser.add_argument( '-T', '--number-of-tables', action='store', - default=1000, + default=10000, type=int, help='Number of tables to be created (default: 1000)') parser.add_argument( '-r', '--number-of-records', action='store', - default=1000, + default=10000, type=int, help='Number of record to be created for each table (default: 1000, -1 for unlimited records)') parser.add_argument( @@ -167,7 +217,18 @@ parser.add_argument( '--out-of-order', action='store_true', help='The order of test data (default: False)') +parser.add_argument( + '-b', + '--table-per-batch', + action='store', + default=1, + type=int, + help='the table per batch (default: 1)') + + args = parser.parse_args() -ri = RestfulInsert(args.host_name, args.db_name, args.number_of_threads, args.number_of_tables, args.number_of_records, args.batch_size, args.table_name_prefix, args.out_of_order) +ri = RestfulInsert( + args.host_name, args.start_timestamp, args.db_name, args.number_of_threads, args.number_of_tables, + args.number_of_records, args.batch_size, args.table_name_prefix, args.out_of_order, args.table_per_batch) ri.run() \ No newline at end of file diff --git a/tests/pytest/pytest_1.sh b/tests/pytest/pytest_1.sh new file mode 100755 index 0000000000000000000000000000000000000000..52f5a30f4e83089bc2e6f230c099e4304a94efd4 --- /dev/null +++ b/tests/pytest/pytest_1.sh @@ -0,0 +1,218 @@ +#!/bin/bash +ulimit -c unlimited + +python3 ./test.py -f insert/basic.py +python3 ./test.py -f insert/int.py +python3 ./test.py -f insert/float.py +python3 ./test.py -f insert/bigint.py +python3 ./test.py -f insert/bool.py +python3 ./test.py -f insert/double.py +python3 ./test.py -f insert/smallint.py +python3 ./test.py -f insert/tinyint.py +python3 ./test.py -f insert/date.py +python3 ./test.py -f insert/binary.py +python3 ./test.py -f insert/nchar.py +#python3 ./test.py -f insert/nchar-boundary.py +python3 ./test.py -f insert/nchar-unicode.py +python3 ./test.py -f insert/multi.py +python3 ./test.py -f insert/randomNullCommit.py +python3 insert/retentionpolicy.py +python3 ./test.py -f insert/alterTableAndInsert.py +python3 ./test.py -f insert/insertIntoTwoTables.py + +python3 ./test.py -f table/alter_wal0.py +python3 ./test.py -f table/column_name.py +python3 ./test.py -f table/column_num.py +python3 ./test.py -f table/db_table.py +python3 ./test.py -f table/create_sensitive.py +#python3 ./test.py -f table/tablename-boundary.py + +# tag +python3 ./test.py -f tag_lite/filter.py +python3 ./test.py -f tag_lite/create-tags-boundary.py +python3 ./test.py -f tag_lite/3.py +python3 ./test.py -f tag_lite/4.py +python3 ./test.py -f tag_lite/5.py +python3 ./test.py -f tag_lite/6.py +python3 ./test.py -f tag_lite/add.py +python3 ./test.py -f tag_lite/bigint.py +python3 ./test.py -f tag_lite/binary_binary.py +python3 ./test.py -f tag_lite/binary.py +python3 ./test.py -f tag_lite/bool_binary.py +python3 ./test.py -f tag_lite/bool_int.py +python3 ./test.py -f tag_lite/bool.py +python3 ./test.py -f tag_lite/change.py +python3 ./test.py -f tag_lite/column.py +python3 ./test.py -f tag_lite/commit.py +python3 ./test.py -f tag_lite/create.py +python3 ./test.py -f tag_lite/datatype.py +python3 ./test.py -f tag_lite/datatype-without-alter.py +python3 ./test.py -f tag_lite/delete.py +python3 ./test.py -f tag_lite/double.py +python3 ./test.py -f tag_lite/float.py +python3 ./test.py -f tag_lite/int_binary.py +python3 ./test.py -f tag_lite/int_float.py +python3 ./test.py -f tag_lite/int.py +python3 ./test.py -f tag_lite/set.py +python3 ./test.py -f tag_lite/smallint.py +python3 ./test.py -f tag_lite/tinyint.py + +#python3 ./test.py -f dbmgmt/database-name-boundary.py + +python3 ./test.py -f import_merge/importBlock1HO.py +python3 ./test.py -f import_merge/importBlock1HPO.py +python3 ./test.py -f import_merge/importBlock1H.py +python3 ./test.py -f import_merge/importBlock1S.py +python3 ./test.py -f import_merge/importBlock1Sub.py +python3 ./test.py -f import_merge/importBlock1TO.py +python3 ./test.py -f import_merge/importBlock1TPO.py +python3 ./test.py -f import_merge/importBlock1T.py +python3 ./test.py -f import_merge/importBlock2HO.py +python3 ./test.py -f import_merge/importBlock2HPO.py +python3 ./test.py -f import_merge/importBlock2H.py +python3 ./test.py -f import_merge/importBlock2S.py +python3 ./test.py -f import_merge/importBlock2Sub.py +python3 ./test.py -f import_merge/importBlock2TO.py +python3 ./test.py -f import_merge/importBlock2TPO.py +python3 ./test.py -f import_merge/importBlock2T.py +python3 ./test.py -f import_merge/importBlockbetween.py +python3 ./test.py -f import_merge/importCacheFileHO.py +python3 ./test.py -f import_merge/importCacheFileHPO.py +python3 ./test.py -f import_merge/importCacheFileH.py +python3 ./test.py -f import_merge/importCacheFileS.py +python3 ./test.py -f import_merge/importCacheFileSub.py +python3 ./test.py -f import_merge/importCacheFileTO.py +python3 ./test.py -f import_merge/importCacheFileTPO.py +python3 ./test.py -f import_merge/importCacheFileT.py +python3 ./test.py -f import_merge/importDataH2.py +python3 ./test.py -f import_merge/importDataHO2.py +python3 ./test.py -f import_merge/importDataHO.py +python3 ./test.py -f import_merge/importDataHPO.py +python3 ./test.py -f import_merge/importDataLastHO.py +python3 ./test.py -f import_merge/importDataLastHPO.py +python3 ./test.py -f import_merge/importDataLastH.py +python3 ./test.py -f import_merge/importDataLastS.py +python3 ./test.py -f import_merge/importDataLastSub.py +python3 ./test.py -f import_merge/importDataLastTO.py +python3 ./test.py -f import_merge/importDataLastTPO.py +python3 ./test.py -f import_merge/importDataLastT.py +python3 ./test.py -f import_merge/importDataS.py +python3 ./test.py -f import_merge/importDataSub.py +python3 ./test.py -f import_merge/importDataTO.py +python3 ./test.py -f import_merge/importDataTPO.py +python3 ./test.py -f import_merge/importDataT.py +python3 ./test.py -f import_merge/importHeadOverlap.py +python3 ./test.py -f import_merge/importHeadPartOverlap.py +python3 ./test.py -f import_merge/importHead.py +python3 ./test.py -f import_merge/importHORestart.py +python3 ./test.py -f import_merge/importHPORestart.py +python3 ./test.py -f import_merge/importHRestart.py +python3 ./test.py -f import_merge/importLastHO.py +python3 ./test.py -f import_merge/importLastHPO.py +python3 ./test.py -f import_merge/importLastH.py +python3 ./test.py -f import_merge/importLastS.py +python3 ./test.py -f import_merge/importLastSub.py +python3 ./test.py -f import_merge/importLastTO.py +python3 ./test.py -f import_merge/importLastTPO.py +python3 ./test.py -f import_merge/importLastT.py +python3 ./test.py -f import_merge/importSpan.py +python3 ./test.py -f import_merge/importSRestart.py +python3 ./test.py -f import_merge/importSubRestart.py +python3 ./test.py -f import_merge/importTailOverlap.py +python3 ./test.py -f import_merge/importTailPartOverlap.py +python3 ./test.py -f import_merge/importTail.py +python3 ./test.py -f import_merge/importToCommit.py +python3 ./test.py -f import_merge/importTORestart.py +python3 ./test.py -f import_merge/importTPORestart.py +python3 ./test.py -f import_merge/importTRestart.py +python3 ./test.py -f import_merge/importInsertThenImport.py +python3 ./test.py -f import_merge/importCSV.py +# user +python3 ./test.py -f user/user_create.py +python3 ./test.py -f user/pass_len.py + +# stable +python3 ./test.py -f stable/query_after_reset.py + +# table +python3 ./test.py -f table/del_stable.py + +#query +python3 ./test.py -f query/filter.py +python3 ./test.py -f query/filterCombo.py +python3 ./test.py -f query/queryNormal.py +python3 ./test.py -f query/queryError.py +python3 ./test.py -f query/filterAllIntTypes.py +python3 ./test.py -f query/filterFloatAndDouble.py +python3 ./test.py -f query/filterOtherTypes.py +python3 ./test.py -f query/querySort.py +python3 ./test.py -f query/queryJoin.py +python3 ./test.py -f query/select_last_crash.py +python3 ./test.py -f query/queryNullValueTest.py +python3 ./test.py -f query/queryInsertValue.py +python3 ./test.py -f query/queryConnection.py +python3 ./test.py -f query/queryCountCSVData.py +python3 ./test.py -f query/natualInterval.py +python3 ./test.py -f query/bug1471.py +#python3 ./test.py -f query/dataLossTest.py +python3 ./test.py -f query/bug1874.py +python3 ./test.py -f query/bug1875.py +python3 ./test.py -f query/bug1876.py +python3 ./test.py -f query/bug2218.py + +#stream +python3 ./test.py -f stream/metric_1.py +python3 ./test.py -f stream/new.py +python3 ./test.py -f stream/stream1.py +python3 ./test.py -f stream/stream2.py +#python3 ./test.py -f stream/parser.py +python3 ./test.py -f stream/history.py + +#alter table +python3 ./test.py -f alter/alter_table_crash.py + +# client +python3 ./test.py -f client/client.py +python3 ./test.py -f client/version.py +python3 ./test.py -f client/alterDatabase.py + +# Misc +python3 testCompress.py +python3 testNoCompress.py +python3 testMinTablesPerVnode.py + +# functions +python3 ./test.py -f functions/function_avg.py -r 1 +python3 ./test.py -f functions/function_bottom.py -r 1 +python3 ./test.py -f functions/function_count.py -r 1 +python3 ./test.py -f functions/function_diff.py -r 1 +python3 ./test.py -f functions/function_first.py -r 1 +python3 ./test.py -f functions/function_last.py -r 1 +python3 ./test.py -f functions/function_last_row.py -r 1 +python3 ./test.py -f functions/function_leastsquares.py -r 1 +python3 ./test.py -f functions/function_max.py -r 1 +python3 ./test.py -f functions/function_min.py -r 1 +python3 ./test.py -f functions/function_operations.py -r 1 +python3 ./test.py -f functions/function_percentile.py -r 1 +python3 ./test.py -f functions/function_spread.py -r 1 +python3 ./test.py -f functions/function_stddev.py -r 1 +python3 ./test.py -f functions/function_sum.py -r 1 +python3 ./test.py -f functions/function_top.py -r 1 +#python3 ./test.py -f functions/function_twa.py -r 1 +python3 queryCount.py +python3 ./test.py -f query/queryGroupbyWithInterval.py +python3 client/twoClients.py +python3 test.py -f query/queryInterval.py +python3 test.py -f query/queryFillTest.py + +# tools +python3 test.py -f tools/taosdemoTest.py +python3 test.py -f tools/taosdumpTest.py +python3 test.py -f tools/lowaTest.py + +# subscribe +python3 test.py -f subscribe/singlemeter.py +#python3 test.py -f subscribe/stability.py +python3 test.py -f subscribe/supertable.py + + diff --git a/tests/pytest/pytest_2.sh b/tests/pytest/pytest_2.sh new file mode 100755 index 0000000000000000000000000000000000000000..fededea3bb34481aa34c37d4a802ba183494e491 --- /dev/null +++ b/tests/pytest/pytest_2.sh @@ -0,0 +1,17 @@ + + +# update +python3 ./test.py -f update/allow_update.py +python3 ./test.py -f update/allow_update-0.py +python3 ./test.py -f update/append_commit_data.py +python3 ./test.py -f update/append_commit_last-0.py +python3 ./test.py -f update/append_commit_last.py +python3 ./test.py -f update/merge_commit_data.py +python3 ./test.py -f update/merge_commit_data-0.py +python3 ./test.py -f update/merge_commit_data2.py +python3 ./test.py -f update/merge_commit_data2_update0.py +python3 ./test.py -f update/merge_commit_last-0.py +python3 ./test.py -f update/merge_commit_last.py + +# wal +python3 ./test.py -f wal/addOldWalTest.py \ No newline at end of file diff --git a/tests/pytest/query/bug2117.py b/tests/pytest/query/bug2117.py new file mode 100644 index 0000000000000000000000000000000000000000..f637558d79fd49d49380c273ee649865d9d3ec4a --- /dev/null +++ b/tests/pytest/query/bug2117.py @@ -0,0 +1,65 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +from util.log import * +from util.cases import * +from util.sql import * +from util.dnodes import * +class TDTestCase: + def init(self, conn, logSql): + tdLog.debug("start to execute %s" % __file__) + tdSql.init(conn.cursor(), logSql) + + def run(self): + tdSql.prepare() + print("==========step1") + print("create table && insert data") + + tdSql.execute("create table mt0 (ts timestamp, c1 int, c2 float, c3 bigint, c4 smallint, c5 tinyint, c6 double, c7 bool,c8 binary(20),c9 nchar(20))") + insertRows = 1000 + t0 = 1604298064000 + tdLog.info("insert %d rows" % (insertRows)) + for i in range(insertRows): + ret = tdSql.execute( + "insert into mt0 values (%d , %d,%d,%d,%d,%d,%d,%d,'%s','%s')" % + (t0+i,i%100,i/2.0,i%41,i%51,i%53,i*1.0,i%2,'taos'+str(i%43),'涛思'+str(i%41))) + print("==========step2") + print("test last with group by normal_col ") + tdSql.query('select last(*) from mt0 group by c3') + tdSql.checkData(0,1,84) + tdSql.checkData(0,9,'涛思0') + tdSql.checkData(1,1,85) + tdSql.checkData(1,9,'涛思1') + tdSql.query('select last(*) from mt0 group by c7') + tdSql.checkData(0,1,98) + tdSql.checkData(0,9,'涛思14') + tdSql.checkData(1,1,99) + tdSql.checkData(1,9,'涛思15') + tdSql.query('select last(*) from mt0 group by c8') + tdSql.checkData(0,3,5) + tdSql.checkData(0,4,20) + tdSql.checkData(3,1,92) + tdSql.checkData(3,9,'涛思8') + tdSql.query('select last(*) from mt0 group by c9') + tdSql.checkData(0,3,0) + tdSql.checkData(0,8,'taos38') + tdSql.checkData(40,1,83) + tdSql.checkData(40,3,40) + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) \ No newline at end of file diff --git a/tests/pytest/query/bug2143.py b/tests/pytest/query/bug2143.py new file mode 100644 index 0000000000000000000000000000000000000000..c28abba53503cbd8c35531c44ca2bd698b2c2d03 --- /dev/null +++ b/tests/pytest/query/bug2143.py @@ -0,0 +1,73 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +from util.log import * +from util.cases import * +from util.sql import * +from util.dnodes import * +class TDTestCase: + def init(self, conn, logSql): + tdLog.debug("start to execute %s" % __file__) + tdSql.init(conn.cursor(), logSql) + + def run(self): + tdSql.prepare() + print("==========step1") + print("create table && insert data") + + tdSql.execute("create table mt0 (ts timestamp, c1 int, c2 float, c3 bigint, c4 smallint, c5 tinyint, c6 double, c7 bool,c8 binary(20),c9 nchar(20))") + insertRows = 1000 + t0 = 1604298064000 + tdLog.info("insert %d rows" % (insertRows)) + for i in range(insertRows): + ret = tdSql.execute( + "insert into mt0 values (%d , %d,%d,%d,%d,%d,%d,%d,'%s','%s')" % + (t0+i,i%100,i/2.0,i%41,i%51,i%53,i*1.0,i%2,'taos'+str(i%43),'涛思'+str(i%41))) + print("==========step2") + print("test group by normal_col with limit offset") + tdSql.query('select max(c1),min(c1),first(c1),last(c1) from mt0 group by c3 limit 3 offset 2') + tdSql.checkData(0,0,99) + tdSql.checkData(0,1,2) + tdSql.checkData(0,2,2) + tdSql.checkData(0,3,86) + tdSql.checkData(1,0,95) + tdSql.checkData(2,1,1) + tdSql.query('select max(c1),min(c1),first(c1),last(c1) from mt0 group by c3 limit 3 offset 40') + tdSql.checkRows(1) + tdSql.query('select max(c1),min(c1),first(c1),last(c1) from mt0 group by c3 limit 3 offset 41') + tdSql.checkRows(0) + tdSql.query('select max(c1),min(c1),first(c1),last(c1) from mt0 group by c3 limit 3 offset 99') + tdSql.checkRows(0) + tdSql.query('select max(c1),min(c1),first(c1),last(c1) from mt0 group by c3 limit 70 offset 3') + tdSql.checkRows(38) + tdSql.query('select max(c1),min(c1),first(c1),last(c1) from mt0 group by c8 limit 3 offset 2') + tdSql.checkData(0,0,91) + tdSql.checkData(0,1,2) + tdSql.checkData(0,2,2) + tdSql.checkData(0,3,91) + tdSql.checkData(1,0,92) + tdSql.checkData(2,1,4) + tdSql.query('select max(c1),min(c1),first(c1),last(c1) from mt0 group by c9 limit 2 offset 9') + tdSql.checkData(0,0,96) + tdSql.checkData(0,1,1) + tdSql.checkData(0,2,9) + tdSql.checkData(0,3,93) + tdSql.checkData(1,0,97) + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) \ No newline at end of file diff --git a/tests/pytest/query/bug2218.py b/tests/pytest/query/bug2218.py index bb92e5d9cee5baae0461cb97fd30de35acd0f5fb..080472383de905c74ca7680d2d7dddd7310daff5 100644 --- a/tests/pytest/query/bug2218.py +++ b/tests/pytest/query/bug2218.py @@ -38,12 +38,12 @@ class TDTestCase: print("test col*1*1 desc ") tdSql.query('select c1,c1*1*1,c2*1*1,c3*1*1,c4*1*1,c5*1*1,c6*1*1 from mt0 order by ts desc limit 2') tdSql.checkData(0,0,99) - tdSql.checkData(0,1,0.0) - tdSql.checkData(0,2,0.0) - tdSql.checkData(0,3,0.0) - tdSql.checkData(0,4,0.0) - tdSql.checkData(0,5,0.0) - tdSql.checkData(0,6,0.0) + tdSql.checkData(0,1,99.0) + tdSql.checkData(0,2,499.0) + tdSql.checkData(0,3,99.0) + tdSql.checkData(0,4,99.0) + tdSql.checkData(0,5,99.0) + tdSql.checkData(0,6,999.0) def stop(self): diff --git a/tests/pytest/query/queryInterval.py b/tests/pytest/query/queryInterval.py index 98b8c9fbef819caebe34fce96bec66da99176131..871c076c0813f0ccd340d33bff992401e522df4a 100644 --- a/tests/pytest/query/queryInterval.py +++ b/tests/pytest/query/queryInterval.py @@ -16,6 +16,7 @@ import taos from util.log import tdLog from util.cases import tdCases from util.sql import tdSql +from util.dnodes import tdDnodes class TDTestCase: @@ -72,6 +73,19 @@ class TDTestCase: tdSql.checkData(6, 0, "2020-09-16 00:00:00") tdSql.checkData(6, 1, 222.0) + # test case for https://jira.taosdata.com:18080/browse/TD-2298 + tdSql.execute("create database test keep 36500") + tdSql.execute("use test") + tdSql.execute("create table t (ts timestamp, voltage int)") + for i in range(10000): + tdSql.execute("insert into t values(%d, 0)" % (1000000 + i * 6000)) + + tdDnodes.stop(1) + tdDnodes.start(1) + tdSql.query("select last(*) from t interval(1s)") + tdSql.checkRows(10000) + + def stop(self): tdSql.close() tdLog.success("%s successfully executed" % __file__) diff --git a/tests/pytest/query/unionAllTest.py b/tests/pytest/query/unionAllTest.py new file mode 100644 index 0000000000000000000000000000000000000000..bb4fb95de676679d448a35b137ee8264cf802ba2 --- /dev/null +++ b/tests/pytest/query/unionAllTest.py @@ -0,0 +1,65 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +import taos +from util.log import tdLog +from util.cases import tdCases +from util.sql import tdSql +import random + + +class TDTestCase: + def init(self, conn, logSql): + tdLog.debug("start to execute %s" % __file__) + tdSql.init(conn.cursor(), logSql) + + self.ts = 1500000000000 + self.num = 10 + + def run(self): + tdSql.prepare() + + tdSql.execute("create table st(ts timestamp, c1 int) tags(loc nchar(20))") + tdSql.execute("create table t0 using st tags('nchar0')") + tdSql.execute("create table t1 using st tags('nchar1')") + tdSql.execute("create table t2 using st tags('nchar2')") + tdSql.execute("create table t3 using st tags('nchar3')") + tdSql.execute("create table t4 using st tags('nchar4')") + tdSql.execute("create table t5 using st tags('nchar5')") + + for i in range(self.num): + tdSql.execute("insert into t0 values(%d, %d)" % (self.ts + i, i)) + tdSql.execute("insert into t1 values(%d, %d)" % (self.ts + i, i)) + tdSql.execute("insert into t2 values(%d, %d)" % (self.ts + i, i)) + tdSql.execute("insert into t3 values(%d, %d)" % (self.ts + i, i)) + tdSql.execute("insert into t4 values(%d, %d)" % (self.ts + i, i)) + tdSql.execute("insert into t5 values(%d, %d)" % (self.ts + i, i)) + + sql = ''' select * from st where loc = 'nchar0' limit 1 union all select * from st where loc = 'nchar1' limit 1 union all select * from st where loc = 'nchar2' limit 1 + union all select * from st where loc = 'nchar3' limit 1 union all select * from st where loc = 'nchar4' limit 1''' + tdSql.query(sql) + tdSql.checkRows(5) + + sql = ''' select * from st where loc = 'nchar0' limit 1 union all select * from st where loc = 'nchar1' limit 1 union all select * from st where loc = 'nchar2' limit 1 + union all select * from st where loc = 'nchar3' limit 1 union all select * from st where loc = 'nchar4' limit 1 union all select * from st where loc = 'nchar5'''' + tdSql.query(sql) + tdSql.checkRows(6) + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) \ No newline at end of file diff --git a/tests/pytest/tools/taosdemoTest2.py b/tests/pytest/tools/taosdemoTest2.py new file mode 100644 index 0000000000000000000000000000000000000000..7d5627be4363467e5e736760eea1e9f0799437d9 --- /dev/null +++ b/tests/pytest/tools/taosdemoTest2.py @@ -0,0 +1,64 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +import os +from util.log import * +from util.cases import * +from util.sql import * +from util.dnodes import * +import threading +import time + + +class TDTestCase: + def init(self, conn, logSql): + tdLog.debug("start to execute %s" % __file__) + tdSql.init(conn.cursor(), logSql) + + self.numberOfTables = 10 + self.numberOfRecords = 1000000 + + def insertDataAndAlterTable(self, threadID): + if(threadID == 0): + os.system("yes | taosdemo -t %d -n %d" % (self.numberOfTables, self.numberOfRecords)) + if(threadID == 1): + print("use test") + tdSql.execute("use test") + print("alter table test.meters add column f4 int") + tdSql.execute("alter table test.meters add column f4 int") + print("insert into test.t0 values (now, 1, 2, 3, 4)") + tdSql.execute("insert into test.t0 values (now, 1, 2, 3, 4)") + + def run(self): + tdSql.prepare() + + t1 = threading.Thread(target=self.insertDataAndAlterTable, args=(0, )) + t2 = threading.Thread(target=self.insertDataAndAlterTable, args=(1, )) + + t1.start() + time.sleep(2) + t2.start() + t1.join() + t2.join() + + tdSql.query("select count(*) from test.meters") + tdSql.checkData(0, 0, self.numberOfRecords * self.numberOfTables + 1) + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) \ No newline at end of file diff --git a/tests/pytest/update/bug_td2279.py b/tests/pytest/update/bug_td2279.py new file mode 100644 index 0000000000000000000000000000000000000000..7e8640dfa09bc904cd49fe88da29bc306fdde6d0 --- /dev/null +++ b/tests/pytest/update/bug_td2279.py @@ -0,0 +1,67 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +import taos +from util.log import * +from util.cases import * +from util.sql import * +from util.dnodes import * + + +class TDTestCase: + def init(self, conn, logSql): + tdLog.debug("start to execute %s" % __file__) + tdSql.init(conn.cursor()) + + self.ts = 1606700000000 + + def restartTaosd(self): + tdDnodes.stop(1) + tdDnodes.startWithoutSleep(1) + tdSql.execute("use db") + + def run(self): + tdSql.prepare() + + print("==============step1") + tdSql.execute("create table t (ts timestamp, a int)") + + for i in range(3276): + tdSql.execute("insert into t values(%d, 0)" % (self.ts + i)) + + newTs = 1606700010000 + for i in range(3275): + tdSql.execute("insert into t values(%d, 0)" % (self.ts + i)) + tdSql.execute("insert into t values(%d, 0)" % 1606700013280) + + self.restartTaosd() + + for i in range(1606700003275, 1606700006609): + tdSql.execute("insert into t values(%d, 0)" % i) + tdSql.execute("insert into t values(%d, 0)" % 1606700006612) + + self.restartTaosd() + + tdSql.execute("insert into t values(%d, 0)" % 1606700006610) + tdSql.query("select * from t") + tdSql.checkRows(6612) + + tdDnodes.stop(1) + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) diff --git a/tests/script/basicSuite.sim b/tests/script/basicSuite.sim index 61a9d68d2658f85256ed4c3cc62dfc53a959594b..5e22e02297bba63ec023780e425019f713dfd8d8 100644 --- a/tests/script/basicSuite.sim +++ b/tests/script/basicSuite.sim @@ -14,7 +14,6 @@ run general/table/vgroup.sim run general/user/authority.sim run general/vector/metrics_mix.sim run general/vector/table_field.sim -run general/user/authority.sim run general/tag/set.sim run general/table/delete_writing.sim run general/stable/disk.sim diff --git a/tests/script/general/alter/dnode.sim b/tests/script/general/alter/dnode.sim index 20ce8799791eaf2aa6198bc07d4abde70a6f31e0..73a095ec054320d0373f137250193bc4d8230234 100644 --- a/tests/script/general/alter/dnode.sim +++ b/tests/script/general/alter/dnode.sim @@ -24,7 +24,7 @@ sql alter dnode 1 debugFlag 135 sql alter dnode 1 debugFlag 131 sql alter dnode 1 monitor 0 sql alter dnode 1 debugFlag 135 -sql alter dnode 1 monitorDebugFlag 135 +sql alter dnode 1 monDebugFlag 135 sql alter dnode 1 vDebugFlag 135 sql alter dnode 1 mDebugFlag 135 sql alter dnode 1 cDebugFlag 135 @@ -44,15 +44,15 @@ sql_error alter dnode 2 tmrDebugFlag 135 print ======== step3 sql_error alter $hostname1 debugFlag 135 -sql_error alter $hostname1 monitorDebugFlag 135 +sql_error alter $hostname1 monDebugFlag 135 sql_error alter $hostname1 vDebugFlag 135 sql_error alter $hostname1 mDebugFlag 135 sql_error alter dnode $hostname2 debugFlag 135 -sql_error alter dnode $hostname2 monitorDebugFlag 135 +sql_error alter dnode $hostname2 monDebugFlag 135 sql_error alter dnode $hostname2 vDebugFlag 135 sql_error alter dnode $hostname2 mDebugFlag 135 sql alter dnode $hostname1 debugFlag 135 -sql alter dnode $hostname1 monitorDebugFlag 135 +sql alter dnode $hostname1 monDebugFlag 135 sql alter dnode $hostname1 vDebugFlag 135 sql alter dnode $hostname1 tmrDebugFlag 131 diff --git a/tests/script/general/db/alter_option.sim b/tests/script/general/db/alter_option.sim index c8aa2480c509353ede3c6ce6269afa9e6efd62b5..1c3f543ffdb149d67cfaf1527a0e88f46f626913 100644 --- a/tests/script/general/db/alter_option.sim +++ b/tests/script/general/db/alter_option.sim @@ -115,31 +115,31 @@ if $data7_db != 20,20,20 then return -1 endi -sql alter database db keep 10 +sql alter database db keep 20 sql show databases print keep $data7_db -if $data7_db != 20,20,10 then +if $data7_db != 20,20,20 then return -1 endi -sql alter database db keep 20 +sql alter database db keep 30 sql show databases print keep $data7_db -if $data7_db != 20,20,20 then +if $data7_db != 20,20,30 then return -1 endi -sql alter database db keep 30 +sql alter database db keep 40 sql show databases print keep $data7_db -if $data7_db != 20,20,30 then +if $data7_db != 20,20,40 then return -1 endi sql alter database db keep 40 sql alter database db keep 30 sql alter database db keep 20 -sql alter database db keep 10 +sql_error alter database db keep 10 sql_error alter database db keep 9 sql_error alter database db keep 1 sql alter database db keep 0 @@ -277,4 +277,4 @@ sql_error alter database db prec 'us' print ============== step status sql_error alter database db status 'delete' -system sh/exec.sh -n dnode1 -s stop -x SIGINT \ No newline at end of file +system sh/exec.sh -n dnode1 -s stop -x SIGINT diff --git a/tests/script/general/parser/col_arithmetic_operation.sim b/tests/script/general/parser/col_arithmetic_operation.sim index 7adae8ef81209aefde853e4fb084b2bc74b84af4..0cc02d088b38ddb915014f306d89c65e7ff41c6c 100644 --- a/tests/script/general/parser/col_arithmetic_operation.sim +++ b/tests/script/general/parser/col_arithmetic_operation.sim @@ -117,16 +117,17 @@ run general/parser/col_arithmetic_query.sim # ================================================================================================ print ====================> crash -# sql select spread(ts )/(1000*3600*24) from ca_stb0 interval(1y) - -sql_error select first(c1, c2) - last(c1, c2) from stb interval(1y) -sql_error select first(ts) - last(ts) from stb interval(1y) -sql_error select top(c1, 2) - last(c1) from stb; -sql_error select stddev(c1) - last(c1) from stb; -sql_error select diff(c1) - last(c1) from stb; -sql_error select first(c7) - last(c7) from stb; -sql_error select first(c8) - last(c8) from stb; -sql_error select first(c9) - last(c9) from stb; +sql use $db +sql select spread(ts )/(1000*3600*24) from $stb interval(1y) + +sql_error select first(c1, c2) - last(c1, c2) from $stb interval(1y) +sql_error select first(ts) - last(ts) from $stb interval(1y) +sql_error select top(c1, 2) - last(c1) from $stb; +sql_error select stddev(c1) - last(c1) from $stb; +sql_error select diff(c1) - last(c1) from $stb; +sql_error select first(c7) - last(c7) from $stb; +sql_error select first(c8) - last(c8) from $stb; +sql_error select first(c9) - last(c9) from $stb; sql_error select max(c2*2) from $tb sql_error select max(c1-c2) from $tb diff --git a/tests/script/general/parser/col_arithmetic_query.sim b/tests/script/general/parser/col_arithmetic_query.sim index 408b039144e1f3de16b2f58fc34b18b39ed9e5e6..53e2c98b56e3c395e810e9edbb1d094baaa20f5c 100644 --- a/tests/script/general/parser/col_arithmetic_query.sim +++ b/tests/script/general/parser/col_arithmetic_query.sim @@ -62,24 +62,73 @@ if $data91 != 1.000000000 then return -1 endi -sql select (c1 * 2) % 7.9 from $tb order by ts desc; +sql select (c1 * 2) % 7.9, c1*1, c1*1*1, c1*c1, c1*c1*c1 from $tb order by ts desc; if $rows != 10000 then return -1 endi -if $data00 != 0.100000000 then - print expect 0.100000000, acutal:$data00 +if $data00 != 2.200000000 then + print expect 2.200000000, actual:$data00 return -1 endi -if $data10 != 2.100000000 then +if $data01 != 9.000000000 then return -1 endi -if $data90 != 6.000000000 then +if $data02 != 9.000000000 then + return -1 +endi + +if $data03 != 81.000000000 then + return -1 +endi + +if $data04 != 729.000000000 then + return -1 +endi + + +if $data10 != 0.200000000 then + return -1 +endi + +if $data11 != 8.000000000 then + return -1 +endi + +if $data12 != 8.000000000 then + return -1 +endi + +if $data13 != 64.000000000 then + return -1 +endi + +if $data14 != 512.000000000 then + return -1 +endi + +if $data90 != 0.000000000 then return -1 endi +if $data91 != 0.000000000 then + return -1 +endi + +if $data92 != 0.000000000 then + return -1 +endi + +if $data93 != 0.000000000 then + return -1 +endi + +if $data94 != 0.000000000 then + return -1 +endi + # [d.3] sql select c1 * c2 /4 from $tb where ts < 1537166000000 and ts > 1537156000000 if $rows != 17 then @@ -95,7 +144,7 @@ if $data10 != 16.000000000 then endi if $data20 != 20.250000000 then - print expect 20.250000000, acutal:$data21 + print expect 20.250000000, actual:$data21 return -1 endi @@ -320,7 +369,9 @@ sql_error select c7-c9 from $tb interval(2y) # multiple retrieve [d.20]=============================================================== sql select c2-c2, 911 from $tb -#======================================= aggregation function arithmetic query cases ================ +#======================================= aggregation function arithmetic query cases =================================== +# on $tb percentile() spread(ts) bug + # asc/desc order [d.2] sql select first(c1) * ( 2 / 3 ) from $stb order by ts asc; if $rows != 1 then @@ -349,11 +400,11 @@ if $data00 != 1.800000000 then return -1 endi -if $data01 != 100000 then +if $data01 != 100000.000000000 then return -1 endi -if $data02 != 200000 then +if $data02 != 200000.000000000 then return -1 endi @@ -374,77 +425,192 @@ if $data02 != 9.000000020 then return -1 endi -# all possible function in the arithmetic expressioin -sql select min(c1) * max(c2) /4, sum(c1) * percentile(c2, 20), apercentile(c4, 33) + 52/9, spread(c5)/min(c2) from $stb where ts < and ts > +# all possible function in the arithmetic expression, add more +sql select min(c1) * max(c2) /4, sum(c1) * apercentile(c2, 20), apercentile(c4, 33) + 52/9, spread(c5)/min(c2), count(1)/sum(c1), avg(c2)*count(c2) from $stb where ts >= '2018-09-17 09:00:00.000' and ts <= '2018-11-25 19:30:00.000'; +if $rows != 1 then + return -1 +endi -# no result return [d.3] +if $data00 != 0.000000000 then + return -1 +endi + +if $data01 != 225000.000000000 then + return -1 +endi + +if $data02 != 8.077777778 then + return -1 +endi + +if $data03 != inf then + return -1 +endi + +if $data04 != 0.444444444 then + return -1 +endi + +if $data05 != 450000.000000000 then + return -1 +endi + +# no result return [d.3]=============================================================== sql select first(c1) * 91 - 7, last(c3) from $stb where ts < 1537146000000 if $rows != 0 then return -1 endi # no result return [d.3] -sql select sum(c2) - avg(c2) from $tb where ts>xxx +sql select sum(c2) - avg(c2) from $stb where ts > '2018-11-25 19:30:00.000' if $rows != 0 then return -1 endi -# single row result aggregation [d.4] -sql select +# single row result aggregation [d.4]=================================================== +# all those cases are aggregation test cases. # error cases -sql_error select first(c1, c2) - last(c1, c2) from $tb +sql_error select first(c1, c2) - last(c1, c2) from $stb +sql_error select top(c1, 5) - bottom(c1, 5) from $stb +sql_error select first(*) - 99 from $stb # multi row result aggregation [d.4] -sql select top(c1, 1) - bottom(c1, 1) from $tb -sql select top(c1, 99) - bottom(c1, 99) from $tb +sql_error select top(c1, 1) - bottom(c1, 1) from $stb +sql_error select top(c1, 99) - bottom(c1, 99) from $stb + +# query on super table [d.5]============================================================= +# all cases in this part are query on super table -# all data types [d.6] -sql select c2-c1, c3/c2, c4*c3, c5%c4, c6+99%22 from $tb +# all data types [d.6]=================================================================== +sql select c2-c1, c3/c2, c4*c3, c5%c4, c6+99%22 from $stb # error case, ts/bool/binary/nchar not support arithmetic expression -sql_error select ts+ts from $tb -sql_error select ts+22 from $tb -sql_error select c7*12 from $tb -sql_error select c8/55 from $tb -sql_error select c9+c8 from $tb +sql_error select first(c7)*12 from $stb +sql_error select last(c8)/55 from $stb +sql_error select last_row(c9) + last_row(c8) from $stb -# arithmetic expression in join [d.7] +# arithmetic expression in join [d.7]=============================================================== -# arithmetic expression in union [d.8] +# arithmetic expression in union [d.8]=============================================================== -# arithmetic expression in group by [d.9] +# arithmetic expression in group by [d.9]=============================================================== # in group by tag -# not support for normal table -sql_error select c5*99 from $tb group by t1 +sql select avg(c4)*99 from $stb group by t1 +if $rows != 10 then + return -1 +endi + +if $data00 != 445.500000000 then + return -1 +endi + +if $data01 != 0 then + return -1 +endi + +if $data90 != 445.500000000 then + return -1 +endi + +if $data91 != 9 then + return -1 +endi # in group by column -sql_error select c6-c6+c3*12 from $tb group by c3; +sql select apercentile(c6, 50)-first(c6)+last(c5)*12, last(c5)*12 from ca_stb0 group by c2; +if $rows != 10 then + return -1 +endi + +if $data00 != 0.000000000 then + return -1 +endi -sql select first(c6) - last(c6) *12 / count(*) from $tb group by c3; +if $data01 != 0.000000000 then + return -1 +endi -# limit offset [d.10] -sql select c6-c6+12 from $tb limit 12 offset 99; -sql select c4/99.123 from $tb limit 1 offset 9999; +if $data10 != 12.000000000 then + return -1 +endi -# slimit/soffset not suport for normal table query. [d.11] -sql_error select sum(c1) from $tb slimit 1 soffset 19; +if $data11 != 12.000000000 then + return -1 +endi + +if $data20 != 24.000000000 then + return -1 +endi + +if $data21 != 24.000000000 then + return -1 +endi + +sql_error select first(c6) - last(c6) *12 / count(*) from $stb group by c3; + +sql select first(c6) - last(c6) *12 / count(*) from $stb group by c5; +if $rows != 10 then + return -1 +endi + +if $data00 != 0.000000000 then + return -1 +endi + +if $data10 != 0.997600000 then + return -1 +endi + +if $data90 != 8.978400000 then + return -1 +endi + +# limit offset [d.10]=============================================================== +sql select first(c6) - sum(c6) + 12 from $stb limit 12 offset 0; +if $rows != 1 then + return -1 +endi + +if $data00 != -449988.000000000 then + return -1 +endi + +sql select apercentile(c4, 21) / 99.123 from $stb limit 1 offset 1; +if $rows != 0 then + return -1 +endi + +sql select apercentile(c4, 21) / sum(c4) from $stb interval(1s) limit 1 offset 1; +if $rows != 1 then + return -1 +endi + +# slimit/soffset not support for normal table query. [d.11]=============================================================== +sql select sum(c1) from $stb slimit 1 soffset 19; +if $rows != 0 then + return -1 +endi -# fill [d.12] -sql_error select c2-c2, c3-c4, c5%c6 from $tb fill(value, 12); +sql select sum(c1) from $stb interval(1s) group by tbname slimit 1 soffset 1 +sql select sum(c1) from ca_stb0 interval(1s) group by tbname slimit 2 soffset 4 limit 10 offset 1 -# constant column. [d.13] +# fill [d.12]=============================================================== +sql_error select first(c1)-last(c1), sum(c3)*count(c3), spread(c5 ) % count(*) from $stb interval(1s) fill(prev); +sql_error select first(c1) from $stb fill(value, 20); +# constant column. [d.13]=============================================================== -# column value filter [d.14] +# column value filter [d.14]=============================================================== -# tag filter(not support for normal table). [d.15] -sql_error select sum(c2)+99 from $tb where t1=12; -# multi-field output [d.16] +# tag filter. [d.15]=============================================================== +sql select sum(c2)+99 from $stb where t1=12; + +# multi-field output [d.16]=============================================================== sql select count(*), sum(c1)*avg(c2), avg(c3)*count(c3), sum(c3), sum(c4), first(c7), last(c8), first(c9), first(c7), last(c8) from $tb sql select c4*1+1/2 from $tb @@ -461,18 +627,30 @@ if $data90 != 9.500000000 then return -1 endi -# interval query [d.17] -sql_error select c2*c2, c3-c3, c4+9 from $tb interval(1s) -sql_error select c7-c9 from $tb interval(2y) +# interval query [d.17]=============================================================== +sql select avg(c2)*count(c2), sum(c3)-first(c3), last(c4)+9 from $stb interval(1s) +if $rows != 10000 then + return -1 +endi -# aggregation query [d.18] -# see test cases below +if $data00 != @18-09-17 09:00:00.000@ then + return -1 +endi -# first/last query [d.19] -# see test cases below +sql_error select first(c7)- last(c1) from $tb interval(2y) -# multiple retrieve [d.20] -sql select c2-c2 from $tb; +# aggregation query [d.18]=============================================================== +# all cases in this part are aggregation query test. + +# first/last query [d.19]=============================================================== + + +# multiple retrieve [d.20]=============================================================== +sql select c2-c2 from $tb sql select first(c1)-last(c1), spread(c2), max(c3) - min(c3), avg(c4)*count(c4) from $tb + + +#====================================================super table query================================================== + diff --git a/tests/script/general/parser/first_last.sim b/tests/script/general/parser/first_last.sim index a934d3bcabf09ebe1584c296d6cab2f5c80654d2..773f92afcfd1fb126f137befd9de1107b050356f 100644 --- a/tests/script/general/parser/first_last.sim +++ b/tests/script/general/parser/first_last.sim @@ -46,7 +46,8 @@ while $i < $tbNum endw $i = $i + 1 -endw +endw + $ts = $ts + 60000 $tb = $tbPrefix . 0 sql insert into $tb (ts) values ( $ts ) @@ -84,4 +85,43 @@ sleep 500 run general/parser/first_last_query.sim +print =================> insert data regression test +sql create database test keep 36500 +sql use test +sql create table tm0 (ts timestamp, k int) + +print =========================> td-2298 +$ts0 = 1537146000000 +$xs = 6000 + +$x = 0 +while $x < 5000 + $ts = $ts0 + $xs + $ts1 = $ts + $xs + $x1 = $x + 1 + + sql insert into tm0 values ( $ts , $x ) ( $ts1 , $x1 ) + $x = $x1 + $ts0 = $ts1 +endw + +system sh/exec.sh -n dnode1 -s stop -x SIGINT +sleep 3000 +system sh/exec.sh -n dnode1 -s start +print ================== server restart completed +sql connect +sleep 500 + +sql use test +sql select count(*), last(ts) from tm0 interval(1s) +if $rows != 10000 then + print expect 10000, actual: $rows + return -1 +endi + +sql select last(ts) from tm0 interval(1s) +if $rows != 10000 then + return -1 +endi + system sh/exec.sh -n dnode1 -s stop -x SIGINT \ No newline at end of file diff --git a/tests/script/general/parser/first_last_query.sim b/tests/script/general/parser/first_last_query.sim index a982f10362a46e7c671e8df4674b06c5c9f60b75..52d888b04d1b055bbbc3c8509832c96579e4320f 100644 --- a/tests/script/general/parser/first_last_query.sim +++ b/tests/script/general/parser/first_last_query.sim @@ -266,4 +266,6 @@ endi if $data14 != @test2@ then print expect test2 , actual: $data14 return -1 -endi \ No newline at end of file +endi + +sql drop table stest \ No newline at end of file diff --git a/tests/script/general/parser/function.sim b/tests/script/general/parser/function.sim new file mode 100644 index 0000000000000000000000000000000000000000..34e9844f711313d9f7c1c2380177bfc2c7aa4740 --- /dev/null +++ b/tests/script/general/parser/function.sim @@ -0,0 +1,228 @@ +system sh/stop_dnodes.sh + +system sh/deploy.sh -n dnode1 -i 1 +system sh/cfg.sh -n dnode1 -c walLevel -v 0 +system sh/cfg.sh -n dnode1 -c tableMetaKeepTimer -v 3 +system sh/exec.sh -n dnode1 -s start +sleep 500 +sql connect + +$dbPrefix = m_func_db +$tbPrefix = m_func_tb +$mtPrefix = m_func_mt + +$tbNum = 10 +$rowNum = 5 +$totalNum = $tbNum * $rowNum +$ts0 = 1537146000000 +$delta = 600000 +print ========== alter.sim +$i = 0 +$db = $dbPrefix . $i +$mt = $mtPrefix . $i + +sql drop database if exists $db +sql create database $db +sql use $db + +print =====================================> test case for twa in single block + +sql create table t1 (ts timestamp, k float); +sql insert into t1 values('2015-08-18 00:00:00', 2.064); +sql insert into t1 values('2015-08-18 00:06:00', 2.116); +sql insert into t1 values('2015-08-18 00:12:00', 2.028); +sql insert into t1 values('2015-08-18 00:18:00', 2.126); +sql insert into t1 values('2015-08-18 00:24:00', 2.041); +sql insert into t1 values('2015-08-18 00:30:00', 2.051); + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:05:00' +if $rows != 1 then + return -1 +endi + +if $data00 != 2.063999891 then + return -1 +endi + +if $data01 != 2.063999891 then + return -1 +endi + +if $data02 != 1 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:07:00' +if $rows != 1 then + return -1 +endi + +if $data00 != 2.089999914 then + return -1 +endi + +if $data01 != 2.089999914 then + return -1 +endi + +if $data02 != 2 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:07:00' interval(1m) order by ts asc +if $rows != 2 then + return -1 +endi + +if $data00 != @15-08-18 00:00:00.000@ then + return -1 +endi + +if $data01 != 2.068333156 then + return -1 +endi + +if $data02 != 2.063999891 then + return -1 +endi + +if $data03 != 1 then + return -1 +endi + +if $data10 != @15-08-18 00:06:00.000@ then + return -1 +endi + +if $data11 != 2.115999937 then + return -1 +endi + +if $data12 != 2.115999937 then + return -1 +endi + +if $data13 != 1 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:07:00' interval(1m) order by ts desc; +if $rows != 2 then + return -1 +endi + +if $data00 != @15-08-18 00:06:00.00@ then + return -1 +endi + +if $data01 != 2.115999937 then + return -1 +endi + +if $data02 != 2.115999937 then + return -1 +endi + +if $data03 != 1 then + return -1 +endi + +if $data11 != 2.068333156 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:27:00' interval(10m) order by ts asc +if $rows != 3 then + return -1 +endi + +if $data01 != 2.088666666 then + return -1 +endi + +if $data02 != 2.089999914 then + return -1 +endi + +if $data03 != 2 then + return -1 +endi + +if $data11 != 2.077099980 then + return -1 +endi + +if $data12 != 2.077000022 then + return -1 +endi + +if $data13 != 2 then + return -1 +endi + +if $data21 != 2.069333235 then + return -1 +endi + +if $data22 != 2.040999889 then + return -1 +endi + +if $data23 != 1 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:27:00' interval(10m) order by ts desc +if $rows != 3 then + return -1 +endi + +if $data01 != 2.069333235 then + return -1 +endi + +if $data11 != 2.077099980 then + return -1 +endi + +if $data21 != 2.088666666 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:30:00' order by ts asc +if $data00 != 2.073699975 then + return -1 +endi + +if $data01 != 2.070999980 then + return -1 +endi + +if $data02 != 6 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:30:00' order by ts desc +if $rows != 1 then + return -1 +endi + +if $data00 != 2.073699975 then + return -1 +endi + +if $data01 != 2.070999980 then + return -1 +endi + +if $data02 != 6 then + return -1 +endi + +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:30:00' interval(10m) order by ts asc +sql select twa(k),avg(k),count(1) from t1 where ts>='2015-8-18 00:00:00' and ts<='2015-8-18 00:30:00' interval(10m) order by ts desc + + +#todo add test case while column filte exists. + +select count(*),TWA(k) from tm0 where ts>='1970-1-1 13:43:00' and ts<='1970-1-1 13:44:10' interval(9s) diff --git a/tests/script/general/parser/groupby.sim b/tests/script/general/parser/groupby.sim index 19d9ae84cbc02e179c2ab060082507529b2bb608..44bae6f242f12e5a6430df64520631d0ac3af073 100644 --- a/tests/script/general/parser/groupby.sim +++ b/tests/script/general/parser/groupby.sim @@ -606,6 +606,44 @@ sql insert into t1 values ('2020-03-27 04:21:16.000', 1)('2020-03-27 04:31:17.00 sql insert into t2 values ('2020-03-27 04:11:16.000', 1)('2020-03-27 04:11:17.000', 2) ('2020-03-27 04:11:18.000', 3) ('2020-03-27 04:11:19.000', 4) ; sql insert into t2 values ('2020-03-27 04:21:16.000', 1)('2020-03-27 04:31:17.000', 2) ('2020-03-27 04:51:18.000', 3) ('2020-03-27 05:10:19.000', 4) ; +print =================>TD-2236 +sql select first(ts),last(ts) from t1 group by c; +if $rows != 4 then + return -1 +endi + +if $data00 != @20-03-27 04:11:16.000@ then + return -1 +endi + +if $data01 != @20-03-27 04:21:16.000@ then + return -1 +endi + +if $data10 != @20-03-27 04:11:17.000@ then + return -1 +endi + +if $data11 != @20-03-27 04:31:17.000@ then + return -1 +endi + +if $data20 != @20-03-27 04:11:18.000@ then + return -1 +endi + +if $data21 != @20-03-27 04:51:18.000@ then + return -1 +endi + +if $data30 != @20-03-27 04:11:19.000@ then + return -1 +endi + +if $data31 != @20-03-27 05:10:19.000@ then + return -1 +endi + #sql select irate(c) from st where t1="1" and ts >= '2020-03-27 04:11:17.732' and ts < '2020-03-27 05:11:17.732' interval(1m) sliding(15s) group by tbname,t1,t2; #if $rows != 40 then # return -1 diff --git a/tests/script/general/parser/limit.sim b/tests/script/general/parser/limit.sim index fb5e704bf1ba3bab457dae2dc06a23628fde316c..2089cd3d2af96e95a910acaf1067b90aa40c26be 100644 --- a/tests/script/general/parser/limit.sim +++ b/tests/script/general/parser/limit.sim @@ -48,8 +48,12 @@ while $i < $halfNum $binary = $binary . ' $nchar = 'nchar . $c $nchar = $nchar . ' - sql insert into $tb values ( $ts , $c , $c , $c , $c , $c , $c , true, $binary , $nchar ) - sql insert into $tb1 values ( $ts , $c , NULL , $c , NULL , $c , $c , true, $binary , $nchar ) + + $ts = $ts + $i + sql insert into $tb values ( $ts , $c , $c , $c , $c , $c , $c , true, $binary , $nchar ) + + $ts = $ts + $halfNum + sql insert into $tb1 values ( $ts , $c , NULL , $c , NULL , $c , $c , true, $binary , $nchar ) $x = $x + 1 endw diff --git a/tests/script/general/parser/limit1_stb.sim b/tests/script/general/parser/limit1_stb.sim index d5846adc45327e6faa7b2bd7cdd10b44ea6fc094..7d61a826aa90c8dd5b3a62de1e9688c0278f59fd 100644 --- a/tests/script/general/parser/limit1_stb.sim +++ b/tests/script/general/parser/limit1_stb.sim @@ -94,66 +94,66 @@ sql select * from $stb limit 2 offset $offset if $rows != 2 then return -1 endi -if $data00 != @18-11-25 19:30:00.000@ then - return -1 -endi -if $data01 != 9 then - return -1 -endi -if $data02 != 9 then - return -1 -endi -if $data03 != 9.00000 then - return -1 -endi -if $data04 != 9.000000000 then - return -1 -endi -if $data05 != 9 then - return -1 -endi -if $data06 != 9 then - return -1 -endi -if $data07 != 1 then - return -1 -endi -if $data08 != binary9 then - return -1 -endi -if $data09 != nchar9 then - return -1 -endi -if $data10 != @18-09-17 09:00:00.000@ then - return -1 -endi -if $data11 != 0 then - return -1 -endi -if $data12 != NULL then - return -1 -endi -if $data13 != 0.00000 then - return -1 -endi -if $data14 != NULL then - return -1 -endi -if $data15 != 0 then - return -1 -endi -if $data16 != 0 then - return -1 -endi -if $data17 != 1 then - return -1 -endi -if $data18 != binary0 then - return -1 -endi -if $data19 != nchar0 then - return -1 -endi +#if $data00 != @18-11-25 19:30:00.000@ then +# return -1 +#endi +#if $data01 != 9 then +# return -1 +#endi +#if $data02 != 9 then +# return -1 +#endi +#if $data03 != 9.00000 then +# return -1 +#endi +#if $data04 != 9.000000000 then +# return -1 +#endi +#if $data05 != 9 then +# return -1 +#endi +#if $data06 != 9 then +# return -1 +#endi +#if $data07 != 1 then +# return -1 +#endi +#if $data08 != binary9 then +# return -1 +#endi +#if $data09 != nchar9 then +# return -1 +#endi +#if $data10 != @18-09-17 09:00:00.000@ then +# return -1 +#endi +#if $data11 != 0 then +# return -1 +#endi +#if $data12 != NULL then +# return -1 +#endi +#if $data13 != 0.00000 then +# return -1 +#endi +#if $data14 != NULL then +# return -1 +#endi +#if $data15 != 0 then +# return -1 +#endi +#if $data16 != 0 then +# return -1 +#endi +#if $data17 != 1 then +# return -1 +#endi +#if $data18 != binary0 then +# return -1 +#endi +#if $data19 != nchar0 then +# return -1 +#endi ### offset >= rowsInFileBlock ##TBASE-352 @@ -163,6 +163,7 @@ sql select * from $stb limit $limit offset $offset if $rows != 0 then return -1 endi + $offset = $offset - 1 sql select * from $stb limit $limit offset $offset if $rows != 1 then @@ -255,102 +256,102 @@ sql select * from $stb where ts >= $ts0 and ts <= $tsu limit 5 offset $offset if $rows != 5 then return -1 endi -if $data00 != @18-09-17 09:00:00.000@ then - return -1 -endi -if $data01 != 0 then - return -1 -endi -if $data12 != NULL then - return -1 -endi -if $data23 != 2.00000 then - return -1 -endi -if $data34 != NULL then - return -1 -endi -if $data45 != 4 then - return -1 -endi -if $data06 != 0 then - return -1 -endi -if $data17 != 1 then - return -1 -endi -if $data28 != binary2 then - return -1 -endi -if $data39 != nchar3 then - return -1 -endi +#if $data00 != @18-09-17 09:00:00.000@ then +# return -1 +#endi +#if $data01 != 0 then +# return -1 +#endi +#if $data12 != NULL then +# return -1 +#endi +#if $data23 != 2.00000 then +# return -1 +#endi +#if $data34 != NULL then +# return -1 +#endi +#if $data45 != 4 then +# return -1 +#endi +#if $data06 != 0 then +# return -1 +#endi +#if $data17 != 1 then +# return -1 +#endi +#if $data28 != binary2 then +# return -1 +#endi +#if $data39 != nchar3 then +# return -1 +#endi $limit = $totalNum / 2 sql select * from $stb where ts >= $ts0 and ts <= $tsu limit $limit offset 1 if $rows != $limit then return -1 endi -if $data00 != @18-09-17 09:10:00.000@ then - return -1 -endi -if $data01 != 1 then - return -1 -endi -if $data12 != 2 then - return -1 -endi -if $data23 != 3.00000 then - return -1 -endi -if $data34 != 4.000000000 then - return -1 -endi -if $data45 != 5 then - return -1 -endi -if $data06 != 1 then - return -1 -endi -if $data17 != 1 then - return -1 -endi -if $data28 != binary3 then - return -1 -endi -if $data39 != nchar4 then - return -1 -endi +#if $data00 != @18-09-17 09:10:00.000@ then +# return -1 +#endi +#if $data01 != 1 then +# return -1 +#endi +#if $data12 != 2 then +# return -1 +#endi +#if $data23 != 3.00000 then +# return -1 +#endi +#if $data34 != 4.000000000 then +# return -1 +#endi +#if $data45 != 5 then +# return -1 +#endi +#if $data06 != 1 then +# return -1 +#endi +#if $data17 != 1 then +# return -1 +#endi +#if $data28 != binary3 then +# return -1 +#endi +#if $data39 != nchar4 then +# return -1 +#endi sql select max(c1), min(c2), avg(c3), sum(c5), spread(c6), first(c7), last(c8), first(c9) from $stb where ts >= $ts0 and ts <= $tsu limit 1 offset 0 if $rows != 1 then return -1 endi -if $data00 != 9 then - return -1 -endi -if $data01 != 0 then - return -1 -endi -if $data02 != 4.500000000 then - return -1 -endi -$val = 45 * $rowNum -if $data03 != $val then - return -1 -endi -if $data04 != 9.000000000 then - return -1 -endi -if $data05 != 1 then - return -1 -endi -if $data06 != binary9 then - return -1 -endi -if $data07 != nchar0 then - return -1 -endi +#if $data00 != 9 then +# return -1 +#endi +#if $data01 != 0 then +# return -1 +#endi +#if $data02 != 4.500000000 then +# return -1 +#endi +#$val = 45 * $rowNum +#if $data03 != $val then +# return -1 +#endi +#if $data04 != 9.000000000 then +# return -1 +#endi +#if $data05 != 1 then +# return -1 +#endi +#if $data06 != binary9 then +# return -1 +#endi +#if $data07 != nchar0 then +# return -1 +#endi sql select max(c1), min(c2), avg(c3), sum(c5), spread(c6), first(c7), last(c8), first(c9) from $stb where ts >= $ts0 and ts <= $tsu and c1 > 1 and c2 < 9 and c3 > 2 and c4 < 8 and c5 > 3 and c6 < 7 and c7 != 0 and c8 like '%5' and t1 > 3 and t1 < 6 limit 1 offset 0; if $rows != 1 then diff --git a/tests/script/general/parser/limit1_tb.sim b/tests/script/general/parser/limit1_tb.sim index 1e473eb8580965b6e777c7fe09cbc607af8938a8..72b63256dba36c799e308b41bebb4a53d5d64334 100644 --- a/tests/script/general/parser/limit1_tb.sim +++ b/tests/script/general/parser/limit1_tb.sim @@ -703,13 +703,13 @@ sql select twa(c1), twa(c2), twa(c3), twa(c4), twa(c5), twa(c6) from $tb where t if $rows != 1 then return -1 endi -if $data00 != 4.499549955 then +if $data00 != 4.500000000 then return -1 endi -if $data02 != 4.499549955 then +if $data02 != 4.500000000 then return -1 endi -if $data05 != 4.499549955 then +if $data05 != 4.500000000 then return -1 endi @@ -717,10 +717,12 @@ sql select first(c1), first(c2), first(c3), first(c4), first(c5), first(c6) from if $rows != 0 then return -1 endi + sql select first(c1), first(c2), first(c3), first(c4), first(c5), first(c6) from $tb where ts >= $ts0 and ts <= $tsu interval(30m) limit 3 offset 1 if $rows != 3 then return -1 endi + if $data01 != 3 then return -1 endi @@ -731,7 +733,6 @@ if $data23 != 9.00000 then return -1 endi - sql select last(c1), last(c2), last(c3), last(c4), last(c5), last(c6) from $tb where ts >= $ts0 and ts <= $tsu limit 5 offset 1 if $rows != 0 then return -1 diff --git a/tests/script/general/parser/limit_stb.sim b/tests/script/general/parser/limit_stb.sim index b41b7b726bc24b83f64cecbec8ac0eb83afa57ea..d929810817927572eb7e1929ce5d98cde0940916 100644 --- a/tests/script/general/parser/limit_stb.sim +++ b/tests/script/general/parser/limit_stb.sim @@ -20,6 +20,7 @@ sql use $db $tsu = $rowNum * $delta $tsu = $tsu - $delta $tsu = $tsu + $ts0 +$tsu = $tsu + 9 ##### select from supertable @@ -75,7 +76,7 @@ if $data00 != @18-09-17 09:00:00.000@ then return -1 endi -if $data40 != @18-09-17 09:00:00.000@ then +if $data40 != @18-09-17 09:00:00.004@ then return -1 endi @@ -84,11 +85,11 @@ if $data01 != 0 then endi print data12 = $data12 -if $data12 != NULL then +if $data12 != 0 then return -1 endi -if $data24 != NULL then +if $data24 != 0.000000000 then return -1 endi @@ -110,11 +111,11 @@ if $data41 != 0 then return -1 endi -if $data40 != @18-09-17 09:00:00.000@ then +if $data40 != @18-09-17 09:00:00.005@ then return -1 endi -if $data00 != @18-09-17 09:00:00.000@ then +if $data00 != @18-09-17 09:00:00.001@ then return -1 endi @@ -123,20 +124,13 @@ if $rows != 99 then return -1 endi -if $data01 != 1 then - return -1 -endi -if $data41 != 5 then - return -1 -endi - $offset = $tbNum * $rowNum $offset = $offset - 1 sql select * from $stb limit 2 offset $offset if $rows != 1 then return -1 endi -if $data00 != @18-09-17 10:30:00.000@ then +if $data00 != @18-09-17 10:30:00.009@ then return -1 endi if $data01 != 9 then @@ -174,7 +168,7 @@ sql select * from $stb limit 2 offset $offset if $rows != 2 then return -1 endi -if $data00 != @18-09-17 10:30:00.000@ then +if $data00 != @18-09-17 10:30:00.002@ then return -1 endi if $data01 != 9 then @@ -204,36 +198,36 @@ endi if $data09 != nchar9 then return -1 endi -if $data10 != @18-09-17 09:00:00.000@ then - return -1 -endi -if $data11 != 0 then - return -1 -endi -if $data12 != NULL then - return -1 -endi -if $data13 != 0.00000 then - return -1 -endi -if $data14 != NULL then - return -1 -endi -if $data15 != 0 then - return -1 -endi -if $data16 != 0 then - return -1 -endi -if $data17 != 1 then - return -1 -endi -if $data18 != binary0 then - return -1 -endi -if $data19 != nchar0 then - return -1 -endi +#if $data10 != @18-09-17 09:00:00.000@ then +# return -1 +#endi +#if $data11 != 0 then +# return -1 +#endi +#if $data12 != NULL then +# return -1 +#endi +#if $data13 != 0.00000 then +# return -1 +#endi +#if $data14 != NULL then +# return -1 +#endi +#if $data15 != 0 then +# return -1 +#endi +#if $data16 != 0 then +# return -1 +#endi +#if $data17 != 1 then +# return -1 +#endi +#if $data18 != binary0 then +# return -1 +#endi +#if $data19 != nchar0 then +# return -1 +#endi $offset = $rowNum * $tbNum sql select * from lm_stb0 limit 2 offset $offset @@ -248,6 +242,7 @@ endi if $data01 != 0 then return -1 endi + sql select ts, c1, c2, c3, c4, c5, c6, c7, c8, c9 from $stb limit 1 offset 1; if $rows != 1 then return -1 @@ -288,52 +283,52 @@ if $data09 != nchar4 then endi ### select from supertable + where + limit offset -sql select * from $stb where ts > '2018-09-17 09:30:00.000' and ts < '2018-09-17 10:30:00.000' limit 5 offset 1 +sql select * from $stb where ts > '2018-09-17 09:30:00.000' and ts < '2018-09-17 10:30:00.000' order by ts asc limit 5 offset 1 if $rows != 5 then return -1 endi -if $data01 != 5 then +if $data01 != 3 then return -1 endi -if $data11 != 6 then +if $data11 != 3 then return -1 endi -if $data21 != 7 then +if $data21 != 3 then return -1 endi -if $data31 != 8 then +if $data31 != 3 then return -1 endi -if $data41 != 4 then +if $data41 != 3 then return -1 endi -sql select * from $stb where ts > '2018-09-17 09:30:00.000' and ts < '2018-09-17 10:30:00.000' limit 5 offset 50 +sql select * from $stb where ts > '2018-09-17 09:30:00.000' and ts < '2018-09-17 10:10:00.000' order by ts asc limit 5 offset 50 if $rows != 0 then return -1 endi -sql select * from $stb where ts > '2018-09-17 09:30:00.000' and ts < '2018-09-17 10:30:00.000' limit 5 offset 1 +sql select * from $stb where ts > '2018-09-17 09:30:00.000' and ts < '2018-09-17 10:30:00.000' order by ts asc limit 5 offset 1 if $rows != 5 then return -1 endi -if $data01 != 5 then +if $data01 != 3 then return -1 endi -if $data11 != 6 then +if $data11 != 3 then return -1 endi -if $data21 != 7 then +if $data21 != 3 then return -1 endi -if $data31 != 8 then +if $data31 != 3 then return -1 endi -if $data41 != 4 then +if $data41 != 3 then return -1 endi -sql select max(c1), min(c2), avg(c3), sum(c5), spread(c6), first(c7), last(c8), first(c9) from lm_stb0 where ts >= '2018-09-17 09:00:00.000' and ts <= '2018-09-17 10:30:00.000' limit 1 offset 0; +sql select max(c1), min(c2), avg(c3), sum(c5), spread(c6), first(c7), last(c8), first(c9) from lm_stb0 where ts >= '2018-09-17 09:00:00.000' and ts <= '2018-09-17 10:30:00.009' order by ts asc limit 1 offset 0; if $rows != 1 then return -1 endi @@ -842,9 +837,6 @@ sql select top(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu order by ts desc if $rows != 3 then return -1 endi -if $data00 != @18-09-17 10:30:00.000@ then - return -1 -endi if $data01 != 9 then return -1 endi @@ -853,9 +845,6 @@ sql select top(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu order by ts asc if $rows != 3 then return -1 endi -if $data00 != @18-09-17 10:30:00.000@ then - return -1 -endi if $data01 != 9 then return -1 endi @@ -864,7 +853,7 @@ sql select top(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu group by t1 orde if $rows != 6 then return -1 endi -if $data00 != @18-09-17 10:00:00.000@ then +if $data00 != @18-09-17 10:00:00.008@ then return -1 endi if $data01 != 6 then @@ -873,7 +862,7 @@ endi if $data02 != 8 then return -1 endi -if $data10 != @18-09-17 10:10:00.000@ then +if $data10 != @18-09-17 10:10:00.008@ then return -1 endi if $data11 != 7 then @@ -882,7 +871,7 @@ endi if $data12 != 8 then return -1 endi -if $data20 != @18-09-17 10:20:00.000@ then +if $data20 != @18-09-17 10:20:00.008@ then return -1 endi if $data21 != 8 then @@ -891,7 +880,7 @@ endi if $data22 != 8 then return -1 endi -if $data30 != @18-09-17 10:00:00.000@ then +if $data30 != @18-09-17 10:00:00.007@ then return -1 endi if $data31 != 6 then @@ -900,7 +889,7 @@ endi if $data32 != 7 then return -1 endi -if $data40 != @18-09-17 10:10:00.000@ then +if $data40 != @18-09-17 10:10:00.007@ then return -1 endi if $data41 != 7 then @@ -909,7 +898,7 @@ endi if $data42 != 7 then return -1 endi -if $data50 != @18-09-17 10:20:00.000@ then +if $data50 != @18-09-17 10:20:00.007@ then return -1 endi if $data51 != 8 then @@ -923,7 +912,7 @@ sql select top(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu group by t1 orde if $rows != 6 then return -1 endi -if $data00 != @18-09-17 10:00:00.000@ then +if $data00 != @18-09-17 10:00:00.001@ then return -1 endi if $data01 != 6 then @@ -932,7 +921,7 @@ endi if $data02 != 1 then return -1 endi -if $data10 != @18-09-17 10:10:00.000@ then +if $data10 != @18-09-17 10:10:00.001@ then return -1 endi if $data11 != 7 then @@ -941,7 +930,7 @@ endi if $data12 != 1 then return -1 endi -if $data20 != @18-09-17 10:20:00.000@ then +if $data20 != @18-09-17 10:20:00.001@ then return -1 endi if $data21 != 8 then @@ -950,7 +939,7 @@ endi if $data22 != 1 then return -1 endi -if $data30 != @18-09-17 10:00:00.000@ then +if $data30 != @18-09-17 10:00:00.002@ then return -1 endi if $data31 != 6 then @@ -959,7 +948,7 @@ endi if $data32 != 2 then return -1 endi -if $data40 != @18-09-17 10:10:00.000@ then +if $data40 != @18-09-17 10:10:00.002@ then return -1 endi if $data41 != 7 then @@ -968,7 +957,7 @@ endi if $data42 != 2 then return -1 endi -if $data50 != @18-09-17 10:20:00.000@ then +if $data50 != @18-09-17 10:20:00.002@ then return -1 endi if $data51 != 8 then @@ -982,7 +971,7 @@ sql select top(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu group by t1 orde if $rows != 6 then return -1 endi -if $data00 != @18-09-17 10:20:00.000@ then +if $data00 != @18-09-17 10:20:00.001@ then return -1 endi if $data01 != 8 then @@ -991,7 +980,7 @@ endi if $data02 != 1 then return -1 endi -if $data10 != @18-09-17 10:10:00.000@ then +if $data10 != @18-09-17 10:10:00.001@ then return -1 endi if $data11 != 7 then @@ -1000,7 +989,7 @@ endi if $data12 != 1 then return -1 endi -if $data20 != @18-09-17 10:00:00.000@ then +if $data20 != @18-09-17 10:00:00.001@ then return -1 endi if $data21 != 6 then @@ -1009,7 +998,7 @@ endi if $data22 != 1 then return -1 endi -if $data30 != @18-09-17 10:20:00.000@ then +if $data30 != @18-09-17 10:20:00.002@ then return -1 endi if $data31 != 8 then @@ -1018,7 +1007,7 @@ endi if $data32 != 2 then return -1 endi -if $data40 != @18-09-17 10:10:00.000@ then +if $data40 != @18-09-17 10:10:00.002@ then return -1 endi if $data41 != 7 then @@ -1027,7 +1016,7 @@ endi if $data42 != 2 then return -1 endi -if $data50 != @18-09-17 10:00:00.000@ then +if $data50 != @18-09-17 10:00:00.002@ then return -1 endi if $data51 != 6 then @@ -1052,9 +1041,9 @@ sql select bottom(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu order by ts d if $rows != 3 then return -1 endi -if $data00 != @18-09-17 09:00:00.000@ then - return -1 -endi +#if $data00 != @18-09-17 09:00:00.000@ then +# return -1 +#endi if $data01 != 0 then return -1 endi @@ -1063,9 +1052,9 @@ sql select bottom(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu order by ts a if $rows != 3 then return -1 endi -if $data00 != @18-09-17 09:00:00.000@ then - return -1 -endi +#if $data00 != @18-09-17 09:00:00.000@ then +# return -1 +#endi if $data01 != 0 then return -1 endi @@ -1074,54 +1063,54 @@ sql select bottom(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu group by t1 o if $rows != 6 then return -1 endi -if $data00 != @18-09-17 09:30:00.000@ then - return -1 -endi +#if $data00 != @18-09-17 09:30:00.000@ then +# return -1 +#endi if $data01 != 3 then return -1 endi if $data02 != 8 then return -1 endi -if $data10 != @18-09-17 09:20:00.000@ then - return -1 -endi +#if $data10 != @18-09-17 09:20:00.000@ then +# return -1 +#endi if $data11 != 2 then return -1 endi if $data12 != 8 then return -1 endi -if $data20 != @18-09-17 09:10:00.000@ then - return -1 -endi +#if $data20 != @18-09-17 09:10:00.000@ then +# return -1 +#endi if $data21 != 1 then return -1 endi if $data22 != 8 then return -1 endi -if $data30 != @18-09-17 09:30:00.000@ then - return -1 -endi +#if $data30 != @18-09-17 09:30:00.000@ then +# return -1 +#endi if $data31 != 3 then return -1 endi if $data32 != 7 then return -1 endi -if $data40 != @18-09-17 09:20:00.000@ then - return -1 -endi +#if $data40 != @18-09-17 09:20:00.000@ then +# return -1 +#endi if $data41 != 2 then return -1 endi if $data42 != 7 then return -1 endi -if $data50 != @18-09-17 09:10:00.000@ then - return -1 -endi +#if $data50 != @18-09-17 09:10:00.000@ then +# return -1 +#endi if $data51 != 1 then return -1 endi @@ -1133,54 +1122,54 @@ sql select bottom(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu group by t1 o if $rows != 6 then return -1 endi -if $data00 != @18-09-17 09:30:00.000@ then - return -1 -endi +#if $data00 != @18-09-17 09:30:00.000@ then +# return -1 +#endi if $data01 != 3 then return -1 endi if $data02 != 1 then return -1 endi -if $data10 != @18-09-17 09:20:00.000@ then - return -1 -endi +#if $data10 != @18-09-17 09:20:00.000@ then +# return -1 +#endi if $data11 != 2 then return -1 endi if $data12 != 1 then return -1 endi -if $data20 != @18-09-17 09:10:00.000@ then - return -1 -endi +#if $data20 != @18-09-17 09:10:00.000@ then +# return -1 +#endi if $data21 != 1 then return -1 endi if $data22 != 1 then return -1 endi -if $data30 != @18-09-17 09:30:00.000@ then - return -1 -endi +#if $data30 != @18-09-17 09:30:00.000@ then +# return -1 +#endi if $data31 != 3 then return -1 endi if $data32 != 2 then return -1 endi -if $data40 != @18-09-17 09:20:00.000@ then - return -1 -endi +#if $data40 != @18-09-17 09:20:00.000@ then +# return -1 +#endi if $data41 != 2 then return -1 endi if $data42 != 2 then return -1 endi -if $data50 != @18-09-17 09:10:00.000@ then - return -1 -endi +#if $data50 != @18-09-17 09:10:00.000@ then +# return -1 +#endi if $data51 != 1 then return -1 endi @@ -1192,54 +1181,54 @@ sql select bottom(c1, 5) from $stb where ts >= $ts0 and ts <= $tsu group by t1 o if $rows != 6 then return -1 endi -if $data00 != @18-09-17 09:30:00.000@ then - return -1 -endi +#if $data00 != @18-09-17 09:30:00.000@ then +# return -1 +#endi if $data01 != 3 then return -1 endi if $data02 != 1 then return -1 endi -if $data10 != @18-09-17 09:20:00.000@ then - return -1 -endi +#if $data10 != @18-09-17 09:20:00.000@ then +# return -1 +#endi if $data11 != 2 then return -1 endi if $data12 != 1 then return -1 endi -if $data20 != @18-09-17 09:10:00.000@ then - return -1 -endi +#if $data20 != @18-09-17 09:10:00.000@ then +# return -1 +#endi if $data21 != 1 then return -1 endi if $data22 != 1 then return -1 endi -if $data30 != @18-09-17 09:30:00.000@ then - return -1 -endi +#if $data30 != @18-09-17 09:30:00.000@ then +# return -1 +#endi if $data31 != 3 then return -1 endi if $data32 != 2 then return -1 endi -if $data40 != @18-09-17 09:20:00.000@ then - return -1 -endi +#if $data40 != @18-09-17 09:20:00.000@ then +# return -1 +#endi if $data41 != 2 then return -1 endi if $data42 != 2 then return -1 endi -if $data50 != @18-09-17 09:10:00.000@ then - return -1 -endi +#if $data50 != @18-09-17 09:10:00.000@ then +# return -1 +#endi if $data51 != 1 then return -1 endi diff --git a/tests/script/general/parser/limit_tb.sim b/tests/script/general/parser/limit_tb.sim index b917627fdf6499fe2dacf948af7badb2b5de43d7..45f55412086c2155a2c7c924603c60d5081a643f 100644 --- a/tests/script/general/parser/limit_tb.sim +++ b/tests/script/general/parser/limit_tb.sim @@ -327,22 +327,22 @@ sql select twa(c1), twa(c2), twa(c3), twa(c4), twa(c5), twa(c6) from $tb where t if $rows != 1 then return -1 endi -if $data00 != 4.000000000 then +if $data00 != 4.500000000 then return -1 endi -if $data01 != 4.000000000 then +if $data01 != 4.500000000 then return -1 endi -if $data02 != 4.000000000 then +if $data02 != 4.500000000 then return -1 endi -if $data03 != 4.000000000 then +if $data03 != 4.500000000 then return -1 endi -if $data04 != 4.000000000 then +if $data04 != 4.500000000 then return -1 endi -if $data05 != 4.000000000 then +if $data05 != 4.500000000 then return -1 endi @@ -690,13 +690,13 @@ sql select twa(c1), twa(c2), twa(c3), twa(c4), twa(c5), twa(c6) from $tb where t if $rows != 1 then return -1 endi -if $data00 != 4.000000000 then +if $data00 != 4.500000000 then return -1 endi -if $data02 != 4.000000000 then +if $data02 != 4.500000000 then return -1 endi -if $data05 != 4.000000000 then +if $data05 != 4.500000000 then return -1 endi diff --git a/tests/script/general/parser/where.sim b/tests/script/general/parser/where.sim index 8e17220b5b38f995c6dc7e662130f7250157aa44..c5b600b514be8047ab0292f23d69c9ddd6ef8399 100644 --- a/tests/script/general/parser/where.sim +++ b/tests/script/general/parser/where.sim @@ -131,7 +131,6 @@ if $data00 != $rowNum then return -1 endi - ## like sql_error select * from $mt where c1 like 1 #sql_error select * from $mt where t1 like 1 @@ -178,7 +177,8 @@ sql create table wh_mt2_tb1 using wh_mt2 tags ('wh_mt2_tb1') # 2019-01-01 09:00:00.000 1546304400000 # 2019-01-01 09:10:00.000 1546305000000 sql insert into wh_mt2_tb1 values ('2019-01-01 00:00:00.000', '2019-01-01 09:00:00.000', 'binary10', 'nchar10') -sql insert into wh_mt2_tb1 values ('2019-01-01 00:10:00.000', '2019-01-01 09:10:00.000', 'binary10', 'nchar10') +sql insert into wh_mt2_tb1 values ('2019-01-01 00:10:00.000', '2019-01-01 09:10:00.000', 'binary10', 'nchar10') + sql select * from wh_mt2_tb1 where c1 > 1546304400000 if $rows != 1 then return -1 diff --git a/tests/script/general/stable/values.sim b/tests/script/general/stable/values.sim index d0c178385103f8fb719ed8d7ec35f55bd9023d22..51488aabef8918b957ca63fd29823b14f772b675 100644 --- a/tests/script/general/stable/values.sim +++ b/tests/script/general/stable/values.sim @@ -18,16 +18,25 @@ sql create table vdb0.vtb00 using vdb0.mt tags( 0 ) sql create table vdb0.vtb01 using vdb0.mt tags( 0 ) sql create database vdb1 -sql create table vdb1.vtb10 using vdb0.mt tags( 1 ) -sql create table vdb1.vtb11 using vdb0.mt tags( 1 ) +sql create table vdb1.mt (ts timestamp, tbcol int) TAGS(tgcol int) +sql_error create table vdb1.vtb10 using vdb0.mt tags( 1 ) +sql_error create table vdb1.vtb11 using vdb0.mt tags( 1 ) +sql create table vdb1.vtb10 using vdb1.mt tags( 1 ) +sql create table vdb1.vtb11 using vdb1.mt tags( 1 ) sql create database vdb2 -sql create table vdb2.vtb20 using vdb0.mt tags( 2 ) -sql create table vdb2.vtb21 using vdb0.mt tags( 2 ) +sql create table vdb2.mt (ts timestamp, tbcol int) TAGS(tgcol int) +sql_error create table vdb2.vtb20 using vdb0.mt tags( 2 ) +sql_error create table vdb2.vtb21 using vdb0.mt tags( 2 ) +sql create table vdb2.vtb20 using vdb2.mt tags( 2 ) +sql create table vdb2.vtb21 using vdb2.mt tags( 2 ) sql create database vdb3 -sql create table vdb3.vtb30 using vdb0.mt tags( 3 ) -sql create table vdb3.vtb31 using vdb0.mt tags( 3 ) +sql create table vdb3.mt (ts timestamp, tbcol int) TAGS(tgcol int) +sql_error create table vdb3.vtb20 using vdb0.mt tags( 2 ) +sql_error create table vdb3.vtb21 using vdb0.mt tags( 2 ) +sql create table vdb3.vtb30 using vdb3.mt tags( 3 ) +sql create table vdb3.vtb31 using vdb3.mt tags( 3 ) print =============== step2 sql insert into vdb0.vtb00 values (1519833600000 , 10) (1519833600001, 20) (1519833600002, 30) @@ -40,7 +49,7 @@ sql insert into vdb3.vtb30 values (1519833600000 , 13) (1519833600001, 23) (1519 sql insert into vdb3.vtb31 values (1519833600000 , 13) (1519833600001, 23) (1519833600002, 33) sql select * from vdb0.mt -if $rows != 24 then +if $rows != 6 then return -1 endi @@ -55,7 +64,7 @@ sql insert into vdb3.vtb30 values (1519833600003 , 43) (1519833600005, 53) (1519 sql insert into vdb3.vtb31 values (1519833600003 , 43) (1519833600005, 53) (1519833600004, 63) sql select * from vdb0.mt -if $rows != 48 then +if $rows != 12 then return -1 endi @@ -66,7 +75,7 @@ sql insert into vdb2.vtb20 values(1519833600006, 62) (1519833600007, 72) vdb2.vt sql insert into vdb3.vtb30 values(1519833600006, 63) (1519833600007, 73) vdb3.vtb31 values(1519833600006, 63) (1519833600007, 73) sql select * from vdb0.mt -if $rows != 64 then +if $rows != 16 then return -1 endi @@ -77,7 +86,7 @@ sql insert into vdb2.vtb20 values(1519833600008, 82) (1519833600007, 72) vdb2.vt sql insert into vdb3.vtb30 values(1519833600008, 83) (1519833600007, 73) vdb3.vtb31 values(1519833600006, 83) (1519833600007, 73) sql select * from vdb0.mt -if $rows != 68 then +if $rows != 17 then return -1 endi @@ -87,7 +96,7 @@ sql insert into vdb0.vtb01 values(1519833600009, 90) (1519833600010, 100) vdb1.v sql select * from vdb0.mt -if $rows != 84 then +if $rows != 21 then return -1 endi @@ -97,7 +106,7 @@ sql insert into vdb0.vtb01 values(1519833600012, 120) (1519833600011, 110) vdb1. sql select * from vdb0.mt -if $rows != 100 then +if $rows != 25 then return -1 endi diff --git a/tests/script/general/wal/sync.sim b/tests/script/general/wal/sync.sim index abaf22f91921e5bcc8effbe6fc1c66766ff92a3f..c6f7402b874115811e6422894c612c341ea5256f 100644 --- a/tests/script/general/wal/sync.sim +++ b/tests/script/general/wal/sync.sim @@ -82,6 +82,7 @@ restful d1 table_rest 1591772800 30000 restful d1 table_rest 1591872800 30000 restful d1 table_rest 1591972800 30000 +sleep 1000 sql select * from table_rest; print rows: $rows if $rows != 300000 then diff --git a/tests/script/sh/deploy.sh b/tests/script/sh/deploy.sh index e26778e86b9876ab6e6f47f4b41207ea35cafbf5..cd2f3772eb11a0c5d3a7a69fcb3d62162eadf14c 100755 --- a/tests/script/sh/deploy.sh +++ b/tests/script/sh/deploy.sh @@ -120,7 +120,7 @@ echo "cDebugFlag 143" >> $TAOS_CFG echo "jnidebugFlag 143" >> $TAOS_CFG echo "odbcdebugFlag 143" >> $TAOS_CFG echo "httpDebugFlag 143" >> $TAOS_CFG -echo "monitorDebugFlag 143" >> $TAOS_CFG +echo "monDebugFlag 143" >> $TAOS_CFG echo "mqttDebugFlag 143" >> $TAOS_CFG echo "qdebugFlag 143" >> $TAOS_CFG echo "rpcDebugFlag 143" >> $TAOS_CFG diff --git a/tests/script/tmp/mnodes.sim b/tests/script/tmp/mnodes.sim index de02ae741b75d4764f280d4aca5eb9cb139aa7ff..e11140028d474b131c472d6c0ff0f9a56587e3a9 100644 --- a/tests/script/tmp/mnodes.sim +++ b/tests/script/tmp/mnodes.sim @@ -20,6 +20,10 @@ system sh/cfg.sh -n dnode1 -c maxTablesPerVnode -v 20000 system sh/cfg.sh -n dnode2 -c maxTablesPerVnode -v 20000 system sh/cfg.sh -n dnode3 -c maxTablesPerVnode -v 20000 +system sh/cfg.sh -n dnode1 -c minTablesPerVnode -v 1000 +system sh/cfg.sh -n dnode2 -c minTablesPerVnode -v 1000 +system sh/cfg.sh -n dnode3 -c minTablesPerVnode -v 1000 + system sh/cfg.sh -n dnode1 -c maxVgroupsPerDb -v 20 system sh/cfg.sh -n dnode2 -c maxVgroupsPerDb -v 20 system sh/cfg.sh -n dnode3 -c maxVgroupsPerDb -v 20 diff --git a/tests/test-all.sh b/tests/test-all.sh index 5897978bce6ab68255b07dd85b3c310526e90e15..14b649eddfe92e8c557f81d27a0d1b95b35602c3 100755 --- a/tests/test-all.sh +++ b/tests/test-all.sh @@ -137,6 +137,12 @@ if [ "$2" != "sim" ]; then elif [ "$1" == "pytest" ]; then echo "### run Python full test ###" runPyCaseOneByOne fulltest.sh + elif [ "$1" == "p1" ]; then + echo "### run Python_1 test ###" + runPyCaseOneByOne pytest_1.sh + elif [ "$1" == "p2" ]; then + echo "### run Python_2 test ###" + runPyCaseOneByOne pytest_2.sh elif [ "$1" == "b2" ] || [ "$1" == "b3" ]; then exit $(($totalFailed + $totalPyFailed)) elif [ "$1" == "smoke" ] || [ -z "$1" ]; then diff --git a/tests/test/c/CMakeLists.txt b/tests/test/c/CMakeLists.txt index 26aa20e647cc166ac991e64bc18511e9b504b335..11480a8ba2f7c2a78b8628e3edabe6e0eb122410 100644 --- a/tests/test/c/CMakeLists.txt +++ b/tests/test/c/CMakeLists.txt @@ -31,8 +31,8 @@ IF (TD_LINUX) #add_executable(createTablePerformance createTablePerformance.c) #target_link_libraries(createTablePerformance taos_static tutil common pthread) - #add_executable(createNormalTable createNormalTable.c) - #target_link_libraries(createNormalTable taos_static tutil common pthread) + add_executable(createNormalTable createNormalTable.c) + target_link_libraries(createNormalTable taos_static tutil common pthread) #add_executable(queryPerformance queryPerformance.c) #target_link_libraries(queryPerformance taos_static tutil common pthread) @@ -45,5 +45,8 @@ IF (TD_LINUX) #add_executable(invalidTableId invalidTableId.c) #target_link_libraries(invalidTableId taos_static tutil common pthread) + + add_executable(hashIterator hashIterator.c) + target_link_libraries(hashIterator taos_static tutil common pthread) ENDIF() diff --git a/tests/test/c/hashIterator.c b/tests/test/c/hashIterator.c new file mode 100644 index 0000000000000000000000000000000000000000..cbd8a0895ebedd4fb0e3c448e4b6865fc0d8a454 --- /dev/null +++ b/tests/test/c/hashIterator.c @@ -0,0 +1,72 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#define _DEFAULT_SOURCE +#include "os.h" +#include "taos.h" +#include "tulog.h" +#include "tutil.h" +#include "hash.h" + +typedef struct HashTestRow { + int32_t keySize; + char key[100]; +} HashTestRow; + +int main(int argc, char *argv[]) { + _hash_fn_t hashFp = taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY); + void * hashHandle = taosHashInit(100, hashFp, true, HASH_ENTRY_LOCK); + + pPrint("insert 3 rows to hash"); + for (int32_t t = 0; t < 3; ++t) { + HashTestRow row = {0}; + row.keySize = sprintf(row.key, "0.db.st%d", t); + + taosHashPut(hashHandle, row.key, row.keySize, &row, sizeof(HashTestRow)); + } + + pPrint("start iterator"); + HashTestRow *row = taosHashIterate(hashHandle, NULL); + while (row) { + pPrint("drop key:%s", row->key); + taosHashRemove(hashHandle, row->key, row->keySize); + + pPrint("get rows from hash"); + for (int32_t t = 0; t < 3; ++t) { + HashTestRow r = {0}; + r.keySize = sprintf(r.key, "0.db.st%d", t); + + void *result = taosHashGet(hashHandle, r.key, r.keySize); + pPrint("get key:%s result:%p", r.key, result); + } + + //Before getting the next iterator, the object just deleted can be obtained + row = taosHashIterate(hashHandle, row); + } + + pPrint("stop iterator"); + taosHashCancelIterate(hashHandle, row); + + pPrint("get rows from hash"); + for (int32_t t = 0; t < 3; ++t) { + HashTestRow r = {0}; + r.keySize = sprintf(r.key, "0.db.st%d", t); + + void *result = taosHashGet(hashHandle, r.key, r.keySize); + pPrint("get key:%s result:%p", r.key, result); + } + + return 0; +} \ No newline at end of file