diff --git a/documentation/webdocs/markdowndocs/administrator-ch.md b/documentation/webdocs/markdowndocs/administrator-ch.md index 44b3ad46712019870be6fefb5234611f55f6e03e..79388a2edb9404a0f7b31b9182eb5ce2cb0d52be 100644 --- a/documentation/webdocs/markdowndocs/administrator-ch.md +++ b/documentation/webdocs/markdowndocs/administrator-ch.md @@ -87,6 +87,7 @@ TDengine系统后台服务由taosd提供，可以在配置文件taos.cfg里修 - httpPort: RESTful服务使用的端口号，所有的HTTP请求（TCP）都需要向该接口发起查询/写入请求。 - dataDir: 数据文件目录，所有的数据文件都将写入该目录。默认值：/var/lib/taos。 - logDir：日志文件目录，客户端和服务器的运行日志文件将写入该目录。默认值：/var/log/taos。 +- tempDir：临时文件目录，客户端和服务器的临时文件（主要是查询时用于保存中间结果的问题）将写入该目录。 默认值：Linux下为 /tmp/，Windows下为环境变量 tmp 或 temp 指向的目录。 - arbitrator：系统中裁决器的end point, 缺省值为空。 - role：dnode的可选角色。0-any; 既可作为mnode，也可分配vnode；1-mgmt;只能作为mnode，不能分配vnode；2-dnode;不能作为mnode，只能分配vnode - debugFlag：运行日志开关。131（输出错误和警告日志），135（ 输出错误、警告和调试日志），143（ 输出错误、警告、调试和跟踪日志）。默认值：131或135（不同模块有不同的默认值）。 diff --git a/documentation20/webdocs/markdowndocs/TAOS SQL-ch.md b/documentation20/webdocs/markdowndocs/TAOS SQL-ch.md index 4082d72f112e8ff62d2e8b2c8b15391dd6f39d8a..191af56bbdf8bf578ca2713f1693e668fee4c2b4 100644 --- a/documentation20/webdocs/markdowndocs/TAOS SQL-ch.md +++ b/documentation20/webdocs/markdowndocs/TAOS SQL-ch.md @@ -844,7 +844,7 @@ TDengine支持针对数据的聚合查询。提供支持的聚合和选择函数 - **PERCENTILE** ```mysql - SELECT PERCENTILE(field_name, P) FROM { tb_name | stb_name } [WHERE clause]; + SELECT PERCENTILE(field_name, P) FROM { tb_name } [WHERE clause]; ``` 功能说明：统计表中某列的值百分比分位数。 返回结果数据类型： 双精度浮点数Double。 diff --git a/documentation20/webdocs/markdowndocs/administrator-ch.md b/documentation20/webdocs/markdowndocs/administrator-ch.md index 4b274e05e6e392879fc887ec1137968270a8e4b8..1e36e6c5e624912e0c70edf5514ce6a112785f7b 100644 --- a/documentation20/webdocs/markdowndocs/administrator-ch.md +++ b/documentation20/webdocs/markdowndocs/administrator-ch.md @@ -35,7 +35,7 @@ TDengine相对于通用数据库，有超高的压缩比，在绝大多数场景 Raw DataSize = numOfTables * rowSizePerTable * rowsPerTable ``` -示例：1000万台智能电表，每台电表每15分钟采集一次数据，每次采集的数据128字节，那么一年的原始数据量是：10000000\*128\*24\*60/15*365 = 44851T。TDengine大概需要消耗44851/5=8970T, 8.9P空间。 +示例：1000万台智能电表，每台电表每15分钟采集一次数据，每次采集的数据128字节，那么一年的原始数据量是：10000000\*128\*24\*60/15*365 = 44.8512T。TDengine大概需要消耗44.851/5=8.97024T空间。 用户可以通过参数keep，设置数据在磁盘中的最大保存时长。为进一步减少存储成本，TDengine还提供多级存储，最冷的数据可以存放在最廉价的存储介质上，应用的访问不用做任何调整，只是读取速度降低了。 diff --git a/documentation20/webdocs/markdowndocs/connector-ch.md b/documentation20/webdocs/markdowndocs/connector-ch.md index 0e29b324871e61d184fde17710c863f06b60bdcc..c5a955f43fcc30e70d8aac3919433c2e767f76ba 100644 --- a/documentation20/webdocs/markdowndocs/connector-ch.md +++ b/documentation20/webdocs/markdowndocs/connector-ch.md @@ -616,6 +616,43 @@ HTTP请求URL采用`sqlutc`时，返回结果集的时间戳将采用UTC时间 - httpEnableCompress: 是否支持压缩，默认不支持，目前TDengine仅支持gzip压缩格式 - httpDebugFlag: 日志开关，131：仅错误和报警信息，135：调试信息，143：非常详细的调试信息，默认131 +## CSharp Connector + +在Windows系统上，C#应用程序可以使用TDengine的原生C接口来执行所有数据库操作，后续版本将提供ORM（dapper）框架驱动。 + +#### 安装TDengine客户端 + +C#连接器需要使用`libtaos.so`和`taos.h`。因此，在使用C#连接器之前，需在程序运行的Windows环境安装TDengine的Windows客户端，以便获得相关驱动文件。 + +安装完成后，在文件夹`C:/TDengine/examples/C#`中，将会看到两个文件 + +- TDengineDriver.cs 调用taos.dll文件的Native C方法 +- TDengineTest.cs 参考程序示例 + +在文件夹`C:\Windows\System32`，将会看到`taos.dll`文件 + +#### 使用方法 + +- 将C#接口文件TDengineDriver.cs加入到应用程序所在.NET项目中 +- 参考TDengineTest.cs来定义数据库连接参数，及执行数据插入、查询等操作的方法 +- 因为C#接口需要用到`taos.dll`文件，用户可以将`taos.dll`文件加入.NET解决方案中 + +#### 注意事项 + +- `taos.dll`文件使用x64平台编译，所以.NET项目在生成.exe文件时，“解决方案”/“项目”的“平台”请均选择“x64”。 +- 此.NET接口目前已经在Visual Studio 2013/2015/2017中验证过，其它VS版本尚待验证。 + +#### 第三方驱动 + +Maikebing.Data.Taos是一个TDengine的ADO.Net提供器，支持linux，windows。该开发包由热心贡献者`麦壳饼@@maikebing`提供，具体请参考 + +``` +//接口下载 +https://github.com/maikebing/Maikebing.EntityFrameworkCore.Taos +//用法说明 +https://www.taosdata.com/blog/2020/11/02/1901.html +``` + ## Go Connector diff --git a/documentation20/webdocs/markdowndocs/faq-ch.md b/documentation20/webdocs/markdowndocs/faq-ch.md index 80deb889ef0eee8e9b47f86b5e58a76c6c070d5b..757b6d9929b68a90e7bc558b553233f24b09ba17 100644 --- a/documentation20/webdocs/markdowndocs/faq-ch.md +++ b/documentation20/webdocs/markdowndocs/faq-ch.md @@ -38,9 +38,9 @@ 6. 检查防火墙设置，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/lib/taos*里, 并且*/usr/local/lib/taos*在系统库函数搜索路径*LD_LIBRARY_PATH*里 +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*driver/c/taos.dll*在你的系统搜索目录里 (建议*taos.dll*放在目录 *C:\Windows\System32*) +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} ` diff --git a/packaging/cfg/taos.cfg b/packaging/cfg/taos.cfg index 974b2b05c12f9fd5a482ef0072a9dacc2d1f09dc..ca88bca3c863dcfd4a95497fc572499dc93f31f9 100644 --- a/packaging/cfg/taos.cfg +++ b/packaging/cfg/taos.cfg @@ -20,6 +20,9 @@ # data file's directory # dataDir /var/lib/taos +# temporary file's directory +# tempDir /tmp/ + # the arbitrator's fully qualified domain name (FQDN) for TDengine system, for cluster only # arbitrator arbitrator_hostname:6042 @@ -256,3 +259,5 @@ # maximum display width of binary and nchar fields in the shell. The parts exceeding this limit will be hidden # maxBinaryDisplayWidth 30 +# enable/disable telemetry reporting +# telemetryReporting 1 \ No newline at end of file diff --git a/packaging/tools/post.sh b/packaging/tools/post.sh index d91daaa5c44488e34dea7ec2ddec0863699446f2..569f316ff3d3edf0af49d9e132044a348b0a6394 100755 --- a/packaging/tools/post.sh +++ b/packaging/tools/post.sh @@ -121,8 +121,11 @@ function install_config() { echo -e -n "${GREEN}Enter FQDN:port (like h1.taosdata.com:6030) of an existing TDengine cluster node to join${NC}" echo echo -e -n "${GREEN}OR leave it blank to build one${NC}:" - read firstEp - while true; do + #read firstEp + if exec < /dev/tty; then + read firstEp; + fi + while true; do if [ ! -z "$firstEp" ]; then # check the format of the firstEp #if [[ $firstEp == $FQDN_PATTERN ]]; then diff --git a/src/balance/src/balance.c b/src/balance/src/balance.c index 0e9bb85b25defd169fea8711d3e0b40304500de4..4c687cb134141e38aef1cb9b17dcb45e89278e1b 100644 --- a/src/balance/src/balance.c +++ b/src/balance/src/balance.c @@ -490,7 +490,7 @@ static bool balanceMontiorDropping() { if (pDnode->status == TAOS_DN_STATUS_OFFLINE) { if (pDnode->lastAccess + tsOfflineThreshold > tsAccessSquence) continue; - if (strcmp(pDnode->dnodeEp, dnodeGetMnodeMasterEp()) == 0) continue; + if (dnodeIsMasterEp(pDnode->dnodeEp)) continue; if (mnodeGetDnodesNum() <= 1) continue; mLInfo("dnode:%d, set to removing state for it offline:%d seconds", pDnode->dnodeId, diff --git a/src/client/inc/tscLocalMerge.h b/src/client/inc/tscLocalMerge.h index 5baa66a9e0229f35c431cea7a0d2dbb9e2ffb0e2..0af8c8b5768c844954a072521e67a3cb6e914ac3 100644 --- a/src/client/inc/tscLocalMerge.h +++ b/src/client/inc/tscLocalMerge.h @@ -64,9 +64,8 @@ typedef struct SLocalReducer { SColumnModel * resColModel; tExtMemBuffer ** pExtMemBuffer; // disk-based buffer SFillInfo* pFillInfo; // interpolation support structure - char * pFinalRes; // result data after interpo - tFilePage * discardData; - SResultInfo * pResInfo; + char* pFinalRes; // result data after interpo + tFilePage* discardData; bool discard; int32_t offset; // limit offset value bool orderPrjOnSTable; // projection query on stable diff --git a/src/client/inc/tscSubquery.h b/src/client/inc/tscSubquery.h index bc01de110345e4c90cf5c15d3d7f6b010cb7308d..3406dcd8589e149f2a62040d1271fbfd40caaac7 100644 --- a/src/client/inc/tscSubquery.h +++ b/src/client/inc/tscSubquery.h @@ -23,7 +23,7 @@ extern "C" { #include "tscUtil.h" #include "tsclient.h" -void tscFetchDatablockFromSubquery(SSqlObj* pSql); +void tscFetchDatablockForSubquery(SSqlObj* pSql); void tscSetupOutputColumnIndex(SSqlObj* pSql); void tscJoinQueryCallback(void* param, TAOS_RES* tres, int code); diff --git a/src/client/inc/tscUtil.h b/src/client/inc/tscUtil.h index 39299d3308df2d2c80c238c06ebd096eb73d2f0a..6628e3087456c52b2daf1b9ef84efa9577ad4f36 100644 --- a/src/client/inc/tscUtil.h +++ b/src/client/inc/tscUtil.h @@ -87,8 +87,8 @@ typedef struct SJoinSupporter { } SJoinSupporter; typedef struct SVgroupTableInfo { - SCMVgroupInfo vgInfo; - SArray* itemList; //SArray + SVgroupInfo vgInfo; + SArray* itemList; //SArray } SVgroupTableInfo; static FORCE_INLINE SQueryInfo* tscGetQueryInfoDetail(SSqlCmd* pCmd, int32_t subClauseIndex) { @@ -228,6 +228,7 @@ void tscClearSubqueryInfo(SSqlCmd* pCmd); void tscFreeVgroupTableInfo(SArray* pVgroupTables); SArray* tscVgroupTableInfoClone(SArray* pVgroupTables); void tscRemoveVgroupTableGroup(SArray* pVgroupTable, int32_t index); +void tscVgroupTableCopy(SVgroupTableInfo* info, SVgroupTableInfo* pInfo); int tscGetSTableVgroupInfo(SSqlObj* pSql, int32_t clauseIndex); int tscGetTableMeta(SSqlObj* pSql, STableMetaInfo* pTableMetaInfo); @@ -238,7 +239,7 @@ void tscDoQuery(SSqlObj* pSql); SVgroupsInfo* tscVgroupInfoClone(SVgroupsInfo *pInfo); void* tscVgroupInfoClear(SVgroupsInfo *pInfo); -void tscSCMVgroupInfoCopy(SCMVgroupInfo* dst, const SCMVgroupInfo* src); +void tscSVgroupInfoCopy(SVgroupInfo* dst, const SVgroupInfo* src); /** * The create object function must be successful expect for the out of memory issue. * diff --git a/src/client/inc/tschemautil.h b/src/client/inc/tschemautil.h index e5bdcecfa8c006f75575d5b197ed29c9cb65fbc9..f6dc45398f35c38598f3f3132b2f6f5601a4ed68 100644 --- a/src/client/inc/tschemautil.h +++ b/src/client/inc/tschemautil.h @@ -107,9 +107,6 @@ SSchema tscGetTbnameColumnSchema(); */ STableMeta* tscCreateTableMetaFromMsg(STableMetaMsg* pTableMetaMsg, size_t* size); -//todo tags value as well as the table id structure needs refactor -char *tsGetTagsValue(STableMeta *pMeta); - #ifdef __cplusplus } #endif diff --git a/src/client/inc/tsclient.h b/src/client/inc/tsclient.h index bbd029d6759d5b77ef00c554d92be0b801d58af9..1c2821638bd24d67c51985a9e67da26e83e768b8 100644 --- a/src/client/inc/tsclient.h +++ b/src/client/inc/tsclient.h @@ -90,12 +90,12 @@ typedef struct STableComInfo { int32_t rowSize; } STableComInfo; -typedef struct SCMCorVgroupInfo { - int32_t version; - int8_t inUse; - int8_t numOfEps; - SEpAddr1 epAddr[TSDB_MAX_REPLICA]; -} SCMCorVgroupInfo; +typedef struct SCorVgroupInfo { + int32_t version; + int8_t inUse; + int8_t numOfEps; + SEpAddr1 epAddr[TSDB_MAX_REPLICA]; +} SCorVgroupInfo; typedef struct STableMeta { STableComInfo tableInfo; @@ -103,8 +103,8 @@ typedef struct STableMeta { int16_t sversion; int16_t tversion; char sTableId[TSDB_TABLE_FNAME_LEN]; - SCMVgroupInfo vgroupInfo; - SCMCorVgroupInfo corVgroupInfo; + SVgroupInfo vgroupInfo; + SCorVgroupInfo corVgroupInfo; STableId id; SSchema schema[]; // if the table is TSDB_CHILD_TABLE, schema is acquired by super table meta info } STableMeta; @@ -339,9 +339,9 @@ typedef struct STscObj { } STscObj; typedef struct SSubqueryState { - int32_t numOfRemain; // the number of remain unfinished subquery - int32_t numOfSub; // the number of total sub-queries - uint64_t numOfRetrievedRows; // total number of points in this query + int32_t numOfRemain; // the number of remain unfinished subquery + int32_t numOfSub; // the number of total sub-queries + uint64_t numOfRetrievedRows; // total number of points in this query } SSubqueryState; typedef struct SSqlObj { @@ -431,14 +431,6 @@ void tscResetSqlCmdObj(SSqlCmd *pCmd, bool removeFromCache); */ void tscFreeSqlResult(SSqlObj *pSql); -/** - * only free part of resources allocated during query. - * TODO remove it later - * Note: this function is multi-thread safe. - * @param pObj - */ -void tscPartiallyFreeSqlObj(SSqlObj *pSql); - /** * free sql object, release allocated resource * @param pObj @@ -523,7 +515,6 @@ extern SRpcCorEpSet tscMgmtEpSet; extern int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo); -int32_t tscCompareTidTags(const void* p1, const void* p2); void tscBuildVgroupTableInfo(SSqlObj* pSql, STableMetaInfo* pTableMetaInfo, SArray* tables); #ifdef __cplusplus diff --git a/src/client/src/tscAsync.c b/src/client/src/tscAsync.c index a85a0ea570f6a3bb7f57c916ff43ac290254fed4..e70cd11fbe03b498d2caad93b086cfc33adf5669 100644 --- a/src/client/src/tscAsync.c +++ b/src/client/src/tscAsync.c @@ -176,7 +176,7 @@ static void tscProcessAsyncRetrieveImpl(void *param, TAOS_RES *tres, int numOfRo } if (pCmd->command == TSDB_SQL_TABLE_JOIN_RETRIEVE) { - tscFetchDatablockFromSubquery(pSql); + tscFetchDatablockForSubquery(pSql); } else { tscProcessSql(pSql); } @@ -226,7 +226,7 @@ void taos_fetch_rows_a(TAOS_RES *taosa, __async_cb_func_t fp, void *param) { // handle the sub queries of join query if (pCmd->command == TSDB_SQL_TABLE_JOIN_RETRIEVE) { - tscFetchDatablockFromSubquery(pSql); + tscFetchDatablockForSubquery(pSql); } else if (pRes->completed) { if(pCmd->command == TSDB_SQL_FETCH || (pCmd->command >= TSDB_SQL_SERV_STATUS && pCmd->command <= TSDB_SQL_CURRENT_USER)) { if (hasMoreVnodesToTry(pSql)) { // sequentially retrieve data from remain vnodes. diff --git a/src/client/src/tscFunctionImpl.c b/src/client/src/tscFunctionImpl.c index 12d3b7dfd38e09b30aed6b8e66e56e7eead61034..e0bd76233399b389c75aba4c28c04bdb8b1ac0f8 100644 --- a/src/client/src/tscFunctionImpl.c +++ b/src/client/src/tscFunctionImpl.c @@ -99,7 +99,7 @@ typedef struct SSumInfo { // the attribute of hasResult is not needed since the num attribute would server as this purpose typedef struct SAvgInfo { double sum; - int64_t num; // num servers as the hasResult attribute in other struct + int64_t num; } SAvgInfo; typedef struct SStddevInfo { @@ -167,7 +167,13 @@ int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionI functionId == TSDB_FUNC_TAG || functionId == TSDB_FUNC_INTERP) { *type = (int16_t)dataType; *bytes = (int16_t)dataBytes; - *interBytes = *bytes + sizeof(SResultInfo); + + if (functionId == TSDB_FUNC_INTERP) { + *interBytes = sizeof(SInterpInfoDetail); + } else { + *interBytes = 0; + } + return TSDB_CODE_SUCCESS; } @@ -175,21 +181,21 @@ int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionI if (functionId == TSDB_FUNC_TID_TAG) { // todo use struct *type = TSDB_DATA_TYPE_BINARY; *bytes = (int16_t)(dataBytes + sizeof(int16_t) + sizeof(int64_t) + sizeof(int32_t) + sizeof(int32_t) + VARSTR_HEADER_SIZE); - *interBytes = *bytes; + *interBytes = 0; return TSDB_CODE_SUCCESS; } if (functionId == TSDB_FUNC_COUNT) { *type = TSDB_DATA_TYPE_BIGINT; *bytes = sizeof(int64_t); - *interBytes = *bytes; + *interBytes = 0; return TSDB_CODE_SUCCESS; } if (functionId == TSDB_FUNC_ARITHM) { *type = TSDB_DATA_TYPE_DOUBLE; *bytes = sizeof(double); - *interBytes = *bytes; + *interBytes = 0; return TSDB_CODE_SUCCESS; } @@ -298,7 +304,7 @@ 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 + sizeof(SResultInfo); + *interBytes = dataBytes; } else if (functionId == TSDB_FUNC_SPREAD) { *type = (int16_t)TSDB_DATA_TYPE_DOUBLE; *bytes = sizeof(double); @@ -310,7 +316,7 @@ int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionI } else if (functionId == TSDB_FUNC_LEASTSQR) { *type = TSDB_DATA_TYPE_BINARY; *bytes = TSDB_AVG_FUNCTION_INTER_BUFFER_SIZE; // string - *interBytes = *bytes + sizeof(SResultInfo); + *interBytes = *bytes; } else if (functionId == TSDB_FUNC_FIRST_DST || functionId == TSDB_FUNC_LAST_DST) { *type = TSDB_DATA_TYPE_BINARY; *bytes = (int16_t)(dataBytes + sizeof(SFirstLastInfo)); @@ -334,28 +340,20 @@ int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionI return TSDB_CODE_SUCCESS; } -void setResultInfoBuf(SResultInfo *pResInfo, int32_t size, bool superTable, char* buf) { - assert(pResInfo->interResultBuf == NULL); - - pResInfo->bufLen = size; - pResInfo->superTableQ = superTable; - pResInfo->interResultBuf = buf; -} - // set the query flag to denote that query is completed static void no_next_step(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->complete = true; } static bool function_setup(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (pResInfo->initialized) { return false; } memset(pCtx->aOutputBuf, 0, (size_t)pCtx->outputBytes); - initResultInfo(pResInfo); + initResultInfo(pResInfo, pCtx->interBufBytes); return true; } @@ -367,7 +365,7 @@ static bool function_setup(SQLFunctionCtx *pCtx) { * @param pCtx */ static void function_finalizer(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (pResInfo->hasResult != DATA_SET_FLAG) { if (pCtx->outputType == TSDB_DATA_TYPE_BINARY || pCtx->outputType == TSDB_DATA_TYPE_NCHAR) { setVardataNull(pCtx->aOutputBuf, pCtx->outputType); @@ -431,7 +429,7 @@ static void count_function_f(SQLFunctionCtx *pCtx, int32_t index) { *((int64_t *)pCtx->aOutputBuf) += 1; // do not need it actually - SResultInfo *pInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pInfo = GET_RES_INFO(pCtx); pInfo->hasResult = DATA_SET_FLAG; } @@ -592,8 +590,8 @@ static void sum_function(SQLFunctionCtx *pCtx) { do_sum(pCtx); // keep the result data in output buffer, not in the intermediate buffer - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - if (pResInfo->hasResult == DATA_SET_FLAG && pResInfo->superTableQ) { + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + if (pResInfo->hasResult == DATA_SET_FLAG && pCtx->stableQuery) { // set the flag for super table query SSumInfo *pSum = (SSumInfo *)pCtx->aOutputBuf; pSum->hasResult = DATA_SET_FLAG; @@ -604,8 +602,8 @@ static void sum_function_f(SQLFunctionCtx *pCtx, int32_t index) { do_sum_f(pCtx, index); // keep the result data in output buffer, not in the intermediate buffer - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - if (pResInfo->hasResult == DATA_SET_FLAG && pResInfo->superTableQ) { + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + if (pResInfo->hasResult == DATA_SET_FLAG && pCtx->stableQuery) { SSumInfo *pSum = (SSumInfo *)pCtx->aOutputBuf; pSum->hasResult = DATA_SET_FLAG; } @@ -615,8 +613,7 @@ static int32_t sum_merge_impl(const SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; GET_TRUE_DATA_TYPE(); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + assert(pCtx->stableQuery); for (int32_t i = 0; i < pCtx->size; ++i) { char * input = GET_INPUT_CHAR_INDEX(pCtx, i); @@ -661,7 +658,7 @@ static void sum_func_second_merge(SQLFunctionCtx *pCtx) { int32_t notNullElems = sum_merge_impl(pCtx); SET_VAL(pCtx, notNullElems, 1); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (notNullElems > 0) { pResInfo->hasResult = DATA_SET_FLAG; @@ -755,9 +752,9 @@ static void avg_function(SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; // NOTE: keep the intermediate result into the interResultBuf - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - SAvgInfo *pAvgInfo = (SAvgInfo *)pResInfo->interResultBuf; + SAvgInfo *pAvgInfo = (SAvgInfo *)GET_ROWCELL_INTERBUF(pResInfo); double * pVal = &pAvgInfo->sum; if (pCtx->preAggVals.isSet) { @@ -800,8 +797,8 @@ static void avg_function(SQLFunctionCtx *pCtx) { } // keep the data into the final output buffer for super table query since this execution may be the last one - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SAvgInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SAvgInfo)); } } @@ -814,9 +811,9 @@ static void avg_function_f(SQLFunctionCtx *pCtx, int32_t index) { SET_VAL(pCtx, 1, 1); // NOTE: keep the intermediate result into the interResultBuf - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - SAvgInfo *pAvgInfo = (SAvgInfo *)pResInfo->interResultBuf; + SAvgInfo *pAvgInfo = (SAvgInfo *)GET_ROWCELL_INTERBUF(pResInfo); if (pCtx->inputType == TSDB_DATA_TYPE_TINYINT) { pAvgInfo->sum += GET_INT8_VAL(pData); @@ -839,16 +836,16 @@ static void avg_function_f(SQLFunctionCtx *pCtx, int32_t index) { pResInfo->hasResult = DATA_SET_FLAG; // keep the data into the final output buffer for super table query since this execution may be the last one - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SAvgInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SAvgInfo)); } } static void avg_func_merge(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + assert(pCtx->stableQuery); - SAvgInfo *pAvgInfo = (SAvgInfo *)pResInfo->interResultBuf; + SAvgInfo *pAvgInfo = (SAvgInfo *)GET_ROWCELL_INTERBUF(pResInfo); char * input = GET_INPUT_CHAR(pCtx); for (int32_t i = 0; i < pCtx->size; ++i, input += pCtx->inputBytes) { @@ -864,12 +861,12 @@ static void avg_func_merge(SQLFunctionCtx *pCtx) { // if the data set hasResult is not set, the result is null if (pAvgInfo->num > 0) { pResInfo->hasResult = DATA_SET_FLAG; - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SAvgInfo)); + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SAvgInfo)); } } static void avg_func_second_merge(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); double *sum = (double*) pCtx->aOutputBuf; char * input = GET_INPUT_CHAR(pCtx); @@ -883,7 +880,7 @@ static void avg_func_second_merge(SQLFunctionCtx *pCtx) { *sum += pInput->sum; // keep the number of data into the temp buffer - *(int64_t *)pResInfo->interResultBuf += pInput->num; + *(int64_t *)GET_ROWCELL_INTERBUF(pResInfo) += pInput->num; } } @@ -891,21 +888,21 @@ static void avg_func_second_merge(SQLFunctionCtx *pCtx) { * the average value is calculated in finalize routine, since current routine does not know the exact number of points */ static void avg_finalizer(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (pCtx->currentStage == SECONDARY_STAGE_MERGE) { assert(pCtx->inputType == TSDB_DATA_TYPE_BINARY); - if (GET_INT64_VAL(pResInfo->interResultBuf) <= 0) { + if (GET_INT64_VAL(GET_ROWCELL_INTERBUF(pResInfo)) <= 0) { setNull(pCtx->aOutputBuf, pCtx->outputType, pCtx->outputBytes); return; // empty table } - *(double *)pCtx->aOutputBuf = (*(double *)pCtx->aOutputBuf) / *(int64_t *)pResInfo->interResultBuf; + *(double *)pCtx->aOutputBuf = (*(double *)pCtx->aOutputBuf) / *(int64_t *)GET_ROWCELL_INTERBUF(pResInfo); } else { // this is the secondary merge, only in the secondary merge, the input type is TSDB_DATA_TYPE_BINARY assert(pCtx->inputType >= TSDB_DATA_TYPE_TINYINT && pCtx->inputType <= TSDB_DATA_TYPE_DOUBLE); - SAvgInfo *pAvgInfo = (SAvgInfo *)pResInfo->interResultBuf; + SAvgInfo *pAvgInfo = (SAvgInfo *)GET_ROWCELL_INTERBUF(pResInfo); if (pAvgInfo->num == 0) { // all data are NULL or empty table setNull(pCtx->aOutputBuf, pCtx->outputType, pCtx->outputBytes); @@ -1116,11 +1113,11 @@ static void min_function(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, notNullElems, 1); if (notNullElems > 0) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; // set the flag for super table query - if (pResInfo->superTableQ) { + if (pCtx->stableQuery) { *(pCtx->aOutputBuf + pCtx->inputBytes) = DATA_SET_FLAG; } } @@ -1133,11 +1130,11 @@ static void max_function(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, notNullElems, 1); if (notNullElems > 0) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; // set the flag for super table query - if (pResInfo->superTableQ) { + if (pCtx->stableQuery) { *(pCtx->aOutputBuf + pCtx->inputBytes) = DATA_SET_FLAG; } } @@ -1148,8 +1145,7 @@ static int32_t minmax_merge_impl(SQLFunctionCtx *pCtx, int32_t bytes, char *outp GET_TRUE_DATA_TYPE(); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + assert(pCtx->stableQuery); for (int32_t i = 0; i < pCtx->size; ++i) { char *input = GET_INPUT_CHAR_INDEX(pCtx, i); @@ -1210,7 +1206,7 @@ static void min_func_merge(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, notNullElems, 1); - if (notNullElems > 0) { // for super table query, SResultInfo is not used + if (notNullElems > 0) { // for super table query, SResultRowCellInfo is not used char *flag = pCtx->aOutputBuf + pCtx->inputBytes; *flag = DATA_SET_FLAG; } @@ -1221,7 +1217,7 @@ static void min_func_second_merge(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, notNullElems, 1); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (notNullElems > 0) { pResInfo->hasResult = DATA_SET_FLAG; } @@ -1242,7 +1238,7 @@ static void max_func_second_merge(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, numOfElem, 1); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (numOfElem > 0) { pResInfo->hasResult = DATA_SET_FLAG; } @@ -1297,8 +1293,8 @@ static void max_function_f(SQLFunctionCtx *pCtx, int32_t index) { SET_VAL(pCtx, 1, 1); minMax_function_f(pCtx, index, 0); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - if (pResInfo->hasResult == DATA_SET_FLAG && pResInfo->superTableQ) { + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + if (pResInfo->hasResult == DATA_SET_FLAG && pCtx->stableQuery) { char *flag = pCtx->aOutputBuf + pCtx->inputBytes; *flag = DATA_SET_FLAG; } @@ -1313,8 +1309,8 @@ static void min_function_f(SQLFunctionCtx *pCtx, int32_t index) { SET_VAL(pCtx, 1, 1); minMax_function_f(pCtx, index, 1); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - if (pResInfo->hasResult == DATA_SET_FLAG && pResInfo->superTableQ) { + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + if (pResInfo->hasResult == DATA_SET_FLAG && pCtx->stableQuery) { char *flag = pCtx->aOutputBuf + pCtx->inputBytes; *flag = DATA_SET_FLAG; } @@ -1330,7 +1326,7 @@ static void min_function_f(SQLFunctionCtx *pCtx, int32_t index) { static void stddev_function(SQLFunctionCtx *pCtx) { // the second stage to calculate standard deviation - SStddevInfo *pStd = GET_RES_INFO(pCtx)->interResultBuf; + SStddevInfo *pStd = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); if (pStd->stage == 0) { // the first stage is to calculate average value avg_function(pCtx); @@ -1381,8 +1377,8 @@ static void stddev_function(SQLFunctionCtx *pCtx) { static void stddev_function_f(SQLFunctionCtx *pCtx, int32_t index) { // the second stage to calculate standard deviation - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SStddevInfo *pStd = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SStddevInfo *pStd = GET_ROWCELL_INTERBUF(pResInfo); /* the first stage is to calculate average value */ if (pStd->stage == 0) { @@ -1433,8 +1429,8 @@ static void stddev_next_step(SQLFunctionCtx *pCtx) { * the stddevInfo and the average info struct share the same buffer area * And the position of each element in their struct is exactly the same matched */ - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SStddevInfo *pStd = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SStddevInfo *pStd = GET_ROWCELL_INTERBUF(pResInfo); if (pStd->stage == 0) { /* @@ -1449,7 +1445,7 @@ static void stddev_next_step(SQLFunctionCtx *pCtx) { pResInfo->initialized = true; // set it initialized to avoid re-initialization // save average value into tmpBuf, for second stage scan - SAvgInfo *pAvg = pResInfo->interResultBuf; + SAvgInfo *pAvg = GET_ROWCELL_INTERBUF(pResInfo); pStd->avg = GET_DOUBLE_VAL(pCtx->aOutputBuf); assert((isnan(pAvg->sum) && pAvg->num == 0) || (pStd->num == pAvg->num && pStd->avg == pAvg->sum)); @@ -1459,7 +1455,7 @@ static void stddev_next_step(SQLFunctionCtx *pCtx) { } static void stddev_finalizer(SQLFunctionCtx *pCtx) { - SStddevInfo *pStd = (SStddevInfo *)GET_RES_INFO(pCtx)->interResultBuf; + SStddevInfo *pStd = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); if (pStd->num <= 0) { setNull(pCtx->aOutputBuf, pCtx->outputType, pCtx->outputBytes); @@ -1505,7 +1501,7 @@ static void first_function(SQLFunctionCtx *pCtx) { TSKEY k = pCtx->ptsList[i]; DO_UPDATE_TAG_COLUMNS(pCtx, k); - SResultInfo *pInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pInfo = GET_RES_INFO(pCtx); pInfo->hasResult = DATA_SET_FLAG; pInfo->complete = true; @@ -1532,7 +1528,7 @@ static void first_function_f(SQLFunctionCtx *pCtx, int32_t index) { TSKEY ts = pCtx->ptsList[index]; DO_UPDATE_TAG_COLUMNS(pCtx, ts); - SResultInfo *pInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pInfo = GET_RES_INFO(pCtx); pInfo->hasResult = DATA_SET_FLAG; pInfo->complete = true; // get the first not-null data, completed } @@ -1576,7 +1572,7 @@ static void first_dist_function(SQLFunctionCtx *pCtx) { first_data_assign_impl(pCtx, data, i); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; notNullElems++; @@ -1604,8 +1600,7 @@ static void first_dist_function_f(SQLFunctionCtx *pCtx, int32_t index) { static void first_dist_func_merge(SQLFunctionCtx *pCtx) { char *pData = GET_INPUT_CHAR(pCtx); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pCtx->size == 1 && pResInfo->superTableQ); + assert(pCtx->size == 1 && pCtx->stableQuery); SFirstLastInfo *pInput = (SFirstLastInfo *)(pData + pCtx->inputBytes); if (pInput->hasResult != DATA_SET_FLAG) { @@ -1620,8 +1615,8 @@ static void first_dist_func_merge(SQLFunctionCtx *pCtx) { } static void first_dist_func_second_merge(SQLFunctionCtx *pCtx) { - assert(pCtx->resultInfo->superTableQ); - + assert(pCtx->stableQuery); + char * pData = GET_INPUT_CHAR(pCtx); SFirstLastInfo *pInput = (SFirstLastInfo*) (pData + pCtx->outputBytes); if (pInput->hasResult != DATA_SET_FLAG) { @@ -1668,7 +1663,7 @@ static void last_function(SQLFunctionCtx *pCtx) { TSKEY ts = pCtx->ptsList[i]; DO_UPDATE_TAG_COLUMNS(pCtx, ts); - SResultInfo *pInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pInfo = GET_RES_INFO(pCtx); pInfo->hasResult = DATA_SET_FLAG; pInfo->complete = true; // set query completed on this column @@ -1691,7 +1686,7 @@ static void last_function_f(SQLFunctionCtx *pCtx, int32_t index) { TSKEY ts = pCtx->ptsList[index]; DO_UPDATE_TAG_COLUMNS(pCtx, ts); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; pResInfo->complete = true; // set query completed } @@ -1740,7 +1735,7 @@ static void last_dist_function(SQLFunctionCtx *pCtx) { last_data_assign_impl(pCtx, data, i); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; notNullElems++; @@ -1776,8 +1771,7 @@ static void last_dist_function_f(SQLFunctionCtx *pCtx, int32_t index) { static void last_dist_func_merge(SQLFunctionCtx *pCtx) { char *pData = GET_INPUT_CHAR(pCtx); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pCtx->size == 1 && pResInfo->superTableQ); + assert(pCtx->size == 1 && pCtx->stableQuery); // the input data is null SFirstLastInfo *pInput = (SFirstLastInfo *)(pData + pCtx->inputBytes); @@ -1833,11 +1827,11 @@ static void last_row_function(SQLFunctionCtx *pCtx) { // assign the last element in current data block assignVal(pCtx->aOutputBuf, pData + (pCtx->size - 1) * pCtx->inputBytes, pCtx->inputBytes, pCtx->inputType); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; // set the result to final result buffer in case of super table query - if (pResInfo->superTableQ) { + if (pCtx->stableQuery) { SLastrowInfo *pInfo1 = (SLastrowInfo *)(pCtx->aOutputBuf + pCtx->inputBytes); pInfo1->ts = pCtx->ptsList[pCtx->size - 1]; pInfo1->hasResult = DATA_SET_FLAG; @@ -1852,7 +1846,7 @@ static void last_row_function(SQLFunctionCtx *pCtx) { static void last_row_finalizer(SQLFunctionCtx *pCtx) { // do nothing at the first stage - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (pResInfo->hasResult != DATA_SET_FLAG) { if (pCtx->outputType == TSDB_DATA_TYPE_BINARY || pCtx->outputType == TSDB_DATA_TYPE_NCHAR) { setVardataNull(pCtx->aOutputBuf, pCtx->outputType); @@ -2044,8 +2038,8 @@ static int32_t resDataAscComparFn(const void *pLeft, const void *pRight) { static int32_t resDataDescComparFn(const void *pLeft, const void *pRight) { return -resDataAscComparFn(pLeft, pRight); } static void copyTopBotRes(SQLFunctionCtx *pCtx, int32_t type) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - STopBotInfo *pRes = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + STopBotInfo *pRes = GET_ROWCELL_INTERBUF(pResInfo); tValuePair **tvp = pRes->res; @@ -2135,18 +2129,18 @@ static void copyTopBotRes(SQLFunctionCtx *pCtx, int32_t type) { * top/bottom use the intermediate result buffer to keep the intermediate result */ static STopBotInfo *getTopBotOutputInfo(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); // only the first_stage_merge is directly written data into final output buffer - if (pResInfo->superTableQ && pCtx->currentStage != SECONDARY_STAGE_MERGE) { + if (pCtx->stableQuery && pCtx->currentStage != SECONDARY_STAGE_MERGE) { return (STopBotInfo*) pCtx->aOutputBuf; } else { // during normal table query and super table at the secondary_stage, result is written to intermediate buffer - return pResInfo->interResultBuf; + return GET_ROWCELL_INTERBUF(pResInfo); } } bool topbot_datablock_filter(SQLFunctionCtx *pCtx, int32_t functionId, const char *minval, const char *maxval) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (pResInfo == NULL) { return true; } @@ -2252,7 +2246,7 @@ static void top_function(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, notNullElems, 1); if (notNullElems > 0) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; } } @@ -2270,7 +2264,7 @@ static void top_function_f(SQLFunctionCtx *pCtx, int32_t index) { do_top_function_add(pRes, (int32_t)pCtx->param[0].i64Key, pData, pCtx->ptsList[index], pCtx->inputType, &pCtx->tagInfo, NULL, 0); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; } @@ -2285,8 +2279,7 @@ static void top_func_merge(SQLFunctionCtx *pCtx) { // remmap the input buffer may cause the struct pointer invalid, so rebuild the STopBotInfo is necessary buildTopBotStruct(pInput, pCtx); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ && pCtx->outputType == TSDB_DATA_TYPE_BINARY && pCtx->size == 1); + assert(pCtx->stableQuery && pCtx->outputType == TSDB_DATA_TYPE_BINARY && pCtx->size == 1); STopBotInfo *pOutput = getTopBotOutputInfo(pCtx); @@ -2314,7 +2307,7 @@ static void top_func_second_merge(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, pInput->num, pOutput->num); if (pOutput->num > 0) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; } } @@ -2343,7 +2336,7 @@ static void bottom_function(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, notNullElems, 1); if (notNullElems > 0) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; } } @@ -2359,7 +2352,7 @@ static void bottom_function_f(SQLFunctionCtx *pCtx, int32_t index) { do_bottom_function_add(pRes, (int32_t)pCtx->param[0].i64Key, pData, pCtx->ptsList[index], pCtx->inputType, &pCtx->tagInfo, NULL, 0); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; } @@ -2374,8 +2367,7 @@ static void bottom_func_merge(SQLFunctionCtx *pCtx) { // remmap the input buffer may cause the struct pointer invalid, so rebuild the STopBotInfo is necessary buildTopBotStruct(pInput, pCtx); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ && pCtx->outputType == TSDB_DATA_TYPE_BINARY && pCtx->size == 1); + assert(pCtx->stableQuery && pCtx->outputType == TSDB_DATA_TYPE_BINARY && pCtx->size == 1); STopBotInfo *pOutput = getTopBotOutputInfo(pCtx); @@ -2403,16 +2395,16 @@ static void bottom_func_second_merge(SQLFunctionCtx *pCtx) { SET_VAL(pCtx, pInput->num, pOutput->num); if (pOutput->num > 0) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; } } static void top_bottom_func_finalizer(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); // data in temporary list is less than the required number of results, not enough qualified number of results - STopBotInfo *pRes = pResInfo->interResultBuf; + STopBotInfo *pRes = GET_ROWCELL_INTERBUF(pResInfo); if (pRes->num == 0) { // no result assert(pResInfo->hasResult != DATA_SET_FLAG); // TODO: @@ -2443,8 +2435,8 @@ static bool percentile_function_setup(SQLFunctionCtx *pCtx) { } // in the first round, get the min-max value of all involved data - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SPercentileInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SPercentileInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); SET_DOUBLE_VAL(&pInfo->minval, DBL_MAX); SET_DOUBLE_VAL(&pInfo->maxval, -DBL_MAX); pInfo->numOfElems = 0; @@ -2455,8 +2447,8 @@ static bool percentile_function_setup(SQLFunctionCtx *pCtx) { static void percentile_function(SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SPercentileInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SPercentileInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); // the first stage, only acquire the min/max value if (pInfo->stage == 0) { @@ -2546,9 +2538,9 @@ static void percentile_function_f(SQLFunctionCtx *pCtx, int32_t index) { return; } - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - SPercentileInfo *pInfo = (SPercentileInfo *)pResInfo->interResultBuf; + SPercentileInfo *pInfo = (SPercentileInfo *)GET_ROWCELL_INTERBUF(pResInfo); if (pInfo->stage == 0) { // TODO extract functions @@ -2595,8 +2587,8 @@ static void percentile_function_f(SQLFunctionCtx *pCtx, int32_t index) { static void percentile_finalizer(SQLFunctionCtx *pCtx) { double v = pCtx->param[0].nType == TSDB_DATA_TYPE_INT ? pCtx->param[0].i64Key : pCtx->param[0].dKey; - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - tMemBucket * pMemBucket = ((SPercentileInfo *)pResInfo->interResultBuf)->pMemBucket; + 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); @@ -2609,8 +2601,8 @@ static void percentile_finalizer(SQLFunctionCtx *pCtx) { } static void percentile_next_step(SQLFunctionCtx *pCtx) { - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SPercentileInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SPercentileInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); if (pInfo->stage == 0) { // all data are null, set it completed @@ -2627,12 +2619,12 @@ static void percentile_next_step(SQLFunctionCtx *pCtx) { ////////////////////////////////////////////////////////////////////////////////// static SAPercentileInfo *getAPerctInfo(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - if (pResInfo->superTableQ && pCtx->currentStage != SECONDARY_STAGE_MERGE) { + if (pCtx->stableQuery && pCtx->currentStage != SECONDARY_STAGE_MERGE) { return (SAPercentileInfo*) pCtx->aOutputBuf; } else { - return pResInfo->interResultBuf; + return GET_ROWCELL_INTERBUF(pResInfo); } } @@ -2651,7 +2643,7 @@ static bool apercentile_function_setup(SQLFunctionCtx *pCtx) { static void apercentile_function(SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; - SResultInfo * pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); SAPercentileInfo *pInfo = getAPerctInfo(pCtx); for (int32_t i = 0; i < pCtx->size; ++i) { @@ -2704,8 +2696,8 @@ static void apercentile_function_f(SQLFunctionCtx *pCtx, int32_t index) { return; } - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SAPercentileInfo *pInfo = getAPerctInfo(pCtx); // pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SAPercentileInfo *pInfo = getAPerctInfo(pCtx); double v = 0; switch (pCtx->inputType) { @@ -2736,8 +2728,8 @@ static void apercentile_function_f(SQLFunctionCtx *pCtx, int32_t index) { } static void apercentile_func_merge(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + assert(pCtx->stableQuery); SAPercentileInfo *pInput = (SAPercentileInfo *)GET_INPUT_CHAR(pCtx); @@ -2794,7 +2786,7 @@ static void apercentile_func_second_merge(SQLFunctionCtx *pCtx) { pOutput->pHisto = pRes; } - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); pResInfo->hasResult = DATA_SET_FLAG; SET_VAL(pCtx, 1, 1); } @@ -2802,8 +2794,8 @@ static void apercentile_func_second_merge(SQLFunctionCtx *pCtx) { static void apercentile_finalizer(SQLFunctionCtx *pCtx) { double v = (pCtx->param[0].nType == TSDB_DATA_TYPE_INT) ? pCtx->param[0].i64Key : pCtx->param[0].dKey; - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SAPercentileInfo *pOutput = pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SAPercentileInfo *pOutput = GET_ROWCELL_INTERBUF(pResInfo); if (pCtx->currentStage == SECONDARY_STAGE_MERGE) { if (pResInfo->hasResult == DATA_SET_FLAG) { // check for null @@ -2840,8 +2832,8 @@ static bool leastsquares_function_setup(SQLFunctionCtx *pCtx) { return false; } - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SLeastsquareInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SLeastsquareInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); // 2*3 matrix pInfo->startVal = pCtx->param[0].dKey; @@ -2867,8 +2859,8 @@ static bool leastsquares_function_setup(SQLFunctionCtx *pCtx) { } static void leastsquares_function(SQLFunctionCtx *pCtx) { - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SLeastsquareInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SLeastsquareInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); double(*param)[3] = pInfo->mat; double x = pInfo->startVal; @@ -2938,8 +2930,8 @@ static void leastsquares_function_f(SQLFunctionCtx *pCtx, int32_t index) { return; } - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SLeastsquareInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SLeastsquareInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); double(*param)[3] = pInfo->mat; @@ -2988,8 +2980,8 @@ static void leastsquares_function_f(SQLFunctionCtx *pCtx, int32_t index) { static void leastsquares_finalizer(SQLFunctionCtx *pCtx) { // no data in query - SResultInfo * pResInfo = GET_RES_INFO(pCtx); - SLeastsquareInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo * pResInfo = GET_RES_INFO(pCtx); + SLeastsquareInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); if (pInfo->num == 0) { if (pCtx->outputType == TSDB_DATA_TYPE_BINARY || pCtx->outputType == TSDB_DATA_TYPE_NCHAR) { @@ -3054,7 +3046,7 @@ static void col_project_function(SQLFunctionCtx *pCtx) { } static void col_project_function_f(SQLFunctionCtx *pCtx, int32_t index) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (pCtx->numOfParams == 2) { // the number of output rows should not affect the final number of rows, so set it to be 0 return; } @@ -3486,7 +3478,7 @@ static bool spread_function_setup(SQLFunctionCtx *pCtx) { return false; } - SSpreadInfo *pInfo = GET_RES_INFO(pCtx)->interResultBuf; + SSpreadInfo *pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); // this is the server-side setup function in client-side, the secondary merge do not need this procedure if (pCtx->currentStage == SECONDARY_STAGE_MERGE) { @@ -3501,8 +3493,8 @@ static bool spread_function_setup(SQLFunctionCtx *pCtx) { } static void spread_function(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SSpreadInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SSpreadInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); int32_t numOfElems = 0; @@ -3568,8 +3560,8 @@ static void spread_function(SQLFunctionCtx *pCtx) { } // keep the data into the final output buffer for super table query since this execution may be the last one - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SSpreadInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SSpreadInfo)); } } @@ -3581,8 +3573,8 @@ static void spread_function_f(SQLFunctionCtx *pCtx, int32_t index) { SET_VAL(pCtx, 1, 1); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SSpreadInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SSpreadInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); double val = 0.0; if (pCtx->inputType == TSDB_DATA_TYPE_TINYINT) { @@ -3611,16 +3603,16 @@ static void spread_function_f(SQLFunctionCtx *pCtx, int32_t index) { pResInfo->hasResult = DATA_SET_FLAG; pInfo->hasResult = DATA_SET_FLAG; - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SSpreadInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SSpreadInfo)); } } void spread_func_merge(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + assert(pCtx->stableQuery); - SSpreadInfo *pResData = pResInfo->interResultBuf; + SSpreadInfo *pResData = GET_ROWCELL_INTERBUF(pResInfo); int32_t notNullElems = 0; for (int32_t i = 0; i < pCtx->size; ++i) { @@ -3644,7 +3636,7 @@ void spread_func_merge(SQLFunctionCtx *pCtx) { } if (notNullElems > 0) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SSpreadInfo)); + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SSpreadInfo)); pResInfo->hasResult = DATA_SET_FLAG; } } @@ -3675,7 +3667,7 @@ void spread_function_finalizer(SQLFunctionCtx *pCtx) { * here we do not check the input data types, because in case of metric query, * the type of intermediate data is binary */ - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); if (pCtx->currentStage == SECONDARY_STAGE_MERGE) { assert(pCtx->inputType == TSDB_DATA_TYPE_BINARY); @@ -3690,7 +3682,7 @@ void spread_function_finalizer(SQLFunctionCtx *pCtx) { assert((pCtx->inputType >= TSDB_DATA_TYPE_TINYINT && pCtx->inputType <= TSDB_DATA_TYPE_DOUBLE) || (pCtx->inputType == TSDB_DATA_TYPE_TIMESTAMP)); - SSpreadInfo *pInfo = GET_RES_INFO(pCtx)->interResultBuf; + SSpreadInfo *pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); if (pInfo->hasResult != DATA_SET_FLAG) { setNull(pCtx->aOutputBuf, pCtx->outputType, pCtx->outputBytes); return; @@ -3714,8 +3706,8 @@ static bool twa_function_setup(SQLFunctionCtx *pCtx) { return false; } - SResultInfo *pResInfo = GET_RES_INFO(pCtx); //->aOutputBuf + pCtx->outputBytes; - STwaInfo * pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); //->aOutputBuf + pCtx->outputBytes; + STwaInfo * pInfo = GET_ROWCELL_INTERBUF(pResInfo); pInfo->lastKey = INT64_MIN; pInfo->type = pCtx->inputType; @@ -3754,8 +3746,8 @@ static void twa_function(SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - STwaInfo * pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + STwaInfo * pInfo = GET_ROWCELL_INTERBUF(pResInfo); int32_t i = 0; @@ -3808,7 +3800,7 @@ static void twa_function(SQLFunctionCtx *pCtx) { pResInfo->hasResult = DATA_SET_FLAG; } - if (pResInfo->superTableQ) { + if (pCtx->stableQuery) { memcpy(pCtx->aOutputBuf, pInfo, sizeof(STwaInfo)); } @@ -3825,8 +3817,8 @@ static void twa_function_f(SQLFunctionCtx *pCtx, int32_t index) { TSKEY *primaryKey = pCtx->ptsList; - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - STwaInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + STwaInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); if (pInfo->lastKey == INT64_MIN) { pInfo->lastKey = pCtx->nStartQueryTimestamp; @@ -3848,14 +3840,13 @@ static void twa_function_f(SQLFunctionCtx *pCtx, int32_t index) { // pCtx->numOfIteratedElems += 1; pResInfo->hasResult = DATA_SET_FLAG; - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(STwaInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(STwaInfo)); } } static void twa_func_merge(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + assert(pCtx->stableQuery); STwaInfo *pBuf = (STwaInfo *)pCtx->aOutputBuf; char * indicator = pCtx->aInputElemBuf; @@ -3895,16 +3886,16 @@ static void twa_func_merge(SQLFunctionCtx *pCtx) { */ void twa_function_copy(SQLFunctionCtx *pCtx) { assert(pCtx->inputType == TSDB_DATA_TYPE_BINARY); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - memcpy(pResInfo->interResultBuf, pCtx->aInputElemBuf, (size_t)pCtx->inputBytes); + memcpy(GET_ROWCELL_INTERBUF(pResInfo), pCtx->aInputElemBuf, (size_t)pCtx->inputBytes); pResInfo->hasResult = ((STwaInfo *)pCtx->aInputElemBuf)->hasResult; } void twa_function_finalizer(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - STwaInfo *pInfo = (STwaInfo *)pResInfo->interResultBuf; + STwaInfo *pInfo = (STwaInfo *)GET_ROWCELL_INTERBUF(pResInfo); assert(pInfo->EKey >= pInfo->lastKey && pInfo->hasResult == pResInfo->hasResult); if (pInfo->hasResult != DATA_SET_FLAG) { @@ -3932,8 +3923,8 @@ void twa_function_finalizer(SQLFunctionCtx *pCtx) { */ static void interp_function(SQLFunctionCtx *pCtx) { // at this point, the value is existed, return directly - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SInterpInfoDetail* pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SInterpInfoDetail* pInfo = GET_ROWCELL_INTERBUF(pResInfo); if (pCtx->size == 1) { char *pData = GET_INPUT_CHAR(pCtx); @@ -4008,7 +3999,6 @@ static void interp_function(SQLFunctionCtx *pCtx) { } } } - } SET_VAL(pCtx, pCtx->size, 1); @@ -4019,8 +4009,8 @@ static bool ts_comp_function_setup(SQLFunctionCtx *pCtx) { return false; // not initialized since it has been initialized } - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - STSCompInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + STSCompInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); pInfo->pTSBuf = tsBufCreate(false, pCtx->order); pInfo->pTSBuf->tsOrder = pCtx->order; @@ -4028,8 +4018,8 @@ static bool ts_comp_function_setup(SQLFunctionCtx *pCtx) { } static void ts_comp_function(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - STSBuf * pTSbuf = ((STSCompInfo *)(pResInfo->interResultBuf))->pTSBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + STSBuf * pTSbuf = ((STSCompInfo *)(GET_ROWCELL_INTERBUF(pResInfo)))->pTSBuf; const char *input = GET_INPUT_CHAR(pCtx); @@ -4053,8 +4043,8 @@ static void ts_comp_function_f(SQLFunctionCtx *pCtx, int32_t index) { return; } - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - STSCompInfo *pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + STSCompInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); STSBuf *pTSbuf = pInfo->pTSBuf; @@ -4065,9 +4055,9 @@ static void ts_comp_function_f(SQLFunctionCtx *pCtx, int32_t index) { } static void ts_comp_finalize(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); - STSCompInfo *pInfo = pResInfo->interResultBuf; + STSCompInfo *pInfo = GET_ROWCELL_INTERBUF(pResInfo); STSBuf * pTSbuf = pInfo->pTSBuf; tsBufFlush(pTSbuf); @@ -4116,8 +4106,8 @@ static bool rate_function_setup(SQLFunctionCtx *pCtx) { return false; } - SResultInfo *pResInfo = GET_RES_INFO(pCtx); //->aOutputBuf + pCtx->outputBytes; - SRateInfo * pInfo = pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); //->aOutputBuf + pCtx->outputBytes; + SRateInfo * pInfo = GET_ROWCELL_INTERBUF(pResInfo); pInfo->CorrectionValue = 0; pInfo->firstKey = INT64_MIN; @@ -4136,8 +4126,8 @@ static bool rate_function_setup(SQLFunctionCtx *pCtx) { static void rate_function(SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SRateInfo *pRateInfo = (SRateInfo *)GET_ROWCELL_INTERBUF(pResInfo); TSKEY *primaryKey = pCtx->ptsList; tscDebug("%p rate_function() size:%d, hasNull:%d", pCtx, pCtx->size, pCtx->hasNull); @@ -4200,8 +4190,8 @@ static void rate_function(SQLFunctionCtx *pCtx) { } // keep the data into the final output buffer for super table query since this execution may be the last one - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SRateInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SRateInfo)); } } @@ -4212,8 +4202,8 @@ static void rate_function_f(SQLFunctionCtx *pCtx, int32_t index) { } // NOTE: keep the intermediate result into the interResultBuf - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SRateInfo *pRateInfo = (SRateInfo *)GET_ROWCELL_INTERBUF(pResInfo); TSKEY *primaryKey = pCtx->ptsList; int64_t v = 0; @@ -4257,20 +4247,18 @@ static void rate_function_f(SQLFunctionCtx *pCtx, int32_t index) { pResInfo->hasResult = DATA_SET_FLAG; // keep the data into the final output buffer for super table query since this execution may be the last one - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SRateInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SRateInfo)); } } static void rate_func_merge(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + assert(pCtx->stableQuery); tscDebug("rate_func_merge() size:%d", pCtx->size); - //SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; SRateInfo *pBuf = (SRateInfo *)pCtx->aOutputBuf; char *indicator = pCtx->aInputElemBuf; @@ -4303,8 +4291,8 @@ static void rate_func_merge(SQLFunctionCtx *pCtx) { static void rate_func_copy(SQLFunctionCtx *pCtx) { assert(pCtx->inputType == TSDB_DATA_TYPE_BINARY); - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - memcpy(pResInfo->interResultBuf, pCtx->aInputElemBuf, (size_t)pCtx->inputBytes); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + memcpy(GET_ROWCELL_INTERBUF(pResInfo), pCtx->aInputElemBuf, (size_t)pCtx->inputBytes); pResInfo->hasResult = ((SRateInfo*)pCtx->aInputElemBuf)->hasResult; SRateInfo* pRateInfo = (SRateInfo*)pCtx->aInputElemBuf; @@ -4315,8 +4303,8 @@ static void rate_func_copy(SQLFunctionCtx *pCtx) { static void rate_finalizer(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SRateInfo *pRateInfo = (SRateInfo *)GET_ROWCELL_INTERBUF(pResInfo); tscDebug("%p isIRate:%d firstKey:%" PRId64 " lastKey:%" PRId64 " firstValue:%" PRId64 " lastValue:%" PRId64 " CorrectionValue:%" PRId64 " hasResult:%d", pCtx, pRateInfo->isIRate, pRateInfo->firstKey, pRateInfo->lastKey, pRateInfo->firstValue, pRateInfo->lastValue, pRateInfo->CorrectionValue, pRateInfo->hasResult); @@ -4341,8 +4329,8 @@ static void rate_finalizer(SQLFunctionCtx *pCtx) { static void irate_function(SQLFunctionCtx *pCtx) { int32_t notNullElems = 0; - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SRateInfo *pRateInfo = (SRateInfo *)GET_ROWCELL_INTERBUF(pResInfo); TSKEY *primaryKey = pCtx->ptsList; tscDebug("%p irate_function() size:%d, hasNull:%d", pCtx, pCtx->size, pCtx->hasNull); @@ -4404,8 +4392,8 @@ static void irate_function(SQLFunctionCtx *pCtx) { } // keep the data into the final output buffer for super table query since this execution may be the last one - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SRateInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SRateInfo)); } } @@ -4416,8 +4404,8 @@ static void irate_function_f(SQLFunctionCtx *pCtx, int32_t index) { } // NOTE: keep the intermediate result into the interResultBuf - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SRateInfo *pRateInfo = (SRateInfo *)GET_ROWCELL_INTERBUF(pResInfo); TSKEY *primaryKey = pCtx->ptsList; int64_t v = 0; @@ -4453,16 +4441,16 @@ static void irate_function_f(SQLFunctionCtx *pCtx, int32_t index) { pResInfo->hasResult = DATA_SET_FLAG; // keep the data into the final output buffer for super table query since this execution may be the last one - if (pResInfo->superTableQ) { - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SRateInfo)); + if (pCtx->stableQuery) { + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SRateInfo)); } } static void do_sumrate_merge(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - assert(pResInfo->superTableQ); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + assert(pCtx->stableQuery); - SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; + SRateInfo *pRateInfo = (SRateInfo *)GET_ROWCELL_INTERBUF(pResInfo); char * input = GET_INPUT_CHAR(pCtx); for (int32_t i = 0; i < pCtx->size; ++i, input += pCtx->inputBytes) { @@ -4486,7 +4474,7 @@ static void do_sumrate_merge(SQLFunctionCtx *pCtx) { if (DATA_SET_FLAG == pRateInfo->hasResult) { pResInfo->hasResult = DATA_SET_FLAG; SET_VAL(pCtx, pRateInfo->num, 1); - memcpy(pCtx->aOutputBuf, pResInfo->interResultBuf, sizeof(SRateInfo)); + memcpy(pCtx->aOutputBuf, GET_ROWCELL_INTERBUF(pResInfo), sizeof(SRateInfo)); } } @@ -4501,10 +4489,10 @@ static void sumrate_func_second_merge(SQLFunctionCtx *pCtx) { } static void sumrate_finalizer(SQLFunctionCtx *pCtx) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - SRateInfo *pRateInfo = (SRateInfo *)pResInfo->interResultBuf; + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); + SRateInfo *pRateInfo = (SRateInfo *)GET_ROWCELL_INTERBUF(pResInfo); - tscDebug("%p sumrate_finalizer() superTableQ:%d num:%" PRId64 " sum:%f hasResult:%d", pCtx, pResInfo->superTableQ, pRateInfo->num, pRateInfo->sum, pRateInfo->hasResult); + tscDebug("%p sumrate_finalizer() superTableQ:%d num:%" PRId64 " sum:%f hasResult:%d", pCtx, pCtx->stableQuery, pRateInfo->num, pRateInfo->sum, pRateInfo->hasResult); if (pRateInfo->hasResult != DATA_SET_FLAG) { setNull(pCtx->aOutputBuf, TSDB_DATA_TYPE_DOUBLE, sizeof(double)); diff --git a/src/client/src/tscLocal.c b/src/client/src/tscLocal.c index b4c3f3549b1576b3ff83300d4f10d9e07778530d..44dffab56f90e2d88604ee96a6cc1ddc8a5cfbef 100644 --- a/src/client/src/tscLocal.c +++ b/src/client/src/tscLocal.c @@ -49,82 +49,6 @@ typedef struct SCreateBuilder { } SCreateBuilder; static void tscSetLocalQueryResult(SSqlObj *pSql, const char *val, const char *columnName, int16_t type, size_t valueLength); -static int32_t getToStringLength(const char *pData, int32_t length, int32_t type) { - char buf[512] = {0}; - - int32_t len = 0; - int32_t MAX_BOOL_TYPE_LENGTH = 5; // max(strlen("true"), strlen("false")); - switch (type) { - case TSDB_DATA_TYPE_BINARY: - return length; - case TSDB_DATA_TYPE_NCHAR: - return length; - case TSDB_DATA_TYPE_DOUBLE: { - double dv = 0; - dv = GET_DOUBLE_VAL(pData); - len = sprintf(buf, "%lf", dv); - if (strncasecmp("nan", buf, 3) == 0) { - len = 4; - } - } break; - case TSDB_DATA_TYPE_FLOAT: { - float fv = 0; - fv = GET_FLOAT_VAL(pData); - len = sprintf(buf, "%f", fv); - if (strncasecmp("nan", buf, 3) == 0) { - len = 4; - } - } break; - case TSDB_DATA_TYPE_TIMESTAMP: - case TSDB_DATA_TYPE_BIGINT: - len = sprintf(buf, "%" PRId64, *(int64_t *)pData); - break; - case TSDB_DATA_TYPE_BOOL: - len = MAX_BOOL_TYPE_LENGTH; - break; - default: - len = sprintf(buf, "%d", *(int32_t *)pData); - break; - }; - return len; -} - -/* - * we need to convert all data into string, so we need to sprintf all kinds of - * non-string data into string, and record its length to get the right - * maximum length. The length may be less or greater than its original binary length: - * For example: - * length((short) 1) == 1, less than sizeof(short) - * length((uint64_t) 123456789011) > 12, greater than sizsof(uint64_t) - */ -static int32_t tscMaxLengthOfTagsFields(SSqlObj *pSql) { - STableMeta *pMeta = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0)->pTableMeta; - - if (pMeta->tableType == TSDB_SUPER_TABLE || pMeta->tableType == TSDB_NORMAL_TABLE || - pMeta->tableType == TSDB_STREAM_TABLE) { - return 0; - } - - char * pTagValue = tsGetTagsValue(pMeta); - SSchema *pTagsSchema = tscGetTableTagSchema(pMeta); - - int32_t len = getToStringLength(pTagValue, pTagsSchema[0].bytes, pTagsSchema[0].type); - - pTagValue += pTagsSchema[0].bytes; - int32_t numOfTags = tscGetNumOfTags(pMeta); - - for (int32_t i = 1; i < numOfTags; ++i) { - int32_t tLen = getToStringLength(pTagValue, pTagsSchema[i].bytes, pTagsSchema[i].type); - if (len < tLen) { - len = tLen; - } - - pTagValue += pTagsSchema[i].bytes; - } - - return len; -} - static int32_t tscSetValueToResObj(SSqlObj *pSql, int32_t rowLen) { SSqlRes *pRes = &pSql->res; @@ -186,8 +110,7 @@ static int32_t tscSetValueToResObj(SSqlObj *pSql, int32_t rowLen) { return 0; } - // the following is handle display tags value for meters created according to metric - char *pTagValue = tsGetTagsValue(pMeta); + // the following is handle display tags for table created according to super table for (int32_t i = numOfRows; i < totalNumOfRows; ++i) { // field name TAOS_FIELD *pField = tscFieldInfoGetField(&pQueryInfo->fieldsInfo, 0); @@ -219,8 +142,6 @@ static int32_t tscSetValueToResObj(SSqlObj *pSql, int32_t rowLen) { char *target = pRes->data + tscFieldInfoGetOffset(pQueryInfo, 3) * totalNumOfRows + pField->bytes * i; const char *src = "TAG"; STR_WITH_MAXSIZE_TO_VARSTR(target, src, pField->bytes); - - pTagValue += pSchema[i].bytes; } return 0; @@ -286,10 +207,10 @@ static int32_t tscProcessDescribeTable(SSqlObj *pSql) { const int32_t TYPE_COLUMN_LENGTH = 16; const int32_t NOTE_COLUMN_MIN_LENGTH = 8; - int32_t noteFieldLen = tscMaxLengthOfTagsFields(pSql); - if (noteFieldLen == 0) { - noteFieldLen = NOTE_COLUMN_MIN_LENGTH; - } + int32_t noteFieldLen = NOTE_COLUMN_MIN_LENGTH;//tscMaxLengthOfTagsFields(pSql); +// if (noteFieldLen == 0) { +// noteFieldLen = NOTE_COLUMN_MIN_LENGTH; +// } int32_t rowLen = tscBuildTableSchemaResultFields(pSql, NUM_OF_DESC_TABLE_COLUMNS, TYPE_COLUMN_LENGTH, noteFieldLen); tscFieldInfoUpdateOffset(pQueryInfo); diff --git a/src/client/src/tscLocalMerge.c b/src/client/src/tscLocalMerge.c index ea073e8192e73f0f2bc1e3d8351decd9177ba403..c67edf5b5a8a026fc10d251218abcf96ab018359 100644 --- a/src/client/src/tscLocalMerge.c +++ b/src/client/src/tscLocalMerge.c @@ -99,12 +99,9 @@ static void tscInitSqlContext(SSqlCmd *pCmd, SLocalReducer *pReducer, tOrderDesc pCtx->param[1].i64Key = pQueryInfo->order.orderColId; } - SResultInfo *pResInfo = &pReducer->pResInfo[i]; - pResInfo->bufLen = pExpr->interBytes; - pResInfo->interResultBuf = calloc(1, (size_t) pResInfo->bufLen); - - pCtx->resultInfo = &pReducer->pResInfo[i]; - pCtx->resultInfo->superTableQ = true; + pCtx->interBufBytes = pExpr->interBytes; + pCtx->resultInfo = calloc(1, pCtx->interBufBytes + sizeof(SResultRowCellInfo)); + pCtx->stableQuery = true; } int16_t n = 0; @@ -345,7 +342,6 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd size_t numOfCols = tscSqlExprNumOfExprs(pQueryInfo); pReducer->pTempBuffer->num = 0; - pReducer->pResInfo = calloc(numOfCols, sizeof(SResultInfo)); tscCreateResPointerInfo(pRes, pQueryInfo); tscInitSqlContext(pCmd, pReducer, pDesc); @@ -489,13 +485,15 @@ void tscDestroyLocalReducer(SSqlObj *pSql) { tscDebug("%p waiting for delete procedure, status: %d", pSql, status); } - pLocalReducer->pFillInfo = taosDestoryFillInfo(pLocalReducer->pFillInfo); + pLocalReducer->pFillInfo = taosDestroyFillInfo(pLocalReducer->pFillInfo); if (pLocalReducer->pCtx != NULL) { for (int32_t i = 0; i < pQueryInfo->fieldsInfo.numOfOutput; ++i) { SQLFunctionCtx *pCtx = &pLocalReducer->pCtx[i]; tVariantDestroy(&pCtx->tag); + taosTFree(pCtx->resultInfo); + if (pCtx->tagInfo.pTagCtxList != NULL) { taosTFree(pCtx->tagInfo.pTagCtxList); } @@ -509,15 +507,6 @@ void tscDestroyLocalReducer(SSqlObj *pSql) { taosTFree(pLocalReducer->pTempBuffer); taosTFree(pLocalReducer->pResultBuf); - if (pLocalReducer->pResInfo != NULL) { - size_t num = tscSqlExprNumOfExprs(pQueryInfo); - for (int32_t i = 0; i < num; ++i) { - taosTFree(pLocalReducer->pResInfo[i].interResultBuf); - } - - taosTFree(pLocalReducer->pResInfo); - } - if (pLocalReducer->pLoserTree) { taosTFree(pLocalReducer->pLoserTree->param); taosTFree(pLocalReducer->pLoserTree); @@ -1072,7 +1061,7 @@ static int64_t getNumOfResultLocal(SQueryInfo *pQueryInfo, SQLFunctionCtx *pCtx) continue; } - SResultInfo* pResInfo = GET_RES_INFO(&pCtx[j]); + SResultRowCellInfo* pResInfo = GET_RES_INFO(&pCtx[j]); if (maxOutput < pResInfo->numOfRes) { maxOutput = pResInfo->numOfRes; } diff --git a/src/client/src/tscProfile.c b/src/client/src/tscProfile.c index eb6843b0e4fd844c9d13da7f55dca47008f62a79..acc5acd786bfe00036e539a104da04af1485a263 100644 --- a/src/client/src/tscProfile.c +++ b/src/client/src/tscProfile.c @@ -222,7 +222,7 @@ void tscKillStream(STscObj *pObj, uint32_t killId) { } int tscBuildQueryStreamDesc(void *pMsg, STscObj *pObj) { - SCMHeartBeatMsg *pHeartbeat = pMsg; + SHeartBeatMsg *pHeartbeat = pMsg; int allocedQueriesNum = pHeartbeat->numOfQueries; int allocedStreamsNum = pHeartbeat->numOfStreams; @@ -277,7 +277,7 @@ int tscBuildQueryStreamDesc(void *pMsg, STscObj *pObj) { } int32_t msgLen = pHeartbeat->numOfQueries * sizeof(SQueryDesc) + pHeartbeat->numOfStreams * sizeof(SStreamDesc) + - sizeof(SCMHeartBeatMsg); + sizeof(SHeartBeatMsg); pHeartbeat->connId = htonl(pObj->connId); pHeartbeat->numOfQueries = htonl(pHeartbeat->numOfQueries); pHeartbeat->numOfStreams = htonl(pHeartbeat->numOfStreams); diff --git a/src/client/src/tscSQLParser.c b/src/client/src/tscSQLParser.c index 8716b5cb02fcc13e834f2672ed58caaab69e8f17..f7aaf830666e586d5ebd485011f3e94ddf47aeeb 100644 --- a/src/client/src/tscSQLParser.c +++ b/src/client/src/tscSQLParser.c @@ -65,7 +65,6 @@ static bool validateTagParams(tFieldList* pTagsList, tFieldList* pFieldList, SSq static int32_t setObjFullName(char* fullName, const char* account, SStrToken* pDB, SStrToken* tableName, int32_t* len); static void getColumnName(tSQLExprItem* pItem, char* resultFieldName, int32_t nameLength); -static void getRevisedName(char* resultFieldName, int32_t functionId, int32_t maxLen, char* columnName); static int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t colIndex, tSQLExprItem* pItem, bool finalResult); static int32_t insertResultField(SQueryInfo* pQueryInfo, int32_t outputIndex, SColumnList* pIdList, int16_t bytes, @@ -80,9 +79,9 @@ static void setColumnOffsetValueInResultset(SQueryInfo* pQueryInfo); static int32_t parseGroupbyClause(SQueryInfo* pQueryInfo, tVariantList* pList, SSqlCmd* pCmd); -static int32_t parseIntervalClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql); +static int32_t parseIntervalClause(SSqlObj* pSql, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql); static int32_t parseOffsetClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql); -static int32_t parseSlidingClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql); +static int32_t parseSlidingClause(SSqlObj* pSql, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql); static int32_t addProjectionExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, tSQLExprItem* pItem); @@ -114,7 +113,7 @@ static int32_t optrToString(tSQLExpr* pExpr, char** exprString); static int32_t getTableIndexImpl(SStrToken* pTableToken, SQueryInfo* pQueryInfo, SColumnIndex* pIndex); static int32_t doFunctionsCompatibleCheck(SSqlCmd* pCmd, SQueryInfo* pQueryInfo); static int32_t doLocalQueryProcess(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql); -static int32_t tscCheckCreateDbParams(SSqlCmd* pCmd, SCMCreateDbMsg* pCreate); +static int32_t tscCheckCreateDbParams(SSqlCmd* pCmd, SCreateDbMsg* pCreate); static SColumnList getColumnList(int32_t num, int16_t tableIndex, int32_t columnIndex); @@ -417,7 +416,7 @@ int32_t tscToSQLCmd(SSqlObj* pSql, struct SSqlInfo* pInfo) { char* pMsg = pCmd->payload; - SCMCfgDnodeMsg* pCfg = (SCMCfgDnodeMsg*)pMsg; + SCfgDnodeMsg* pCfg = (SCfgDnodeMsg*)pMsg; pDCL->a[0].n = strdequote(pDCL->a[0].z); strncpy(pCfg->ep, pDCL->a[0].z, pDCL->a[0].n); @@ -616,14 +615,20 @@ static bool isTopBottomQuery(SQueryInfo* pQueryInfo) { return false; } -int32_t parseIntervalClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql) { +int32_t parseIntervalClause(SSqlObj* pSql, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql) { const char* msg1 = "invalid query expression"; const char* msg2 = "interval cannot be less than 10 ms"; + const char* msg3 = "sliding cannot be used without interval"; + + SSqlCmd* pCmd = &pSql->cmd; STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); STableComInfo tinfo = tscGetTableInfo(pTableMetaInfo->pTableMeta); if (pQuerySql->interval.type == 0 || pQuerySql->interval.n == 0) { + if (pQuerySql->sliding.n > 0) { + return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg3); + } return TSDB_CODE_SUCCESS; } @@ -656,7 +661,7 @@ int32_t parseIntervalClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQ return TSDB_CODE_TSC_INVALID_SQL; } - if (parseSlidingClause(pCmd, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { + if (parseSlidingClause(pSql, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { return TSDB_CODE_TSC_INVALID_SQL; } @@ -709,7 +714,7 @@ int32_t parseIntervalClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQ return TSDB_CODE_TSC_INVALID_SQL; } - if (parseSlidingClause(pCmd, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { + if (parseSlidingClause(pSql, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { return TSDB_CODE_TSC_INVALID_SQL; } @@ -767,13 +772,15 @@ int32_t parseOffsetClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQue return TSDB_CODE_SUCCESS; } -int32_t parseSlidingClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql) { +int32_t parseSlidingClause(SSqlObj* pSql, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql) { const char* msg0 = "sliding value too small"; const char* msg1 = "sliding value no larger than the interval value"; const char* msg2 = "sliding value can not less than 1% of interval value"; - const char* msg3 = "does not support sliding when interval is natual month/year"; + const char* msg3 = "does not support sliding when interval is natural month/year"; + const char* msg4 = "sliding not support yet in ordinary query"; const static int32_t INTERVAL_SLIDING_FACTOR = 100; + SSqlCmd* pCmd = &pSql->cmd; STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); STableComInfo tinfo = tscGetTableInfo(pTableMetaInfo->pTableMeta); @@ -806,6 +813,10 @@ int32_t parseSlidingClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQu return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg2); } + if (pQueryInfo->interval.sliding != pQueryInfo->interval.interval && pSql->pStream == NULL) { + return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg4); + } + return TSDB_CODE_SUCCESS; } @@ -1211,7 +1222,7 @@ static void tscInsertPrimaryTSSourceColumn(SQueryInfo* pQueryInfo, SColumnIndex* tscColumnListInsert(pQueryInfo->colList, &tsCol); } static int32_t handleArithmeticExpr(SSqlCmd* pCmd, int32_t clauseIndex, int32_t exprIndex, tSQLExprItem* pItem) { - const char* msg1 = "invalid column name, or illegal column type"; + const char* msg1 = "invalid column name, illegal column type, or columns in arithmetic expression from two tables"; const char* msg2 = "invalid arithmetic expression in select clause"; const char* msg3 = "tag columns can not be used in arithmetic expression"; const char* msg4 = "columns from different table mixed up in arithmetic expression"; @@ -1625,16 +1636,15 @@ int32_t addProjectionExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, t } static int32_t setExprInfoForFunctions(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SSchema* pSchema, SConvertFunc cvtFunc, - char* aliasName, int32_t resColIdx, SColumnIndex* pColIndex, bool finalResult) { + const char* name, int32_t resColIdx, SColumnIndex* pColIndex, bool finalResult) { const char* msg1 = "not support column types"; int16_t type = 0; int16_t bytes = 0; - char columnName[TSDB_COL_NAME_LEN] = {0}; int32_t functionID = cvtFunc.execFuncId; if (functionID == TSDB_FUNC_SPREAD) { - int32_t t1 = pSchema[pColIndex->columnIndex].type; + int32_t t1 = pSchema->type; if (t1 == TSDB_DATA_TYPE_BINARY || t1 == TSDB_DATA_TYPE_NCHAR || t1 == TSDB_DATA_TYPE_BOOL) { invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg1); return -1; @@ -1643,18 +1653,12 @@ static int32_t setExprInfoForFunctions(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SS bytes = tDataTypeDesc[type].nSize; } } else { - type = pSchema[pColIndex->columnIndex].type; - bytes = pSchema[pColIndex->columnIndex].bytes; + type = pSchema->type; + bytes = pSchema->bytes; } - if (aliasName != NULL) { - tstrncpy(columnName, aliasName, sizeof(columnName)); - } else { - getRevisedName(columnName, cvtFunc.originFuncId, sizeof(columnName) - 1, pSchema[pColIndex->columnIndex].name); - } - SSqlExpr* pExpr = tscSqlExprAppend(pQueryInfo, functionID, pColIndex, type, bytes, bytes, false); - tstrncpy(pExpr->aliasName, columnName, sizeof(pExpr->aliasName)); + tstrncpy(pExpr->aliasName, name, tListLen(pExpr->aliasName)); if (cvtFunc.originFuncId == TSDB_FUNC_LAST_ROW && cvtFunc.originFuncId != functionID) { pExpr->colInfo.flag |= TSDB_COL_NULL; @@ -1674,7 +1678,7 @@ static int32_t setExprInfoForFunctions(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SS // if it is not in the final result, do not add it SColumnList ids = getColumnList(1, pColIndex->tableIndex, pColIndex->columnIndex); if (finalResult) { - insertResultField(pQueryInfo, resColIdx, &ids, bytes, (int8_t)type, columnName, pExpr); + insertResultField(pQueryInfo, resColIdx, &ids, bytes, (int8_t)type, pExpr->aliasName, pExpr); } else { tscColumnListInsert(pQueryInfo->colList, &(ids.ids[0])); } @@ -1682,6 +1686,23 @@ static int32_t setExprInfoForFunctions(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SS return TSDB_CODE_SUCCESS; } +void setResultColName(char* name, tSQLExprItem* pItem, int32_t functionId, SStrToken* pToken) { + if (pItem->aliasName != NULL) { + tstrncpy(name, pItem->aliasName, TSDB_COL_NAME_LEN); + } else { + char uname[TSDB_COL_NAME_LEN] = {0}; + int32_t len = MIN(pToken->n + 1, TSDB_COL_NAME_LEN); + tstrncpy(uname, pToken->z, len); + + int32_t size = TSDB_COL_NAME_LEN + tListLen(aAggs[functionId].aName) + 2 + 1; + char tmp[TSDB_COL_NAME_LEN + tListLen(aAggs[functionId].aName) + 2 + 1] = {0}; + + snprintf(tmp, size, "%s(%s)", aAggs[functionId].aName, uname); + + tstrncpy(name, tmp, TSDB_COL_NAME_LEN); + } +} + int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t colIndex, tSQLExprItem* pItem, bool finalResult) { STableMetaInfo* pTableMetaInfo = NULL; int32_t optr = pItem->pNode->nSQLOptr; @@ -1939,8 +1960,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col SColumnIndex index = COLUMN_INDEX_INITIALIZER; - if (pParamElem->pNode->nSQLOptr == TK_ALL) { - // select table.* + if (pParamElem->pNode->nSQLOptr == TK_ALL) { // select table.* SStrToken tmpToken = pParamElem->pNode->colInfo; if (getTableIndexByName(&tmpToken, pQueryInfo, &index) != TSDB_CODE_SUCCESS) { @@ -1950,9 +1970,13 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col pTableMetaInfo = tscGetMetaInfo(pQueryInfo, index.tableIndex); SSchema* pSchema = tscGetTableSchema(pTableMetaInfo->pTableMeta); + char name[TSDB_COL_NAME_LEN] = {0}; for (int32_t j = 0; j < tscGetNumOfColumns(pTableMetaInfo->pTableMeta); ++j) { index.columnIndex = j; - if (setExprInfoForFunctions(pCmd, pQueryInfo, pSchema, cvtFunc, pItem->aliasName, colIndex++, &index, finalResult) != 0) { + SStrToken t = {.z = pSchema[j].name, .n = (uint32_t)strnlen(pSchema[j].name, TSDB_COL_NAME_LEN)}; + setResultColName(name, pItem, cvtFunc.originFuncId, &t); + + if (setExprInfoForFunctions(pCmd, pQueryInfo, &pSchema[j], cvtFunc, name, colIndex++, &index, finalResult) != 0) { return TSDB_CODE_TSC_INVALID_SQL; } } @@ -1963,14 +1987,18 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col } pTableMetaInfo = tscGetMetaInfo(pQueryInfo, index.tableIndex); - SSchema* pSchema = tscGetTableSchema(pTableMetaInfo->pTableMeta); // functions can not be applied to tags if ((index.columnIndex >= tscGetNumOfColumns(pTableMetaInfo->pTableMeta)) || (index.columnIndex < 0)) { return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg6); } - if (setExprInfoForFunctions(pCmd, pQueryInfo, pSchema, cvtFunc, pItem->aliasName, colIndex + i, &index, finalResult) != 0) { + char name[TSDB_COL_NAME_LEN] = {0}; + + SSchema* pSchema = tscGetTableColumnSchema(pTableMetaInfo->pTableMeta, index.columnIndex); + setResultColName(name, pItem, cvtFunc.originFuncId, &pParamElem->pNode->colInfo); + + if (setExprInfoForFunctions(pCmd, pQueryInfo, pSchema, cvtFunc, name, colIndex + i, &index, finalResult) != 0) { return TSDB_CODE_TSC_INVALID_SQL; } @@ -2007,7 +2035,12 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col for (int32_t i = 0; i < tscGetNumOfColumns(pTableMetaInfo->pTableMeta); ++i) { SColumnIndex index = {.tableIndex = j, .columnIndex = i}; - if (setExprInfoForFunctions(pCmd, pQueryInfo, pSchema, cvtFunc, pItem->aliasName, colIndex, &index, finalResult) != 0) { + + char name[TSDB_COL_NAME_LEN] = {0}; + SStrToken t = {.z = pSchema->name, .n = (uint32_t)strnlen(pSchema->name, TSDB_COL_NAME_LEN)}; + setResultColName(name, pItem, cvtFunc.originFuncId, &t); + + if (setExprInfoForFunctions(pCmd, pQueryInfo, &pSchema[index.columnIndex], cvtFunc, name, colIndex, &index, finalResult) != 0) { return TSDB_CODE_TSC_INVALID_SQL; } @@ -2240,10 +2273,6 @@ void getColumnName(tSQLExprItem* pItem, char* resultFieldName, int32_t nameLengt } } -void getRevisedName(char* resultFieldName, int32_t functionId, int32_t maxLen, char* columnName) { - snprintf(resultFieldName, maxLen, "%s(%s)", aAggs[functionId].aName, columnName); -} - static bool isTablenameToken(SStrToken* token) { SStrToken tmpToken = *token; SStrToken tableToken = {0}; @@ -2728,6 +2757,7 @@ static bool functionCompatibleCheck(SQueryInfo* pQueryInfo, bool joinQuery) { int32_t parseGroupbyClause(SQueryInfo* pQueryInfo, tVariantList* pList, SSqlCmd* pCmd) { const char* msg1 = "too many columns in group by clause"; const char* msg2 = "invalid column name in group by clause"; + const char* msg3 = "columns from one table allowed as group by columns"; const char* msg7 = "not support group by expression"; const char* msg8 = "not allowed column type for group by"; const char* msg9 = "tags not allowed for table query"; @@ -2763,7 +2793,11 @@ int32_t parseGroupbyClause(SQueryInfo* pQueryInfo, tVariantList* pList, SSqlCmd* return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg2); } - tableIndex = index.tableIndex; + if (tableIndex == COLUMN_INDEX_INITIAL_VAL) { + tableIndex = index.tableIndex; + } else if (tableIndex != index.tableIndex) { + return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg3); + } pTableMetaInfo = tscGetMetaInfo(pQueryInfo, index.tableIndex); pTableMeta = pTableMetaInfo->pTableMeta; @@ -3348,7 +3382,8 @@ int32_t doArithmeticExprToString(tSQLExpr* pExpr, char** exprString) { return TSDB_CODE_SUCCESS; } -static int32_t validateSQLExpr(SSqlCmd* pCmd, tSQLExpr* pExpr, SQueryInfo* pQueryInfo, SColumnList* pList, int32_t* type) { +static int32_t validateSQLExpr(SSqlCmd* pCmd, tSQLExpr* pExpr, SQueryInfo* pQueryInfo, SColumnList* pList, + int32_t* type, uint64_t* uid) { if (pExpr->nSQLOptr == TK_ID) { if (*type == NON_ARITHMEIC_EXPR) { *type = NORMAL_ARITHMETIC; @@ -3397,13 +3432,22 @@ static int32_t validateSQLExpr(SSqlCmd* pCmd, tSQLExpr* pExpr, SQueryInfo* pQuer } // Not supported data type in arithmetic expression + uint64_t id = -1; for(int32_t i = 0; i < inc; ++i) { SSqlExpr* p1 = tscSqlExprGet(pQueryInfo, i + outputIndex); int16_t t = p1->resType; if (t == TSDB_DATA_TYPE_BINARY || t == TSDB_DATA_TYPE_NCHAR || t == TSDB_DATA_TYPE_BOOL || t == TSDB_DATA_TYPE_TIMESTAMP) { return TSDB_CODE_TSC_INVALID_SQL; } + + if (i == 0) { + id = p1->uid; + } else if (id != p1->uid){ + return TSDB_CODE_TSC_INVALID_SQL; + } } + + *uid = id; } return TSDB_CODE_SUCCESS; @@ -3415,13 +3459,16 @@ static int32_t validateArithmeticSQLExpr(SSqlCmd* pCmd, tSQLExpr* pExpr, SQueryI } tSQLExpr* pLeft = pExpr->pLeft; + uint64_t uidLeft = 0; + uint64_t uidRight = 0; + if (pLeft->nSQLOptr >= TK_PLUS && pLeft->nSQLOptr <= TK_REM) { int32_t ret = validateArithmeticSQLExpr(pCmd, pLeft, pQueryInfo, pList, type); if (ret != TSDB_CODE_SUCCESS) { return ret; } } else { - int32_t ret = validateSQLExpr(pCmd, pLeft, pQueryInfo, pList, type); + int32_t ret = validateSQLExpr(pCmd, pLeft, pQueryInfo, pList, type, &uidLeft); if (ret != TSDB_CODE_SUCCESS) { return ret; } @@ -3434,10 +3481,15 @@ static int32_t validateArithmeticSQLExpr(SSqlCmd* pCmd, tSQLExpr* pExpr, SQueryI return ret; } } else { - int32_t ret = validateSQLExpr(pCmd, pRight, pQueryInfo, pList, type); + int32_t ret = validateSQLExpr(pCmd, pRight, pQueryInfo, pList, type, &uidRight); if (ret != TSDB_CODE_SUCCESS) { return ret; } + + // the expression not from the same table, return error + if (uidLeft != uidRight && uidLeft != 0 && uidRight != 0) { + return TSDB_CODE_TSC_INVALID_SQL; + } } return TSDB_CODE_SUCCESS; @@ -4423,8 +4475,8 @@ static void setDefaultOrderInfo(SQueryInfo* pQueryInfo) { int32_t parseOrderbyClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql, SSchema* pSchema) { const char* msg0 = "only support order by primary timestamp"; const char* msg1 = "invalid column name"; - const char* msg2 = "only support order by primary timestamp and queried column"; - const char* msg3 = "only support order by primary timestamp and first tag in groupby clause"; + const char* msg2 = "only support order by primary timestamp or queried column"; + const char* msg3 = "only support order by primary timestamp or first tag in groupby clause"; setDefaultOrderInfo(pQueryInfo); STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); @@ -5185,7 +5237,7 @@ int32_t parseLimitClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t clauseIn return TSDB_CODE_SUCCESS; } -static int32_t setKeepOption(SSqlCmd* pCmd, SCMCreateDbMsg* pMsg, SCreateDBInfo* pCreateDb) { +static int32_t setKeepOption(SSqlCmd* pCmd, SCreateDbMsg* pMsg, SCreateDBInfo* pCreateDb) { const char* msg = "invalid number of options"; pMsg->daysToKeep = htonl(-1); @@ -5216,7 +5268,7 @@ static int32_t setKeepOption(SSqlCmd* pCmd, SCMCreateDbMsg* pMsg, SCreateDBInfo* return TSDB_CODE_SUCCESS; } -static int32_t setTimePrecision(SSqlCmd* pCmd, SCMCreateDbMsg* pMsg, SCreateDBInfo* pCreateDbInfo) { +static int32_t setTimePrecision(SSqlCmd* pCmd, SCreateDbMsg* pMsg, SCreateDBInfo* pCreateDbInfo) { const char* msg = "invalid time precision"; pMsg->precision = TSDB_TIME_PRECISION_MILLI; // millisecond by default @@ -5240,7 +5292,7 @@ static int32_t setTimePrecision(SSqlCmd* pCmd, SCMCreateDbMsg* pMsg, SCreateDBIn return TSDB_CODE_SUCCESS; } -static void setCreateDBOption(SCMCreateDbMsg* pMsg, SCreateDBInfo* pCreateDb) { +static void setCreateDBOption(SCreateDbMsg* pMsg, SCreateDBInfo* pCreateDb) { pMsg->maxTables = htonl(-1); // max tables can not be set anymore pMsg->cacheBlockSize = htonl(pCreateDb->cacheBlockSize); pMsg->totalBlocks = htonl(pCreateDb->numOfBlocks); @@ -5257,7 +5309,7 @@ static void setCreateDBOption(SCMCreateDbMsg* pMsg, SCreateDBInfo* pCreateDb) { } int32_t parseCreateDBOptions(SSqlCmd* pCmd, SCreateDBInfo* pCreateDbSql) { - SCMCreateDbMsg* pMsg = (SCMCreateDbMsg*)(pCmd->payload); + SCreateDbMsg* pMsg = (SCreateDbMsg *)(pCmd->payload); setCreateDBOption(pMsg, pCreateDbSql); if (setKeepOption(pCmd, pMsg, pCreateDbSql) != TSDB_CODE_SUCCESS) { @@ -5775,7 +5827,7 @@ int32_t doLocalQueryProcess(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQuerySQL* pQ } // can only perform the parameters based on the macro definitation -int32_t tscCheckCreateDbParams(SSqlCmd* pCmd, SCMCreateDbMsg* pCreate) { +int32_t tscCheckCreateDbParams(SSqlCmd* pCmd, SCreateDbMsg* pCreate) { char msg[512] = {0}; if (pCreate->walLevel != -1 && (pCreate->walLevel < TSDB_MIN_WAL_LEVEL || pCreate->walLevel > TSDB_MAX_WAL_LEVEL)) { @@ -6099,7 +6151,7 @@ int32_t doCheckForStream(SSqlObj* pSql, SSqlInfo* pInfo) { } // set interval value - if (parseIntervalClause(pCmd, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { + if (parseIntervalClause(pSql, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { return TSDB_CODE_TSC_INVALID_SQL; } else { if ((pQueryInfo->interval.interval > 0) && @@ -6307,7 +6359,7 @@ int32_t doCheckForQuery(SSqlObj* pSql, SQuerySQL* pQuerySql, int32_t index) { } // set interval value - if (parseIntervalClause(pCmd, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { + if (parseIntervalClause(pSql, pQueryInfo, pQuerySql) != TSDB_CODE_SUCCESS) { return TSDB_CODE_TSC_INVALID_SQL; } else { if ((pQueryInfo->interval.interval > 0) && diff --git a/src/client/src/tscSchemaUtil.c b/src/client/src/tscSchemaUtil.c index 2a9a9de84b6b4377c0ca99ee4e5594d234a2d1e4..441ad8d8382c667c164f7c797fc344cb9ab9118d 100644 --- a/src/client/src/tscSchemaUtil.c +++ b/src/client/src/tscSchemaUtil.c @@ -140,7 +140,7 @@ struct SSchema tscGetTbnameColumnSchema() { strcpy(s.name, TSQL_TBNAME_L); return s; } -static void tscInitCorVgroupInfo(SCMCorVgroupInfo *corVgroupInfo, SCMVgroupInfo *vgroupInfo) { +static void tscInitCorVgroupInfo(SCorVgroupInfo *corVgroupInfo, SVgroupInfo *vgroupInfo) { corVgroupInfo->version = 0; corVgroupInfo->inUse = 0; corVgroupInfo->numOfEps = vgroupInfo->numOfEps; @@ -166,7 +166,7 @@ STableMeta* tscCreateTableMetaFromMsg(STableMetaMsg* pTableMetaMsg, size_t* size pTableMeta->id.tid = pTableMetaMsg->tid; pTableMeta->id.uid = pTableMetaMsg->uid; - SCMVgroupInfo* pVgroupInfo = &pTableMeta->vgroupInfo; + SVgroupInfo* pVgroupInfo = &pTableMeta->vgroupInfo; pVgroupInfo->numOfEps = pTableMetaMsg->vgroup.numOfEps; pVgroupInfo->vgId = pTableMetaMsg->vgroup.vgId; @@ -197,28 +197,6 @@ STableMeta* tscCreateTableMetaFromMsg(STableMetaMsg* pTableMetaMsg, size_t* size return pTableMeta; } -/** - * the TableMeta data format in memory is as follows: - * - * +--------------------+ - * |STableMeta Body data| sizeof(STableMeta) - * +--------------------+ - * |Schema data | numOfTotalColumns * sizeof(SSchema) - * +--------------------+ - * |Tags data | tag_col_1.bytes + tag_col_2.bytes + .... - * +--------------------+ - * - * @param pTableMeta - * @return - */ -char* tsGetTagsValue(STableMeta* pTableMeta) { - int32_t offset = 0; -// int32_t numOfTotalCols = pTableMeta->numOfColumns + pTableMeta->numOfTags; -// uint32_t offset = sizeof(STableMeta) + numOfTotalCols * sizeof(SSchema); - - return ((char*)pTableMeta + offset); -} - // todo refactor UNUSED_FUNC static FORCE_INLINE char* skipSegments(char* input, char delim, int32_t num) { for (int32_t i = 0; i < num; ++i) { diff --git a/src/client/src/tscServer.c b/src/client/src/tscServer.c index 2e7a99f2fb89bf9cddc6f45217197b77f9eb8df5..1584afe706e21d1a16e4ab914297aadaccbbf1b3 100644 --- a/src/client/src/tscServer.c +++ b/src/client/src/tscServer.c @@ -48,7 +48,7 @@ static int32_t getWaitingTimeInterval(int32_t count) { return initial * (2<<(count - 2)); } -static void tscSetDnodeEpSet(SSqlObj* pSql, SCMVgroupInfo* pVgroupInfo) { +static void tscSetDnodeEpSet(SSqlObj* pSql, SVgroupInfo* pVgroupInfo) { assert(pSql != NULL && pVgroupInfo != NULL && pVgroupInfo->numOfEps > 0); SRpcEpSet* pEpSet = &pSql->epSet; @@ -100,7 +100,7 @@ void tscUpdateMgmtEpSet(SRpcEpSet *pEpSet) { tscMgmtEpSet.epSet = *pEpSet; taosCorEndWrite(&tscMgmtEpSet.version); } -static void tscDumpEpSetFromVgroupInfo(SCMCorVgroupInfo *pVgroupInfo, SRpcEpSet *pEpSet) { +static void tscDumpEpSetFromVgroupInfo(SCorVgroupInfo *pVgroupInfo, SRpcEpSet *pEpSet) { if (pVgroupInfo == NULL) { return;} taosCorBeginRead(&pVgroupInfo->version); int8_t inUse = pVgroupInfo->inUse; @@ -117,7 +117,7 @@ static void tscUpdateVgroupInfo(SSqlObj *pObj, SRpcEpSet *pEpSet) { SSqlCmd *pCmd = &pObj->cmd; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); if (pTableMetaInfo == NULL || pTableMetaInfo->pTableMeta == NULL) { return;} - SCMCorVgroupInfo *pVgroupInfo = &pTableMetaInfo->pTableMeta->corVgroupInfo; + SCorVgroupInfo *pVgroupInfo = &pTableMetaInfo->pTableMeta->corVgroupInfo; taosCorBeginWrite(&pVgroupInfo->version); tscDebug("before: Endpoint in use: %d", pVgroupInfo->inUse); @@ -158,12 +158,12 @@ void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) { SSqlRes *pRes = &pSql->res; if (code == 0) { - SCMHeartBeatRsp *pRsp = (SCMHeartBeatRsp *)pRes->pRsp; - SRpcEpSet * epSet = &pRsp->epSet; + SHeartBeatRsp *pRsp = (SHeartBeatRsp *)pRes->pRsp; + SRpcEpSet * epSet = &pRsp->epSet; if (epSet->numOfEps > 0) { tscEpSetHtons(epSet); tscUpdateMgmtEpSet(epSet); - } + } pSql->pTscObj->connId = htonl(pRsp->connId); @@ -584,7 +584,7 @@ static char *doSerializeTableInfo(SQueryTableMsg* pQueryMsg, SSqlObj *pSql, char STableMeta * pTableMeta = pTableMetaInfo->pTableMeta; if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo) || pTableMetaInfo->pVgroupTables == NULL) { - SCMVgroupInfo* pVgroupInfo = NULL; + SVgroupInfo* pVgroupInfo = NULL; if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { int32_t index = pTableMetaInfo->vgroupIndex; assert(index >= 0); @@ -878,37 +878,20 @@ int tscBuildQueryMsg(SSqlObj *pSql, SSqlInfo *pInfo) { // compressed ts block pQueryMsg->tsOffset = htonl((int32_t)(pMsg - pCmd->payload)); - int32_t tsLen = 0; - int32_t numOfBlocks = 0; if (pQueryInfo->tsBuf != NULL) { - int32_t vnodeId = htonl(pQueryMsg->head.vgId); - STSVnodeBlockInfo *pBlockInfo = tsBufGetVnodeBlockInfo(pQueryInfo->tsBuf, vnodeId); - assert(QUERY_IS_JOIN_QUERY(pQueryInfo->type) && pBlockInfo != NULL); // this query should not be sent - - // todo refactor - if (fseek(pQueryInfo->tsBuf->f, pBlockInfo->offset, SEEK_SET) != 0) { - int code = TAOS_SYSTEM_ERROR(ferror(pQueryInfo->tsBuf->f)); - tscError("%p: fseek failed: %s", pSql, tstrerror(code)); + // note: here used the index instead of actual vnode id. + int32_t vnodeIndex = pTableMetaInfo->vgroupIndex; + int32_t code = dumpFileBlockByVnodeId(pQueryInfo->tsBuf, vnodeIndex, pMsg, &pQueryMsg->tsLen, &pQueryMsg->tsNumOfBlocks); + if (code != TSDB_CODE_SUCCESS) { return code; } - size_t s = fread(pMsg, 1, pBlockInfo->compLen, pQueryInfo->tsBuf->f); - if (s != pBlockInfo->compLen) { - int code = TAOS_SYSTEM_ERROR(ferror(pQueryInfo->tsBuf->f)); - tscError("%p: fread didn't return expected data: %s", pSql, tstrerror(code)); - return code; - } + pMsg += pQueryMsg->tsLen; - pMsg += pBlockInfo->compLen; - tsLen = pBlockInfo->compLen; - numOfBlocks = pBlockInfo->numOfBlocks; - } - - pQueryMsg->tsLen = htonl(tsLen); - pQueryMsg->tsNumOfBlocks = htonl(numOfBlocks); - if (pQueryInfo->tsBuf != NULL) { pQueryMsg->tsOrder = htonl(pQueryInfo->tsBuf->tsOrder); + pQueryMsg->tsLen = htonl(pQueryMsg->tsLen); + pQueryMsg->tsNumOfBlocks = htonl(pQueryMsg->tsNumOfBlocks); } int32_t msgLen = (int32_t)(pMsg - pCmd->payload); @@ -925,10 +908,10 @@ int tscBuildQueryMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildCreateDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMCreateDbMsg); + pCmd->payloadLen = sizeof(SCreateDbMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_CREATE_DB; - SCMCreateDbMsg *pCreateDbMsg = (SCMCreateDbMsg*)pCmd->payload; + SCreateDbMsg *pCreateDbMsg = (SCreateDbMsg *)pCmd->payload; assert(pCmd->numOfClause == 1); STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); @@ -939,13 +922,13 @@ int32_t tscBuildCreateDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildCreateDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMCreateDnodeMsg); + pCmd->payloadLen = sizeof(SCreateDnodeMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMCreateDnodeMsg *pCreate = (SCMCreateDnodeMsg *)pCmd->payload; + SCreateDnodeMsg *pCreate = (SCreateDnodeMsg *)pCmd->payload; strncpy(pCreate->ep, pInfo->pDCLInfo->a[0].z, pInfo->pDCLInfo->a[0].n); pCmd->msgType = TSDB_MSG_TYPE_CM_CREATE_DNODE; @@ -955,13 +938,13 @@ int32_t tscBuildCreateDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildAcctMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMCreateAcctMsg); + pCmd->payloadLen = sizeof(SCreateAcctMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMCreateAcctMsg *pAlterMsg = (SCMCreateAcctMsg *)pCmd->payload; + SCreateAcctMsg *pAlterMsg = (SCreateAcctMsg *)pCmd->payload; SStrToken *pName = &pInfo->pDCLInfo->user.user; SStrToken *pPwd = &pInfo->pDCLInfo->user.passwd; @@ -1000,14 +983,14 @@ int32_t tscBuildAcctMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildUserMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMCreateUserMsg); + pCmd->payloadLen = sizeof(SCreateUserMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMCreateUserMsg *pAlterMsg = (SCMCreateUserMsg*)pCmd->payload; + SCreateUserMsg *pAlterMsg = (SCreateUserMsg *)pCmd->payload; SUserInfo *pUser = &pInfo->pDCLInfo->user; strncpy(pAlterMsg->user, pUser->user.z, pUser->user.n); @@ -1032,21 +1015,21 @@ int32_t tscBuildUserMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildCfgDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMCfgDnodeMsg); + pCmd->payloadLen = sizeof(SCfgDnodeMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_CONFIG_DNODE; return TSDB_CODE_SUCCESS; } int32_t tscBuildDropDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMDropDbMsg); + pCmd->payloadLen = sizeof(SDropDbMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMDropDbMsg *pDropDbMsg = (SCMDropDbMsg*)pCmd->payload; + SDropDbMsg *pDropDbMsg = (SDropDbMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDropDbMsg->db, pTableMetaInfo->name, sizeof(pDropDbMsg->db)); @@ -1076,13 +1059,13 @@ int32_t tscBuildDropTableMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildDropDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMDropDnodeMsg); + pCmd->payloadLen = sizeof(SDropDnodeMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMDropDnodeMsg *pDrop = (SCMDropDnodeMsg *)pCmd->payload; + SDropDnodeMsg * pDrop = (SDropDnodeMsg *)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDrop->ep, pTableMetaInfo->name, sizeof(pDrop->ep)); pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_DNODE; @@ -1092,7 +1075,7 @@ int32_t tscBuildDropDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildDropUserMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMDropUserMsg); + pCmd->payloadLen = sizeof(SDropUserMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_USER; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { @@ -1100,7 +1083,7 @@ int32_t tscBuildDropUserMsg(SSqlObj *pSql, SSqlInfo *pInfo) { return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMDropUserMsg *pDropMsg = (SCMDropUserMsg*)pCmd->payload; + SDropUserMsg * pDropMsg = (SDropUserMsg *)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDropMsg->user, pTableMetaInfo->name, sizeof(pDropMsg->user)); @@ -1109,7 +1092,7 @@ int32_t tscBuildDropUserMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildDropAcctMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMDropUserMsg); + pCmd->payloadLen = sizeof(SDropUserMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_ACCT; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { @@ -1117,7 +1100,7 @@ int32_t tscBuildDropAcctMsg(SSqlObj *pSql, SSqlInfo *pInfo) { return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMDropUserMsg *pDropMsg = (SCMDropUserMsg*)pCmd->payload; + SDropUserMsg * pDropMsg = (SDropUserMsg *)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDropMsg->user, pTableMetaInfo->name, sizeof(pDropMsg->user)); @@ -1126,14 +1109,14 @@ int32_t tscBuildDropAcctMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int32_t tscBuildUseDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMUseDbMsg); + pCmd->payloadLen = sizeof(SUseDbMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMUseDbMsg *pUseDbMsg = (SCMUseDbMsg*)pCmd->payload; + SUseDbMsg *pUseDbMsg = (SUseDbMsg *)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); strcpy(pUseDbMsg->db, pTableMetaInfo->name); pCmd->msgType = TSDB_MSG_TYPE_CM_USE_DB; @@ -1145,14 +1128,14 @@ int32_t tscBuildShowMsg(SSqlObj *pSql, SSqlInfo *pInfo) { STscObj *pObj = pSql->pTscObj; SSqlCmd *pCmd = &pSql->cmd; pCmd->msgType = TSDB_MSG_TYPE_CM_SHOW; - pCmd->payloadLen = sizeof(SCMShowMsg) + 100; + pCmd->payloadLen = sizeof(SShowMsg) + 100; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMShowMsg *pShowMsg = (SCMShowMsg*)pCmd->payload; + SShowMsg *pShowMsg = (SShowMsg *)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); size_t nameLen = strlen(pTableMetaInfo->name); @@ -1179,13 +1162,13 @@ int32_t tscBuildShowMsg(SSqlObj *pSql, SSqlInfo *pInfo) { pShowMsg->payloadLen = htons(pEpAddr->n); } - pCmd->payloadLen = sizeof(SCMShowMsg) + pShowMsg->payloadLen; + pCmd->payloadLen = sizeof(SShowMsg) + pShowMsg->payloadLen; return TSDB_CODE_SUCCESS; } int32_t tscBuildKillMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMKillQueryMsg); + pCmd->payloadLen = sizeof(SKillQueryMsg); switch (pCmd->command) { case TSDB_SQL_KILL_QUERY: @@ -1297,8 +1280,7 @@ int tscBuildCreateTableMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int tscEstimateAlterTableMsgLength(SSqlCmd *pCmd) { SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); - return minMsgSize() + sizeof(SCMAlterTableMsg) + sizeof(SSchema) * tscNumOfFields(pQueryInfo) + - TSDB_EXTRA_PAYLOAD_SIZE; + return minMsgSize() + sizeof(SAlterTableMsg) + sizeof(SSchema) * tscNumOfFields(pQueryInfo) + TSDB_EXTRA_PAYLOAD_SIZE; } int tscBuildAlterTableMsg(SSqlObj *pSql, SSqlInfo *pInfo) { @@ -1317,7 +1299,7 @@ int tscBuildAlterTableMsg(SSqlObj *pSql, SSqlInfo *pInfo) { return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMAlterTableMsg *pAlterTableMsg = (SCMAlterTableMsg *)pCmd->payload; + SAlterTableMsg *pAlterTableMsg = (SAlterTableMsg *)pCmd->payload; tscGetDBInfoFromTableFullName(pTableMetaInfo->name, pAlterTableMsg->db); strcpy(pAlterTableMsg->tableId, pTableMetaInfo->name); @@ -1366,10 +1348,10 @@ int tscBuildUpdateTagMsg(SSqlObj* pSql, SSqlInfo *pInfo) { int tscAlterDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; - pCmd->payloadLen = sizeof(SCMAlterDbMsg); + pCmd->payloadLen = sizeof(SAlterDbMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_ALTER_DB; - SCMAlterDbMsg *pAlterDbMsg = (SCMAlterDbMsg*)pCmd->payload; + SAlterDbMsg *pAlterDbMsg = (SAlterDbMsg* )pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pAlterDbMsg->db, pTableMetaInfo->name, sizeof(pAlterDbMsg->db)); @@ -1494,14 +1476,14 @@ int tscBuildConnectMsg(SSqlObj *pSql, SSqlInfo *pInfo) { STscObj *pObj = pSql->pTscObj; SSqlCmd *pCmd = &pSql->cmd; pCmd->msgType = TSDB_MSG_TYPE_CM_CONNECT; - pCmd->payloadLen = sizeof(SCMConnectMsg); + pCmd->payloadLen = sizeof(SConnectMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } - SCMConnectMsg *pConnect = (SCMConnectMsg*)pCmd->payload; + SConnectMsg *pConnect = (SConnectMsg*)pCmd->payload; // TODO refactor full_name char *db; // ugly code to move the space @@ -1523,11 +1505,11 @@ int tscBuildTableMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); - SCMTableInfoMsg* pInfoMsg = (SCMTableInfoMsg *)pCmd->payload; + STableInfoMsg *pInfoMsg = (STableInfoMsg *)pCmd->payload; strcpy(pInfoMsg->tableId, pTableMetaInfo->name); pInfoMsg->createFlag = htons(pSql->cmd.autoCreated ? 1 : 0); - char* pMsg = (char*)pInfoMsg + sizeof(SCMTableInfoMsg); + char *pMsg = (char *)pInfoMsg + sizeof(STableInfoMsg); size_t len = htonl(pCmd->tagData.dataLen); if (pSql->cmd.autoCreated) { @@ -1546,7 +1528,7 @@ int tscBuildTableMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { /** * multi table meta req pkg format: - * | SMgmtHead | SCMMultiTableInfoMsg | tableId0 | tableId1 | tableId2 | ...... + * | SMgmtHead | SMultiTableInfoMsg | tableId0 | tableId1 | tableId2 | ...... * no used 4B **/ int tscBuildMultiMeterMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { @@ -1564,7 +1546,7 @@ int tscBuildMultiMeterMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SMgmtHead *pMgmt = (SMgmtHead *)(pCmd->payload + tsRpcHeadSize); memset(pMgmt->db, 0, TSDB_TABLE_FNAME_LEN); // server don't need the db - SCMMultiTableInfoMsg *pInfoMsg = (SCMMultiTableInfoMsg *)(pCmd->payload + tsRpcHeadSize + sizeof(SMgmtHead)); + SMultiTableInfoMsg *pInfoMsg = (SMultiTableInfoMsg *)(pCmd->payload + tsRpcHeadSize + sizeof(SMgmtHead)); pInfoMsg->numOfTables = htonl((int32_t)pCmd->count); if (pCmd->payloadLen > 0) { @@ -1573,7 +1555,7 @@ int tscBuildMultiMeterMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { taosTFree(tmpData); - pCmd->payloadLen += sizeof(SMgmtHead) + sizeof(SCMMultiTableInfoMsg); + pCmd->payloadLen += sizeof(SMgmtHead) + sizeof(SMultiTableInfoMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_TABLES_META; assert(pCmd->payloadLen + minMsgSize() <= pCmd->allocSize); @@ -1618,12 +1600,12 @@ int tscBuildSTableVgroupMsg(SSqlObj *pSql, SSqlInfo *pInfo) { char* pMsg = pCmd->payload; SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); - - SCMSTableVgroupMsg *pStableVgroupMsg = (SCMSTableVgroupMsg *) pMsg; + + SSTableVgroupMsg *pStableVgroupMsg = (SSTableVgroupMsg *)pMsg; pStableVgroupMsg->numOfTables = htonl(pQueryInfo->numOfTables); - pMsg += sizeof(SCMSTableVgroupMsg); - - for(int32_t i = 0; i < pQueryInfo->numOfTables; ++i) { + pMsg += sizeof(SSTableVgroupMsg); + + for (int32_t i = 0; i < pQueryInfo->numOfTables; ++i) { STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, i); size_t size = sizeof(pTableMetaInfo->name); tstrncpy(pMsg, pTableMetaInfo->name, size); @@ -1656,7 +1638,7 @@ int tscBuildHeartBeatMsg(SSqlObj *pSql, SSqlInfo *pInfo) { numOfStreams++; } - int size = numOfQueries * sizeof(SQueryDesc) + numOfStreams * sizeof(SStreamDesc) + sizeof(SCMHeartBeatMsg) + 100; + int size = numOfQueries * sizeof(SQueryDesc) + numOfStreams * sizeof(SStreamDesc) + sizeof(SHeartBeatMsg) + 100; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, size)) { pthread_mutex_unlock(&pObj->mutex); tscError("%p failed to create heartbeat msg", pSql); @@ -1664,7 +1646,7 @@ int tscBuildHeartBeatMsg(SSqlObj *pSql, SSqlInfo *pInfo) { } // TODO the expired hb and client can not be identified by server till now. - SCMHeartBeatMsg *pHeartbeat = (SCMHeartBeatMsg *)pCmd->payload; + SHeartBeatMsg *pHeartbeat = (SHeartBeatMsg *)pCmd->payload; tstrncpy(pHeartbeat->clientVer, version, tListLen(pHeartbeat->clientVer)); pHeartbeat->numOfQueries = numOfQueries; @@ -1755,7 +1737,7 @@ int tscProcessTableMetaRsp(SSqlObj *pSql) { /** * multi table meta rsp pkg format: - * | STaosRsp | ieType | SCMMultiTableInfoMsg | SMeterMeta0 | SSchema0 | SMeterMeta1 | SSchema1 | SMeterMeta2 | SSchema2 + * | STaosRsp | ieType | SMultiTableInfoMsg | SMeterMeta0 | SSchema0 | SMeterMeta1 | SSchema1 | SMeterMeta2 | SSchema2 * |...... 1B 1B 4B **/ int tscProcessMultiMeterMetaRsp(SSqlObj *pSql) { @@ -1772,9 +1754,9 @@ int tscProcessMultiMeterMetaRsp(SSqlObj *pSql) { rsp++; - SCMMultiTableInfoMsg *pInfo = (SCMMultiTableInfoMsg *)rsp; + SMultiTableInfoMsg *pInfo = (SMultiTableInfoMsg *)rsp; totalNum = htonl(pInfo->numOfTables); - rsp += sizeof(SCMMultiTableInfoMsg); + rsp += sizeof(SMultiTableInfoMsg); for (i = 0; i < totalNum; i++) { SMultiTableMeta *pMultiMeta = (SMultiTableMeta *)rsp; @@ -1866,10 +1848,10 @@ int tscProcessSTableVgroupRsp(SSqlObj *pSql) { SSqlRes* pRes = &pSql->res; // NOTE: the order of several table must be preserved. - SCMSTableVgroupRspMsg *pStableVgroup = (SCMSTableVgroupRspMsg *)pRes->pRsp; + SSTableVgroupRspMsg *pStableVgroup = (SSTableVgroupRspMsg *)pRes->pRsp; pStableVgroup->numOfTables = htonl(pStableVgroup->numOfTables); - char* pMsg = pRes->pRsp + sizeof(SCMSTableVgroupRspMsg); - + char *pMsg = pRes->pRsp + sizeof(SSTableVgroupRspMsg); + // master sqlObj locates in param SSqlObj* parent = pSql->param; assert(parent != NULL); @@ -1881,18 +1863,18 @@ int tscProcessSTableVgroupRsp(SSqlObj *pSql) { SVgroupsMsg * pVgroupMsg = (SVgroupsMsg *) pMsg; pVgroupMsg->numOfVgroups = htonl(pVgroupMsg->numOfVgroups); - size_t size = sizeof(SCMVgroupMsg) * pVgroupMsg->numOfVgroups + sizeof(SVgroupsMsg); + size_t size = sizeof(SVgroupMsg) * pVgroupMsg->numOfVgroups + sizeof(SVgroupsMsg); - size_t vgroupsz = sizeof(SCMVgroupInfo) * pVgroupMsg->numOfVgroups + sizeof(SVgroupsInfo); + size_t vgroupsz = sizeof(SVgroupInfo) * pVgroupMsg->numOfVgroups + sizeof(SVgroupsInfo); pInfo->vgroupList = calloc(1, vgroupsz); assert(pInfo->vgroupList != NULL); pInfo->vgroupList->numOfVgroups = pVgroupMsg->numOfVgroups; for (int32_t j = 0; j < pInfo->vgroupList->numOfVgroups; ++j) { //just init, no need to lock - SCMVgroupInfo *pVgroups = &pInfo->vgroupList->vgroups[j]; + SVgroupInfo *pVgroups = &pInfo->vgroupList->vgroups[j]; - SCMVgroupMsg *vmsg = &pVgroupMsg->vgroups[j]; + SVgroupMsg *vmsg = &pVgroupMsg->vgroups[j]; pVgroups->vgId = htonl(vmsg->vgId); pVgroups->numOfEps = vmsg->numOfEps; @@ -1914,10 +1896,10 @@ int tscProcessSTableVgroupRsp(SSqlObj *pSql) { * current process do not use the cache at all */ int tscProcessShowRsp(SSqlObj *pSql) { - STableMetaMsg * pMetaMsg; - SCMShowRsp *pShow; - SSchema * pSchema; - char key[20]; + STableMetaMsg *pMetaMsg; + SShowRsp * pShow; + SSchema * pSchema; + char key[20]; SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; @@ -1926,7 +1908,7 @@ int tscProcessShowRsp(SSqlObj *pSql) { STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); - pShow = (SCMShowRsp *)pRes->pRsp; + pShow = (SShowRsp *)pRes->pRsp; pShow->qhandle = htobe64(pShow->qhandle); pRes->qhandle = pShow->qhandle; @@ -2024,7 +2006,7 @@ int tscProcessConnectRsp(SSqlObj *pSql) { char temp[TSDB_TABLE_FNAME_LEN * 2] = {0}; - SCMConnectRsp *pConnect = (SCMConnectRsp *)pRes->pRsp; + SConnectRsp *pConnect = (SConnectRsp *)pRes->pRsp; tstrncpy(pObj->acctId, pConnect->acctId, sizeof(pObj->acctId)); // copy acctId from response int32_t len = sprintf(temp, "%s%s%s", pObj->acctId, TS_PATH_DELIMITER, pObj->db); diff --git a/src/client/src/tscSubquery.c b/src/client/src/tscSubquery.c index c747eb2fe8405e899c67eaad9476badfda9b2077..de0d899f8c363b5a3dd226f3f5ab69eeb8fc3aff 100644 --- a/src/client/src/tscSubquery.c +++ b/src/client/src/tscSubquery.c @@ -32,11 +32,26 @@ typedef struct SInsertSupporter { static void freeJoinSubqueryObj(SSqlObj* pSql); static bool tscHasRemainDataInSubqueryResultSet(SSqlObj *pSql); -static bool tsCompare(int32_t order, int64_t left, int64_t right) { +static int32_t tsCompare(int32_t order, int64_t left, int64_t right) { + if (left == right) { + return 0; + } + if (order == TSDB_ORDER_ASC) { - return left < right; + return left < right? -1:1; } else { - return left > right; + return left > right? -1:1; + } +} + +static void skipRemainValue(STSBuf* pTSBuf, tVariant* tag1) { + while (tsBufNextPos(pTSBuf)) { + STSElem el1 = tsBufGetElem(pTSBuf); + + int32_t res = tVariantCompare(el1.tag, tag1); + if (res != 0) { // it is a record with new tag + return; + } } } @@ -51,10 +66,10 @@ static int64_t doTSBlockIntersect(SSqlObj* pSql, SJoinSupporter* pSupporter1, SJ SLimitVal* pLimit = &pQueryInfo->limit; int32_t order = pQueryInfo->order.order; - + SQueryInfo* pSubQueryInfo1 = tscGetQueryInfoDetail(&pSql->pSubs[0]->cmd, 0); SQueryInfo* pSubQueryInfo2 = tscGetQueryInfoDetail(&pSql->pSubs[1]->cmd, 0); - + pSubQueryInfo1->tsBuf = output1; pSubQueryInfo2->tsBuf = output2; @@ -88,59 +103,74 @@ static int64_t doTSBlockIntersect(SSqlObj* pSql, SJoinSupporter* pSupporter1, SJ int64_t numOfInput1 = 1; int64_t numOfInput2 = 1; - while (1) { - STSElem elem1 = tsBufGetElem(pSupporter1->pTSBuf); - STSElem elem2 = tsBufGetElem(pSupporter2->pTSBuf); + while(1) { + STSElem elem = tsBufGetElem(pSupporter1->pTSBuf); -#ifdef _DEBUG_VIEW - tscInfo("%" PRId64 ", tags:%"PRId64" \t %" PRId64 ", tags:%"PRId64, elem1.ts, elem1.tag.i64Key, elem2.ts, elem2.tag.i64Key); -#endif + // no data in pSupporter1 anymore, jump out of loop + if (!tsBufIsValidElem(&elem)) { + break; + } - int32_t res = tVariantCompare(elem1.tag, elem2.tag); - if (res == -1 || (res == 0 && tsCompare(order, elem1.ts, elem2.ts))) { - if (!tsBufNextPos(pSupporter1->pTSBuf)) { - break; - } + // find the data in supporter2 with the same tag value + STSElem e2 = tsBufFindElemStartPosByTag(pSupporter2->pTSBuf, elem.tag); - numOfInput1++; - } else if ((res > 0) || (res == 0 && tsCompare(order, elem2.ts, elem1.ts))) { - if (!tsBufNextPos(pSupporter2->pTSBuf)) { - break; - } + /** + * there are elements in pSupporter2 with the same tag, continue + */ + tVariant tag1 = {0}; + tVariantAssign(&tag1, elem.tag); - numOfInput2++; - } else { - /* - * in case of stable query, limit/offset is not applied here. the limit/offset is applied to the - * final results which is acquired after the secondry merge of in the client. - */ - if (pLimit->offset == 0 || pQueryInfo->interval.interval > 0 || QUERY_IS_STABLE_QUERY(pQueryInfo->type)) { - if (win->skey > elem1.ts) { - win->skey = elem1.ts; - } - - if (win->ekey < elem1.ts) { - win->ekey = elem1.ts; + if (tsBufIsValidElem(&e2)) { + while (1) { + STSElem elem1 = tsBufGetElem(pSupporter1->pTSBuf); + STSElem elem2 = tsBufGetElem(pSupporter2->pTSBuf); + + // data with current are exhausted + if (!tsBufIsValidElem(&elem1) || tVariantCompare(elem1.tag, &tag1) != 0) { + break; } - - tsBufAppend(output1, elem1.vnode, elem1.tag, (const char*)&elem1.ts, sizeof(elem1.ts)); - tsBufAppend(output2, elem2.vnode, elem2.tag, (const char*)&elem2.ts, sizeof(elem2.ts)); - } else { - pLimit->offset -= 1; - } + if (!tsBufIsValidElem(&elem2) || tVariantCompare(elem2.tag, &tag1) != 0) { // ignore all records with the same tag + skipRemainValue(pSupporter1->pTSBuf, &tag1); + break; + } - if (!tsBufNextPos(pSupporter1->pTSBuf)) { - break; - } + /* + * in case of stable query, limit/offset is not applied here. the limit/offset is applied to the + * final results which is acquired after the secondary merge of in the client. + */ + int32_t re = tsCompare(order, elem1.ts, elem2.ts); + if (re < 0) { + tsBufNextPos(pSupporter1->pTSBuf); + numOfInput1++; + } else if (re > 0) { + tsBufNextPos(pSupporter2->pTSBuf); + numOfInput2++; + } else { + if (pLimit->offset == 0 || pQueryInfo->interval.interval > 0 || QUERY_IS_STABLE_QUERY(pQueryInfo->type)) { + if (win->skey > elem1.ts) { + win->skey = elem1.ts; + } + + if (win->ekey < elem1.ts) { + win->ekey = elem1.ts; + } + + tsBufAppend(output1, elem1.vnode, elem1.tag, (const char*)&elem1.ts, sizeof(elem1.ts)); + tsBufAppend(output2, elem2.vnode, elem2.tag, (const char*)&elem2.ts, sizeof(elem2.ts)); + } else { + pLimit->offset -= 1;//offset apply to projection? + } - numOfInput1++; + tsBufNextPos(pSupporter1->pTSBuf); + numOfInput1++; - if (!tsBufNextPos(pSupporter2->pTSBuf)) { - break; + tsBufNextPos(pSupporter2->pTSBuf); + numOfInput2++; + } } - - numOfInput2++; + } else { // no data in pSupporter2, ignore current data in pSupporter2 + skipRemainValue(pSupporter1->pTSBuf, &tag1); } } @@ -162,8 +192,9 @@ static int64_t doTSBlockIntersect(SSqlObj* pSql, SJoinSupporter* pSupporter1, SJ TSKEY et = taosGetTimestampUs(); tscDebug("%p input1:%" PRId64 ", input2:%" PRId64 ", final:%" PRId64 " in %d vnodes for secondary query after ts blocks " - "intersecting, skey:%" PRId64 ", ekey:%" PRId64 ", numOfVnode:%d, elasped time:%"PRId64" us", pSql, numOfInput1, numOfInput2, output1->numOfTotal, - output1->numOfVnodes, win->skey, win->ekey, tsBufGetNumOfVnodes(output1), et - st); + "intersecting, skey:%" PRId64 ", ekey:%" PRId64 ", numOfVnode:%d, elapsed time:%" PRId64 " us", + pSql, numOfInput1, numOfInput2, output1->numOfTotal, output1->numOfVnodes, win->skey, win->ekey, + tsBufGetNumOfVnodes(output1), et - st); return output1->numOfTotal; } @@ -248,6 +279,68 @@ static UNUSED_FUNC bool needSecondaryQuery(SQueryInfo* pQueryInfo) { return false; } +static void filterVgroupTables(SQueryInfo* pQueryInfo, SArray* pVgroupTables) { + int32_t num = 0; + int32_t* list = NULL; + tsBufGetVnodeIdList(pQueryInfo->tsBuf, &num, &list); + + // The virtual node, of which all tables are disqualified after the timestamp intersection, + // is removed to avoid next stage query. + // TODO: If tables from some vnodes are not qualified for next stage query, discard them. + for (int32_t k = 0; k < taosArrayGetSize(pVgroupTables);) { + SVgroupTableInfo* p = taosArrayGet(pVgroupTables, k); + + bool found = false; + for (int32_t f = 0; f < num; ++f) { + if (p->vgInfo.vgId == list[f]) { + found = true; + break; + } + } + + if (!found) { + tscRemoveVgroupTableGroup(pVgroupTables, k); + } else { + k++; + } + } + + assert(taosArrayGetSize(pVgroupTables) > 0); + TSDB_QUERY_SET_TYPE(pQueryInfo->type, TSDB_QUERY_TYPE_MULTITABLE_QUERY); + + taosTFree(list); +} + +static SArray* buildVgroupTableByResult(SQueryInfo* pQueryInfo, SArray* pVgroupTables) { + int32_t num = 0; + int32_t* list = NULL; + tsBufGetVnodeIdList(pQueryInfo->tsBuf, &num, &list); + + size_t numOfGroups = taosArrayGetSize(pVgroupTables); + + SArray* pNew = taosArrayInit(num, sizeof(SVgroupTableInfo)); + + SVgroupTableInfo info; + for (int32_t i = 0; i < num; ++i) { + int32_t vnodeId = list[i]; + + for (int32_t j = 0; j < numOfGroups; ++j) { + SVgroupTableInfo* p1 = taosArrayGet(pVgroupTables, j); + if (p1->vgInfo.vgId == vnodeId) { + tscVgroupTableCopy(&info, p1); + break; + } + } + + taosArrayPush(pNew, &info); + } + + taosTFree(list); + TSDB_QUERY_SET_TYPE(pQueryInfo->type, TSDB_QUERY_TYPE_MULTITABLE_QUERY); + + return pNew; +} + /* * launch secondary stage query to fetch the result that contains timestamp in set */ @@ -322,12 +415,11 @@ static int32_t tscLaunchRealSubqueries(SSqlObj* pSql) { pQueryInfo->fieldsInfo = pSupporter->fieldsInfo; pQueryInfo->groupbyExpr = pSupporter->groupInfo; - SQueryInfo *pNewQueryInfo = tscGetQueryInfoDetail(&pNew->cmd, 0); - assert(pNew->subState.numOfSub == 0 && pNew->cmd.numOfClause == 1 && pNewQueryInfo->numOfTables == 1); + assert(pNew->subState.numOfSub == 0 && pNew->cmd.numOfClause == 1 && pQueryInfo->numOfTables == 1); - tscFieldInfoUpdateOffset(pNewQueryInfo); + tscFieldInfoUpdateOffset(pQueryInfo); - STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pNewQueryInfo, 0); + STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); pTableMetaInfo->pVgroupTables = pSupporter->pVgroupTables; pSupporter->exprList = NULL; @@ -341,7 +433,7 @@ static int32_t tscLaunchRealSubqueries(SSqlObj* pSql) { * during the timestamp intersection. */ pSupporter->limit = pQueryInfo->limit; - pNewQueryInfo->limit = pSupporter->limit; + pQueryInfo->limit = pSupporter->limit; SColumnIndex index = {.tableIndex = 0, .columnIndex = PRIMARYKEY_TIMESTAMP_COL_INDEX}; SSchema* s = tscGetTableColumnSchema(pTableMetaInfo->pTableMeta, 0); @@ -356,7 +448,7 @@ static int32_t tscLaunchRealSubqueries(SSqlObj* pSql) { tscAddSpecialColumnForSelect(pQueryInfo, 0, functionId, &index, s, TSDB_COL_NORMAL); tscPrintSelectClause(pNew, 0); - tscFieldInfoUpdateOffset(pNewQueryInfo); + tscFieldInfoUpdateOffset(pQueryInfo); pExpr = tscSqlExprGet(pQueryInfo, 0); } @@ -371,39 +463,21 @@ static int32_t tscLaunchRealSubqueries(SSqlObj* pSql) { pExpr->numOfParams = 1; } - int32_t num = 0; - int32_t *list = NULL; - tsBufGetVnodeIdList(pNewQueryInfo->tsBuf, &num, &list); - - if (pTableMetaInfo->pVgroupTables != NULL) { - for(int32_t k = 0; k < taosArrayGetSize(pTableMetaInfo->pVgroupTables);) { - SVgroupTableInfo* p = taosArrayGet(pTableMetaInfo->pVgroupTables, k); - - bool found = false; - for(int32_t f = 0; f < num; ++f) { - if (p->vgInfo.vgId == list[f]) { - found = true; - break; - } - } - - if (!found) { - tscRemoveVgroupTableGroup(pTableMetaInfo->pVgroupTables, k); - } else { - k++; - } + if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { + assert(pTableMetaInfo->pVgroupTables != NULL); + if (tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0)) { + SArray* p = buildVgroupTableByResult(pQueryInfo, pTableMetaInfo->pVgroupTables); + tscFreeVgroupTableInfo(pTableMetaInfo->pVgroupTables); + pTableMetaInfo->pVgroupTables = p; + } else { + filterVgroupTables(pQueryInfo, pTableMetaInfo->pVgroupTables); } - - assert(taosArrayGetSize(pTableMetaInfo->pVgroupTables) > 0); - TSDB_QUERY_SET_TYPE(pQueryInfo->type, TSDB_QUERY_TYPE_MULTITABLE_QUERY); } - taosTFree(list); - - size_t numOfCols = taosArrayGetSize(pNewQueryInfo->colList); + size_t numOfCols = taosArrayGetSize(pQueryInfo->colList); tscDebug("%p subquery:%p tableIndex:%d, vgroupIndex:%d, type:%d, exprInfo:%" PRIzu ", colList:%" PRIzu ", fieldsInfo:%d, name:%s", - pSql, pNew, 0, pTableMetaInfo->vgroupIndex, pNewQueryInfo->type, taosArrayGetSize(pNewQueryInfo->exprList), - numOfCols, pNewQueryInfo->fieldsInfo.numOfOutput, pTableMetaInfo->name); + pSql, pNew, 0, pTableMetaInfo->vgroupIndex, pQueryInfo->type, taosArrayGetSize(pQueryInfo->exprList), + numOfCols, pQueryInfo->fieldsInfo.numOfOutput, pTableMetaInfo->name); } //prepare the subqueries object failed, abort @@ -509,7 +583,7 @@ void tscBuildVgroupTableInfo(SSqlObj* pSql, STableMetaInfo* pTableMetaInfo, SArr SVgroupTableInfo info = {{0}}; for (int32_t m = 0; m < pvg->numOfVgroups; ++m) { if (tt->vgId == pvg->vgroups[m].vgId) { - tscSCMVgroupInfoCopy(&info.vgInfo, &pvg->vgroups[m]); + tscSVgroupInfoCopy(&info.vgInfo, &pvg->vgroups[m]); break; } } @@ -517,17 +591,29 @@ void tscBuildVgroupTableInfo(SSqlObj* pSql, STableMetaInfo* pTableMetaInfo, SArr vgTables = taosArrayInit(4, sizeof(STableIdInfo)); info.itemList = vgTables; + + if (taosArrayGetSize(result) > 0) { + SVgroupTableInfo* prevGroup = taosArrayGet(result, taosArrayGetSize(result) - 1); + tscDebug("%p vgId:%d, tables:%"PRId64, pSql, prevGroup->vgInfo.vgId, taosArrayGetSize(prevGroup->itemList)); + } + taosArrayPush(result, &info); } - tscDebug("%p tid:%d, uid:%"PRIu64",vgId:%d added for vnode query", pSql, tt->tid, tt->uid, tt->vgId) STableIdInfo item = {.uid = tt->uid, .tid = tt->tid, .key = INT64_MIN}; taosArrayPush(vgTables, &item); + + tscTrace("%p tid:%d, uid:%"PRIu64",vgId:%d added", pSql, tt->tid, tt->uid, tt->vgId); prev = tt; } pTableMetaInfo->pVgroupTables = result; pTableMetaInfo->vgroupIndex = 0; + + if (taosArrayGetSize(result) > 0) { + SVgroupTableInfo* g = taosArrayGet(result, taosArrayGetSize(result) - 1); + tscDebug("%p vgId:%d, tables:%"PRId64, pSql, g->vgInfo.vgId, taosArrayGetSize(g->itemList)); + } } static void issueTSCompQuery(SSqlObj* pSql, SJoinSupporter* pSupporter, SSqlObj* pParent) { @@ -602,11 +688,11 @@ static bool checkForDuplicateTagVal(SSchema* pColSchema, SJoinSupporter* p1, SSq } static int32_t getIntersectionOfTableTuple(SQueryInfo* pQueryInfo, SSqlObj* pParentSql, SArray** s1, SArray** s2) { - tscDebug("%p all subqueries retrieve complete, do tags match", pParentSql); - SJoinSupporter* p1 = pParentSql->pSubs[0]->param; SJoinSupporter* p2 = pParentSql->pSubs[1]->param; + tscDebug("%p all subquery retrieve complete, do tags match, %d, %d", pParentSql, p1->num, p2->num); + // sort according to the tag value qsort(p1->pIdTagList, p1->num, p1->tagSize, tagValCompar); qsort(p2->pIdTagList, p2->num, p2->tagSize, tagValCompar); @@ -655,6 +741,19 @@ static int32_t getIntersectionOfTableTuple(SQueryInfo* pQueryInfo, SSqlObj* pPar qsort((*s1)->pData, t1, size, tidTagsCompar); qsort((*s2)->pData, t2, size, tidTagsCompar); +#if 0 + for(int32_t k = 0; k < t1; ++k) { + STidTags* p = (*s1)->pData + size * k; + printf("%d, tag:%s\n", p->vgId, ((tstr*)(p->tag))->data); + } + + for(int32_t k = 0; k < t1; ++k) { + STidTags* p = (*s2)->pData + size * k; + printf("%d, tag:%s\n", p->vgId, ((tstr*)(p->tag))->data); + } +#endif + + tscDebug("%p tags match complete, result: %"PRId64", %"PRId64, pParentSql, t1, t2); return TSDB_CODE_SUCCESS; } @@ -958,6 +1057,7 @@ static void joinRetrieveFinalResCallback(void* param, TAOS_RES* tres, int numOfR } } + assert(pState->numOfRemain > 0); if (atomic_sub_fetch_32(&pState->numOfRemain, 1) > 0) { tscDebug("%p sub:%p completed, remain:%d, total:%d", pParentSql, tres, pState->numOfRemain, pState->numOfSub); return; @@ -971,6 +1071,7 @@ static void joinRetrieveFinalResCallback(void* param, TAOS_RES* tres, int numOfR } // update the records for each subquery in parent sql object. + bool stableQuery = tscIsTwoStageSTableQuery(pQueryInfo, 0); for (int32_t i = 0; i < pState->numOfSub; ++i) { if (pParentSql->pSubs[i] == NULL) { tscDebug("%p %p sub:%d not retrieve data", pParentSql, NULL, i); @@ -984,7 +1085,10 @@ static void joinRetrieveFinalResCallback(void* param, TAOS_RES* tres, int numOfR pRes1->numOfRows, pRes1->numOfTotal); assert(pRes1->row < pRes1->numOfRows); } else { - pRes1->numOfClauseTotal += pRes1->numOfRows; + if (!stableQuery) { + pRes1->numOfClauseTotal += pRes1->numOfRows; + } + tscDebug("%p sub:%p index:%d numOfRows:%"PRId64" total:%"PRId64, pParentSql, pParentSql->pSubs[i], i, pRes1->numOfRows, pRes1->numOfTotal); } @@ -994,7 +1098,7 @@ static void joinRetrieveFinalResCallback(void* param, TAOS_RES* tres, int numOfR tscBuildResFromSubqueries(pParentSql); } -void tscFetchDatablockFromSubquery(SSqlObj* pSql) { +void tscFetchDatablockForSubquery(SSqlObj* pSql) { assert(pSql->subState.numOfSub >= 1); int32_t numOfFetch = 0; @@ -1056,11 +1160,22 @@ void tscFetchDatablockFromSubquery(SSqlObj* pSql) { if (numOfFetch <= 0) { bool tryNextVnode = false; - SSqlObj* pp = pSql->pSubs[0]; - SQueryInfo* pi = tscGetQueryInfoDetail(&pp->cmd, 0); + bool orderedPrjQuery = false; + for(int32_t i = 0; i < pSql->subState.numOfSub; ++i) { + SSqlObj* pSub = pSql->pSubs[i]; + if (pSub == NULL) { + continue; + } + + SQueryInfo* p = tscGetQueryInfoDetail(&pSub->cmd, 0); + orderedPrjQuery = tscNonOrderedProjectionQueryOnSTable(p, 0); + if (orderedPrjQuery) { + break; + } + } // get the number of subquery that need to retrieve the next vnode. - if (tscNonOrderedProjectionQueryOnSTable(pi, 0)) { + if (orderedPrjQuery) { for (int32_t i = 0; i < pSql->subState.numOfSub; ++i) { SSqlObj* pSub = pSql->pSubs[i]; if (pSub != NULL && pSub->res.row >= pSub->res.numOfRows && pSub->res.completed) { @@ -1164,7 +1279,6 @@ void tscSetupOutputColumnIndex(SSqlObj* pSql) { SSqlCmd* pCmd = &pSql->cmd; SSqlRes* pRes = &pSql->res; - tscDebug("%p all subquery response, retrieve data for subclause:%d", pSql, pCmd->clauseIndex); // the column transfer support struct has been built if (pRes->pColumnIndex != NULL) { @@ -1260,21 +1374,23 @@ void tscJoinQueryCallback(void* param, TAOS_RES* tres, int code) { return; } - // wait for the other subqueries response from vnode - if (atomic_sub_fetch_32(&pParentSql->subState.numOfRemain, 1) > 0) { - return; + + STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); + + // In case of consequence query from other vnode, do not wait for other query response here. + if (!(pTableMetaInfo->vgroupIndex > 0 && tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0))) { + if (atomic_sub_fetch_32(&pParentSql->subState.numOfRemain, 1) > 0) { + return; + } } tscSetupOutputColumnIndex(pParentSql); - STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); /** * if the query is a continue query (vgroupIndex > 0 for projection query) for next vnode, do the retrieval of * data instead of returning to its invoker */ if (pTableMetaInfo->vgroupIndex > 0 && tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0)) { -// pParentSql->subState.numOfRemain = pParentSql->subState.numOfSub; // reset the record value - pSql->fp = joinRetrieveFinalResCallback; // continue retrieve data pSql->cmd.command = TSDB_SQL_FETCH; tscProcessSql(pSql); @@ -1671,8 +1787,8 @@ static int32_t tscReissueSubquery(SRetrieveSupport *trsupport, SSqlObj *pSql, in SSqlObj *pParentSql = trsupport->pParentSql; int32_t subqueryIndex = trsupport->subqueryIndex; - STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); - SCMVgroupInfo* pVgroup = &pTableMetaInfo->vgroupList->vgroups[0]; + STableMetaInfo* pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); + SVgroupInfo* pVgroup = &pTableMetaInfo->vgroupList->vgroups[0]; tExtMemBufferClear(trsupport->pExtMemBuffer[subqueryIndex]); @@ -1874,7 +1990,7 @@ static void tscRetrieveFromDnodeCallBack(void *param, TAOS_RES *tres, int numOfR assert(pState->numOfRemain <= pState->numOfSub && pState->numOfRemain >= 0); STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); - SCMVgroupInfo *pVgroup = &pTableMetaInfo->vgroupList->vgroups[0]; + SVgroupInfo *pVgroup = &pTableMetaInfo->vgroupList->vgroups[0]; if (pParentSql->res.code != TSDB_CODE_SUCCESS) { trsupport->numOfRetry = MAX_NUM_OF_SUBQUERY_RETRY; @@ -1985,7 +2101,7 @@ void tscRetrieveDataRes(void *param, TAOS_RES *tres, int code) { assert(pSql->cmd.numOfClause == 1 && pQueryInfo->numOfTables == 1); STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); - SCMVgroupInfo* pVgroup = &pTableMetaInfo->vgroupList->vgroups[trsupport->subqueryIndex]; + SVgroupInfo* pVgroup = &pTableMetaInfo->vgroupList->vgroups[trsupport->subqueryIndex]; // stable query killed or other subquery failed, all query stopped if (pParentSql->res.code != TSDB_CODE_SUCCESS) { diff --git a/src/client/src/tscSystem.c b/src/client/src/tscSystem.c index 77d668c5af152eb96ae2f0e698b806444faed667..839d5889f3beb5401ba91790fed1c4a925c600bc 100644 --- a/src/client/src/tscSystem.c +++ b/src/client/src/tscSystem.c @@ -104,7 +104,6 @@ void taos_init_imp(void) { taosReadGlobalCfg(); taosCheckGlobalCfg(); - taosPrintGlobalCfg(); tscDebug("starting to initialize TAOS client ..."); tscDebug("Local End Point is:%s", tsLocalEp); diff --git a/src/client/src/tscUtil.c b/src/client/src/tscUtil.c index 117874475ba9c120d612643d14250810bd0f5e8c..879eeeadedf03afc2a6dbd240c5500c801efc483 100644 --- a/src/client/src/tscUtil.c +++ b/src/client/src/tscUtil.c @@ -338,34 +338,6 @@ void tscFreeSqlResult(SSqlObj* pSql) { memset(&pSql->res, 0, sizeof(SSqlRes)); } -void tscPartiallyFreeSqlObj(SSqlObj* pSql) { - if (pSql == NULL || pSql->signature != pSql) { - return; - } - - SSqlCmd* pCmd = &pSql->cmd; - int32_t cmd = pCmd->command; - if (cmd < TSDB_SQL_INSERT || cmd == TSDB_SQL_RETRIEVE_LOCALMERGE || cmd == TSDB_SQL_RETRIEVE_EMPTY_RESULT || - cmd == TSDB_SQL_TABLE_JOIN_RETRIEVE) { - tscRemoveFromSqlList(pSql); - } - - // pSql->sqlstr will be used by tscBuildQueryStreamDesc -// if (pObj->signature == pObj) { - //pthread_mutex_lock(&pObj->mutex); - taosTFree(pSql->sqlstr); - //pthread_mutex_unlock(&pObj->mutex); -// } - - tscFreeSqlResult(pSql); - - taosTFree(pSql->pSubs); - pSql->subState.numOfSub = 0; - pSql->self = 0; - - tscResetSqlCmdObj(pCmd, false); -} - static void tscFreeSubobj(SSqlObj* pSql) { if (pSql->subState.numOfSub == 0) { return; @@ -434,22 +406,32 @@ void tscFreeSqlObj(SSqlObj* pSql) { tscDebug("%p start to free sqlObj", pSql); pSql->res.code = TSDB_CODE_TSC_QUERY_CANCELLED; + tscFreeSubobj(pSql); - tscPartiallyFreeSqlObj(pSql); + SSqlCmd* pCmd = &pSql->cmd; + int32_t cmd = pCmd->command; + if (cmd < TSDB_SQL_INSERT || cmd == TSDB_SQL_RETRIEVE_LOCALMERGE || cmd == TSDB_SQL_RETRIEVE_EMPTY_RESULT || + cmd == TSDB_SQL_TABLE_JOIN_RETRIEVE) { + tscRemoveFromSqlList(pSql); + } pSql->signature = NULL; pSql->fp = NULL; - - SSqlCmd* pCmd = &pSql->cmd; + taosTFree(pSql->sqlstr); + + taosTFree(pSql->pSubs); + pSql->subState.numOfSub = 0; + pSql->self = 0; + + tscFreeSqlResult(pSql); + tscResetSqlCmdObj(pCmd, false); memset(pCmd->payload, 0, (size_t)pCmd->allocSize); taosTFree(pCmd->payload); pCmd->allocSize = 0; - taosTFree(pSql->sqlstr); tsem_destroy(&pSql->rspSem); - free(pSql); } @@ -1714,6 +1696,17 @@ void tscRemoveVgroupTableGroup(SArray* pVgroupTable, int32_t index) { taosArrayRemove(pVgroupTable, index); } +void tscVgroupTableCopy(SVgroupTableInfo* info, SVgroupTableInfo* pInfo) { + memset(info, 0, sizeof(SVgroupTableInfo)); + + info->vgInfo = pInfo->vgInfo; + for(int32_t j = 0; j < pInfo->vgInfo.numOfEps; ++j) { + info->vgInfo.epAddr[j].fqdn = strdup(pInfo->vgInfo.epAddr[j].fqdn); + } + + info->itemList = taosArrayClone(pInfo->itemList); +} + SArray* tscVgroupTableInfoClone(SArray* pVgroupTables) { if (pVgroupTables == NULL) { return NULL; @@ -1725,14 +1718,8 @@ SArray* tscVgroupTableInfoClone(SArray* pVgroupTables) { SVgroupTableInfo info; for (size_t i = 0; i < num; i++) { SVgroupTableInfo* pInfo = taosArrayGet(pVgroupTables, i); - memset(&info, 0, sizeof(SVgroupTableInfo)); - - info.vgInfo = pInfo->vgInfo; - for(int32_t j = 0; j < pInfo->vgInfo.numOfEps; ++j) { - info.vgInfo.epAddr[j].fqdn = strdup(pInfo->vgInfo.epAddr[j].fqdn); - } + tscVgroupTableCopy(&info, pInfo); - info.itemList = taosArrayClone(pInfo->itemList); taosArrayPush(pa, &info); } @@ -2441,7 +2428,7 @@ SVgroupsInfo* tscVgroupInfoClone(SVgroupsInfo *vgroupList) { return NULL; } - size_t size = sizeof(SVgroupsInfo) + sizeof(SCMVgroupInfo) * vgroupList->numOfVgroups; + size_t size = sizeof(SVgroupsInfo) + sizeof(SVgroupInfo) * vgroupList->numOfVgroups; SVgroupsInfo* pNew = calloc(1, size); if (pNew == NULL) { return NULL; @@ -2450,9 +2437,9 @@ SVgroupsInfo* tscVgroupInfoClone(SVgroupsInfo *vgroupList) { pNew->numOfVgroups = vgroupList->numOfVgroups; for(int32_t i = 0; i < vgroupList->numOfVgroups; ++i) { - SCMVgroupInfo* pNewVInfo = &pNew->vgroups[i]; + SVgroupInfo* pNewVInfo = &pNew->vgroups[i]; - SCMVgroupInfo* pvInfo = &vgroupList->vgroups[i]; + SVgroupInfo* pvInfo = &vgroupList->vgroups[i]; pNewVInfo->vgId = pvInfo->vgId; pNewVInfo->numOfEps = pvInfo->numOfEps; @@ -2471,7 +2458,7 @@ void* tscVgroupInfoClear(SVgroupsInfo *vgroupList) { } for(int32_t i = 0; i < vgroupList->numOfVgroups; ++i) { - SCMVgroupInfo* pVgroupInfo = &vgroupList->vgroups[i]; + SVgroupInfo* pVgroupInfo = &vgroupList->vgroups[i]; for(int32_t j = 0; j < pVgroupInfo->numOfEps; ++j) { taosTFree(pVgroupInfo->epAddr[j].fqdn); @@ -2482,7 +2469,7 @@ void* tscVgroupInfoClear(SVgroupsInfo *vgroupList) { return NULL; } -void tscSCMVgroupInfoCopy(SCMVgroupInfo* dst, const SCMVgroupInfo* src) { +void tscSVgroupInfoCopy(SVgroupInfo* dst, const SVgroupInfo* src) { dst->vgId = src->vgId; dst->numOfEps = src->numOfEps; for(int32_t i = 0; i < dst->numOfEps; ++i) { diff --git a/src/common/inc/tglobal.h b/src/common/inc/tglobal.h index 4636eaac08db4943e1837b5d6e8db40341ba6546..f059ec1566a237a1efaf7acf4c76aa8dcbd10753 100644 --- a/src/common/inc/tglobal.h +++ b/src/common/inc/tglobal.h @@ -51,6 +51,7 @@ extern char tsLocale[]; extern char tsCharset[]; // default encode string extern int32_t tsEnableCoreFile; extern int32_t tsCompressMsgSize; +extern char tsTempDir[]; //query buffer management extern int32_t tsQueryBufferSize; // maximum allowed usage buffer for each data node during query processing @@ -183,13 +184,13 @@ extern int32_t debugFlag; #define NEEDTO_COMPRESSS_MSG(size) (tsCompressMsgSize != -1 && (size) > tsCompressMsgSize) -void taosInitGlobalCfg(); -bool taosCheckGlobalCfg(); -void taosSetAllDebugFlag(); -bool taosCfgDynamicOptions(char *msg); -int taosGetFqdnPortFromEp(const char *ep, char *fqdn, uint16_t *port); -bool taosCheckBalanceCfgOptions(const char *option, int32_t *vnodeId, int32_t *dnodeId); - +void taosInitGlobalCfg(); +int32_t taosCheckGlobalCfg(); +void taosSetAllDebugFlag(); +bool taosCfgDynamicOptions(char *msg); +int taosGetFqdnPortFromEp(const char *ep, char *fqdn, uint16_t *port); +bool taosCheckBalanceCfgOptions(const char *option, int32_t *vnodeId, int32_t *dnodeId); + #ifdef __cplusplus } #endif diff --git a/src/common/src/tglobal.c b/src/common/src/tglobal.c index 32569e39823416a25bc0f0bb88d1feb9faafcff1..cbeac421c95faa03fd668a4f5742e47a289999e0 100644 --- a/src/common/src/tglobal.c +++ b/src/common/src/tglobal.c @@ -58,6 +58,7 @@ char tsLocale[TSDB_LOCALE_LEN] = {0}; char tsCharset[TSDB_LOCALE_LEN] = {0}; // default encode string int32_t tsEnableCoreFile = 0; int32_t tsMaxBinaryDisplayWidth = 30; +char tsTempDir[TSDB_FILENAME_LEN] = "/tmp/"; /* * denote if the server needs to compress response message at the application layer to client, including query rsp, @@ -1310,13 +1311,23 @@ static void doInitGlobalConfig(void) { cfg.ptrLength = 0; cfg.unitType = TAOS_CFG_UTYPE_NONE; taosInitConfigOption(cfg); + + cfg.option = "tempDir"; + cfg.ptr = tsTempDir; + cfg.valType = TAOS_CFG_VTYPE_STRING; + cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_CLIENT; + cfg.minValue = 0; + cfg.maxValue = 0; + cfg.ptrLength = tListLen(tsTempDir); + cfg.unitType = TAOS_CFG_UTYPE_NONE; + taosInitConfigOption(cfg); } void taosInitGlobalCfg() { pthread_once(&tsInitGlobalCfgOnce, doInitGlobalConfig); } -bool taosCheckGlobalCfg() { +int32_t taosCheckGlobalCfg() { char fqdn[TSDB_FQDN_LEN]; uint16_t port; @@ -1375,7 +1386,9 @@ bool taosCheckGlobalCfg() { tsSyncPort = tsServerPort + TSDB_PORT_SYNC; tsHttpPort = tsServerPort + TSDB_PORT_HTTP; - return true; + taosPrintGlobalCfg(); + + return 0; } int taosGetFqdnPortFromEp(const char *ep, char *fqdn, uint16_t *port) { diff --git a/src/common/src/tvariant.c b/src/common/src/tvariant.c index 9eb9924932c7757c3c999eb5afadd8c719dc16bc..ca1644c0a24472428097467ded7601294c4dcd87 100644 --- a/src/common/src/tvariant.c +++ b/src/common/src/tvariant.c @@ -108,7 +108,7 @@ void tVariantCreateFromBinary(tVariant *pVar, const char *pz, size_t len, uint32 break; } case TSDB_DATA_TYPE_BINARY: { // todo refactor, extract a method - pVar->pz = calloc(len, sizeof(char)); + pVar->pz = calloc(len + 1, sizeof(char)); memcpy(pVar->pz, pz, len); pVar->nLen = (int32_t)len; break; diff --git a/src/dnode/inc/dnodeCfg.h b/src/dnode/inc/dnodeCfg.h new file mode 100644 index 0000000000000000000000000000000000000000..35d889646021a079abe18c01a26fa5891718aaa0 --- /dev/null +++ b/src/dnode/inc/dnodeCfg.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_CFG_H +#define TDENGINE_DNODE_CFG_H + +#ifdef __cplusplus +extern "C" { +#endif + +int32_t dnodeInitCfg(); +void dnodeCleanupCfg(); +void dnodeUpdateCfg(SDnodeCfg *cfg); +int32_t dnodeGetDnodeId(); +void dnodeGetCfg(int32_t *dnodeId, char *clusterId); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/dnode/inc/dnodeEps.h b/src/dnode/inc/dnodeEps.h new file mode 100644 index 0000000000000000000000000000000000000000..2a203498c1f270c7e456694a4e1e195cbd9022cd --- /dev/null +++ b/src/dnode/inc/dnodeEps.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_DNODE_EP_H +#define TDENGINE_DNODE_EP_H + +#ifdef __cplusplus +extern "C" { +#endif + +#include "taosmsg.h" + +int32_t dnodeInitEps(); +void dnodeCleanupEps(); +void dnodeUpdateEps(SDnodeEps *eps); +void dnodeUpdateEp(int32_t dnodeId, char *epstr, char *fqdn, uint16_t *port); +bool dnodeCheckEpChanged(int32_t dnodeId, char *epstr); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/dnode/inc/dnodeMInfos.h b/src/dnode/inc/dnodeMInfos.h new file mode 100644 index 0000000000000000000000000000000000000000..9c3c85c47e2dbcc11c5b5a80fbf091bd93855149 --- /dev/null +++ b/src/dnode/inc/dnodeMInfos.h @@ -0,0 +1,36 @@ +/* + * 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_MINFOS_H +#define TDENGINE_DNODE_MINFOS_H + +#ifdef __cplusplus +extern "C" { +#endif + +#include "taosmsg.h" + +int32_t dnodeInitMInfos(); +void dnodeCleanupMInfos(); +void dnodeUpdateMInfos(SMnodeInfos *minfos); +void dnodeUpdateEpSetForPeer(SRpcEpSet *epSet); +void dnodeGetMInfos(SMnodeInfos *minfos); +bool dnodeIsMasterEp(char *ep); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/dnode/inc/dnodeMgmt.h b/src/dnode/inc/dnodeMgmt.h index e8f4a0823f3cc9846b85804d323e73a8cfc0476e..2038ef5286b32522b11409ba5a253b33228b984d 100644 --- a/src/dnode/inc/dnodeMgmt.h +++ b/src/dnode/inc/dnodeMgmt.h @@ -20,6 +20,8 @@ extern "C" { #endif +#include "trpc.h" + int32_t dnodeInitMgmt(); void dnodeCleanupMgmt(); int32_t dnodeInitMgmtTimer(); @@ -35,8 +37,8 @@ void* dnodeGetVnodeTsdb(void *pVnode); void dnodeReleaseVnode(void *pVnode); void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell); -void dnodeGetMnodeEpSetForPeer(void *epSet); -void dnodeGetMnodeEpSetForShell(void *epSet); +void dnodeGetEpSetForPeer(SRpcEpSet *epSet); +void dnodeGetEpSetForShell(SRpcEpSet *epSet); #ifdef __cplusplus } diff --git a/src/dnode/inc/dnodeVWrite.h b/src/dnode/inc/dnodeVWrite.h index 7da701a8e270239ee2f48e090138f048cc81f880..323405143fd10b7b4f73bac5103ad63ed870044f 100644 --- a/src/dnode/inc/dnodeVWrite.h +++ b/src/dnode/inc/dnodeVWrite.h @@ -20,9 +20,12 @@ extern "C" { #endif -int32_t dnodeInitVnodeWrite(); -void dnodeCleanupVnodeWrite(); -void dnodeDispatchToVnodeWriteQueue(SRpcMsg *pMsg); +int32_t dnodeInitVWrite(); +void dnodeCleanupVWrite(); +void dnodeDispatchToVWriteQueue(SRpcMsg *pMsg); +void * dnodeAllocVWriteQueue(void *pVnode); +void dnodeFreeVWriteQueue(void *wqueue); +void dnodeSendRpcVWriteRsp(void *pVnode, void *param, int32_t code); #ifdef __cplusplus } diff --git a/src/dnode/src/dnodeCfg.c b/src/dnode/src/dnodeCfg.c new file mode 100644 index 0000000000000000000000000000000000000000..16d109a13a7600c5414b1f1edd175a8f15aa4cf0 --- /dev/null +++ b/src/dnode/src/dnodeCfg.c @@ -0,0 +1,163 @@ +/* + * 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 "tglobal.h" +#include "dnode.h" +#include "dnodeInt.h" +#include "dnodeCfg.h" + +static SDnodeCfg tsCfg = {0}; +static pthread_mutex_t tsCfgMutex; + +static int32_t dnodeReadCfg(); +static int32_t dnodeWriteCfg(); +static void dnodeResetCfg(SDnodeCfg *cfg); +static void dnodePrintCfg(SDnodeCfg *cfg); + +int32_t dnodeInitCfg() { + pthread_mutex_init(&tsCfgMutex, NULL); + dnodeResetCfg(NULL); + int32_t ret = dnodeReadCfg(); + if (ret == 0) { + dInfo("dnode cfg is initialized"); + } + return ret; +} + +void dnodeCleanupCfg() { pthread_mutex_destroy(&tsCfgMutex); } + +void dnodeUpdateCfg(SDnodeCfg *cfg) { + if (tsCfg.dnodeId != 0) return; + dnodeResetCfg(cfg); +} + +int32_t dnodeGetDnodeId() { + int32_t dnodeId = 0; + pthread_mutex_lock(&tsCfgMutex); + dnodeId = tsCfg.dnodeId; + pthread_mutex_unlock(&tsCfgMutex); + return dnodeId; +} + +void dnodeGetCfg(int32_t *dnodeId, char *clusterId) { + pthread_mutex_lock(&tsCfgMutex); + *dnodeId = tsCfg.dnodeId; + tstrncpy(clusterId, tsCfg.clusterId, TSDB_CLUSTER_ID_LEN); + pthread_mutex_unlock(&tsCfgMutex); +} + +static void dnodeResetCfg(SDnodeCfg *cfg) { + if (cfg == NULL) return; + if (cfg->dnodeId == 0) return; + + pthread_mutex_lock(&tsCfgMutex); + tsCfg.dnodeId = cfg->dnodeId; + tstrncpy(tsCfg.clusterId, cfg->clusterId, TSDB_CLUSTER_ID_LEN); + dnodePrintCfg(cfg); + dnodeWriteCfg(); + pthread_mutex_unlock(&tsCfgMutex); +} + +static void dnodePrintCfg(SDnodeCfg *cfg) { + dInfo("dnodeId is set to %d, clusterId is set to %s", cfg->dnodeId, cfg->clusterId); +} + +static int32_t dnodeReadCfg() { + int32_t len = 0; + int32_t maxLen = 200; + char * content = calloc(1, maxLen + 1); + cJSON * root = NULL; + FILE * fp = NULL; + SDnodeCfg cfg = {0}; + + char file[TSDB_FILENAME_LEN + 20] = {0}; + sprintf(file, "%s/dnodeCfg.json", tsDnodeDir); + + fp = fopen(file, "r"); + if (!fp) { + dDebug("failed to read %s, file not exist", file); + goto PARSE_CFG_OVER; + } + + len = fread(content, 1, maxLen, fp); + if (len <= 0) { + dError("failed to read %s, content is null", file); + goto PARSE_CFG_OVER; + } + + content[len] = 0; + root = cJSON_Parse(content); + if (root == NULL) { + dError("failed to read %s, invalid json format", file); + goto PARSE_CFG_OVER; + } + + cJSON *dnodeId = cJSON_GetObjectItem(root, "dnodeId"); + if (!dnodeId || dnodeId->type != cJSON_Number) { + dError("failed to read %s, dnodeId not found", file); + goto PARSE_CFG_OVER; + } + cfg.dnodeId = dnodeId->valueint; + + cJSON *clusterId = cJSON_GetObjectItem(root, "clusterId"); + if (!clusterId || clusterId->type != cJSON_String) { + dError("failed to read %s, clusterId not found", file); + goto PARSE_CFG_OVER; + } + tstrncpy(cfg.clusterId, clusterId->valuestring, TSDB_CLUSTER_ID_LEN); + + dInfo("read file %s successed", file); + +PARSE_CFG_OVER: + if (content != NULL) free(content); + if (root != NULL) cJSON_Delete(root); + if (fp != NULL) fclose(fp); + terrno = 0; + + dnodeResetCfg(&cfg); + return 0; +} + +static int32_t dnodeWriteCfg() { + char file[TSDB_FILENAME_LEN + 20] = {0}; + sprintf(file, "%s/dnodeCfg.json", tsDnodeDir); + + FILE *fp = fopen(file, "w"); + if (!fp) { + dError("failed to write %s, reason:%s", file, strerror(errno)); + return -1; + } + + int32_t len = 0; + int32_t maxLen = 200; + char * content = calloc(1, maxLen + 1); + + len += snprintf(content + len, maxLen - len, "{\n"); + len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", tsCfg.dnodeId); + len += snprintf(content + len, maxLen - len, " \"clusterId\": \"%s\"\n", tsCfg.clusterId); + len += snprintf(content + len, maxLen - len, "}\n"); + + fwrite(content, 1, len, fp); + fflush(fp); + fclose(fp); + free(content); + terrno = 0; + + dInfo("successed to write %s", file); + return 0; +} diff --git a/src/dnode/src/dnodeCheck.c b/src/dnode/src/dnodeCheck.c index 9b68fc1f6c7a4db266eeb36cc20459bc71870739..a9ee4ac649c2f4b2734ce7f0dd59004b08e2fb67 100644 --- a/src/dnode/src/dnodeCheck.c +++ b/src/dnode/src/dnodeCheck.c @@ -15,9 +15,7 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taosdef.h" #include "tglobal.h" -#include "mnode.h" #include "dnodeInt.h" #include "dnodeCheck.h" @@ -30,8 +28,8 @@ typedef struct { void (*stopFp)(); } SCheckItem; -static SCheckItem tsCheckItem[TSDB_CHECK_ITEM_MAX] = {{0}}; -int64_t tsMinFreeMemSizeForStart = 0; +static SCheckItem tsCheckItem[TSDB_CHECK_ITEM_MAX] = {{0}}; +int64_t tsMinFreeMemSizeForStart = 0; static int bindTcpPort(int port) { int serverSocket; @@ -264,8 +262,6 @@ int32_t dnodeInitCheck() { } } + dInfo("dnode check is initialized"); return 0; } - - - diff --git a/src/dnode/src/dnodeEps.c b/src/dnode/src/dnodeEps.c new file mode 100644 index 0000000000000000000000000000000000000000..9c90c391813f57024ecf090136dfa86f5e5f91e6 --- /dev/null +++ b/src/dnode/src/dnodeEps.c @@ -0,0 +1,283 @@ +/* + * 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 "tglobal.h" +#include "hash.h" +#include "dnode.h" +#include "dnodeInt.h" +#include "dnodeEps.h" + +static SDnodeEps *tsEps = NULL; +static SHashObj * tsEpsHash = NULL; +static pthread_mutex_t tsEpsMutex; + +static int32_t dnodeReadEps(); +static int32_t dnodeWriteEps(); +static void dnodeResetEps(SDnodeEps *eps); +static void dnodePrintEps(SDnodeEps *eps); + +int32_t dnodeInitEps() { + pthread_mutex_init(&tsEpsMutex, NULL); + tsEpsHash = taosHashInit(4, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, true); + dnodeResetEps(NULL); + int32_t ret = dnodeReadEps(); + if (ret == 0) { + dInfo("dnode eps is initialized"); + } + return ret; +} + +void dnodeCleanupEps() { + pthread_mutex_lock(&tsEpsMutex); + if (tsEps) { + free(tsEps); + tsEps = NULL; + } + if (tsEpsHash) { + taosHashCleanup(tsEpsHash); + tsEpsHash = NULL; + } + pthread_mutex_unlock(&tsEpsMutex); + pthread_mutex_destroy(&tsEpsMutex); +} + +void dnodeUpdateEps(SDnodeEps *eps) { + if (eps == NULL) return; + + eps->dnodeNum = htonl(eps->dnodeNum); + for (int32_t i = 0; i < eps->dnodeNum; ++i) { + eps->dnodeEps[i].dnodeId = htonl(eps->dnodeEps[i].dnodeId); + eps->dnodeEps[i].dnodePort = htons(eps->dnodeEps[i].dnodePort); + } + + pthread_mutex_lock(&tsEpsMutex); + if (eps->dnodeNum != tsEps->dnodeNum) { + dnodeResetEps(eps); + dnodeWriteEps(); + } else { + int32_t size = sizeof(SDnodeEps) + eps->dnodeNum * sizeof(SDnodeEp); + if (memcmp(eps, tsEps, size) != 0) { + dnodeResetEps(eps); + dnodeWriteEps(); + } + } + pthread_mutex_unlock(&tsEpsMutex); +} + +bool dnodeCheckEpChanged(int32_t dnodeId, char *epstr) { + bool changed = false; + pthread_mutex_lock(&tsEpsMutex); + SDnodeEp *ep = taosHashGet(tsEpsHash, &dnodeId, sizeof(int32_t)); + if (ep != NULL) { + char epSaved[TSDB_EP_LEN + 1]; + snprintf(epSaved, TSDB_EP_LEN, "%s:%u", ep->dnodeFqdn, ep->dnodePort); + changed = strcmp(epstr, epSaved) != 0; + tstrncpy(epstr, epSaved, TSDB_EP_LEN); + } + pthread_mutex_unlock(&tsEpsMutex); + return changed; +} + +void dnodeUpdateEp(int32_t dnodeId, char *epstr, char *fqdn, uint16_t *port) { + pthread_mutex_lock(&tsEpsMutex); + SDnodeEp *ep = taosHashGet(tsEpsHash, &dnodeId, sizeof(int32_t)); + if (ep != NULL) { + if (port) *port = ep->dnodePort; + if (fqdn) tstrncpy(fqdn, ep->dnodeFqdn, TSDB_FQDN_LEN); + if (epstr) snprintf(epstr, TSDB_EP_LEN, "%s:%u", ep->dnodeFqdn, ep->dnodePort); + } + pthread_mutex_unlock(&tsEpsMutex); +} + +static void dnodeResetEps(SDnodeEps *eps) { + if (eps == NULL) { + int32_t size = sizeof(SDnodeEps) + sizeof(SDnodeEp); + if (tsEps == NULL) { + tsEps = calloc(1, size); + } else { + tsEps->dnodeNum = 0; + } + } else { + assert(tsEps); + + int32_t size = sizeof(SDnodeEps) + sizeof(SDnodeEp) * eps->dnodeNum; + if (eps->dnodeNum > tsEps->dnodeNum) { + tsEps = realloc(tsEps, size); + } + memcpy(tsEps, eps, size); + dnodePrintEps(eps); + } + + for (int32_t i = 0; i < tsEps->dnodeNum; ++i) { + SDnodeEp *ep = &tsEps->dnodeEps[i]; + taosHashPut(tsEpsHash, &ep->dnodeId, sizeof(int32_t), ep, sizeof(SDnodeEp)); + } +} + +static void dnodePrintEps(SDnodeEps *eps) { + dDebug("print dnodeEp, dnodeNum:%d", eps->dnodeNum); + for (int32_t i = 0; i < eps->dnodeNum; i++) { + SDnodeEp *ep = &eps->dnodeEps[i]; + dDebug("dnodeId:%d, dnodeFqdn:%s dnodePort:%u", ep->dnodeId, ep->dnodeFqdn, ep->dnodePort); + } +} + +static int32_t dnodeReadEps() { + int32_t ret = -1; + int32_t len = 0; + int32_t maxLen = 30000; + char * content = calloc(1, maxLen + 1); + cJSON * root = NULL; + FILE * fp = NULL; + SDnodeEps *eps = NULL; + + char file[TSDB_FILENAME_LEN + 20] = {0}; + sprintf(file, "%s/dnodeEps.json", tsDnodeDir); + + fp = fopen(file, "r"); + if (!fp) { + dDebug("failed to read %s, file not exist", file); + goto PRASE_EPS_OVER; + } + + len = fread(content, 1, maxLen, fp); + if (len <= 0) { + dError("failed to read %s, content is null", file); + goto PRASE_EPS_OVER; + } + + content[len] = 0; + root = cJSON_Parse(content); + if (root == NULL) { + dError("failed to read %s, invalid json format", file); + goto PRASE_EPS_OVER; + } + + cJSON *dnodeNum = cJSON_GetObjectItem(root, "dnodeNum"); + if (!dnodeNum || dnodeNum->type != cJSON_Number) { + dError("failed to read %s, dnodeNum not found", file); + goto PRASE_EPS_OVER; + } + + cJSON *dnodeInfos = cJSON_GetObjectItem(root, "dnodeInfos"); + if (!dnodeInfos || dnodeInfos->type != cJSON_Array) { + dError("failed to read %s, dnodeInfos not found", file); + goto PRASE_EPS_OVER; + } + + int32_t dnodeInfosSize = cJSON_GetArraySize(dnodeInfos); + if (dnodeInfosSize != dnodeNum->valueint) { + dError("failed to read %s, dnodeInfos size:%d not matched dnodeNum:%d", file, dnodeInfosSize, + (int32_t)dnodeNum->valueint); + goto PRASE_EPS_OVER; + } + + int32_t epsSize = sizeof(SDnodeEps) + dnodeInfosSize * sizeof(SDnodeEp); + eps = calloc(1, epsSize); + eps->dnodeNum = dnodeInfosSize; + + for (int32_t i = 0; i < dnodeInfosSize; ++i) { + cJSON *dnodeInfo = cJSON_GetArrayItem(dnodeInfos, i); + if (dnodeInfo == NULL) break; + + SDnodeEp *ep = &eps->dnodeEps[i]; + + cJSON *dnodeId = cJSON_GetObjectItem(dnodeInfo, "dnodeId"); + if (!dnodeId || dnodeId->type != cJSON_Number) { + dError("failed to read %s, dnodeId not found", file); + goto PRASE_EPS_OVER; + } + ep->dnodeId = dnodeId->valueint; + + cJSON *dnodeFqdn = cJSON_GetObjectItem(dnodeInfo, "dnodeFqdn"); + if (!dnodeFqdn || dnodeFqdn->type != cJSON_String || dnodeFqdn->valuestring == NULL) { + dError("failed to read %s, dnodeFqdn not found", file); + goto PRASE_EPS_OVER; + } + strncpy(ep->dnodeFqdn, dnodeFqdn->valuestring, TSDB_FQDN_LEN); + + cJSON *dnodePort = cJSON_GetObjectItem(dnodeInfo, "dnodePort"); + if (!dnodePort || dnodePort->type != cJSON_Number) { + dError("failed to read %s, dnodePort not found", file); + goto PRASE_EPS_OVER; + } + ep->dnodePort = (uint16_t)dnodePort->valueint; + } + + ret = 0; + + dInfo("read file %s successed", file); + dnodePrintEps(eps); + +PRASE_EPS_OVER: + if (content != NULL) free(content); + if (root != NULL) cJSON_Delete(root); + if (fp != NULL) fclose(fp); + if (ret != 0) { + if (eps) free(eps); + eps = NULL; + } + + dnodeResetEps(eps); + if (eps) free(eps); + + dnodeUpdateEp(dnodeGetDnodeId(), tsLocalEp, tsLocalFqdn, &tsServerPort); + + terrno = 0; + return 0; +} + +static int32_t dnodeWriteEps() { + char file[TSDB_FILENAME_LEN + 20] = {0}; + sprintf(file, "%s/dnodeEps.json", tsDnodeDir); + + FILE *fp = fopen(file, "w"); + if (!fp) { + dError("failed to write %s, reason:%s", file, strerror(errno)); + return -1; + } + + int32_t len = 0; + int32_t maxLen = 30000; + char * content = calloc(1, maxLen + 1); + + len += snprintf(content + len, maxLen - len, "{\n"); + len += snprintf(content + len, maxLen - len, " \"dnodeNum\": %d,\n", tsEps->dnodeNum); + len += snprintf(content + len, maxLen - len, " \"dnodeInfos\": [{\n"); + for (int32_t i = 0; i < tsEps->dnodeNum; ++i) { + SDnodeEp *ep = &tsEps->dnodeEps[i]; + len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", ep->dnodeId); + len += snprintf(content + len, maxLen - len, " \"dnodeFqdn\": \"%s\",\n", ep->dnodeFqdn); + len += snprintf(content + len, maxLen - len, " \"dnodePort\": %u\n", ep->dnodePort); + if (i < tsEps->dnodeNum - 1) { + len += snprintf(content + len, maxLen - len, " },{\n"); + } else { + len += snprintf(content + len, maxLen - len, " }]\n"); + } + } + len += snprintf(content + len, maxLen - len, "}\n"); + + fwrite(content, 1, len, fp); + fflush(fp); + fclose(fp); + free(content); + terrno = 0; + + dInfo("successed to write %s", file); + return 0; +} diff --git a/src/dnode/src/dnodeMInfos.c b/src/dnode/src/dnodeMInfos.c new file mode 100644 index 0000000000000000000000000000000000000000..c985db371d96be5d78d71a50f9602485157a84a4 --- /dev/null +++ b/src/dnode/src/dnodeMInfos.c @@ -0,0 +1,288 @@ +/* + * 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 "tglobal.h" +#include "mnode.h" +#include "dnode.h" +#include "dnodeInt.h" +#include "dnodeMInfos.h" + +static SMnodeInfos tsMInfos; +static SRpcEpSet tsMEpSet; +static pthread_mutex_t tsMInfosMutex; + +static void dnodeResetMInfos(SMnodeInfos *minfos); +static void dnodePrintMInfos(SMnodeInfos *minfos); +static int32_t dnodeReadMInfos(); +static int32_t dnodeWriteMInfos(); + +int32_t dnodeInitMInfos() { + pthread_mutex_init(&tsMInfosMutex, NULL); + dnodeResetMInfos(NULL); + int32_t ret = dnodeReadMInfos(); + if (ret == 0) { + dInfo("dnode minfos is initialized"); + } + + return ret; +} + +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); + return; + } + + for (int32_t i = 0; i < minfos->mnodeNum; ++i) { + SMnodeInfo *minfo = &minfos->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); + return; + } + } + + pthread_mutex_lock(&tsMInfosMutex); + if (minfos->mnodeNum != tsMInfos.mnodeNum) { + dnodeResetMInfos(minfos); + dnodeWriteMInfos(); + sdbUpdateAsync(); + } else { + int32_t size = sizeof(SMnodeInfos); + if (memcmp(minfos, &tsMInfos, size) != 0) { + dnodeResetMInfos(minfos); + dnodeWriteMInfos(); + sdbUpdateAsync(); + } + } + pthread_mutex_unlock(&tsMInfosMutex); +} + +void dnodeUpdateEpSetForPeer(SRpcEpSet *ep) { + if (ep->numOfEps <= 0) { + dError("mnode EP list for peer is changed, but content is invalid, discard it"); + return; + } + + pthread_mutex_lock(&tsMInfosMutex); + dInfo("mnode EP list for peer is changed, numOfEps:%d inUse:%d", ep->numOfEps, ep->inUse); + for (int i = 0; i < ep->numOfEps; ++i) { + ep->port[i] -= TSDB_PORT_DNODEDNODE; + dInfo("mnode index:%d %s:%u", i, ep->fqdn[i], ep->port[i]); + } + tsMEpSet = *ep; + pthread_mutex_unlock(&tsMInfosMutex); +} + +bool dnodeIsMasterEp(char *ep) { + pthread_mutex_lock(&tsMInfosMutex); + bool isMaster = strcmp(ep, tsMInfos.mnodeInfos[tsMEpSet.inUse].mnodeEp) == 0; + pthread_mutex_unlock(&tsMInfosMutex); + + return isMaster; +} + +void dnodeGetMInfos(SMnodeInfos *minfos) { + pthread_mutex_lock(&tsMInfosMutex); + memcpy(minfos, &tsMInfos, sizeof(SMnodeInfos)); + for (int32_t i = 0; i < tsMInfos.mnodeNum; ++i) { + minfos->mnodeInfos[i].mnodeId = htonl(tsMInfos.mnodeInfos[i].mnodeId); + } + pthread_mutex_unlock(&tsMInfosMutex); +} + +void dnodeGetEpSetForPeer(SRpcEpSet *epSet) { + pthread_mutex_lock(&tsMInfosMutex); + *epSet = tsMEpSet; + for (int i = 0; i < epSet->numOfEps; ++i) { + epSet->port[i] += TSDB_PORT_DNODEDNODE; + } + pthread_mutex_unlock(&tsMInfosMutex); +} + +void dnodeGetEpSetForShell(SRpcEpSet *epSet) { + pthread_mutex_lock(&tsMInfosMutex); + *epSet = tsMEpSet; + 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 dnodeResetMInfos(SMnodeInfos *minfos) { + if (minfos == NULL) { + tsMEpSet.numOfEps = 1; + taosGetFqdnPortFromEp(tsFirst, tsMEpSet.fqdn[0], &tsMEpSet.port[0]); + + if (strcmp(tsSecond, tsFirst) != 0) { + tsMEpSet.numOfEps = 2; + taosGetFqdnPortFromEp(tsSecond, tsMEpSet.fqdn[1], &tsMEpSet.port[1]); + } + return; + } + + if (minfos->mnodeNum == 0) return; + + int32_t size = sizeof(SMnodeInfos); + memcpy(&tsMInfos, minfos, size); + + tsMEpSet.inUse = tsMInfos.inUse; + tsMEpSet.numOfEps = tsMInfos.mnodeNum; + for (int32_t i = 0; i < tsMInfos.mnodeNum; i++) { + taosGetFqdnPortFromEp(tsMInfos.mnodeInfos[i].mnodeEp, tsMEpSet.fqdn[i], &tsMEpSet.port[i]); + } + + dnodePrintMInfos(minfos); +} + +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}; + + char file[TSDB_FILENAME_LEN + 20] = {0}; + sprintf(file, "%s/mnodeEpSet.json", tsDnodeDir); + + fp = fopen(file, "r"); + if (!fp) { + dDebug("failed to read %s, file not exist", file); + goto PARSE_MINFOS_OVER; + } + + len = fread(content, 1, maxLen, fp); + if (len <= 0) { + dError("failed to read %s, content is null", file); + goto PARSE_MINFOS_OVER; + } + + content[len] = 0; + root = cJSON_Parse(content); + if (root == NULL) { + dError("failed to read %s, invalid json format", file); + goto PARSE_MINFOS_OVER; + } + + cJSON *inUse = cJSON_GetObjectItem(root, "inUse"); + if (!inUse || inUse->type != cJSON_Number) { + dError("failed to read mnodeEpSet.json, inUse not found"); + goto PARSE_MINFOS_OVER; + } + tsMInfos.inUse = inUse->valueint; + + cJSON *nodeNum = cJSON_GetObjectItem(root, "nodeNum"); + if (!nodeNum || nodeNum->type != cJSON_Number) { + dError("failed to read mnodeEpSet.json, nodeNum not found"); + goto PARSE_MINFOS_OVER; + } + minfos.mnodeNum = nodeNum->valueint; + + cJSON *nodeInfos = cJSON_GetObjectItem(root, "nodeInfos"); + if (!nodeInfos || nodeInfos->type != cJSON_Array) { + dError("failed to read mnodeEpSet.json, nodeInfos not found"); + goto PARSE_MINFOS_OVER; + } + + int size = cJSON_GetArraySize(nodeInfos); + if (size != minfos.mnodeNum) { + dError("failed to read mnodeEpSet.json, nodeInfos size not matched"); + goto PARSE_MINFOS_OVER; + } + + for (int i = 0; i < size; ++i) { + cJSON *nodeInfo = cJSON_GetArrayItem(nodeInfos, i); + if (nodeInfo == NULL) continue; + + cJSON *nodeId = cJSON_GetObjectItem(nodeInfo, "nodeId"); + if (!nodeId || nodeId->type != cJSON_Number) { + 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); + } + + dInfo("read file %s successed", file); + dnodePrintMInfos(&minfos); + +PARSE_MINFOS_OVER: + if (content != NULL) free(content); + if (root != NULL) cJSON_Delete(root); + if (fp != NULL) fclose(fp); + terrno = 0; + + for (int32_t i = 0; i < minfos.mnodeNum; ++i) { + SMnodeInfo *mInfo = &minfos.mnodeInfos[i]; + dnodeUpdateEp(mInfo->mnodeId, mInfo->mnodeEp, NULL, NULL); + } + dnodeResetMInfos(&minfos); + return 0; +} + +static int32_t dnodeWriteMInfos() { + char file[TSDB_FILENAME_LEN + 20] = {0}; + sprintf(file, "%s/mnodeEpSet.json", tsDnodeDir); + + FILE *fp = fopen(file, "w"); + if (!fp) { + dError("failed to write %s, reason:%s", file, strerror(errno)); + return -1; + } + + int32_t len = 0; + int32_t maxLen = 2000; + char * content = calloc(1, maxLen + 1); + + len += snprintf(content + len, maxLen - len, "{\n"); + len += snprintf(content + len, maxLen - len, " \"inUse\": %d,\n", tsMInfos.inUse); + len += snprintf(content + len, maxLen - len, " \"nodeNum\": %d,\n", tsMInfos.mnodeNum); + len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n"); + for (int32_t i = 0; i < tsMInfos.mnodeNum; i++) { + len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", tsMInfos.mnodeInfos[i].mnodeId); + len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", tsMInfos.mnodeInfos[i].mnodeEp); + if (i < tsMInfos.mnodeNum - 1) { + len += snprintf(content + len, maxLen - len, " },{\n"); + } else { + len += snprintf(content + len, maxLen - len, " }]\n"); + } + } + len += snprintf(content + len, maxLen - len, "}\n"); + + fwrite(content, 1, len, fp); + fflush(fp); + fclose(fp); + free(content); + terrno = 0; + + dInfo("successed to write %s", file); + return 0; +} diff --git a/src/dnode/src/dnodeMPeer.c b/src/dnode/src/dnodeMPeer.c index 8414d79a9815287dfdfae3af5cc123304745c56d..0da7ab74e45a27a15186014197ab3bf03ef867d0 100644 --- a/src/dnode/src/dnodeMPeer.c +++ b/src/dnode/src/dnodeMPeer.c @@ -58,7 +58,7 @@ int32_t dnodeInitMnodePeer() { dDebug("dnode mpeer worker:%d is created", i); } - dDebug("dnode mpeer is opened, workers:%d qset:%p", tsMPeerPool.maxNum, tsMPeerQset); + dDebug("dnode mpeer is initialized, workers:%d qset:%p", tsMPeerPool.maxNum, tsMPeerQset); return 0; } diff --git a/src/dnode/src/dnodeMRead.c b/src/dnode/src/dnodeMRead.c index fdcbb5889f766ddebdb3f1e56ccffa0b5b129552..4ad787f26e1b9e31a043644d710c6e6de5e354aa 100644 --- a/src/dnode/src/dnodeMRead.c +++ b/src/dnode/src/dnodeMRead.c @@ -60,7 +60,7 @@ int32_t dnodeInitMnodeRead() { dDebug("dnode mread worker:%d is created", i); } - dDebug("dnode mread is opened, workers:%d qset:%p", tsMReadPool.maxNum, tsMReadQset); + dDebug("dnode mread is initialized, workers:%d qset:%p", tsMReadPool.maxNum, tsMReadQset); return 0; } diff --git a/src/dnode/src/dnodeMWrite.c b/src/dnode/src/dnodeMWrite.c index 384a0fae75088197d7fb01dd67c6e1b9d38739cd..600688b9fd2cb5623b6bed492bbc6c420b55aa81 100644 --- a/src/dnode/src/dnodeMWrite.c +++ b/src/dnode/src/dnodeMWrite.c @@ -60,7 +60,7 @@ int32_t dnodeInitMnodeWrite() { dDebug("dnode mwrite worker:%d is created", i); } - dDebug("dnode mwrite is opened, workers:%d qset:%p", tsMWritePool.maxNum, tsMWriteQset); + dDebug("dnode mwrite is initialized, workers:%d qset:%p", tsMWritePool.maxNum, tsMWriteQset); return 0; } diff --git a/src/dnode/src/dnodeMain.c b/src/dnode/src/dnodeMain.c index 97e6f2ce6debe7ffc273ca103e3f55576da0eb31..b4c174e29bc15be2796c6fab60e4ba0ab9e7b621 100644 --- a/src/dnode/src/dnodeMain.c +++ b/src/dnode/src/dnodeMain.c @@ -19,11 +19,15 @@ #include "tutil.h" #include "tconfig.h" #include "tglobal.h" +#include "twal.h" #include "dnode.h" #include "dnodeInt.h" #include "dnodeMgmt.h" #include "dnodePeer.h" #include "dnodeModule.h" +#include "dnodeEps.h" +#include "dnodeMInfos.h" +#include "dnodeCfg.h" #include "dnodeCheck.h" #include "dnodeVRead.h" #include "dnodeVWrite.h" @@ -33,26 +37,32 @@ #include "dnodeShell.h" #include "dnodeTelemetry.h" -static int32_t dnodeInitStorage(); -static void dnodeCleanupStorage(); -static void dnodeSetRunStatus(SDnodeRunStatus status); -static void dnodeCheckDataDirOpenned(char *dir); static SDnodeRunStatus tsDnodeRunStatus = TSDB_DNODE_RUN_STATUS_STOPPED; + +static int32_t dnodeInitStorage(); +static void dnodeCleanupStorage(); +static void dnodeSetRunStatus(SDnodeRunStatus status); +static void dnodeCheckDataDirOpenned(char *dir); static int32_t dnodeInitComponents(); -static void dnodeCleanupComponents(int32_t stepId); -static int dnodeCreateDir(const char *dir); +static void dnodeCleanupComponents(int32_t stepId); +static int dnodeCreateDir(const char *dir); typedef struct { const char *const name; - int (*init)(); - void (*cleanup)(); + int32_t (*init)(); + void (*cleanup)(); } SDnodeComponent; static const SDnodeComponent tsDnodeComponents[] = { {"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", dnodeInitVnodeRead, dnodeCleanupVnodeRead}, - {"vwrite", dnodeInitVnodeWrite, dnodeCleanupVnodeWrite}, + {"vwrite", dnodeInitVWrite, dnodeCleanupVWrite}, {"mread", dnodeInitMnodeRead, dnodeCleanupMnodeRead}, {"mwrite", dnodeInitMnodeWrite, dnodeCleanupMnodeWrite}, {"mpeer", dnodeInitMnodePeer, dnodeCleanupMnodePeer}, @@ -75,7 +85,9 @@ static int dnodeCreateDir(const char *dir) { static void dnodeCleanupComponents(int32_t stepId) { for (int32_t i = stepId; i >= 0; i--) { - tsDnodeComponents[i].cleanup(); + if (tsDnodeComponents[i].cleanup) { + (*tsDnodeComponents[i].cleanup)(); + } } } @@ -112,14 +124,13 @@ int32_t dnodeInitSystem() { printf("failed to init log file\n"); } - if (!taosReadGlobalCfg() || !taosCheckGlobalCfg()) { + if (!taosReadGlobalCfg()) { taosPrintGlobalCfg(); dError("TDengine read global config failed"); return -1; } - taosPrintGlobalCfg(); - dInfo("start to initialize TDengine on %s", tsLocalEp); + dInfo("start to initialize TDengine"); if (dnodeInitComponents() != 0) { return -1; @@ -198,7 +209,7 @@ static int32_t dnodeInitStorage() { dnodeCheckDataDirOpenned(tsDnodeDir); - dInfo("storage directory is initialized"); + dInfo("dnode storage is initialized at %s", tsDnodeDir); return 0; } diff --git a/src/dnode/src/dnodeMgmt.c b/src/dnode/src/dnodeMgmt.c index 968a8d9759e5618753996476b40efc3be77f7925..4a5dc31a9fc5008b1a1c8288d9798d23acfb98b6 100644 --- a/src/dnode/src/dnodeMgmt.c +++ b/src/dnode/src/dnodeMgmt.c @@ -31,12 +31,13 @@ #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" -#define MPEER_CONTENT_LEN 2000 - typedef struct { pthread_t thread; int32_t threadIndex; @@ -46,23 +47,13 @@ typedef struct { int32_t * vnodeList; } SOpenVnodeThread; -void * tsDnodeTmr = NULL; -static void * tsStatusTimer = NULL; -static uint32_t tsRebootTime; - -static SRpcEpSet tsDMnodeEpSet = {0}; -static SDMMnodeInfos tsDMnodeInfos = {0}; -static SDMDnodeCfg tsDnodeCfg = {0}; -static taos_qset tsMgmtQset = NULL; -static taos_queue tsMgmtQueue = NULL; -static pthread_t tsQthread; - -static void dnodeUpdateMnodeInfos(SDMMnodeInfos *pMnodes); -static bool dnodeReadMnodeInfos(); -static void dnodeSaveMnodeInfos(); -static void dnodeUpdateDnodeCfg(SDMDnodeCfg *pCfg); -static bool dnodeReadDnodeCfg(); -static void dnodeSaveDnodeCfg(); +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); @@ -74,7 +65,7 @@ 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 dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg); static int32_t (*dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *pMsg); int32_t dnodeInitMgmt() { @@ -86,28 +77,8 @@ int32_t dnodeInitMgmt() { dnodeProcessMgmtMsgFp[TSDB_MSG_TYPE_MD_CREATE_MNODE] = dnodeProcessCreateMnodeMsg; dnodeAddClientRspHandle(TSDB_MSG_TYPE_DM_STATUS_RSP, dnodeProcessStatusRsp); - dnodeReadDnodeCfg(); tsRebootTime = taosGetTimestampSec(); - if (!dnodeReadMnodeInfos()) { - memset(&tsDMnodeEpSet, 0, sizeof(SRpcEpSet)); - memset(&tsDMnodeInfos, 0, sizeof(SDMMnodeInfos)); - - tsDMnodeEpSet.numOfEps = 1; - taosGetFqdnPortFromEp(tsFirst, tsDMnodeEpSet.fqdn[0], &tsDMnodeEpSet.port[0]); - - if (strcmp(tsSecond, tsFirst) != 0) { - tsDMnodeEpSet.numOfEps = 2; - taosGetFqdnPortFromEp(tsSecond, tsDMnodeEpSet.fqdn[1], &tsDMnodeEpSet.port[1]); - } - } else { - tsDMnodeEpSet.inUse = tsDMnodeInfos.inUse; - tsDMnodeEpSet.numOfEps = tsDMnodeInfos.nodeNum; - for (int32_t i = 0; i < tsDMnodeInfos.nodeNum; i++) { - taosGetFqdnPortFromEp(tsDMnodeInfos.nodeInfos[i].nodeEp, tsDMnodeEpSet.fqdn[i], &tsDMnodeEpSet.port[i]); - } - } - int32_t code = vnodeInitResources(); if (code != TSDB_CODE_SUCCESS) { dnodeCleanupMgmt(); @@ -381,7 +352,7 @@ static void dnodeCloseVnodes() { } static void* dnodeParseVnodeMsg(SRpcMsg *rpcMsg) { - SMDCreateVnodeMsg *pCreate = rpcMsg->pCont; + 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); @@ -404,7 +375,7 @@ static void* dnodeParseVnodeMsg(SRpcMsg *rpcMsg) { } static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) { - SMDCreateVnodeMsg *pCreate = dnodeParseVnodeMsg(rpcMsg); + SCreateVnodeMsg *pCreate = dnodeParseVnodeMsg(rpcMsg); void *pVnode = vnodeAcquire(pCreate->cfg.vgId); if (pVnode != NULL) { @@ -418,7 +389,7 @@ static int32_t dnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) { } static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) { - SMDAlterVnodeMsg *pAlter = dnodeParseVnodeMsg(rpcMsg); + SAlterVnodeMsg *pAlter = dnodeParseVnodeMsg(rpcMsg); void *pVnode = vnodeAcquire(pAlter->cfg.vgId); if (pVnode != NULL) { @@ -433,14 +404,14 @@ static int32_t dnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) { } static int32_t dnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) { - SMDDropVnodeMsg *pDrop = rpcMsg->pCont; + SDropVnodeMsg *pDrop = rpcMsg->pCont; pDrop->vgId = htonl(pDrop->vgId); return vnodeDrop(pDrop->vgId); } static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg) { -// SMDAlterStreamMsg *pStream = pCont; +// SAlterStreamMsg *pStream = pCont; // pStream->uid = htobe64(pStream->uid); // pStream->stime = htobe64(pStream->stime); // pStream->vnode = htonl(pStream->vnode); @@ -453,12 +424,12 @@ static int32_t dnodeProcessAlterStreamMsg(SRpcMsg *pMsg) { } static int32_t dnodeProcessConfigDnodeMsg(SRpcMsg *pMsg) { - SMDCfgDnodeMsg *pCfg = pMsg->pCont; + SCfgDnodeMsg *pCfg = pMsg->pCont; return taosCfgDynamicOptions(pCfg->config); } static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) { - SMDCreateMnodeMsg *pCfg = pMsg->pCont; + 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()); @@ -470,10 +441,10 @@ static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) { return TSDB_CODE_MND_DNODE_EP_NOT_CONFIGURED; } - dDebug("dnodeId:%d, create mnode msg is received from mnodes, numOfMnodes:%d", pCfg->dnodeId, pCfg->mnodes.nodeNum); - for (int i = 0; i < pCfg->mnodes.nodeNum; ++i) { - pCfg->mnodes.nodeInfos[i].nodeId = htonl(pCfg->mnodes.nodeInfos[i].nodeId); - dDebug("mnode index:%d, mnode:%d:%s", i, pCfg->mnodes.nodeInfos[i].nodeId, pCfg->mnodes.nodeInfos[i].nodeEp); + 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); @@ -481,34 +452,6 @@ static int32_t dnodeProcessCreateMnodeMsg(SRpcMsg *pMsg) { return TSDB_CODE_SUCCESS; } -void dnodeUpdateMnodeEpSetForPeer(SRpcEpSet *pEpSet) { - if (pEpSet->numOfEps <= 0) { - dError("mnode EP list for peer is changed, but content is invalid, discard it"); - return; - } - - dInfo("mnode EP list for peer is changed, numOfEps:%d inUse:%d", pEpSet->numOfEps, pEpSet->inUse); - for (int i = 0; i < pEpSet->numOfEps; ++i) { - pEpSet->port[i] -= TSDB_PORT_DNODEDNODE; - dInfo("mnode index:%d %s:%u", i, pEpSet->fqdn[i], pEpSet->port[i]); - } - - tsDMnodeEpSet = *pEpSet; -} - -void dnodeGetMnodeEpSetForPeer(void *epSetRaw) { - SRpcEpSet *epSet = epSetRaw; - *epSet = tsDMnodeEpSet; - - for (int i=0; inumOfEps; ++i) - epSet->port[i] += TSDB_PORT_DNODEDNODE; -} - -void dnodeGetMnodeEpSetForShell(void *epSetRaw) { - SRpcEpSet *epSet = epSetRaw; - *epSet = tsDMnodeEpSet; -} - static void dnodeProcessStatusRsp(SRpcMsg *pMsg) { if (pMsg->code != TSDB_CODE_SUCCESS) { dError("status rsp is received, error:%s", tstrerror(pMsg->code)); @@ -516,202 +459,24 @@ static void dnodeProcessStatusRsp(SRpcMsg *pMsg) { return; } - SDMStatusRsp *pStatusRsp = pMsg->pCont; - SDMMnodeInfos *pMnodes = &pStatusRsp->mnodes; - if (pMnodes->nodeNum <= 0) { - dError("status msg is invalid, num of ips is %d", pMnodes->nodeNum); - taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); - return; - } + SStatusRsp *pStatusRsp = pMsg->pCont; + SMnodeInfos *minfos = &pStatusRsp->mnodes; + dnodeUpdateMInfos(minfos); - SDMDnodeCfg *pCfg = &pStatusRsp->dnodeCfg; - pCfg->numOfVnodes = htonl(pCfg->numOfVnodes); + SDnodeCfg *pCfg = &pStatusRsp->dnodeCfg; + pCfg->numOfVnodes = htonl(pCfg->numOfVnodes); pCfg->moduleStatus = htonl(pCfg->moduleStatus); pCfg->dnodeId = htonl(pCfg->dnodeId); - - for (int32_t i = 0; i < pMnodes->nodeNum; ++i) { - SDMMnodeInfo *pMnodeInfo = &pMnodes->nodeInfos[i]; - pMnodeInfo->nodeId = htonl(pMnodeInfo->nodeId); - } + dnodeUpdateCfg(pCfg); vnodeSetAccess(pStatusRsp->vgAccess, pCfg->numOfVnodes); - // will not set mnode in status msg - // dnodeProcessModuleStatus(pCfg->moduleStatus); - dnodeUpdateDnodeCfg(pCfg); + SDnodeEps *pEps = (SDnodeEps *)((char *)pStatusRsp->vgAccess + pCfg->numOfVnodes * sizeof(SVgroupAccess)); + dnodeUpdateEps(pEps); - dnodeUpdateMnodeInfos(pMnodes); taosTmrReset(dnodeSendStatusMsg, tsStatusInterval * 1000, NULL, tsDnodeTmr, &tsStatusTimer); } -static bool dnodeCheckMnodeInfos(SDMMnodeInfos *pMnodes) { - if (pMnodes->nodeNum <= 0 || pMnodes->nodeNum > 3) { - dError("invalid mnode infos, num:%d", pMnodes->nodeNum); - return false; - } - - for (int32_t i = 0; i < pMnodes->nodeNum; ++i) { - SDMMnodeInfo *pMnodeInfo = &pMnodes->nodeInfos[i]; - if (pMnodeInfo->nodeId <= 0 || strlen(pMnodeInfo->nodeEp) <= 5) { - dError("invalid mnode info:%d, nodeId:%d nodeEp:%s", i, pMnodeInfo->nodeId, pMnodeInfo->nodeEp); - return false; - } - } - - return true; -} - -static void dnodeUpdateMnodeInfos(SDMMnodeInfos *pMnodes) { - bool mnodesChanged = (memcmp(&tsDMnodeInfos, pMnodes, sizeof(SDMMnodeInfos)) != 0); - bool mnodesNotInit = (tsDMnodeInfos.nodeNum == 0); - if (!(mnodesChanged || mnodesNotInit)) return; - - if (!dnodeCheckMnodeInfos(pMnodes)) return; - - memcpy(&tsDMnodeInfos, pMnodes, sizeof(SDMMnodeInfos)); - dInfo("mnode infos is changed, nodeNum:%d inUse:%d", tsDMnodeInfos.nodeNum, tsDMnodeInfos.inUse); - for (int32_t i = 0; i < tsDMnodeInfos.nodeNum; i++) { - dInfo("mnode index:%d, %s", tsDMnodeInfos.nodeInfos[i].nodeId, tsDMnodeInfos.nodeInfos[i].nodeEp); - } - - tsDMnodeEpSet.inUse = tsDMnodeInfos.inUse; - tsDMnodeEpSet.numOfEps = tsDMnodeInfos.nodeNum; - for (int32_t i = 0; i < tsDMnodeInfos.nodeNum; i++) { - taosGetFqdnPortFromEp(tsDMnodeInfos.nodeInfos[i].nodeEp, tsDMnodeEpSet.fqdn[i], &tsDMnodeEpSet.port[i]); - } - - dnodeSaveMnodeInfos(); - sdbUpdateAsync(); -} - -static bool dnodeReadMnodeInfos() { - char ipFile[TSDB_FILENAME_LEN*2] = {0}; - - sprintf(ipFile, "%s/mnodeEpSet.json", tsDnodeDir); - FILE *fp = fopen(ipFile, "r"); - if (!fp) { - dDebug("failed to read mnodeEpSet.json, file not exist"); - return false; - } - - bool ret = false; - int maxLen = 2000; - char *content = calloc(1, maxLen + 1); - int len = fread(content, 1, maxLen, fp); - if (len <= 0) { - free(content); - fclose(fp); - dError("failed to read mnodeEpSet.json, content is null"); - return false; - } - - content[len] = 0; - cJSON* root = cJSON_Parse(content); - if (root == NULL) { - dError("failed to read mnodeEpSet.json, invalid json format"); - goto PARSE_OVER; - } - - cJSON* inUse = cJSON_GetObjectItem(root, "inUse"); - if (!inUse || inUse->type != cJSON_Number) { - dError("failed to read mnodeEpSet.json, inUse not found"); - goto PARSE_OVER; - } - tsDMnodeInfos.inUse = inUse->valueint; - - cJSON* nodeNum = cJSON_GetObjectItem(root, "nodeNum"); - if (!nodeNum || nodeNum->type != cJSON_Number) { - dError("failed to read mnodeEpSet.json, nodeNum not found"); - goto PARSE_OVER; - } - tsDMnodeInfos.nodeNum = nodeNum->valueint; - - cJSON* nodeInfos = cJSON_GetObjectItem(root, "nodeInfos"); - if (!nodeInfos || nodeInfos->type != cJSON_Array) { - dError("failed to read mnodeEpSet.json, nodeInfos not found"); - goto PARSE_OVER; - } - - int size = cJSON_GetArraySize(nodeInfos); - if (size != tsDMnodeInfos.nodeNum) { - dError("failed to read mnodeEpSet.json, nodeInfos size not matched"); - goto PARSE_OVER; - } - - for (int i = 0; i < size; ++i) { - cJSON *nodeInfo = cJSON_GetArrayItem(nodeInfos, i); - if (nodeInfo == NULL) continue; - - cJSON *nodeId = cJSON_GetObjectItem(nodeInfo, "nodeId"); - if (!nodeId || nodeId->type != cJSON_Number) { - dError("failed to read mnodeEpSet.json, nodeId not found"); - goto PARSE_OVER; - } - tsDMnodeInfos.nodeInfos[i].nodeId = 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_OVER; - } - strncpy(tsDMnodeInfos.nodeInfos[i].nodeEp, nodeEp->valuestring, TSDB_EP_LEN); - } - - ret = true; - - dInfo("read mnode epSet successed, numOfEps:%d inUse:%d", tsDMnodeInfos.nodeNum, tsDMnodeInfos.inUse); - for (int32_t i = 0; i < tsDMnodeInfos.nodeNum; i++) { - dInfo("mnode:%d, %s", tsDMnodeInfos.nodeInfos[i].nodeId, tsDMnodeInfos.nodeInfos[i].nodeEp); - } - -PARSE_OVER: - free(content); - cJSON_Delete(root); - fclose(fp); - return ret; -} - -static void dnodeSaveMnodeInfos() { - char ipFile[TSDB_FILENAME_LEN] = {0}; - sprintf(ipFile, "%s/mnodeEpSet.json", tsDnodeDir); - FILE *fp = fopen(ipFile, "w"); - if (!fp) return; - - int32_t len = 0; - int32_t maxLen = 2000; - char * content = calloc(1, maxLen + 1); - - len += snprintf(content + len, maxLen - len, "{\n"); - len += snprintf(content + len, maxLen - len, " \"inUse\": %d,\n", tsDMnodeInfos.inUse); - len += snprintf(content + len, maxLen - len, " \"nodeNum\": %d,\n", tsDMnodeInfos.nodeNum); - len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n"); - for (int32_t i = 0; i < tsDMnodeInfos.nodeNum; i++) { - len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", tsDMnodeInfos.nodeInfos[i].nodeId); - len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", tsDMnodeInfos.nodeInfos[i].nodeEp); - if (i < tsDMnodeInfos.nodeNum -1) { - len += snprintf(content + len, maxLen - len, " },{\n"); - } else { - len += snprintf(content + len, maxLen - len, " }]\n"); - } - } - len += snprintf(content + len, maxLen - len, "}\n"); - - fwrite(content, 1, len, fp); - fflush(fp); - fclose(fp); - free(content); - - dInfo("save mnode epSet successed"); -} - -char *dnodeGetMnodeMasterEp() { - return tsDMnodeInfos.nodeInfos[tsDMnodeEpSet.inUse].nodeEp; -} - -void* dnodeGetMnodeInfos() { - return &tsDMnodeInfos; -} - static void dnodeSendStatusMsg(void *handle, void *tmrId) { if (tsDnodeTmr == NULL) { dError("dnode timer is already released"); @@ -724,22 +489,21 @@ static void dnodeSendStatusMsg(void *handle, void *tmrId) { return; } - int32_t contLen = sizeof(SDMStatusMsg) + TSDB_MAX_VNODES * sizeof(SVnodeLoad); - SDMStatusMsg *pStatus = rpcMallocCont(contLen); + 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; } - //strcpy(pStatus->dnodeName, tsDnodeName); + dnodeGetCfg(&pStatus->dnodeId, pStatus->clusterId); + pStatus->dnodeId = htonl(dnodeGetDnodeId()); pStatus->version = htonl(tsVersion); - pStatus->dnodeId = htonl(tsDnodeCfg.dnodeId); pStatus->lastReboot = htonl(tsRebootTime); pStatus->numOfCores = htons((uint16_t) tsNumOfCores); pStatus->diskAvailable = tsAvailDataDirGB; pStatus->alternativeRole = (uint8_t) tsAlternativeRole; - tstrncpy(pStatus->clusterId, tsDnodeCfg.clusterId, TSDB_CLUSTER_ID_LEN); tstrncpy(pStatus->dnodeEp, tsLocalEp, TSDB_EP_LEN); // fill cluster cfg parameters @@ -759,7 +523,7 @@ static void dnodeSendStatusMsg(void *handle, void *tmrId) { tstrncpy(pStatus->clusterCfg.charset, tsCharset, TSDB_LOCALE_LEN); vnodeBuildStatusMsg(pStatus); - contLen = sizeof(SDMStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad); + contLen = sizeof(SStatusMsg) + pStatus->openVnodes * sizeof(SVnodeLoad); pStatus->openVnodes = htons(pStatus->openVnodes); SRpcMsg rpcMsg = { @@ -769,110 +533,19 @@ static void dnodeSendStatusMsg(void *handle, void *tmrId) { }; SRpcEpSet epSet; - dnodeGetMnodeEpSetForPeer(&epSet); + dnodeGetEpSetForPeer(&epSet); dnodeSendMsgToDnode(&epSet, &rpcMsg); } -static bool dnodeReadDnodeCfg() { - char dnodeCfgFile[TSDB_FILENAME_LEN*2] = {0}; - - sprintf(dnodeCfgFile, "%s/dnodeCfg.json", tsDnodeDir); - - FILE *fp = fopen(dnodeCfgFile, "r"); - if (!fp) { - dDebug("failed to read dnodeCfg.json, file not exist"); - return false; - } - - bool ret = false; - int maxLen = 100; - char *content = calloc(1, maxLen + 1); - int len = fread(content, 1, maxLen, fp); - if (len <= 0) { - free(content); - fclose(fp); - dError("failed to read dnodeCfg.json, content is null"); - return false; - } - - content[len] = 0; - cJSON* root = cJSON_Parse(content); - if (root == NULL) { - dError("failed to read dnodeCfg.json, invalid json format"); - goto PARSE_CFG_OVER; - } - - cJSON* dnodeId = cJSON_GetObjectItem(root, "dnodeId"); - if (!dnodeId || dnodeId->type != cJSON_Number) { - dError("failed to read dnodeCfg.json, dnodeId not found"); - goto PARSE_CFG_OVER; - } - tsDnodeCfg.dnodeId = dnodeId->valueint; - - cJSON* clusterId = cJSON_GetObjectItem(root, "clusterId"); - if (!clusterId || clusterId->type != cJSON_String) { - dError("failed to read dnodeCfg.json, clusterId not found"); - goto PARSE_CFG_OVER; - } - tstrncpy(tsDnodeCfg.clusterId, clusterId->valuestring, TSDB_CLUSTER_ID_LEN); - - ret = true; - - dInfo("read numOfVnodes successed, dnodeId:%d", tsDnodeCfg.dnodeId); - -PARSE_CFG_OVER: - free(content); - cJSON_Delete(root); - fclose(fp); - return ret; -} - -static void dnodeSaveDnodeCfg() { - char dnodeCfgFile[TSDB_FILENAME_LEN] = {0}; - sprintf(dnodeCfgFile, "%s/dnodeCfg.json", tsDnodeDir); - - FILE *fp = fopen(dnodeCfgFile, "w"); - if (!fp) return; - - int32_t len = 0; - int32_t maxLen = 200; - char * content = calloc(1, maxLen + 1); - - len += snprintf(content + len, maxLen - len, "{\n"); - len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", tsDnodeCfg.dnodeId); - len += snprintf(content + len, maxLen - len, " \"clusterId\": \"%s\"\n", tsDnodeCfg.clusterId); - len += snprintf(content + len, maxLen - len, "}\n"); - - fwrite(content, 1, len, fp); - fflush(fp); - fclose(fp); - free(content); - - dInfo("save dnodeId successed"); -} - -void dnodeUpdateDnodeCfg(SDMDnodeCfg *pCfg) { - if (tsDnodeCfg.dnodeId == 0) { - dInfo("dnodeId is set to %d, clusterId is set to %s", pCfg->dnodeId, pCfg->clusterId); - tsDnodeCfg.dnodeId = pCfg->dnodeId; - tstrncpy(tsDnodeCfg.clusterId, pCfg->clusterId, TSDB_CLUSTER_ID_LEN); - dnodeSaveDnodeCfg(); - } -} - -int32_t dnodeGetDnodeId() { - return tsDnodeCfg.dnodeId; -} - void dnodeSendRedirectMsg(SRpcMsg *rpcMsg, bool forShell) { SRpcConnInfo connInfo = {0}; rpcGetConnInfo(rpcMsg->handle, &connInfo); SRpcEpSet epSet = {0}; if (forShell) { - dnodeGetMnodeEpSetForShell(&epSet); + dnodeGetEpSetForShell(&epSet); } else { - dnodeGetMnodeEpSetForPeer(&epSet); + dnodeGetEpSetForPeer(&epSet); } dDebug("msg:%s will be redirected, dnodeIp:%s user:%s, numOfEps:%d inUse:%d", taosMsg[rpcMsg->msgType], diff --git a/src/dnode/src/dnodeModule.c b/src/dnode/src/dnodeModule.c index 46376159c6782efde7adbd19c75af83aa4cde397..bd9500ba51226d138fd3fe52f027d144d289681b 100644 --- a/src/dnode/src/dnodeModule.c +++ b/src/dnode/src/dnodeModule.c @@ -114,6 +114,7 @@ int32_t dnodeInitModules() { } } + dInfo("dnode modules is initialized"); return 0; } @@ -146,8 +147,8 @@ void dnodeProcessModuleStatus(uint32_t moduleStatus) { } } -bool dnodeStartMnode(void *pMnodes) { - SDMMnodeInfos *mnodes = pMnodes; +bool dnodeStartMnode(SMnodeInfos *minfos) { + SMnodeInfos *mnodes = minfos; if (tsModuleStatus & (1 << TSDB_MOD_MNODE)) { dDebug("mnode module is already started, module status:%d", tsModuleStatus); diff --git a/src/dnode/src/dnodePeer.c b/src/dnode/src/dnodePeer.c index 3bc2f7b48b319f3c9e6215a05463ebedc74035fa..c6fc2b9e36e1c51374d8a7b164ff32a97c0cf9d0 100644 --- a/src/dnode/src/dnodePeer.c +++ b/src/dnode/src/dnodePeer.c @@ -28,8 +28,8 @@ #include "dnodeMgmt.h" #include "dnodeVWrite.h" #include "dnodeMPeer.h" +#include "dnodeMInfos.h" -extern void dnodeUpdateMnodeEpSetForPeer(SRpcEpSet *pEpSet); static void (*dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *); static void dnodeProcessReqMsgFromDnode(SRpcMsg *pMsg, SRpcEpSet *); static void (*dnodeProcessRspMsgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *rpcMsg); @@ -38,10 +38,10 @@ static void *tsDnodeServerRpc = NULL; static void *tsDnodeClientRpc = NULL; int32_t dnodeInitServer() { - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = dnodeDispatchToVnodeWriteQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = dnodeDispatchToVnodeWriteQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = dnodeDispatchToVnodeWriteQueue; - dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = dnodeDispatchToVnodeWriteQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = dnodeDispatchToVWriteQueue; + dnodeProcessReqMsgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = dnodeDispatchToVWriteQueue; + 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; @@ -72,7 +72,7 @@ int32_t dnodeInitServer() { return -1; } - dInfo("inter-dnodes RPC server is opened"); + dInfo("dnode inter-dnodes RPC server is initialized"); return 0; } @@ -137,7 +137,7 @@ int32_t dnodeInitClient() { return -1; } - dInfo("inter-dnodes rpc client is opened"); + dInfo("dnode inter-dnodes rpc client is initialized"); return 0; } @@ -145,13 +145,13 @@ void dnodeCleanupClient() { if (tsDnodeClientRpc) { rpcClose(tsDnodeClientRpc); tsDnodeClientRpc = NULL; - dInfo("inter-dnodes rpc client is closed"); + dInfo("dnode inter-dnodes rpc client is closed"); } } static void dnodeProcessRspFromDnode(SRpcMsg *pMsg, SRpcEpSet *pEpSet) { if (pMsg->msgType == TSDB_MSG_TYPE_DM_STATUS_RSP && pEpSet) { - dnodeUpdateMnodeEpSetForPeer(pEpSet); + dnodeUpdateEpSetForPeer(pEpSet); } if (dnodeProcessRspMsgFp[pMsg->msgType]) { @@ -173,7 +173,7 @@ void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg) { void dnodeSendMsgToMnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp) { SRpcEpSet epSet = {0}; - dnodeGetMnodeEpSetForPeer(&epSet); + dnodeGetEpSetForPeer(&epSet); rpcSendRecv(tsDnodeClientRpc, &epSet, rpcMsg, rpcRsp); } diff --git a/src/dnode/src/dnodeShell.c b/src/dnode/src/dnodeShell.c index 4c6c2100e048e9aaf23f4155d3c54604104a8f9b..33cda607937bac0032ca202ff1e746aec401f618 100644 --- a/src/dnode/src/dnodeShell.c +++ b/src/dnode/src/dnodeShell.c @@ -38,10 +38,10 @@ static int32_t tsDnodeQueryReqNum = 0; static int32_t tsDnodeSubmitReqNum = 0; int32_t dnodeInitShell() { - dnodeProcessShellMsgFp[TSDB_MSG_TYPE_SUBMIT] = dnodeDispatchToVnodeWriteQueue; + dnodeProcessShellMsgFp[TSDB_MSG_TYPE_SUBMIT] = dnodeDispatchToVWriteQueue; dnodeProcessShellMsgFp[TSDB_MSG_TYPE_QUERY] = dnodeDispatchToVnodeReadQueue; dnodeProcessShellMsgFp[TSDB_MSG_TYPE_FETCH] = dnodeDispatchToVnodeReadQueue; - dnodeProcessShellMsgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = dnodeDispatchToVnodeWriteQueue; + dnodeProcessShellMsgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = dnodeDispatchToVWriteQueue; // the following message shall be treated as mnode write dnodeProcessShellMsgFp[TSDB_MSG_TYPE_CM_CREATE_ACCT] = dnodeDispatchToMnodeWriteQueue; @@ -97,7 +97,7 @@ int32_t dnodeInitShell() { return -1; } - dInfo("shell rpc server is opened"); + dInfo("dnode shell rpc server is initialized"); return 0; } @@ -146,12 +146,12 @@ static int dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char int code = mnodeRetriveAuth(user, spi, encrypt, secret, ckey); if (code != TSDB_CODE_APP_NOT_READY) return code; - SDMAuthMsg *pMsg = rpcMallocCont(sizeof(SDMAuthMsg)); + SAuthMsg *pMsg = rpcMallocCont(sizeof(SAuthMsg)); tstrncpy(pMsg->user, user, sizeof(pMsg->user)); SRpcMsg rpcMsg = {0}; rpcMsg.pCont = pMsg; - rpcMsg.contLen = sizeof(SDMAuthMsg); + rpcMsg.contLen = sizeof(SAuthMsg); rpcMsg.msgType = TSDB_MSG_TYPE_DM_AUTH; dDebug("user:%s, send auth msg to mnodes", user); @@ -161,7 +161,7 @@ static int dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char if (rpcRsp.code != 0) { dError("user:%s, auth msg received from mnodes, error:%s", user, tstrerror(rpcRsp.code)); } else { - SDMAuthRsp *pRsp = rpcRsp.pCont; + SAuthRsp *pRsp = rpcRsp.pCont; dDebug("user:%s, auth msg received from mnodes", user); memcpy(secret, pRsp->secret, TSDB_KEY_LEN); memcpy(ckey, pRsp->ckey, TSDB_KEY_LEN); @@ -176,8 +176,8 @@ static int dnodeRetrieveUserAuthInfo(char *user, char *spi, char *encrypt, char void *dnodeSendCfgTableToRecv(int32_t vgId, int32_t tid) { dDebug("vgId:%d, tid:%d send config table msg to mnode", vgId, tid); - int32_t contLen = sizeof(SDMConfigTableMsg); - SDMConfigTableMsg *pMsg = rpcMallocCont(contLen); + int32_t contLen = sizeof(SConfigTableMsg); + SConfigTableMsg *pMsg = rpcMallocCont(contLen); pMsg->dnodeId = htonl(dnodeGetDnodeId()); pMsg->vgId = htonl(vgId); diff --git a/src/dnode/src/dnodeTelemetry.c b/src/dnode/src/dnodeTelemetry.c index 4fdc0b8a73b58efa801f33ea99ffb7d123a72dcf..e973f9901f19b7aa6f4d4958d88e01ba258a82bb 100644 --- a/src/dnode/src/dnodeTelemetry.c +++ b/src/dnode/src/dnodeTelemetry.c @@ -268,7 +268,7 @@ static void dnodeGetEmail(char* filepath) { return; } - if (taosTRead(fd, (void *)tsEmail, TSDB_FQDN_LEN) < 0) { + if (taosRead(fd, (void *)tsEmail, TSDB_FQDN_LEN) < 0) { dError("failed to read %d bytes from file %s since %s", TSDB_FQDN_LEN, filepath, strerror(errno)); } @@ -299,6 +299,7 @@ int32_t dnodeInitTelemetry() { dTrace("failed to create telemetry thread, reason:%s", strerror(errno)); } + dInfo("dnode telemetry is initialized"); return 0; } diff --git a/src/dnode/src/dnodeVRead.c b/src/dnode/src/dnodeVRead.c index e61158ef30dddd6037af36c61be1fd522f8af10b..abf3cb527d3cc918161e2df10fd9f939adabfc30 100644 --- a/src/dnode/src/dnodeVRead.c +++ b/src/dnode/src/dnodeVRead.c @@ -61,7 +61,7 @@ int32_t dnodeInitVnodeRead() { pWorker->workerId = i; } - dInfo("dnode read is opened, min worker:%d max worker:%d", readPool.min, readPool.max); + dInfo("dnode read is initialized, min worker:%d max worker:%d", readPool.min, readPool.max); return 0; } @@ -132,7 +132,7 @@ void dnodeDispatchToVnodeReadQueue(SRpcMsg *pMsg) { } } -void *dnodeAllocateVnodeRqueue(void *pVnode) { +void *dnodeAllocVReadQueue(void *pVnode) { pthread_mutex_lock(&readPool.mutex); taos_queue queue = taosOpenQueue(); if (queue == NULL) { @@ -167,7 +167,7 @@ void *dnodeAllocateVnodeRqueue(void *pVnode) { return queue; } -void dnodeFreeVnodeRqueue(void *rqueue) { +void dnodeFreeVReadQueue(void *rqueue) { taosCloseQueue(rqueue); // dynamically adjust the number of threads diff --git a/src/dnode/src/dnodeVWrite.c b/src/dnode/src/dnodeVWrite.c index f2740bf6b810197283926a602db4fc423067e231..3a820f180c73b856a4693d668f91d675deb5b75b 100644 --- a/src/dnode/src/dnodeVWrite.c +++ b/src/dnode/src/dnodeVWrite.c @@ -15,74 +15,65 @@ #define _DEFAULT_SOURCE #include "os.h" -#include "taosmsg.h" -#include "taoserror.h" -#include "tutil.h" +#include "tglobal.h" #include "tqueue.h" -#include "trpc.h" #include "tsdb.h" #include "twal.h" -#include "tdataformat.h" -#include "tglobal.h" #include "tsync.h" #include "vnode.h" -#include "dnodeInt.h" #include "syncInt.h" -#include "dnodeVWrite.h" -#include "dnodeMgmt.h" +#include "dnodeInt.h" typedef struct { - taos_qall qall; - taos_qset qset; // queue set - pthread_t thread; // thread - int32_t workerId; // worker ID + taos_qall qall; + taos_qset qset; // queue set + int32_t workerId; // worker ID + pthread_t thread; // thread } SWriteWorker; typedef struct { - SRspRet rspRet; - int32_t processedCount; - int32_t code; - void *pCont; - int32_t contLen; - SRpcMsg rpcMsg; + SRspRet rspRet; + SRpcMsg rpcMsg; + int32_t processedCount; + int32_t code; + int32_t contLen; + void * pCont; } SWriteMsg; typedef struct { - int32_t max; // max number of workers - int32_t nextId; // from 0 to max-1, cyclic - SWriteWorker *writeWorker; + int32_t max; // max number of workers + int32_t nextId; // from 0 to max-1, cyclic + SWriteWorker *worker; pthread_mutex_t mutex; } SWriteWorkerPool; +static SWriteWorkerPool tsVWriteWP; static void *dnodeProcessWriteQueue(void *param); -static void dnodeHandleIdleWorker(SWriteWorker *pWorker); -SWriteWorkerPool wWorkerPool; +int32_t dnodeInitVWrite() { + tsVWriteWP.max = tsNumOfCores; + tsVWriteWP.worker = (SWriteWorker *)tcalloc(sizeof(SWriteWorker), tsVWriteWP.max); + if (tsVWriteWP.worker == NULL) return -1; + pthread_mutex_init(&tsVWriteWP.mutex, NULL); -int32_t dnodeInitVnodeWrite() { - wWorkerPool.max = tsNumOfCores; - wWorkerPool.writeWorker = (SWriteWorker *)calloc(sizeof(SWriteWorker), wWorkerPool.max); - if (wWorkerPool.writeWorker == NULL) return -1; - pthread_mutex_init(&wWorkerPool.mutex, NULL); - - for (int32_t i = 0; i < wWorkerPool.max; ++i) { - wWorkerPool.writeWorker[i].workerId = i; + for (int32_t i = 0; i < tsVWriteWP.max; ++i) { + tsVWriteWP.worker[i].workerId = i; } - dInfo("dnode write is opened, max worker %d", wWorkerPool.max); + dInfo("dnode vwrite is initialized, max worker %d", tsVWriteWP.max); return 0; } -void dnodeCleanupVnodeWrite() { - for (int32_t i = 0; i < wWorkerPool.max; ++i) { - SWriteWorker *pWorker = wWorkerPool.writeWorker + i; +void dnodeCleanupVWrite() { + for (int32_t i = 0; i < tsVWriteWP.max; ++i) { + SWriteWorker *pWorker = tsVWriteWP.worker + i; if (pWorker->thread) { taosQsetThreadResume(pWorker->qset); } } - for (int32_t i = 0; i < wWorkerPool.max; ++i) { - SWriteWorker *pWorker = wWorkerPool.writeWorker + i; + for (int32_t i = 0; i < tsVWriteWP.max; ++i) { + SWriteWorker *pWorker = tsVWriteWP.worker + i; if (pWorker->thread) { pthread_join(pWorker->thread, NULL); taosFreeQall(pWorker->qall); @@ -90,13 +81,13 @@ void dnodeCleanupVnodeWrite() { } } - pthread_mutex_destroy(&wWorkerPool.mutex); - free(wWorkerPool.writeWorker); - dInfo("dnode write is closed"); + pthread_mutex_destroy(&tsVWriteWP.mutex); + tfree(tsVWriteWP.worker); + dInfo("dnode vwrite is closed"); } -void dnodeDispatchToVnodeWriteQueue(SRpcMsg *pMsg) { - char *pCont = (char *)pMsg->pCont; +void dnodeDispatchToVWriteQueue(SRpcMsg *pMsg) { + char *pCont = pMsg->pCont; if (pMsg->msgType == TSDB_MSG_TYPE_SUBMIT) { SMsgDesc *pDesc = (SMsgDesc *)pCont; @@ -111,7 +102,7 @@ void dnodeDispatchToVnodeWriteQueue(SRpcMsg *pMsg) { taos_queue queue = vnodeAcquireWqueue(pHead->vgId); if (queue) { // put message into queue - SWriteMsg *pWrite = (SWriteMsg *)taosAllocateQitem(sizeof(SWriteMsg)); + SWriteMsg *pWrite = taosAllocateQitem(sizeof(SWriteMsg)); pWrite->rpcMsg = *pMsg; pWrite->pCont = pCont; pWrite->contLen = pHead->contLen; @@ -130,12 +121,12 @@ void dnodeDispatchToVnodeWriteQueue(SRpcMsg *pMsg) { } } -void *dnodeAllocateVnodeWqueue(void *pVnode) { - pthread_mutex_lock(&wWorkerPool.mutex); - SWriteWorker *pWorker = wWorkerPool.writeWorker + wWorkerPool.nextId; +void *dnodeAllocVWriteQueue(void *pVnode) { + pthread_mutex_lock(&tsVWriteWP.mutex); + SWriteWorker *pWorker = tsVWriteWP.worker + tsVWriteWP.nextId; void *queue = taosOpenQueue(); if (queue == NULL) { - pthread_mutex_unlock(&wWorkerPool.mutex); + pthread_mutex_unlock(&tsVWriteWP.mutex); return NULL; } @@ -143,7 +134,7 @@ void *dnodeAllocateVnodeWqueue(void *pVnode) { pWorker->qset = taosOpenQset(); if (pWorker->qset == NULL) { taosCloseQueue(queue); - pthread_mutex_unlock(&wWorkerPool.mutex); + pthread_mutex_unlock(&tsVWriteWP.mutex); return NULL; } @@ -152,7 +143,7 @@ void *dnodeAllocateVnodeWqueue(void *pVnode) { if (pWorker->qall == NULL) { taosCloseQset(pWorker->qset); taosCloseQueue(queue); - pthread_mutex_unlock(&wWorkerPool.mutex); + pthread_mutex_unlock(&tsVWriteWP.mutex); return NULL; } pthread_attr_t thAttr; @@ -160,37 +151,35 @@ void *dnodeAllocateVnodeWqueue(void *pVnode) { pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); if (pthread_create(&pWorker->thread, &thAttr, dnodeProcessWriteQueue, pWorker) != 0) { - dError("failed to create thread to process read queue, reason:%s", strerror(errno)); + dError("failed to create thread to process vwrite queue since %s", strerror(errno)); taosFreeQall(pWorker->qall); taosCloseQset(pWorker->qset); taosCloseQueue(queue); queue = NULL; } else { - dDebug("write worker:%d is launched", pWorker->workerId); - wWorkerPool.nextId = (wWorkerPool.nextId + 1) % wWorkerPool.max; + dDebug("dnode vwrite worker:%d is launched", pWorker->workerId); + tsVWriteWP.nextId = (tsVWriteWP.nextId + 1) % tsVWriteWP.max; } pthread_attr_destroy(&thAttr); } else { taosAddIntoQset(pWorker->qset, queue, pVnode); - wWorkerPool.nextId = (wWorkerPool.nextId + 1) % wWorkerPool.max; + tsVWriteWP.nextId = (tsVWriteWP.nextId + 1) % tsVWriteWP.max; } - pthread_mutex_unlock(&wWorkerPool.mutex); - dDebug("pVnode:%p, write queue:%p is allocated", pVnode, queue); + pthread_mutex_unlock(&tsVWriteWP.mutex); + dDebug("pVnode:%p, dnode vwrite queue:%p is allocated", pVnode, queue); return queue; } -void dnodeFreeVnodeWqueue(void *wqueue) { +void dnodeFreeVWriteQueue(void *wqueue) { taosCloseQueue(wqueue); - - // dynamically adjust the number of threads } -void dnodeSendRpcVnodeWriteRsp(void *pVnode, void *param, int32_t code) { - SWriteMsg *pWrite = (SWriteMsg *)param; - if (pWrite == NULL) return; +void dnodeSendRpcVWriteRsp(void *pVnode, void *param, int32_t code) { + if (param == NULL) return; + SWriteMsg *pWrite = param; if (code < 0) pWrite->code = code; int32_t count = atomic_add_fetch_32(&pWrite->processedCount, 1); @@ -215,44 +204,45 @@ static void *dnodeProcessWriteQueue(void *param) { SWriteWorker *pWorker = (SWriteWorker *)param; SWriteMsg * pWrite; SWalHead * pHead; - int32_t numOfMsgs; - int type; - void * pVnode, *item; SRspRet * pRspRet; + void * pVnode; + void * pItem; + int32_t numOfMsgs; + int32_t qtype; - dDebug("write worker:%d is running", pWorker->workerId); + dDebug("dnode vwrite worker:%d is running", pWorker->workerId); while (1) { numOfMsgs = taosReadAllQitemsFromQset(pWorker->qset, pWorker->qall, &pVnode); if (numOfMsgs == 0) { - dDebug("qset:%p, dnode write got no message from qset, exiting", pWorker->qset); + dDebug("qset:%p, dnode vwrite got no message from qset, exiting", pWorker->qset); break; } for (int32_t i = 0; i < numOfMsgs; ++i) { pWrite = NULL; pRspRet = NULL; - taosGetQitem(pWorker->qall, &type, &item); - if (type == TAOS_QTYPE_RPC) { - pWrite = (SWriteMsg *)item; + taosGetQitem(pWorker->qall, &qtype, &pItem); + if (qtype == TAOS_QTYPE_RPC) { + pWrite = pItem; pRspRet = &pWrite->rspRet; - pHead = (SWalHead *)(pWrite->pCont - sizeof(SWalHead)); + pHead = (SWalHead *)((char *)pWrite->pCont - sizeof(SWalHead)); pHead->msgType = pWrite->rpcMsg.msgType; pHead->version = 0; pHead->len = pWrite->contLen; dDebug("%p, rpc msg:%s will be processed in vwrite queue", pWrite->rpcMsg.ahandle, taosMsg[pWrite->rpcMsg.msgType]); - } else if (type == TAOS_QTYPE_CQ) { - pHead = (SWalHead *)((char*)item + sizeof(SSyncHead)); + } else if (qtype == TAOS_QTYPE_CQ) { + pHead = (SWalHead *)((char *)pItem + sizeof(SSyncHead)); dTrace("%p, CQ wal msg:%s will be processed in vwrite queue, version:%" PRIu64, pHead, taosMsg[pHead->msgType], pHead->version); } else { - pHead = (SWalHead *)item; + pHead = pItem; dTrace("%p, wal msg:%s will be processed in vwrite queue, version:%" PRIu64, pHead, taosMsg[pHead->msgType], pHead->version); } - int32_t code = vnodeProcessWrite(pVnode, type, pHead, pRspRet); + int32_t code = vnodeProcessWrite(pVnode, qtype, pHead, pRspRet); dTrace("%p, msg:%s is processed in vwrite queue, version:%" PRIu64 ", result:%s", pHead, taosMsg[pHead->msgType], pHead->version, tstrerror(code)); @@ -267,17 +257,17 @@ static void *dnodeProcessWriteQueue(void *param) { // browse all items, and process them one by one taosResetQitems(pWorker->qall); for (int32_t i = 0; i < numOfMsgs; ++i) { - taosGetQitem(pWorker->qall, &type, &item); - if (type == TAOS_QTYPE_RPC) { - pWrite = (SWriteMsg *)item; - dnodeSendRpcVnodeWriteRsp(pVnode, item, pWrite->rpcMsg.code); - } else if (type == TAOS_QTYPE_FWD) { - pHead = (SWalHead *)item; + taosGetQitem(pWorker->qall, &qtype, &pItem); + if (qtype == TAOS_QTYPE_RPC) { + pWrite = pItem; + dnodeSendRpcVWriteRsp(pVnode, pItem, pWrite->rpcMsg.code); + } else if (qtype == TAOS_QTYPE_FWD) { + pHead = pItem; vnodeConfirmForward(pVnode, pHead->version, 0); - taosFreeQitem(item); + taosFreeQitem(pItem); vnodeRelease(pVnode); } else { - taosFreeQitem(item); + taosFreeQitem(pItem); vnodeRelease(pVnode); } } @@ -285,19 +275,3 @@ static void *dnodeProcessWriteQueue(void *param) { return NULL; } - -UNUSED_FUNC -static void dnodeHandleIdleWorker(SWriteWorker *pWorker) { - int32_t num = taosGetQueueNumber(pWorker->qset); - - if (num > 0) { - usleep(30000); - sched_yield(); - } else { - taosFreeQall(pWorker->qall); - taosCloseQset(pWorker->qset); - pWorker->qset = NULL; - dDebug("write worker:%d is released", pWorker->workerId); - pthread_exit(NULL); - } -} diff --git a/src/inc/dnode.h b/src/inc/dnode.h index e84545be1753f3fedb4ee78acf397c91d824ad8b..b4973cc672546dc99531e3f6349d9761067f94f8 100644 --- a/src/inc/dnode.h +++ b/src/inc/dnode.h @@ -21,6 +21,7 @@ extern "C" { #endif #include "trpc.h" +#include "taosmsg.h" typedef struct { int32_t queryReqNum; @@ -38,12 +39,13 @@ SDnodeRunStatus dnodeGetRunStatus(); SDnodeStatisInfo dnodeGetStatisInfo(); bool dnodeIsFirstDeploy(); -char * dnodeGetMnodeMasterEp(); -void dnodeGetMnodeEpSetForPeer(void *epSet); -void dnodeGetMnodeEpSetForShell(void *epSet); -void * dnodeGetMnodeInfos(); +bool dnodeIsMasterEp(char *ep); +void dnodeGetEpSetForPeer(SRpcEpSet *epSet); +void dnodeGetEpSetForShell(SRpcEpSet *epSet); int32_t dnodeGetDnodeId(); -bool dnodeStartMnode(void *pModes); +void dnodeUpdateEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port); +bool dnodeCheckEpChanged(int32_t dnodeId, char *epstr); +bool dnodeStartMnode(SMnodeInfos *minfos); void dnodeAddClientRspHandle(uint8_t msgType, void (*fp)(SRpcMsg *rpcMsg)); void dnodeSendMsgToDnode(SRpcEpSet *epSet, SRpcMsg *rpcMsg); @@ -51,11 +53,11 @@ void dnodeSendMsgToMnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp); void dnodeSendMsgToDnodeRecv(SRpcMsg *rpcMsg, SRpcMsg *rpcRsp, SRpcEpSet *epSet); void *dnodeSendCfgTableToRecv(int32_t vgId, int32_t tid); -void *dnodeAllocateVnodeWqueue(void *pVnode); -void dnodeFreeVnodeWqueue(void *queue); -void *dnodeAllocateVnodeRqueue(void *pVnode); -void dnodeFreeVnodeRqueue(void *rqueue); -void dnodeSendRpcVnodeWriteRsp(void *pVnode, void *param, int32_t code); +void *dnodeAllocVWriteQueue(void *pVnode); +void dnodeFreeVWriteQueue(void *wqueue); +void *dnodeAllocVReadQueue(void *pVnode); +void dnodeFreeVReadQueue(void *rqueue); +void dnodeSendRpcVWriteRsp(void *pVnode, void *param, int32_t code); int32_t dnodeAllocateMnodePqueue(); void dnodeFreeMnodePqueue(); diff --git a/src/inc/taosmsg.h b/src/inc/taosmsg.h index ddc72eab140d663f71f0da2a1383df73e76c32c6..8139c9ab17b66e4f4bf6ac8e6eced5b0cdc30f63 100644 --- a/src/inc/taosmsg.h +++ b/src/inc/taosmsg.h @@ -295,7 +295,7 @@ typedef struct { SSchema schema[]; // tagVal is padded after schema // char tagVal[]; -} SCMAlterTableMsg; +} SAlterTableMsg; typedef struct { SMsgHead head; @@ -312,12 +312,12 @@ typedef struct { } SUpdateTableTagValMsg; typedef struct { - char clientVersion[TSDB_VERSION_LEN]; - char msgVersion[TSDB_VERSION_LEN]; - char db[TSDB_TABLE_FNAME_LEN]; - char appName[TSDB_APPNAME_LEN]; + char clientVersion[TSDB_VERSION_LEN]; + char msgVersion[TSDB_VERSION_LEN]; + char db[TSDB_TABLE_FNAME_LEN]; + char appName[TSDB_APPNAME_LEN]; int32_t pid; -} SCMConnectMsg; +} SConnectMsg; typedef struct { char acctId[TSDB_ACCT_LEN]; @@ -328,7 +328,7 @@ typedef struct { int8_t reserved2; int32_t connId; SRpcEpSet epSet; -} SCMConnectRsp; +} SConnectRsp; typedef struct { int32_t maxUsers; @@ -348,18 +348,18 @@ typedef struct { char user[TSDB_USER_LEN]; char pass[TSDB_KEY_LEN]; SAcctCfg cfg; -} SCMCreateAcctMsg, SCMAlterAcctMsg; +} SCreateAcctMsg, SAlterAcctMsg; typedef struct { char user[TSDB_USER_LEN]; -} SCMDropUserMsg, SCMDropAcctMsg; +} SDropUserMsg, SDropAcctMsg; typedef struct { char user[TSDB_USER_LEN]; char pass[TSDB_KEY_LEN]; int8_t privilege; int8_t flag; -} SCMCreateUserMsg, SCMAlterUserMsg; +} SCreateUserMsg, SAlterUserMsg; typedef struct { int32_t contLen; @@ -374,11 +374,11 @@ typedef struct { int32_t vgId; uint64_t uid; char tableId[TSDB_TABLE_FNAME_LEN]; -} SMDDropSTableMsg; +} SDropSTableMsg; typedef struct { int32_t vgId; -} SMDDropVnodeMsg; +} SDropVnodeMsg; typedef struct SColIndex { int16_t colId; // column id @@ -544,12 +544,12 @@ typedef struct { int8_t replications; int8_t quorum; int8_t ignoreExist; -} SCMCreateDbMsg, SCMAlterDbMsg; +} SCreateDbMsg, SAlterDbMsg; typedef struct { char db[TSDB_TABLE_FNAME_LEN]; uint8_t ignoreNotExists; -} SCMDropDbMsg, SCMUseDbMsg; +} SDropDbMsg, SUseDbMsg; // IMPORTANT: sizeof(SVnodeStatisticInfo) should not exceed // TSDB_FILE_HEADER_LEN/4 - TSDB_FILE_HEADER_VERSION_SIZE @@ -564,7 +564,7 @@ typedef struct { typedef struct { int32_t vgId; int8_t accessState; -} SDMVgroupAccess; +} SVgroupAccess; typedef struct { int32_t dnodeId; @@ -572,18 +572,29 @@ typedef struct { uint32_t numOfVnodes; char clusterId[TSDB_CLUSTER_ID_LEN]; char reserved[16]; -} SDMDnodeCfg; +} SDnodeCfg; + +typedef struct { + int32_t dnodeId; + uint16_t dnodePort; + char dnodeFqdn[TSDB_FQDN_LEN]; +} SDnodeEp; typedef struct { - int32_t nodeId; - char nodeEp[TSDB_EP_LEN]; -} SDMMnodeInfo; + int32_t dnodeNum; + SDnodeEp dnodeEps[]; +} SDnodeEps; typedef struct { - int8_t inUse; - int8_t nodeNum; - SDMMnodeInfo nodeInfos[TSDB_MAX_REPLICA]; -} SDMMnodeInfos; + int32_t mnodeId; + char mnodeEp[TSDB_EP_LEN]; +} SMnodeInfo; + +typedef struct { + int8_t inUse; + int8_t mnodeNum; + SMnodeInfo mnodeInfos[TSDB_MAX_REPLICA]; +} SMnodeInfos; typedef struct { int32_t numOfMnodes; // tsNumOfMnodes @@ -615,13 +626,13 @@ typedef struct { uint8_t reserve2[15]; SClusterCfg clusterCfg; SVnodeLoad load[]; -} SDMStatusMsg; +} SStatusMsg; typedef struct { - SDMMnodeInfos mnodes; - SDMDnodeCfg dnodeCfg; - SDMVgroupAccess vgAccess[]; -} SDMStatusRsp; + SMnodeInfos mnodes; + SDnodeCfg dnodeCfg; + SVgroupAccess vgAccess[]; +} SStatusRsp; typedef struct { uint32_t vgId; @@ -644,54 +655,54 @@ typedef struct { int8_t wals; int8_t quorum; int8_t reserved[16]; -} SMDVnodeCfg; +} SVnodeCfg; typedef struct { int32_t nodeId; char nodeEp[TSDB_EP_LEN]; -} SMDVnodeDesc; +} SVnodeDesc; typedef struct { - char db[TSDB_ACCT_LEN + TSDB_DB_NAME_LEN]; - SMDVnodeCfg cfg; - SMDVnodeDesc nodes[TSDB_MAX_REPLICA]; -} SMDCreateVnodeMsg, SMDAlterVnodeMsg; + char db[TSDB_ACCT_LEN + TSDB_DB_NAME_LEN]; + SVnodeCfg cfg; + SVnodeDesc nodes[TSDB_MAX_REPLICA]; +} SCreateVnodeMsg, SAlterVnodeMsg; typedef struct { char tableId[TSDB_TABLE_FNAME_LEN]; int16_t createFlag; char tags[]; -} SCMTableInfoMsg; +} STableInfoMsg; typedef struct { int32_t numOfTables; char tableIds[]; -} SCMMultiTableInfoMsg; +} SMultiTableInfoMsg; -typedef struct SCMSTableVgroupMsg { +typedef struct SSTableVgroupMsg { int32_t numOfTables; -} SCMSTableVgroupMsg, SCMSTableVgroupRspMsg; +} SSTableVgroupMsg, SSTableVgroupRspMsg; typedef struct { int32_t vgId; int8_t numOfEps; SEpAddr1 epAddr[TSDB_MAX_REPLICA]; -} SCMVgroupInfo; +} SVgroupInfo; typedef struct { int32_t vgId; int8_t numOfEps; SEpAddrMsg epAddr[TSDB_MAX_REPLICA]; -} SCMVgroupMsg; +} SVgroupMsg; typedef struct { int32_t numOfVgroups; - SCMVgroupInfo vgroups[]; + SVgroupInfo vgroups[]; } SVgroupsInfo; typedef struct { int32_t numOfVgroups; - SCMVgroupMsg vgroups[]; + SVgroupMsg vgroups[]; } SVgroupsMsg; typedef struct STableMetaMsg { @@ -706,7 +717,7 @@ typedef struct STableMetaMsg { int16_t tversion; int32_t tid; uint64_t uid; - SCMVgroupMsg vgroup; + SVgroupMsg vgroup; SSchema schema[]; } STableMetaMsg; @@ -732,38 +743,38 @@ typedef struct { char db[TSDB_ACCT_LEN + TSDB_DB_NAME_LEN]; uint16_t payloadLen; char payload[]; -} SCMShowMsg; +} SShowMsg; -typedef struct SCMShowRsp { +typedef struct SShowRsp { uint64_t qhandle; STableMetaMsg tableMeta; -} SCMShowRsp; +} SShowRsp; typedef struct { - char ep[TSDB_EP_LEN]; // end point, hostname:port -} SCMCreateDnodeMsg, SCMDropDnodeMsg; + char ep[TSDB_EP_LEN]; // end point, hostname:port +} SCreateDnodeMsg, SDropDnodeMsg; typedef struct { int32_t dnodeId; char dnodeEp[TSDB_EP_LEN]; // end point, hostname:port - SDMMnodeInfos mnodes; -} SMDCreateMnodeMsg; + SMnodeInfos mnodes; +} SCreateMnodeMsg; typedef struct { int32_t dnodeId; int32_t vgId; int32_t tid; -} SDMConfigTableMsg; +} SConfigTableMsg; typedef struct { uint32_t dnodeId; int32_t vgId; -} SDMConfigVnodeMsg; +} SConfigVnodeMsg; typedef struct { char ep[TSDB_EP_LEN]; // end point, hostname:port char config[64]; -} SMDCfgDnodeMsg, SCMCfgDnodeMsg; +} SCfgDnodeMsg; typedef struct { char sql[TSDB_SHOW_SQL_LEN]; @@ -792,7 +803,7 @@ typedef struct { int32_t numOfStreams; char appName[TSDB_APPNAME_LEN]; char pData[]; -} SCMHeartBeatMsg; +} SHeartBeatMsg; typedef struct { uint32_t queryId; @@ -802,11 +813,11 @@ typedef struct { uint32_t connId; int8_t killConnection; SRpcEpSet epSet; -} SCMHeartBeatRsp; +} SHeartBeatRsp; typedef struct { char queryId[TSDB_KILL_MSG_LEN + 1]; -} SCMKillQueryMsg, SCMKillStreamMsg, SCMKillConnMsg; +} SKillQueryMsg, SKillStreamMsg, SKillConnMsg; typedef struct { int32_t vnode; @@ -815,7 +826,7 @@ typedef struct { uint64_t stime; // stream starting time int32_t status; char tableId[TSDB_TABLE_FNAME_LEN]; -} SMDAlterStreamMsg; +} SAlterStreamMsg; typedef struct { char user[TSDB_USER_LEN]; @@ -823,7 +834,7 @@ typedef struct { char encrypt; char secret[TSDB_KEY_LEN]; char ckey[TSDB_KEY_LEN]; -} SDMAuthMsg, SDMAuthRsp; +} SAuthMsg, SAuthRsp; #pragma pack(pop) diff --git a/src/inc/tsync.h b/src/inc/tsync.h index ca0f70d104d603d176d89dd5b92979433f390466..671adefab83fed42f7bc3fb0ad9591050f118e73 100644 --- a/src/inc/tsync.h +++ b/src/inc/tsync.h @@ -68,7 +68,7 @@ typedef uint32_t (*FGetFileInfo)(void *ahandle, char *name, uint32_t *index, uin // get the wal file from index or after // return value, -1: error, 1:more wal files, 0:last WAL. if name[0]==0, no WAL file -typedef int (*FGetWalInfo)(void *ahandle, char *name, uint32_t *index); +typedef int32_t (*FGetWalInfo)(void *ahandle, char *fileName, int64_t *fileId); // when a forward pkt is received, call this to handle data typedef int (*FWriteToCache)(void *ahandle, void *pHead, int type); diff --git a/src/inc/twal.h b/src/inc/twal.h index 1ce7b132b068526f87fa594845bb591ade6f4966..3a229ed8350624e1c69772abbe9b2f30ac232b80 100644 --- a/src/inc/twal.h +++ b/src/inc/twal.h @@ -19,9 +19,11 @@ extern "C" { #endif -#define TAOS_WAL_NOLOG 0 -#define TAOS_WAL_WRITE 1 -#define TAOS_WAL_FSYNC 2 +typedef enum { + TAOS_WAL_NOLOG = 0, + TAOS_WAL_WRITE = 1, + TAOS_WAL_FSYNC = 2 +} EWalType; typedef struct { int8_t msgType; @@ -34,8 +36,9 @@ typedef struct { } SWalHead; typedef struct { - int8_t walLevel; // wal level + int32_t vgId; int32_t fsyncPeriod; // millisecond + int8_t walLevel; // wal level int8_t wals; // number of WAL files; int8_t keep; // keep the wal file when closed } SWalCfg; @@ -43,14 +46,18 @@ typedef struct { typedef void* twalh; // WAL HANDLE typedef int (*FWalWrite)(void *ahandle, void *pHead, int type); -twalh walOpen(const char *path, const SWalCfg *pCfg); -int walAlter(twalh pWal, const SWalCfg *pCfg); +int32_t walInit(); +void walCleanUp(); + +twalh walOpen(char *path, SWalCfg *pCfg); +int32_t walAlter(twalh pWal, SWalCfg *pCfg); +void walStop(twalh); void walClose(twalh); -int walRenew(twalh); -int walWrite(twalh, SWalHead *); +int32_t walRenew(twalh); +int32_t walWrite(twalh, SWalHead *); void walFsync(twalh); -int walRestore(twalh, void *pVnode, FWalWrite writeFp); -int walGetWalFile(twalh, char *name, uint32_t *index); +int32_t walRestore(twalh, void *pVnode, FWalWrite writeFp); +int32_t walGetWalFile(twalh, char *fileName, int64_t *fileId); int64_t walGetVersion(twalh); #ifdef __cplusplus diff --git a/src/inc/vnode.h b/src/inc/vnode.h index fdce4d62794075bd2e7027b125780fbd7a2deaed..8c387065ccd8c8502df35ef2f576ad738049a421 100644 --- a/src/inc/vnode.h +++ b/src/inc/vnode.h @@ -43,10 +43,10 @@ typedef struct { extern char *vnodeStatus[]; -int32_t vnodeCreate(SMDCreateVnodeMsg *pVnodeCfg); +int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg); int32_t vnodeDrop(int32_t vgId); int32_t vnodeOpen(int32_t vgId, char *rootDir); -int32_t vnodeAlter(void *pVnode, SMDCreateVnodeMsg *pVnodeCfg); +int32_t vnodeAlter(void *pVnode, SCreateVnodeMsg *pVnodeCfg); int32_t vnodeClose(int32_t vgId); void* vnodeAcquire(int32_t vgId); // add refcount @@ -60,7 +60,7 @@ int32_t vnodeCheckWrite(void *pVnode); int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes); void vnodeBuildStatusMsg(void *param); void vnodeConfirmForward(void *param, uint64_t version, int32_t code); -void vnodeSetAccess(SDMVgroupAccess *pAccess, int32_t numOfVnodes); +void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes); int32_t vnodeInitResources(); void vnodeCleanupResources(); diff --git a/src/kit/shell/src/shellEngine.c b/src/kit/shell/src/shellEngine.c index 748b7e792982352de611227d82a844448d251ee2..9b166f935190a0f09cc4cb92032b9e83a59e5d17 100644 --- a/src/kit/shell/src/shellEngine.c +++ b/src/kit/shell/src/shellEngine.c @@ -244,7 +244,7 @@ int32_t shellRunCommand(TAOS* con, char* command) { } *p++ = c; - if (c == ';') { + if (c == ';' && quote == 0) { c = *p; *p = 0; if (shellRunSingleCommand(con, cmd) < 0) { diff --git a/src/mnode/inc/mnodeProfile.h b/src/mnode/inc/mnodeProfile.h index 1e5b1c0f9c7c1c220a88339f00d46538c78cf938..ee57c5da1f66b620e8624a74ba774643c589e818 100644 --- a/src/mnode/inc/mnodeProfile.h +++ b/src/mnode/inc/mnodeProfile.h @@ -45,7 +45,7 @@ void mnodeCleanupProfile(); SConnObj *mnodeCreateConn(char *user, uint32_t ip, uint16_t port, int32_t pid, const char* app); SConnObj *mnodeAccquireConn(int32_t connId, char *user, uint32_t ip, uint16_t port); void mnodeReleaseConn(SConnObj *pConn); -int32_t mnodeSaveQueryStreamList(SConnObj *pConn, SCMHeartBeatMsg *pHBMsg); +int32_t mnodeSaveQueryStreamList(SConnObj *pConn, SHeartBeatMsg *pHBMsg); #ifdef __cplusplus } diff --git a/src/mnode/src/mnodeDb.c b/src/mnode/src/mnodeDb.c index 8d7c267ab7cb328bb187518a88be3d344b0d53e2..748b856ba42c2d1e80154b2df4007be5953bb95d 100644 --- a/src/mnode/src/mnodeDb.c +++ b/src/mnode/src/mnodeDb.c @@ -41,7 +41,7 @@ static void * tsDbSdb = NULL; static int32_t tsDbUpdateSize; -static int32_t mnodeCreateDb(SAcctObj *pAcct, SCMCreateDbMsg *pCreate, SMnodeMsg *pMsg); +static int32_t mnodeCreateDb(SAcctObj *pAcct, SCreateDbMsg *pCreate, SMnodeMsg *pMsg); static int32_t mnodeDropDb(SMnodeMsg *newMsg); static int32_t mnodeSetDbDropping(SDbObj *pDb); static int32_t mnodeGetDbMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); @@ -352,7 +352,7 @@ static int32_t mnodeCreateDbCb(SMnodeMsg *pMsg, int32_t code) { return code; } -static int32_t mnodeCreateDb(SAcctObj *pAcct, SCMCreateDbMsg *pCreate, SMnodeMsg *pMsg) { +static int32_t mnodeCreateDb(SAcctObj *pAcct, SCreateDbMsg *pCreate, SMnodeMsg *pMsg) { int32_t code = acctCheck(pAcct, ACCT_GRANT_DB); if (code != 0) return code; @@ -805,7 +805,7 @@ static int32_t mnodeSetDbDropping(SDbObj *pDb) { } static int32_t mnodeProcessCreateDbMsg(SMnodeMsg *pMsg) { - SCMCreateDbMsg *pCreate = pMsg->rpcMsg.pCont; + SCreateDbMsg *pCreate = pMsg->rpcMsg.pCont; pCreate->maxTables = htonl(pCreate->maxTables); pCreate->cacheBlockSize = htonl(pCreate->cacheBlockSize); pCreate->totalBlocks = htonl(pCreate->totalBlocks); @@ -830,7 +830,7 @@ static int32_t mnodeProcessCreateDbMsg(SMnodeMsg *pMsg) { return code; } -static SDbCfg mnodeGetAlterDbOption(SDbObj *pDb, SCMAlterDbMsg *pAlter) { +static SDbCfg mnodeGetAlterDbOption(SDbObj *pDb, SAlterDbMsg *pAlter) { SDbCfg newCfg = pDb->cfg; int32_t maxTables = htonl(pAlter->maxTables); int32_t cacheBlockSize = htonl(pAlter->cacheBlockSize); @@ -977,7 +977,7 @@ static int32_t mnodeAlterDbCb(SMnodeMsg *pMsg, int32_t code) { return TSDB_CODE_SUCCESS; } -static int32_t mnodeAlterDb(SDbObj *pDb, SCMAlterDbMsg *pAlter, void *pMsg) { +static int32_t mnodeAlterDb(SDbObj *pDb, SAlterDbMsg *pAlter, void *pMsg) { SDbCfg newCfg = mnodeGetAlterDbOption(pDb, pAlter); if (terrno != TSDB_CODE_SUCCESS) { return terrno; @@ -1009,7 +1009,7 @@ static int32_t mnodeAlterDb(SDbObj *pDb, SCMAlterDbMsg *pAlter, void *pMsg) { } static int32_t mnodeProcessAlterDbMsg(SMnodeMsg *pMsg) { - SCMAlterDbMsg *pAlter = pMsg->rpcMsg.pCont; + SAlterDbMsg *pAlter = pMsg->rpcMsg.pCont; mDebug("db:%s, alter db msg is received from thandle:%p", pAlter->db, pMsg->rpcMsg.handle); if (pMsg->pDb == NULL) pMsg->pDb = mnodeGetDb(pAlter->db); @@ -1060,7 +1060,7 @@ static int32_t mnodeDropDb(SMnodeMsg *pMsg) { } static int32_t mnodeProcessDropDbMsg(SMnodeMsg *pMsg) { - SCMDropDbMsg *pDrop = pMsg->rpcMsg.pCont; + SDropDbMsg *pDrop = pMsg->rpcMsg.pCont; mDebug("db:%s, drop db msg is received from thandle:%p", pDrop->db, pMsg->rpcMsg.handle); if (pMsg->pDb == NULL) pMsg->pDb = mnodeGetDb(pDrop->db); diff --git a/src/mnode/src/mnodeDnode.c b/src/mnode/src/mnodeDnode.c index 1cd861e22303d5f6f0dcb75c7ef83cf66696578f..bf8647c1930bd345a3aabbffcdbf1ab53318a7f4 100644 --- a/src/mnode/src/mnodeDnode.c +++ b/src/mnode/src/mnodeDnode.c @@ -39,11 +39,15 @@ #include "mnodeCluster.h" int32_t tsAccessSquence = 0; -static void *tsDnodeSdb = NULL; +static void * tsDnodeSdb = NULL; static int32_t tsDnodeUpdateSize = 0; extern void * tsMnodeSdb; extern void * tsVgroupSdb; +static SDnodeEps*tsDnodeEps; +static int32_t tsDnodeEpsSize; +static pthread_mutex_t tsDnodeEpsMutex; + static int32_t mnodeCreateDnode(char *ep, SMnodeMsg *pMsg); static int32_t mnodeProcessCreateDnodeMsg(SMnodeMsg *pMsg); static int32_t mnodeProcessDropDnodeMsg(SMnodeMsg *pMsg); @@ -59,6 +63,7 @@ static int32_t mnodeRetrieveVnodes(SShowObj *pShow, char *data, int32_t rows, vo static int32_t mnodeGetDnodeMeta(STableMetaMsg *pMeta, SShowObj *pShow, void *pConn); static int32_t mnodeRetrieveDnodes(SShowObj *pShow, char *data, int32_t rows, void *pConn); static char* mnodeGetDnodeAlternativeRoleStr(int32_t alternativeRole); +static void mnodeUpdateDnodeEps(); static char* offlineReason[] = { "", @@ -95,6 +100,9 @@ static int32_t mnodeDnodeActionInsert(SSdbOper *pOper) { pDnode->offlineReason = TAOS_DN_OFF_STATUS_NOT_RECEIVED; } + dnodeUpdateEp(pDnode->dnodeId, pDnode->dnodeEp, pDnode->dnodeFqdn, &pDnode->dnodePort); + mnodeUpdateDnodeEps(); + mInfo("dnode:%d, fqdn:%s ep:%s port:%d, do insert action", pDnode->dnodeId, pDnode->dnodeFqdn, pDnode->dnodeEp, pDnode->dnodePort); return TSDB_CODE_SUCCESS; } @@ -107,6 +115,7 @@ static int32_t mnodeDnodeActionDelete(SSdbOper *pOper) { #endif mnodeDropMnodeLocal(pDnode->dnodeId); balanceAsyncNotify(); + mnodeUpdateDnodeEps(); mDebug("dnode:%d, all vgroups is dropped from sdb", pDnode->dnodeId); return TSDB_CODE_SUCCESS; @@ -121,6 +130,7 @@ static int32_t mnodeDnodeActionUpdate(SSdbOper *pOper) { } mnodeDecDnodeRef(pDnode); + mnodeUpdateDnodeEps(); return TSDB_CODE_SUCCESS; } @@ -152,12 +162,14 @@ static int32_t mnodeDnodeActionRestored() { } } + mnodeUpdateDnodeEps(); return TSDB_CODE_SUCCESS; } int32_t mnodeInitDnodes() { SDnodeObj tObj; tsDnodeUpdateSize = (int8_t *)tObj.updateEnd - (int8_t *)&tObj; + pthread_mutex_init(&tsDnodeEpsMutex, NULL); SSdbTableDesc tableDesc = { .tableId = SDB_TABLE_DNODE, @@ -201,6 +213,9 @@ int32_t mnodeInitDnodes() { void mnodeCleanupDnodes() { sdbCloseTable(tsDnodeSdb); + pthread_mutex_destroy(&tsDnodeEpsMutex); + free(tsDnodeEps); + tsDnodeEps = NULL; tsDnodeSdb = NULL; } @@ -299,7 +314,7 @@ static int32_t mnodeProcessCfgDnodeMsg(SMnodeMsg *pMsg) { return TSDB_CODE_MND_NO_RIGHTS; } - SCMCfgDnodeMsg *pCmCfgDnode = pMsg->rpcMsg.pCont; + SCfgDnodeMsg *pCmCfgDnode = pMsg->rpcMsg.pCont; if (pCmCfgDnode->ep[0] == 0) { tstrncpy(pCmCfgDnode->ep, tsLocalEp, TSDB_EP_LEN); } @@ -334,7 +349,7 @@ static int32_t mnodeProcessCfgDnodeMsg(SMnodeMsg *pMsg) { mnodeDecDnodeRef(pDnode); return code; } else { - SMDCfgDnodeMsg *pMdCfgDnode = rpcMallocCont(sizeof(SMDCfgDnodeMsg)); + SCfgDnodeMsg *pMdCfgDnode = rpcMallocCont(sizeof(SCfgDnodeMsg)); strcpy(pMdCfgDnode->ep, pCmCfgDnode->ep); strcpy(pMdCfgDnode->config, pCmCfgDnode->config); @@ -343,7 +358,7 @@ static int32_t mnodeProcessCfgDnodeMsg(SMnodeMsg *pMsg) { .code = 0, .msgType = TSDB_MSG_TYPE_MD_CONFIG_DNODE, .pCont = pMdCfgDnode, - .contLen = sizeof(SMDCfgDnodeMsg) + .contLen = sizeof(SCfgDnodeMsg) }; mInfo("dnode:%s, is configured by %s", pCmCfgDnode->ep, pMsg->pUser->user); @@ -418,9 +433,54 @@ static int32_t mnodeCheckClusterCfgPara(const SClusterCfg *clusterCfg) { return 0; } +static int32_t mnodeGetDnodeEpsSize() { + pthread_mutex_lock(&tsDnodeEpsMutex); + int32_t size = tsDnodeEpsSize; + pthread_mutex_unlock(&tsDnodeEpsMutex); + return size; +} + +static void mnodeGetDnodeEpsData(SDnodeEps *pEps, int32_t epsSize) { + pthread_mutex_lock(&tsDnodeEpsMutex); + if (epsSize == tsDnodeEpsSize) { + memcpy(pEps, tsDnodeEps, tsDnodeEpsSize); + } + pthread_mutex_unlock(&tsDnodeEpsMutex); +} + +static void mnodeUpdateDnodeEps() { + pthread_mutex_lock(&tsDnodeEpsMutex); + + int32_t totalDnodes = mnodeGetDnodesNum(); + tsDnodeEpsSize = sizeof(SDnodeEps) + totalDnodes * sizeof(SDnodeEp); + free(tsDnodeEps); + tsDnodeEps = calloc(1, tsDnodeEpsSize); + tsDnodeEps->dnodeNum = htonl(totalDnodes); + + SDnodeObj *pDnode = NULL; + void * pIter = NULL; + int32_t dnodesNum = 0; + + while (1) { + pIter = mnodeGetNextDnode(pIter, &pDnode); + if (pDnode == NULL) break; + if (dnodesNum >= totalDnodes) break; + + SDnodeEp *pEp = &tsDnodeEps->dnodeEps[dnodesNum]; + dnodesNum++; + pEp->dnodeId = htonl(pDnode->dnodeId); + pEp->dnodePort = htons(pDnode->dnodePort); + tstrncpy(pEp->dnodeFqdn, pDnode->dnodeFqdn, TSDB_FQDN_LEN); + mnodeDecDnodeRef(pDnode); + } + + sdbFreeIter(pIter); + pthread_mutex_unlock(&tsDnodeEpsMutex); +} + static int32_t mnodeProcessDnodeStatusMsg(SMnodeMsg *pMsg) { SDnodeObj *pDnode = NULL; - SDMStatusMsg *pStatus = pMsg->rpcMsg.pCont; + SStatusMsg *pStatus = pMsg->rpcMsg.pCont; pStatus->dnodeId = htonl(pStatus->dnodeId); pStatus->moduleStatus = htonl(pStatus->moduleStatus); pStatus->lastReboot = htonl(pStatus->lastReboot); @@ -477,8 +537,11 @@ static int32_t mnodeProcessDnodeStatusMsg(SMnodeMsg *pMsg) { } int32_t openVnodes = htons(pStatus->openVnodes); - int32_t contLen = sizeof(SDMStatusRsp) + openVnodes * sizeof(SDMVgroupAccess); - SDMStatusRsp *pRsp = rpcMallocCont(contLen); + int32_t epsSize = mnodeGetDnodeEpsSize(); + int32_t vgAccessSize = openVnodes * sizeof(SVgroupAccess); + int32_t contLen = sizeof(SStatusRsp) + vgAccessSize + epsSize; + + SStatusRsp *pRsp = rpcMallocCont(contLen); if (pRsp == NULL) { mnodeDecDnodeRef(pDnode); return TSDB_CODE_MND_OUT_OF_MEMORY; @@ -488,8 +551,8 @@ static int32_t mnodeProcessDnodeStatusMsg(SMnodeMsg *pMsg) { pRsp->dnodeCfg.moduleStatus = htonl((int32_t)pDnode->isMgmt); pRsp->dnodeCfg.numOfVnodes = htonl(openVnodes); tstrncpy(pRsp->dnodeCfg.clusterId, mnodeGetClusterId(), TSDB_CLUSTER_ID_LEN); - SDMVgroupAccess *pAccess = (SDMVgroupAccess *)((char *)pRsp + sizeof(SDMStatusRsp)); - + SVgroupAccess *pAccess = (SVgroupAccess *)((char *)pRsp + sizeof(SStatusRsp)); + for (int32_t j = 0; j < openVnodes; ++j) { SVnodeLoad *pVload = &pStatus->load[j]; pVload->vgId = htonl(pVload->vgId); @@ -539,6 +602,9 @@ static int32_t mnodeProcessDnodeStatusMsg(SMnodeMsg *pMsg) { mnodeDecDnodeRef(pDnode); + SDnodeEps *pEps = (SDnodeEps *)((char *)pRsp + sizeof(SStatusRsp) + vgAccessSize); + mnodeGetDnodeEpsData(pEps, epsSize); + pMsg->rpcRsp.len = contLen; pMsg->rpcRsp.rsp = pRsp; @@ -645,7 +711,7 @@ static int32_t mnodeDropDnodeByEp(char *ep, SMnodeMsg *pMsg) { } static int32_t mnodeProcessCreateDnodeMsg(SMnodeMsg *pMsg) { - SCMCreateDnodeMsg *pCreate = pMsg->rpcMsg.pCont; + SCreateDnodeMsg *pCreate = pMsg->rpcMsg.pCont; if (strcmp(pMsg->pUser->user, TSDB_DEFAULT_USER) != 0) { return TSDB_CODE_MND_NO_RIGHTS; @@ -655,7 +721,7 @@ static int32_t mnodeProcessCreateDnodeMsg(SMnodeMsg *pMsg) { } static int32_t mnodeProcessDropDnodeMsg(SMnodeMsg *pMsg) { - SCMDropDnodeMsg *pDrop = pMsg->rpcMsg.pCont; + SDropDnodeMsg *pDrop = pMsg->rpcMsg.pCont; if (strcmp(pMsg->pUser->user, TSDB_DEFAULT_USER) != 0) { return TSDB_CODE_MND_NO_RIGHTS; diff --git a/src/mnode/src/mnodeMnode.c b/src/mnode/src/mnodeMnode.c index 89b2f50b731c90c1261957da93623ba16cebf2ab..5a1825a2a33a23c4a0c9e647a561613a2162dc84 100644 --- a/src/mnode/src/mnodeMnode.c +++ b/src/mnode/src/mnodeMnode.c @@ -38,7 +38,7 @@ static void * tsMnodeSdb = NULL; static int32_t tsMnodeUpdateSize = 0; static SRpcEpSet tsMnodeEpSetForShell; static SRpcEpSet tsMnodeEpSetForPeer; -static SDMMnodeInfos tsMnodeInfos; +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); @@ -70,8 +70,9 @@ static int32_t mnodeMnodeActionInsert(SSdbOper *pOper) { pDnode->isMgmt = true; mnodeDecDnodeRef(pDnode); - - mInfo("mnode:%d, fqdn:%s ep:%s port:%d, do insert action", pMnode->mnodeId, pDnode->dnodeFqdn, pDnode->dnodeEp, pDnode->dnodePort); + + mInfo("mnode:%d, fqdn:%s ep:%s port:%u, do insert action", pMnode->mnodeId, pDnode->dnodeFqdn, pDnode->dnodeEp, + pDnode->dnodePort); return TSDB_CODE_SUCCESS; } @@ -202,7 +203,7 @@ void mnodeUpdateMnodeEpSet() { memset(&tsMnodeEpSetForShell, 0, sizeof(SRpcEpSet)); memset(&tsMnodeEpSetForPeer, 0, sizeof(SRpcEpSet)); - memset(&tsMnodeInfos, 0, sizeof(SDMMnodeInfos)); + memset(&tsMnodeInfos, 0, sizeof(SMnodeInfos)); int32_t index = 0; void * pIter = NULL; @@ -221,8 +222,8 @@ void mnodeUpdateMnodeEpSet() { 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])); - tsMnodeInfos.nodeInfos[index].nodeId = htonl(pMnode->mnodeId); - strcpy(tsMnodeInfos.nodeInfos[index].nodeEp, pDnode->dnodeEp); + tsMnodeInfos.mnodeInfos[index].mnodeId = htonl(pMnode->mnodeId); + strcpy(tsMnodeInfos.mnodeInfos[index].mnodeEp, pDnode->dnodeEp); if (pMnode->role == TAOS_SYNC_ROLE_MASTER) { tsMnodeEpSetForShell.inUse = index; @@ -238,7 +239,7 @@ void mnodeUpdateMnodeEpSet() { mnodeDecMnodeRef(pMnode); } - tsMnodeInfos.nodeNum = index; + tsMnodeInfos.mnodeNum = index; tsMnodeEpSetForShell.numOfEps = index; tsMnodeEpSetForPeer.numOfEps = index; @@ -260,19 +261,19 @@ void mnodeGetMnodeEpSetForShell(SRpcEpSet *epSet) { } char* mnodeGetMnodeMasterEp() { - return tsMnodeInfos.nodeInfos[tsMnodeInfos.inUse].nodeEp; + return tsMnodeInfos.mnodeInfos[tsMnodeInfos.inUse].mnodeEp; } void mnodeGetMnodeInfos(void *mnodeInfos) { mnodeMnodeRdLock(); - *(SDMMnodeInfos *)mnodeInfos = tsMnodeInfos; + *(SMnodeInfos *)mnodeInfos = tsMnodeInfos; mnodeMnodeUnLock(); } static int32_t mnodeSendCreateMnodeMsg(int32_t dnodeId, char *dnodeEp) { mDebug("dnode:%d, send create mnode msg to dnode %s", dnodeId, dnodeEp); - SMDCreateMnodeMsg *pCreate = rpcMallocCont(sizeof(SMDCreateMnodeMsg)); + SCreateMnodeMsg *pCreate = rpcMallocCont(sizeof(SCreateMnodeMsg)); if (pCreate == NULL) { return TSDB_CODE_MND_OUT_OF_MEMORY; } else { @@ -280,21 +281,21 @@ static int32_t mnodeSendCreateMnodeMsg(int32_t dnodeId, char *dnodeEp) { tstrncpy(pCreate->dnodeEp, dnodeEp, sizeof(pCreate->dnodeEp)); pCreate->mnodes = tsMnodeInfos; bool found = false; - for (int i = 0; i < pCreate->mnodes.nodeNum; ++i) { - if (pCreate->mnodes.nodeInfos[i].nodeId == htonl(dnodeId)) { + for (int i = 0; i < pCreate->mnodes.mnodeNum; ++i) { + if (pCreate->mnodes.mnodeInfos[i].mnodeId == htonl(dnodeId)) { found = true; } } if (!found) { - pCreate->mnodes.nodeInfos[pCreate->mnodes.nodeNum].nodeId = htonl(dnodeId); - tstrncpy(pCreate->mnodes.nodeInfos[pCreate->mnodes.nodeNum].nodeEp, dnodeEp, sizeof(pCreate->dnodeEp)); - pCreate->mnodes.nodeNum++; + pCreate->mnodes.mnodeInfos[pCreate->mnodes.mnodeNum].mnodeId = htonl(dnodeId); + tstrncpy(pCreate->mnodes.mnodeInfos[pCreate->mnodes.mnodeNum].mnodeEp, dnodeEp, sizeof(pCreate->dnodeEp)); + pCreate->mnodes.mnodeNum++; } } SRpcMsg rpcMsg = {0}; rpcMsg.pCont = pCreate; - rpcMsg.contLen = sizeof(SMDCreateMnodeMsg); + rpcMsg.contLen = sizeof(SCreateMnodeMsg); rpcMsg.msgType = TSDB_MSG_TYPE_MD_CREATE_MNODE; SRpcMsg rpcRsp = {0}; diff --git a/src/mnode/src/mnodeProfile.c b/src/mnode/src/mnodeProfile.c index c29d1ec0b7c863ec86d78f28997d40ba86b4dd34..9831a95c5f07298db17c15cdc6ceed6ffd05bfd0 100644 --- a/src/mnode/src/mnodeProfile.c +++ b/src/mnode/src/mnodeProfile.c @@ -280,7 +280,7 @@ static int32_t mnodeRetrieveConns(SShowObj *pShow, char *data, int32_t rows, voi } // not thread safe, need optimized -int32_t mnodeSaveQueryStreamList(SConnObj *pConn, SCMHeartBeatMsg *pHBMsg) { +int32_t mnodeSaveQueryStreamList(SConnObj *pConn, SHeartBeatMsg *pHBMsg) { pConn->numOfQueries = htonl(pHBMsg->numOfQueries); if (pConn->numOfQueries > 0) { if (pConn->pQueries == NULL) { @@ -561,7 +561,7 @@ static int32_t mnodeProcessKillQueryMsg(SMnodeMsg *pMsg) { SUserObj *pUser = pMsg->pUser; if (strcmp(pUser->user, TSDB_DEFAULT_USER) != 0) return TSDB_CODE_MND_NO_RIGHTS; - SCMKillQueryMsg *pKill = pMsg->rpcMsg.pCont; + SKillQueryMsg *pKill = pMsg->rpcMsg.pCont; mInfo("kill query msg is received, queryId:%s", pKill->queryId); const char delim = ':'; @@ -592,7 +592,7 @@ static int32_t mnodeProcessKillStreamMsg(SMnodeMsg *pMsg) { SUserObj *pUser = pMsg->pUser; if (strcmp(pUser->user, TSDB_DEFAULT_USER) != 0) return TSDB_CODE_MND_NO_RIGHTS; - SCMKillQueryMsg *pKill = pMsg->rpcMsg.pCont; + SKillQueryMsg *pKill = pMsg->rpcMsg.pCont; mInfo("kill stream msg is received, streamId:%s", pKill->queryId); const char delim = ':'; @@ -623,7 +623,7 @@ static int32_t mnodeProcessKillConnectionMsg(SMnodeMsg *pMsg) { SUserObj *pUser = pMsg->pUser; if (strcmp(pUser->user, TSDB_DEFAULT_USER) != 0) return TSDB_CODE_MND_NO_RIGHTS; - SCMKillConnMsg *pKill = pMsg->rpcMsg.pCont; + SKillConnMsg *pKill = pMsg->rpcMsg.pCont; int32_t connId = atoi(pKill->queryId); SConnObj *pConn = taosCacheAcquireByKey(tsMnodeConnCache, &connId, sizeof(int32_t)); if (pConn == NULL) { diff --git a/src/mnode/src/mnodeSdb.c b/src/mnode/src/mnodeSdb.c index 8c61c61a10f6a5dd46c188acf32ef1242f3e7c50..8fb0b330605ec6cd63a60ad9e64a5e8dc62b67d4 100644 --- a/src/mnode/src/mnodeSdb.c +++ b/src/mnode/src/mnodeSdb.c @@ -175,7 +175,7 @@ static void *sdbGetTableFromId(int32_t tableId) { } static int32_t sdbInitWal() { - SWalCfg walCfg = {.walLevel = 2, .wals = 2, .keep = 1, .fsyncPeriod = 0}; + SWalCfg walCfg = {.vgId = 1, .walLevel = 2, .wals = 2, .keep = 1, .fsyncPeriod = 0}; char temp[TSDB_FILENAME_LEN]; sprintf(temp, "%s/wal", tsMnodeDir); tsSdbObj.wal = walOpen(temp, &walCfg); @@ -237,8 +237,8 @@ static uint32_t sdbGetFileInfo(void *ahandle, char *name, uint32_t *index, uint3 return 0; } -static int sdbGetWalInfo(void *ahandle, char *name, uint32_t *index) { - return walGetWalFile(tsSdbObj.wal, name, index); +static int32_t sdbGetWalInfo(void *ahandle, char *fileName, int64_t *fileId) { + return walGetWalFile(tsSdbObj.wal, fileName, fileId); } static void sdbNotifyRole(void *ahandle, int8_t role) { @@ -312,7 +312,7 @@ void sdbUpdateAsync() { } void sdbUpdateSync(void *pMnodes) { - SDMMnodeInfos *mnodes = pMnodes; + SMnodeInfos *mnodes = pMnodes; if (!mnodeIsRunning()) { mDebug("mnode not start yet, update sync config later"); return; @@ -346,10 +346,10 @@ void sdbUpdateSync(void *pMnodes) { syncCfg.replica = index; mDebug("mnodes info not input, use infos in sdb, numOfMnodes:%d", syncCfg.replica); } else { - for (index = 0; index < mnodes->nodeNum; ++index) { - SDMMnodeInfo *node = &mnodes->nodeInfos[index]; - syncCfg.nodeInfo[index].nodeId = node->nodeId; - taosGetFqdnPortFromEp(node->nodeEp, syncCfg.nodeInfo[index].nodeFqdn, &syncCfg.nodeInfo[index].nodePort); + for (index = 0; index < mnodes->mnodeNum; ++index) { + SMnodeInfo *node = &mnodes->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; diff --git a/src/mnode/src/mnodeShow.c b/src/mnode/src/mnodeShow.c index 655fe7125977c5a19061527a16eb1b1d6758490f..2c9d404e1d99ea4705363caba363b2d6380dca07 100644 --- a/src/mnode/src/mnodeShow.c +++ b/src/mnode/src/mnodeShow.c @@ -110,7 +110,7 @@ static char *mnodeGetShowType(int32_t showType) { } static int32_t mnodeProcessShowMsg(SMnodeMsg *pMsg) { - SCMShowMsg *pShowMsg = pMsg->rpcMsg.pCont; + SShowMsg *pShowMsg = pMsg->rpcMsg.pCont; if (pShowMsg->type >= TSDB_MGMT_TABLE_MAX) { return TSDB_CODE_MND_INVALID_MSG_TYPE; } @@ -132,8 +132,8 @@ static int32_t mnodeProcessShowMsg(SMnodeMsg *pMsg) { return TSDB_CODE_MND_OUT_OF_MEMORY; } - int32_t size = sizeof(SCMShowRsp) + sizeof(SSchema) * TSDB_MAX_COLUMNS + TSDB_EXTRA_PAYLOAD_SIZE; - SCMShowRsp *pShowRsp = rpcMallocCont(size); + int32_t size = sizeof(SShowRsp) + sizeof(SSchema) * TSDB_MAX_COLUMNS + TSDB_EXTRA_PAYLOAD_SIZE; + SShowRsp *pShowRsp = rpcMallocCont(size); if (pShowRsp == NULL) { mnodeReleaseShowObj(pShow, true); return TSDB_CODE_MND_OUT_OF_MEMORY; @@ -146,7 +146,7 @@ static int32_t mnodeProcessShowMsg(SMnodeMsg *pMsg) { if (code == TSDB_CODE_SUCCESS) { pMsg->rpcRsp.rsp = pShowRsp; - pMsg->rpcRsp.len = sizeof(SCMShowRsp) + sizeof(SSchema) * pShow->numOfColumns; + pMsg->rpcRsp.len = sizeof(SShowRsp) + sizeof(SSchema) * pShow->numOfColumns; mnodeReleaseShowObj(pShow, false); return TSDB_CODE_SUCCESS; } else { @@ -232,13 +232,14 @@ static int32_t mnodeProcessRetrieveMsg(SMnodeMsg *pMsg) { } static int32_t mnodeProcessHeartBeatMsg(SMnodeMsg *pMsg) { - SCMHeartBeatRsp *pRsp = (SCMHeartBeatRsp *) rpcMallocCont(sizeof(SCMHeartBeatRsp)); + SHeartBeatRsp *pRsp = (SHeartBeatRsp *)rpcMallocCont(sizeof(SHeartBeatRsp)); if (pRsp == NULL) { return TSDB_CODE_MND_OUT_OF_MEMORY; } - SCMHeartBeatMsg *pHBMsg = pMsg->rpcMsg.pCont; + SHeartBeatMsg *pHBMsg = pMsg->rpcMsg.pCont; if (taosCheckVersion(pHBMsg->clientVer, version, 3) != TSDB_CODE_SUCCESS) { + rpcFreeCont(pRsp); return TSDB_CODE_TSC_INVALID_VERSION; // todo change the error code } @@ -280,15 +281,15 @@ static int32_t mnodeProcessHeartBeatMsg(SMnodeMsg *pMsg) { mnodeGetMnodeEpSetForShell(&pRsp->epSet); pMsg->rpcRsp.rsp = pRsp; - pMsg->rpcRsp.len = sizeof(SCMHeartBeatRsp); + pMsg->rpcRsp.len = sizeof(SHeartBeatRsp); mnodeReleaseConn(pConn); return TSDB_CODE_SUCCESS; } static int32_t mnodeProcessConnectMsg(SMnodeMsg *pMsg) { - SCMConnectMsg *pConnectMsg = pMsg->rpcMsg.pCont; - SCMConnectRsp *pConnectRsp = NULL; + SConnectMsg *pConnectMsg = pMsg->rpcMsg.pCont; + SConnectRsp *pConnectRsp = NULL; int32_t code = TSDB_CODE_SUCCESS; SRpcConnInfo connInfo = {0}; @@ -324,7 +325,7 @@ static int32_t mnodeProcessConnectMsg(SMnodeMsg *pMsg) { mnodeDecDbRef(pDb); } - pConnectRsp = rpcMallocCont(sizeof(SCMConnectRsp)); + pConnectRsp = rpcMallocCont(sizeof(SConnectRsp)); if (pConnectRsp == NULL) { code = TSDB_CODE_MND_OUT_OF_MEMORY; goto connect_over; @@ -353,14 +354,14 @@ connect_over: } else { mLInfo("user:%s login from %s, result:%s", connInfo.user, taosIpStr(connInfo.clientIp), tstrerror(code)); pMsg->rpcRsp.rsp = pConnectRsp; - pMsg->rpcRsp.len = sizeof(SCMConnectRsp); + pMsg->rpcRsp.len = sizeof(SConnectRsp); } return code; } static int32_t mnodeProcessUseMsg(SMnodeMsg *pMsg) { - SCMUseDbMsg *pUseDbMsg = pMsg->rpcMsg.pCont; + SUseDbMsg *pUseDbMsg = pMsg->rpcMsg.pCont; int32_t code = TSDB_CODE_SUCCESS; if (pMsg->pDb == NULL) pMsg->pDb = mnodeGetDb(pUseDbMsg->db); diff --git a/src/mnode/src/mnodeTable.c b/src/mnode/src/mnodeTable.c index 82a062169a591d26d665437273183a4c5e1f27b4..28dc2c6cff7b5e8eabbe78de89051cc9a5cddde4 100644 --- a/src/mnode/src/mnodeTable.c +++ b/src/mnode/src/mnodeTable.c @@ -778,7 +778,7 @@ static int32_t mnodeProcessDropTableMsg(SMnodeMsg *pMsg) { } static int32_t mnodeProcessTableMetaMsg(SMnodeMsg *pMsg) { - SCMTableInfoMsg *pInfo = pMsg->rpcMsg.pCont; + STableInfoMsg *pInfo = pMsg->rpcMsg.pCont; pInfo->createFlag = htons(pInfo->createFlag); mDebug("app:%p:%p, table:%s, table meta msg is received from thandle:%p, createFlag:%d", pMsg->rpcMsg.ahandle, pMsg, pInfo->tableId, pMsg->rpcMsg.handle, pInfo->createFlag); @@ -915,8 +915,8 @@ static int32_t mnodeProcessDropSuperTableMsg(SMnodeMsg *pMsg) { SVgObj *pVgroup = mnodeGetVgroup(*pVgId); if (pVgroup == NULL) break; - SMDDropSTableMsg *pDrop = rpcMallocCont(sizeof(SMDDropSTableMsg)); - pDrop->contLen = htonl(sizeof(SMDDropSTableMsg)); + SDropSTableMsg *pDrop = rpcMallocCont(sizeof(SDropSTableMsg)); + pDrop->contLen = htonl(sizeof(SDropSTableMsg)); pDrop->vgId = htonl(pVgroup->vgId); pDrop->uid = htobe64(pStable->uid); mnodeExtractTableName(pStable->info.tableId, pDrop->tableId); @@ -924,7 +924,7 @@ static int32_t mnodeProcessDropSuperTableMsg(SMnodeMsg *pMsg) { mInfo("app:%p:%p, stable:%s, send drop stable msg to vgId:%d", pMsg->rpcMsg.ahandle, pMsg, pStable->info.tableId, pVgroup->vgId); SRpcEpSet epSet = mnodeGetEpSetFromVgroup(pVgroup); - SRpcMsg rpcMsg = {.pCont = pDrop, .contLen = sizeof(SMDDropSTableMsg), .msgType = TSDB_MSG_TYPE_MD_DROP_STABLE}; + SRpcMsg rpcMsg = {.pCont = pDrop, .contLen = sizeof(SDropSTableMsg), .msgType = TSDB_MSG_TYPE_MD_DROP_STABLE}; dnodeSendMsgToDnode(&epSet, &rpcMsg); mnodeDecVgroupRef(pVgroup); } @@ -1472,31 +1472,31 @@ static int32_t mnodeGetSuperTableMeta(SMnodeMsg *pMsg) { } static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) { - SCMSTableVgroupMsg *pInfo = pMsg->rpcMsg.pCont; + SSTableVgroupMsg *pInfo = pMsg->rpcMsg.pCont; int32_t numOfTable = htonl(pInfo->numOfTables); // reserve space - int32_t contLen = sizeof(SCMSTableVgroupRspMsg) + 32 * sizeof(SCMVgroupMsg) + sizeof(SVgroupsMsg); + int32_t contLen = sizeof(SSTableVgroupRspMsg) + 32 * sizeof(SVgroupMsg) + sizeof(SVgroupsMsg); for (int32_t i = 0; i < numOfTable; ++i) { - char *stableName = (char*)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_FNAME_LEN) * i; + char *stableName = (char *)pInfo + sizeof(SSTableVgroupMsg) + (TSDB_TABLE_FNAME_LEN)*i; SSuperTableObj *pTable = mnodeGetSuperTable(stableName); if (pTable != NULL && pTable->vgHash != NULL) { - contLen += (taosHashGetSize(pTable->vgHash) * sizeof(SCMVgroupMsg) + sizeof(SVgroupsMsg)); - } - + contLen += (taosHashGetSize(pTable->vgHash) * sizeof(SVgroupMsg) + sizeof(SVgroupsMsg)); + } + mnodeDecTableRef(pTable); } - SCMSTableVgroupRspMsg *pRsp = rpcMallocCont(contLen); + SSTableVgroupRspMsg *pRsp = rpcMallocCont(contLen); if (pRsp == NULL) { return TSDB_CODE_MND_OUT_OF_MEMORY; } pRsp->numOfTables = 0; - char *msg = (char *)pRsp + sizeof(SCMSTableVgroupRspMsg); + char *msg = (char *)pRsp + sizeof(SSTableVgroupRspMsg); for (int32_t i = 0; i < numOfTable; ++i) { - char * stableName = (char *)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_FNAME_LEN)*i; + char * stableName = (char *)pInfo + sizeof(SSTableVgroupMsg) + (TSDB_TABLE_FNAME_LEN)*i; SSuperTableObj *pTable = mnodeGetSuperTable(stableName); if (pTable == NULL) { mError("app:%p:%p, stable:%s, not exist while get stable vgroup info", pMsg->rpcMsg.ahandle, pMsg, stableName); @@ -1548,7 +1548,7 @@ static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) { pVgroupMsg->numOfVgroups = htonl(vgSize); // one table is done, try the next table - msg += sizeof(SVgroupsMsg) + vgSize * sizeof(SCMVgroupMsg); + msg += sizeof(SVgroupsMsg) + vgSize * sizeof(SVgroupMsg); pRsp->numOfTables++; } } @@ -2146,7 +2146,7 @@ static int32_t mnodeDoGetChildTableMeta(SMnodeMsg *pMsg, STableMetaMsg *pMeta) { } static int32_t mnodeAutoCreateChildTable(SMnodeMsg *pMsg) { - SCMTableInfoMsg *pInfo = pMsg->rpcMsg.pCont; + STableInfoMsg *pInfo = pMsg->rpcMsg.pCont; STagData *pTags = (STagData *)pInfo->tags; int32_t tagLen = htonl(pTags->dataLen); if (pTags->name[0] == 0) { @@ -2307,7 +2307,7 @@ static SChildTableObj* mnodeGetTableByPos(int32_t vnode, int32_t tid) { static int32_t mnodeProcessTableCfgMsg(SMnodeMsg *pMsg) { return TSDB_CODE_COM_OPS_NOT_SUPPORT; #if 0 - SDMConfigTableMsg *pCfg = pMsg->rpcMsg.pCont; + SConfigTableMsg *pCfg = pMsg->rpcMsg.pCont; pCfg->dnodeId = htonl(pCfg->dnodeId); pCfg->vgId = htonl(pCfg->vgId); pCfg->sid = htonl(pCfg->sid); @@ -2469,7 +2469,7 @@ static void mnodeProcessAlterTableRsp(SRpcMsg *rpcMsg) { } static int32_t mnodeProcessMultiTableMetaMsg(SMnodeMsg *pMsg) { - SCMMultiTableInfoMsg *pInfo = pMsg->rpcMsg.pCont; + SMultiTableInfoMsg *pInfo = pMsg->rpcMsg.pCont; pInfo->numOfTables = htonl(pInfo->numOfTables); int32_t totalMallocLen = 4 * 1024 * 1024; // first malloc 4 MB, subsequent reallocation as twice @@ -2705,7 +2705,7 @@ static int32_t mnodeRetrieveShowTables(SShowObj *pShow, char *data, int32_t rows } static int32_t mnodeProcessAlterTableMsg(SMnodeMsg *pMsg) { - SCMAlterTableMsg *pAlter = pMsg->rpcMsg.pCont; + SAlterTableMsg *pAlter = pMsg->rpcMsg.pCont; mDebug("app:%p:%p, table:%s, alter table msg is received from thandle:%p", pMsg->rpcMsg.ahandle, pMsg, pAlter->tableId, pMsg->rpcMsg.handle); diff --git a/src/mnode/src/mnodeUser.c b/src/mnode/src/mnodeUser.c index 779e25254897ff37f1ca884e83469233d7197b8c..8e9d2908c5df2b83cae49461b9bd40732d864ce3 100644 --- a/src/mnode/src/mnodeUser.c +++ b/src/mnode/src/mnodeUser.c @@ -414,7 +414,7 @@ static int32_t mnodeProcessCreateUserMsg(SMnodeMsg *pMsg) { SUserObj *pOperUser = pMsg->pUser; if (pOperUser->superAuth) { - SCMCreateUserMsg *pCreate = pMsg->rpcMsg.pCont; + SCreateUserMsg *pCreate = pMsg->rpcMsg.pCont; return mnodeCreateUser(pOperUser->pAcct, pCreate->user, pCreate->pass, pMsg); } else { mError("user:%s, no rights to create user", pOperUser->user); @@ -426,7 +426,7 @@ static int32_t mnodeProcessAlterUserMsg(SMnodeMsg *pMsg) { int32_t code; SUserObj *pOperUser = pMsg->pUser; - SCMAlterUserMsg *pAlter = pMsg->rpcMsg.pCont; + SAlterUserMsg *pAlter = pMsg->rpcMsg.pCont; SUserObj *pUser = mnodeGetUser(pAlter->user); if (pUser == NULL) { return TSDB_CODE_MND_INVALID_USER; @@ -514,7 +514,7 @@ static int32_t mnodeProcessDropUserMsg(SMnodeMsg *pMsg) { int32_t code; SUserObj *pOperUser = pMsg->pUser; - SCMDropUserMsg *pDrop = pMsg->rpcMsg.pCont; + SDropUserMsg *pDrop = pMsg->rpcMsg.pCont; SUserObj *pUser = mnodeGetUser(pDrop->user); if (pUser == NULL) { return TSDB_CODE_MND_INVALID_USER; @@ -604,11 +604,11 @@ int32_t mnodeRetriveAuth(char *user, char *spi, char *encrypt, char *secret, cha } static int32_t mnodeProcessAuthMsg(SMnodeMsg *pMsg) { - SDMAuthMsg *pAuthMsg = pMsg->rpcMsg.pCont; - SDMAuthRsp *pAuthRsp = rpcMallocCont(sizeof(SDMAuthRsp)); + SAuthMsg *pAuthMsg = pMsg->rpcMsg.pCont; + SAuthRsp *pAuthRsp = rpcMallocCont(sizeof(SAuthRsp)); pMsg->rpcRsp.rsp = pAuthRsp; - pMsg->rpcRsp.len = sizeof(SDMAuthRsp); + pMsg->rpcRsp.len = sizeof(SAuthRsp); return mnodeRetriveAuth(pAuthMsg->user, &pAuthRsp->spi, &pAuthRsp->encrypt, pAuthRsp->secret, pAuthRsp->ckey); } diff --git a/src/mnode/src/mnodeVgroup.c b/src/mnode/src/mnodeVgroup.c index df703241c2612294fb5ac5f970f9ed68b0ac4e0e..1d1f21c3d7b2a3092ca13b699a08bed332a985d4 100644 --- a/src/mnode/src/mnodeVgroup.c +++ b/src/mnode/src/mnodeVgroup.c @@ -818,11 +818,11 @@ void mnodeRemoveTableFromVgroup(SVgObj *pVgroup, SChildTableObj *pTable) { } } -static SMDCreateVnodeMsg *mnodeBuildVnodeMsg(SVgObj *pVgroup) { +static SCreateVnodeMsg *mnodeBuildVnodeMsg(SVgObj *pVgroup) { SDbObj *pDb = pVgroup->pDb; if (pDb == NULL) return NULL; - SMDCreateVnodeMsg *pVnode = rpcMallocCont(sizeof(SMDCreateVnodeMsg)); + SCreateVnodeMsg *pVnode = rpcMallocCont(sizeof(SCreateVnodeMsg)); if (pVnode == NULL) return NULL; strcpy(pVnode->db, pVgroup->dbName); @@ -830,7 +830,7 @@ static SMDCreateVnodeMsg *mnodeBuildVnodeMsg(SVgObj *pVgroup) { //TODO: dynamic alloc tables in tsdb maxTables = MAX(10000, tsMaxTablePerVnode); - SMDVnodeCfg *pCfg = &pVnode->cfg; + SVnodeCfg *pCfg = &pVnode->cfg; pCfg->vgId = htonl(pVgroup->vgId); pCfg->cfgVersion = htonl(pDb->cfgVersion); pCfg->cacheBlockSize = htonl(pDb->cfg.cacheBlockSize); @@ -851,7 +851,7 @@ static SMDCreateVnodeMsg *mnodeBuildVnodeMsg(SVgObj *pVgroup) { pCfg->wals = 3; pCfg->quorum = pDb->cfg.quorum; - SMDVnodeDesc *pNodes = pVnode->nodes; + SVnodeDesc *pNodes = pVnode->nodes; for (int32_t j = 0; j < pVgroup->numOfVnodes; ++j) { SDnodeObj *pDnode = pVgroup->vnodeGid[j].pDnode; if (pDnode != NULL) { @@ -886,11 +886,11 @@ SRpcEpSet mnodeGetEpSetFromIp(char *ep) { } static void mnodeSendAlterVnodeMsg(SVgObj *pVgroup, SRpcEpSet *epSet) { - SMDAlterVnodeMsg *pAlter = mnodeBuildVnodeMsg(pVgroup); + SAlterVnodeMsg *pAlter = mnodeBuildVnodeMsg(pVgroup); SRpcMsg rpcMsg = { .ahandle = NULL, .pCont = pAlter, - .contLen = pAlter ? sizeof(SMDAlterVnodeMsg) : 0, + .contLen = pAlter ? sizeof(SAlterVnodeMsg) : 0, .code = 0, .msgType = TSDB_MSG_TYPE_MD_ALTER_VNODE }; @@ -909,11 +909,11 @@ void mnodeSendAlterVgroupMsg(SVgObj *pVgroup) { } static void mnodeSendCreateVnodeMsg(SVgObj *pVgroup, SRpcEpSet *epSet, void *ahandle) { - SMDCreateVnodeMsg *pCreate = mnodeBuildVnodeMsg(pVgroup); + SCreateVnodeMsg *pCreate = mnodeBuildVnodeMsg(pVgroup); SRpcMsg rpcMsg = { .ahandle = ahandle, .pCont = pCreate, - .contLen = pCreate ? sizeof(SMDCreateVnodeMsg) : 0, + .contLen = pCreate ? sizeof(SCreateVnodeMsg) : 0, .code = 0, .msgType = TSDB_MSG_TYPE_MD_CREATE_VNODE }; @@ -982,8 +982,8 @@ static void mnodeProcessCreateVnodeRsp(SRpcMsg *rpcMsg) { } } -static SMDDropVnodeMsg *mnodeBuildDropVnodeMsg(int32_t vgId) { - SMDDropVnodeMsg *pDrop = rpcMallocCont(sizeof(SMDDropVnodeMsg)); +static SDropVnodeMsg *mnodeBuildDropVnodeMsg(int32_t vgId) { + SDropVnodeMsg *pDrop = rpcMallocCont(sizeof(SDropVnodeMsg)); if (pDrop == NULL) return NULL; pDrop->vgId = htonl(vgId); @@ -991,11 +991,11 @@ static SMDDropVnodeMsg *mnodeBuildDropVnodeMsg(int32_t vgId) { } void mnodeSendDropVnodeMsg(int32_t vgId, SRpcEpSet *epSet, void *ahandle) { - SMDDropVnodeMsg *pDrop = mnodeBuildDropVnodeMsg(vgId); + SDropVnodeMsg *pDrop = mnodeBuildDropVnodeMsg(vgId); SRpcMsg rpcMsg = { .ahandle = ahandle, .pCont = pDrop, - .contLen = pDrop ? sizeof(SMDDropVnodeMsg) : 0, + .contLen = pDrop ? sizeof(SDropVnodeMsg) : 0, .code = 0, .msgType = TSDB_MSG_TYPE_MD_DROP_VNODE }; @@ -1044,7 +1044,7 @@ static void mnodeProcessDropVnodeRsp(SRpcMsg *rpcMsg) { } static int32_t mnodeProcessVnodeCfgMsg(SMnodeMsg *pMsg) { - SDMConfigVnodeMsg *pCfg = pMsg->rpcMsg.pCont; + SConfigVnodeMsg *pCfg = pMsg->rpcMsg.pCont; pCfg->dnodeId = htonl(pCfg->dnodeId); pCfg->vgId = htonl(pCfg->vgId); diff --git a/src/os/inc/os.h b/src/os/inc/os.h index 86e16db8b1446308060945d3a7db2531287c62ec..97204990043ae9ceacf9c9f3e1c7cdd52d5060e0 100644 --- a/src/os/inc/os.h +++ b/src/os/inc/os.h @@ -52,9 +52,10 @@ extern "C" { #include "osWindows.h" #endif +#include "osDef.h" +#include "osAlloc.h" #include "osAtomic.h" #include "osCommon.h" -#include "osDef.h" #include "osDir.h" #include "osFile.h" #include "osLz4.h" diff --git a/src/os/inc/osAlloc.h b/src/os/inc/osAlloc.h new file mode 100644 index 0000000000000000000000000000000000000000..2d970174800373300832706b2b16225c106e5585 --- /dev/null +++ b/src/os/inc/osAlloc.h @@ -0,0 +1,43 @@ +/* + * 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_OS_ALLOC_H +#define TDENGINE_OS_ALLOC_H + +#ifdef __cplusplus +extern "C" { +#endif + +#ifndef TAOS_OS_FUNC_ALLOC + #define tmalloc(size) malloc(size) + #define tcalloc(nmemb, size) calloc(nmemb, size) + #define trealloc(p, size) realloc(p, size) + #define tmemalign(alignment, size) malloc(size) + #define tfree(p) free(p) + #define tmemzero(p, size) memset(p, 0, size) +#else + void *tmalloc(int32_t size); + void *tcalloc(int32_t nmemb, int32_t size); + void *trealloc(void *p, int32_t size); + void *tmemalign(int32_t alignment, int32_t size); + void tfree(void *p); + void tmemzero(void *p, int32_t size); +#endif + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/os/inc/osDarwin.h b/src/os/inc/osDarwin.h index c1a950fbe6375b14e3cf25277eb933b96e518f10..7bb844831e03dd1e21abd389a61976e14927ebbd 100644 --- a/src/os/inc/osDarwin.h +++ b/src/os/inc/osDarwin.h @@ -72,8 +72,6 @@ extern "C" { #include #define TAOS_OS_FUNC_FILE_SENDIFLE - #define taosFSendFile(outfile, infile, offset, count) taosFSendFileImp(outfile, infile, offset, size) - #define taosTSendFile(dfd, sfd, offset, size) taosTSendFileImp(dfd, sfd, offset, size) #define TAOS_OS_FUNC_SEMPHONE #define tsem_t dispatch_semaphore_t diff --git a/src/os/inc/osFile.h b/src/os/inc/osFile.h index dc19c8177c62307ca110486c130f6a6b56047ef4..62e44d8eb0b70fb1526693895847637daa72247a 100644 --- a/src/os/inc/osFile.h +++ b/src/os/inc/osFile.h @@ -20,46 +20,52 @@ extern "C" { #endif -ssize_t taosTReadImp(int fd, void *buf, size_t count); -ssize_t taosTWriteImp(int fd, void *buf, size_t count); +#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; \ + } \ + } -ssize_t taosTSendFileImp(int dfd, int sfd, off_t *offset, size_t size); -int taosFSendFileImp(FILE* out_file, FILE* in_file, int64_t* offset, int32_t count); +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); +int32_t taosRenameFile(char *fullPath, char *suffix, char delimiter, char **dstPath); -#ifndef TAOS_OS_FUNC_FILE_SENDIFLE - #define taosTSendFile(dfd, sfd, offset, size) taosTSendFileImp(dfd, sfd, offset, size) - #define taosFSendFile(outfile, infile, offset, count) taosTSendFileImp(fileno(outfile), fileno(infile), offset, size) -#endif +#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 taosTRead(fd, buf, count) taosTReadImp(fd, buf, count) -#define taosTWrite(fd, buf, count) taosTWriteImp(fd, buf, count) -#define taosLSeek(fd, offset, whence) lseek(fd, offset, whence) +// TAOS_OS_FUNC_FILE_SENDIFLE +int64_t taosSendFile(int32_t dfd, int32_t sfd, int64_t *offset, int64_t size); +int64_t taosFSendFile(FILE *outfile, FILE *infile, int64_t *offset, int64_t size); #ifdef TAOS_RANDOM_FILE_FAIL - void taosSetRandomFileFailFactor(int factor); + void taosSetRandomFileFailFactor(int32_t factor); void taosSetRandomFileFailOutput(const char *path); #ifdef TAOS_RANDOM_FILE_FAIL_TEST - ssize_t taosReadFileRandomFail(int fd, void *buf, size_t count, const char *file, uint32_t line); - ssize_t taosWriteFileRandomFail(int fd, void *buf, size_t count, const char *file, uint32_t line); - off_t taosLSeekRandomFail(int fd, off_t offset, int whence, const char *file, uint32_t line); - #undef taosTRead - #undef taosTWrite + int64_t taosReadFileRandomFail(int32_t fd, void *buf, int32_t count, const char *file, uint32_t line); + int64_t taosWriteFileRandomFail(int32_t fd, void *buf, int32_t count, const char *file, uint32_t line); + int64_t taosLSeekRandomFail(int32_t fd, int64_t offset, int32_t whence, const char *file, uint32_t line); + #undef taosRead + #undef taosWrite #undef taosLSeek - #define taosTRead(fd, buf, count) taosReadFileRandomFail(fd, buf, count, __FILE__, __LINE__) - #define taosTWrite(fd, buf, count) taosWriteFileRandomFail(fd, buf, count, __FILE__, __LINE__) + #define taosRead(fd, buf, count) taosReadFileRandomFail(fd, buf, count, __FILE__, __LINE__) + #define taosWrite(fd, buf, count) taosWriteFileRandomFail(fd, buf, count, __FILE__, __LINE__) #define taosLSeek(fd, offset, whence) taosLSeekRandomFail(fd, offset, whence, __FILE__, __LINE__) #endif #endif -int32_t taosFileRename(char *fullPath, char *suffix, char delimiter, char **dstPath); - // TAOS_OS_FUNC_FILE_GETTMPFILEPATH void taosGetTmpfilePath(const char *fileNamePrefix, char *dstPath); -#ifndef TAOS_OS_FUNC_FILE_FTRUNCATE - #define taosFtruncate ftruncate -#endif - +// TAOS_OS_FUNC_FILE_FTRUNCATE +int32_t taosFtruncate(int32_t fd, int64_t length); #ifdef __cplusplus } #endif diff --git a/src/os/inc/osSocket.h b/src/os/inc/osSocket.h index 0ab3ff0fcafa43a404533f145a0bb1595bb29061..cbfdedef485f73f0005b2da1e1d9cc3adf9dd377 100644 --- a/src/os/inc/osSocket.h +++ b/src/os/inc/osSocket.h @@ -33,21 +33,19 @@ extern "C" { x = FD_INITIALIZER; \ } \ } - typedef int SOCKET; + typedef int32_t SOCKET; #endif #ifndef TAOS_OS_DEF_EPOLL #define TAOS_EPOLL_WAIT_TIME -1 #endif -#define taosClose(x) taosCloseSocket(x) - #ifdef TAOS_RANDOM_NETWORK_FAIL #ifdef TAOS_RANDOM_NETWORK_FAIL_TEST - ssize_t taosSendRandomFail(int sockfd, const void *buf, size_t len, int flags); - ssize_t taosSendToRandomFail(int sockfd, const void *buf, size_t len, int flags, const struct sockaddr *dest_addr, socklen_t addrlen); - ssize_t taosReadSocketRandomFail(int fd, void *buf, size_t count); - ssize_t taosWriteSocketRandomFail(int fd, const void *buf, size_t count); + int64_t taosSendRandomFail(int32_t sockfd, const void *buf, size_t len, int32_t flags); + int64_t taosSendToRandomFail(int32_t sockfd, const void *buf, size_t len, int32_t flags, const struct sockaddr *dest_addr, socklen_t addrlen); + int64_t taosReadSocketRandomFail(int32_t fd, void *buf, size_t count); + int64_t taosWriteSocketRandomFail(int32_t fd, const void *buf, size_t count); #undef taosSend #undef taosSendto #undef taosReadSocket @@ -60,14 +58,14 @@ extern "C" { #endif // TAOS_OS_FUNC_SOCKET -int taosSetNonblocking(SOCKET sock, int on); -void taosBlockSIGPIPE(); +int32_t taosSetNonblocking(SOCKET sock, int32_t on); +void taosBlockSIGPIPE(); // TAOS_OS_FUNC_SOCKET_SETSOCKETOPT -int taosSetSockOpt(SOCKET socketfd, int level, int optname, void *optval, int optlen); +int32_t taosSetSockOpt(SOCKET socketfd, int32_t level, int32_t optname, void *optval, int32_t optlen); // TAOS_OS_FUNC_SOCKET_INET -uint32_t taosInetAddr(char *ipAddr); +uint32_t taosInetAddr(char *ipAddr); const char *taosInetNtoa(struct in_addr ipInt); #ifdef __cplusplus diff --git a/src/os/inc/osWindows.h b/src/os/inc/osWindows.h index dc1da35037dc06c258856a1a143146492eeed0e5..36e30528bf9d84739fd391442145d7e65e1f7df2 100644 --- a/src/os/inc/osWindows.h +++ b/src/os/inc/osWindows.h @@ -62,11 +62,8 @@ extern "C" { #define TAOS_OS_FUNC_FILE_ISDIR #define TAOS_OS_FUNC_FILE_ISLNK #define TAOS_OS_FUNC_FILE_SENDIFLE - #define taosFSendFile(outfile, infile, offset, count) taosFSendFileImp(outfile, infile, offset, size) - #define taosTSendFile(dfd, sfd, offset, size) taosTSendFileImp(dfd, sfd, offset, size) #define TAOS_OS_FUNC_FILE_GETTMPFILEPATH -#define TAOS_OS_FUNC_FILE_FTRUNCATE - extern int taosFtruncate(int fd, int64_t length); +#define TAOS_OS_FUNC_FILE_FTRUNCATE #define TAOS_OS_FUNC_MATH #define SWAP(a, b, c) \ @@ -139,7 +136,6 @@ typedef int (*__compar_fn_t)(const void *, const void *); #define in_addr_t unsigned long #define socklen_t int #define htobe64 htonll -#define twrite write #define getpid _getpid struct tm *localtime_r(const time_t *timep, struct tm *result); diff --git a/src/os/src/darwin/darwinFile.c b/src/os/src/darwin/darwinFile.c index 66bdb5b939d77ab03b30f024466fe3e3ff4e9388..dacf4db74137b8cc58db2b390ba971cf9dfbc012 100644 --- a/src/os/src/darwin/darwinFile.c +++ b/src/os/src/darwin/darwinFile.c @@ -19,21 +19,19 @@ #define _SEND_FILE_STEP_ 1000 -int taosFSendFileImp(FILE* out_file, FILE* in_file, int64_t* offset, int32_t count) { +int64_t taosFSendFile(FILE *out_file, FILE *in_file, int64_t *offset, int64_t count) { fseek(in_file, (int32_t)(*offset), 0); - int writeLen = 0; - uint8_t buffer[_SEND_FILE_STEP_] = { 0 }; + int writeLen = 0; + uint8_t buffer[_SEND_FILE_STEP_] = {0}; for (int len = 0; len < (count - _SEND_FILE_STEP_); len += _SEND_FILE_STEP_) { size_t rlen = fread(buffer, 1, _SEND_FILE_STEP_, in_file); if (rlen <= 0) { return writeLen; - } - else if (rlen < _SEND_FILE_STEP_) { + } else if (rlen < _SEND_FILE_STEP_) { fwrite(buffer, 1, rlen, out_file); return (int)(writeLen + rlen); - } - else { + } else { fwrite(buffer, 1, _SEND_FILE_STEP_, in_file); writeLen += _SEND_FILE_STEP_; } @@ -44,8 +42,7 @@ int taosFSendFileImp(FILE* out_file, FILE* in_file, int64_t* offset, int32_t cou size_t rlen = fread(buffer, 1, remain, in_file); if (rlen <= 0) { return writeLen; - } - else { + } else { fwrite(buffer, 1, remain, out_file); writeLen += remain; } @@ -54,7 +51,7 @@ int taosFSendFileImp(FILE* out_file, FILE* in_file, int64_t* offset, int32_t cou return writeLen; } -ssize_t taosTSendFileImp(int dfd, int sfd, off_t *offset, size_t size) { - uError("not implemented yet"); +int64_t taosSendFile(int32_t dfd, int32_t sfd, int64_t* offset, int64_t size) { + uError("taosSendFile not implemented yet"); return -1; } \ No newline at end of file diff --git a/src/os/src/detail/osAlloc.c b/src/os/src/detail/osAlloc.c new file mode 100644 index 0000000000000000000000000000000000000000..4ca35793e7bd16e024c4dddbfdcc9138ab70d3f8 --- /dev/null +++ b/src/os/src/detail/osAlloc.c @@ -0,0 +1,79 @@ +/* + * 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 "tulog.h" +#include "osAlloc.h" + +#define TSDB_HAVE_MEMALIGN +#ifdef TAOS_OS_FUNC_ALLOC + +void *tmalloc(int32_t size) { + void *p = malloc(size); + if (p == NULL) { + terrno = TAOS_SYSTEM_ERROR(errno); + uError("failed to malloc memory, size:%d reason:%s", size, strerror(errno)); + } + + return p; +} + +void *tcalloc(int32_t nmemb, int32_t size) { + void *p = calloc(nmemb, size); + if (p == NULL) { + terrno = TAOS_SYSTEM_ERROR(errno); + uError("failed to calloc memory, nmemb:%d size:%d reason:%s", nmemb, size, strerror(errno)); + } + + return p; +} + +void *trealloc(void *p, int32_t size) { + p = realloc(p, size); + if (p == NULL) { + terrno = TAOS_SYSTEM_ERROR(errno); + uError("failed to realloc memory, size:%d reason:%s", size, strerror(errno)); + } + + return p; +} + +void tfree(void *p) { free(p); } + +void tmemzero(void *p, int32_t size) { memset(p, 0, size); } + +#ifdef TSDB_HAVE_MEMALIGN + +void *tmemalign(int32_t alignment, int32_t size) { + void *p; + + int err = posix_memalign(&p, alignment, size); + if (err) { + terrno = TAOS_SYSTEM_ERROR(errno); + uError("failed to memalign memory, alignment:%d size:%d reason:%s", alignment, size, strerror(err)); + p = NULL; + } + + return p; +} + +#else + +void *tmemalign(int32_t alignment, int32_t size) { return tmalloc(size); } + +#endif +#endif \ No newline at end of file diff --git a/src/os/src/detail/osFail.c b/src/os/src/detail/osFail.c index e0eb200851f976e1c9dad47afbc152bfb021e4e5..a99bcd01dbccd0290de1fc38a1220b573ab74b22 100644 --- a/src/os/src/detail/osFail.c +++ b/src/os/src/detail/osFail.c @@ -20,7 +20,7 @@ #ifdef TAOS_RANDOM_NETWORK_FAIL -ssize_t taosSendRandomFail(int sockfd, const void *buf, size_t len, int flags) { +int64_t taosSendRandomFail(int32_t sockfd, const void *buf, size_t len, int32_t flags) { if (rand() % RANDOM_NETWORK_FAIL_FACTOR == 0) { errno = ECONNRESET; return -1; @@ -29,8 +29,8 @@ ssize_t taosSendRandomFail(int sockfd, const void *buf, size_t len, int flags) { return send(sockfd, buf, len, flags); } -ssize_t taosSendToRandomFail(int sockfd, const void *buf, size_t len, int flags, const struct sockaddr *dest_addr, - socklen_t addrlen) { +int64_t taosSendToRandomFail(int32_t sockfd, const void *buf, size_t len, int32_t flags, + const struct sockaddr *dest_addr, socklen_t addrlen) { if (rand() % RANDOM_NETWORK_FAIL_FACTOR == 0) { errno = ECONNRESET; return -1; @@ -39,7 +39,7 @@ ssize_t taosSendToRandomFail(int sockfd, const void *buf, size_t len, int flags, return sendto(sockfd, buf, len, flags, dest_addr, addrlen); } -ssize_t taosReadSocketRandomFail(int fd, void *buf, size_t count) { +int64_t taosReadSocketRandomFail(int32_t fd, void *buf, size_t count) { if (rand() % RANDOM_NETWORK_FAIL_FACTOR == 0) { errno = ECONNRESET; return -1; @@ -48,7 +48,7 @@ ssize_t taosReadSocketRandomFail(int fd, void *buf, size_t count) { return read(fd, buf, count); } -ssize_t taosWriteSocketRandomFail(int fd, const void *buf, size_t count) { +int64_t taosWriteSocketRandomFail(int32_t fd, const void *buf, size_t count) { if (rand() % RANDOM_NETWORK_FAIL_FACTOR == 0) { errno = EINTR; return -1; @@ -61,10 +61,10 @@ ssize_t taosWriteSocketRandomFail(int fd, const void *buf, size_t count) { #ifdef TAOS_RANDOM_FILE_FAIL -static int random_file_fail_factor = 20; +static int32_t random_file_fail_factor = 20; static FILE *fpRandomFileFailOutput = NULL; -void taosSetRandomFileFailFactor(int factor) { +void taosSetRandomFileFailFactor(int32_t factor) { random_file_fail_factor = factor; } @@ -77,7 +77,7 @@ static void close_random_file_fail_output() { } } -static void random_file_fail_output_sig(int sig) { +static void random_file_fail_output_sig(int32_t sig) { fprintf(fpRandomFileFailOutput, "signal %d received.\n", sig); struct sigaction act = {0}; @@ -105,7 +105,7 @@ void taosSetRandomFileFailOutput(const char *path) { sigaction(SIGILL, &act, NULL); } -ssize_t taosReadFileRandomFail(int fd, void *buf, size_t count, const char *file, uint32_t line) { +int64_t taosReadFileRandomFail(int32_t fd, void *buf, int32_t count, const char *file, uint32_t line) { if (random_file_fail_factor > 0) { if (rand() % random_file_fail_factor == 0) { errno = EIO; @@ -113,10 +113,10 @@ ssize_t taosReadFileRandomFail(int fd, void *buf, size_t count, const char *file } } - return taosTReadImp(fd, buf, count); + return taosReadImp(fd, buf, count); } -ssize_t taosWriteFileRandomFail(int fd, void *buf, size_t count, const char *file, uint32_t line) { +int64_t taosWriteFileRandomFail(int32_t fd, void *buf, int32_t count, const char *file, uint32_t line) { if (random_file_fail_factor > 0) { if (rand() % random_file_fail_factor == 0) { errno = EIO; @@ -124,10 +124,10 @@ ssize_t taosWriteFileRandomFail(int fd, void *buf, size_t count, const char *fil } } - return taosTWriteImp(fd, buf, count); + return taosWriteImp(fd, buf, count); } -off_t taosLSeekRandomFail(int fd, off_t offset, int whence, const char *file, uint32_t line) { +int64_t taosLSeekRandomFail(int32_t fd, int64_t offset, int32_t whence, const char *file, uint32_t line) { if (random_file_fail_factor > 0) { if (rand() % random_file_fail_factor == 0) { errno = EIO; @@ -135,7 +135,7 @@ off_t taosLSeekRandomFail(int fd, off_t offset, int whence, const char *file, ui } } - return lseek(fd, offset, whence); + return taosLSeekImp(fd, offset, whence); } #endif //TAOS_RANDOM_FILE_FAIL diff --git a/src/os/src/detail/osFile.c b/src/os/src/detail/osFile.c index 8f055dd8129f340f267e64cdd905505a7b675a2d..6eb4515f3098f89991640d2fdee2b0aca47c1703 100644 --- a/src/os/src/detail/osFile.c +++ b/src/os/src/detail/osFile.c @@ -15,16 +15,22 @@ #define _DEFAULT_SOURCE #include "os.h" +#include "tglobal.h" #ifndef TAOS_OS_FUNC_FILE_GETTMPFILEPATH + void taosGetTmpfilePath(const char *fileNamePrefix, char *dstPath) { const char *tdengineTmpFileNamePrefix = "tdengine-"; char tmpPath[PATH_MAX]; - char *tmpDir = "/tmp/"; + int32_t len = strlen(tsTempDir); + memcpy(tmpPath, tsTempDir, len); + + if (tmpPath[len - 1] != '/') { + tmpPath[len++] = '/'; + } - strcpy(tmpPath, tmpDir); - strcat(tmpPath, tdengineTmpFileNamePrefix); + strcpy(tmpPath + len, tdengineTmpFileNamePrefix); if (strlen(tmpPath) + strlen(fileNamePrefix) + strlen("-%d-%s") < PATH_MAX) { strcat(tmpPath, fileNamePrefix); strcat(tmpPath, "-%d-%s"); @@ -34,10 +40,10 @@ void taosGetTmpfilePath(const char *fileNamePrefix, char *dstPath) { taosRandStr(rand, tListLen(rand) - 1); snprintf(dstPath, PATH_MAX, tmpPath, getpid(), rand); } + #endif -// rename file name -int32_t taosFileRename(char *fullPath, char *suffix, char delimiter, char **dstPath) { +int32_t taosRenameFile(char *fullPath, char *suffix, char delimiter, char **dstPath) { int32_t ts = taosGetTimestampSec(); char fname[PATH_MAX] = {0}; // max file name length must be less than 255 @@ -46,12 +52,13 @@ int32_t taosFileRename(char *fullPath, char *suffix, char delimiter, char **dstP if (delimiterPos == NULL) return -1; int32_t fileNameLen = 0; - if (suffix) + if (suffix) { fileNameLen = snprintf(fname, PATH_MAX, "%s.%d.%s", delimiterPos + 1, ts, suffix); - else + } else { fileNameLen = snprintf(fname, PATH_MAX, "%s.%d", delimiterPos + 1, ts); + } - size_t len = (size_t)((delimiterPos - fullPath) + fileNameLen + 1); + int32_t len = (int32_t)((delimiterPos - fullPath) + fileNameLen + 1); if (*dstPath == NULL) { *dstPath = calloc(1, len + 1); if (*dstPath == NULL) return -1; @@ -64,9 +71,9 @@ int32_t taosFileRename(char *fullPath, char *suffix, char delimiter, char **dstP return rename(fullPath, *dstPath); } -ssize_t taosTReadImp(int fd, void *buf, size_t count) { - size_t leftbytes = count; - ssize_t readbytes; +int64_t taosReadImp(int32_t fd, void *buf, int64_t count) { + int64_t leftbytes = count; + int64_t readbytes; char * tbuf = (char *)buf; while (leftbytes > 0) { @@ -78,19 +85,19 @@ ssize_t taosTReadImp(int fd, void *buf, size_t count) { return -1; } } else if (readbytes == 0) { - return (ssize_t)(count - leftbytes); + return (int64_t)(count - leftbytes); } leftbytes -= readbytes; tbuf += readbytes; } - return (ssize_t)count; + return count; } -ssize_t taosTWriteImp(int fd, void *buf, size_t n) { - size_t nleft = n; - ssize_t nwritten = 0; +int64_t taosWriteImp(int32_t fd, void *buf, int64_t n) { + int64_t nleft = n; + int64_t nwritten = 0; char * tbuf = (char *)buf; while (nleft > 0) { @@ -105,13 +112,18 @@ ssize_t taosTWriteImp(int fd, void *buf, size_t n) { tbuf += nwritten; } - return (ssize_t)n; + return n; +} + +int64_t taosLSeekImp(int32_t fd, int64_t offset, int32_t whence) { + return (int64_t)tlseek(fd, (long)offset, whence); } #ifndef TAOS_OS_FUNC_FILE_SENDIFLE -ssize_t taosTSendFileImp(int dfd, int sfd, off_t *offset, size_t size) { - size_t leftbytes = size; - ssize_t sentbytes; + +int64_t taosSendFile(int32_t dfd, int32_t sfd, int64_t *offset, int64_t size) { + int64_t leftbytes = size; + int64_t sentbytes; while (leftbytes > 0) { /* @@ -126,7 +138,7 @@ ssize_t taosTSendFileImp(int dfd, int sfd, off_t *offset, size_t size) { return -1; } } else if (sentbytes == 0) { - return (ssize_t)(size - leftbytes); + return (int64_t)(size - leftbytes); } leftbytes -= sentbytes; @@ -134,4 +146,17 @@ ssize_t taosTSendFileImp(int dfd, int sfd, off_t *offset, size_t size) { return size; } + +int64_t taosFSendFile(FILE *outfile, FILE *infile, int64_t *offset, int64_t size) { + return taosSendFile(fileno(outfile), fileno(infile), offset, size); +} + +#endif + +#ifndef TAOS_OS_FUNC_FILE_FTRUNCATE + +int32_t taosFtruncate(int32_t fd, int64_t length) { + return ftruncate(fd, length); +} + #endif \ No newline at end of file diff --git a/src/os/src/detail/osSocket.c b/src/os/src/detail/osSocket.c index 8a51c389e9f2efdcf3eaad97d434269ff3c0061f..c7c9d774271555ae5989aa18b7e00d325a0eddde 100644 --- a/src/os/src/detail/osSocket.c +++ b/src/os/src/detail/osSocket.c @@ -19,8 +19,8 @@ #ifndef TAOS_OS_FUNC_SOCKET -int taosSetNonblocking(SOCKET sock, int on) { - int flags = 0; +int32_t taosSetNonblocking(SOCKET sock, int32_t on) { + int32_t flags = 0; if ((flags = fcntl(sock, F_GETFL, 0)) < 0) { uError("fcntl(F_GETFL) error: %d (%s)\n", errno, strerror(errno)); return 1; @@ -43,7 +43,7 @@ void taosBlockSIGPIPE() { sigset_t signal_mask; sigemptyset(&signal_mask); sigaddset(&signal_mask, SIGPIPE); - int rc = pthread_sigmask(SIG_BLOCK, &signal_mask, NULL); + int32_t rc = pthread_sigmask(SIG_BLOCK, &signal_mask, NULL); if (rc != 0) { uError("failed to block SIGPIPE"); } @@ -53,7 +53,7 @@ void taosBlockSIGPIPE() { #ifndef TAOS_OS_FUNC_SOCKET_SETSOCKETOPT -int taosSetSockOpt(SOCKET socketfd, int level, int optname, void *optval, int optlen) { +int32_t taosSetSockOpt(SOCKET socketfd, int32_t level, int32_t optname, void *optval, int32_t optlen) { return setsockopt(socketfd, level, optname, optval, (socklen_t)optlen); } diff --git a/src/os/src/windows/wEnv.c b/src/os/src/windows/wEnv.c index 8110a194904bbbc1166ac57de187de284db34a6b..19351eb7c964a4c2a8a4d1d5d4d1c8ec669908dc 100644 --- a/src/os/src/windows/wEnv.c +++ b/src/os/src/windows/wEnv.c @@ -46,5 +46,16 @@ void osInit() { strcpy(tsDnodeDir, ""); strcpy(tsMnodeDir, ""); strcpy(tsOsName, "Windows"); + + const char *tmpDir = getenv("tmp"); + if (tmpDir != NULL) { + tmpDir = getenv("temp"); + } + if (tmpDir != NULL) { + strcpy(tsTempDir, tmpDir); + } else { + strcpy(tsTempDir, "C:\\Windows\\Temp"); + } + taosWinSocketInit(); } diff --git a/src/os/src/windows/wFile.c b/src/os/src/windows/wFile.c index 5549c078a5220d3e493ce21e2c19e38fb9611cbb..734ed9916d4004a0e776f53e590e9d467ea5055a 100644 --- a/src/os/src/windows/wFile.c +++ b/src/os/src/windows/wFile.c @@ -16,17 +16,20 @@ #define _DEFAULT_SOURCE #include "os.h" #include "tulog.h" +#include "tglobal.h" void taosGetTmpfilePath(const char *fileNamePrefix, char *dstPath) { const char* tdengineTmpFileNamePrefix = "tdengine-"; - char tmpPath[PATH_MAX]; + char tmpPath[PATH_MAX]; - char *tmpDir = getenv("tmp"); - if (tmpDir == NULL) { - tmpDir = ""; + int32_t len = (int32_t)strlen(tsTempDir); + memcpy(tmpPath, tsTempDir, len); + + if (tmpPath[len - 1] != '/' && tmpPath[len - 1] != '\\') { + tmpPath[len++] = '\\'; } - - strcpy(tmpPath, tmpDir); + + strcpy(tmpPath + len, tdengineTmpFileNamePrefix); strcat(tmpPath, tdengineTmpFileNamePrefix); if (strlen(tmpPath) + strlen(fileNamePrefix) + strlen("-%d-%s") < PATH_MAX) { strcat(tmpPath, fileNamePrefix); @@ -40,19 +43,19 @@ void taosGetTmpfilePath(const char *fileNamePrefix, char *dstPath) { #define _SEND_FILE_STEP_ 1000 -int taosFSendFileImp(FILE* out_file, FILE* in_file, int64_t* offset, int32_t count) { +int64_t taosFSendFile(FILE *out_file, FILE *in_file, int64_t *offset, int64_t count) { fseek(in_file, (int32_t)(*offset), 0); - int writeLen = 0; + int64_t writeLen = 0; uint8_t buffer[_SEND_FILE_STEP_] = { 0 }; - for (int len = 0; len < (count - _SEND_FILE_STEP_); len += _SEND_FILE_STEP_) { + for (int64_t len = 0; len < (count - _SEND_FILE_STEP_); len += _SEND_FILE_STEP_) { size_t rlen = fread(buffer, 1, _SEND_FILE_STEP_, in_file); if (rlen <= 0) { return writeLen; } else if (rlen < _SEND_FILE_STEP_) { fwrite(buffer, 1, rlen, out_file); - return (int)(writeLen + rlen); + return (int64_t)(writeLen + rlen); } else { fwrite(buffer, 1, _SEND_FILE_STEP_, in_file); @@ -60,7 +63,7 @@ int taosFSendFileImp(FILE* out_file, FILE* in_file, int64_t* offset, int32_t cou } } - int remain = count - writeLen; + int64_t remain = count - writeLen; if (remain > 0) { size_t rlen = fread(buffer, 1, remain, in_file); if (rlen <= 0) { @@ -75,12 +78,12 @@ int taosFSendFileImp(FILE* out_file, FILE* in_file, int64_t* offset, int32_t cou return writeLen; } -ssize_t taosTSendFileImp(int dfd, int sfd, off_t *offset, size_t size) { - uError("taosTSendFileImp no implemented yet"); +int64_t taosSendFile(int32_t dfd, int32_t sfd, int64_t* offset, int64_t size) { + uError("taosSendFile no implemented yet"); return 0; } -int taosFtruncate(int fd, int64_t length) { +int32_t taosFtruncate(int32_t fd, int64_t length) { uError("taosFtruncate no implemented yet"); return 0; } \ No newline at end of file diff --git a/src/os/src/windows/wSocket.c b/src/os/src/windows/wSocket.c index da9242d6a3bff471c4e749f0faf7c95b6c3a9d8c..3b091b269931e644d9f8c2c01ba3c9cb9ddc520c 100644 --- a/src/os/src/windows/wSocket.c +++ b/src/os/src/windows/wSocket.c @@ -34,7 +34,7 @@ void taosWinSocketInit() { } } -int taosSetNonblocking(SOCKET sock, int on) { +int32_t taosSetNonblocking(SOCKET sock, int32_t on) { u_long mode; if (on) { mode = 1; @@ -48,7 +48,7 @@ int taosSetNonblocking(SOCKET sock, int on) { void taosBlockSIGPIPE() {} -int taosSetSockOpt(SOCKET socketfd, int level, int optname, void *optval, int optlen) { +int32_t taosSetSockOpt(SOCKET socketfd, int32_t level, int32_t optname, void *optval, int32_t optlen) { if (level == SOL_SOCKET && optname == TCP_KEEPCNT) { return 0; } @@ -72,7 +72,7 @@ int taosSetSockOpt(SOCKET socketfd, int level, int optname, void *optval, int op uint32_t taosInetAddr(char *ipAddr) { uint32_t value; - int ret = inet_pton(AF_INET, ipAddr, &value); + int32_t ret = inet_pton(AF_INET, ipAddr, &value); if (ret <= 0) { return INADDR_NONE; } else { diff --git a/src/query/inc/qExecutor.h b/src/query/inc/qExecutor.h index b474bea98717034c68ca4b3beb5a3a61288a064b..d7db897cdf9973019536a7b9bc8f8a4d7a8bc85d 100644 --- a/src/query/inc/qExecutor.h +++ b/src/query/inc/qExecutor.h @@ -40,6 +40,19 @@ typedef struct SGroupResInfo { int32_t rowId; } SGroupResInfo; +typedef struct SResultRowPool { + int32_t elemSize; + int32_t blockSize; + int32_t numOfElemPerBlock; + + struct { + int32_t blockIndex; + int32_t pos; + } position; + + SArray* pData; // SArray +} SResultRowPool; + typedef struct SSqlGroupbyExpr { int16_t tableIndex; SArray* columnInfo; // SArray, group by columns information @@ -48,14 +61,14 @@ typedef struct SSqlGroupbyExpr { int16_t orderType; // order by type: asc/desc } SSqlGroupbyExpr; -typedef struct SWindowResult { +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 - SResultInfo* resultInfo; // For each result column, there is a resultInfo + SResultRowCellInfo* pCellInfo; // For each result column, there is a resultInfo union {STimeWindow win; char* key;}; // start key of current time window -} SWindowResult; +} SResultRow; /** * If the number of generated results is greater than this value, @@ -69,16 +82,14 @@ typedef struct SResultRec { } SResultRec; typedef struct SWindowResInfo { - SWindowResult* pResult; // result list - SHashObj* hashList; // hash list for quick access - int16_t type; // data type for hash key + 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 - int32_t size; // number of result set int64_t startTime; // start time of the first time window for sliding query int64_t prevSKey; // previous (not completed) sliding window start key - int64_t threshold; // threshold to halt query and return the generated results. - int64_t interval; // time window interval } SWindowResInfo; typedef struct SColumnFilterElem { @@ -94,7 +105,7 @@ typedef struct SSingleColumnFilterInfo { SColumnFilterElem* pFilters; } SSingleColumnFilterInfo; -typedef struct STableQueryInfo { // todo merge with the STableQueryInfo struct +typedef struct STableQueryInfo { TSKEY lastKey; int32_t groupIndex; // group id in table list int16_t queryRangeSet; // denote if the query range is set, only available for interval query @@ -122,7 +133,9 @@ typedef struct SQueryCostInfo { uint32_t discardBlocks; uint64_t elapsedTime; uint64_t firstStageMergeTime; - uint64_t internalSupSize; + uint64_t winInfoSize; + uint64_t tableInfoSize; + uint64_t hashSize; uint64_t numOfTimeWindows; } SQueryCostInfo; @@ -155,11 +168,11 @@ typedef struct SQuery { typedef struct SQueryRuntimeEnv { jmp_buf env; - SResultInfo* resultInfo; // todo refactor to merge with SWindowResInfo + SResultRow* pResultRow; // todo refactor to merge with SWindowResInfo SQuery* pQuery; SQLFunctionCtx* pCtx; int32_t numOfRowsPerPage; - int16_t offset[TSDB_MAX_COLUMNS]; + uint16_t offset[TSDB_MAX_COLUMNS]; uint16_t scanFlag; // denotes reversed scan of data or not SFillInfo* pFillInfo; SWindowResInfo windowResInfo; @@ -175,6 +188,11 @@ typedef struct SQueryRuntimeEnv { int32_t interBufSize; // intermediate buffer sizse int32_t prevGroupId; // previous executed group id SDiskbasedResultBuf* pResultBuf; // query result buffer based on blocked-wised disk file + SHashObj* pResultRowHashTable; // quick locate the window object for each result + char* keyBuf; // window key buffer + SResultRowPool* pool; // window result object pool + + int32_t* rowCellInfoOffset;// offset value for each row result cell info } SQueryRuntimeEnv; enum { diff --git a/src/query/inc/qFill.h b/src/query/inc/qFill.h index 6d44fee09557fa8b8d73527677a2c7b238ecb42c..329ea9a789cf7e5c8963788c68981b63ccb234e3 100644 --- a/src/query/inc/qFill.h +++ b/src/query/inc/qFill.h @@ -71,7 +71,7 @@ SFillInfo* taosInitFillInfo(int32_t order, TSKEY skey, int32_t numOfTags, int32_ void taosResetFillInfo(SFillInfo* pFillInfo, TSKEY startTimestamp); -void* taosDestoryFillInfo(SFillInfo *pFillInfo); +void* taosDestroyFillInfo(SFillInfo *pFillInfo); void taosFillSetStartInfo(SFillInfo* pFillInfo, int32_t numOfRows, TSKEY endKey); diff --git a/src/query/inc/qSqlparser.h b/src/query/inc/qSqlparser.h index bc8f9a5e23df72f82bcb3bb5140bec42fe426268..25da04710d08336796e6b63415e89d438c97fb42 100644 --- a/src/query/inc/qSqlparser.h +++ b/src/query/inc/qSqlparser.h @@ -223,13 +223,6 @@ typedef struct tSQLExprList { tSQLExprItem *a; /* One entry for each expression */ } tSQLExprList; -typedef struct tSQLExprListList { - int32_t nList; /* Number of expressions on the list */ - int32_t nAlloc; /* Number of entries allocated below */ - tSQLExprList **a; /* one entry for each row */ -} tSQLExprListList; - - /** * * @param yyp The parser diff --git a/src/query/inc/qTsbuf.h b/src/query/inc/qTsbuf.h index 6c2a955f47577a0da19f6520b2addc44e66b73ba..73b18f8915122305cca4a45604986f2e5593e274 100644 --- a/src/query/inc/qTsbuf.h +++ b/src/query/inc/qTsbuf.h @@ -135,6 +135,12 @@ int32_t tsBufGetNumOfVnodes(STSBuf* pTSBuf); void tsBufGetVnodeIdList(STSBuf* pTSBuf, int32_t* num, int32_t** vnodeId); +int32_t dumpFileBlockByVnodeId(STSBuf* pTSBuf, int32_t vnodeId, void* buf, int32_t* len, int32_t* numOfBlocks); + +STSElem tsBufFindElemStartPosByTag(STSBuf* pTSBuf, tVariant* pTag); + +bool tsBufIsValidElem(STSElem* pElem); + #ifdef __cplusplus } #endif diff --git a/src/query/inc/qUtil.h b/src/query/inc/qUtil.h index 32f26f66f5a46f98477db7e67c7a3a6d988fdfc8..5d649261a6e9fbf8b9eab44aab699ea792f86138 100644 --- a/src/query/inc/qUtil.h +++ b/src/query/inc/qUtil.h @@ -15,13 +15,22 @@ #ifndef TDENGINE_QUERYUTIL_H #define TDENGINE_QUERYUTIL_H +#define SET_RES_WINDOW_KEY(_k, _ori, _len, _uid) \ + do { \ + assert(sizeof(_uid) == sizeof(uint64_t)); \ + *(uint64_t *)(_k) = (_uid); \ + memcpy((_k) + sizeof(uint64_t), (_ori), (_len)); \ + } while (0) + +#define GET_RES_WINDOW_KEY_LEN(_l) ((_l) + sizeof(uint64_t)) + int32_t getOutputInterResultBufSize(SQuery* pQuery); -void clearTimeWindowResBuf(SQueryRuntimeEnv* pRuntimeEnv, SWindowResult* pOneOutputRes); -void copyTimeWindowResBuf(SQueryRuntimeEnv* pRuntimeEnv, SWindowResult* dst, const SWindowResult* src); +void clearResultRow(SQueryRuntimeEnv* pRuntimeEnv, SResultRow* pRow); +void copyResultRow(SQueryRuntimeEnv* pRuntimeEnv, SResultRow* dst, const SResultRow* src); +SResultRowCellInfo* getResultCell(SQueryRuntimeEnv* pRuntimeEnv, const SResultRow* pRow, int32_t index); -int32_t initWindowResInfo(SWindowResInfo* pWindowResInfo, SQueryRuntimeEnv* pRuntimeEnv, int32_t size, - int32_t threshold, int16_t type); +int32_t initWindowResInfo(SWindowResInfo* pWindowResInfo, int32_t size, int32_t threshold, int16_t type); void cleanupTimeWindowInfo(SWindowResInfo* pWindowResInfo); void resetTimeWindowInfo(SQueryRuntimeEnv* pRuntimeEnv, SWindowResInfo* pWindowResInfo); @@ -33,9 +42,9 @@ void closeTimeWindow(SWindowResInfo* pWindowResInfo, int32_t slot); void closeAllTimeWindow(SWindowResInfo* pWindowResInfo); void removeRedundantWindow(SWindowResInfo *pWindowResInfo, TSKEY lastKey, int32_t order); -static FORCE_INLINE SWindowResult *getWindowResult(SWindowResInfo *pWindowResInfo, int32_t slot) { +static FORCE_INLINE SResultRow *getResultRow(SWindowResInfo *pWindowResInfo, int32_t slot) { assert(pWindowResInfo != NULL && slot >= 0 && slot < pWindowResInfo->size); - return &pWindowResInfo->pResult[slot]; + return pWindowResInfo->pResult[slot]; } #define curTimeWindowIndex(_winres) ((_winres)->curIndex) @@ -43,9 +52,9 @@ static FORCE_INLINE SWindowResult *getWindowResult(SWindowResInfo *pWindowResInf bool isWindowResClosed(SWindowResInfo *pWindowResInfo, int32_t slot); -int32_t createQueryResultInfo(SQuery *pQuery, SWindowResult *pResultRow, bool isSTableQuery, size_t interBufSize); +int32_t initResultRow(SResultRow *pResultRow); -static FORCE_INLINE char *getPosInResultPage(SQueryRuntimeEnv *pRuntimeEnv, int32_t columnIndex, SWindowResult *pResult, +static FORCE_INLINE char *getPosInResultPage(SQueryRuntimeEnv *pRuntimeEnv, int32_t columnIndex, SResultRow *pResult, tFilePage* page) { assert(pResult != NULL && pRuntimeEnv != NULL); @@ -62,4 +71,14 @@ bool notNull_filter(SColumnFilterElem *pFilter, char* minval, char* maxval); __filter_func_t *getRangeFilterFuncArray(int32_t type); __filter_func_t *getValueFilterFuncArray(int32_t type); +size_t getWindowResultSize(SQueryRuntimeEnv* pRuntimeEnv); + +SResultRowPool* initResultRowPool(size_t size); +SResultRow* getNewResultRow(SResultRowPool* p); +int64_t getResultRowPoolMemSize(SResultRowPool* p); +void* destroyResultRowPool(SResultRowPool* p); +int32_t getNumOfAllocatedResultRows(SResultRowPool* p); +int32_t getNumOfUsedResultRows(SResultRowPool* p); + + #endif // TDENGINE_QUERYUTIL_H diff --git a/src/query/inc/tsqlfunction.h b/src/query/inc/tsqlfunction.h index 28b9a60102d41a11fa02fee1ca51cbf8e263bf3b..3bd4aad2766954709231f55ecdccfe03209a91bc 100644 --- a/src/query/inc/tsqlfunction.h +++ b/src/query/inc/tsqlfunction.h @@ -145,15 +145,14 @@ typedef struct SInterpInfoDetail { int8_t primaryCol; } SInterpInfoDetail; -typedef struct SResultInfo { +typedef struct SResultRowCellInfo { int8_t hasResult; // result generated, not NULL value - bool initialized; // output buffer has been initialized - bool complete; // query has completed - bool superTableQ; // is super table query - uint32_t bufLen; // buffer size - uint64_t numOfRes; // num of output result in current buffer - void* interResultBuf; // output result buffer -} SResultInfo; + bool initialized; // output buffer has been initialized + bool complete; // query has completed + uint16_t numOfRes; // num of output result in current buffer +} SResultRowCellInfo; + +#define GET_ROWCELL_INTERBUF(_c) ((void*) ((char*)(_c) + sizeof(SResultRowCellInfo))) struct SQLFunctionCtx; @@ -175,9 +174,11 @@ typedef struct SQLFunctionCtx { int16_t inputBytes; int16_t outputType; - int16_t outputBytes; // size of results, determined by function and input column data type - bool hasNull; // null value exist in current block + int16_t outputBytes; // size of results, determined by function and input column data type + int32_t interBufBytes; // internal buffer size + bool hasNull; // null value exist in current block bool requireNull; // require null in some function + bool stableQuery; int16_t functionId; // function id void * aInputElemBuf; char * aOutputBuf; // final result output buffer, point to sdata->data @@ -189,7 +190,8 @@ typedef struct SQLFunctionCtx { void * ptsOutputBuf; // corresponding output buffer for timestamp of each result, e.g., top/bottom*/ SQLPreAggVal preAggVals; tVariant tag; - SResultInfo *resultInfo; + + SResultRowCellInfo *resultInfo; SExtTagsInfo tagInfo; } SQLFunctionCtx; @@ -274,16 +276,16 @@ bool topbot_datablock_filter(SQLFunctionCtx *pCtx, int32_t functionId, const cha (_r)->initialized = false; \ } while (0) -void setResultInfoBuf(SResultInfo *pResInfo, int32_t size, bool superTable, char* buf); +//void setResultInfoBuf(SResultRowCellInfo *pResInfo, char* buf); -static FORCE_INLINE void initResultInfo(SResultInfo *pResInfo) { +static FORCE_INLINE void initResultInfo(SResultRowCellInfo *pResInfo, uint32_t bufLen) { pResInfo->initialized = true; // the this struct has been initialized flag pResInfo->complete = false; pResInfo->hasResult = false; pResInfo->numOfRes = 0; - memset(pResInfo->interResultBuf, 0, (size_t)pResInfo->bufLen); + memset(GET_ROWCELL_INTERBUF(pResInfo), 0, (size_t)bufLen); } #ifdef __cplusplus diff --git a/src/query/src/qExecutor.c b/src/query/src/qExecutor.c index 7a7c1ee2d0fc9fe48824ee5fc4fd74e2c2f6834a..d65d01d1558795767a6ad497dea7312444f3686e 100644 --- a/src/query/src/qExecutor.c +++ b/src/query/src/qExecutor.c @@ -178,9 +178,9 @@ static void getNextTimeWindow(SQuery* pQuery, STimeWindow* tw) { // todo move to utility static int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *group); -static void setWindowResOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult); -static void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult); -static void resetMergeResultBuf(SQuery *pQuery, SQLFunctionCtx *pCtx, SResultInfo *pResultInfo); +static void setResultOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *pResult); +static void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *pResult); +static void resetMergeResultBuf(SQueryRuntimeEnv* pRuntimeEnv, SQLFunctionCtx *pCtx, SResultRow *pRow); static bool functionNeedToExecute(SQueryRuntimeEnv *pRuntimeEnv, SQLFunctionCtx *pCtx, int32_t functionId); static void setExecParams(SQuery *pQuery, SQLFunctionCtx *pCtx, void* inputData, TSKEY *tsCol, SDataBlockInfo* pBlockInfo, @@ -255,7 +255,7 @@ int64_t getNumOfResult(SQueryRuntimeEnv *pRuntimeEnv) { continue; } - SResultInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); + SResultRowCellInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); if (pResInfo != NULL && maxOutput < pResInfo->numOfRes) { maxOutput = pResInfo->numOfRes; } @@ -272,7 +272,7 @@ void updateNumOfResult(SQueryRuntimeEnv *pRuntimeEnv, int32_t numOfRes) { SQuery *pQuery = pRuntimeEnv->pQuery; for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { - SResultInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); + SResultRowCellInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); int16_t functionId = pRuntimeEnv->pCtx[j].functionId; if (functionId == TSDB_FUNC_TS || functionId == TSDB_FUNC_TAG || functionId == TSDB_FUNC_TAGPRJ || @@ -447,11 +447,11 @@ static bool hasNullValue(SColIndex* pColIndex, SDataStatis *pStatis, SDataStatis return true; } -static SWindowResult *doSetTimeWindowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowResInfo, char *pData, - int16_t bytes, bool masterscan) { - SQuery *pQuery = pRuntimeEnv->pQuery; - - int32_t *p1 = (int32_t *) taosHashGet(pWindowResInfo->hashList, pData, bytes); +static SResultRow *doPrepareResultRowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowResInfo, char *pData, + int16_t bytes, bool masterscan, uint64_t uid) { + SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, pData, bytes, uid); + int32_t *p1 = + (int32_t *)taosHashGet(pRuntimeEnv->pResultRowHashTable, pRuntimeEnv->keyBuf, GET_RES_WINDOW_KEY_LEN(bytes)); if (p1 != NULL) { pWindowResInfo->curIndex = *p1; } else { @@ -461,41 +461,37 @@ static SWindowResult *doSetTimeWindowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SWin // more than the capacity, reallocate the resources if (pWindowResInfo->size >= pWindowResInfo->capacity) { - int64_t newCap = 0; + int64_t newCapacity = 0; if (pWindowResInfo->capacity > 10000) { - newCap = (int64_t)(pWindowResInfo->capacity * 1.25); + newCapacity = (int64_t)(pWindowResInfo->capacity * 1.25); } else { - newCap = (int64_t)(pWindowResInfo->capacity * 1.5); + newCapacity = (int64_t)(pWindowResInfo->capacity * 1.5); } - char *t = realloc(pWindowResInfo->pResult, (size_t)(newCap * sizeof(SWindowResult))); - pRuntimeEnv->summary.internalSupSize += (newCap - pWindowResInfo->capacity) * sizeof(SWindowResult); - pRuntimeEnv->summary.numOfTimeWindows += (newCap - pWindowResInfo->capacity); - + char *t = realloc(pWindowResInfo->pResult, (size_t)(newCapacity * POINTER_BYTES)); if (t == NULL) { longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_OUT_OF_MEMORY); } - pWindowResInfo->pResult = (SWindowResult *)t; - - int32_t inc = (int32_t)newCap - pWindowResInfo->capacity; - memset(&pWindowResInfo->pResult[pWindowResInfo->capacity], 0, sizeof(SWindowResult) * inc); + pWindowResInfo->pResult = (SResultRow **)t; - pRuntimeEnv->summary.internalSupSize += (pQuery->numOfOutput * sizeof(SResultInfo) + pRuntimeEnv->interBufSize) * inc; + int32_t inc = (int32_t)newCapacity - pWindowResInfo->capacity; + memset(&pWindowResInfo->pResult[pWindowResInfo->capacity], 0, POINTER_BYTES * inc); - for (int32_t i = pWindowResInfo->capacity; i < newCap; ++i) { - int32_t ret = createQueryResultInfo(pQuery, &pWindowResInfo->pResult[i], pRuntimeEnv->stableQuery, pRuntimeEnv->interBufSize); - if (ret != TSDB_CODE_SUCCESS) { - longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_OUT_OF_MEMORY); - } - } + pWindowResInfo->capacity = (int32_t)newCapacity; + } - pWindowResInfo->capacity = (int32_t)newCap; + SResultRow *pResult = getNewResultRow(pRuntimeEnv->pool); + pWindowResInfo->pResult[pWindowResInfo->size] = pResult; + int32_t ret = initResultRow(pResult); + if (ret != TSDB_CODE_SUCCESS) { + longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_OUT_OF_MEMORY); } // add a new result set for a new group pWindowResInfo->curIndex = pWindowResInfo->size++; - taosHashPut(pWindowResInfo->hashList, pData, bytes, (char *)&pWindowResInfo->curIndex, sizeof(int32_t)); + taosHashPut(pRuntimeEnv->pResultRowHashTable, pRuntimeEnv->keyBuf, GET_RES_WINDOW_KEY_LEN(bytes), + (char *)&pWindowResInfo->curIndex, sizeof(int32_t)); } // too many time window in query @@ -503,7 +499,7 @@ static SWindowResult *doSetTimeWindowFromKey(SQueryRuntimeEnv *pRuntimeEnv, SWin longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_TOO_MANY_TIMEWINDOW); } - return getWindowResult(pWindowResInfo, pWindowResInfo->curIndex); + return getResultRow(pWindowResInfo, pWindowResInfo->curIndex); } // get the correct time window according to the handled timestamp @@ -519,7 +515,7 @@ static STimeWindow getActiveTimeWindow(SWindowResInfo *pWindowResInfo, int64_t t } } else { int32_t slot = curTimeWindowIndex(pWindowResInfo); - SWindowResult* pWindowRes = getWindowResult(pWindowResInfo, slot); + SResultRow* pWindowRes = getResultRow(pWindowResInfo, slot); w = pWindowRes->win; } @@ -555,7 +551,7 @@ static STimeWindow getActiveTimeWindow(SWindowResInfo *pWindowResInfo, int64_t t return w; } -static int32_t addNewWindowResultBuf(SWindowResult *pWindowRes, SDiskbasedResultBuf *pResultBuf, int32_t sid, +static int32_t addNewWindowResultBuf(SResultRow *pWindowRes, SDiskbasedResultBuf *pResultBuf, int32_t tid, int32_t numOfRowsPerPage) { if (pWindowRes->pageId != -1) { return 0; @@ -565,10 +561,10 @@ static int32_t addNewWindowResultBuf(SWindowResult *pWindowRes, SDiskbasedResult // in the first scan, new space needed for results int32_t pageId = -1; - SIDList list = getDataBufPagesIdList(pResultBuf, sid); + SIDList list = getDataBufPagesIdList(pResultBuf, tid); if (taosArrayGetSize(list) == 0) { - pData = getNewDataBuf(pResultBuf, sid, &pageId); + pData = getNewDataBuf(pResultBuf, tid, &pageId); } else { SPageInfo* pi = getLastPageInfo(list); pData = getResBufPage(pResultBuf, pi->pageId); @@ -578,7 +574,7 @@ static int32_t addNewWindowResultBuf(SWindowResult *pWindowRes, SDiskbasedResult // release current page first, and prepare the next one releaseResBufPageInfo(pResultBuf, pi); - pData = getNewDataBuf(pResultBuf, sid, &pageId); + pData = getNewDataBuf(pResultBuf, tid, &pageId); if (pData != NULL) { assert(pData->num == 0); // number of elements must be 0 for new allocated buffer } @@ -600,14 +596,13 @@ static int32_t addNewWindowResultBuf(SWindowResult *pWindowRes, SDiskbasedResult return 0; } -static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowResInfo, int32_t sid, +static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowResInfo, SDataBlockInfo* pBockInfo, STimeWindow *win, bool masterscan, bool* newWind) { assert(win->skey <= win->ekey); SDiskbasedResultBuf *pResultBuf = pRuntimeEnv->pResultBuf; - SWindowResult *pWindowRes = doSetTimeWindowFromKey(pRuntimeEnv, pWindowResInfo, (char *)&win->skey, - TSDB_KEYSIZE, masterscan); - if (pWindowRes == NULL) { + SResultRow *pResultRow = doPrepareResultRowFromKey(pRuntimeEnv, pWindowResInfo, (char *)&win->skey, TSDB_KEYSIZE, masterscan, pBockInfo->uid); + if (pResultRow == NULL) { *newWind = false; return masterscan? -1:0; @@ -616,23 +611,22 @@ static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowRes *newWind = true; // not assign result buffer yet, add new result buffer - if (pWindowRes->pageId == -1) { - int32_t ret = addNewWindowResultBuf(pWindowRes, pResultBuf, sid, pRuntimeEnv->numOfRowsPerPage); + if (pResultRow->pageId == -1) { + int32_t ret = addNewWindowResultBuf(pResultRow, pResultBuf, pBockInfo->tid, pRuntimeEnv->numOfRowsPerPage); if (ret != TSDB_CODE_SUCCESS) { return -1; } } // set time window for current result - pWindowRes->win = (*win); - - setWindowResOutputBufInitCtx(pRuntimeEnv, pWindowRes); + pResultRow->win = (*win); + setWindowResOutputBufInitCtx(pRuntimeEnv, pResultRow); return TSDB_CODE_SUCCESS; } static bool getTimeWindowResStatus(SWindowResInfo *pWindowResInfo, int32_t slot) { assert(slot >= 0 && slot < pWindowResInfo->size); - return pWindowResInfo->pResult[slot].closed; + return pWindowResInfo->pResult[slot]->closed; } static FORCE_INLINE int32_t getForwardStepsInBlock(int32_t numOfRows, __block_search_fn_t searchFn, TSKEY ekey, int16_t pos, @@ -691,7 +685,7 @@ static int32_t doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKe int64_t skey = TSKEY_INITIAL_VAL; for (i = 0; i < pWindowResInfo->size; ++i) { - SWindowResult *pResult = &pWindowResInfo->pResult[i]; + SResultRow *pResult = pWindowResInfo->pResult[i]; if (pResult->closed) { numOfClosed += 1; continue; @@ -715,7 +709,7 @@ static int32_t doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKe pWindowResInfo->curIndex = i; } - pWindowResInfo->prevSKey = pWindowResInfo->pResult[pWindowResInfo->curIndex].win.skey; + 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) { @@ -1024,8 +1018,7 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * bool hasTimeWindow = false; STimeWindow win = getActiveTimeWindow(pWindowResInfo, ts, pQuery); - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo->tid, &win, masterScan, &hasTimeWindow) != - TSDB_CODE_SUCCESS) { + if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { taosTFree(sasArray); return; } @@ -1053,8 +1046,7 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * // null data, failed to allocate more memory buffer hasTimeWindow = false; - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo->tid, &nextWin, masterScan, - &hasTimeWindow) != TSDB_CODE_SUCCESS) { + if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &nextWin, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { break; } @@ -1095,7 +1087,7 @@ static void blockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis * taosTFree(sasArray); } -static int32_t setGroupResultOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, char *pData, int16_t type, int16_t bytes) { +static int32_t setGroupResultOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, char *pData, int16_t type, int16_t bytes, int32_t groupIndex) { if (isNull(pData, type)) { // ignore the null value return -1; } @@ -1112,13 +1104,14 @@ static int32_t setGroupResultOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, char *pDat len = varDataLen(pData); } else if (type == TSDB_DATA_TYPE_FLOAT || type == TSDB_DATA_TYPE_DOUBLE) { SQInfo* pQInfo = GET_QINFO_ADDR(pRuntimeEnv); - qError("QInfo:%p group by not supported on double/float/binary/nchar columns, abort", pQInfo); + qError("QInfo:%p group by not supported on double/float columns, abort", pQInfo); longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_APP_ERROR); } - SWindowResult *pWindowRes = doSetTimeWindowFromKey(pRuntimeEnv, &pRuntimeEnv->windowResInfo, d, len, true); - if (pWindowRes == NULL) { + uint64_t uid = groupIndex; // uid is always set to be 0. + SResultRow *pResultRow = doPrepareResultRowFromKey(pRuntimeEnv, &pRuntimeEnv->windowResInfo, d, len, true, uid); + if (pResultRow == NULL) { return -1; } @@ -1132,23 +1125,21 @@ static int32_t setGroupResultOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, char *pDat } if (type == TSDB_DATA_TYPE_BINARY || type == TSDB_DATA_TYPE_NCHAR) { - pWindowRes->key = malloc(varDataTLen(pData)); - varDataCopy(pWindowRes->key, pData); + pResultRow->key = malloc(varDataTLen(pData)); + varDataCopy(pResultRow->key, pData); } else { - pWindowRes->win.skey = v; - pWindowRes->win.ekey = v; + pResultRow->win.skey = v; + pResultRow->win.ekey = v; } - assert(pRuntimeEnv->windowResInfo.interval == 0); - - if (pWindowRes->pageId == -1) { - int32_t ret = addNewWindowResultBuf(pWindowRes, pResultBuf, GROUPRESULTID, pRuntimeEnv->numOfRowsPerPage); + if (pResultRow->pageId == -1) { + int32_t ret = addNewWindowResultBuf(pResultRow, pResultBuf, GROUPRESULTID, pRuntimeEnv->numOfRowsPerPage); if (ret != 0) { return -1; } } - setWindowResOutputBuf(pRuntimeEnv, pWindowRes); + setResultOutputBuf(pRuntimeEnv, pResultRow); initCtxOutputBuf(pRuntimeEnv); return TSDB_CODE_SUCCESS; } @@ -1231,7 +1222,7 @@ static int32_t doTSJoinFilter(SQueryRuntimeEnv *pRuntimeEnv, int32_t offset) { } static bool functionNeedToExecute(SQueryRuntimeEnv *pRuntimeEnv, SQLFunctionCtx *pCtx, int32_t functionId) { - SResultInfo *pResInfo = GET_RES_INFO(pCtx); + SResultRowCellInfo *pResInfo = GET_RES_INFO(pCtx); SQuery* pQuery = pRuntimeEnv->pQuery; // in case of timestamp column, always generated results. @@ -1335,7 +1326,7 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS STimeWindow win = getActiveTimeWindow(pWindowResInfo, ts, pQuery); bool hasTimeWindow = false; - int32_t ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo->tid, &win, masterScan, &hasTimeWindow); + int32_t ret = setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &win, masterScan, &hasTimeWindow); if (ret != TSDB_CODE_SUCCESS) { // null data, too many state code continue; } @@ -1363,7 +1354,7 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS // null data, failed to allocate more memory buffer hasTimeWindow = false; - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo->tid, &nextWin, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { + if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pDataBlockInfo, &nextWin, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { break; } @@ -1379,7 +1370,7 @@ static void rowwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SDataStatis *pS if (groupbyColumnValue) { char *val = groupbyColumnData + bytes * offset; - int32_t ret = setGroupResultOutputBuf(pRuntimeEnv, val, type, bytes); + int32_t ret = setGroupResultOutputBuf(pRuntimeEnv, val, type, bytes, item->groupIndex); if (ret != TSDB_CODE_SUCCESS) { // null data, too many state code continue; } @@ -1518,7 +1509,8 @@ void setExecParams(SQuery *pQuery, SQLFunctionCtx *pCtx, void* inputData, TSKEY * top/bottom values emerge, so does diff function */ if (functionId == TSDB_FUNC_TWA) { - STwaInfo *pTWAInfo = GET_RES_INFO(pCtx)->interResultBuf; + SResultRowCellInfo* pInfo = GET_RES_INFO(pCtx); + STwaInfo *pTWAInfo = (STwaInfo*) GET_ROWCELL_INTERBUF(pInfo); pTWAInfo->SKey = pQuery->window.skey; pTWAInfo->EKey = pQuery->window.ekey; } @@ -1532,7 +1524,9 @@ void setExecParams(SQuery *pQuery, SQLFunctionCtx *pCtx, void* inputData, TSKEY pCtx->preAggVals.statis.max = pBlockInfo->window.ekey; } } else if (functionId == TSDB_FUNC_INTERP) { - SInterpInfoDetail *pInterpInfo = GET_RES_INFO(pCtx)->interResultBuf; + SResultRowCellInfo* pInfo = GET_RES_INFO(pCtx); + + SInterpInfoDetail *pInterpInfo = (SInterpInfoDetail *)GET_ROWCELL_INTERBUF(pInfo); pInterpInfo->type = (int8_t)pQuery->fillType; pInterpInfo->ts = pQuery->window.skey; pInterpInfo->primaryCol = (colId == PRIMARYKEY_TIMESTAMP_COL_INDEX); @@ -1609,26 +1603,15 @@ static int32_t setCtxTagColumnInfo(SQueryRuntimeEnv *pRuntimeEnv, SQLFunctionCtx return TSDB_CODE_SUCCESS; } -static FORCE_INLINE void setWindowResultInfo(SResultInfo *pResultInfo, SQuery *pQuery, bool isStableQuery, char* buf) { - char* p = buf; - for (int32_t i = 0; i < pQuery->numOfOutput; ++i) { - int32_t size = pQuery->pSelectExpr[i].interBytes; - setResultInfoBuf(&pResultInfo[i], size, isStableQuery, p); - - p += size; - } -} - static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int16_t order) { qDebug("QInfo:%p setup runtime env", GET_QINFO_ADDR(pRuntimeEnv)); SQuery *pQuery = pRuntimeEnv->pQuery; - size_t size = pRuntimeEnv->interBufSize + pQuery->numOfOutput * sizeof(SResultInfo); - - pRuntimeEnv->resultInfo = calloc(1, size); pRuntimeEnv->pCtx = (SQLFunctionCtx *)calloc(pQuery->numOfOutput, sizeof(SQLFunctionCtx)); + pRuntimeEnv->rowCellInfoOffset = calloc(pQuery->numOfOutput, sizeof(int32_t)); + pRuntimeEnv->pResultRow = getNewResultRow(pRuntimeEnv->pool);//calloc(1, sizeof(SResultRow)); - if (pRuntimeEnv->resultInfo == NULL || pRuntimeEnv->pCtx == NULL) { + if (pRuntimeEnv->pResultRow == NULL || pRuntimeEnv->pCtx == NULL || pRuntimeEnv->rowCellInfoOffset == NULL) { goto _clean; } @@ -1665,7 +1648,6 @@ static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int16_t order pCtx->inputType = pQuery->colList[index].type; } - assert(isValidDataType(pCtx->inputType)); pCtx->ptsOutputBuf = NULL; @@ -1674,6 +1656,8 @@ static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int16_t order pCtx->order = pQuery->order.order; pCtx->functionId = pSqlFuncMsg->functionId; + pCtx->stableQuery = pRuntimeEnv->stableQuery; + pCtx->interBufBytes = pQuery->pSelectExpr[i].interBytes; pCtx->numOfParams = pSqlFuncMsg->numOfParams; for (int32_t j = 0; j < pCtx->numOfParams; ++j) { @@ -1703,16 +1687,14 @@ static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int16_t order if (i > 0) { pRuntimeEnv->offset[i] = pRuntimeEnv->offset[i - 1] + pRuntimeEnv->pCtx[i - 1].outputBytes; + pRuntimeEnv->rowCellInfoOffset[i] = pRuntimeEnv->rowCellInfoOffset[i - 1] + sizeof(SResultRowCellInfo) + pQuery->pSelectExpr[i - 1].interBytes; } - } - char* buf = (char*) pRuntimeEnv->resultInfo + sizeof(SResultInfo) * pQuery->numOfOutput; - - // set the intermediate result output buffer - setWindowResultInfo(pRuntimeEnv->resultInfo, pQuery, pRuntimeEnv->stableQuery, buf); + } // if it is group by normal column, do not set output buffer, the output buffer is pResult - if (!pRuntimeEnv->groupbyNormalCol && !pRuntimeEnv->stableQuery) { + // fixed output query/multi-output query for normal table + if (!pRuntimeEnv->groupbyNormalCol && !pRuntimeEnv->stableQuery && !QUERY_IS_INTERVAL_QUERY(pRuntimeEnv->pQuery)) { resetCtxOutputBuf(pRuntimeEnv); } @@ -1724,7 +1706,6 @@ static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int16_t order return TSDB_CODE_SUCCESS; _clean: - taosTFree(pRuntimeEnv->resultInfo); taosTFree(pRuntimeEnv->pCtx); return TSDB_CODE_QRY_OUT_OF_MEMORY; @@ -1753,17 +1734,22 @@ static void teardownQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv) { taosTFree(pCtx->tagInfo.pTagCtxList); } - taosTFree(pRuntimeEnv->resultInfo); taosTFree(pRuntimeEnv->pCtx); } - pRuntimeEnv->pFillInfo = taosDestoryFillInfo(pRuntimeEnv->pFillInfo); + pRuntimeEnv->pFillInfo = taosDestroyFillInfo(pRuntimeEnv->pFillInfo); destroyResultBuf(pRuntimeEnv->pResultBuf); tsdbCleanupQueryHandle(pRuntimeEnv->pQueryHandle); tsdbCleanupQueryHandle(pRuntimeEnv->pSecQueryHandle); pRuntimeEnv->pTSBuf = tsBufDestroy(pRuntimeEnv->pTSBuf); + taosTFree(pRuntimeEnv->keyBuf); + + taosHashCleanup(pRuntimeEnv->pResultRowHashTable); + pRuntimeEnv->pResultRowHashTable = NULL; + + pRuntimeEnv->pool = destroyResultRowPool(pRuntimeEnv->pool); } #define IS_QUERY_KILLED(_q) ((_q)->code == TSDB_CODE_TSC_QUERY_CANCELLED) @@ -2262,7 +2248,7 @@ int32_t loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo * pW TSKEY k = QUERY_IS_ASC_QUERY(pQuery)? pBlockInfo->window.skey:pBlockInfo->window.ekey; STimeWindow win = getActiveTimeWindow(pWindowResInfo, k, pQuery); - if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pBlockInfo->tid, &win, masterScan, &hasTimeWindow) != + if (setWindowOutputBufByKey(pRuntimeEnv, pWindowResInfo, pBlockInfo, &win, masterScan, &hasTimeWindow) != TSDB_CODE_SUCCESS) { // todo handle error in set result for timewindow } @@ -2648,7 +2634,7 @@ void setTagVal(SQueryRuntimeEnv *pRuntimeEnv, void *pTable, void *tsdb) { } } -static void doMerge(SQueryRuntimeEnv *pRuntimeEnv, int64_t timestamp, SWindowResult *pWindowRes, bool mergeFlag) { +static void doMerge(SQueryRuntimeEnv *pRuntimeEnv, int64_t timestamp, SResultRow *pWindowRes, bool mergeFlag) { SQuery * pQuery = pRuntimeEnv->pQuery; SQLFunctionCtx *pCtx = pRuntimeEnv->pCtx; @@ -2822,14 +2808,14 @@ int32_t tableResultComparFn(const void *pLeft, const void *pRight, void *param) } SWindowResInfo *pWindowResInfo1 = &supporter->pTableQueryInfo[left]->windowResInfo; - SWindowResult * pWindowRes1 = getWindowResult(pWindowResInfo1, leftPos); + 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; - SWindowResult * pWindowRes2 = getWindowResult(pWindowResInfo2, rightPos); + SResultRow * pWindowRes2 = getResultRow(pWindowResInfo2, rightPos); tFilePage *page2 = getResBufPage(pRuntimeEnv->pResultBuf, pWindowRes2->pageId); char *b2 = getPosInResultPage(pRuntimeEnv, PRIMARYKEY_TIMESTAMP_COL_INDEX, pWindowRes2, page2); @@ -2950,7 +2936,9 @@ void copyResToQueryResultBuf(SQInfo *pQInfo, SQuery *pQuery) { pQuery->rec.rows += offset; } -int64_t getNumOfResultWindowRes(SQuery *pQuery, SWindowResult *pWindowRes) { +int64_t getNumOfResultWindowRes(SQueryRuntimeEnv* pRuntimeEnv, SResultRow *pResultRow) { + SQuery* pQuery = pRuntimeEnv->pQuery; + for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { int32_t functionId = pQuery->pSelectExpr[j].base.functionId; @@ -2962,7 +2950,7 @@ int64_t getNumOfResultWindowRes(SQuery *pQuery, SWindowResult *pWindowRes) { continue; } - SResultInfo *pResultInfo = &pWindowRes->resultInfo[j]; + SResultRowCellInfo *pResultInfo = getResultCell(pRuntimeEnv, pResultRow, j); assert(pResultInfo != NULL); if (pResultInfo->numOfRes > 0) { @@ -3031,18 +3019,8 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) { SLoserTreeInfo *pTree = NULL; tLoserTreeCreate(&pTree, numOfTables, &cs, tableResultComparFn); - SResultInfo *pResultInfo = calloc(pQuery->numOfOutput, sizeof(SResultInfo)); - if (pResultInfo == NULL) { - longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_OUT_OF_MEMORY); - } - - char* buf = calloc(1, pRuntimeEnv->interBufSize); - if (buf == NULL) { - longjmp(pRuntimeEnv->env, TSDB_CODE_QRY_OUT_OF_MEMORY); - } - - setWindowResultInfo(pResultInfo, pQuery, pRuntimeEnv->stableQuery, buf); - resetMergeResultBuf(pQuery, pRuntimeEnv->pCtx, pResultInfo); + SResultRow* pRow = getNewResultRow(pRuntimeEnv->pool); + resetMergeResultBuf(pRuntimeEnv, pRuntimeEnv->pCtx, pRow); pQInfo->groupResInfo.groupId = getGroupResultId(pQInfo->groupIndex); @@ -3057,23 +3035,20 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) { taosTFree(pTableList); taosTFree(posList); taosTFree(pTree); - taosTFree(pResultInfo); - taosTFree(buf); - longjmp(pRuntimeEnv->env, TSDB_CODE_TSC_QUERY_CANCELLED); } int32_t pos = pTree->pNode[0].index; SWindowResInfo *pWindowResInfo = &pTableList[pos]->windowResInfo; - SWindowResult *pWindowRes = getWindowResult(pWindowResInfo, cs.position[pos]); + SResultRow *pWindowRes = getResultRow(pWindowResInfo, cs.position[pos]); tFilePage *page = getResBufPage(pRuntimeEnv->pResultBuf, pWindowRes->pageId); char *b = getPosInResultPage(pRuntimeEnv, PRIMARYKEY_TIMESTAMP_COL_INDEX, pWindowRes, page); TSKEY ts = GET_INT64_VAL(b); assert(ts == pWindowRes->win.skey); - int64_t num = getNumOfResultWindowRes(pQuery, pWindowRes); + int64_t num = getNumOfResultWindowRes(pRuntimeEnv, pWindowRes); if (num <= 0) { cs.position[pos] += 1; @@ -3094,7 +3069,7 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) { return -1; } - resetMergeResultBuf(pQuery, pRuntimeEnv->pCtx, pResultInfo); + resetMergeResultBuf(pRuntimeEnv, pRuntimeEnv->pCtx, pRow); } doMerge(pRuntimeEnv, ts, pWindowRes, false); @@ -3116,7 +3091,7 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) { } } else { // current page is not needed anymore - SWindowResult *pNextWindowRes = getWindowResult(pWindowResInfo, cs.position[pos]); + SResultRow *pNextWindowRes = getResultRow(pWindowResInfo, cs.position[pos]); if (pNextWindowRes->pageId != currentPageId) { releaseResBufPage(pRuntimeEnv->pResultBuf, page); } @@ -3133,8 +3108,6 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) { taosTFree(pTree); taosTFree(pTableList); taosTFree(posList); - taosTFree(pResultInfo); - return -1; } } @@ -3151,8 +3124,8 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) { taosTFree(posList); taosTFree(pTree); - taosTFree(pResultInfo); - taosTFree(buf); +// taosTFree(pResultInfo); +// taosTFree(buf); return pQInfo->groupResInfo.numOfDataPages; } @@ -3195,12 +3168,14 @@ int32_t flushFromResultBuf(SQueryRuntimeEnv* pRuntimeEnv, SGroupResInfo* pGroupR return TSDB_CODE_SUCCESS; } -void resetMergeResultBuf(SQuery *pQuery, SQLFunctionCtx *pCtx, SResultInfo *pResultInfo) { +void resetMergeResultBuf(SQueryRuntimeEnv* pRuntimeEnv, SQLFunctionCtx *pCtx, SResultRow *pRow) { + SQuery* pQuery = pRuntimeEnv->pQuery; + for (int32_t k = 0; k < pQuery->numOfOutput; ++k) { pCtx[k].aOutputBuf = pQuery->sdata[k]->data - pCtx[k].outputBytes; pCtx[k].size = 1; pCtx[k].startOffset = 0; - pCtx[k].resultInfo = &pResultInfo[k]; + pCtx[k].resultInfo = getResultCell(pRuntimeEnv, pRow, k); pQuery->sdata[k]->num = 0; } @@ -3237,8 +3212,8 @@ static void updateTableQueryInfoForReverseScan(SQuery *pQuery, STableQueryInfo * pTableQueryInfo->windowResInfo.curIndex = pTableQueryInfo->windowResInfo.size - 1; } -static void disableFuncInReverseScanImpl(SQInfo* pQInfo, SWindowResInfo *pWindowResInfo, int32_t order) { - SQuery* pQuery = pQInfo->runtimeEnv.pQuery; +static void disableFuncInReverseScanImpl(SQueryRuntimeEnv* pRuntimeEnv, SWindowResInfo *pWindowResInfo, int32_t order) { + SQuery* pQuery = pRuntimeEnv->pQuery; for (int32_t i = 0; i < pWindowResInfo->size; ++i) { bool closed = getTimeWindowResStatus(pWindowResInfo, i); @@ -3246,17 +3221,18 @@ static void disableFuncInReverseScanImpl(SQInfo* pQInfo, SWindowResInfo *pWindow continue; } - SWindowResult *buf = getWindowResult(pWindowResInfo, i); + SResultRow *pRow = getResultRow(pWindowResInfo, i); // open/close the specified query for each group result for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { int32_t functId = pQuery->pSelectExpr[j].base.functionId; + SResultRowCellInfo* pInfo = getResultCell(pRuntimeEnv, pRow, j); if (((functId == TSDB_FUNC_FIRST || functId == TSDB_FUNC_FIRST_DST) && order == TSDB_ORDER_ASC) || ((functId == TSDB_FUNC_LAST || functId == TSDB_FUNC_LAST_DST) && order == TSDB_ORDER_DESC)) { - buf->resultInfo[j].complete = false; + pInfo->complete = false; } else if (functId != TSDB_FUNC_TS && functId != TSDB_FUNC_TAG) { - buf->resultInfo[j].complete = true; + pInfo->complete = true; } } } @@ -3270,7 +3246,7 @@ void disableFuncInReverseScan(SQInfo *pQInfo) { // group by normal columns and interval query on normal table SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; if (pRuntimeEnv->groupbyNormalCol || QUERY_IS_INTERVAL_QUERY(pQuery)) { - disableFuncInReverseScanImpl(pQInfo, pWindowResInfo, order); + disableFuncInReverseScanImpl(pRuntimeEnv, pWindowResInfo, order); } else { // for simple result of table query, for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { // todo refactor int32_t functId = pQuery->pSelectExpr[j].base.functionId; @@ -3320,27 +3296,16 @@ void switchCtxOrder(SQueryRuntimeEnv *pRuntimeEnv) { } } -int32_t createQueryResultInfo(SQuery *pQuery, SWindowResult *pResultRow, bool isSTableQuery, size_t interBufSize) { - int32_t numOfCols = pQuery->numOfOutput; - - size_t size = numOfCols * sizeof(SResultInfo) + interBufSize; - pResultRow->resultInfo = calloc(1, size); - if (pResultRow->resultInfo == NULL) { - return TSDB_CODE_QRY_OUT_OF_MEMORY; - } - +int32_t initResultRow(SResultRow *pResultRow) { + pResultRow->pCellInfo = (SResultRowCellInfo*)((char*)pResultRow + sizeof(SResultRow)); pResultRow->pageId = -1; pResultRow->rowId = -1; - - char* buf = (char*) pResultRow->resultInfo + numOfCols * sizeof(SResultInfo); - - // set the intermediate result output buffer - setWindowResultInfo(pResultRow->resultInfo, pQuery, isSTableQuery, buf); return TSDB_CODE_SUCCESS; } void resetCtxOutputBuf(SQueryRuntimeEnv *pRuntimeEnv) { SQuery *pQuery = pRuntimeEnv->pQuery; + SResultRow* pRow = pRuntimeEnv->pResultRow; for (int32_t i = 0; i < pQuery->numOfOutput; ++i) { SQLFunctionCtx *pCtx = &pRuntimeEnv->pCtx[i]; @@ -3350,8 +3315,9 @@ void resetCtxOutputBuf(SQueryRuntimeEnv *pRuntimeEnv) { * set the output buffer information and intermediate buffer * not all queries require the interResultBuf, such as COUNT/TAGPRJ/PRJ/TAG etc. */ - RESET_RESULT_INFO(&pRuntimeEnv->resultInfo[i]); - pCtx->resultInfo = &pRuntimeEnv->resultInfo[i]; + SResultRowCellInfo* pCellInfo = getResultCell(pRuntimeEnv, pRow, i); + RESET_RESULT_INFO(pCellInfo); + pCtx->resultInfo = pCellInfo; // set the timestamp output buffer for top/bottom/diff query int32_t functionId = pQuery->pSelectExpr[i].base.functionId; @@ -3400,7 +3366,7 @@ void initCtxOutputBuf(SQueryRuntimeEnv *pRuntimeEnv) { int32_t functionId = pQuery->pSelectExpr[j].base.functionId; pRuntimeEnv->pCtx[j].currentStage = 0; - SResultInfo* pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); + SResultRowCellInfo* pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); if (pResInfo->initialized) { continue; } @@ -3469,12 +3435,12 @@ bool needScanDataBlocksAgain(SQueryRuntimeEnv *pRuntimeEnv) { SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo; for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - SWindowResult *pResult = getWindowResult(pWindowResInfo, i); + SResultRow *pResult = getResultRow(pWindowResInfo, i); if (!pResult->closed) { continue; } - setWindowResOutputBuf(pRuntimeEnv, pResult); + setResultOutputBuf(pRuntimeEnv, pResult); for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { int16_t functId = pQuery->pSelectExpr[j].base.functionId; @@ -3483,7 +3449,7 @@ bool needScanDataBlocksAgain(SQueryRuntimeEnv *pRuntimeEnv) { } aAggs[functId].xNextStep(&pRuntimeEnv->pCtx[j]); - SResultInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); + SResultRowCellInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); toContinue |= (!pResInfo->complete); } @@ -3496,7 +3462,7 @@ bool needScanDataBlocksAgain(SQueryRuntimeEnv *pRuntimeEnv) { } aAggs[functId].xNextStep(&pRuntimeEnv->pCtx[j]); - SResultInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); + SResultRowCellInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]); toContinue |= (!pResInfo->complete); } @@ -3695,12 +3661,12 @@ void finalizeQueryResult(SQueryRuntimeEnv *pRuntimeEnv) { } for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - SWindowResult *buf = &pWindowResInfo->pResult[i]; + SResultRow *buf = pWindowResInfo->pResult[i]; if (!isWindowResClosed(pWindowResInfo, i)) { continue; } - setWindowResOutputBuf(pRuntimeEnv, buf); + setResultOutputBuf(pRuntimeEnv, buf); for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { aAggs[pQuery->pSelectExpr[j].base.functionId].xFinalize(&pRuntimeEnv->pCtx[j]); @@ -3747,7 +3713,7 @@ static STableQueryInfo *createTableQueryInfo(SQueryRuntimeEnv *pRuntimeEnv, void if (QUERY_IS_INTERVAL_QUERY(pQuery) || pRuntimeEnv->groupbyNormalCol) { int32_t initialSize = 16; int32_t initialThreshold = 100; - int32_t code = initWindowResInfo(&pTableQueryInfo->windowResInfo, pRuntimeEnv, initialSize, initialThreshold, TSDB_DATA_TYPE_INT); + int32_t code = initWindowResInfo(&pTableQueryInfo->windowResInfo, initialSize, initialThreshold, TSDB_DATA_TYPE_INT); if (code != TSDB_CODE_SUCCESS) { return NULL; } @@ -3762,6 +3728,7 @@ void destroyTableQueryInfoImpl(STableQueryInfo *pTableQueryInfo) { return; } + tVariantDestroy(&pTableQueryInfo->tag); cleanupTimeWindowInfo(&pTableQueryInfo->windowResInfo); } @@ -3786,9 +3753,10 @@ void setExecutionContext(SQInfo *pQInfo, int32_t groupIndex, TSKEY nextKey) { return; } - SWindowResult *pWindowRes = doSetTimeWindowFromKey(pRuntimeEnv, pWindowResInfo, (char *)&groupIndex, - sizeof(groupIndex), true); - if (pWindowRes == NULL) { + uint64_t uid = 0; // uid is always set to be 0 + SResultRow *pResultRow = doPrepareResultRowFromKey(pRuntimeEnv, pWindowResInfo, (char *)&groupIndex, + sizeof(groupIndex), true, uid); + if (pResultRow == NULL) { return; } @@ -3796,8 +3764,8 @@ void setExecutionContext(SQInfo *pQInfo, int32_t groupIndex, TSKEY nextKey) { * not assign result buffer yet, add new result buffer * all group belong to one result set, and each group result has different group id so set the id to be one */ - if (pWindowRes->pageId == -1) { - if (addNewWindowResultBuf(pWindowRes, pRuntimeEnv->pResultBuf, groupIndex, pRuntimeEnv->numOfRowsPerPage) != + if (pResultRow->pageId == -1) { + if (addNewWindowResultBuf(pResultRow, pRuntimeEnv->pResultBuf, groupIndex, pRuntimeEnv->numOfRowsPerPage) != TSDB_CODE_SUCCESS) { return; } @@ -3805,11 +3773,11 @@ void setExecutionContext(SQInfo *pQInfo, int32_t groupIndex, TSKEY nextKey) { // record the current active group id pRuntimeEnv->prevGroupId = groupIndex; - setWindowResOutputBuf(pRuntimeEnv, pWindowRes); + setResultOutputBuf(pRuntimeEnv, pResultRow); initCtxOutputBuf(pRuntimeEnv); } -void setWindowResOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult) { +void setResultOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *pResult) { SQuery *pQuery = pRuntimeEnv->pQuery; // Note: pResult->pos[i]->num == 0, there is only fixed number of results for each group @@ -3825,18 +3793,14 @@ void setWindowResOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult } /* - * set the output buffer information and intermediate buffer + * set the output buffer information and intermediate buffer, * not all queries require the interResultBuf, such as COUNT */ - pCtx->resultInfo = &pResult->resultInfo[i]; - - // set super table query flag - SResultInfo *pResInfo = GET_RES_INFO(pCtx); - pResInfo->superTableQ = pRuntimeEnv->stableQuery; + pCtx->resultInfo = getResultCell(pRuntimeEnv, pResult, i); } } -void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult) { +void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *pResult) { SQuery *pQuery = pRuntimeEnv->pQuery; // Note: pResult->pos[i]->num == 0, there is only fixed number of results for each group @@ -3845,7 +3809,7 @@ void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult * for (int32_t i = 0; i < pQuery->numOfOutput; ++i) { SQLFunctionCtx *pCtx = &pRuntimeEnv->pCtx[i]; - pCtx->resultInfo = &pResult->resultInfo[i]; + pCtx->resultInfo = getResultCell(pRuntimeEnv, pResult, i); if (pCtx->resultInfo->initialized && pCtx->resultInfo->complete) { continue; } @@ -3858,12 +3822,6 @@ void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult * pCtx->ptsOutputBuf = pRuntimeEnv->pCtx[0].aOutputBuf; } - /* - * set the output buffer information and intermediate buffer - * not all queries require the interResultBuf, such as COUNT - */ - pCtx->resultInfo->superTableQ = pRuntimeEnv->stableQuery; // set super table query flag - if (!pCtx->resultInfo->initialized) { aAggs[functionId].init(pCtx); } @@ -4004,7 +3962,7 @@ static int32_t doCopyToSData(SQInfo *pQInfo, SWindowResInfo *pResultInfo, int32_ qDebug("QInfo:%p start to copy data from windowResInfo to query buf", pQInfo); int32_t totalSet = numOfClosedTimeWindow(pResultInfo); - SWindowResult* result = pResultInfo->pResult; + SResultRow** result = pResultInfo->pResult; if (orderType == TSDB_ORDER_ASC) { startIdx = pQInfo->groupIndex; @@ -4017,13 +3975,13 @@ static int32_t doCopyToSData(SQInfo *pQInfo, SWindowResInfo *pResultInfo, int32_ SGroupResInfo* pGroupResInfo = &pQInfo->groupResInfo; for (int32_t i = startIdx; (i < totalSet) && (i >= 0); i += step) { - if (result[i].numOfRows == 0) { + if (result[i]->numOfRows == 0) { pQInfo->groupIndex += 1; pGroupResInfo->rowId = 0; continue; } - int32_t numOfRowsToCopy = result[i].numOfRows - pGroupResInfo->rowId; + int32_t numOfRowsToCopy = result[i]->numOfRows - pGroupResInfo->rowId; int32_t oldOffset = pGroupResInfo->rowId; /* @@ -4038,13 +3996,13 @@ static int32_t doCopyToSData(SQInfo *pQInfo, SWindowResInfo *pResultInfo, int32_ pQInfo->groupIndex += 1; } - tFilePage *page = getResBufPage(pRuntimeEnv->pResultBuf, result[i].pageId); + tFilePage *page = getResBufPage(pRuntimeEnv->pResultBuf, result[i]->pageId); for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { int32_t size = pRuntimeEnv->pCtx[j].outputBytes; char *out = pQuery->sdata[j]->data + numOfResult * size; - char *in = getPosInResultPage(pRuntimeEnv, j, &result[i], page); + char *in = getPosInResultPage(pRuntimeEnv, j, result[i], page); memcpy(out, in + oldOffset * size, size * numOfRowsToCopy); } @@ -4091,7 +4049,7 @@ static void updateWindowResNumOfRes(SQueryRuntimeEnv *pRuntimeEnv) { } for (int32_t i = 0; i < pRuntimeEnv->windowResInfo.size; ++i) { - SWindowResult *pResult = &pRuntimeEnv->windowResInfo.pResult[i]; + SResultRow *pResult = pRuntimeEnv->windowResInfo.pResult[i]; for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { int32_t functionId = pRuntimeEnv->pCtx[j].functionId; @@ -4099,7 +4057,8 @@ static void updateWindowResNumOfRes(SQueryRuntimeEnv *pRuntimeEnv) { continue; } - pResult->numOfRows = (uint16_t)(MAX(pResult->numOfRows, pResult->resultInfo[j].numOfRes)); + SResultRowCellInfo* pCell = getResultCell(pRuntimeEnv, pResult, j); + pResult->numOfRows = (uint16_t)(MAX(pResult->numOfRows, pCell->numOfRes)); } } } @@ -4245,16 +4204,24 @@ static void queryCostStatis(SQInfo *pQInfo) { SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv; SQueryCostInfo *pSummary = &pRuntimeEnv->summary; + uint64_t hashSize = taosHashGetMemSize(pQInfo->runtimeEnv.pResultRowHashTable); + hashSize += taosHashGetMemSize(pQInfo->tableqinfoGroupInfo.map); + pSummary->hashSize = hashSize; + // add the merge time pSummary->elapsedTime += pSummary->firstStageMergeTime; + SResultRowPool* p = pQInfo->runtimeEnv.pool; + pSummary->winInfoSize = getResultRowPoolMemSize(p); + pSummary->numOfTimeWindows = getNumOfAllocatedResultRows(p); + qDebug("QInfo:%p :cost summary: elapsed time:%"PRId64" us, first merge:%"PRId64" us, total blocks:%d, " "load block statis:%d, load data block:%d, total rows:%"PRId64 ", check rows:%"PRId64, pQInfo, pSummary->elapsedTime, pSummary->firstStageMergeTime, pSummary->totalBlocks, pSummary->loadBlockStatis, pSummary->loadBlocks, pSummary->totalRows, pSummary->totalCheckedRows); - qDebug("QInfo:%p :cost summary: internal size:%"PRId64"B, numOfWin:%"PRId64, pQInfo, pSummary->internalSupSize, - pSummary->numOfTimeWindows); + qDebug("QInfo:%p :cost summary: winResPool size:%.2f Kb, numOfWin:%"PRId64", tableInfoSize:%.2f Kb, hashTable:%.2f Kb", pQInfo, pSummary->winInfoSize/1024.0, + pSummary->numOfTimeWindows, pSummary->tableInfoSize/1024.0, pSummary->hashSize/1024.0); } static void updateOffsetVal(SQueryRuntimeEnv *pRuntimeEnv, SDataBlockInfo *pBlockInfo) { @@ -4612,7 +4579,7 @@ int32_t doInitQInfo(SQInfo *pQInfo, STSBuf *pTsBuf, void *tsdb, int32_t vgId, bo } } - code = initWindowResInfo(&pRuntimeEnv->windowResInfo, pRuntimeEnv, 8, threshold, type); + code = initWindowResInfo(&pRuntimeEnv->windowResInfo, 8, threshold, type); if (code != TSDB_CODE_SUCCESS) { return code; } @@ -4632,7 +4599,7 @@ int32_t doInitQInfo(SQInfo *pQInfo, STSBuf *pTsBuf, void *tsdb, int32_t vgId, bo type = TSDB_DATA_TYPE_TIMESTAMP; } - code = initWindowResInfo(&pRuntimeEnv->windowResInfo, pRuntimeEnv, numOfResultRows, 4096, type); + code = initWindowResInfo(&pRuntimeEnv->windowResInfo, numOfResultRows, 1024, type); if (code != TSDB_CODE_SUCCESS) { return code; } @@ -4659,7 +4626,7 @@ static void enableExecutionForNextTable(SQueryRuntimeEnv *pRuntimeEnv) { SQuery *pQuery = pRuntimeEnv->pQuery; for (int32_t i = 0; i < pQuery->numOfOutput; ++i) { - SResultInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[i]); + SResultRowCellInfo *pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[i]); if (pResInfo != NULL) { pResInfo->complete = false; } @@ -4882,7 +4849,7 @@ static void sequentialTableProcess(SQInfo *pQInfo) { size_t numOfGroups = GET_NUM_OF_TABLEGROUP(pQInfo); - if (isPointInterpoQuery(pQuery) || isFirstLastRowQuery(pQuery)) { + if (isPointInterpoQuery(pQuery)) { resetCtxOutputBuf(pRuntimeEnv); assert(pQuery->limit.offset == 0 && pQuery->limit.limit != 0); @@ -4912,11 +4879,7 @@ static void sequentialTableProcess(SQInfo *pQInfo) { pRuntimeEnv->pQueryHandle = NULL; } - if (isFirstLastRowQuery(pQuery)) { - assert(0); // last_row query switch to other routine to handle - } else { - pRuntimeEnv->pQueryHandle = tsdbQueryRowsInExternalWindow(pQInfo->tsdb, &cond, &gp, pQInfo); - } + pRuntimeEnv->pQueryHandle = tsdbQueryRowsInExternalWindow(pQInfo->tsdb, &cond, &gp, pQInfo); taosArrayDestroy(tx); taosArrayDestroy(g1); @@ -4930,10 +4893,6 @@ static void sequentialTableProcess(SQInfo *pQInfo) { assert(taosArrayGetSize(s) >= 1); setTagVal(pRuntimeEnv, taosArrayGetP(s, 0), pQInfo->tsdb); - if (isFirstLastRowQuery(pQuery)) { - assert(taosArrayGetSize(s) == 1); - } - taosArrayDestroy(s); // here we simply set the first table as current table @@ -5012,11 +4971,12 @@ static void sequentialTableProcess(SQInfo *pQInfo) { } 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 - SWindowResult *pResult = &pWindowResInfo->pResult[i]; + SResultRow *pResult = pWindowResInfo->pResult[i]; for (int32_t j = 0; j < pQuery->numOfOutput; ++j) { - pResult->numOfRows = (uint16_t)(MAX(pResult->numOfRows, pResult->resultInfo[j].numOfRes)); + SResultRowCellInfo* pCell = getResultCell(pRuntimeEnv, pResult, j); + pResult->numOfRows = (uint16_t)(MAX(pResult->numOfRows, pCell->numOfRes)); } } @@ -5287,7 +5247,6 @@ static void multiTableQueryProcess(SQInfo *pQInfo) { if (pQInfo->code != TSDB_CODE_SUCCESS || IS_QUERY_KILLED(pQInfo)) { qDebug("QInfo:%p query killed or error occurred, code:%s, abort", pQInfo, tstrerror(pQInfo->code)); //TODO finalizeQueryResult may cause SEGSEV, since the memory may not allocated yet, add a cleanup function instead -// finalizeQueryResult(pRuntimeEnv); // clean up allocated resource during query longjmp(pRuntimeEnv->env, TSDB_CODE_TSC_QUERY_CANCELLED); } @@ -6278,8 +6237,6 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou calResultBufSize(pQuery); for (int32_t col = 0; col < pQuery->numOfOutput; ++col) { - assert(pExprs[col].interBytes >= pExprs[col].bytes); - // allocate additional memory for interResults that are usually larger then final results size_t size = (size_t)((pQuery->rec.capacity + 1) * pExprs[col].bytes + pExprs[col].interBytes + sizeof(tFilePage)); pQuery->sdata[col] = (tFilePage *)calloc(1, size); @@ -6305,12 +6262,18 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou pQInfo->tableqinfoGroupInfo.pGroupList = taosArrayInit(numOfGroups, POINTER_BYTES); pQInfo->tableqinfoGroupInfo.numOfTables = pTableGroupInfo->numOfTables; pQInfo->tableqinfoGroupInfo.map = taosHashInit(pTableGroupInfo->numOfTables, - taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, false); + 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->pBuf = calloc(pTableGroupInfo->numOfTables, sizeof(STableQueryInfo)); if (pQInfo->pBuf == NULL) { goto _cleanup; @@ -6345,7 +6308,7 @@ 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); + void* buf = (char*) pQInfo->pBuf + index * sizeof(STableQueryInfo); window.skey = info->lastKey; STableQueryInfo* item = createTableQueryInfo(&pQInfo->runtimeEnv, info->pTable, window, buf); @@ -6410,12 +6373,13 @@ static int32_t initQInfo(SQueryTableMsg *pQueryMsg, void *tsdb, int32_t vgId, SQ SQuery *pQuery = pQInfo->runtimeEnv.pQuery; STSBuf *pTSBuf = NULL; - if (pQueryMsg->tsLen > 0) { // open new file to save the result + if (pQueryMsg->tsLen > 0) { // open new file to save the result char *tsBlock = (char *) pQueryMsg + pQueryMsg->tsOffset; pTSBuf = tsBufCreateFromCompBlocks(tsBlock, pQueryMsg->tsNumOfBlocks, pQueryMsg->tsLen, pQueryMsg->tsOrder, vgId); tsBufResetPos(pTSBuf); bool ret = tsBufNextPos(pTSBuf); + UNUSED(ret); } diff --git a/src/query/src/qFill.c b/src/query/src/qFill.c index f186726c0120f2e2e34580fec0da00c2083e5da9..99fa3a8e0ffc9b53ae370d4e5860bf16a8988367 100644 --- a/src/query/src/qFill.c +++ b/src/query/src/qFill.c @@ -91,7 +91,7 @@ void taosResetFillInfo(SFillInfo* pFillInfo, TSKEY startTimestamp) { pFillInfo->numOfTotal = 0; } -void* taosDestoryFillInfo(SFillInfo* pFillInfo) { +void* taosDestroyFillInfo(SFillInfo* pFillInfo) { if (pFillInfo == NULL) { return NULL; } diff --git a/src/query/src/qResultbuf.c b/src/query/src/qResultbuf.c index b3e97459d36c68d48c04447b159473bd179a5dbb..edb2ca687f476faf01f1a464f80957bd3916407f 100644 --- a/src/query/src/qResultbuf.c +++ b/src/query/src/qResultbuf.c @@ -407,14 +407,14 @@ void destroyResultBuf(SDiskbasedResultBuf* pResultBuf) { } if (pResultBuf->file != NULL) { - qDebug("QInfo:%p res output buffer closed, total:%" PRId64 " bytes, inmem size:%dbytes, file size:%"PRId64" bytes", - pResultBuf->handle, pResultBuf->totalBufSize, listNEles(pResultBuf->lruList) * pResultBuf->pageSize, - pResultBuf->fileSize); + qDebug("QInfo:%p res output buffer closed, total:%.2f Kb, inmem size:%.2f Kb, file size:%.2f", + pResultBuf->handle, pResultBuf->totalBufSize/1024.0, listNEles(pResultBuf->lruList) * pResultBuf->pageSize / 1024.0, + pResultBuf->fileSize/1024.0); fclose(pResultBuf->file); } else { - qDebug("QInfo:%p res output buffer closed, total:%" PRId64 " bytes, no file created", pResultBuf->handle, - pResultBuf->totalBufSize); + qDebug("QInfo:%p res output buffer closed, total:%.2f Kb, no file created", pResultBuf->handle, + pResultBuf->totalBufSize/1024.0); } unlink(pResultBuf->path); diff --git a/src/query/src/qTsbuf.c b/src/query/src/qTsbuf.c index ad29cef5c290f4a6bcd45f8a79828b0c14727dc0..df9ea1db0367f18cf8d436636dc72d2ec67c4648 100644 --- a/src/query/src/qTsbuf.c +++ b/src/query/src/qTsbuf.c @@ -156,7 +156,8 @@ void* tsBufDestroy(STSBuf* pTSBuf) { } else { // tscDebug("tsBuf %p destroyed, tmp file:%s, remains", pTSBuf, pTSBuf->path); } - + + tVariantDestroy(&pTSBuf->block.tag); free(pTSBuf); return NULL; } @@ -218,6 +219,15 @@ static void shrinkBuffer(STSList* ptsData) { } } +static int32_t getTagAreaLength(tVariant* pa) { + int32_t t = sizeof(pa->nLen) * 2 + sizeof(pa->nType); + if (pa->nType != TSDB_DATA_TYPE_NULL) { + t += pa->nLen; + } + + return t; +} + static void writeDataToDisk(STSBuf* pTSBuf) { if (pTSBuf->tsData.len == 0) { return; @@ -243,19 +253,27 @@ static void writeDataToDisk(STSBuf* pTSBuf) { */ int32_t metaLen = 0; metaLen += (int32_t)fwrite(&pBlock->tag.nType, 1, sizeof(pBlock->tag.nType), pTSBuf->f); - metaLen += (int32_t)fwrite(&pBlock->tag.nLen, 1, sizeof(pBlock->tag.nLen), pTSBuf->f); + int32_t trueLen = pBlock->tag.nLen; if (pBlock->tag.nType == TSDB_DATA_TYPE_BINARY || pBlock->tag.nType == TSDB_DATA_TYPE_NCHAR) { + metaLen += (int32_t)fwrite(&pBlock->tag.nLen, 1, sizeof(pBlock->tag.nLen), pTSBuf->f); metaLen += (int32_t)fwrite(pBlock->tag.pz, 1, (size_t)pBlock->tag.nLen, pTSBuf->f); } else if (pBlock->tag.nType != TSDB_DATA_TYPE_NULL) { - metaLen += (int32_t)fwrite(&pBlock->tag.i64Key, 1, sizeof(int64_t), pTSBuf->f); + metaLen += (int32_t)fwrite(&pBlock->tag.nLen, 1, sizeof(pBlock->tag.nLen), pTSBuf->f); + metaLen += (int32_t)fwrite(&pBlock->tag.i64Key, 1, (size_t) pBlock->tag.nLen, pTSBuf->f); + } else { + trueLen = 0; + metaLen += (int32_t)fwrite(&trueLen, 1, sizeof(pBlock->tag.nLen), pTSBuf->f); } fwrite(&pBlock->numOfElem, sizeof(pBlock->numOfElem), 1, pTSBuf->f); fwrite(&pBlock->compLen, sizeof(pBlock->compLen), 1, pTSBuf->f); fwrite(pBlock->payload, (size_t)pBlock->compLen, 1, pTSBuf->f); fwrite(&pBlock->compLen, sizeof(pBlock->compLen), 1, pTSBuf->f); - + + metaLen += (int32_t) fwrite(&trueLen, 1, sizeof(pBlock->tag.nLen), pTSBuf->f); + assert(metaLen == getTagAreaLength(&pBlock->tag)); + int32_t blockSize = metaLen + sizeof(pBlock->numOfElem) + sizeof(pBlock->compLen) * 2 + pBlock->compLen; pTSBuf->fileSize += blockSize; @@ -284,23 +302,28 @@ static void expandBuffer(STSList* ptsData, int32_t inputSize) { STSBlock* readDataFromDisk(STSBuf* pTSBuf, int32_t order, bool decomp) { STSBlock* pBlock = &pTSBuf->block; - + // clear the memory buffer - void* tmp = pBlock->payload; - memset(pBlock, 0, sizeof(STSBlock)); - pBlock->payload = tmp; - + pBlock->compLen = 0; + pBlock->padding = 0; + pBlock->numOfElem = 0; + + int32_t offset = -1; + if (order == TSDB_ORDER_DESC) { /* * set the right position for the reversed traverse, the reversed traverse is started from * the end of each comp data block */ - int32_t ret = fseek(pTSBuf->f, -(int32_t)(sizeof(pBlock->padding)), SEEK_CUR); - size_t sz = fread(&pBlock->padding, sizeof(pBlock->padding), 1, pTSBuf->f); + int32_t prev = -(int32_t) (sizeof(pBlock->padding) + sizeof(pBlock->tag.nLen)); + int32_t ret = fseek(pTSBuf->f, prev, SEEK_CUR); + size_t sz = fread(&pBlock->padding, 1, sizeof(pBlock->padding), pTSBuf->f); + sz = fread(&pBlock->tag.nLen, 1, sizeof(pBlock->tag.nLen), pTSBuf->f); UNUSED(sz); - + pBlock->compLen = pBlock->padding; - int32_t offset = pBlock->compLen + sizeof(pBlock->compLen) * 2 + sizeof(pBlock->numOfElem) + sizeof(pBlock->tag); + + offset = pBlock->compLen + sizeof(pBlock->compLen) * 2 + sizeof(pBlock->numOfElem) + getTagAreaLength(&pBlock->tag); ret = fseek(pTSBuf->f, -offset, SEEK_CUR); UNUSED(ret); } @@ -319,7 +342,7 @@ STSBlock* readDataFromDisk(STSBuf* pTSBuf, int32_t order, bool decomp) { sz = fread(pBlock->tag.pz, (size_t)pBlock->tag.nLen, 1, pTSBuf->f); } else if (pBlock->tag.nType != TSDB_DATA_TYPE_NULL) { - sz = fread(&pBlock->tag.i64Key, sizeof(int64_t), 1, pTSBuf->f); + sz = fread(&pBlock->tag.i64Key, (size_t) pBlock->tag.nLen, 1, pTSBuf->f); } sz = fread(&pBlock->numOfElem, sizeof(pBlock->numOfElem), 1, pTSBuf->f); @@ -327,8 +350,7 @@ STSBlock* readDataFromDisk(STSBuf* pTSBuf, int32_t order, bool decomp) { sz = fread(&pBlock->compLen, sizeof(pBlock->compLen), 1, pTSBuf->f); UNUSED(sz); sz = fread(pBlock->payload, (size_t)pBlock->compLen, 1, pTSBuf->f); - UNUSED(sz); - + if (decomp) { pTSBuf->tsData.len = tsDecompressTimestamp(pBlock->payload, pBlock->compLen, pBlock->numOfElem, pTSBuf->tsData.rawBuf, @@ -337,11 +359,20 @@ STSBlock* readDataFromDisk(STSBuf* pTSBuf, int32_t order, bool decomp) { // read the comp length at the length of comp block sz = fread(&pBlock->padding, sizeof(pBlock->padding), 1, pTSBuf->f); + assert(pBlock->padding == pBlock->compLen); + + int32_t n = 0; + sz = fread(&n, sizeof(pBlock->tag.nLen), 1, pTSBuf->f); + if (pBlock->tag.nType == TSDB_DATA_TYPE_NULL) { + assert(n == 0); + } else { + assert(n == pBlock->tag.nLen); + } + UNUSED(sz); // for backwards traverse, set the start position at the end of previous block if (order == TSDB_ORDER_DESC) { - int32_t offset = pBlock->compLen + sizeof(pBlock->compLen) * 2 + sizeof(pBlock->numOfElem) + sizeof(pBlock->tag); int32_t r = fseek(pTSBuf->f, -offset, SEEK_CUR); UNUSED(r); } @@ -446,7 +477,7 @@ void tsBufFlush(STSBuf* pTSBuf) { fsync(fileno(pTSBuf->f)); } -static int32_t tsBufFindVnodeIndexFromId(STSVnodeBlockInfoEx* pVnodeInfoEx, int32_t numOfVnodes, int32_t vnodeId) { +static int32_t tsBufFindVnodeById(STSVnodeBlockInfoEx* pVnodeInfoEx, int32_t numOfVnodes, int32_t vnodeId) { int32_t j = -1; for (int32_t i = 0; i < numOfVnodes; ++i) { if (pVnodeInfoEx[i].info.vnode == vnodeId) { @@ -478,7 +509,7 @@ static int32_t tsBufFindBlock(STSBuf* pTSBuf, STSVnodeBlockInfo* pBlockInfo, int if (pTSBuf->cur.order == TSDB_ORDER_DESC) { STSBlock* pBlock = &pTSBuf->block; int32_t compBlockSize = - pBlock->compLen + sizeof(pBlock->compLen) * 2 + sizeof(pBlock->numOfElem) + sizeof(pBlock->tag); + pBlock->compLen + sizeof(pBlock->compLen) * 2 + sizeof(pBlock->numOfElem) + getTagAreaLength(&pBlock->tag); int32_t ret = fseek(pTSBuf->f, -compBlockSize, SEEK_CUR); UNUSED(ret); } @@ -506,7 +537,7 @@ static int32_t tsBufFindBlockByTag(STSBuf* pTSBuf, STSVnodeBlockInfo* pBlockInfo } if (tVariantCompare(&pTSBuf->block.tag, tag) == 0) { - return i; + return (pTSBuf->cur.order == TSDB_ORDER_ASC)? i: (pBlockInfo->numOfBlocks - (i + 1)); } } @@ -575,7 +606,7 @@ static int32_t doUpdateVnodeInfo(STSBuf* pTSBuf, int64_t offset, STSVnodeBlockIn } STSVnodeBlockInfo* tsBufGetVnodeBlockInfo(STSBuf* pTSBuf, int32_t vnodeId) { - int32_t j = tsBufFindVnodeIndexFromId(pTSBuf->pData, pTSBuf->numOfVnodes, vnodeId); + int32_t j = tsBufFindVnodeById(pTSBuf->pData, pTSBuf->numOfVnodes, vnodeId); if (j == -1) { return NULL; } @@ -772,7 +803,7 @@ int32_t tsBufMerge(STSBuf* pDestBuf, const STSBuf* pSrcBuf) { int64_t offset = getDataStartOffset(); int32_t size = (int32_t)pSrcBuf->fileSize - (int32_t)offset; - ssize_t rc = taosFSendFile(pDestBuf->f, pSrcBuf->f, &offset, size); + int64_t rc = taosFSendFile(pDestBuf->f, pSrcBuf->f, &offset, size); if (rc == -1) { // tscError("failed to merge tsBuf from:%s to %s, reason:%s\n", pSrcBuf->path, pDestBuf->path, strerror(errno)); @@ -839,7 +870,7 @@ STSElem tsBufGetElemStartPos(STSBuf* pTSBuf, int32_t vnodeId, tVariant* tag) { return elem; } - int32_t j = tsBufFindVnodeIndexFromId(pTSBuf->pData, pTSBuf->numOfVnodes, vnodeId); + int32_t j = tsBufFindVnodeById(pTSBuf->pData, pTSBuf->numOfVnodes, vnodeId); if (j == -1) { return elem; } @@ -992,4 +1023,47 @@ void tsBufGetVnodeIdList(STSBuf* pTSBuf, int32_t* num, int32_t** vnodeId) { for(int32_t i = 0; i < size; ++i) { (*vnodeId)[i] = pTSBuf->pData[i].info.vnode; } -} \ No newline at end of file +} + +int32_t dumpFileBlockByVnodeId(STSBuf* pTSBuf, int32_t vnodeIndex, void* buf, int32_t* len, int32_t* numOfBlocks) { + assert(vnodeIndex >= 0 && vnodeIndex < pTSBuf->numOfVnodes); + STSVnodeBlockInfo *pBlockInfo = &pTSBuf->pData[vnodeIndex].info; + + *len = 0; + *numOfBlocks = 0; + + if (fseek(pTSBuf->f, pBlockInfo->offset, SEEK_SET) != 0) { + int code = TAOS_SYSTEM_ERROR(ferror(pTSBuf->f)); +// qError("%p: fseek failed: %s", pSql, tstrerror(code)); + return code; + } + + size_t s = fread(buf, 1, pBlockInfo->compLen, pTSBuf->f); + if (s != pBlockInfo->compLen) { + int code = TAOS_SYSTEM_ERROR(ferror(pTSBuf->f)); +// tscError("%p: fread didn't return expected data: %s", pSql, tstrerror(code)); + return code; + } + + *len = pBlockInfo->compLen; + *numOfBlocks = pBlockInfo->numOfBlocks; + + return TSDB_CODE_SUCCESS; +} + +STSElem tsBufFindElemStartPosByTag(STSBuf* pTSBuf, tVariant* pTag) { + STSElem el = {.vnode = -1}; + + for (int32_t i = 0; i < pTSBuf->numOfVnodes; ++i) { + el = tsBufGetElemStartPos(pTSBuf, pTSBuf->pData[i].info.vnode, pTag); + if (el.vnode == pTSBuf->pData[i].info.vnode) { + return el; + } + } + + return el; +} + +bool tsBufIsValidElem(STSElem* pElem) { + return pElem->vnode >= 0; +} diff --git a/src/query/src/qUtil.c b/src/query/src/qUtil.c index ac95afffb157847269122bf14f65a3308ef5d9a2..dc968bad06f99ab58bfeb90e7dc143a450ae34b9 100644 --- a/src/query/src/qUtil.c +++ b/src/query/src/qUtil.c @@ -14,8 +14,9 @@ */ #include "os.h" -#include "hash.h" #include "taosmsg.h" +#include "hash.h" + #include "qExecutor.h" #include "qUtil.h" @@ -26,74 +27,36 @@ int32_t getOutputInterResultBufSize(SQuery* pQuery) { size += pQuery->pSelectExpr[i].interBytes; } - assert(size > 0); + assert(size >= 0); return size; } -int32_t initWindowResInfo(SWindowResInfo *pWindowResInfo, SQueryRuntimeEnv *pRuntimeEnv, int32_t size, - int32_t threshold, int16_t type) { +int32_t initWindowResInfo(SWindowResInfo *pWindowResInfo, int32_t size, int32_t threshold, int16_t type) { pWindowResInfo->capacity = size; pWindowResInfo->threshold = threshold; pWindowResInfo->type = type; - _hash_fn_t fn = taosGetDefaultHashFunction(type); - pWindowResInfo->hashList = taosHashInit(threshold, fn, true, false); - if (pWindowResInfo->hashList == NULL) { - return TSDB_CODE_QRY_OUT_OF_MEMORY; - } - pWindowResInfo->curIndex = -1; pWindowResInfo->size = 0; pWindowResInfo->prevSKey = TSKEY_INITIAL_VAL; - SQueryCostInfo* pSummary = &pRuntimeEnv->summary; - - // use the pointer arraylist - pWindowResInfo->pResult = calloc(threshold, sizeof(SWindowResult)); + pWindowResInfo->pResult = calloc(pWindowResInfo->capacity, POINTER_BYTES); if (pWindowResInfo->pResult == NULL) { return TSDB_CODE_QRY_OUT_OF_MEMORY; } - pWindowResInfo->interval = pRuntimeEnv->pQuery->interval.interval; - - pSummary->internalSupSize += sizeof(SWindowResult) * threshold; - pSummary->internalSupSize += (pRuntimeEnv->pQuery->numOfOutput * sizeof(SResultInfo) + pRuntimeEnv->interBufSize) * pWindowResInfo->capacity; - pSummary->numOfTimeWindows = threshold; - - for (int32_t i = 0; i < pWindowResInfo->capacity; ++i) { - int32_t code = createQueryResultInfo(pRuntimeEnv->pQuery, &pWindowResInfo->pResult[i], pRuntimeEnv->stableQuery, pRuntimeEnv->interBufSize); - if (code != TSDB_CODE_SUCCESS) { - return code; - } - } - return TSDB_CODE_SUCCESS; } -void destroyTimeWindowRes(SWindowResult *pWindowRes) { - if (pWindowRes == NULL) { - return; - } - - free(pWindowRes->resultInfo); -} - void cleanupTimeWindowInfo(SWindowResInfo *pWindowResInfo) { if (pWindowResInfo == NULL) { return; } if (pWindowResInfo->capacity == 0) { - assert(pWindowResInfo->hashList == NULL && pWindowResInfo->pResult == NULL); + assert(/*pWindowResInfo->hashList == NULL && */pWindowResInfo->pResult == NULL); return; } - if (pWindowResInfo->pResult != NULL) { - for (int32_t i = 0; i < pWindowResInfo->capacity; ++i) { - destroyTimeWindowRes(&pWindowResInfo->pResult[i]); - } - } - - taosHashCleanup(pWindowResInfo->hashList); taosTFree(pWindowResInfo->pResult); } @@ -103,17 +66,13 @@ void resetTimeWindowInfo(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowR } for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - SWindowResult *pWindowRes = &pWindowResInfo->pResult[i]; - clearTimeWindowResBuf(pRuntimeEnv, pWindowRes); + SResultRow *pWindowRes = pWindowResInfo->pResult[i]; + clearResultRow(pRuntimeEnv, pWindowRes); } pWindowResInfo->curIndex = -1; - taosHashCleanup(pWindowResInfo->hashList); pWindowResInfo->size = 0; - _hash_fn_t fn = taosGetDefaultHashFunction(pWindowResInfo->type); - pWindowResInfo->hashList = taosHashInit(pWindowResInfo->capacity, fn, true, false); - pWindowResInfo->startTime = TSKEY_INITIAL_VAL; pWindowResInfo->prevSKey = TSKEY_INITIAL_VAL; } @@ -127,13 +86,13 @@ void clearFirstNTimeWindow(SQueryRuntimeEnv *pRuntimeEnv, int32_t num) { int32_t numOfClosed = numOfClosedTimeWindow(pWindowResInfo); assert(num >= 0 && num <= numOfClosed); - int16_t type = pWindowResInfo->type; - - char *key = NULL; - int16_t bytes = -1; + int16_t type = pWindowResInfo->type; + STableId* id = TSDB_TABLEID(pRuntimeEnv->pQuery->current->pTable); // uid is always set to be 0. + char *key = NULL; + int16_t bytes = -1; for (int32_t i = 0; i < num; ++i) { - SWindowResult *pResult = &pWindowResInfo->pResult[i]; + SResultRow *pResult = pWindowResInfo->pResult[i]; if (pResult->closed) { // remove the window slot from hash table // todo refactor @@ -145,7 +104,8 @@ void clearFirstNTimeWindow(SQueryRuntimeEnv *pRuntimeEnv, int32_t num) { bytes = tDataTypeDesc[pWindowResInfo->type].nSize; } - taosHashRemove(pWindowResInfo->hashList, (const char *)key, bytes); + SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, key, bytes, id->uid); + taosHashRemove(pRuntimeEnv->pResultRowHashTable, (const char *)pRuntimeEnv->keyBuf, GET_RES_WINDOW_KEY_LEN(bytes)); } else { break; } @@ -155,19 +115,19 @@ void clearFirstNTimeWindow(SQueryRuntimeEnv *pRuntimeEnv, int32_t num) { // clear all the closed windows from the window list for (int32_t k = 0; k < remain; ++k) { - copyTimeWindowResBuf(pRuntimeEnv, &pWindowResInfo->pResult[k], &pWindowResInfo->pResult[num + k]); + copyResultRow(pRuntimeEnv, pWindowResInfo->pResult[k], pWindowResInfo->pResult[num + k]); } // move the unclosed window in the front of the window list for (int32_t k = remain; k < pWindowResInfo->size; ++k) { - SWindowResult *pWindowRes = &pWindowResInfo->pResult[k]; - clearTimeWindowResBuf(pRuntimeEnv, pWindowRes); + SResultRow *pWindowRes = pWindowResInfo->pResult[k]; + clearResultRow(pRuntimeEnv, pWindowRes); } pWindowResInfo->size = remain; for (int32_t k = 0; k < pWindowResInfo->size; ++k) { - SWindowResult *pResult = &pWindowResInfo->pResult[k]; + SResultRow *pResult = pWindowResInfo->pResult[k]; if (type == TSDB_DATA_TYPE_BINARY || type == TSDB_DATA_TYPE_NCHAR) { key = varDataVal(pResult->key); @@ -177,12 +137,15 @@ void clearFirstNTimeWindow(SQueryRuntimeEnv *pRuntimeEnv, int32_t num) { bytes = tDataTypeDesc[pWindowResInfo->type].nSize; } - int32_t *p = (int32_t *)taosHashGet(pWindowResInfo->hashList, (const char *)key, bytes); + SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, key, bytes, id->uid); + int32_t *p = (int32_t *)taosHashGet(pRuntimeEnv->pResultRowHashTable, (const char *)pRuntimeEnv->keyBuf, GET_RES_WINDOW_KEY_LEN(bytes)); assert(p != NULL); int32_t v = (*p - num); assert(v >= 0 && v <= pWindowResInfo->size); - taosHashPut(pWindowResInfo->hashList, (char *)key, bytes, (char *)&v, sizeof(int32_t)); + + SET_RES_WINDOW_KEY(pRuntimeEnv->keyBuf, key, bytes, id->uid); + taosHashPut(pRuntimeEnv->pResultRowHashTable, pRuntimeEnv->keyBuf, GET_RES_WINDOW_KEY_LEN(bytes), (char *)&v, sizeof(int32_t)); } pWindowResInfo->curIndex = -1; @@ -200,7 +163,7 @@ void clearClosedTimeWindow(SQueryRuntimeEnv *pRuntimeEnv) { int32_t numOfClosedTimeWindow(SWindowResInfo *pWindowResInfo) { int32_t i = 0; - while (i < pWindowResInfo->size && pWindowResInfo->pResult[i].closed) { + while (i < pWindowResInfo->size && pWindowResInfo->pResult[i]->closed) { ++i; } @@ -211,11 +174,11 @@ void closeAllTimeWindow(SWindowResInfo *pWindowResInfo) { assert(pWindowResInfo->size >= 0 && pWindowResInfo->capacity >= pWindowResInfo->size); for (int32_t i = 0; i < pWindowResInfo->size; ++i) { - if (pWindowResInfo->pResult[i].closed) { + if (pWindowResInfo->pResult[i]->closed) { continue; } - pWindowResInfo->pResult[i].closed = true; + pWindowResInfo->pResult[i]->closed = true; } } @@ -231,19 +194,19 @@ void removeRedundantWindow(SWindowResInfo *pWindowResInfo, TSKEY lastKey, int32_ } // get the result order - int32_t resultOrder = (pWindowResInfo->pResult[0].win.skey < pWindowResInfo->pResult[1].win.skey)? 1:-1; + int32_t resultOrder = (pWindowResInfo->pResult[0]->win.skey < pWindowResInfo->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; + TSKEY ekey = pWindowResInfo->pResult[i]->win.ekey; while (i < pWindowResInfo->size && (ekey < lastKey)) { ++i; } } else if (order == QUERY_DESC_FORWARD_STEP) { - while (i < pWindowResInfo->size && (pWindowResInfo->pResult[i].win.skey > lastKey)) { + while (i < pWindowResInfo->size && (pWindowResInfo->pResult[i]->win.skey > lastKey)) { ++i; } } @@ -254,30 +217,33 @@ void removeRedundantWindow(SWindowResInfo *pWindowResInfo, TSKEY lastKey, int32_ } bool isWindowResClosed(SWindowResInfo *pWindowResInfo, int32_t slot) { - return (getWindowResult(pWindowResInfo, slot)->closed == true); + return (getResultRow(pWindowResInfo, slot)->closed == true); } void closeTimeWindow(SWindowResInfo *pWindowResInfo, int32_t slot) { - getWindowResult(pWindowResInfo, slot)->closed = true; + getResultRow(pWindowResInfo, slot)->closed = true; } -void clearTimeWindowResBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pWindowRes) { +void clearResultRow(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *pWindowRes) { if (pWindowRes == NULL) { return; } - tFilePage *page = getResBufPage(pRuntimeEnv->pResultBuf, pWindowRes->pageId); + // the result does not put into the SDiskbasedResultBuf, ignore it. + if (pWindowRes->pageId >= 0) { + tFilePage *page = getResBufPage(pRuntimeEnv->pResultBuf, pWindowRes->pageId); - for (int32_t i = 0; i < pRuntimeEnv->pQuery->numOfOutput; ++i) { - SResultInfo *pResultInfo = &pWindowRes->resultInfo[i]; - - char * s = getPosInResultPage(pRuntimeEnv, i, pWindowRes, page); - size_t size = pRuntimeEnv->pQuery->pSelectExpr[i].bytes; - memset(s, 0, size); - - RESET_RESULT_INFO(pResultInfo); + for (int32_t i = 0; i < pRuntimeEnv->pQuery->numOfOutput; ++i) { + SResultRowCellInfo *pResultInfo = &pWindowRes->pCellInfo[i]; + + char * s = getPosInResultPage(pRuntimeEnv, i, pWindowRes, page); + size_t size = pRuntimeEnv->pQuery->pSelectExpr[i].bytes; + memset(s, 0, size); + + RESET_RESULT_INFO(pResultInfo); + } } - + pWindowRes->numOfRows = 0; pWindowRes->pageId = -1; pWindowRes->rowId = -1; @@ -290,7 +256,7 @@ void clearTimeWindowResBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pWindow * since the attribute of "Pos" is bound to each window result when the window result is created in the * disk-based result buffer. */ -void copyTimeWindowResBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *dst, const SWindowResult *src) { +void copyResultRow(SQueryRuntimeEnv *pRuntimeEnv, SResultRow *dst, const SResultRow *src) { dst->numOfRows = src->numOfRows; dst->win = src->win; dst->closed = src->closed; @@ -298,25 +264,105 @@ void copyTimeWindowResBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *dst, con int32_t nOutputCols = pRuntimeEnv->pQuery->numOfOutput; for (int32_t i = 0; i < nOutputCols; ++i) { - SResultInfo *pDst = &dst->resultInfo[i]; - SResultInfo *pSrc = &src->resultInfo[i]; + SResultRowCellInfo *pDst = getResultCell(pRuntimeEnv, dst, i); + SResultRowCellInfo *pSrc = getResultCell(pRuntimeEnv, src, i); - char *buf = pDst->interResultBuf; - memcpy(pDst, pSrc, sizeof(SResultInfo)); - pDst->interResultBuf = buf; // restore the allocated buffer +// char *buf = pDst->interResultBuf; + memcpy(pDst, pSrc, sizeof(SResultRowCellInfo) + pRuntimeEnv->pCtx[i].interBufBytes); +// pDst->interResultBuf = buf; // restore the allocated buffer // copy the result info struct - memcpy(pDst->interResultBuf, pSrc->interResultBuf, pDst->bufLen); +// memcpy(pDst->interResultBuf, pSrc->interResultBuf, pRuntimeEnv->pCtx[i].interBufBytes); // copy the output buffer data from src to dst, the position info keep unchanged tFilePage *dstpage = getResBufPage(pRuntimeEnv->pResultBuf, dst->pageId); char * dstBuf = getPosInResultPage(pRuntimeEnv, i, dst, dstpage); tFilePage *srcpage = getResBufPage(pRuntimeEnv->pResultBuf, src->pageId); - char * srcBuf = getPosInResultPage(pRuntimeEnv, i, (SWindowResult *)src, srcpage); + char * srcBuf = getPosInResultPage(pRuntimeEnv, i, (SResultRow *)src, srcpage); size_t s = pRuntimeEnv->pQuery->pSelectExpr[i].bytes; memcpy(dstBuf, srcBuf, s); } } +SResultRowCellInfo* getResultCell(SQueryRuntimeEnv* pRuntimeEnv, const SResultRow* pRow, int32_t index) { + assert(index >= 0 && index < pRuntimeEnv->pQuery->numOfOutput); + return (SResultRowCellInfo*)((char*) pRow->pCellInfo + pRuntimeEnv->rowCellInfoOffset[index]); +} + +size_t getWindowResultSize(SQueryRuntimeEnv* pRuntimeEnv) { + return (pRuntimeEnv->pQuery->numOfOutput * sizeof(SResultRowCellInfo)) + pRuntimeEnv->interBufSize + sizeof(SResultRow); +} + +SResultRowPool* initResultRowPool(size_t size) { + SResultRowPool* p = calloc(1, sizeof(SResultRowPool)); + if (p == NULL) { + return NULL; + } + + p->numOfElemPerBlock = 128; + + p->elemSize = (int32_t) size; + p->blockSize = p->numOfElemPerBlock * p->elemSize; + p->position.pos = 0; + + p->pData = taosArrayInit(8, POINTER_BYTES); + return p; +} + +SResultRow* getNewResultRow(SResultRowPool* p) { + if (p == NULL) { + return NULL; + } + + void* ptr = NULL; + if (p->position.pos == 0) { + ptr = calloc(1, p->blockSize); + taosArrayPush(p->pData, &ptr); + + } else { + size_t last = taosArrayGetSize(p->pData); + + void** pBlock = taosArrayGet(p->pData, last - 1); + ptr = ((char*) (*pBlock)) + p->elemSize * p->position.pos; + } + + p->position.pos = (p->position.pos + 1)%p->numOfElemPerBlock; + initResultRow(ptr); + + return ptr; +} + +int64_t getResultRowPoolMemSize(SResultRowPool* p) { + if (p == NULL) { + return 0; + } + + return taosArrayGetSize(p->pData) * p->blockSize; +} + +int32_t getNumOfAllocatedResultRows(SResultRowPool* p) { + return (int32_t) taosArrayGetSize(p->pData) * p->numOfElemPerBlock; +} + +int32_t getNumOfUsedResultRows(SResultRowPool* p) { + return getNumOfAllocatedResultRows(p) - p->numOfElemPerBlock + p->position.pos; +} + +void* destroyResultRowPool(SResultRowPool* p) { + if (p == NULL) { + return NULL; + } + + size_t size = taosArrayGetSize(p->pData); + for(int32_t i = 0; i < size; ++i) { + void** ptr = taosArrayGet(p->pData, i); + taosTFree(*ptr); + } + + taosArrayDestroy(p->pData); + + taosTFree(p); + return NULL; +} diff --git a/src/query/tests/tsBufTest.cpp b/src/query/tests/tsBufTest.cpp index 8cd3a9cbef0bbf268db074a935a7b25cb389944e..2892af397960df6c1730f17209a02aa227887439 100644 --- a/src/query/tests/tsBufTest.cpp +++ b/src/query/tests/tsBufTest.cpp @@ -472,13 +472,20 @@ void mergeIdenticalVnodeBufferTest() { tsBufFlush(pTSBuf2); tsBufMerge(pTSBuf1, pTSBuf2); - EXPECT_EQ(pTSBuf1->numOfVnodes, 1); + EXPECT_EQ(pTSBuf1->numOfVnodes, 2); EXPECT_EQ(pTSBuf1->numOfTotal, numOfTags * 2 * num); tsBufResetPos(pTSBuf1); + + int32_t count = 0; while (tsBufNextPos(pTSBuf1)) { STSElem elem = tsBufGetElem(pTSBuf1); - EXPECT_EQ(elem.vnode, 12); + + if (count++ < numOfTags * num) { + EXPECT_EQ(elem.vnode, 12); + } else { + EXPECT_EQ(elem.vnode, 77); + } printf("%d-%" PRIu64 "-%" PRIu64 "\n", elem.vnode, elem.tag->i64Key, elem.ts); } diff --git a/src/rpc/src/rpcMain.c b/src/rpc/src/rpcMain.c index d170fed30c437f8566853ff6cfd92c6032347d87..d01c34c810a5c5d8dd172bf9065565ab0dfa7ae5 100644 --- a/src/rpc/src/rpcMain.c +++ b/src/rpc/src/rpcMain.c @@ -341,7 +341,7 @@ void *rpcMallocCont(int contLen) { tError("failed to malloc msg, size:%d", size); return NULL; } else { - tTrace("malloc mem: %p", start); + tTrace("malloc mem:%p size:%d", start, size); } return start + sizeof(SRpcReqContext) + sizeof(SRpcHead); @@ -1479,7 +1479,7 @@ static SRpcHead *rpcDecompressRpcMsg(SRpcHead *pHead) { pNewHead->msgLen = rpcMsgLenFromCont(origLen); rpcFreeMsg(pHead); // free the compressed message buffer pHead = pNewHead; - tTrace("decomp malloc mem: %p", temp); + tTrace("decomp malloc mem:%p", temp); } else { tError("failed to allocate memory to decompress msg, contLen:%d", contLen); } diff --git a/src/rpc/src/rpcTcp.c b/src/rpc/src/rpcTcp.c index bc8d360d39509ce9a2fdbe6a9dd883c5c5c99190..c6de398608e419f47cc734053d64bdc2d247f90a 100644 --- a/src/rpc/src/rpcTcp.c +++ b/src/rpc/src/rpcTcp.c @@ -437,12 +437,13 @@ static int taosReadTcpData(SFdObj *pFdObj, SRecvInfo *pInfo) { } msgLen = (int32_t)htonl((uint32_t)rpcHead.msgLen); - buffer = malloc(msgLen + tsRpcOverhead); + int32_t size = msgLen + tsRpcOverhead; + buffer = malloc(size); if (NULL == buffer) { tError("%s %p TCP malloc(size:%d) fail", pThreadObj->label, pFdObj->thandle, msgLen); return -1; } else { - tTrace("TCP malloc mem: %p", buffer); + tTrace("TCP malloc mem:%p size:%d", buffer, size); } msg = buffer + tsRpcOverhead; diff --git a/src/rpc/src/rpcUdp.c b/src/rpc/src/rpcUdp.c index 6f653046615f162c516b5eebf08995d30c6214d7..9bac2b1c192e1010fd3f796a635b401ae7372719 100644 --- a/src/rpc/src/rpcUdp.c +++ b/src/rpc/src/rpcUdp.c @@ -209,12 +209,13 @@ static void *taosRecvUdpData(void *param) { continue; } - char *tmsg = malloc(dataLen + tsRpcOverhead); + int32_t size = dataLen + tsRpcOverhead; + char *tmsg = malloc(size); if (NULL == tmsg) { tError("%s failed to allocate memory, size:%" PRId64, pConn->label, (int64_t)dataLen); continue; } else { - tTrace("UDP malloc mem: %p", tmsg); + tTrace("UDP malloc mem:%p size:%d", tmsg, size); } tmsg += tsRpcOverhead; // overhead for SRpcReqContext diff --git a/src/sync/src/syncRetrieve.c b/src/sync/src/syncRetrieve.c index 60625d75eccdbe6bbb29f97b31ecd8e9855480a7..0fe189db7a5fa15fb05b2182acfac5dac1160cfe 100644 --- a/src/sync/src/syncRetrieve.c +++ b/src/sync/src/syncRetrieve.c @@ -149,7 +149,7 @@ static int syncRetrieveFile(SSyncPeer *pPeer) { int sfd = open(name, O_RDONLY); if (sfd < 0) break; - ret = taosTSendFile(pPeer->syncFd, sfd, NULL, fileInfo.size); + ret = taosSendFile(pPeer->syncFd, sfd, NULL, fileInfo.size); close(sfd); if (ret < 0) break; @@ -287,7 +287,7 @@ static int syncRetrieveLastWal(SSyncPeer *pPeer, char *name, uint64_t fversion, return -1; } -static int syncProcessLastWal(SSyncPeer *pPeer, char *wname, uint32_t index) { +static int syncProcessLastWal(SSyncPeer *pPeer, char *wname, int64_t index) { SSyncNode *pNode = pPeer->pSyncNode; int code = -1; char fname[TSDB_FILENAME_LEN * 2]; // full path to wal file @@ -377,7 +377,7 @@ static int syncRetrieveWal(SSyncPeer *pPeer) { int32_t size; struct stat fstat; int code = -1; - uint32_t index = 0; + int64_t index = 0; while (1) { // retrieve wal info @@ -406,7 +406,7 @@ static int syncRetrieveWal(SSyncPeer *pPeer) { int sfd = open(fname, O_RDONLY); if (sfd < 0) break; - code = taosTSendFile(pPeer->syncFd, sfd, NULL, size); + code = taosSendFile(pPeer->syncFd, sfd, NULL, size); close(sfd); if (code < 0) break; diff --git a/src/sync/test/syncServer.c b/src/sync/test/syncServer.c index 380b971fa89bd1726e138c973a974bc995500693..0cf752da97848056a7e5e3892025e87c3aef4c5d 100644 --- a/src/sync/test/syncServer.c +++ b/src/sync/test/syncServer.c @@ -254,7 +254,7 @@ uint32_t getFileInfo(void *ahandle, char *name, uint32_t *index, uint32_t eindex return magic; } -int getWalInfo(void *ahandle, char *name, uint32_t *index) { +int getWalInfo(void *ahandle, char *name, int64_t *index) { struct stat fstat; char aname[280]; diff --git a/src/tsdb/src/tsdbFile.c b/src/tsdb/src/tsdbFile.c index 5633e97cdf56bea8f2c24fc862d720bea2c376db..3a5416dd30b61f3e8925b3ef67c5533d18d19021 100644 --- a/src/tsdb/src/tsdbFile.c +++ b/src/tsdb/src/tsdbFile.c @@ -13,10 +13,8 @@ * along with this program. If not, see . */ #define _DEFAULT_SOURCE -#include - #define TAOS_RANDOM_FILE_FAIL_TEST - +#include #include "os.h" #include "talgo.h" #include "tchecksum.h" @@ -428,7 +426,7 @@ int tsdbUpdateFileHeader(SFile *pFile) { terrno = TAOS_SYSTEM_ERROR(errno); return -1; } - if (taosTWrite(pFile->fd, (void *)buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { + if (taosWrite(pFile->fd, (void *)buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { tsdbError("failed to write %d bytes to file %s since %s", TSDB_FILE_HEAD_SIZE, pFile->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); return -1; @@ -493,7 +491,7 @@ int tsdbLoadFileHeader(SFile *pFile, uint32_t *version) { return -1; } - if (taosTRead(pFile->fd, buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { + if (taosRead(pFile->fd, buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { tsdbError("failed to read file %s header part with %d bytes, reason:%s", pFile->fname, TSDB_FILE_HEAD_SIZE, strerror(errno)); terrno = TSDB_CODE_TDB_FILE_CORRUPTED; diff --git a/src/tsdb/src/tsdbMain.c b/src/tsdb/src/tsdbMain.c index a1e6376304a54ea810f3662bd30e3f97fd53765c..f714aadc36b45a940268a59493067ac0892b4058 100644 --- a/src/tsdb/src/tsdbMain.c +++ b/src/tsdb/src/tsdbMain.c @@ -579,7 +579,7 @@ static int32_t tsdbSaveConfig(char *rootDir, STsdbCfg *pCfg) { taosCalcChecksumAppend(0, (uint8_t *)buf, TSDB_FILE_HEAD_SIZE); - if (taosTWrite(fd, (void *)buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { + if (taosWrite(fd, (void *)buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { tsdbError("vgId:%d failed to write %d bytes to file %s since %s", pCfg->tsdbId, TSDB_FILE_HEAD_SIZE, fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -620,7 +620,7 @@ static int tsdbLoadConfig(char *rootDir, STsdbCfg *pCfg) { goto _err; } - if (taosTRead(fd, (void *)buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { + if (taosRead(fd, (void *)buf, TSDB_FILE_HEAD_SIZE) < TSDB_FILE_HEAD_SIZE) { tsdbError("failed to read %d bytes from file %s since %s", TSDB_FILE_HEAD_SIZE, fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); goto _err; diff --git a/src/tsdb/src/tsdbRWHelper.c b/src/tsdb/src/tsdbRWHelper.c index 357093bd9e125567ef3d3629c00877f54778b500..9940ec40efc8742054ed036b4cbe32ffc48d0297 100644 --- a/src/tsdb/src/tsdbRWHelper.c +++ b/src/tsdb/src/tsdbRWHelper.c @@ -14,9 +14,7 @@ */ #define _DEFAULT_SOURCE - #define TAOS_RANDOM_FILE_FAIL_TEST - #include "os.h" #include "talgo.h" #include "tchecksum.h" @@ -335,7 +333,7 @@ int tsdbMoveLastBlockIfNeccessary(SRWHelper *pHelper) { return -1; } - if (taosTSendFile(helperNewLastF(pHelper)->fd, helperLastF(pHelper)->fd, NULL, pCompBlock->len) < pCompBlock->len) { + if (taosSendFile(helperNewLastF(pHelper)->fd, helperLastF(pHelper)->fd, NULL, pCompBlock->len) < pCompBlock->len) { tsdbError("vgId:%d failed to sendfile from file %s to file %s since %s", REPO_ID(pHelper->pRepo), helperLastF(pHelper)->fname, helperNewLastF(pHelper)->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -380,7 +378,7 @@ int tsdbWriteCompInfo(SRWHelper *pHelper) { pIdx->tid = pHelper->tableInfo.tid; ASSERT(pIdx->offset >= TSDB_FILE_HEAD_SIZE); - if (taosTWrite(pFile->fd, (void *)(pHelper->pCompInfo), pIdx->len) < (int)pIdx->len) { + if (taosWrite(pFile->fd, (void *)(pHelper->pCompInfo), pIdx->len) < (int)pIdx->len) { tsdbError("vgId:%d failed to write %d bytes to file %s since %s", REPO_ID(pHelper->pRepo), pIdx->len, pFile->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -432,7 +430,7 @@ int tsdbWriteCompIdx(SRWHelper *pHelper) { ASSERT(offset == pFile->info.size); - if (taosTWrite(pFile->fd, (void *)pHelper->pWIdx, pFile->info.len) < (int)pFile->info.len) { + if (taosWrite(pFile->fd, (void *)pHelper->pWIdx, pFile->info.len) < (int)pFile->info.len) { tsdbError("vgId:%d failed to write %d bytes to file %s since %s", REPO_ID(pHelper->pRepo), pFile->info.len, pFile->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -454,7 +452,7 @@ int tsdbLoadCompIdxImpl(SFile *pFile, uint32_t offset, uint32_t len, void *buffe return -1; } - if (taosTRead(pFile->fd, buffer, len) < len) { + if (taosRead(pFile->fd, buffer, len) < len) { tsdbError("%s: read file %s offset %u len %u failed since %s", prefixMsg, pFile->fname, offset, len, strerror(errno)); terrno = TSDB_CODE_TDB_FILE_CORRUPTED; @@ -551,7 +549,7 @@ int tsdbLoadCompInfoImpl(SFile *pFile, SCompIdx *pIdx, SCompInfo **ppCompInfo) { return -1; } - if (taosTRead(pFile->fd, (void *)(*ppCompInfo), pIdx->len) < (int)pIdx->len) { + if (taosRead(pFile->fd, (void *)(*ppCompInfo), pIdx->len) < (int)pIdx->len) { tsdbError("%s: read file %s offset %u len %u failed since %s", prefixMsg, pFile->fname, pIdx->offset, pIdx->len, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -608,7 +606,7 @@ int tsdbLoadCompData(SRWHelper *pHelper, SCompBlock *pCompBlock, void *target) { return -1; } - if (taosTRead(pFile->fd, (void *)pHelper->pCompData, tsize) < tsize) { + if (taosRead(pFile->fd, (void *)pHelper->pCompData, tsize) < tsize) { tsdbError("vgId:%d failed to read %" PRIzu " bytes from file %s since %s", REPO_ID(pHelper->pRepo), tsize, pFile->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -823,7 +821,7 @@ static int tsdbWriteBlockToFile(SRWHelper *pHelper, SFile *pFile, SDataCols *pDa sizeof(TSCKSUM)); // Write the whole block to file - if (taosTWrite(pFile->fd, (void *)pCompData, lsize) < lsize) { + if (taosWrite(pFile->fd, (void *)pCompData, lsize) < lsize) { tsdbError("vgId:%d failed to write %d bytes to file %s since %s", REPO_ID(helperRepo(pHelper)), lsize, pFile->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -1210,7 +1208,7 @@ static int tsdbLoadColData(SRWHelper *pHelper, SFile *pFile, SCompBlock *pCompBl return -1; } - if (taosTRead(pFile->fd, pHelper->pBuffer, pCompCol->len) < pCompCol->len) { + if (taosRead(pFile->fd, pHelper->pBuffer, pCompCol->len) < pCompCol->len) { tsdbError("vgId:%d failed to read %d bytes from file %s since %s", REPO_ID(pHelper->pRepo), pCompCol->len, pFile->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); @@ -1325,7 +1323,7 @@ static int tsdbLoadBlockDataImpl(SRWHelper *pHelper, SCompBlock *pCompBlock, SDa terrno = TAOS_SYSTEM_ERROR(errno); goto _err; } - if (taosTRead(fd, (void *)pCompData, pCompBlock->len) < pCompBlock->len) { + if (taosRead(fd, (void *)pCompData, pCompBlock->len) < pCompBlock->len) { tsdbError("vgId:%d failed to read %d bytes from file %s since %s", REPO_ID(pHelper->pRepo), pCompBlock->len, pFile->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); diff --git a/src/tsdb/src/tsdbRead.c b/src/tsdb/src/tsdbRead.c index 90e16aee53f6978be3b4b0d81172152c0b74138e..3daefc1f5f0b74da98f926d650006f2bce46e23d 100644 --- a/src/tsdb/src/tsdbRead.c +++ b/src/tsdb/src/tsdbRead.c @@ -72,8 +72,8 @@ typedef struct STableCheckInfo { STable* pTableObj; SCompInfo* pCompInfo; int32_t compSize; - int32_t numOfBlocks; // number of qualified data blocks not the original blocks - int32_t chosen; // indicate which iterator should move forward + int32_t numOfBlocks:29; // number of qualified data blocks not the original blocks + int8_t chosen:2; // indicate which iterator should move forward bool initBuf; // whether to initialize the in-memory skip list iterator or not SSkipListIterator* iter; // mem buffer skip list iterator SSkipListIterator* iiter; // imem buffer skip list iterator diff --git a/src/util/inc/hash.h b/src/util/inc/hash.h index 2a58bec8830faab4cf9909bd417a44e2e8881570..42bc136584d618e62826c307b698746ca387ff41 100644 --- a/src/util/inc/hash.h +++ b/src/util/inc/hash.h @@ -31,14 +31,16 @@ extern "C" { typedef void (*_hash_free_fn_t)(void *param); typedef struct SHashNode { - char *key; -// struct SHashNode *prev; +// char *key; struct SHashNode *next; - uint32_t hashVal; // the hash value of key, if hashVal == HASH_VALUE_IN_TRASH, this node is moved to trash + uint32_t hashVal; // the hash value of key uint32_t keyLen; // length of the key - char *data; +// 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) + typedef enum SHashLockTypeE { HASH_NO_LOCK = 0, HASH_ENTRY_LOCK = 1, @@ -175,6 +177,8 @@ void* taosHashDestroyIter(SHashMutableIterator* iter); */ int32_t taosHashGetMaxOverflowLinkLength(const SHashObj *pHashObj); +size_t taosHashGetMemSize(const SHashObj *pHashObj); + #ifdef __cplusplus } #endif diff --git a/src/util/inc/tsocket.h b/src/util/inc/tsocket.h index f14e8dbb356e131bbd580f8c5310bea944e71033..391cc44accadd785fb46e02aa79cde7617834c70 100644 --- a/src/util/inc/tsocket.h +++ b/src/util/inc/tsocket.h @@ -20,21 +20,21 @@ extern "C" { #endif -int taosReadn(SOCKET sock, char *buffer, int len); -int taosWriteMsg(SOCKET fd, void *ptr, int nbytes); -int taosReadMsg(SOCKET fd, void *ptr, int nbytes); -int taosNonblockwrite(SOCKET fd, char *ptr, int nbytes); -int taosCopyFds(SOCKET sfd, SOCKET dfd, int64_t len); -int taosSetNonblocking(SOCKET sock, int on); +int32_t taosReadn(SOCKET sock, char *buffer, int32_t len); +int32_t taosWriteMsg(SOCKET fd, void *ptr, int32_t nbytes); +int32_t taosReadMsg(SOCKET fd, void *ptr, int32_t nbytes); +int32_t taosNonblockwrite(SOCKET fd, char *ptr, int32_t nbytes); +int32_t taosCopyFds(SOCKET sfd, SOCKET dfd, int64_t len); +int32_t taosSetNonblocking(SOCKET sock, int32_t on); -SOCKET taosOpenUdpSocket(uint32_t localIp, uint16_t localPort); -SOCKET taosOpenTcpClientSocket(uint32_t ip, uint16_t port, uint32_t localIp); -SOCKET taosOpenTcpServerSocket(uint32_t ip, uint16_t port); -int taosKeepTcpAlive(SOCKET sockFd); +SOCKET taosOpenUdpSocket(uint32_t localIp, uint16_t localPort); +SOCKET taosOpenTcpClientSocket(uint32_t ip, uint16_t port, uint32_t localIp); +SOCKET taosOpenTcpServerSocket(uint32_t ip, uint16_t port); +int32_t taosKeepTcpAlive(SOCKET sockFd); -int taosGetFqdn(char *); +int32_t taosGetFqdn(char *); uint32_t taosGetIpFromFqdn(const char *); -void tinet_ntoa(char *ipstr, unsigned int ip); +void tinet_ntoa(char *ipstr, uint32_t ip); uint32_t ip2uint(const char *const ip_addr); #ifdef __cplusplus diff --git a/src/util/src/hash.c b/src/util/src/hash.c index 625d4af1ac981c5d1f5f079f5b15533a4d63ef24..22b5da1491854926d53f2f56ee402e6269fa624c 100644 --- a/src/util/src/hash.c +++ b/src/util/src/hash.c @@ -22,14 +22,13 @@ #define DO_FREE_HASH_NODE(_n) \ do { \ - taosTFree((_n)->data); \ taosTFree(_n); \ } while (0) #define FREE_HASH_NODE(_h, _n) \ do { \ if ((_h)->freeFp) { \ - (_h)->freeFp((_n)->data); \ + (_h)->freeFp(GET_HASH_NODE_DATA(_n)); \ } \ \ DO_FREE_HASH_NODE(_n); \ @@ -77,7 +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(pNode->key, key, keyLen) == 0)) { + if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0)) { assert(pNode->hashVal == hashVal); break; } @@ -115,11 +114,15 @@ 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(SHashNode *pNode, SHashNode *pNewNode) { +static FORCE_INLINE SHashNode *doUpdateHashNode(SHashEntry* pe, SHashNode* prev, SHashNode *pNode, SHashNode *pNewNode) { assert(pNode->keyLen == pNewNode->keyLen); - SWAP(pNode->key, pNewNode->key, void *); - SWAP(pNode->data, pNewNode->data, void *); + if (prev != NULL) { + prev->next = pNewNode; + } else { + pe->next = pNewNode; + } + pNewNode->next = pNode->next; return pNewNode; } @@ -208,12 +211,14 @@ int32_t taosHashPut(SHashObj *pHashObj, const void *key, size_t keyLen, void *da assert(pNode == NULL); } + SHashNode* prev = NULL; while (pNode) { - if ((pNode->keyLen == keyLen) && (memcmp(pNode->key, key, keyLen) == 0)) { + if ((pNode->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY(pNode), key, keyLen) == 0)) { assert(pNode->hashVal == hashVal); break; } + prev = pNode; pNode = pNode->next; } @@ -239,7 +244,10 @@ 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(pNode, pNewNode); + doUpdateHashNode(pe, prev, pNode, pNewNode); + DO_FREE_HASH_NODE(pNode); + } else { + DO_FREE_HASH_NODE(pNewNode); } if (pHashObj->type == HASH_ENTRY_LOCK) { @@ -249,7 +257,6 @@ int32_t taosHashPut(SHashObj *pHashObj, const void *key, size_t keyLen, void *da // enable resize __rd_unlock(&pHashObj->lock, pHashObj->type); - DO_FREE_HASH_NODE(pNewNode); return pHashObj->enableUpdate ? 0 : -1; } } @@ -293,13 +300,13 @@ void* taosHashGetCB(SHashObj *pHashObj, const void *key, size_t keyLen, void (*f SHashNode *pNode = doSearchInEntryList(pe, key, keyLen, hashVal); if (pNode != NULL) { if (fp != NULL) { - fp(pNode->data); + fp(GET_HASH_NODE_DATA(pNode)); } if (d != NULL) { - memcpy(d, pNode->data, dsize); + memcpy(d, GET_HASH_NODE_DATA(pNode), dsize); } else { - data = pNode->data; + data = GET_HASH_NODE_DATA(pNode); } } @@ -357,13 +364,13 @@ int32_t taosHashRemoveWithData(SHashObj *pHashObj, const void *key, size_t keyLe SHashNode *pRes = NULL; // remove it - if ((pNode->keyLen == keyLen) && (memcmp(pNode->key, key, keyLen) == 0)) { + 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((pNode->next)->key, key, keyLen) == 0)) { + if (((pNode->next)->keyLen == keyLen) && (memcmp(GET_HASH_NODE_KEY((pNode->next)), key, keyLen) == 0)) { assert((pNode->next)->hashVal == hashVal); break; } @@ -392,7 +399,7 @@ int32_t taosHashRemoveWithData(SHashObj *pHashObj, const void *key, size_t keyLe __rd_unlock(&pHashObj->lock, pHashObj->type); if (data != NULL && pRes != NULL) { - memcpy(data, pRes->data, dsize); + memcpy(data, GET_HASH_NODE_DATA(pRes), dsize); } if (pRes != NULL) { @@ -426,7 +433,7 @@ int32_t taosHashCondTraverse(SHashObj *pHashObj, bool (*fp)(void *, void *), voi // todo remove the first node SHashNode *pNode = NULL; while((pNode = pEntry->next) != NULL) { - if (fp && (!fp(param, pNode->data))) { + if (fp && (!fp(param, GET_HASH_NODE_DATA(pNode)))) { pEntry->num -= 1; atomic_sub_fetch_64(&pHashObj->size, 1); @@ -451,7 +458,7 @@ int32_t taosHashCondTraverse(SHashObj *pHashObj, bool (*fp)(void *, void *), voi while ((pNext = pNode->next) != NULL) { // not qualified, remove it - if (fp && (!fp(param, pNext->data))) { + if (fp && (!fp(param, GET_HASH_NODE_DATA(pNext)))) { pNode->next = pNext->next; pEntry->num -= 1; atomic_sub_fetch_64(&pHashObj->size, 1); @@ -605,7 +612,7 @@ bool taosHashIterNext(SHashMutableIterator *pIter) { } } -void *taosHashIterGet(SHashMutableIterator *iter) { return (iter == NULL) ? NULL : iter->pCur->data; } +void *taosHashIterGet(SHashMutableIterator *iter) { return (iter == NULL) ? NULL : GET_HASH_NODE_DATA(iter->pCur); } void *taosHashDestroyIter(SHashMutableIterator *iter) { if (iter == NULL) { @@ -743,21 +750,19 @@ void taosHashTableResize(SHashObj *pHashObj) { } SHashNode *doCreateHashNode(const void *key, size_t keyLen, const void *pData, size_t dsize, uint32_t hashVal) { - SHashNode *pNewNode = calloc(1, sizeof(SHashNode)); + SHashNode *pNewNode = calloc(1, sizeof(SHashNode) + keyLen + dsize); if (pNewNode == NULL) { uError("failed to allocate memory, reason:%s", strerror(errno)); return NULL; } - pNewNode->data = malloc(dsize + keyLen); - memcpy(pNewNode->data, pData, dsize); - - pNewNode->key = pNewNode->data + dsize; - memcpy(pNewNode->key, key, keyLen); - pNewNode->keyLen = (uint32_t)keyLen; pNewNode->hashVal = hashVal; + + memcpy(GET_HASH_NODE_DATA(pNewNode), pData, dsize); + memcpy(GET_HASH_NODE_KEY(pNewNode), key, keyLen); + return pNewNode; } @@ -798,3 +803,11 @@ SHashNode *getNextHashNode(SHashMutableIterator *pIter) { return 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); +} diff --git a/src/util/src/tconfig.c b/src/util/src/tconfig.c index 0ec55841a060a49f4aa9e29981fa426e42d29d5c..f449bfb68b6f9beecaf9f1a041a4c6c1c73ee09f 100644 --- a/src/util/src/tconfig.c +++ b/src/util/src/tconfig.c @@ -355,7 +355,11 @@ bool taosReadGlobalCfg() { fclose(fp); taosTFree(line); - + + if (debugFlag & DEBUG_TRACE || debugFlag & DEBUG_DEBUG || debugFlag & DEBUG_DUMP) { + taosSetAllDebugFlag(); + } + return true; } diff --git a/src/util/src/tkvstore.c b/src/util/src/tkvstore.c index 6ba1d87d92ecf216bfde346f9d3ff1563515d34d..0806c29ff84b2a1e44792964ed959b7370611373 100644 --- a/src/util/src/tkvstore.c +++ b/src/util/src/tkvstore.c @@ -14,9 +14,7 @@ */ #define _DEFAULT_SOURCE - #define TAOS_RANDOM_FILE_FAIL_TEST - #include "os.h" #include "hash.h" #include "taoserror.h" @@ -188,7 +186,7 @@ int tdKVStoreStartCommit(SKVStore *pStore) { goto _err; } - if (taosTSendFile(pStore->sfd, pStore->fd, NULL, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) { + if (taosSendFile(pStore->sfd, pStore->fd, NULL, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) { uError("failed to send file %d bytes since %s", TD_KVSTORE_HEADER_SIZE, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); goto _err; @@ -248,13 +246,13 @@ int tdUpdateKVStoreRecord(SKVStore *pStore, uint64_t uid, void *cont, int contLe ASSERT(tlen == POINTER_DISTANCE(pBuf, buf)); ASSERT(tlen == sizeof(SKVRecord)); - if (taosTWrite(pStore->fd, buf, tlen) < tlen) { + if (taosWrite(pStore->fd, buf, tlen) < tlen) { uError("failed to write %d bytes to file %s since %s", tlen, pStore->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); return -1; } - if (taosTWrite(pStore->fd, cont, contLen) < contLen) { + if (taosWrite(pStore->fd, cont, contLen) < contLen) { uError("failed to write %d bytes to file %s since %s", contLen, pStore->fname, strerror(errno)); return -1; } @@ -292,7 +290,7 @@ int tdDropKVStoreRecord(SKVStore *pStore, uint64_t uid) { void *pBuf = buf; tdEncodeKVRecord(&pBuf, &rInfo); - if (taosTWrite(pStore->fd, buf, POINTER_DISTANCE(pBuf, buf)) < POINTER_DISTANCE(pBuf, buf)) { + if (taosWrite(pStore->fd, buf, POINTER_DISTANCE(pBuf, buf)) < POINTER_DISTANCE(pBuf, buf)) { uError("failed to write %" PRId64 " bytes to file %s since %s", (int64_t)(POINTER_DISTANCE(pBuf, buf)), pStore->fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); return -1; @@ -339,7 +337,7 @@ void tsdbGetStoreInfo(char *fname, uint32_t *magic, int64_t *size) { int fd = open(fname, O_RDONLY); if (fd < 0) goto _err; - if (taosTRead(fd, buf, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) goto _err; + if (taosRead(fd, buf, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) goto _err; if (!taosCheckChecksumWhole((uint8_t *)buf, TD_KVSTORE_HEADER_SIZE)) goto _err; void *pBuf = (void *)buf; @@ -368,7 +366,7 @@ static int tdLoadKVStoreHeader(int fd, char *fname, SStoreInfo *pInfo, uint32_t return -1; } - if (taosTRead(fd, buf, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) { + if (taosRead(fd, buf, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) { uError("failed to read %d bytes from file %s since %s", TD_KVSTORE_HEADER_SIZE, fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); return -1; @@ -402,7 +400,7 @@ static int tdUpdateKVStoreHeader(int fd, char *fname, SStoreInfo *pInfo) { ASSERT(POINTER_DISTANCE(pBuf, buf) + sizeof(TSCKSUM) <= TD_KVSTORE_HEADER_SIZE); taosCalcChecksumAppend(0, (uint8_t *)buf, TD_KVSTORE_HEADER_SIZE); - if (taosTWrite(fd, buf, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) { + if (taosWrite(fd, buf, TD_KVSTORE_HEADER_SIZE) < TD_KVSTORE_HEADER_SIZE) { uError("failed to write %d bytes to file %s since %s", TD_KVSTORE_HEADER_SIZE, fname, strerror(errno)); terrno = TAOS_SYSTEM_ERROR(errno); return -1; @@ -535,7 +533,7 @@ static int tdRestoreKVStore(SKVStore *pStore) { ASSERT(pStore->info.size == TD_KVSTORE_HEADER_SIZE); while (true) { - ssize_t tsize = taosTRead(pStore->fd, tbuf, sizeof(SKVRecord)); + int64_t tsize = taosRead(pStore->fd, tbuf, sizeof(SKVRecord)); if (tsize == 0) break; if (tsize < sizeof(SKVRecord)) { uError("failed to read %" PRIzu " bytes from file %s at offset %" PRId64 "since %s", sizeof(SKVRecord), pStore->fname, @@ -598,7 +596,7 @@ static int tdRestoreKVStore(SKVStore *pStore) { goto _err; } - if (taosTRead(pStore->fd, buf, (size_t)pRecord->size) < pRecord->size) { + if (taosRead(pStore->fd, buf, (size_t)pRecord->size) < pRecord->size) { uError("failed to read %" PRId64 " bytes from file %s since %s, offset %" PRId64, pRecord->size, pStore->fname, strerror(errno), pRecord->offset); terrno = TAOS_SYSTEM_ERROR(errno); diff --git a/src/util/src/tlog.c b/src/util/src/tlog.c index 09b0933fd6e32e9b65c8c7acbb81fbfe7d5c005b..0ad7917b393b7a2ddcb055ebb2c3b5a098a804d4 100644 --- a/src/util/src/tlog.c +++ b/src/util/src/tlog.c @@ -336,11 +336,11 @@ static int32_t taosOpenLogFile(char *fn, int32_t maxLines, int32_t maxFileNum) { lseek(tsLogObj.logHandle->fd, 0, SEEK_END); sprintf(name, "==================================================\n"); - taosTWrite(tsLogObj.logHandle->fd, name, (uint32_t)strlen(name)); + taosWrite(tsLogObj.logHandle->fd, name, (uint32_t)strlen(name)); sprintf(name, " new log file \n"); - taosTWrite(tsLogObj.logHandle->fd, name, (uint32_t)strlen(name)); + taosWrite(tsLogObj.logHandle->fd, name, (uint32_t)strlen(name)); sprintf(name, "==================================================\n"); - taosTWrite(tsLogObj.logHandle->fd, name, (uint32_t)strlen(name)); + taosWrite(tsLogObj.logHandle->fd, name, (uint32_t)strlen(name)); return 0; } @@ -390,7 +390,7 @@ void taosPrintLog(const char *flags, int32_t dflag, const char *format, ...) { if (tsAsyncLog) { taosPushLogBuffer(tsLogObj.logHandle, buffer, len); } else { - taosTWrite(tsLogObj.logHandle->fd, buffer, len); + taosWrite(tsLogObj.logHandle->fd, buffer, len); } if (tsLogObj.maxLines > 0) { @@ -400,7 +400,7 @@ void taosPrintLog(const char *flags, int32_t dflag, const char *format, ...) { } } - if (dflag & DEBUG_SCREEN) taosTWrite(1, buffer, (uint32_t)len); + if (dflag & DEBUG_SCREEN) taosWrite(1, buffer, (uint32_t)len); } void taosDumpData(unsigned char *msg, int32_t len) { @@ -419,7 +419,7 @@ void taosDumpData(unsigned char *msg, int32_t len) { pos += 3; if (c >= 16) { temp[pos++] = '\n'; - taosTWrite(tsLogObj.logHandle->fd, temp, (uint32_t)pos); + taosWrite(tsLogObj.logHandle->fd, temp, (uint32_t)pos); c = 0; pos = 0; } @@ -427,9 +427,7 @@ void taosDumpData(unsigned char *msg, int32_t len) { temp[pos++] = '\n'; - taosTWrite(tsLogObj.logHandle->fd, temp, (uint32_t)pos); - - return; + taosWrite(tsLogObj.logHandle->fd, temp, (uint32_t)pos); } void taosPrintLongString(const char *flags, int32_t dflag, const char *format, ...) { @@ -467,7 +465,7 @@ void taosPrintLongString(const char *flags, int32_t dflag, const char *format, . if (tsAsyncLog) { taosPushLogBuffer(tsLogObj.logHandle, buffer, len); } else { - taosTWrite(tsLogObj.logHandle->fd, buffer, len); + taosWrite(tsLogObj.logHandle->fd, buffer, len); } if (tsLogObj.maxLines > 0) { @@ -477,7 +475,7 @@ void taosPrintLongString(const char *flags, int32_t dflag, const char *format, . } } - if (dflag & DEBUG_SCREEN) taosTWrite(1, buffer, (uint32_t)len); + if (dflag & DEBUG_SCREEN) taosWrite(1, buffer, (uint32_t)len); } #if 0 @@ -606,7 +604,7 @@ static void *taosAsyncOutputLog(void *param) { while (1) { log_size = taosPollLogBuffer(tLogBuff, tempBuffer, TSDB_DEFAULT_LOG_BUF_UNIT); if (log_size) { - taosTWrite(tLogBuff->fd, tempBuffer, log_size); + taosWrite(tLogBuff->fd, tempBuffer, log_size); LOG_BUF_START(tLogBuff) = (LOG_BUF_START(tLogBuff) + log_size) % LOG_BUF_SIZE(tLogBuff); } else { break; diff --git a/src/util/src/tnote.c b/src/util/src/tnote.c index 4f05277a847a07831371278915011ebc4da400c6..9536f6fb70f9fe6d74c981274c849cab0c00ef1f 100644 --- a/src/util/src/tnote.c +++ b/src/util/src/tnote.c @@ -265,7 +265,7 @@ void taosNotePrint(taosNoteInfo * pNote, const char * const format, ...) buffer[len] = 0; if (pNote->taosNoteFd >= 0) { - taosTWrite(pNote->taosNoteFd, buffer, (unsigned int)len); + taosWrite(pNote->taosNoteFd, buffer, (unsigned int)len); if (pNote->taosNoteMaxLines > 0) { pNote->taosNoteLines++; diff --git a/src/util/src/tsocket.c b/src/util/src/tsocket.c index 4cf73e6dff8aa534e87145a3e1cb04e48d759c65..1003bc6178754e2eb6a3f8dbc61d0a878022e8f5 100644 --- a/src/util/src/tsocket.c +++ b/src/util/src/tsocket.c @@ -18,7 +18,7 @@ #include "tsocket.h" #include "taoserror.h" -int taosGetFqdn(char *fqdn) { +int32_t taosGetFqdn(char *fqdn) { char hostname[1024]; hostname[1023] = '\0'; if (gethostname(hostname, 1023) == -1) { @@ -26,10 +26,10 @@ int taosGetFqdn(char *fqdn) { return -1; } - struct addrinfo hints = {0}; + struct addrinfo hints = {0}; struct addrinfo *result = NULL; hints.ai_flags = AI_CANONNAME; - int ret = getaddrinfo(hostname, NULL, &hints, &result); + int32_t ret = getaddrinfo(hostname, NULL, &hints, &result); if (!result) { uError("failed to get fqdn, code:%d, reason:%s", ret, gai_strerror(ret)); return -1; @@ -49,10 +49,10 @@ uint32_t taosGetIpFromFqdn(const char *fqdn) { int32_t ret = getaddrinfo(fqdn, NULL, &hints, &result); if (result) { - struct sockaddr *sa = result->ai_addr; - struct sockaddr_in *si = (struct sockaddr_in*)sa; - struct in_addr ia = si->sin_addr; - uint32_t ip = ia.s_addr; + struct sockaddr * sa = result->ai_addr; + struct sockaddr_in *si = (struct sockaddr_in *)sa; + struct in_addr ia = si->sin_addr; + uint32_t ip = ia.s_addr; freeaddrinfo(result); return ip; } else { @@ -70,7 +70,7 @@ uint32_t taosGetIpFromFqdn(const char *fqdn) { } } -// Function converting an IP address string to an unsigned int. +// Function converting an IP address string to an uint32_t. uint32_t ip2uint(const char *const ip_addr) { char ip_addr_cpy[20]; char ip[5]; @@ -81,7 +81,7 @@ uint32_t ip2uint(const char *const ip_addr) { s_start = ip_addr_cpy; s_end = ip_addr_cpy; - int k; + int32_t k; for (k = 0; *s_start != '\0'; s_start = s_end) { for (s_end = s_start; *s_end != '.' && *s_end != '\0'; s_end++) { @@ -95,17 +95,17 @@ uint32_t ip2uint(const char *const ip_addr) { ip[k] = '\0'; - return *((unsigned int *)ip); + return *((uint32_t *)ip); } -int taosWriteMsg(SOCKET fd, void *buf, int nbytes) { - int nleft, nwritten; - char *ptr = (char *)buf; +int32_t taosWriteMsg(SOCKET fd, void *buf, int32_t nbytes) { + int32_t nleft, nwritten; + char * ptr = (char *)buf; nleft = nbytes; while (nleft > 0) { - nwritten = (int)taosWriteSocket(fd, (char *)ptr, (size_t)nleft); + nwritten = (int32_t)taosWriteSocket(fd, (char *)ptr, (size_t)nleft); if (nwritten <= 0) { if (errno == EINTR) continue; @@ -120,16 +120,16 @@ int taosWriteMsg(SOCKET fd, void *buf, int nbytes) { return (nbytes - nleft); } -int taosReadMsg(SOCKET fd, void *buf, int nbytes) { - int nleft, nread; - char *ptr = (char *)buf; +int32_t taosReadMsg(SOCKET fd, void *buf, int32_t nbytes) { + int32_t nleft, nread; + char * ptr = (char *)buf; nleft = nbytes; if (fd < 0) return -1; while (nleft > 0) { - nread = (int)taosReadSocket(fd, ptr, (size_t)nleft); + nread = (int32_t)taosReadSocket(fd, ptr, (size_t)nleft); if (nread == 0) { break; } else if (nread < 0) { @@ -147,11 +147,11 @@ int taosReadMsg(SOCKET fd, void *buf, int nbytes) { return (nbytes - nleft); } -int taosNonblockwrite(SOCKET fd, char *ptr, int nbytes) { +int32_t taosNonblockwrite(SOCKET fd, char *ptr, int32_t nbytes) { taosSetNonblocking(fd, 1); - int nleft, nwritten, nready; - fd_set fset; + int32_t nleft, nwritten, nready; + fd_set fset; struct timeval tv; nleft = nbytes; @@ -160,7 +160,7 @@ int taosNonblockwrite(SOCKET fd, char *ptr, int nbytes) { tv.tv_usec = 0; FD_ZERO(&fset); FD_SET(fd, &fset); - if ((nready = select((int)(fd + 1), NULL, &fset, NULL, &tv)) == 0) { + if ((nready = select((int32_t)(fd + 1), NULL, &fset, NULL, &tv)) == 0) { errno = ETIMEDOUT; uError("fd %d timeout, no enough space to write", fd); break; @@ -172,7 +172,7 @@ int taosNonblockwrite(SOCKET fd, char *ptr, int nbytes) { return -1; } - nwritten = (int)taosSend(fd, ptr, (size_t)nleft, MSG_NOSIGNAL); + nwritten = (int32_t)taosSend(fd, ptr, (size_t)nleft, MSG_NOSIGNAL); if (nwritten <= 0) { if (errno == EAGAIN || errno == EINTR) continue; @@ -189,10 +189,10 @@ int taosNonblockwrite(SOCKET fd, char *ptr, int nbytes) { return (nbytes - nleft); } -int taosReadn(SOCKET fd, char *ptr, int nbytes) { - int nread, nready, nleft = nbytes; +int32_t taosReadn(SOCKET fd, char *ptr, int32_t nbytes) { + int32_t nread, nready, nleft = nbytes; - fd_set fset; + fd_set fset; struct timeval tv; while (nleft > 0) { @@ -200,7 +200,7 @@ int taosReadn(SOCKET fd, char *ptr, int nbytes) { tv.tv_usec = 0; FD_ZERO(&fset); FD_SET(fd, &fset); - if ((nready = select((int)(fd + 1), NULL, &fset, NULL, &tv)) == 0) { + if ((nready = select((int32_t)(fd + 1), NULL, &fset, NULL, &tv)) == 0) { errno = ETIMEDOUT; uError("fd %d timeout\n", fd); break; @@ -210,7 +210,7 @@ int taosReadn(SOCKET fd, char *ptr, int nbytes) { return -1; } - if ((nread = (int)taosReadSocket(fd, ptr, (size_t)nleft)) < 0) { + if ((nread = (int32_t)taosReadSocket(fd, ptr, (size_t)nleft)) < 0) { if (errno == EINTR) continue; uError("read error, %d (%s)", errno, strerror(errno)); return -1; @@ -229,8 +229,8 @@ int taosReadn(SOCKET fd, char *ptr, int nbytes) { SOCKET taosOpenUdpSocket(uint32_t ip, uint16_t port) { struct sockaddr_in localAddr; - SOCKET sockFd; - int bufSize = 1024000; + SOCKET sockFd; + int32_t bufSize = 1024000; uDebug("open udp socket:0x%x:%hu", ip, port); @@ -239,7 +239,7 @@ SOCKET taosOpenUdpSocket(uint32_t ip, uint16_t port) { localAddr.sin_addr.s_addr = ip; localAddr.sin_port = (uint16_t)htons(port); - if ((sockFd = (int)socket(AF_INET, SOCK_DGRAM, 0)) <= 2) { + if ((sockFd = (int32_t)socket(AF_INET, SOCK_DGRAM, 0)) <= 2) { uError("failed to open udp socket: %d (%s)", errno, strerror(errno)); taosCloseSocketNoCheck(sockFd); return -1; @@ -268,9 +268,9 @@ SOCKET taosOpenUdpSocket(uint32_t ip, uint16_t port) { } SOCKET taosOpenTcpClientSocket(uint32_t destIp, uint16_t destPort, uint32_t clientIp) { - SOCKET sockFd = 0; + SOCKET sockFd = 0; + int32_t ret; struct sockaddr_in serverAddr, clientAddr; - int ret; sockFd = socket(PF_INET, SOCK_STREAM, IPPROTO_TCP); @@ -281,7 +281,7 @@ SOCKET taosOpenTcpClientSocket(uint32_t destIp, uint16_t destPort, uint32_t clie } /* set REUSEADDR option, so the portnumber can be re-used */ - int reuse = 1; + int32_t reuse = 1; if (taosSetSockOpt(sockFd, SOL_SOCKET, SO_REUSEADDR, (void *)&reuse, sizeof(reuse)) < 0) { uError("setsockopt SO_REUSEADDR failed: %d (%s)", errno, strerror(errno)); taosCloseSocket(sockFd); @@ -296,8 +296,8 @@ SOCKET taosOpenTcpClientSocket(uint32_t destIp, uint16_t destPort, uint32_t clie /* bind socket to client address */ if (bind(sockFd, (struct sockaddr *)&clientAddr, sizeof(clientAddr)) < 0) { - uError("bind tcp client socket failed, client(0x%x:0), dest(0x%x:%d), reason:(%s)", - clientIp, destIp, destPort, strerror(errno)); + uError("bind tcp client socket failed, client(0x%x:0), dest(0x%x:%d), reason:(%s)", clientIp, destIp, destPort, + strerror(errno)); taosCloseSocket(sockFd); return -1; } @@ -311,7 +311,7 @@ SOCKET taosOpenTcpClientSocket(uint32_t destIp, uint16_t destPort, uint32_t clie ret = connect(sockFd, (struct sockaddr *)&serverAddr, sizeof(serverAddr)); if (ret != 0) { - //uError("failed to connect socket, ip:0x%x, port:%hu(%s)", destIp, destPort, strerror(errno)); + // uError("failed to connect socket, ip:0x%x, port:%hu(%s)", destIp, destPort, strerror(errno)); taosCloseSocket(sockFd); sockFd = -1; } else { @@ -321,36 +321,36 @@ SOCKET taosOpenTcpClientSocket(uint32_t destIp, uint16_t destPort, uint32_t clie return sockFd; } -int taosKeepTcpAlive(SOCKET sockFd) { - int alive = 1; +int32_t taosKeepTcpAlive(SOCKET sockFd) { + int32_t alive = 1; if (taosSetSockOpt(sockFd, SOL_SOCKET, SO_KEEPALIVE, (void *)&alive, sizeof(alive)) < 0) { uError("fd:%d setsockopt SO_KEEPALIVE failed: %d (%s)", sockFd, errno, strerror(errno)); taosCloseSocket(sockFd); return -1; } - int probes = 3; + int32_t probes = 3; if (taosSetSockOpt(sockFd, SOL_TCP, TCP_KEEPCNT, (void *)&probes, sizeof(probes)) < 0) { uError("fd:%d setsockopt SO_KEEPCNT failed: %d (%s)", sockFd, errno, strerror(errno)); taosCloseSocket(sockFd); return -1; } - int alivetime = 10; + int32_t alivetime = 10; if (taosSetSockOpt(sockFd, SOL_TCP, TCP_KEEPIDLE, (void *)&alivetime, sizeof(alivetime)) < 0) { uError("fd:%d setsockopt SO_KEEPIDLE failed: %d (%s)", sockFd, errno, strerror(errno)); taosCloseSocket(sockFd); return -1; } - int interval = 3; + int32_t interval = 3; if (taosSetSockOpt(sockFd, SOL_TCP, TCP_KEEPINTVL, (void *)&interval, sizeof(interval)) < 0) { uError("fd:%d setsockopt SO_KEEPINTVL failed: %d (%s)", sockFd, errno, strerror(errno)); taosCloseSocket(sockFd); return -1; } - int nodelay = 1; + int32_t nodelay = 1; if (taosSetSockOpt(sockFd, IPPROTO_TCP, TCP_NODELAY, (void *)&nodelay, sizeof(nodelay)) < 0) { uError("fd:%d setsockopt TCP_NODELAY failed %d (%s)", sockFd, errno, strerror(errno)); taosCloseSocket(sockFd); @@ -371,8 +371,8 @@ int taosKeepTcpAlive(SOCKET sockFd) { SOCKET taosOpenTcpServerSocket(uint32_t ip, uint16_t port) { struct sockaddr_in serverAdd; - SOCKET sockFd; - int reuse; + SOCKET sockFd; + int32_t reuse; uDebug("open tcp server socket:0x%x:%hu", ip, port); @@ -381,7 +381,7 @@ SOCKET taosOpenTcpServerSocket(uint32_t ip, uint16_t port) { serverAdd.sin_addr.s_addr = ip; serverAdd.sin_port = (uint16_t)htons(port); - if ((sockFd = (int)socket(AF_INET, SOCK_STREAM, IPPROTO_TCP)) <= 2) { + if ((sockFd = (int32_t)socket(AF_INET, SOCK_STREAM, IPPROTO_TCP)) <= 2) { uError("failed to open TCP socket: %d (%s)", errno, strerror(errno)); taosCloseSocketNoCheck(sockFd); return -1; @@ -417,38 +417,38 @@ SOCKET taosOpenTcpServerSocket(uint32_t ip, uint16_t port) { return sockFd; } -void tinet_ntoa(char *ipstr, unsigned int ip) { +void tinet_ntoa(char *ipstr, uint32_t ip) { sprintf(ipstr, "%d.%d.%d.%d", ip & 0xFF, (ip >> 8) & 0xFF, (ip >> 16) & 0xFF, ip >> 24); } #define COPY_SIZE 32768 // sendfile shall be used -int taosCopyFds(SOCKET sfd, SOCKET dfd, int64_t len) { +int32_t taosCopyFds(SOCKET sfd, SOCKET dfd, int64_t len) { int64_t leftLen; - int readLen, writeLen; + int32_t readLen, writeLen; char temp[COPY_SIZE]; leftLen = len; while (leftLen > 0) { if (leftLen < COPY_SIZE) - readLen = (int)leftLen; + readLen = (int32_t)leftLen; else readLen = COPY_SIZE; // 4K - int retLen = taosReadMsg(sfd, temp, (int)readLen); + int32_t retLen = taosReadMsg(sfd, temp, (int32_t)readLen); if (readLen != retLen) { - uError("read error, readLen:%d retLen:%d len:%" PRId64 " leftLen:%" PRId64 ", reason:%s", readLen, retLen, len, leftLen, - strerror(errno)); + uError("read error, readLen:%d retLen:%d len:%" PRId64 " leftLen:%" PRId64 ", reason:%s", readLen, retLen, len, + leftLen, strerror(errno)); return -1; } writeLen = taosWriteMsg(dfd, temp, readLen); if (readLen != writeLen) { - uError("copy error, readLen:%d writeLen:%d len:%" PRId64 " leftLen:%" PRId64 ", reason:%s", readLen, writeLen, len, leftLen, - strerror(errno)); + uError("copy error, readLen:%d writeLen:%d len:%" PRId64 " leftLen:%" PRId64 ", reason:%s", readLen, writeLen, + len, leftLen, strerror(errno)); return -1; } diff --git a/src/wal/inc/walMgmt.h b/src/vnode/inc/vnodeCfg.h similarity index 82% rename from src/wal/inc/walMgmt.h rename to src/vnode/inc/vnodeCfg.h index a23c7f8ec3d0064de23c03a62a81388b63b0f319..c5887fef5d53ca32c215a5f2da9d182386ebc4f0 100644 --- a/src/wal/inc/walMgmt.h +++ b/src/vnode/inc/vnodeCfg.h @@ -13,13 +13,16 @@ * along with this program. If not, see . */ -#ifndef TDENGINE_WAL_MGMT_H -#define TDENGINE_WAL_MGMT_H +#ifndef TDENGINE_VNODE_CFG_H +#define TDENGINE_VNODE_CFG_H #ifdef __cplusplus extern "C" { #endif +int32_t vnodeReadCfg(SVnodeObj *pVnode); +int32_t vnodeWriteCfg(SCreateVnodeMsg *pVnodeCfg); + #ifdef __cplusplus } #endif diff --git a/src/vnode/inc/vnodeVersion.h b/src/vnode/inc/vnodeVersion.h new file mode 100644 index 0000000000000000000000000000000000000000..1d086cb21fdab0247038c7f3d32d89b38c19d871 --- /dev/null +++ b/src/vnode/inc/vnodeVersion.h @@ -0,0 +1,30 @@ +/* + * 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_VERSION_H +#define TDENGINE_VNODE_VERSION_H + +#ifdef __cplusplus +extern "C" { +#endif + +int32_t vnodeReadVersion(SVnodeObj *pVnode); +int32_t vnodeSaveVersion(SVnodeObj *pVnode); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/vnode/src/vnodeCfg.c b/src/vnode/src/vnodeCfg.c new file mode 100644 index 0000000000000000000000000000000000000000..f0040f8cdf8cf52d6cca66602c7bbc61ae931cca --- /dev/null +++ b/src/vnode/src/vnodeCfg.c @@ -0,0 +1,326 @@ +/* + * 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 "taoserror.h" +#include "cJSON.h" +#include "tglobal.h" +#include "tsdb.h" +#include "dnode.h" +#include "vnodeInt.h" +#include "vnodeVersion.h" +#include "vnodeCfg.h" + +static void vnodeLoadCfg(SVnodeObj *pVnode, SCreateVnodeMsg* vnodeMsg) { + strcpy(pVnode->db, vnodeMsg->db); + pVnode->cfgVersion = vnodeMsg->cfg.cfgVersion; + pVnode->tsdbCfg.cacheBlockSize = vnodeMsg->cfg.cacheBlockSize; + pVnode->tsdbCfg.totalBlocks = vnodeMsg->cfg.totalBlocks; + pVnode->tsdbCfg.daysPerFile = vnodeMsg->cfg.daysPerFile; + pVnode->tsdbCfg.keep = vnodeMsg->cfg.daysToKeep; + pVnode->tsdbCfg.keep1 = vnodeMsg->cfg.daysToKeep1; + pVnode->tsdbCfg.keep2 = vnodeMsg->cfg.daysToKeep2; + pVnode->tsdbCfg.minRowsPerFileBlock = vnodeMsg->cfg.minRowsPerFileBlock; + pVnode->tsdbCfg.maxRowsPerFileBlock = vnodeMsg->cfg.maxRowsPerFileBlock; + pVnode->tsdbCfg.precision = vnodeMsg->cfg.precision; + pVnode->tsdbCfg.compression = vnodeMsg->cfg.compression; + pVnode->walCfg.walLevel = vnodeMsg->cfg.walLevel; + pVnode->walCfg.fsyncPeriod = vnodeMsg->cfg.fsyncPeriod; + pVnode->walCfg.wals = vnodeMsg->cfg.wals; + pVnode->walCfg.keep = 0; + pVnode->syncCfg.replica = vnodeMsg->cfg.replications; + pVnode->syncCfg.quorum = vnodeMsg->cfg.quorum; + + for (int i = 0; i < pVnode->syncCfg.replica; ++i) { + SVnodeDesc *node = &vnodeMsg->nodes[i]; + pVnode->syncCfg.nodeInfo[i].nodeId = node->nodeId; + taosGetFqdnPortFromEp(node->nodeEp, pVnode->syncCfg.nodeInfo[i].nodeFqdn, &pVnode->syncCfg.nodeInfo[i].nodePort); + pVnode->syncCfg.nodeInfo[i].nodePort += TSDB_PORT_SYNC; + } + + vInfo("vgId:%d, load vnode cfg successfully, replcia:%d", pVnode->vgId, pVnode->syncCfg.replica); + for (int32_t i = 0; i < pVnode->syncCfg.replica; i++) { + SNodeInfo *node = &pVnode->syncCfg.nodeInfo[i]; + vInfo("vgId:%d, dnode:%d, %s:%u", pVnode->vgId, node->nodeId, node->nodeFqdn, node->nodePort); + } +} + +int32_t vnodeReadCfg(SVnodeObj *pVnode) { + int32_t ret = TSDB_CODE_VND_APP_ERROR; + int32_t len = 0; + int maxLen = 1000; + char * content = calloc(1, maxLen + 1); + cJSON * root = NULL; + FILE * fp = NULL; + bool nodeChanged = false; + SCreateVnodeMsg vnodeMsg; + + char file[TSDB_FILENAME_LEN + 30] = {0}; + sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, pVnode->vgId); + + vnodeMsg.cfg.vgId = pVnode->vgId; + + fp = fopen(file, "r"); + if (!fp) { + vError("vgId:%d, failed to open vnode cfg file:%s to read, error:%s", pVnode->vgId, file, strerror(errno)); + ret = TAOS_SYSTEM_ERROR(errno); + goto PARSE_VCFG_ERROR; + } + + len = fread(content, 1, maxLen, fp); + if (len <= 0) { + vError("vgId:%d, failed to read %s, content is null", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + + content[len] = 0; + root = cJSON_Parse(content); + if (root == NULL) { + vError("vgId:%d, failed to read %s, invalid json format", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + + cJSON *db = cJSON_GetObjectItem(root, "db"); + if (!db || db->type != cJSON_String || db->valuestring == NULL) { + vError("vgId:%d, failed to read %s, db not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + strcpy(vnodeMsg.db, db->valuestring); + + cJSON *cfgVersion = cJSON_GetObjectItem(root, "cfgVersion"); + if (!cfgVersion || cfgVersion->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, cfgVersion not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.cfgVersion = cfgVersion->valueint; + + cJSON *cacheBlockSize = cJSON_GetObjectItem(root, "cacheBlockSize"); + if (!cacheBlockSize || cacheBlockSize->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, cacheBlockSize not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.cacheBlockSize = cacheBlockSize->valueint; + + cJSON *totalBlocks = cJSON_GetObjectItem(root, "totalBlocks"); + if (!totalBlocks || totalBlocks->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, totalBlocks not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.totalBlocks = totalBlocks->valueint; + + cJSON *daysPerFile = cJSON_GetObjectItem(root, "daysPerFile"); + if (!daysPerFile || daysPerFile->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, daysPerFile not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.daysPerFile = daysPerFile->valueint; + + cJSON *daysToKeep = cJSON_GetObjectItem(root, "daysToKeep"); + if (!daysToKeep || daysToKeep->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, daysToKeep not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.daysToKeep = daysToKeep->valueint; + + cJSON *daysToKeep1 = cJSON_GetObjectItem(root, "daysToKeep1"); + if (!daysToKeep1 || daysToKeep1->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, daysToKeep1 not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.daysToKeep1 = daysToKeep1->valueint; + + cJSON *daysToKeep2 = cJSON_GetObjectItem(root, "daysToKeep2"); + if (!daysToKeep2 || daysToKeep2->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, daysToKeep2 not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.daysToKeep2 = daysToKeep2->valueint; + + cJSON *minRowsPerFileBlock = cJSON_GetObjectItem(root, "minRowsPerFileBlock"); + if (!minRowsPerFileBlock || minRowsPerFileBlock->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, minRowsPerFileBlock not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.minRowsPerFileBlock = minRowsPerFileBlock->valueint; + + cJSON *maxRowsPerFileBlock = cJSON_GetObjectItem(root, "maxRowsPerFileBlock"); + if (!maxRowsPerFileBlock || maxRowsPerFileBlock->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, maxRowsPerFileBlock not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.maxRowsPerFileBlock = maxRowsPerFileBlock->valueint; + + cJSON *precision = cJSON_GetObjectItem(root, "precision"); + if (!precision || precision->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, precision not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.precision = (int8_t)precision->valueint; + + cJSON *compression = cJSON_GetObjectItem(root, "compression"); + if (!compression || compression->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, compression not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.compression = (int8_t)compression->valueint; + + cJSON *walLevel = cJSON_GetObjectItem(root, "walLevel"); + if (!walLevel || walLevel->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, walLevel not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.walLevel = (int8_t)walLevel->valueint; + + cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsync"); + if (!walLevel || walLevel->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, fsyncPeriod not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.fsyncPeriod = fsyncPeriod->valueint; + + cJSON *wals = cJSON_GetObjectItem(root, "wals"); + if (!wals || wals->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, wals not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.wals = (int8_t)wals->valueint; + + cJSON *replica = cJSON_GetObjectItem(root, "replica"); + if (!replica || replica->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, replica not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.replications = (int8_t)replica->valueint; + + cJSON *quorum = cJSON_GetObjectItem(root, "quorum"); + if (!quorum || quorum->type != cJSON_Number) { + vError("vgId: %d, failed to read %s, quorum not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + vnodeMsg.cfg.quorum = (int8_t)quorum->valueint; + + cJSON *nodeInfos = cJSON_GetObjectItem(root, "nodeInfos"); + if (!nodeInfos || nodeInfos->type != cJSON_Array) { + vError("vgId:%d, failed to read %s, nodeInfos not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + + int size = cJSON_GetArraySize(nodeInfos); + if (size != vnodeMsg.cfg.replications) { + vError("vgId:%d, failed to read %s, nodeInfos size not matched", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + + for (int i = 0; i < size; ++i) { + cJSON *nodeInfo = cJSON_GetArrayItem(nodeInfos, i); + if (nodeInfo == NULL) continue; + SVnodeDesc *node = &vnodeMsg.nodes[i]; + + cJSON *nodeId = cJSON_GetObjectItem(nodeInfo, "nodeId"); + if (!nodeId || nodeId->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, nodeId not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + node->nodeId = nodeId->valueint; + + cJSON *nodeEp = cJSON_GetObjectItem(nodeInfo, "nodeEp"); + if (!nodeEp || nodeEp->type != cJSON_String || nodeEp->valuestring == NULL) { + vError("vgId:%d, failed to read %s, nodeFqdn not found", pVnode->vgId, file); + goto PARSE_VCFG_ERROR; + } + tstrncpy(node->nodeEp, nodeEp->valuestring, TSDB_EP_LEN); + + if (!nodeChanged) { + nodeChanged = dnodeCheckEpChanged(node->nodeId, node->nodeEp); + } + } + + ret = TSDB_CODE_SUCCESS; + +PARSE_VCFG_ERROR: + if (content != NULL) free(content); + if (root != NULL) cJSON_Delete(root); + if (fp != NULL) fclose(fp); + + if (nodeChanged) { + vnodeWriteCfg(&vnodeMsg); + } + + if (ret == TSDB_CODE_SUCCESS) { + vnodeLoadCfg(pVnode, &vnodeMsg); + } + + terrno = 0; + return ret; +} + +int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) { + char file[TSDB_FILENAME_LEN + 30] = {0}; + sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, pMsg->cfg.vgId); + + FILE *fp = fopen(file, "w"); + if (!fp) { + vError("vgId:%d, failed to write %s error:%s", pMsg->cfg.vgId, file, strerror(errno)); + terrno = TAOS_SYSTEM_ERROR(errno); + return terrno; + } + + int32_t len = 0; + int32_t maxLen = 1000; + char * content = calloc(1, maxLen + 1); + + len += snprintf(content + len, maxLen - len, "{\n"); + len += snprintf(content + len, maxLen - len, " \"db\": \"%s\",\n", pMsg->db); + len += snprintf(content + len, maxLen - len, " \"cfgVersion\": %d,\n", pMsg->cfg.cfgVersion); + len += snprintf(content + len, maxLen - len, " \"cacheBlockSize\": %d,\n", pMsg->cfg.cacheBlockSize); + len += snprintf(content + len, maxLen - len, " \"totalBlocks\": %d,\n", pMsg->cfg.totalBlocks); + len += snprintf(content + len, maxLen - len, " \"daysPerFile\": %d,\n", pMsg->cfg.daysPerFile); + len += snprintf(content + len, maxLen - len, " \"daysToKeep\": %d,\n", pMsg->cfg.daysToKeep); + len += snprintf(content + len, maxLen - len, " \"daysToKeep1\": %d,\n", pMsg->cfg.daysToKeep1); + len += snprintf(content + len, maxLen - len, " \"daysToKeep2\": %d,\n", pMsg->cfg.daysToKeep2); + len += snprintf(content + len, maxLen - len, " \"minRowsPerFileBlock\": %d,\n", pMsg->cfg.minRowsPerFileBlock); + len += snprintf(content + len, maxLen - len, " \"maxRowsPerFileBlock\": %d,\n", pMsg->cfg.maxRowsPerFileBlock); + len += snprintf(content + len, maxLen - len, " \"precision\": %d,\n", pMsg->cfg.precision); + len += snprintf(content + len, maxLen - len, " \"compression\": %d,\n", pMsg->cfg.compression); + len += snprintf(content + len, maxLen - len, " \"walLevel\": %d,\n", pMsg->cfg.walLevel); + len += snprintf(content + len, maxLen - len, " \"fsync\": %d,\n", pMsg->cfg.fsyncPeriod); + len += snprintf(content + len, maxLen - len, " \"replica\": %d,\n", pMsg->cfg.replications); + len += snprintf(content + len, maxLen - len, " \"wals\": %d,\n", pMsg->cfg.wals); + len += snprintf(content + len, maxLen - len, " \"quorum\": %d,\n", pMsg->cfg.quorum); + len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n"); + for (int32_t i = 0; i < pMsg->cfg.replications; i++) { + SVnodeDesc *node = &pMsg->nodes[i]; + dnodeUpdateEp(node->nodeId, node->nodeEp, NULL, NULL); + len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", node->nodeId); + len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", node->nodeEp); + if (i < pMsg->cfg.replications - 1) { + len += snprintf(content + len, maxLen - len, " },{\n"); + } else { + len += snprintf(content + len, maxLen - len, " }]\n"); + } + } + len += snprintf(content + len, maxLen - len, "}\n"); + + fwrite(content, 1, len, fp); + fflush(fp); + fclose(fp); + free(content); + terrno = 0; + + vInfo("vgId:%d, successed to write %s", pMsg->cfg.vgId, file); + return TSDB_CODE_SUCCESS; +} diff --git a/src/vnode/src/vnodeMain.c b/src/vnode/src/vnodeMain.c index 9c283dd9c4a941fe2fbbfa43fe95e549b92ae898..e2069331168b57af64155b968453a3e4eb120056 100644 --- a/src/vnode/src/vnodeMain.c +++ b/src/vnode/src/vnodeMain.c @@ -31,18 +31,14 @@ #include "vnodeInt.h" #include "query.h" #include "dnode.h" - -#define TSDB_VNODE_VERSION_CONTENT_LEN 31 +#include "vnodeCfg.h" +#include "vnodeVersion.h" static SHashObj*tsDnodeVnodesHash; static void vnodeCleanUp(SVnodeObj *pVnode); -static int32_t vnodeSaveCfg(SMDCreateVnodeMsg *pVnodeCfg); -static int32_t vnodeReadCfg(SVnodeObj *pVnode); -static int32_t vnodeSaveVersion(SVnodeObj *pVnode); -static int32_t vnodeReadVersion(SVnodeObj *pVnode); static int vnodeProcessTsdbStatus(void *arg, int status); static uint32_t vnodeGetFileInfo(void *ahandle, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fversion); -static int vnodeGetWalInfo(void *ahandle, char *name, uint32_t *index); +static int vnodeGetWalInfo(void *ahandle, char *fileName, int64_t *fileId); static void vnodeNotifyRole(void *ahandle, int8_t role); static void vnodeCtrlFlow(void *handle, int32_t mseconds); static int vnodeNotifyFileSynced(void *ahandle, uint64_t fversion); @@ -90,7 +86,7 @@ void vnodeCleanupResources() { syncCleanUp(); } -int32_t vnodeCreate(SMDCreateVnodeMsg *pVnodeCfg) { +int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) { int32_t code; SVnodeObj *pVnode = vnodeAcquire(pVnodeCfg->cfg.vgId); @@ -128,7 +124,7 @@ int32_t vnodeCreate(SMDCreateVnodeMsg *pVnodeCfg) { } } - code = vnodeSaveCfg(pVnodeCfg); + code = vnodeWriteCfg(pVnodeCfg); if (code != TSDB_CODE_SUCCESS) { vError("vgId:%d, failed to save vnode cfg, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(code)); return code; @@ -138,7 +134,6 @@ int32_t vnodeCreate(SMDCreateVnodeMsg *pVnodeCfg) { tsdbCfg.tsdbId = pVnodeCfg->cfg.vgId; tsdbCfg.cacheBlockSize = pVnodeCfg->cfg.cacheBlockSize; tsdbCfg.totalBlocks = pVnodeCfg->cfg.totalBlocks; - // tsdbCfg.maxTables = pVnodeCfg->cfg.maxTables; tsdbCfg.daysPerFile = pVnodeCfg->cfg.daysPerFile; tsdbCfg.keep = pVnodeCfg->cfg.daysToKeep; tsdbCfg.minRowsPerFileBlock = pVnodeCfg->cfg.minRowsPerFileBlock; @@ -176,7 +171,7 @@ int32_t vnodeDrop(int32_t vgId) { return TSDB_CODE_SUCCESS; } -int32_t vnodeAlter(void *param, SMDCreateVnodeMsg *pVnodeCfg) { +int32_t vnodeAlter(void *param, SCreateVnodeMsg *pVnodeCfg) { SVnodeObj *pVnode = param; // vnode in non-ready state and still needs to return success instead of TSDB_CODE_VND_INVALID_STATUS @@ -186,7 +181,7 @@ int32_t vnodeAlter(void *param, SMDCreateVnodeMsg *pVnodeCfg) { return TSDB_CODE_SUCCESS; } - int32_t code = vnodeSaveCfg(pVnodeCfg); + int32_t code = vnodeWriteCfg(pVnodeCfg); if (code != TSDB_CODE_SUCCESS) { pVnode->status = TAOS_VN_STATUS_READY; return code; @@ -259,8 +254,8 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) { pVnode->fversion = pVnode->version; - pVnode->wqueue = dnodeAllocateVnodeWqueue(pVnode); - pVnode->rqueue = dnodeAllocateVnodeRqueue(pVnode); + pVnode->wqueue = dnodeAllocVWriteQueue(pVnode); + pVnode->rqueue = dnodeAllocVReadQueue(pVnode); if (pVnode->wqueue == NULL || pVnode->rqueue == NULL) { vnodeCleanUp(pVnode); return terrno; @@ -303,6 +298,7 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) { } sprintf(temp, "%s/wal", rootDir); + pVnode->walCfg.vgId = pVnode->vgId; pVnode->wal = walOpen(temp, &pVnode->walCfg); if (pVnode->wal == NULL) { vnodeCleanUp(pVnode); @@ -314,6 +310,8 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) { pVnode->version = walGetVersion(pVnode->wal); } + walRenew(pVnode->wal); + SSyncInfo syncInfo; syncInfo.vgId = pVnode->vgId; syncInfo.version = pVnode->version; @@ -323,7 +321,7 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) { syncInfo.getWalInfo = vnodeGetWalInfo; syncInfo.getFileInfo = vnodeGetFileInfo; syncInfo.writeToCache = vnodeWriteToQueue; - syncInfo.confirmForward = dnodeSendRpcVnodeWriteRsp; + syncInfo.confirmForward = dnodeSendRpcVWriteRsp; syncInfo.notifyRole = vnodeNotifyRole; syncInfo.notifyFlowCtrl = vnodeCtrlFlow; syncInfo.notifyFileSynced = vnodeNotifyFileSynced; @@ -388,6 +386,10 @@ void vnodeRelease(void *pVnodeRaw) { pVnode->qMgmt = NULL; } + if (pVnode->wal) { + walStop(pVnode->wal); + } + if (pVnode->tsdb) { tsdbCloseRepo(pVnode->tsdb, 1); pVnode->tsdb = NULL; @@ -406,12 +408,12 @@ void vnodeRelease(void *pVnodeRaw) { } if (pVnode->wqueue) { - dnodeFreeVnodeWqueue(pVnode->wqueue); + dnodeFreeVWriteQueue(pVnode->wqueue); pVnode->wqueue = NULL; } if (pVnode->rqueue) { - dnodeFreeVnodeRqueue(pVnode->rqueue); + dnodeFreeVReadQueue(pVnode->rqueue); pVnode->rqueue = NULL; } @@ -499,7 +501,7 @@ void *vnodeGetWal(void *pVnode) { return ((SVnodeObj *)pVnode)->wal; } -static void vnodeBuildVloadMsg(SVnodeObj *pVnode, SDMStatusMsg *pStatus) { +static void vnodeBuildVloadMsg(SVnodeObj *pVnode, SStatusMsg *pStatus) { int64_t totalStorage = 0; int64_t compStorage = 0; int64_t pointsWritten = 0; @@ -543,7 +545,7 @@ int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) { } void vnodeBuildStatusMsg(void *param) { - SDMStatusMsg *pStatus = param; + SStatusMsg *pStatus = param; SHashMutableIterator *pIter = taosHashCreateIter(tsDnodeVnodesHash); while (taosHashIterNext(pIter)) { @@ -557,7 +559,7 @@ void vnodeBuildStatusMsg(void *param) { taosHashDestroyIter(pIter); } -void vnodeSetAccess(SDMVgroupAccess *pAccess, int32_t numOfVnodes) { +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); @@ -605,25 +607,27 @@ static int vnodeProcessTsdbStatus(void *arg, int status) { SVnodeObj *pVnode = arg; if (status == TSDB_STATUS_COMMIT_START) { - pVnode->fversion = pVnode->version; + pVnode->fversion = pVnode->version; return walRenew(pVnode->wal); } - if (status == TSDB_STATUS_COMMIT_OVER) + if (status == TSDB_STATUS_COMMIT_OVER) { return vnodeSaveVersion(pVnode); + } - return 0; + return 0; } -static uint32_t vnodeGetFileInfo(void *ahandle, char *name, uint32_t *index, uint32_t eindex, int64_t *size, uint64_t *fversion) { +static uint32_t vnodeGetFileInfo(void *ahandle, char *name, uint32_t *index, uint32_t eindex, int64_t *size, + uint64_t *fversion) { SVnodeObj *pVnode = ahandle; *fversion = pVnode->fversion; return tsdbGetFileInfo(pVnode->tsdb, name, index, eindex, size); } -static int vnodeGetWalInfo(void *ahandle, char *name, uint32_t *index) { +static int vnodeGetWalInfo(void *ahandle, char *fileName, int64_t *fileId) { SVnodeObj *pVnode = ahandle; - return walGetWalFile(pVnode->wal, name, index); + return walGetWalFile(pVnode->wal, fileName, fileId); } static void vnodeNotifyRole(void *ahandle, int8_t role) { @@ -640,17 +644,18 @@ static void vnodeNotifyRole(void *ahandle, int8_t role) { static void vnodeCtrlFlow(void *ahandle, int32_t mseconds) { SVnodeObj *pVnode = ahandle; - if (pVnode->delay != mseconds) + if (pVnode->delay != mseconds) { vInfo("vgId:%d, sync flow control, mseconds:%d", pVnode->vgId, mseconds); + } pVnode->delay = mseconds; } -static int vnodeResetTsdb(SVnodeObj *pVnode) -{ +static int vnodeResetTsdb(SVnodeObj *pVnode) { char rootDir[128] = "\0"; sprintf(rootDir, "%s/tsdb", pVnode->rootDir); - if (atomic_val_compare_exchange_8(&pVnode->status, TAOS_VN_STATUS_READY, TAOS_VN_STATUS_RESET) != TAOS_VN_STATUS_READY) { + if (atomic_val_compare_exchange_8(&pVnode->status, TAOS_VN_STATUS_READY, TAOS_VN_STATUS_RESET) != + TAOS_VN_STATUS_READY) { return -1; } @@ -675,7 +680,7 @@ static int vnodeResetTsdb(SVnodeObj *pVnode) pVnode->tsdb = tsdbOpenRepo(rootDir, &appH); pVnode->status = TAOS_VN_STATUS_READY; - vnodeRelease(pVnode); + vnodeRelease(pVnode); return 0; } @@ -690,360 +695,3 @@ static int vnodeNotifyFileSynced(void *ahandle, uint64_t fversion) { return vnodeResetTsdb(pVnode); } - -static int32_t vnodeSaveCfg(SMDCreateVnodeMsg *pVnodeCfg) { - char cfgFile[TSDB_FILENAME_LEN + 30] = {0}; - sprintf(cfgFile, "%s/vnode%d/config.json", tsVnodeDir, pVnodeCfg->cfg.vgId); - FILE *fp = fopen(cfgFile, "w"); - if (!fp) { - vError("vgId:%d, failed to open vnode cfg file for write, file:%s error:%s", pVnodeCfg->cfg.vgId, cfgFile, - strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - return terrno; - } - - int32_t len = 0; - int32_t maxLen = 1000; - char * content = calloc(1, maxLen + 1); - if (content == NULL) { - fclose(fp); - return TSDB_CODE_VND_OUT_OF_MEMORY; - } - - len += snprintf(content + len, maxLen - len, "{\n"); - len += snprintf(content + len, maxLen - len, " \"db\": \"%s\",\n", pVnodeCfg->db); - len += snprintf(content + len, maxLen - len, " \"cfgVersion\": %d,\n", pVnodeCfg->cfg.cfgVersion); - len += snprintf(content + len, maxLen - len, " \"cacheBlockSize\": %d,\n", pVnodeCfg->cfg.cacheBlockSize); - len += snprintf(content + len, maxLen - len, " \"totalBlocks\": %d,\n", pVnodeCfg->cfg.totalBlocks); - // len += snprintf(content + len, maxLen - len, " \"maxTables\": %d,\n", pVnodeCfg->cfg.maxTables); - len += snprintf(content + len, maxLen - len, " \"daysPerFile\": %d,\n", pVnodeCfg->cfg.daysPerFile); - len += snprintf(content + len, maxLen - len, " \"daysToKeep\": %d,\n", pVnodeCfg->cfg.daysToKeep); - len += snprintf(content + len, maxLen - len, " \"daysToKeep1\": %d,\n", pVnodeCfg->cfg.daysToKeep1); - len += snprintf(content + len, maxLen - len, " \"daysToKeep2\": %d,\n", pVnodeCfg->cfg.daysToKeep2); - len += snprintf(content + len, maxLen - len, " \"minRowsPerFileBlock\": %d,\n", pVnodeCfg->cfg.minRowsPerFileBlock); - len += snprintf(content + len, maxLen - len, " \"maxRowsPerFileBlock\": %d,\n", pVnodeCfg->cfg.maxRowsPerFileBlock); - // len += snprintf(content + len, maxLen - len, " \"commitTime\": %d,\n", pVnodeCfg->cfg.commitTime); - len += snprintf(content + len, maxLen - len, " \"precision\": %d,\n", pVnodeCfg->cfg.precision); - len += snprintf(content + len, maxLen - len, " \"compression\": %d,\n", pVnodeCfg->cfg.compression); - len += snprintf(content + len, maxLen - len, " \"walLevel\": %d,\n", pVnodeCfg->cfg.walLevel); - len += snprintf(content + len, maxLen - len, " \"fsync\": %d,\n", pVnodeCfg->cfg.fsyncPeriod); - len += snprintf(content + len, maxLen - len, " \"replica\": %d,\n", pVnodeCfg->cfg.replications); - len += snprintf(content + len, maxLen - len, " \"wals\": %d,\n", pVnodeCfg->cfg.wals); - len += snprintf(content + len, maxLen - len, " \"quorum\": %d,\n", pVnodeCfg->cfg.quorum); - - len += snprintf(content + len, maxLen - len, " \"nodeInfos\": [{\n"); - - vInfo("vgId:%d, save vnode cfg, replica:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.replications); - for (int32_t i = 0; i < pVnodeCfg->cfg.replications; i++) { - len += snprintf(content + len, maxLen - len, " \"nodeId\": %d,\n", pVnodeCfg->nodes[i].nodeId); - len += snprintf(content + len, maxLen - len, " \"nodeEp\": \"%s\"\n", pVnodeCfg->nodes[i].nodeEp); - vInfo("vgId:%d, save vnode cfg, nodeId:%d nodeEp:%s", pVnodeCfg->cfg.vgId, pVnodeCfg->nodes[i].nodeId, - pVnodeCfg->nodes[i].nodeEp); - - if (i < pVnodeCfg->cfg.replications - 1) { - len += snprintf(content + len, maxLen - len, " },{\n"); - } else { - len += snprintf(content + len, maxLen - len, " }]\n"); - } - } - len += snprintf(content + len, maxLen - len, "}\n"); - - fwrite(content, 1, len, fp); - fflush(fp); - fclose(fp); - free(content); - - vInfo("vgId:%d, save vnode cfg successed", pVnodeCfg->cfg.vgId); - - return TSDB_CODE_SUCCESS; -} - -static int32_t vnodeReadCfg(SVnodeObj *pVnode) { - cJSON *root = NULL; - char *content = NULL; - char cfgFile[TSDB_FILENAME_LEN + 30] = {0}; - int maxLen = 1000; - - terrno = TSDB_CODE_VND_APP_ERROR; - sprintf(cfgFile, "%s/vnode%d/config.json", tsVnodeDir, pVnode->vgId); - FILE *fp = fopen(cfgFile, "r"); - if (!fp) { - vError("vgId:%d, failed to open vnode cfg file:%s to read, error:%s", pVnode->vgId, - cfgFile, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - goto PARSE_OVER; - } - - content = calloc(1, maxLen + 1); - if (content == NULL) goto PARSE_OVER; - int len = fread(content, 1, maxLen, fp); - if (len <= 0) { - vError("vgId:%d, failed to read vnode cfg, content is null", pVnode->vgId); - free(content); - fclose(fp); - return errno; - } - - root = cJSON_Parse(content); - if (root == NULL) { - vError("vgId:%d, failed to read vnode cfg, invalid json format", pVnode->vgId); - goto PARSE_OVER; - } - - cJSON *db = cJSON_GetObjectItem(root, "db"); - if (!db || db->type != cJSON_String || db->valuestring == NULL) { - vError("vgId:%d, failed to read vnode cfg, db not found", pVnode->vgId); - goto PARSE_OVER; - } - strcpy(pVnode->db, db->valuestring); - - cJSON *cfgVersion = cJSON_GetObjectItem(root, "cfgVersion"); - if (!cfgVersion || cfgVersion->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, cfgVersion not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->cfgVersion = cfgVersion->valueint; - - cJSON *cacheBlockSize = cJSON_GetObjectItem(root, "cacheBlockSize"); - if (!cacheBlockSize || cacheBlockSize->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, cacheBlockSize not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.cacheBlockSize = cacheBlockSize->valueint; - - cJSON *totalBlocks = cJSON_GetObjectItem(root, "totalBlocks"); - if (!totalBlocks || totalBlocks->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, totalBlocks not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.totalBlocks = totalBlocks->valueint; - - // cJSON *maxTables = cJSON_GetObjectItem(root, "maxTables"); - // if (!maxTables || maxTables->type != cJSON_Number) { - // vError("vgId:%d, failed to read vnode cfg, maxTables not found", pVnode->vgId); - // goto PARSE_OVER; - // } - // pVnode->tsdbCfg.maxTables = maxTables->valueint; - - cJSON *daysPerFile = cJSON_GetObjectItem(root, "daysPerFile"); - if (!daysPerFile || daysPerFile->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, daysPerFile not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.daysPerFile = daysPerFile->valueint; - - cJSON *daysToKeep = cJSON_GetObjectItem(root, "daysToKeep"); - if (!daysToKeep || daysToKeep->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, daysToKeep not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.keep = daysToKeep->valueint; - - cJSON *daysToKeep1 = cJSON_GetObjectItem(root, "daysToKeep1"); - if (!daysToKeep1 || daysToKeep1->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, daysToKeep1 not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.keep1 = daysToKeep1->valueint; - - cJSON *daysToKeep2 = cJSON_GetObjectItem(root, "daysToKeep2"); - if (!daysToKeep2 || daysToKeep2->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, daysToKeep2 not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.keep2 = daysToKeep2->valueint; - - cJSON *minRowsPerFileBlock = cJSON_GetObjectItem(root, "minRowsPerFileBlock"); - if (!minRowsPerFileBlock || minRowsPerFileBlock->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, minRowsPerFileBlock not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.minRowsPerFileBlock = minRowsPerFileBlock->valueint; - - cJSON *maxRowsPerFileBlock = cJSON_GetObjectItem(root, "maxRowsPerFileBlock"); - if (!maxRowsPerFileBlock || maxRowsPerFileBlock->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, maxRowsPerFileBlock not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.maxRowsPerFileBlock = maxRowsPerFileBlock->valueint; - - // cJSON *commitTime = cJSON_GetObjectItem(root, "commitTime"); - // if (!commitTime || commitTime->type != cJSON_Number) { - // vError("vgId:%d, failed to read vnode cfg, commitTime not found", pVnode->vgId); - // goto PARSE_OVER; - // } - // pVnode->tsdbCfg.commitTime = (int8_t)commitTime->valueint; - - cJSON *precision = cJSON_GetObjectItem(root, "precision"); - if (!precision || precision->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, precision not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.precision = (int8_t)precision->valueint; - - cJSON *compression = cJSON_GetObjectItem(root, "compression"); - if (!compression || compression->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, compression not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->tsdbCfg.compression = (int8_t)compression->valueint; - - cJSON *walLevel = cJSON_GetObjectItem(root, "walLevel"); - if (!walLevel || walLevel->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, walLevel not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->walCfg.walLevel = (int8_t) walLevel->valueint; - - cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsync"); - if (!walLevel || walLevel->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, fsyncPeriod not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->walCfg.fsyncPeriod = fsyncPeriod->valueint; - - cJSON *wals = cJSON_GetObjectItem(root, "wals"); - if (!wals || wals->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, wals not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->walCfg.wals = (int8_t)wals->valueint; - pVnode->walCfg.keep = 0; - - cJSON *replica = cJSON_GetObjectItem(root, "replica"); - if (!replica || replica->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, replica not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->syncCfg.replica = (int8_t)replica->valueint; - - cJSON *quorum = cJSON_GetObjectItem(root, "quorum"); - if (!quorum || quorum->type != cJSON_Number) { - vError("vgId: %d, failed to read vnode cfg, quorum not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->syncCfg.quorum = (int8_t)quorum->valueint; - - cJSON *nodeInfos = cJSON_GetObjectItem(root, "nodeInfos"); - if (!nodeInfos || nodeInfos->type != cJSON_Array) { - vError("vgId:%d, failed to read vnode cfg, nodeInfos not found", pVnode->vgId); - goto PARSE_OVER; - } - - int size = cJSON_GetArraySize(nodeInfos); - if (size != pVnode->syncCfg.replica) { - vError("vgId:%d, failed to read vnode cfg, nodeInfos size not matched", pVnode->vgId); - goto PARSE_OVER; - } - - for (int i = 0; i < size; ++i) { - cJSON *nodeInfo = cJSON_GetArrayItem(nodeInfos, i); - if (nodeInfo == NULL) continue; - - cJSON *nodeId = cJSON_GetObjectItem(nodeInfo, "nodeId"); - if (!nodeId || nodeId->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode cfg, nodeId not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->syncCfg.nodeInfo[i].nodeId = nodeId->valueint; - - cJSON *nodeEp = cJSON_GetObjectItem(nodeInfo, "nodeEp"); - if (!nodeEp || nodeEp->type != cJSON_String || nodeEp->valuestring == NULL) { - vError("vgId:%d, failed to read vnode cfg, nodeFqdn not found", pVnode->vgId); - goto PARSE_OVER; - } - - taosGetFqdnPortFromEp(nodeEp->valuestring, pVnode->syncCfg.nodeInfo[i].nodeFqdn, &pVnode->syncCfg.nodeInfo[i].nodePort); - pVnode->syncCfg.nodeInfo[i].nodePort += TSDB_PORT_SYNC; - } - - terrno = TSDB_CODE_SUCCESS; - - vInfo("vgId:%d, read vnode cfg successfully, replcia:%d", pVnode->vgId, pVnode->syncCfg.replica); - for (int32_t i = 0; i < pVnode->syncCfg.replica; i++) { - vInfo("vgId:%d, dnode:%d, %s:%d", pVnode->vgId, pVnode->syncCfg.nodeInfo[i].nodeId, - pVnode->syncCfg.nodeInfo[i].nodeFqdn, pVnode->syncCfg.nodeInfo[i].nodePort); - } - -PARSE_OVER: - taosTFree(content); - cJSON_Delete(root); - if (fp) fclose(fp); - return terrno; -} - -static int32_t vnodeSaveVersion(SVnodeObj *pVnode) { - char versionFile[TSDB_FILENAME_LEN + 30] = {0}; - sprintf(versionFile, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId); - FILE *fp = fopen(versionFile, "w"); - if (!fp) { - vError("vgId:%d, failed to open vnode version file for write, file:%s error:%s", pVnode->vgId, - versionFile, strerror(errno)); - return TAOS_SYSTEM_ERROR(errno); - } - - int32_t len = 0; - int32_t maxLen = 30; - char content[TSDB_VNODE_VERSION_CONTENT_LEN] = {0}; - - len += snprintf(content + len, maxLen - len, "{\n"); - len += snprintf(content + len, maxLen - len, " \"version\": %" PRId64 "\n", pVnode->fversion); - len += snprintf(content + len, maxLen - len, "}\n"); - - fwrite(content, 1, len, fp); - fflush(fp); - fclose(fp); - - vInfo("vgId:%d, save vnode version:%" PRId64 " succeed", pVnode->vgId, pVnode->fversion); - - return TSDB_CODE_SUCCESS; -} - -static int32_t vnodeReadVersion(SVnodeObj *pVnode) { - char versionFile[TSDB_FILENAME_LEN + 30] = {0}; - char *content = NULL; - cJSON *root = NULL; - int maxLen = 100; - - terrno = TSDB_CODE_VND_INVALID_VRESION_FILE; - sprintf(versionFile, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId); - FILE *fp = fopen(versionFile, "r"); - if (!fp) { - if (errno != ENOENT) { - vError("vgId:%d, failed to open version file:%s error:%s", pVnode->vgId, versionFile, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - } else { - terrno = TSDB_CODE_SUCCESS; - } - goto PARSE_OVER; - } - - content = calloc(1, maxLen + 1); - int len = fread(content, 1, maxLen, fp); - if (len <= 0) { - vError("vgId:%d, failed to read vnode version, content is null", pVnode->vgId); - goto PARSE_OVER; - } - - root = cJSON_Parse(content); - if (root == NULL) { - vError("vgId:%d, failed to read vnode version, invalid json format", pVnode->vgId); - goto PARSE_OVER; - } - - cJSON *ver = cJSON_GetObjectItem(root, "version"); - if (!ver || ver->type != cJSON_Number) { - vError("vgId:%d, failed to read vnode version, version not found", pVnode->vgId); - goto PARSE_OVER; - } - pVnode->version = ver->valueint; - - terrno = TSDB_CODE_SUCCESS; - vInfo("vgId:%d, read vnode version successfully, version:%" PRId64, pVnode->vgId, pVnode->version); - -PARSE_OVER: - taosTFree(content); - cJSON_Delete(root); - if (fp) fclose(fp); - return terrno; -} diff --git a/src/vnode/src/vnodeVersion.c b/src/vnode/src/vnodeVersion.c new file mode 100644 index 0000000000000000000000000000000000000000..1d0695fb53406da3aa1efca93406a9d9015b0527 --- /dev/null +++ b/src/vnode/src/vnodeVersion.c @@ -0,0 +1,103 @@ +/* + * 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 "cJSON.h" +#include "tglobal.h" +#include "tsdb.h" +#include "vnodeInt.h" +#include "vnodeVersion.h" + +int32_t vnodeReadVersion(SVnodeObj *pVnode) { + int32_t len = 0; + int32_t maxLen = 100; + char * content = calloc(1, maxLen + 1); + cJSON * root = NULL; + FILE * fp = NULL; + + terrno = TSDB_CODE_VND_INVALID_VRESION_FILE; + char file[TSDB_FILENAME_LEN + 30] = {0}; + sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId); + + fp = fopen(file, "r"); + if (!fp) { + if (errno != ENOENT) { + vError("vgId:%d, failed to read %s, error:%s", pVnode->vgId, file, strerror(errno)); + terrno = TAOS_SYSTEM_ERROR(errno); + } else { + terrno = TSDB_CODE_SUCCESS; + } + goto PARSE_VER_ERROR; + } + + len = fread(content, 1, maxLen, fp); + if (len <= 0) { + vError("vgId:%d, failed to read %s, content is null", pVnode->vgId, file); + goto PARSE_VER_ERROR; + } + + root = cJSON_Parse(content); + if (root == NULL) { + vError("vgId:%d, failed to read %s, invalid json format", pVnode->vgId, file); + goto PARSE_VER_ERROR; + } + + cJSON *ver = cJSON_GetObjectItem(root, "version"); + if (!ver || ver->type != cJSON_Number) { + vError("vgId:%d, failed to read %s, version not found", pVnode->vgId, file); + goto PARSE_VER_ERROR; + } + pVnode->version = ver->valueint; + + terrno = TSDB_CODE_SUCCESS; + vInfo("vgId:%d, read %s successfully, version:%" PRId64, pVnode->vgId, file, pVnode->version); + +PARSE_VER_ERROR: + if (content != NULL) free(content); + if (root != NULL) cJSON_Delete(root); + if (fp != NULL) fclose(fp); + + return terrno; +} + +int32_t vnodeSaveVersion(SVnodeObj *pVnode) { + char file[TSDB_FILENAME_LEN + 30] = {0}; + sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId); + + FILE *fp = fopen(file, "w"); + if (!fp) { + vError("vgId:%d, failed to write %s, reason:%s", pVnode->vgId, file, strerror(errno)); + return -1; + } + + int32_t len = 0; + int32_t maxLen = 100; + char * content = calloc(1, maxLen + 1); + + len += snprintf(content + len, maxLen - len, "{\n"); + len += snprintf(content + len, maxLen - len, " \"version\": %" PRId64 "\n", pVnode->fversion); + len += snprintf(content + len, maxLen - len, "}\n"); + + fwrite(content, 1, len, fp); + fflush(fp); + fclose(fp); + free(content); + terrno = 0; + + vInfo("vgId:%d, successed to write %s, version:%" PRId64, pVnode->vgId, file, pVnode->fversion); + return TSDB_CODE_SUCCESS; +} \ No newline at end of file diff --git a/src/vnode/src/vnodeWrite.c b/src/vnode/src/vnodeWrite.c index 855df81a1b9029d7ddd2f5e5ca6d42e8c680f7ac..c2771c73c695244ff78fe3293dc7a91f940fed6f 100644 --- a/src/vnode/src/vnodeWrite.c +++ b/src/vnode/src/vnodeWrite.c @@ -183,7 +183,7 @@ static int32_t vnodeProcessAlterTableMsg(SVnodeObj *pVnode, void *pCont, SRspRet } static int32_t vnodeProcessDropStableMsg(SVnodeObj *pVnode, void *pCont, SRspRet *pRet) { - SMDDropSTableMsg *pTable = pCont; + SDropSTableMsg *pTable = pCont; int32_t code = TSDB_CODE_SUCCESS; vDebug("vgId:%d, stable:%s, start to drop", pVnode->vgId, pTable->tableId); @@ -204,7 +204,6 @@ static int32_t vnodeProcessUpdateTagValMsg(SVnodeObj *pVnode, void *pCont, SRspR return TSDB_CODE_SUCCESS; } - int vnodeWriteCqMsgToQueue(void *param, void *data, int type) { SVnodeObj *pVnode = param; SWalHead * pHead = data; @@ -222,7 +221,6 @@ int vnodeWriteCqMsgToQueue(void *param, void *data, int type) { return 0; } - int vnodeWriteToQueue(void *param, void *data, int type) { SVnodeObj *pVnode = param; SWalHead * pHead = data; diff --git a/src/wal/inc/walInt.h b/src/wal/inc/walInt.h index 593611589d6e1eb5d90a0eb8986f21bb32d5d1c6..5273eb5b1c2c56d300243f8bbd724a39031aa164 100644 --- a/src/wal/inc/walInt.h +++ b/src/wal/inc/walInt.h @@ -31,8 +31,34 @@ extern int32_t wDebugFlag; #define wDebug(...) { if (wDebugFlag & DEBUG_DEBUG) { taosPrintLog("WAL ", wDebugFlag, __VA_ARGS__); }} #define wTrace(...) { if (wDebugFlag & DEBUG_TRACE) { taosPrintLog("WAL ", wDebugFlag, __VA_ARGS__); }} -#define walPrefix "wal" -#define walSignature (uint32_t)(0xFAFBFDFE) +#define WAL_PREFIX "wal" +#define WAL_PREFIX_LEN 3 +#define WAL_REFRESH_MS 1000 +#define WAL_MAX_SIZE (1024 * 1024) +#define WAL_SIGNATURE ((uint32_t)(0xFAFBFDFE)) +#define WAL_PATH_LEN (TSDB_FILENAME_LEN + 12) +#define WAL_FILE_LEN (TSDB_FILENAME_LEN + 32) +#define WAL_FILE_NUM 3 + +typedef struct { + uint64_t version; + int64_t fileId; + int32_t vgId; + int32_t fd; + int32_t keep; + int32_t level; + int32_t fsyncPeriod; + int32_t fsyncSeq; + int8_t stop; + int8_t reserved[3]; + char path[WAL_PATH_LEN]; + char name[WAL_FILE_LEN]; + pthread_mutex_t mutex; +} SWal; + +int32_t walGetNextFile(SWal *pWal, int64_t *nextFileId); +int32_t walGetOldFile(SWal *pWal, int64_t curFileId, int32_t minDiff, int64_t *oldFileId); +int32_t walGetNewFile(SWal *pWal, int64_t *newFileId); #ifdef __cplusplus } diff --git a/src/wal/src/walMgmt.c b/src/wal/src/walMgmt.c index 2dd094d860528fa95084b7d3a0f404007cb757ed..c8f0274174dc518ea36500d093023ec53c5fb75d 100644 --- a/src/wal/src/walMgmt.c +++ b/src/wal/src/walMgmt.c @@ -16,6 +16,219 @@ #define _DEFAULT_SOURCE #include "os.h" #include "taoserror.h" +#include "tref.h" #include "twal.h" #include "walInt.h" -#include "walMgmt.h" \ No newline at end of file + +typedef struct { + int32_t refId; + int32_t seq; + int8_t stop; + pthread_t thread; + pthread_mutex_t mutex; +} SWalMgmt; + +static SWalMgmt tsWal; +static int32_t walCreateThread(); +static void walStopThread(); +static int32_t walInitObj(SWal *pWal); +static void walFreeObj(void *pWal); + +int32_t walInit() { + tmemzero(&tsWal, sizeof(SWalMgmt)); + tsWal.refId = taosOpenRef(TSDB_MIN_VNODES, walFreeObj); + + int32_t code = walCreateThread(); + if (code != TSDB_CODE_SUCCESS) { + wError("failed to init wal module since %s", tstrerror(code)); + return code; + } + + wInfo("wal module is initialized, refId:%d", tsWal.refId); + return code; +} + +void walCleanUp() { + walStopThread(); + taosCloseRef(tsWal.refId); + wInfo("wal module is cleaned up"); +} + +void *walOpen(char *path, SWalCfg *pCfg) { + SWal *pWal = tcalloc(1, sizeof(SWal)); + if (pWal == NULL) { + terrno = TAOS_SYSTEM_ERROR(errno); + return NULL; + } + + pWal->vgId = pCfg->vgId; + pWal->fd = -1; + pWal->fileId = -1; + pWal->level = pCfg->walLevel; + pWal->keep = pCfg->keep; + pWal->fsyncPeriod = pCfg->fsyncPeriod; + tstrncpy(pWal->path, path, sizeof(pWal->path)); + pthread_mutex_init(&pWal->mutex, NULL); + + pWal->fsyncSeq = pCfg->fsyncPeriod % 1000; + if (pWal->fsyncSeq <= 0) pWal->fsyncSeq = 1; + + if (walInitObj(pWal) != TSDB_CODE_SUCCESS) { + walFreeObj(pWal); + return NULL; + } + + if (taosAddRef(tsWal.refId, pWal) != TSDB_CODE_SUCCESS) { + walFreeObj(pWal); + return NULL; + } + + wDebug("vgId:%d, wal:%p is opened, level:%d fsyncPeriod:%d", pWal->vgId, pWal, pWal->level, pWal->fsyncPeriod); + + return pWal; +} + +int32_t walAlter(void *handle, SWalCfg *pCfg) { + if (handle == NULL) return TSDB_CODE_WAL_APP_ERROR; + SWal *pWal = handle; + + if (pWal->level == pCfg->walLevel && pWal->fsyncPeriod == pCfg->fsyncPeriod) { + wDebug("vgId:%d, old walLevel:%d fsync:%d, new walLevel:%d fsync:%d not change", pWal->vgId, pWal->level, + pWal->fsyncPeriod, pCfg->walLevel, pCfg->fsyncPeriod); + return TSDB_CODE_SUCCESS; + } + + wInfo("vgId:%d, change old walLevel:%d fsync:%d, new walLevel:%d fsync:%d", pWal->vgId, pWal->level, + pWal->fsyncPeriod, pCfg->walLevel, pCfg->fsyncPeriod); + + pWal->level = pCfg->walLevel; + pWal->fsyncPeriod = pCfg->fsyncPeriod; + pWal->fsyncSeq = pCfg->fsyncPeriod % 1000; + if (pWal->fsyncSeq <= 0) pWal->fsyncSeq = 1; + + return TSDB_CODE_SUCCESS; +} + +void walStop(void *handle) { + if (handle == NULL) return; + SWal *pWal = handle; + + pthread_mutex_lock(&pWal->mutex); + pWal->stop = 1; + pthread_mutex_unlock(&pWal->mutex); + wDebug("vgId:%d, stop write wal", pWal->vgId); +} + +void walClose(void *handle) { + if (handle == NULL) return; + + SWal *pWal = handle; + pthread_mutex_lock(&pWal->mutex); + + taosClose(pWal->fd); + + if (!pWal->keep) { + int64_t fileId = -1; + while (walGetNextFile(pWal, &fileId) >= 0) { + snprintf(pWal->name, sizeof(pWal->name), "%s/%s%" PRId64, pWal->path, WAL_PREFIX, fileId); + + if (remove(pWal->name) < 0) { + wError("vgId:%d, wal:%p file:%s, failed to remove", pWal->vgId, pWal, pWal->name); + } else { + wDebug("vgId:%d, wal:%p file:%s, it is removed", pWal->vgId, pWal, pWal->name); + } + } + } else { + wDebug("vgId:%d, wal:%p file:%s, it is closed and kept", pWal->vgId, pWal, pWal->name); + } + + pthread_mutex_unlock(&pWal->mutex); + taosRemoveRef(tsWal.refId, pWal); +} + +static int32_t walInitObj(SWal *pWal) { + if (taosMkDir(pWal->path, 0755) != 0) { + wError("vgId:%d, path:%s, failed to create directory since %s", pWal->vgId, pWal->path, strerror(errno)); + return TAOS_SYSTEM_ERROR(errno); + } + + wDebug("vgId:%d, object is initialized", pWal->vgId); + return TSDB_CODE_SUCCESS; +} + +static void walFreeObj(void *wal) { + SWal *pWal = wal; + wDebug("vgId:%d, wal:%p is freed", pWal->vgId, pWal); + + taosClose(pWal->fd); + pthread_mutex_destroy(&pWal->mutex); + tfree(pWal); +} + +static bool walNeedFsync(SWal *pWal) { + if (pWal->fsyncPeriod <= 0 || pWal->level != TAOS_WAL_FSYNC) { + return false; + } + + if (tsWal.seq % pWal->fsyncSeq == 0) { + return true; + } + + return false; +} + +static void walUpdateSeq() { + taosMsleep(WAL_REFRESH_MS); + if (++tsWal.seq <= 0) { + tsWal.seq = 1; + } +} + +static void walFsyncAll() { + SWal *pWal = taosIterateRef(tsWal.refId, NULL); + 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); + if (code != 0) { + wError("vgId:%d, file:%s, failed to fsync since %s", pWal->vgId, pWal->name, strerror(code)); + } + } + pWal = taosIterateRef(tsWal.refId, pWal); + } +} + +static void *walThreadFunc(void *param) { + while (1) { + walUpdateSeq(); + walFsyncAll(); + if (tsWal.stop) break; + } + + return NULL; +} + +static int32_t walCreateThread() { + pthread_attr_t thAttr; + pthread_attr_init(&thAttr); + pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE); + + if (pthread_create(&tsWal.thread, &thAttr, walThreadFunc, NULL) != 0) { + wError("failed to create wal thread since %s", strerror(errno)); + return TAOS_SYSTEM_ERROR(errno); + } + + pthread_attr_destroy(&thAttr); + wDebug("wal thread is launched"); + + return TSDB_CODE_SUCCESS; +} + +static void walStopThread() { + tsWal.stop = 1; + if (tsWal.thread) { + pthread_join(tsWal.thread, NULL); + } + + wDebug("wal thread is stopped"); +} diff --git a/src/wal/src/walUtil.c b/src/wal/src/walUtil.c new file mode 100644 index 0000000000000000000000000000000000000000..e4d9a555b3a60cb6be1e6584652ec4a309b1c301 --- /dev/null +++ b/src/wal/src/walUtil.c @@ -0,0 +1,118 @@ +/* + * 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 "walInt.h" + +int32_t walGetNextFile(SWal *pWal, int64_t *nextFileId) { + int64_t curFileId = *nextFileId; + int64_t minFileId = INT64_MAX; + + DIR *dir = opendir(pWal->path); + if (dir == NULL) { + wError("vgId:%d, path:%s, failed to open since %s", pWal->vgId, pWal->path, strerror(errno)); + return -1; + } + + struct dirent *ent; + while ((ent = readdir(dir)) != NULL) { + char *name = ent->d_name; + + if (strncmp(name, WAL_PREFIX, WAL_PREFIX_LEN) == 0) { + int64_t id = atoll(name + WAL_PREFIX_LEN); + if (id <= curFileId) continue; + + if (id < minFileId) { + minFileId = id; + } + } + } + closedir(dir); + + if (minFileId == INT64_MAX) return -1; + + *nextFileId = minFileId; + wTrace("vgId:%d, path:%s, curFileId:%" PRId64 " nextFileId:%" PRId64, pWal->vgId, pWal->path, curFileId, *nextFileId); + + return 0; +} + +int32_t walGetOldFile(SWal *pWal, int64_t curFileId, int32_t minDiff, int64_t *oldFileId) { + int64_t minFileId = INT64_MAX; + + DIR *dir = opendir(pWal->path); + if (dir == NULL) { + wError("vgId:%d, path:%s, failed to open since %s", pWal->vgId, pWal->path, strerror(errno)); + return -1; + } + + struct dirent *ent; + while ((ent = readdir(dir)) != NULL) { + char *name = ent->d_name; + + if (strncmp(name, WAL_PREFIX, WAL_PREFIX_LEN) == 0) { + int64_t id = atoll(name + WAL_PREFIX_LEN); + if (id >= curFileId) continue; + + minDiff--; + if (id < minFileId) { + minFileId = id; + } + } + } + closedir(dir); + + if (minFileId == INT64_MAX) return -1; + if (minDiff > 0) return -1; + + *oldFileId = minFileId; + wTrace("vgId:%d, path:%s, curFileId:%" PRId64 " oldFildId:%" PRId64, pWal->vgId, pWal->path, curFileId, *oldFileId); + + return 0; +} + +int32_t walGetNewFile(SWal *pWal, int64_t *newFileId) { + int64_t maxFileId = INT64_MIN; + + DIR *dir = opendir(pWal->path); + if (dir == NULL) { + wError("vgId:%d, path:%s, failed to open since %s", pWal->vgId, pWal->path, strerror(errno)); + return -1; + } + + struct dirent *ent; + while ((ent = readdir(dir)) != NULL) { + char *name = ent->d_name; + + if (strncmp(name, WAL_PREFIX, WAL_PREFIX_LEN) == 0) { + int64_t id = atoll(name + WAL_PREFIX_LEN); + if (id > maxFileId) { + maxFileId = id; + } + } + } + closedir(dir); + + if (maxFileId == INT64_MIN) { + *newFileId = 0; + } else { + *newFileId = maxFileId; + } + + wTrace("vgId:%d, path:%s, newFileId:%" PRId64, pWal->vgId, pWal->path, *newFileId); + + return 0; +} \ No newline at end of file diff --git a/src/wal/src/walWrite.c b/src/wal/src/walWrite.c index 95587caa141a8964a845a9a0b47485bfc20fa998..e57cb0e042f321ed4d4d3c463a84e6917d4e0fc7 100644 --- a/src/wal/src/walWrite.c +++ b/src/wal/src/walWrite.c @@ -14,239 +14,101 @@ */ #define _DEFAULT_SOURCE +#define TAOS_RANDOM_FILE_FAIL_TEST #include "os.h" -#include "twal.h" -#include "walInt.h" -#include "walMgmt.h" -#include "tchecksum.h" -#include "tutil.h" -#include "ttimer.h" #include "taoserror.h" +#include "tchecksum.h" #include "twal.h" -#include "tqueue.h" - - -typedef struct { - uint64_t version; - int fd; - int keep; - int level; - int32_t fsyncPeriod; - void *timer; - void *signature; - int max; // maximum number of wal files - uint32_t id; // increase continuously - int num; // number of wal files - char path[TSDB_FILENAME_LEN]; - char name[TSDB_FILENAME_LEN+16]; - pthread_mutex_t mutex; -} SWal; - -static void *walTmrCtrl = NULL; -static int tsWalNum = 0; -static pthread_once_t walModuleInit = PTHREAD_ONCE_INIT; -static int walHandleExistingFiles(const char *path); -static int walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp); -static int walRemoveWalFiles(const char *path); -static void walProcessFsyncTimer(void *param, void *tmrId); -static void walRelease(SWal *pWal); -static int walGetMaxOldFileId(char *odir); - -static void walModuleInitFunc() { - walTmrCtrl = taosTmrInit(1000, 100, 300000, "WAL"); - if (walTmrCtrl == NULL) - walModuleInit = PTHREAD_ONCE_INIT; - else - wDebug("WAL module is initialized"); -} - -static inline bool walNeedFsyncTimer(SWal *pWal) { - if (pWal->fsyncPeriod > 0 && pWal->level == TAOS_WAL_FSYNC) { - return true; - } - return false; -} - -void *walOpen(const char *path, const SWalCfg *pCfg) { - SWal *pWal = calloc(sizeof(SWal), 1); - if (pWal == NULL) { - terrno = TAOS_SYSTEM_ERROR(errno); - return NULL; - } - - pthread_once(&walModuleInit, walModuleInitFunc); - if (walTmrCtrl == NULL) { - free(pWal); - terrno = TAOS_SYSTEM_ERROR(errno); - return NULL; - } - - atomic_add_fetch_32(&tsWalNum, 1); - pWal->fd = -1; - pWal->max = pCfg->wals; - pWal->id = 0; - pWal->num = 0; - pWal->level = pCfg->walLevel; - pWal->keep = pCfg->keep; - pWal->fsyncPeriod = pCfg->fsyncPeriod; - pWal->signature = pWal; - tstrncpy(pWal->path, path, sizeof(pWal->path)); - pthread_mutex_init(&pWal->mutex, NULL); - - if (walNeedFsyncTimer(pWal)) { - pWal->timer = taosTmrStart(walProcessFsyncTimer, pWal->fsyncPeriod, pWal, walTmrCtrl); - if (pWal->timer == NULL) { - terrno = TAOS_SYSTEM_ERROR(errno); - walRelease(pWal); - return NULL; - } - } +#include "walInt.h" - if (taosMkDir(path, 0755) != 0) { - terrno = TAOS_SYSTEM_ERROR(errno); - wError("wal:%s, failed to create directory(%s)", path, strerror(errno)); - walRelease(pWal); - pWal = NULL; - } +static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, char *name, int64_t fileId); - if (pCfg->keep == 1) return pWal; +int32_t walRenew(void *handle) { + if (handle == NULL) return 0; - if (walHandleExistingFiles(path) == 0) walRenew(pWal); + SWal * pWal = handle; + int32_t code = 0; - if (pWal && pWal->fd < 0) { - terrno = TAOS_SYSTEM_ERROR(errno); - wError("wal:%s, failed to open(%s)", path, strerror(errno)); - walRelease(pWal); - pWal = NULL; + if (pWal->stop) { + wDebug("vgId:%d, do not create a new wal file", pWal->vgId); + return 0; } - if (pWal) wDebug("wal:%s, it is open, level:%d fsyncPeriod:%d", path, pWal->level, pWal->fsyncPeriod); - return pWal; -} - -int walAlter(twalh wal, const SWalCfg *pCfg) { - SWal *pWal = wal; - if (pWal == NULL) { - return TSDB_CODE_WAL_APP_ERROR; - } + pthread_mutex_lock(&pWal->mutex); - if (pWal->level == pCfg->walLevel && pWal->fsyncPeriod == pCfg->fsyncPeriod) { - wDebug("wal:%s, old walLevel:%d fsync:%d, new walLevel:%d fsync:%d not change", pWal->name, pWal->level, - pWal->fsyncPeriod, pCfg->walLevel, pCfg->fsyncPeriod); - return TSDB_CODE_SUCCESS; + if (pWal->fd >= 0) { + tclose(pWal->fd); + wDebug("vgId:%d, file:%s, it is closed", pWal->vgId, pWal->name); } - wInfo("wal:%s, change old walLevel:%d fsync:%d, new walLevel:%d fsync:%d", pWal->name, pWal->level, pWal->fsyncPeriod, - pCfg->walLevel, pCfg->fsyncPeriod); - - pthread_mutex_lock(&pWal->mutex); - pWal->level = pCfg->walLevel; - pWal->fsyncPeriod = pCfg->fsyncPeriod; - if (walNeedFsyncTimer(pWal)) { - wInfo("wal:%s, reset fsync timer, walLevel:%d fsyncPeriod:%d", pWal->name, pWal->level, pWal->fsyncPeriod); - taosTmrReset(walProcessFsyncTimer, pWal->fsyncPeriod, pWal, &pWal->timer, walTmrCtrl); + if (pWal->keep) { + pWal->fileId = 0; } else { - wInfo("wal:%s, stop fsync timer, walLevel:%d fsyncPeriod:%d", pWal->name, pWal->level, pWal->fsyncPeriod); - taosTmrStop(pWal->timer); - pWal->timer = NULL; + if (walGetNewFile(pWal, &pWal->fileId) != 0) pWal->fileId = 0; + pWal->fileId++; } - pthread_mutex_unlock(&pWal->mutex); - return TSDB_CODE_SUCCESS; -} - -void walClose(void *handle) { - if (handle == NULL) return; + 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); - SWal *pWal = handle; - taosClose(pWal->fd); - if (pWal->timer) taosTmrStopA(&pWal->timer); - - if (pWal->keep == 0) { - // remove all files in the directory - for (int i = 0; i < pWal->num; ++i) { - snprintf(pWal->name, sizeof(pWal->name), "%s/%s%d", pWal->path, walPrefix, pWal->id - i); - if (remove(pWal->name) < 0) { - wError("wal:%s, failed to remove", pWal->name); - } else { - wDebug("wal:%s, it is removed", pWal->name); - } - } + if (pWal->fd < 0) { + code = TAOS_SYSTEM_ERROR(errno); + wError("vgId:%d, file:%s, failed to open since %s", pWal->vgId, pWal->name, strerror(errno)); } else { - wDebug("wal:%s, it is closed and kept", pWal->name); + wDebug("vgId:%d, file:%s, it is created", pWal->vgId, pWal->name); } - walRelease(pWal); -} - -int walRenew(void *handle) { - if (handle == NULL) return 0; - SWal *pWal = handle; - - terrno = 0; - - pthread_mutex_lock(&pWal->mutex); - - if (pWal->fd >= 0) { - close(pWal->fd); - pWal->id++; - wDebug("wal:%s, it is closed", pWal->name); - } - - pWal->num++; + if (!pWal->keep) { + // remove the oldest wal file + int64_t oldFileId = -1; + if (walGetOldFile(pWal, pWal->fileId, WAL_FILE_NUM, &oldFileId) == 0) { + char walName[WAL_FILE_LEN] = {0}; + snprintf(walName, sizeof(walName), "%s/%s%" PRId64, pWal->path, WAL_PREFIX, oldFileId); - snprintf(pWal->name, sizeof(pWal->name), "%s/%s%d", pWal->path, walPrefix, pWal->id); - pWal->fd = open(pWal->name, O_WRONLY | O_CREAT, S_IRWXU | S_IRWXG | S_IRWXO); - - if (pWal->fd < 0) { - wError("wal:%s, failed to open(%s)", pWal->name, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - } else { - wDebug("wal:%s, it is created", pWal->name); - - if (pWal->num > pWal->max) { - // remove the oldest wal file - char name[TSDB_FILENAME_LEN * 3]; - snprintf(name, sizeof(name), "%s/%s%d", pWal->path, walPrefix, pWal->id - pWal->max); - if (remove(name) < 0) { - wError("wal:%s, failed to remove(%s)", name, strerror(errno)); + if (remove(walName) < 0) { + wError("vgId:%d, file:%s, failed to remove since %s", pWal->vgId, walName, strerror(errno)); } else { - wDebug("wal:%s, it is removed", name); + wDebug("vgId:%d, file:%s, it is removed", pWal->vgId, walName); } - - pWal->num--; } } pthread_mutex_unlock(&pWal->mutex); - return terrno; + return code; } -int walWrite(void *handle, SWalHead *pHead) { - SWal *pWal = handle; - if (pWal == NULL) return -1; +int32_t walWrite(void *handle, SWalHead *pHead) { + if (handle == NULL) return -1; - terrno = 0; + SWal * pWal = handle; + int32_t code = 0; // no wal + if (pWal->fd <= 0) return 0; if (pWal->level == TAOS_WAL_NOLOG) return 0; if (pHead->version <= pWal->version) return 0; - pHead->signature = walSignature; + pHead->signature = WAL_SIGNATURE; taosCalcChecksumAppend(0, (uint8_t *)pHead, sizeof(SWalHead)); - int contLen = pHead->len + sizeof(SWalHead); + int32_t contLen = pHead->len + sizeof(SWalHead); + + pthread_mutex_lock(&pWal->mutex); - if (taosTWrite(pWal->fd, pHead, contLen) != contLen) { - wError("wal:%s, failed to write(%s)", pWal->name, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - return terrno; + if (taosWrite(pWal->fd, 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, fileId:%" PRId64 " fd:%d, write wal ver:%" PRId64 ", head ver:%" PRIu64 ", len:%d ", pWal->vgId, + pWal->fileId, pWal->fd, pWal->version, pHead->version, pHead->len); pWal->version = pHead->version; } + + pthread_mutex_unlock(&pWal->mutex); + ASSERT(contLen == pHead->len + sizeof(SWalHead)); - return 0; + return code; } void walFsync(void *handle) { @@ -255,295 +117,203 @@ void walFsync(void *handle) { if (pWal->fsyncPeriod == 0) { if (fsync(pWal->fd) < 0) { - wError("wal:%s, fsync failed(%s)", pWal->name, strerror(errno)); + wError("vgId:%d, file:%s, fsync failed since %s", pWal->vgId, pWal->name, strerror(errno)); } } } -int walRestore(void *handle, void *pVnode, int (*writeFp)(void *, void *, int)) { - SWal *pWal = handle; - struct dirent *ent; - int count = 0; - uint32_t maxId = 0, minId = -1, index =0; +int32_t walRestore(void *handle, void *pVnode, int32_t (*writeFp)(void *, void *, int32_t)) { + if (handle == NULL) return -1; - terrno = 0; - int plen = strlen(walPrefix); - char opath[TSDB_FILENAME_LEN + 5]; + SWal * pWal = handle; + int32_t count = 0; + int32_t code = 0; + int64_t fileId = -1; - int slen = snprintf(opath, sizeof(opath), "%s", pWal->path); - if (pWal->keep == 0) strcpy(opath + slen, "/old"); + while ((code = walGetNextFile(pWal, &fileId)) >= 0) { + if (fileId == pWal->fileId) continue; - DIR *dir = opendir(opath); - if (dir == NULL && errno == ENOENT) return 0; - if (dir == NULL) { - terrno = TAOS_SYSTEM_ERROR(errno); - return terrno; - } + char walName[WAL_FILE_LEN]; + snprintf(walName, sizeof(pWal->name), "%s/%s%" PRId64, pWal->path, WAL_PREFIX, fileId); - while ((ent = readdir(dir)) != NULL) { - if (strncmp(ent->d_name, walPrefix, plen) == 0) { - index = atol(ent->d_name + plen); - if (index > maxId) maxId = index; - if (index < minId) minId = index; - count++; + wDebug("vgId:%d, file:%s, will be restored", pWal->vgId, walName); + int32_t code = walRestoreWalFile(pWal, pVnode, writeFp, walName, fileId); + if (code != TSDB_CODE_SUCCESS) { + wError("vgId:%d, file:%s, failed to restore since %s", pWal->vgId, walName, tstrerror(code)); + continue; } - } - closedir(dir); + if (!pWal->keep) { + wDebug("vgId:%d, file:%s, restore success, remove this file", pWal->vgId, walName); + remove(walName); + } else { + wDebug("vgId:%d, file:%s, restore success and keep it", pWal->vgId, walName); + } - if (count == 0) { - if (pWal->keep) terrno = walRenew(pWal); - return terrno; + count++; } - if (count != (maxId - minId + 1)) { - wError("wal:%s, messed up, count:%d max:%d min:%d", opath, count, maxId, minId); - terrno = TSDB_CODE_WAL_APP_ERROR; - } else { - wDebug("wal:%s, %d files will be restored", opath, count); - - for (index = minId; index <= maxId; ++index) { - snprintf(pWal->name, sizeof(pWal->name), "%s/%s%d", opath, walPrefix, index); - terrno = walRestoreWalFile(pWal, pVnode, writeFp); - if (terrno < 0) continue; - } - } + if (!pWal->keep) return TSDB_CODE_SUCCESS; - if (terrno == 0) { - if (pWal->keep == 0) { - terrno = walRemoveWalFiles(opath); - if (terrno == 0) { - if (remove(opath) < 0) { - wError("wal:%s, failed to remove directory(%s)", opath, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - } - } - } else { - // open the existing WAL file in append mode - pWal->num = count; - pWal->id = maxId; - snprintf(pWal->name, sizeof(pWal->name), "%s/%s%d", opath, walPrefix, maxId); - pWal->fd = open(pWal->name, O_WRONLY | O_CREAT | O_APPEND, S_IRWXU | S_IRWXG | S_IRWXO); - if (pWal->fd < 0) { - wError("wal:%s, failed to open file(%s)", pWal->name, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - } + if (count == 0) { + wDebug("vgId:%d, wal file not exist, renew it", pWal->vgId); + return walRenew(pWal); + } else { + // 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) { + wError("vgId:%d, file:%s, failed to open since %s", pWal->vgId, pWal->name, strerror(errno)); + return TAOS_SYSTEM_ERROR(errno); } + wDebug("vgId:%d, file:%s open success", pWal->vgId, pWal->name); } - return terrno; + return TSDB_CODE_SUCCESS; } -int walGetWalFile(void *handle, char *name, uint32_t *index) { - SWal * pWal = handle; - int code = 1; - int32_t first = 0; +int32_t walGetWalFile(void *handle, char *fileName, int64_t *fileId) { + if (handle == NULL) return -1; + SWal *pWal = handle; - name[0] = 0; - if (pWal == NULL || pWal->num == 0) return 0; + // for keep + if (*fileId == 0) *fileId = -1; pthread_mutex_lock(&(pWal->mutex)); - first = pWal->id + 1 - pWal->num; - if (*index == 0) *index = first; // set to first one - - if (*index < first && *index > pWal->id) { - code = -1; // index out of range - } else { - sprintf(name, "wal/%s%d", walPrefix, *index); - code = (*index == pWal->id) ? 0 : 1; + int32_t code = walGetNextFile(pWal, fileId); + if (code >= 0) { + sprintf(fileName, "wal/%s%" PRId64, WAL_PREFIX, *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); pthread_mutex_unlock(&(pWal->mutex)); return code; } -static void walRelease(SWal *pWal) { - pthread_mutex_destroy(&pWal->mutex); - pWal->signature = NULL; - free(pWal); +static void walFtruncate(SWal *pWal, int32_t fd, int64_t offset) { + taosFtruncate(fd, offset); + fsync(fd); +} + +static int32_t walSkipCorruptedRecord(SWal *pWal, SWalHead *pHead, int32_t fd, int64_t *offset) { + int64_t pos = *offset; + while (1) { + pos++; + + if (lseek(fd, 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) { + wError("vgId:%d, read to end of corrupted wal file, offset:%" PRId64, pWal->vgId, pos); + return TSDB_CODE_WAL_FILE_CORRUPTED; + } + + if (pHead->signature != WAL_SIGNATURE) { + continue; + } - if (atomic_sub_fetch_32(&tsWalNum, 1) == 0) { - if (walTmrCtrl) taosTmrCleanUp(walTmrCtrl); - walTmrCtrl = NULL; - walModuleInit = PTHREAD_ONCE_INIT; - wDebug("WAL module is cleaned up"); + if (taosCheckChecksumWhole((uint8_t *)pHead, sizeof(SWalHead))) { + wInfo("vgId:%d, wal head cksum check passed, offset:%" PRId64, pWal->vgId, pos); + *offset = pos; + return TSDB_CODE_SUCCESS; + } } -} -static int walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp) { - char *name = pWal->name; - int size = 1024 * 1024; // default 1M buffer size + return TSDB_CODE_WAL_FILE_CORRUPTED; +} - terrno = 0; - char *buffer = malloc(size); +static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, char *name, int64_t fileId) { + int32_t size = WAL_MAX_SIZE; + void * buffer = tmalloc(size); if (buffer == NULL) { - terrno = TAOS_SYSTEM_ERROR(errno); - return terrno; + wError("vgId:%d, file:%s, failed to open for restore since %s", pWal->vgId, name, strerror(errno)); + return TAOS_SYSTEM_ERROR(errno); } - SWalHead *pHead = (SWalHead *)buffer; - - int fd = open(name, O_RDWR); + int32_t fd = open(name, O_RDWR); if (fd < 0) { - wError("wal:%s, failed to open for restore(%s)", name, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - free(buffer); - return terrno; + wError("vgId:%d, file:%s, failed to open for restore since %s", pWal->vgId, name, strerror(errno)); + tfree(buffer); + return TAOS_SYSTEM_ERROR(errno); } - wDebug("wal:%s, start to restore", name); + wDebug("vgId:%d, file:%s, start to restore", pWal->vgId, name); + + int32_t code = TSDB_CODE_SUCCESS; + int64_t offset = 0; + SWalHead *pHead = buffer; - size_t offset = 0; while (1) { - int ret = taosTRead(fd, pHead, sizeof(SWalHead)); + int32_t ret = taosRead(fd, pHead, sizeof(SWalHead)); if (ret == 0) break; if (ret < 0) { - wError("wal:%s, failed to read wal head part since %s", name, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); + wError("vgId:%d, file:%s, failed to read wal head since %s", pWal->vgId, name, strerror(errno)); + code = TAOS_SYSTEM_ERROR(errno); break; } if (ret < sizeof(SWalHead)) { - wError("wal:%s, failed to read head, ret:%d, skip the rest of file", name, ret); - taosFtruncate(fd, offset); - fsync(fd); + wError("vgId:%d, file:%s, failed to read wal head, ret is %d", pWal->vgId, name, ret); + walFtruncate(pWal, fd, offset); break; } if (!taosCheckChecksumWhole((uint8_t *)pHead, sizeof(SWalHead))) { - wWarn("wal:%s, cksum is messed up, skip the rest of file", name); - terrno = TSDB_CODE_WAL_FILE_CORRUPTED; - ASSERT(false); - break; + wError("vgId:%d, file:%s, wal head cksum is messed up, ver:%" PRIu64 " len:%d offset:%" PRId64, pWal->vgId, name, + pHead->version, pHead->len, offset); + code = walSkipCorruptedRecord(pWal, pHead, fd, &offset); + if (code != TSDB_CODE_SUCCESS) { + walFtruncate(pWal, fd, offset); + break; + } } if (pHead->len > size - sizeof(SWalHead)) { size = sizeof(SWalHead) + pHead->len; buffer = realloc(buffer, size); if (buffer == NULL) { - terrno = TAOS_SYSTEM_ERROR(errno); + wError("vgId:%d, file:%s, failed to open for restore since %s", pWal->vgId, name, strerror(errno)); + code = TAOS_SYSTEM_ERROR(errno); break; } - pHead = (SWalHead *)buffer; + pHead = buffer; } - ret = taosTRead(fd, pHead->cont, pHead->len); + ret = taosRead(fd, pHead->cont, pHead->len); if (ret < 0) { - wError("wal:%s failed to read wal body part since %s", name, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); + wError("vgId:%d, file:%s, failed to read wal body since %s", pWal->vgId, name, strerror(errno)); + code = TAOS_SYSTEM_ERROR(errno); break; } if (ret < pHead->len) { - wError("wal:%s, failed to read body, len:%d ret:%d, skip the rest of file", name, pHead->len, ret); - taosFtruncate(fd, offset); - fsync(fd); - break; + wError("vgId:%d, file:%s, failed to read wal body, ret:%d len:%d", pWal->vgId, name, ret, pHead->len); + offset += sizeof(SWalHead); + continue; } offset = offset + sizeof(SWalHead) + pHead->len; - if (pWal->keep) pWal->version = pHead->version; - (*writeFp)(pVnode, pHead, TAOS_QTYPE_WAL); - } - - close(fd); - free(buffer); + wTrace("vgId:%d, fileId:%" PRId64 ", restore wal ver:%" PRIu64 ", head ver:%" PRIu64 " len:%d", pWal->vgId, fileId, + pWal->version, pHead->version, pHead->len); - return terrno; -} - -int walHandleExistingFiles(const char *path) { - char oname[TSDB_FILENAME_LEN * 3]; - char nname[TSDB_FILENAME_LEN * 3]; - char opath[TSDB_FILENAME_LEN]; - - snprintf(opath, sizeof(opath), "%s/old", path); - - struct dirent *ent; - DIR *dir = opendir(path); - int plen = strlen(walPrefix); - terrno = 0; - - int midx = walGetMaxOldFileId(opath); - int count = 0; - while ((ent = readdir(dir)) != NULL) { - if (strncmp(ent->d_name, walPrefix, plen) == 0) { - midx++; - snprintf(oname, sizeof(oname), "%s/%s", path, ent->d_name); - snprintf(nname, sizeof(nname), "%s/old/wal%d", path, midx); - if (taosMkDir(opath, 0755) != 0) { - wError("wal:%s, failed to create directory:%s(%s)", oname, opath, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - break; - } - - if (rename(oname, nname) < 0) { - wError("wal:%s, failed to move to new:%s", oname, nname); - terrno = TAOS_SYSTEM_ERROR(errno); - break; - } - - count++; - } - - wDebug("wal:%s, %d files are moved for restoration", path, count); - } - - closedir(dir); - return terrno; -} - -static int walRemoveWalFiles(const char *path) { - int plen = strlen(walPrefix); - char name[TSDB_FILENAME_LEN * 3]; - - terrno = 0; - - struct dirent *ent; - DIR *dir = opendir(path); - if (dir == NULL && errno == ENOENT) return 0; - if (dir == NULL) { - terrno = TAOS_SYSTEM_ERROR(errno); - return terrno; - } + if (pWal->keep) pWal->version = pHead->version; - while ((ent = readdir(dir)) != NULL) { - if (strncmp(ent->d_name, walPrefix, plen) == 0) { - snprintf(name, sizeof(name), "%s/%s", path, ent->d_name); - if (remove(name) < 0) { - wError("wal:%s, failed to remove(%s)", name, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - } - } + (*writeFp)(pVnode, pHead, TAOS_QTYPE_WAL); } - closedir(dir); + tclose(fd); + tfree(buffer); - return terrno; -} - -static void walProcessFsyncTimer(void *param, void *tmrId) { - SWal *pWal = param; - - if (pWal->signature != pWal) return; - if (pWal->fd < 0) return; - - if (fsync(pWal->fd) < 0) { - wError("wal:%s, fsync failed(%s)", pWal->name, strerror(errno)); - } - - if (walNeedFsyncTimer(pWal)) { - pWal->timer = taosTmrStart(walProcessFsyncTimer, pWal->fsyncPeriod, pWal, walTmrCtrl); - } else { - wInfo("wal:%s, stop fsync timer for walLevel:%d fsyncPeriod:%d", pWal->name, pWal->level, pWal->fsyncPeriod); - taosTmrStop(pWal->timer); - pWal->timer = NULL; - } + return code; } int64_t walGetVersion(twalh param) { @@ -552,29 +322,3 @@ int64_t walGetVersion(twalh param) { return pWal->version; } - -static int walGetMaxOldFileId(char *odir) { - int midx = 0; - DIR * dir = NULL; - struct dirent *dp = NULL; - int plen = strlen(walPrefix); - - if (access(odir, F_OK) != 0) return midx; - - dir = opendir(odir); - if (dir == NULL) { - wError("failed to open directory %s since %s", odir, strerror(errno)); - terrno = TAOS_SYSTEM_ERROR(errno); - return -1; - } - - while ((dp = readdir(dir)) != NULL) { - if (strncmp(dp->d_name, walPrefix, plen) == 0) { - int idx = atol(dp->d_name + plen); - if (midx < idx) midx = idx; - } - } - - closedir(dir); - return midx; -} \ No newline at end of file diff --git a/src/wal/test/waltest.c b/src/wal/test/waltest.c index bbee1347b8f92aa6cfad448fdfb369de8f5a6301..186f2ef5ffe9fed68f9c9205b251cbb3e10dfeab 100644 --- a/src/wal/test/waltest.c +++ b/src/wal/test/waltest.c @@ -115,17 +115,17 @@ int main(int argc, char *argv[]) { printf("%d wal files are written\n", total); - uint32_t index = 0; - char name[256]; + int64_t index = 0; + char name[256]; while (1) { int code = walGetWalFile(pWal, name, &index); if (code == -1) { - printf("failed to get wal file, index:%d\n", index); + printf("failed to get wal file, index:%" PRId64 "\n", index); break; } - printf("index:%d wal:%s\n", index, name); + printf("index:%" PRId64 " wal:%s\n", index, name); if (code == 0) break; index++; diff --git a/tests/pytest/insert/restfulInsert.py b/tests/pytest/insert/restfulInsert.py new file mode 100644 index 0000000000000000000000000000000000000000..a6c9b074e12c571e7ef715a520e20295c936362c --- /dev/null +++ b/tests/pytest/insert/restfulInsert.py @@ -0,0 +1,83 @@ +################################################################### +# 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 requests +import threading +import random +import time + +class RestfulInsert: + def init(self): + self.header = {'Authorization': 'Basic cm9vdDp0YW9zZGF0YQ=='} + self.url = "http://127.0.0.1:6041/rest/sql" + self.ts = 1500000000000 + self.numOfThreads = 20 + self.numOfTables = 10000 + self.recordsPerTable = 10000 + self.batchSize = 1000 + self.tableNamePerfix = 't' + + 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): + tableID = threadID * tablesPerThread + name = 'beijing' if tableID % 2 == 0 else 'shanghai' + data = "create table test.%s%d using test.meters tags(%d, '%s')" % (self.tableNamePerfix, tableID + i, tableID + i, name) + requests.post(self.url, data, headers = self.header) + + 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 test.%s%d values" % (self.tableNamePerfix, tableID) + 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) + + def run(self): + data = "drop database if exists test" + requests.post(self.url, data, headers = self.header) + data = "create database test" + requests.post(self.url, data, headers = self.header) + data = "create table test.meters(ts timestamp, f1 int, f2 int, f3 int) tags(id int, loc nchar(20))" + requests.post(self.url, data, headers = self.header) + + threads = [] + startTime = time.time() + for i in range(self.numOfThreads): + thread = threading.Thread(target=self.createTable, args=(i,)) + thread.start() + threads.append(thread) + for i in range(self.numOfThreads): + threads[i].join() + print("createing %d tables takes %d seconds" % (self.numOfTables, (time.time() - startTime))) + + print("inserting data =======") + threads = [] + startTime = time.time() + for i in range(self.numOfThreads): + thread = threading.Thread(target=self.insertData, args=(i,)) + thread.start() + threads.append(thread) + + for i in range(self.numOfThreads): + threads[i].join() + print("inserting %d records takes %d seconds" % (self.numOfTables * self.recordsPerTable, (time.time() - startTime))) + +ri = RestfulInsert() +ri.init() +ri.run() \ No newline at end of file diff --git a/tests/pytest/query/sliding.py b/tests/pytest/query/sliding.py new file mode 100644 index 0000000000000000000000000000000000000000..810d90117a73bba81ea010f81aa605196a496dc4 --- /dev/null +++ b/tests/pytest/query/sliding.py @@ -0,0 +1,63 @@ +################################################################### +# 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 + + def run(self): + tdSql.prepare() + + tdSql.execute("create table meters(ts timestamp, col1 int) tags(id int, loc nchar(20))") + sql = "insert into t0 using meters tags(1, 'beijing') values" + for i in range(100): + sql += "(%d, %d)" % (self.ts + i * 1000, random.randint(1, 100)) + tdSql.execute(sql) + + sql = "insert into t1 using meters tags(2, 'shanghai') values" + for i in range(100): + sql += "(%d, %d)" % (self.ts + i * 1000, random.randint(1, 100)) + tdSql.execute(sql) + + tdSql.query("select count(*) from meters interval(10s) sliding(5s)") + tdSql.checkRows(21) + + tdSql.error("select count(*) from meters sliding(5s)") + + tdSql.error("select count(*) from meters sliding(5s) interval(10s)") + + tdSql.error("select * from meters sliding(5s) order by ts desc") + + tdSql.query("select count(*) from meters group by loc") + tdSql.checkRows(2) + + tdSql.error("select * from meters group by loc sliding(5s)") + + def stop(self): + tdSql.close() + tdLog.success("%s successfully executed" % __file__) + + +tdCases.addWindows(__file__, TDTestCase()) +tdCases.addLinux(__file__, TDTestCase()) diff --git a/tests/script/general/insert/basic.sim b/tests/script/general/insert/basic.sim index ba8cff83fa8b470af1b1f7fccf74d0e1d042049e..3f0f25a95ba68c8099c02e6daab933411089a632 100644 --- a/tests/script/general/insert/basic.sim +++ b/tests/script/general/insert/basic.sim @@ -8,8 +8,8 @@ sleep 3000 sql connect $i = 0 -$dbPrefix = tb_in_db -$tbPrefix = tb_in_tb +$dbPrefix = d +$tbPrefix = t $db = $dbPrefix . $i $tb = $tbPrefix . $i @@ -22,28 +22,27 @@ sql create table $tb (ts timestamp, speed int) $x = 0 while $x < 10 - $ms = $x . m - sql insert into $tb values (now + $ms , $x ) + $cc = $x * 60000 + $ms = 1601481600000 + $cc + + sql insert into $tb values ($ms , $x ) $x = $x + 1 endw print =============== step 2 -sql insert into $tb values (now - 5m , 10) -sql insert into $tb values (now - 6m , 10) -sql insert into $tb values (now - 7m , 10) -sql insert into $tb values (now - 8m , 10) +$x = 0 +while $x < 5 + $cc = $x * 60000 + $ms = 1551481600000 + $cc + + sql insert into $tb values ($ms , $x ) + $x = $x + 1 +endw sql select * from $tb print $rows points data are retrieved -if $rows != 14 then - return -1 -endi - -sql drop database $db -sleep 1000 -sql show databases -if $rows != 0 then +if $rows != 15 then return -1 endi diff --git a/tests/script/general/parser/fill.sim b/tests/script/general/parser/fill.sim index f89c27d71fd8ba02105f502afd4df26a1870ffb5..92aa6a922cc901c88ad68f87f61e8befbed5e87c 100644 --- a/tests/script/general/parser/fill.sim +++ b/tests/script/general/parser/fill.sim @@ -116,7 +116,7 @@ if $data81 != 1 then endi # avg_with_fill -print avg_witt_constant_fill +print avg_with_constant_fill sql select avg(c1), avg(c2), avg(c3), avg(c4), avg(c5) from $tb where ts >= $ts0 and ts <= $tsu interval(5m) fill(value, 6, 6, 6, 6, 6) if $rows != 9 then return -1 @@ -371,12 +371,10 @@ if $data11 != 99 then endi sql select * from $tb -#print data08 = $data08 if $data08 != NCHAR then + print expect NCHAR, actual:$data08 return -1 endi -#return -1 - # fill_into_nonarithmetic_fieds sql select first(c6), first(c7), first(c8) from $tb where ts >= $ts0 and ts <= $tsu interval(5m) fill(value, 20000000, 20000000, 20000000) diff --git a/tests/script/general/parser/groupby.sim b/tests/script/general/parser/groupby.sim index b70fe88e815eec020b657f00a972eba74c6e2976..145c4a008f1ae26378baeabdfd5103cc807266f3 100644 --- a/tests/script/general/parser/groupby.sim +++ b/tests/script/general/parser/groupby.sim @@ -435,53 +435,53 @@ 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) ; -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 -endi - -if $data01 != 1.000000000 then - return -1 -endi -if $data02 != t1 then - return -1 -endi -if $data03 != 1 then - return -1 -endi -if $data04 != 1 then - return -1 -endi - -if $data11 != 1.000000000 then - return -1 -endi -if $data12 != t1 then - return -1 -endi -if $data13 != 1 then - return -1 -endi -if $data14 != 1 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 limit 1; -if $rows != 2 then - return -1 -endi - -if $data11 != 1.000000000 then - return -1 -endi -if $data12 != t2 then - return -1 -endi -if $data13 != 1 then - return -1 -endi -if $data14 != 2 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 +#endi +# +#if $data01 != 1.000000000 then +# return -1 +#endi +#if $data02 != t1 then +# return -1 +#endi +#if $data03 != 1 then +# return -1 +#endi +#if $data04 != 1 then +# return -1 +#endi +# +#if $data11 != 1.000000000 then +# return -1 +#endi +#if $data12 != t1 then +# return -1 +#endi +#if $data13 != 1 then +# return -1 +#endi +#if $data14 != 1 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 limit 1; +#if $rows != 2 then +# return -1 +#endi +# +#if $data11 != 1.000000000 then +# return -1 +#endi +#if $data12 != t2 then +# return -1 +#endi +#if $data13 != 1 then +# return -1 +#endi +#if $data14 != 2 then +# return -1 +#endi system sh/exec.sh -n dnode1 -s stop -x SIGINT diff --git a/tests/script/general/parser/join.sim b/tests/script/general/parser/join.sim index 79b30ffe922af1e80ab242c4c95c835c9a4a790f..624c2e72c40bfee35f8b0b17646c97149eda65c1 100644 --- a/tests/script/general/parser/join.sim +++ b/tests/script/general/parser/join.sim @@ -444,6 +444,9 @@ if $rows != $val then return -1 endi +#=============================================================== +sql select first(join_tb0.c8),first(join_tb0.c9) from join_tb1 , join_tb0 where join_tb1.ts = join_tb0.ts and join_tb1.ts <= 100002 and join_tb0.c7 = true + #====================group by========================================= diff --git a/tests/script/general/parser/join_multivnode.sim b/tests/script/general/parser/join_multivnode.sim index 102d5ba93fb79ff8874660b240c26937cce6dce7..c5fcf575ae551db7dd9ae25c06e815e7889b856d 100644 --- a/tests/script/general/parser/join_multivnode.sim +++ b/tests/script/general/parser/join_multivnode.sim @@ -315,4 +315,9 @@ if $data03 != 0 then return -1 endi +sql_error select count(join_mt0.c1), first(join_mt0.c1)-last(join_mt1.c1), first(join_mt1.c9) from join_mt0, join_mt1 where join_mt0.t1=join_mt1.t1 and join_mt0.ts=join_mt1.ts;", NULL); +sql select count(join_mt0.c1), first(join_mt0.c1)/count(*), first(join_mt1.c9) from join_mt0, join_mt1 where join_mt0.t1=join_mt1.t1 and join_mt0.ts=join_mt1.ts;", NULL); +sql select count(join_mt0.c1), first(join_mt0.c1)-last(join_mt0.c1), first(join_mt1.c9) from join_mt0, join_mt1 where join_mt0.t1=join_mt1.t1 and join_mt0.ts=join_mt1.ts;", NULL); +sql select last(join_mt0.c1) from join_mt0, join_mt1 where join_mt0.t1=join_mt1.t1 and join_mt0.ts=join_mt1.ts;", NULL); + system sh/exec.sh -n dnode1 -s stop -x SIGINT \ No newline at end of file diff --git a/tests/script/general/parser/testSuite.sim b/tests/script/general/parser/testSuite.sim index b8484089250cbf0b8dd61e48ddb1e044878cb56e..3dd80b8e38c8144fee218ec372e59fe262ac7a15 100644 --- a/tests/script/general/parser/testSuite.sim +++ b/tests/script/general/parser/testSuite.sim @@ -1,51 +1,51 @@ -#sleep 2000 -#run general/parser/alter.sim -#sleep 2000 -#run general/parser/alter1.sim -#sleep 2000 -#run general/parser/alter_stable.sim -#sleep 2000 -#run general/parser/auto_create_tb.sim -#sleep 2000 -#run general/parser/auto_create_tb_drop_tb.sim -#sleep 2000 -#run general/parser/col_arithmetic_operation.sim -#sleep 2000 -#run general/parser/columnValue.sim -#sleep 2000 -#run general/parser/commit.sim -#sleep 2000 -#run general/parser/create_db.sim -#sleep 2000 -#run general/parser/create_mt.sim -#sleep 2000 -#run general/parser/create_tb.sim -#sleep 2000 -#run general/parser/dbtbnameValidate.sim -#sleep 2000 -#run general/parser/fill.sim -#sleep 2000 -#run general/parser/fill_stb.sim -#sleep 2000 -##run general/parser/fill_us.sim # -#sleep 2000 -#run general/parser/first_last.sim -#sleep 2000 -#run general/parser/import_commit1.sim -#sleep 2000 -#run general/parser/import_commit2.sim -#sleep 2000 -#run general/parser/import_commit3.sim -#sleep 2000 -##run general/parser/import_file.sim -#sleep 2000 -#run general/parser/insert_tb.sim -#sleep 2000 -#run general/parser/tags_dynamically_specifiy.sim -#sleep 2000 -#run general/parser/interp.sim -#sleep 2000 -#run general/parser/lastrow.sim +sleep 2000 +run general/parser/alter.sim +sleep 2000 +run general/parser/alter1.sim +sleep 2000 +run general/parser/alter_stable.sim +sleep 2000 +run general/parser/auto_create_tb.sim +sleep 2000 +run general/parser/auto_create_tb_drop_tb.sim +sleep 2000 +run general/parser/col_arithmetic_operation.sim +sleep 2000 +run general/parser/columnValue.sim +sleep 2000 +run general/parser/commit.sim +sleep 2000 +run general/parser/create_db.sim +sleep 2000 +run general/parser/create_mt.sim +sleep 2000 +run general/parser/create_tb.sim +sleep 2000 +run general/parser/dbtbnameValidate.sim +sleep 2000 +run general/parser/fill.sim +sleep 2000 +run general/parser/fill_stb.sim +sleep 2000 +#run general/parser/fill_us.sim # +sleep 2000 +run general/parser/first_last.sim +sleep 2000 +run general/parser/import_commit1.sim +sleep 2000 +run general/parser/import_commit2.sim +sleep 2000 +run general/parser/import_commit3.sim +sleep 2000 +#run general/parser/import_file.sim +sleep 2000 +run general/parser/insert_tb.sim +sleep 2000 +run general/parser/tags_dynamically_specifiy.sim +sleep 2000 +run general/parser/interp.sim +sleep 2000 +run general/parser/lastrow.sim sleep 2000 run general/parser/limit.sim sleep 2000 diff --git a/tests/script/unique/dnode/offline1.sim b/tests/script/unique/dnode/offline1.sim index 02d03dee97be0f2de62b3f8eb18194c595e7b050..beebbfda60c03d24a774347c8ecf8c2f9a4c6c9e 100644 --- a/tests/script/unique/dnode/offline1.sim +++ b/tests/script/unique/dnode/offline1.sim @@ -49,7 +49,7 @@ print dnode1 $data4_2 if $data4_1 != ready then return -1 endi -if $data4_2 != offline then +if $data4_2 == ready then return -1 endi diff --git a/tests/test-all.sh b/tests/test-all.sh index e45dd15fedc999c08037544254f13f78607ee638..f4e992eb5a8da3cd29d34ff6b4e6ebba1e4ff8c2 100755 --- a/tests/test-all.sh +++ b/tests/test-all.sh @@ -9,7 +9,7 @@ NC='\033[0m' function runSimCaseOneByOne { while read -r line; do - if [[ $line =~ ^./test.sh* ]]; then + if [[ $line =~ ^./test.sh* ]] || [[ $line =~ ^run* ]]; then case=`echo $line | grep sim$ |awk '{print $NF}'` start_time=`date +%s`