Merge branch '3.0' into cpwu/3.0

.gitignore

@@ -118,3 +118,4 @@ contrib/*
 !contrib/test
 sql
 debug*/
+.env
\ No newline at end of file

cmake/taostools_CMakeLists.txt.in

@@ -2,7 +2,7 @@
 # taos-tools
 ExternalProject_Add(taos-tools
         GIT_REPOSITORY https://github.com/taosdata/taos-tools.git
-        GIT_TAG 8a5e336
+        GIT_TAG 3c7dafe
         SOURCE_DIR "${TD_SOURCE_DIR}/tools/taos-tools"
         BINARY_DIR ""
         #BUILD_IN_SOURCE TRUE

cmake/taosws_CMakeLists.txt.in

@@ -2,7 +2,7 @@
 # taosws-rs
 ExternalProject_Add(taosws-rs
         GIT_REPOSITORY https://github.com/taosdata/taosws-rs.git
-        GIT_TAG 648cc62
+        GIT_TAG 29424d5
         SOURCE_DIR "${TD_SOURCE_DIR}/tools/taosws-rs"
         BINARY_DIR ""
         #BUILD_IN_SOURCE TRUE

docs/examples/go/connect/afconn/main.go

@@ -2,6 +2,7 @@ package main
 
 import (
 	"fmt"
+	"log"
 
 	"github.com/taosdata/driver-go/v3/af"
 )
@@ -10,7 +11,7 @@ func main() {
 	conn, err := af.Open("localhost", "root", "taosdata", "", 6030)
 	defer conn.Close()
 	if err != nil {
-		fmt.Println("failed to connect, err:", err)
+		log.Fatalln("failed to connect, err:", err)
 	} else {
 		fmt.Println("connected")
 	}

docs/examples/go/connect/cgoexample/main.go

@@ -3,6 +3,7 @@ package main
 import (
 	"database/sql"
 	"fmt"
+	"log"
 
 	_ "github.com/taosdata/driver-go/v3/taosSql"
 )
@@ -11,7 +12,7 @@ func main() {
 	var taosDSN = "root:taosdata@tcp(localhost:6030)/"
 	taos, err := sql.Open("taosSql", taosDSN)
 	if err != nil {
-		fmt.Println("failed to connect TDengine, err:", err)
+		log.Fatalln("failed to connect TDengine, err:", err)
 		return
 	}
 	fmt.Println("Connected")

docs/examples/go/connect/restexample/main.go

@@ -3,6 +3,7 @@ package main
 import (
 	"database/sql"
 	"fmt"
+	"log"
 
 	_ "github.com/taosdata/driver-go/v3/taosRestful"
 )
@@ -11,7 +12,7 @@ func main() {
 	var taosDSN = "root:taosdata@http(localhost:6041)/"
 	taos, err := sql.Open("taosRestful", taosDSN)
 	if err != nil {
-		fmt.Println("failed to connect TDengine, err:", err)
+		log.Fatalln("failed to connect TDengine, err:", err)
 		return
 	}
 	fmt.Println("Connected")

docs/examples/go/connect/wrapper/main.go

-package main
-
-import (
-	"fmt"
-
-	"github.com/taosdata/driver-go/v3/wrapper"
-)
-
-func main() {
-	conn, err := wrapper.TaosConnect("localhost", "root", "taosdata", "", 6030)
-	defer wrapper.TaosClose(conn)
-	if err != nil {
-		fmt.Println("fail to connect, err:", err)
-	} else {
-		fmt.Println("connected")
-	}
-}

docs/examples/go/insert/json/main.go

 package main
 
 import (
-	"fmt"
+	"log"
 
 	"github.com/taosdata/driver-go/v3/af"
 )
@@ -20,7 +20,7 @@ func prepareDatabase(conn *af.Connector) {
 func main() {
 	conn, err := af.Open("localhost", "root", "taosdata", "", 6030)
 	if err != nil {
-		fmt.Println("fail to connect, err:", err)
+		log.Fatalln("fail to connect, err:", err)
 	}
 	defer conn.Close()
 	prepareDatabase(conn)
@@ -32,6 +32,6 @@ func main() {
 
 	err = conn.OpenTSDBInsertJsonPayload(payload)
 	if err != nil {
-		fmt.Println("insert error:", err)
+		log.Fatalln("insert error:", err)
 	}
 }

docs/examples/go/insert/line/main.go

@@ -2,6 +2,7 @@ package main
 
 import (
 	"fmt"
+	"log"
 
 	"github.com/taosdata/driver-go/v3/af"
 )
@@ -33,6 +34,6 @@ func main() {
 
 	err = conn.InfluxDBInsertLines(lines, "ms")
 	if err != nil {
-		fmt.Println("insert error:", err)
+		log.Fatalln("insert error:", err)
 	}
 }

docs/examples/go/insert/sql/main.go

@@ -3,6 +3,7 @@ package main
 import (
 	"database/sql"
 	"fmt"
+	"log"
 
 	_ "github.com/taosdata/driver-go/v3/taosRestful"
 )
@@ -10,28 +11,26 @@ import (
 func createStable(taos *sql.DB) {
 	_, err := taos.Exec("CREATE DATABASE power")
 	if err != nil {
-		fmt.Println("failed to create database, err:", err)
+		log.Fatalln("failed to create database, err:", err)
 	}
 	_, err = taos.Exec("CREATE STABLE power.meters (ts TIMESTAMP, current FLOAT, voltage INT, phase FLOAT) TAGS (location BINARY(64), groupId INT)")
 	if err != nil {
-		fmt.Println("failed to create stable, err:", err)
+		log.Fatalln("failed to create stable, err:", err)
 	}
 }
 
 func insertData(taos *sql.DB) {
-	sql := `INSERT INTO power.d1001 USING power.meters TAGS(California.SanFrancisco, 2) VALUES ('2018-10-03 14:38:05.000', 10.30000, 219, 0.31000) ('2018-10-03 14:38:15.000', 12.60000, 218, 0.33000) ('2018-10-03 14:38:16.800', 12.30000, 221, 0.31000)
-	power.d1002 USING power.meters TAGS(California.SanFrancisco, 3) VALUES ('2018-10-03 14:38:16.650', 10.30000, 218, 0.25000)
-	power.d1003 USING power.meters TAGS(California.LosAngeles, 2) VALUES ('2018-10-03 14:38:05.500', 11.80000, 221, 0.28000) ('2018-10-03 14:38:16.600', 13.40000, 223, 0.29000)
-	power.d1004 USING power.meters TAGS(California.LosAngeles, 3) VALUES ('2018-10-03 14:38:05.000', 10.80000, 223, 0.29000) ('2018-10-03 14:38:06.500', 11.50000, 221, 0.35000)`
+	sql := `INSERT INTO power.d1001 USING power.meters TAGS('California.SanFrancisco', 2) VALUES ('2018-10-03 14:38:05.000', 10.30000, 219, 0.31000) ('2018-10-03 14:38:15.000', 12.60000, 218, 0.33000) ('2018-10-03 14:38:16.800', 12.30000, 221, 0.31000)
+	power.d1002 USING power.meters TAGS('California.SanFrancisco', 3) VALUES ('2018-10-03 14:38:16.650', 10.30000, 218, 0.25000)
+	power.d1003 USING power.meters TAGS('California.LosAngeles', 2) VALUES ('2018-10-03 14:38:05.500', 11.80000, 221, 0.28000) ('2018-10-03 14:38:16.600', 13.40000, 223, 0.29000)
+	power.d1004 USING power.meters TAGS('California.LosAngeles', 3) VALUES ('2018-10-03 14:38:05.000', 10.80000, 223, 0.29000) ('2018-10-03 14:38:06.500', 11.50000, 221, 0.35000)`
 	result, err := taos.Exec(sql)
 	if err != nil {
-		fmt.Println("failed to insert, err:", err)
-		return
+		log.Fatalln("failed to insert, err:", err)
 	}
 	rowsAffected, err := result.RowsAffected()
 	if err != nil {
-		fmt.Println("failed to get affected rows, err:", err)
-		return
+		log.Fatalln("failed to get affected rows, err:", err)
 	}
 	fmt.Println("RowsAffected", rowsAffected)
 }
@@ -40,8 +39,7 @@ func main() {
 	var taosDSN = "root:taosdata@http(localhost:6041)/"
 	taos, err := sql.Open("taosRestful", taosDSN)
 	if err != nil {
-		fmt.Println("failed to connect TDengine, err:", err)
-		return
+		log.Fatalln("failed to connect TDengine, err:", err)
 	}
 	defer taos.Close()
 	createStable(taos)

docs/examples/go/insert/stmt/main.go

@@ -5,8 +5,8 @@ import (
 	"time"
 
 	"github.com/taosdata/driver-go/v3/af"
-	"github.com/taosdata/driver-go/v3/af/param"
 	"github.com/taosdata/driver-go/v3/common"
+	"github.com/taosdata/driver-go/v3/common/param"
 )
 
 func checkErr(err error, prompt string) {

docs/examples/go/insert/telnet/main.go

 package main
 
 import (
-	"fmt"
+	"log"
 
 	"github.com/taosdata/driver-go/v3/af"
 )
@@ -20,7 +20,7 @@ func prepareDatabase(conn *af.Connector) {
 func main() {
 	conn, err := af.Open("localhost", "root", "taosdata", "", 6030)
 	if err != nil {
-		fmt.Println("fail to connect, err:", err)
+		log.Fatalln("fail to connect, err:", err)
 	}
 	defer conn.Close()
 	prepareDatabase(conn)
@@ -37,6 +37,6 @@ func main() {
 
 	err = conn.OpenTSDBInsertTelnetLines(lines)
 	if err != nil {
-		fmt.Println("insert error:", err)
+		log.Fatalln("insert error:", err)
 	}
 }

docs/examples/go/query/sync/main.go

@@ -2,7 +2,7 @@ package main
 
 import (
 	"database/sql"
-	"fmt"
+	"log"
 	"time"
 
 	_ "github.com/taosdata/driver-go/v3/taosRestful"
@@ -12,14 +12,12 @@ func main() {
 	var taosDSN = "root:taosdata@http(localhost:6041)/power"
 	taos, err := sql.Open("taosRestful", taosDSN)
 	if err != nil {
-		fmt.Println("failed to connect TDengine, err:", err)
-		return
+		log.Fatalln("failed to connect TDengine, err:", err)
 	}
 	defer taos.Close()
 	rows, err := taos.Query("SELECT ts, current FROM meters LIMIT 2")
 	if err != nil {
-		fmt.Println("failed to select from table, err:", err)
-		return
+		log.Fatalln("failed to select from table, err:", err)
 	}
 
 	defer rows.Close()
@@ -30,9 +28,9 @@ func main() {
 		}
 		err := rows.Scan(&r.ts, &r.current)
 		if err != nil {
-			fmt.Println("scan error:\n", err)
+			log.Fatalln("scan error:\n", err)
 			return
 		}
-		fmt.Println(r.ts, r.current)
+		log.Fatalln(r.ts, r.current)
 	}
 }

docs/examples/node/nativeexample/async_query_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 const conn = taos.connect({ host: "localhost", database: "power" });
 const cursor = conn.cursor();
 
@@ -18,4 +18,3 @@ try {
     conn.close();
   }, 2000);
 }
-// bug here: jira 14506

docs/examples/node/nativeexample/connect.js

-const taos = require("td2.0-connector");
+const { options, connect } = require("@tdengine/rest");
 
-var conn = taos.connect({
-  host: "localhost",
-  port: 6030,
-  user: "root",
-  password: "taosdata",
-});
-conn.close();
+async function test() {
+  options.path = "/rest/sql";
+  options.host = "localhost";
+  let conn = connect(options);
+  let cursor = conn.cursor();
+  try {
+    let res = await cursor.query("SELECT server_version()");
+    res.toString();
+  } catch (err) {
+    console.log(err);
+  }
+}
+test();
 
-// run with: node connect.js
 // output:
-// Successfully connected to TDengine
+// server_version() |
+// ===================
+// 3.0.0.0          |

docs/examples/node/nativeexample/influxdb_line_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({
   host: "localhost",

docs/examples/node/nativeexample/insert_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({
   host: "localhost",
@@ -11,11 +11,11 @@ try {
   cursor.execute(
     "CREATE STABLE meters (ts TIMESTAMP, current FLOAT, voltage INT, phase FLOAT) TAGS (location BINARY(64), groupId INT)"
   );
-  var sql = `INSERT INTO power.d1001 USING power.meters TAGS(California.SanFrancisco, 2) VALUES ('2018-10-03 14:38:05.000', 10.30000, 219, 0.31000) ('2018-10-03 14:38:15.000', 12.60000, 218, 0.33000) ('2018-10-03 14:38:16.800', 12.30000, 221, 0.31000)
-power.d1002 USING power.meters TAGS(California.SanFrancisco, 3) VALUES ('2018-10-03 14:38:16.650', 10.30000, 218, 0.25000)
-power.d1003 USING power.meters TAGS(California.LosAngeles, 2) VALUES ('2018-10-03 14:38:05.500', 11.80000, 221, 0.28000) ('2018-10-03 14:38:16.600', 13.40000, 223, 0.29000)
-power.d1004 USING power.meters TAGS(California.LosAngeles, 3) VALUES ('2018-10-03 14:38:05.000', 10.80000, 223, 0.29000) ('2018-10-03 14:38:06.500', 11.50000, 221, 0.35000)`;
-  cursor.execute(sql);
+  var sql = `INSERT INTO power.d1001 USING power.meters TAGS('California.SanFrancisco', 2) VALUES ('2018-10-03 14:38:05.000', 10.30000, 219, 0.31000) ('2018-10-03 14:38:15.000', 12.60000, 218, 0.33000) ('2018-10-03 14:38:16.800', 12.30000, 221, 0.31000)
+power.d1002 USING power.meters TAGS('California.SanFrancisco', 3) VALUES ('2018-10-03 14:38:16.650', 10.30000, 218, 0.25000)
+power.d1003 USING power.meters TAGS('California.LosAngeles', 2) VALUES ('2018-10-03 14:38:05.500', 11.80000, 221, 0.28000) ('2018-10-03 14:38:16.600', 13.40000, 223, 0.29000)
+power.d1004 USING power.meters TAGS('California.LosAngeles', 3) VALUES ('2018-10-03 14:38:05.000', 10.80000, 223, 0.29000) ('2018-10-03 14:38:06.500', 11.50000, 221, 0.35000)`;
+  cursor.execute(sql,{'quiet':false});
 } finally {
   cursor.close();
   conn.close();

docs/examples/node/nativeexample/multi_bind_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({
   host: "localhost",
@@ -24,10 +24,10 @@ function insertData() {
   );
 
   // bind table name and tags
-  let tagBind = new taos.TaosBind(2);
-  tagBind.bindBinary("California.SanFrancisco");
-  tagBind.bindInt(2);
-  cursor.stmtSetTbnameTags("d1001", tagBind.getBind());
+  let tagBind = new taos.TaosMultiBindArr(2);
+  tagBind.multiBindBinary(["California.SanFrancisco"]);
+  tagBind.multiBindInt([2]);
+  cursor.stmtSetTbnameTags("d1001", tagBind.getMultiBindArr());
 
   // bind values
   let valueBind = new taos.TaosMultiBindArr(4);

docs/examples/node/nativeexample/opentsdb_json_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({
   host: "localhost",

docs/examples/node/nativeexample/opentsdb_telnet_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({
   host: "localhost",

docs/examples/node/nativeexample/param_bind_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({
   host: "localhost",
@@ -23,25 +23,22 @@ function insertData() {
   );
 
   // bind table name and tags
-  let tagBind = new taos.TaosBind(2);
-  tagBind.bindBinary("California.SanFrancisco");
-  tagBind.bindInt(2);
-  cursor.stmtSetTbnameTags("d1001", tagBind.getBind());
+  let tagBind = new taos.TaosMultiBindArr(2);
+  tagBind.multiBindBinary(["California.SanFrancisco"]);
+  tagBind.multiBindInt([2]);
+  cursor.stmtSetTbnameTags("d1001", tagBind.getMultiBindArr());
 
   // bind values
-  let rows = [
-    [1648432611249, 10.3, 219, 0.31],
-    [1648432611749, 12.6, 218, 0.33],
-  ];
-  for (let row of rows) {
-    let valueBind = new taos.TaosBind(4);
-    valueBind.bindTimestamp(row[0]);
-    valueBind.bindFloat(row[1]);
-    valueBind.bindInt(row[2]);
-    valueBind.bindFloat(row[3]);
-    cursor.stmtBindParam(valueBind.getBind());
+  let rows = [[1648432611249, 1648432611749], [10.3, 12.6], [219, 218], [0.31, 0.33]];
+
+  let valueBind = new taos.TaosMultiBindArr(4);
+  valueBind.multiBindTimestamp(rows[0]);
+  valueBind.multiBindFloat(rows[1]);
+  valueBind.multiBindInt(rows[2]);
+  valueBind.multiBindFloat(rows[3]);
+  cursor.stmtBindParamBatch(valueBind.getMultiBindArr());
   cursor.stmtAddBatch();
-  }
+
 
   // execute
   cursor.stmtExecute();

docs/examples/node/nativeexample/query_example.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({ host: "localhost", database: "power" });
 const cursor = conn.cursor();
@@ -9,8 +9,6 @@ query.execute().then(function (result) {
 
 // output:
 // Successfully connected to TDengine
-// Query OK, 2 row(s) in set (0.00317767s)
-
 //            ts             |         current          |
 // =======================================================
 // 2018-10-03 14:38:05.000   | 10.3                     |

docs/examples/node/nativeexample/subscribe_demo.js

-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 
 const conn = taos.connect({ host: "localhost", database: "power" });
-// 未完成
+var cursor = conn.cursor();
\ No newline at end of file
+
+function runConsumer() {
+
+    // create topic 
+    cursor.execute("create topic topic_name_example as select * from meters");
+
+    let consumer = taos.consumer({
+        'group.id': 'tg2',
+        'td.connect.user': 'root',
+        'td.connect.pass': 'taosdata',
+        'msg.with.table.name': 'true',
+        'enable.auto.commit': 'true'
+    });
+    
+    // subscribe the topic just created.
+    consumer.subscribe("topic_name_example");
+
+    // get subscribe topic list
+    let topicList = consumer.subscription();
+    console.log(topicList);
+
+    for (let i = 0; i < 5; i++) {
+        let msg = consumer.consume(100);
+        console.log(msg.topicPartition);
+        console.log(msg.block);
+        console.log(msg.fields)
+        consumer.commit(msg);
+        console.log(`=======consumer ${i} done`)
+    }
+
+    consumer.unsubscribe();
+    consumer.close();
+
+    // drop topic
+    cursor.execute("drop topic topic_name_example");
+}
+
+
+try {
+    runConsumer();
+} finally {
+
+    setTimeout(() => {
+        cursor.close();
+        conn.close();
+    }, 2000);
+}
\ No newline at end of file

docs/examples/node/package.json

@@ -4,7 +4,7 @@
   "main": "index.js",
   "license": "MIT",
   "dependencies": {
-    "td2.0-connector": "^2.0.12",
-    "td2.0-rest-connector": "^1.0.0"
+    "@tdengine/client": "^3.0.0",
+    "@tdengine/rest": "^3.0.0"
   }
 }

docs/examples/node/restexample/connect.js

-const { options, connect } = require("td2.0-rest-connector");
+const { options, connect } = require("@tdengine/rest");
 
 async function test() {
   options.path = "/rest/sqlt";
@@ -17,4 +17,4 @@ test();
 // output:
 // server_version() |
 // ===================
-// 2.4.0.12         |
+// 3.0.0.0          |

docs/zh/07-develop/03-insert-data/_go_stmt.mdx

@@ -3,6 +3,6 @@
 ```
 
 :::tip
-driver-go 的模块 `github.com/taosdata/driver-go/v2/wrapper` 是 C 接口的底层封装。使用这个模块也可以实现参数绑定写入。
+driver-go 的模块 `github.com/taosdata/driver-go/v3/wrapper` 是 C 接口的底层封装。使用这个模块也可以实现参数绑定写入。
 
 :::

docs/zh/07-develop/07-tmq.md

 ---
-sidebar_label: 消息队列
-description: "数据订阅与推送服务。连续写入到 TDengine 中的时序数据能够被自动推送到订阅客户端。"
-title: 消息队列
+sidebar_label: 数据订阅
+description: "数据订阅与推送服务。写入到 TDengine 中的时序数据能够被自动推送到订阅客户端。"
+title: 数据订阅
 ---
 
-基于数据天然的时间序列特性，TDengine 的数据写入（insert）与消息系统的数据发布（pub）逻辑上一致，均可视为系统中插入一条带时间戳的新记录。同时，TDengine 在内部严格按照数据时间序列单调递增的方式保存数据。本质上来说，TDengine 中每一张表均可视为一个标准的消息队列。
+为了帮助应用实时获取写入 TDengine 的数据，或者以事件到达顺序处理数据，TDengine提供了类似消息队列产品的数据订阅、消费接口。这样在很多场景下，采用 TDengine 的时序数据处理系统不再需要集成消息队列产品，比如 kafka, 从而简化系统设计的复杂度，降低运营维护成本。
 
-TDengine 内嵌支持消息订阅与推送服务（下文都简称TMQ）。使用系统提供的 API，用户可使用普通查询语句订阅数据库中的一张或多张表，或整个库。客户端启动订阅后，定时或按需轮询服务器是否有新的记录到达，有新的记录到达就会将结果反馈到客户。
+与 kafka 一样，你需要定义 topic, 但 TDengine 的 topic 可以是一张超级表，或一张子表。不仅如此，你可以通过标签、表名、列、表达式等多种方法过滤所需数据，并且支持对数据进行函数变换、预处理（包括标量udf计算）。与其他消息队列软件相比，这是 TDengine 数据订阅功能的最大的优势，它提供了更大的灵活性，数据的颗粒度可以由应用随时调整，而且数据的过滤交给 TDengine，而不是应用完成，有效的减少传输的数据量。
 
-TMQ提供了提交机制来保证消息队列的可靠性和正确性。在调用方法上，支持自动提交和手动提交。
+消费者订阅 topic 后，可以实时获得最新的数据。多个消费者可以组成一个消费者组 (consumer group), 一个消费者组里的多个消费者共享消费进度，便于多线程分布式的消费数据，提高数据通吐率。但不同消费者组即使消费同一个topic, 并不共享消费进度。一个消费者组可以订阅多个 topic。如果订阅的是超级表，数据可能会分布在多个不同的vnode上，也就是多个shard上，这样一个消费组里有多个消费者可以提高消费效率。TDengine 的消息队列提供了消息的ACK机制，在宕机、重启等复杂环境下确保at least once消费。
+
+为了实现上述功能，TDengine 采用了灵活的 WAL (Write-Ahead-Log) 文件切换与保留机制：可以按照时间或文件大小来保留WAL文件（详见create database语句）。在消费时，TDengine 从 WAL 中获取数据，并经过过滤、变换等操作，将数据推送给消费者。
+
+本文档不对消息队列本身的基础知识做介绍，如果需要了解，请自行搜索。
+
+## 主要数据结构和API
 
 TMQ 的 API 中，与订阅相关的主要数据结构和API如下：
 
@@ -47,7 +53,9 @@ DLL_EXPORT void           tmq_conf_set_auto_commit_cb(tmq_conf_t *conf, tmq_comm
 
 这些 API 的文档请见 [C/C++ Connector](/reference/connector/cpp)，下面介绍一下它们的具体用法（超级表和子表结构请参考“数据建模”一节），完整的示例代码可以在 [tmq.c](https://github.com/taosdata/TDengine/blob/3.0/examples/c/tmq.c) 看到。
 
-一、首先完成建库、建一张超级表和多张子表，并每个子表插入若干条数据记录：
+## 写入数据
+
+首先完成建库、建一张超级表和多张子表操作，然后就可以写入数据了，比如：
 
 ```sql
 drop database if exists tmqdb;
@@ -63,14 +71,15 @@ insert into tmqdb.ctb2 values(now, 2, 2, 'a1')(now+1s, 22, 22, 'a22');
 insert into tmqdb.ctb3 values(now, 3, 3, 'a1')(now+1s, 33, 33, 'a33');
 ```
 
-二、创建topic：
+## 创建topic：
 
 ```sql
 create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
 ```
 
-注：TMQ支持多种订阅类型：
-1、列订阅
+TMQ支持多种订阅类型：
+
+### 列订阅
 
 语法：CREATE TOPIC topic_name as subquery
 通过select语句订阅（包括select *，或select ts, c1等指定列描述订阅，可以带条件过滤、标量函数计算，但不支持聚合函数、不支持时间窗口聚合）
@@ -79,25 +88,18 @@ create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
 - 被订阅或用于计算的column和tag不可被删除、修改
 - 若发生schema变更，新增的column不出现在结果中
 
-2、超级表订阅
+### 超级表订阅
 语法：CREATE TOPIC topic_name AS STABLE stbName
 
-- 订阅某超级表的全部数据，schema变更不受限，schema变更后写入的数据将以最新schema返回
-- 在tmq的返回消息中schema是块级别的，每块的schema可能不一样
-- 列变更后写入的数据若未落盘，将以写入时的schema返回
-- 列变更后写入的数据若已落盘，将以落盘时的schema返回
-
-3、db订阅
-语法：CREATE TOPIC topic_name AS DATABASE db_name
-
-- 订阅某一db的全部数据，schema变更不受限
-- 在tmq的返回消息中schema是块级别的，每块的schema可能不一样
-- 列变更后写入的数据若未落盘，将以写入时的schema返回
-- 列变更后写入的数据若已落盘，将以落盘时的schema返回
+与select * from stbName订阅的区别是：
+- 不会限制用户的schema变更
+- 返回的是非结构化的数据：返回数据的schema会随之超级表的schema变化而变化
+- 用户对于要处理的每一个数据块都可能有不同的schema，因此，必须重新获取schema
+- 返回数据不带有tag
 
-三、创建consumer
+## 创建 consumer 以及consumer group
 
-目前支持的config：
+对于consumer, 目前支持的config包括：
 
 | 参数名称                     | 参数值                         | 备注                                                   |
 | ---------------------------- | ------------------------------ | ------------------------------------------------------ |
@@ -131,7 +133,12 @@ create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
   return tmq;
 ```
 
-四、创建订阅主题列表
+上述配置中包括consumer group ID，如果多个 consumer 指定的 consumer group ID一样，则自动形成一个consumer group，共享消费进度。
+
+
+## 创建 topic 列表
+
+单个consumer支持同时订阅多个topic。
 
 ```sql
   tmq_list_t* topicList = tmq_list_new();
@@ -139,9 +146,7 @@ create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
   return topicList;
 ```
 
-单个consumer支持同时订阅多个topic。
-
-五、启动订阅并开始消费
+## 启动订阅并开始消费
 
 ```sql
   /* 启动订阅 */
@@ -151,9 +156,9 @@ create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
   /* 循环poll消息 */
   int32_t totalRows = 0;
   int32_t msgCnt = 0;
-  int32_t consumeDelay = 5000;
+  int32_t timeOut = 5000;
   while (running) {
-    TAOS_RES* tmqmsg = tmq_consumer_poll(tmq, consumeDelay);
+    TAOS_RES* tmqmsg = tmq_consumer_poll(tmq, timeOut);
     if (tmqmsg) {
       msgCnt++;
       totalRows += msg_process(tmqmsg);
@@ -199,7 +204,7 @@ create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
 }
 ```
 
-五、结束消费
+## 结束消费
 
 ```sql
   /* 取消订阅 */
@@ -209,7 +214,7 @@ create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
   tmq_consumer_close(tmq);
 ```
 
-六、删除topic
+## 删除topic
 
 如果不再需要，可以删除创建topic，但注意：只有没有被订阅的topic才能别删除。
 
@@ -218,7 +223,7 @@ create topic topicName as select ts, c1, c2, c3 from tmqdb.stb where c1 > 1;
   drop topic topicName;
 ```
 
-七、状态查看
+## 状态查看
 
 1、topics：查询已经创建的topic
 

docs/zh/14-reference/03-connector/node.mdx

@@ -15,11 +15,11 @@ import NodeOpenTSDBTelnet from "../../07-develop/03-insert-data/_js_opts_telnet.
 import NodeOpenTSDBJson from "../../07-develop/03-insert-data/_js_opts_json.mdx";
 import NodeQuery from "../../07-develop/04-query-data/_js.mdx";
 
-`td2.0-connector` 和 `td2.0-rest-connector` 是 TDengine 的官方 Node.js 语言连接器。Node.js 开发人员可以通过它开发可以存取 TDengine 集群数据的应用软件。
+`@tdengine/client` 和 `@tdengine/rest` 是 TDengine 的官方 Node.js 语言连接器。 Node.js 开发人员可以通过它开发可以存取 TDengine 集群数据的应用软件。注意：从 TDengine 3.0 开始 Node.js 原生连接器的包名由 `td2.0-connector` 改名为 `@tdengine/client` 而 rest 连接器的包名由 `td2.0-rest-connector` 改为 `@tdengine/rest`。并且不与 TDengine 2.x 兼容。
 
-`td2.0-connector` 是**原生连接器**，它通过 TDengine 客户端驱动程序（taosc）连接 TDengine 运行实例，支持数据写入、查询、订阅、schemaless 接口和参数绑定接口等功能。`td2.0-rest-connector` 是 **REST 连接器**，它通过 taosAdapter 提供的 REST 接口连接 TDengine 的运行实例。REST 连接器可以在任何平台运行，但性能略为下降，接口实现的功能特性集合和原生接口有少量不同。
+`@tdengine/client` 是**原生连接器**，它通过 TDengine 客户端驱动程序（taosc）连接 TDengine 运行实例，支持数据写入、查询、订阅、schemaless 接口和参数绑定接口等功能。`@tdengine/rest` 是 **REST 连接器**，它通过 taosAdapter 提供的 REST 接口连接 TDengine 的运行实例。REST 连接器可以在任何平台运行，但性能略为下降，接口实现的功能特性集合和原生接口有少量不同。
 
-Node.js 连接器源码托管在 [GitHub](https://github.com/taosdata/taos-connector-node)。
+Node.js 连接器源码托管在 [GitHub](https://github.com/taosdata/taos-connector-node/tree/3.0)。
 
 ## 支持的平台
 
@@ -58,7 +58,7 @@ REST 连接器支持所有能运行 Node.js 的平台。
 <TabItem value="Linux" label="Linux 系统安装依赖工具">
 
 - `python` (建议`v2.7` , `v3.x.x` 目前还不支持)
-- `td2.0-connector` 2.0.6 支持 Node.js LTS v10.9.0 或更高版本, Node.js LTS v12.8.0 或更高版本；2.0.5 及更早版本支持 Node.js LTS v10.x 版本。其他版本可能存在包兼容性的问题
+- `@tdengine/client` 3.0.0 支持 Node.js LTS v10.9.0 或更高版本, Node.js LTS v12.8.0 或更高版本；其他版本可能存在包兼容性的问题
 - `make`
 - C 语言编译器，[GCC](https://gcc.gnu.org) v4.8.5 或更高版本
 
@@ -90,14 +90,14 @@ REST 连接器支持所有能运行 Node.js 的平台。
 <TabItem value="install_native" label="安装原生连接器">
 
 ```bash
-npm install td2.0-connector
+npm install @tdengine/client
 ```
 
 </TabItem>
 <TabItem value="install_rest" label="安装 REST 连接器">
 
 ```bash
-npm i td2.0-rest-connector
+npm install @tdengine/rest
 ```
 
 </TabItem>
@@ -109,13 +109,13 @@ npm i td2.0-rest-connector
 
 验证方法：
 
-- 新建安装验证目录，例如：`~/tdengine-test`，下载 GitHub 上 [nodejsChecker.js 源代码](https://github.com/taosdata/TDengine/tree/develop/examples/nodejs/nodejsChecker.js)到本地。
+- 新建安装验证目录，例如：`~/tdengine-test`，下载 GitHub 上 [nodejsChecker.js 源代码](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/nodejsChecker.js)到本地。
 
 - 在命令行中执行以下命令。
 
 ```bash
 npm init -y
-npm install td2.0-connector
+npm install @tdengine/client
 node nodejsChecker.js host=localhost
 ```
 
@@ -128,11 +128,11 @@ node nodejsChecker.js host=localhost
 <Tabs defaultValue="native">
 <TabItem value="native" label="原生连接">
 
-安装并引用 `td2.0-connector` 包。
+安装并引用 `@tdengine/client` 包。
 
 ```javascript
 //A cursor also needs to be initialized in order to interact with TDengine from Node.js.
-const taos = require("td2.0-connector");
+const taos = require("@tdengine/client");
 var conn = taos.connect({
   host: "127.0.0.1",
   user: "root",
@@ -149,12 +149,12 @@ conn.close();
 </TabItem>
 <TabItem value="rest" label="REST 连接">
 
-安装并引用 `td2.0-rest-connector` 包。
+安装并引用 `@tdengine/rest` 包。
 
 ```javascript
 //A cursor also needs to be initialized in order to interact with TDengine from Node.js.
-import { options, connect } from "td2.0-rest-connector";
-options.path = "/rest/sqlt";
+import { options, connect } from "@tdengine/rest";
+options.path = "/rest/sql";
 // set host
 options.host = "localhost";
 // set other options like user/passwd
@@ -190,26 +190,23 @@ let cursor = conn.cursor();
 
 <NodeQuery />
 
+
 ## 更多示例程序
 
 | 示例程序                                                                                                                                   | 示例程序描述                           |
 | ------------------------------------------------------------------------------------------------------------------------------------------ | -------------------------------------- |
-| [connection](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/cursorClose.js)                                  | 建立连接的示例。                       |
-| [stmtBindBatch](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/stmtBindParamBatchSample.js)                  | 绑定多行参数插入的示例。               |
-| [stmtBind](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/stmtBindParamSample.js)                            | 一行一行绑定参数插入的示例。           |
-| [stmtBindSingleParamBatch](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/stmtBindSingleParamBatchSample.js) | 按列绑定参数插入的示例。               |
-| [stmtUseResult](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/stmtUseResultSample.js)                       | 绑定参数查询的示例。                   |
-| [json tag](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/testJsonTag.js)                                    | Json tag 的使用示例。                  |
-| [Nanosecond](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/testNanoseconds.js)                              | 时间戳为纳秒精度的使用的示例。         |
-| [Microsecond](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/testMicroseconds.js)                            | 时间戳为微秒精度的使用的示例。         |
-| [schemless insert](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/testSchemalessInsert.js)                   | schemless 插入的示例。                 |
-| [subscribe](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/testSubscribe.js)                                 | 订阅的使用示例。                       |
-| [asyncQuery](https://github.com/taosdata/taos-connector-node/tree/develop/nodejs/examples/tset.js)                                         | 异步查询的使用示例。                   |
-| [REST](https://github.com/taosdata/taos-connector-node/blob/develop/typescript-rest/example/example.ts)                                    | 使用 REST 连接的 TypeScript 使用示例。 |
+| [basicUse](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/queryExample.js)                                  | 基本的使用如如建立连接，执行 SQL 等操作。                       |
+| [stmtBindBatch](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/bindParamBatch.js)                  | 绑定多行参数插入的示例。               | |
+| [stmtBindSingleParamBatch](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/bindSingleParamBatch.js) | 按列绑定参数插入的示例。               |
+| [stmtQuery](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/stmtQuery.js)                       | 绑定参数查询的示例。                   |
+| [schemless insert](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/schemaless.js)                   | schemless 插入的示例。                 |
+| [TMQ](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/tmq.js)                                 | 订阅的使用示例。                       |
+| [asyncQuery](https://github.com/taosdata/taos-connector-node/blob/3.0/nodejs/examples/asyncQueryExample.js)                                         | 异步查询的使用示例。                   |
+| [REST](https://github.com/taosdata/taos-connector-node/blob/3.0/typescript-rest/example/example.ts)                                    | 使用 REST 连接的 TypeScript 使用示例。 |
 
 ## 使用限制
 
-Node.js 连接器 >= v2.0.6 目前支持 node 的版本为：支持 >=v12.8.0 <= v12.9.1 || >=v10.20.0 <= v10.9.0 ；2.0.5 及更早版本支持 v10.x 版本，其他版本可能存在包兼容性的问题。
+native 连接器（`@tdengine/client`） >= v3.0.0 目前支持 node 的版本为：支持 >=v12.8.0 <= v12.9.1 || >=v10.20.0 <= v10.9.0 ；2.0.5 及更早版本支持 v10.x 版本，其他版本可能存在包兼容性的问题。
 
 ## 其他说明
 
@@ -225,7 +222,7 @@ Node.js 连接器的使用参见[视频教程](https://www.taosdata.com/blog/202
 
 2. Node.js 版本
 
-  连接器 >v2.0.6 目前兼容的 Node.js 版本为：>=v10.20.0 <= v10.9.0 || >=v12.8.0 <= v12.9.1
+   原生连接器 `@tdengine/client` 目前兼容的 Node.js 版本为：>=v10.20.0 <= v10.9.0 || >=v12.8.0 <= v12.9.1
 
 3. "Unable to establish connection"，"Unable to resolve FQDN"
 
@@ -235,17 +232,21 @@ Node.js 连接器的使用参见[视频教程](https://www.taosdata.com/blog/202
 
 ### 原生连接器
 
-| td2.0-connector 版本 | 说明                                                             |
-| -------------------- | ---------------------------------------------------------------- |
-| 2.0.12               | 修复 cursor.close() 报错的 bug。                                 |
-| 2.0.11               | 支持绑定参数、json tag、schemaless 接口等功能。                  |
-| 2.0.10               | 支持连接管理，普通查询、连续查询、获取系统信息、订阅功能等功能。 |
-
+| package name     | version | TDengine version    | 说明                                                             |
+|------------------|---------|---------------------|------------------------------------------------------------------|
+| @tdengine/client | 3.0.0   | 3.0.0               | 支持TDengine 3.0 且不与2.x 兼容。                                                          |
+| td2.0-connector  | 2.0.12  | 2.4.x；2.5.x；2.6.x | 修复 cursor.close() 报错的 bug。                                 |
+| td2.0-connector  | 2.0.11  | 2.4.x；2.5.x；2.6.x | 支持绑定参数、json tag、schemaless 接口等功能。                  |
+| td2.0-connector  | 2.0.10  | 2.4.x；2.5.x；2.6.x | 支持连接管理，普通查询、连续查询、获取系统信息、订阅功能等功能。 |
 ### REST 连接器
 
-| td2.0-rest-connector 版本 | 说明                                                             |
-| ------------------------- | ---------------------------------------------------------------- |
-| 1.0.3                     | 支持连接管理、普通查询、获取系统信息、错误信息、连续查询等功能。 |
+| package name         | version | TDengine version    | 说明                                                                      |
+|----------------------|---------|---------------------|---------------------------------------------------------------------------|
+| @tdengine/rest       | 3.0.0   | 3.0.0               | 支持 TDegnine 3.0，且不与2.x 兼容。                                               |
+| td2.0-rest-connector | 1.0.7   | 2.4.x；2.5.x；2.6.x | 移除默认端口 6041。                                                       |
+| td2.0-rest-connector | 1.0.6   | 2.4.x；2.5.x；2.6.x | 修复create，insert，update，alter 等SQL 执行返回的 affectRows 错误的bug。 |
+| td2.0-rest-connector | 1.0.5   | 2.4.x；2.5.x；2.6.x | 支持云服务 cloud Token；                                                  |
+| td2.0-rest-connector | 1.0.3   | 2.4.x；2.5.x；2.6.x | 支持连接管理、普通查询、获取系统信息、错误信息、连续查询等功能。          |
 
 ## API 参考
 

docs/zh/14-reference/11-docker/index.md

@@ -147,7 +147,7 @@ import (
     "fmt"
     "time"
 
-    _ "github.com/taosdata/driver-go/v2/taosSql"
+    _ "github.com/taosdata/driver-go/v3/taosSql"
 )
 
 type config struct {

docs/zh/28-releases.md

+---
+sidebar_label: 发布历史
+title: 发布历史
+---
+
+import Release from "/components/Release";
+
+
+<Release versionPrefix="3.0" />

include/common/tmsg.h

@@ -2657,6 +2657,34 @@ typedef struct {
   SEpSet  epSet;
 } SVgEpSet;
 
+typedef struct {
+  int64_t refId;
+  int64_t suid;
+  int8_t  level;
+} SRSmaFetchMsg;
+
+static FORCE_INLINE int32_t tEncodeSRSmaFetchMsg(SEncoder* pCoder, const SRSmaFetchMsg* pReq) {
+  if (tStartEncode(pCoder) < 0) return -1;
+
+  if (tEncodeI64(pCoder, pReq->refId) < 0) return -1;
+  if (tEncodeI64(pCoder, pReq->suid) < 0) return -1;
+  if (tEncodeI8(pCoder, pReq->level) < 0) return -1;
+
+  tEndEncode(pCoder);
+  return 0;
+}
+
+static FORCE_INLINE int32_t tDecodeSRSmaFetchMsg(SDecoder* pCoder, SRSmaFetchMsg* pReq) {
+  if (tStartDecode(pCoder) < 0) return -1;
+
+  if (tDecodeI64(pCoder, &pReq->refId) < 0) return -1;
+  if (tDecodeI64(pCoder, &pReq->suid) < 0) return -1;
+  if (tDecodeI8(pCoder, &pReq->level) < 0) return -1;
+
+  tEndDecode(pCoder);
+  return 0;
+}
+
 typedef struct {
   int8_t         version;       // for compatibility(default 0)
   int8_t         intervalUnit;  // MACRO: TIME_UNIT_XXX
@@ -3055,6 +3083,7 @@ int32_t tEncodeDeleteRes(SEncoder* pCoder, const SDeleteRes* pRes);
 int32_t tDecodeDeleteRes(SDecoder* pCoder, SDeleteRes* pRes);
 
 typedef struct {
+  int32_t msgIdx;
   int32_t msgType;
   int32_t msgLen;
   void*   msg;
@@ -3068,6 +3097,7 @@ typedef struct {
 
 typedef struct {
   int32_t reqType;
+  int32_t msgIdx;
   int32_t msgLen;
   int32_t rspCode;
   void*   msg;

include/libs/catalog/catalog.h

@@ -58,12 +58,17 @@ typedef struct SDbInfo {
   int64_t dbId;
 } SDbInfo;
 
+typedef struct STablesReq {
+  char    dbFName[TSDB_DB_FNAME_LEN];
+  SArray* pTables;
+} STablesReq;
+
 typedef struct SCatalogReq {
   SArray* pDbVgroup;      // element is db full name
   SArray* pDbCfg;         // element is db full name
   SArray* pDbInfo;        // element is db full name
-  SArray* pTableMeta;     // element is SNAME
-  SArray* pTableHash;     // element is SNAME
+  SArray* pTableMeta;     // element is STablesReq
+  SArray* pTableHash;     // element is STablesReq
   SArray* pUdf;           // element is udf name
   SArray* pIndex;         // element is index name
   SArray* pUser;          // element is SUserAuthInfo

include/libs/executor/executor.h

@@ -103,7 +103,7 @@ int32_t qUpdateQualifiedTableId(qTaskInfo_t tinfo, const SArray* tableIdList, bo
  * @return
  */
 int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, struct SSubplan* pPlan,
-                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, const char* sql, EOPTR_EXEC_MODEL model);
+                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, char* sql, EOPTR_EXEC_MODEL model);
 
 /**
  *

include/libs/function/functionMgt.h

@@ -202,6 +202,7 @@ bool fmIsForbidStreamFunc(int32_t funcId);
 bool fmIsIntervalInterpoFunc(int32_t funcId);
 bool fmIsInterpFunc(int32_t funcId);
 bool fmIsLastRowFunc(int32_t funcId);
+bool fmIsNotNullOutputFunc(int32_t funcId);
 bool fmIsSelectValueFunc(int32_t funcId);
 bool fmIsSystemInfoFunc(int32_t funcId);
 bool fmIsImplicitTsFunc(int32_t funcId);

include/libs/transport/trpc.h

@@ -43,11 +43,12 @@ typedef struct SRpcHandleInfo {
   // rpc info
   void   *handle;         // rpc handle returned to app
   int64_t refId;          // refid, used by server
-  int32_t  noResp;         // has response or not(default 0, 0: resp, 1: no resp);
-  int32_t  persistHandle;  // persist handle or not
-  STraceId traceId;
+  int8_t  noResp;         // has response or not(default 0, 0: resp, 1: no resp)
+  int8_t  persistHandle;  // persist handle or not
   int8_t  hasEpSet;
 
+  STraceId traceId;
+
   // app info
   void *ahandle;  // app handle set by client
   void *wrapper;  // wrapper handle
@@ -69,8 +70,9 @@ typedef struct SRpcMsg {
   SRpcHandleInfo info;
 } SRpcMsg;
 
-typedef void (*RpcCfp)(void *parent, SRpcMsg *, SEpSet *rf);
+typedef void (*RpcCfp)(void *parent, SRpcMsg *, SEpSet *epset);
 typedef bool (*RpcRfp)(int32_t code, tmsg_t msgType);
+typedef bool (*RpcTfp)(int32_t code, tmsg_t msgType);
 
 typedef struct SRpcInit {
   char     localFqdn[TSDB_FQDN_LEN];
@@ -84,12 +86,15 @@ typedef struct SRpcInit {
   // the following is for client app ecurity only
   char *user;  // user name
 
-  // call back to process incoming msg, code shall be ignored by server app
+  // call back to process incoming msg
   RpcCfp cfp;
 
-  // user defined retry func
+  // retry not not for particular msg
   RpcRfp rfp;
 
+  // set up timeout for particular msg
+  RpcTfp tfp;
+
   void *parent;
 } SRpcInit;
 

include/util/taoserror.h

@@ -46,6 +46,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_RPC_FQDN_ERROR                TAOS_DEF_ERROR_CODE(0, 0x0015)
 #define TSDB_CODE_RPC_PORT_EADDRINUSE           TAOS_DEF_ERROR_CODE(0, 0x0017)
 #define TSDB_CODE_RPC_BROKEN_LINK               TAOS_DEF_ERROR_CODE(0, 0x0018)
+#define TSDB_CODE_RPC_TIMEOUT                   TAOS_DEF_ERROR_CODE(0, 0x0019)
 
 //common & util
 #define TSDB_CODE_TIME_UNSYNCED                 TAOS_DEF_ERROR_CODE(0, 0x0013) 

include/util/tlockfree.h

@@ -27,7 +27,7 @@ typedef void (*_ref_fn_t)(const void *pObj);
 
 #define T_REF_DECLARE()   \
   struct {                \
-    int32_t val;        \
+    volatile int32_t val; \
   } _ref;
 
 #define T_REF_REGISTER_FUNC(s, e) \

source/client/src/clientEnv.c

@@ -109,6 +109,14 @@ static bool clientRpcRfp(int32_t code, tmsg_t msgType) {
   }
 }
 
+// start timer for particular msgType
+static bool clientRpcTfp(int32_t code, tmsg_t msgType) {
+  if (msgType == TDMT_VND_SUBMIT || msgType == TDMT_VND_CREATE_TABLE) {
+    return true;
+  }
+  return false;
+}
+
 // TODO refactor
 void *openTransporter(const char *user, const char *auth, int32_t numOfThread) {
   SRpcInit rpcInit;
@@ -118,6 +126,7 @@ void *openTransporter(const char *user, const char *auth, int32_t numOfThread) {
   rpcInit.numOfThreads = numOfThread;
   rpcInit.cfp = processMsgFromServer;
   rpcInit.rfp = clientRpcRfp;
+  rpcInit.tfp = clientRpcTfp;
   rpcInit.sessions = 1024;
   rpcInit.connType = TAOS_CONN_CLIENT;
   rpcInit.user = (char *)user;

source/client/src/clientImpl.c

@@ -1308,8 +1308,8 @@ int32_t doProcessMsgFromServer(void* param) {
   char      tbuf[40] = {0};
   TRACE_TO_STR(trace, tbuf);
 
-  tscDebug("processMsgFromServer handle %p, message: %s, code: %s, gtid: %s", pMsg->info.handle,
-           TMSG_INFO(pMsg->msgType), tstrerror(pMsg->code), tbuf);
+  tscDebug("processMsgFromServer handle %p, message: %s, size:%d, code: %s, gtid: %s", pMsg->info.handle,
+           TMSG_INFO(pMsg->msgType), pMsg->contLen, tstrerror(pMsg->code), tbuf);
 
   if (pSendInfo->requestObjRefId != 0) {
     SRequestObj* pRequest = (SRequestObj*)taosAcquireRef(clientReqRefPool, pSendInfo->requestObjRefId);
@@ -1922,7 +1922,7 @@ _OVER:
   return code;
 }
 
-int32_t appendTbToReq(SArray* pList, int32_t pos1, int32_t len1, int32_t pos2, int32_t len2, const char* str,
+int32_t appendTbToReq(SHashObj* pHash, int32_t pos1, int32_t len1, int32_t pos2, int32_t len2, const char* str,
                       int32_t acctId, char* db) {
   SName name;
 
@@ -1957,20 +1957,33 @@ int32_t appendTbToReq(SArray* pList, int32_t pos1, int32_t len1, int32_t pos2, i
     return -1;
   }
 
-  taosArrayPush(pList, &name);
+  char dbFName[TSDB_DB_FNAME_LEN];
+  sprintf(dbFName, "%d.%.*s", acctId, dbLen, dbName);
+
+  STablesReq* pDb = taosHashGet(pHash, dbFName, strlen(dbFName));
+  if (pDb) {
+    taosArrayPush(pDb->pTables, &name);
+  } else {
+    STablesReq db;
+    db.pTables = taosArrayInit(20, sizeof(SName));
+    strcpy(db.dbFName, dbFName);
+    taosArrayPush(db.pTables, &name);
+    taosHashPut(pHash, dbFName, strlen(dbFName), &db, sizeof(db));
+  }
 
   return TSDB_CODE_SUCCESS;
 }
 
 int32_t transferTableNameList(const char* tbList, int32_t acctId, char* dbName, SArray** pReq) {
-  *pReq = taosArrayInit(10, sizeof(SName));
-  if (NULL == *pReq) {
+  SHashObj* pHash = taosHashInit(3, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
+  if (NULL == pHash) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return terrno;
   }
 
   bool    inEscape = false;
   int32_t code = 0;
+  void *pIter = NULL;
 
   int32_t vIdx = 0;
   int32_t vPos[2];
@@ -1985,7 +1998,7 @@ int32_t transferTableNameList(const char* tbList, int32_t acctId, char* dbName,
         vLen[vIdx] = i - vPos[vIdx];
       }
 
-      code = appendTbToReq(*pReq, vPos[0], vLen[0], vPos[1], vLen[1], tbList, acctId, dbName);
+      code = appendTbToReq(pHash, vPos[0], vLen[0], vPos[1], vLen[1], tbList, acctId, dbName);
       if (code) {
         goto _return;
       }
@@ -2035,7 +2048,7 @@ int32_t transferTableNameList(const char* tbList, int32_t acctId, char* dbName,
         vLen[vIdx] = i - vPos[vIdx];
       }
 
-      code = appendTbToReq(*pReq, vPos[0], vLen[0], vPos[1], vLen[1], tbList, acctId, dbName);
+      code = appendTbToReq(pHash, vPos[0], vLen[0], vPos[1], vLen[1], tbList, acctId, dbName);
       if (code) {
         goto _return;
       }
@@ -2067,14 +2080,31 @@ int32_t transferTableNameList(const char* tbList, int32_t acctId, char* dbName,
     goto _return;
   }
 
+  int32_t dbNum = taosHashGetSize(pHash);
+  *pReq = taosArrayInit(dbNum, sizeof(STablesReq));
+  pIter = taosHashIterate(pHash, NULL);
+  while (pIter) {
+    STablesReq* pDb = (STablesReq*)pIter;
+    taosArrayPush(*pReq, pDb);
+    pIter = taosHashIterate(pHash, pIter);
+  }
+
+  taosHashCleanup(pHash);
+
   return TSDB_CODE_SUCCESS;
 
 _return:
 
   terrno = TSDB_CODE_TSC_INVALID_OPERATION;
 
-  taosArrayDestroy(*pReq);
-  *pReq = NULL;
+  pIter = taosHashIterate(pHash, NULL);
+  while (pIter) {
+    STablesReq* pDb = (STablesReq*)pIter;
+    taosArrayDestroy(pDb->pTables);
+    pIter = taosHashIterate(pHash, pIter);
+  }
+
+  taosHashCleanup(pHash);
 
   return terrno;
 }

source/common/src/systable.c

@@ -308,9 +308,9 @@ static const SSysDbTableSchema offsetSchema[] = {
 };
 
 static const SSysDbTableSchema querySchema[] = {
-    {.name = "query_id", .bytes = TSDB_QUERY_ID_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR},
-    {.name = "req_id", .bytes = 8, .type = TSDB_DATA_TYPE_UBIGINT},
-    {.name = "connId", .bytes = 4, .type = TSDB_DATA_TYPE_UINT},
+    {.name = "kill_id", .bytes = TSDB_QUERY_ID_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR},
+    {.name = "query_id", .bytes = 8, .type = TSDB_DATA_TYPE_UBIGINT},
+    {.name = "conn_id", .bytes = 4, .type = TSDB_DATA_TYPE_UINT},
     {.name = "app", .bytes = TSDB_APP_NAME_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR},
     {.name = "pid", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "user", .bytes = TSDB_USER_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR},
@@ -354,11 +354,19 @@ static const SSysTableMeta perfsMeta[] = {
     {TSDB_PERFS_TABLE_APPS, appSchema, tListLen(appSchema)}};
 
 void getInfosDbMeta(const SSysTableMeta** pInfosTableMeta, size_t* size) {
+  if (pInfosTableMeta) {
     *pInfosTableMeta = infosMeta;
+  }
+  if (size) {
     *size = tListLen(infosMeta);
+  }
 }
 
 void getPerfDbMeta(const SSysTableMeta** pPerfsTableMeta, size_t* size) {
+  if (pPerfsTableMeta) {
     *pPerfsTableMeta = perfsMeta;
+  }
+  if (size) {
     *size = tListLen(perfsMeta);
+  }
 }

source/dnode/mnode/impl/src/mndDb.c

@@ -1691,13 +1691,17 @@ static int32_t mndRetrieveDbs(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBloc
   if (!pShow->sysDbRsp) {
     SDbObj infoschemaDb = {0};
     setInformationSchemaDbCfg(&infoschemaDb);
-    dumpDbInfoData(pBlock, &infoschemaDb, pShow, numOfRows, 14, true, 0, 1);
+    size_t numOfTables = 0;
+    getInfosDbMeta(NULL, &numOfTables);
+    dumpDbInfoData(pBlock, &infoschemaDb, pShow, numOfRows, numOfTables, true, 0, 1);
 
     numOfRows += 1;
 
     SDbObj perfschemaDb = {0};
     setPerfSchemaDbCfg(&perfschemaDb);
-    dumpDbInfoData(pBlock, &perfschemaDb, pShow, numOfRows, 3, true, 0, 1);
+    numOfTables = 0;
+    getPerfDbMeta(NULL, &numOfTables);
+    dumpDbInfoData(pBlock, &perfschemaDb, pShow, numOfRows, numOfTables, true, 0, 1);
 
     numOfRows += 1;
     pShow->sysDbRsp = true;

source/dnode/mnode/impl/src/mndQuery.c

@@ -84,6 +84,9 @@ int32_t mndProcessBatchMetaMsg(SRpcMsg *pMsg) {
   }
   
   for (int32_t i = 0; i < msgNum; ++i) {
+    req.msgIdx = ntohl(*(int32_t*)((char*)pMsg->pCont + offset));
+    offset += sizeof(req.msgIdx);
+
     req.msgType = ntohl(*(int32_t*)((char*)pMsg->pCont + offset));
     offset += sizeof(req.msgType);
 
@@ -111,6 +114,7 @@ int32_t mndProcessBatchMetaMsg(SRpcMsg *pMsg) {
     } else {
       rsp.rspCode = 0;
     }
+    rsp.msgIdx = req.msgIdx;
     rsp.reqType = reqMsg.msgType;
     rsp.msgLen = reqMsg.info.rspLen;
     rsp.msg = reqMsg.info.rsp;
@@ -136,6 +140,8 @@ int32_t mndProcessBatchMetaMsg(SRpcMsg *pMsg) {
     
     *(int32_t*)((char*)pRsp + offset) = htonl(p->reqType);
     offset += sizeof(p->reqType);
+    *(int32_t*)((char*)pRsp + offset) = htonl(p->msgIdx);
+    offset += sizeof(p->msgIdx);
     *(int32_t*)((char*)pRsp + offset) = htonl(p->msgLen);
     offset += sizeof(p->msgLen);
     *(int32_t*)((char*)pRsp + offset) = htonl(p->rspCode);

source/dnode/mnode/impl/src/mndSma.c

@@ -489,7 +489,7 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   smaObj.uid = mndGenerateUid(pCreate->name, TSDB_TABLE_FNAME_LEN);
   ASSERT(smaObj.uid != 0);
   char resultTbName[TSDB_TABLE_FNAME_LEN + 16] = {0};
-  snprintf(resultTbName, TSDB_TABLE_FNAME_LEN + 16, "td.tsma.rst.tb.%s", pCreate->name);
+  snprintf(resultTbName, TSDB_TABLE_FNAME_LEN + 16, "%s_td_tsma_rst_tb",pCreate->name);
   memcpy(smaObj.dstTbName, resultTbName, TSDB_TABLE_FNAME_LEN);
   smaObj.dstTbUid = mndGenerateUid(smaObj.dstTbName, TSDB_TABLE_FNAME_LEN);
   smaObj.stbUid = pStb->uid;
@@ -603,6 +603,9 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   if (mndPersistStream(pMnode, pTrans, &streamObj) != 0) goto _OVER;
   if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
 
+  mDebug("mndSma: create sma index %s %" PRIi64 " on stb:%" PRIi64 ", dstSuid:%" PRIi64 " dstTb:%s dstVg:%d",
+         pCreate->name, smaObj.uid, smaObj.stbUid, smaObj.dstTbUid, smaObj.dstTbName, smaObj.dstVgId);
+
   code = 0;
 
 _OVER:

source/dnode/vnode/src/inc/sma.h

@@ -115,24 +115,29 @@ struct SSmaStat {
 #define RSMA_FS_LOCK(r)       (&(r)->lock)
 
 struct SRSmaInfoItem {
-  void   *taskInfo;  // qTaskInfo_t
-  int64_t refId;
-  tmr_h   tmrId;
-  int32_t maxDelay;
   int8_t  level;
   int8_t  triggerStat;
+  int32_t maxDelay;
+  tmr_h   tmrId;
 };
 
 struct SRSmaInfo {
   STSchema *pTSchema;
   int64_t   suid;
+  int64_t   refId;  // refId of SRSmaStat
   int8_t    delFlag;
   T_REF_DECLARE()
   SRSmaInfoItem items[TSDB_RETENTION_L2];
+  void         *taskInfo[TSDB_RETENTION_L2];   // qTaskInfo_t
+  void         *iTaskInfo[TSDB_RETENTION_L2];  // immutable
 };
-#define RSMA_INFO_HEAD_LEN   24
+
+#define RSMA_INFO_HEAD_LEN     32
 #define RSMA_INFO_IS_DEL(r)    ((r)->delFlag == 1)
 #define RSMA_INFO_SET_DEL(r)   ((r)->delFlag = 1)
+#define RSMA_INFO_QTASK(r, i)  ((r)->taskInfo[i])
+#define RSMA_INFO_IQTASK(r, i) ((r)->iTaskInfo[i])
+#define RSMA_INFO_ITEM(r, i)   (&(r)->items[i])
 
 enum {
   TASK_TRIGGER_STAT_INIT = 0,
@@ -168,8 +173,8 @@ int32_t tdUnRefSmaStat(SSma *pSma, SSmaStat *pStat);
 int32_t tdRefRSmaInfo(SSma *pSma, SRSmaInfo *pRSmaInfo);
 int32_t tdUnRefRSmaInfo(SSma *pSma, SRSmaInfo *pRSmaInfo);
 
-void   *tdAcquireSmaRef(int32_t rsetId, int64_t refId, const char *tags, int32_t ln);
-int32_t tdReleaseSmaRef(int32_t rsetId, int64_t refId, const char *tags, int32_t ln);
+void   *tdAcquireSmaRef(int32_t rsetId, int64_t refId);
+int32_t tdReleaseSmaRef(int32_t rsetId, int64_t refId);
 
 int32_t tdCheckAndInitSmaEnv(SSma *pSma, int8_t smaType);
 
@@ -223,12 +228,11 @@ static FORCE_INLINE void tdSmaStatSetDropped(STSmaStat *pTStat) {
 
 void           tdRSmaQTaskInfoGetFileName(int32_t vid, int64_t version, char *outputName);
 void           tdRSmaQTaskInfoGetFullName(int32_t vid, int64_t version, const char *path, char *outputName);
-int32_t        tdCloneRSmaInfo(SSma *pSma, SRSmaInfo *pDest, SRSmaInfo *pSrc);
+int32_t        tdCloneRSmaInfo(SSma *pSma, SRSmaInfo **pDest, SRSmaInfo *pSrc);
 void           tdFreeQTaskInfo(qTaskInfo_t *taskHandle, int32_t vgId, int32_t level);
 static int32_t tdDestroySmaState(SSmaStat *pSmaStat, int8_t smaType);
 void          *tdFreeSmaState(SSmaStat *pSmaStat, int8_t smaType);
 void          *tdFreeRSmaInfo(SSma *pSma, SRSmaInfo *pInfo, bool isDeepFree);
-void           tdRemoveRSmaInfoBySuid(SSma *pSma, int64_t suid);
 int32_t        tdRSmaPersistExecImpl(SRSmaStat *pRSmaStat, SHashObj *pInfoHash);
 
 int32_t tdProcessRSmaCreateImpl(SSma *pSma, SRSmaParam *param, int64_t suid, const char *tbName);

source/dnode/vnode/src/meta/metaTable.c

@@ -381,6 +381,8 @@ int metaCreateTable(SMeta *pMeta, int64_t version, SVCreateTbReq *pReq) {
     terrno = TSDB_CODE_TDB_TABLE_ALREADY_EXIST;
     metaReaderClear(&mr);
     return -1;
+  } else if (terrno == TSDB_CODE_PAR_TABLE_NOT_EXIST) {
+    terrno = TSDB_CODE_SUCCESS;
   }
   metaReaderClear(&mr);
 

source/dnode/vnode/src/sma/smaCommit.c

@@ -308,12 +308,12 @@ static int32_t tdProcessRSmaSyncPostCommitImpl(SSma *pSma) {
  * @return int32_t
  */
 static int32_t tdProcessRSmaAsyncPreCommitImpl(SSma *pSma) {
-  SSmaEnv *pSmaEnv = SMA_RSMA_ENV(pSma);
-  if (!pSmaEnv) {
+  SSmaEnv *pEnv = SMA_RSMA_ENV(pSma);
+  if (!pEnv) {
     return TSDB_CODE_SUCCESS;
   }
 
-  SSmaStat  *pStat = SMA_ENV_STAT(pSmaEnv);
+  SSmaStat  *pStat = SMA_ENV_STAT(pEnv);
   SRSmaStat *pRSmaStat = SMA_RSMA_STAT(pStat);
 
   // step 1: set rsma stat
@@ -337,18 +337,26 @@ static int32_t tdProcessRSmaAsyncPreCommitImpl(SSma *pSma) {
   }
 
   // step 3:  swap rsmaInfoHash and iRsmaInfoHash
-  ASSERT(!RSMA_IMU_INFO_HASH(pRSmaStat));
+  // lock
+  taosWLockLatch(SMA_ENV_LOCK(pEnv));
+
   ASSERT(RSMA_INFO_HASH(pRSmaStat));
+  ASSERT(!RSMA_IMU_INFO_HASH(pRSmaStat));
 
   RSMA_IMU_INFO_HASH(pRSmaStat) = RSMA_INFO_HASH(pRSmaStat);
   RSMA_INFO_HASH(pRSmaStat) =
       taosHashInit(RSMA_TASK_INFO_HASH_SLOT, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), true, HASH_ENTRY_LOCK);
 
   if (!RSMA_INFO_HASH(pRSmaStat)) {
+    // unlock
+    taosWUnLockLatch(SMA_ENV_LOCK(pEnv));
     smaError("vgId:%d, rsma async commit failed since %s", SMA_VID(pSma), terrstr());
     return TSDB_CODE_FAILED;
   }
 
+  // unlock
+  taosWUnLockLatch(SMA_ENV_LOCK(pEnv));
+
   // step 4: others
   pRSmaStat->commitAppliedVer = pSma->pVnode->state.applied;
 
@@ -383,26 +391,52 @@ static int32_t tdProcessRSmaAsyncCommitImpl(SSma *pSma) {
  * @return int32_t
  */
 static int32_t tdProcessRSmaAsyncPostCommitImpl(SSma *pSma) {
-  SSmaEnv *pSmaEnv = SMA_RSMA_ENV(pSma);
-  if (!pSmaEnv) {
+  SSmaEnv *pEnv = SMA_RSMA_ENV(pSma);
+  if (!pEnv) {
     return TSDB_CODE_SUCCESS;
   }
 
-  SSmaStat  *pStat = SMA_ENV_STAT(pSmaEnv);
+  SSmaStat  *pStat = SMA_ENV_STAT(pEnv);
   SRSmaStat *pRSmaStat = SMA_RSMA_STAT(pStat);
 
   // step 1: merge rsmaInfoHash and iRsmaInfoHash
-  taosWLockLatch(SMA_ENV_LOCK(pSmaEnv));
-
+  // lock
+  taosWLockLatch(SMA_ENV_LOCK(pEnv));
+#if 0
   if (taosHashGetSize(RSMA_INFO_HASH(pRSmaStat)) <= 0) {
-    // TODO: optimization - just switch the hash pointer if rsmaInfoHash is empty
+    // just switch the hash pointer if rsmaInfoHash is empty
+    if (taosHashGetSize(RSMA_IMU_INFO_HASH(pRSmaStat)) > 0) {
+      SHashObj *infoHash = RSMA_INFO_HASH(pRSmaStat);
+      RSMA_INFO_HASH(pRSmaStat) = RSMA_IMU_INFO_HASH(pRSmaStat);
+      RSMA_IMU_INFO_HASH(pRSmaStat) = infoHash;
     }
-
+  } else {
+#endif
+#if 1
   void *pIter = taosHashIterate(RSMA_IMU_INFO_HASH(pRSmaStat), NULL);
   while (pIter) {
     tb_uid_t *pSuid = (tb_uid_t *)taosHashGetKey(pIter, NULL);
 
     if (!taosHashGet(RSMA_INFO_HASH(pRSmaStat), pSuid, sizeof(tb_uid_t))) {
+      SRSmaInfo *pRSmaInfo = *(SRSmaInfo **)pIter;
+      if (RSMA_INFO_IS_DEL(pRSmaInfo)) {
+        int32_t refVal = T_REF_VAL_GET(pRSmaInfo);
+        if (refVal == 0) {
+          tdFreeRSmaInfo(pSma, pRSmaInfo, true);
+          smaDebug(
+              "vgId:%d, rsma async post commit, free rsma info since already deleted and ref is 0 for "
+              "table:%" PRIi64,
+              SMA_VID(pSma), *pSuid);
+        } else {
+          smaDebug(
+              "vgId:%d, rsma async post commit, not free rsma info since ref is %d although already deleted for "
+              "table:%" PRIi64,
+              SMA_VID(pSma), refVal, *pSuid);
+        }
+
+        pIter = taosHashIterate(RSMA_IMU_INFO_HASH(pRSmaStat), pIter);
+        continue;
+      }
       taosHashPut(RSMA_INFO_HASH(pRSmaStat), pSuid, sizeof(tb_uid_t), pIter, sizeof(pIter));
       smaDebug("vgId:%d, rsma async post commit, migrated from iRsmaInfoHash for table:%" PRIi64, SMA_VID(pSma),
                *pSuid);
@@ -416,11 +450,14 @@ static int32_t tdProcessRSmaAsyncPostCommitImpl(SSma *pSma) {
 
     pIter = taosHashIterate(RSMA_IMU_INFO_HASH(pRSmaStat), pIter);
   }
+#endif
+  // }
 
   taosHashCleanup(RSMA_IMU_INFO_HASH(pRSmaStat));
   RSMA_IMU_INFO_HASH(pRSmaStat) = NULL;
 
-  taosWUnLockLatch(SMA_ENV_LOCK(pSmaEnv));
+  // unlock
+  taosWUnLockLatch(SMA_ENV_LOCK(pEnv));
 
   // step 2: cleanup outdated qtaskinfo files
   tdCleanupQTaskInfoFiles(pSma, pRSmaStat);

source/dnode/vnode/src/sma/smaEnv.c

@@ -183,9 +183,6 @@ int32_t tdUnRefRSmaInfo(SSma *pSma, SRSmaInfo *pRSmaInfo) {
   int ref = T_REF_DEC(pRSmaInfo);
   smaDebug("vgId:%d, unref rsma info:%p, val:%d", SMA_VID(pSma), pRSmaInfo, ref);
 
-  if (ref == 0) {
-    tdRemoveRSmaInfoBySuid(pSma, pRSmaInfo->suid);
-  }
   return 0;
 }
 

source/dnode/vnode/src/sma/smaRollup.c

@@ -32,11 +32,13 @@ static int32_t    tdUidStorePut(STbUidStore *pStore, tb_uid_t suid, tb_uid_t *ui
 static int32_t    tdUpdateTbUidListImpl(SSma *pSma, tb_uid_t *suid, SArray *tbUids);
 static int32_t    tdSetRSmaInfoItemParams(SSma *pSma, SRSmaParam *param, SRSmaStat *pStat, SRSmaInfo *pRSmaInfo,
                                           int8_t idx);
-static int32_t    tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, SRSmaInfoItem *rsmaItem,
-                                    STSchema *pTSchema, tb_uid_t suid, int8_t level);
-static SRSmaInfo *tdGetRSmaInfoBySuid(SSma *pSma, int64_t suid);
-static int32_t    tdRSmaFetchAndSubmitResult(SRSmaInfoItem *pItem, STSchema *pTSchema, int64_t suid, SRSmaStat *pStat,
-                                             int8_t blkType);
+static int32_t    tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, SRSmaInfo *pInfo, tb_uid_t suid,
+                                    int8_t level);
+static SRSmaInfo *tdAcquireRSmaInfoBySuid(SSma *pSma, int64_t suid);
+static void       tdReleaseRSmaInfo(SSma *pSma, SRSmaInfo *pInfo);
+
+static int32_t tdRSmaFetchAndSubmitResult(qTaskInfo_t taskInfo, SRSmaInfoItem *pItem, STSchema *pTSchema, int64_t suid,
+                                          SRSmaStat *pStat, int8_t blkType);
 static void    tdRSmaFetchTrigger(void *param, void *tmrId);
 
 static int32_t tdRSmaQTaskInfoIterInit(SRSmaQTaskInfoIter *pIter, STFile *pTFile);
@@ -115,17 +117,26 @@ void *tdFreeRSmaInfo(SSma *pSma, SRSmaInfo *pInfo, bool isDeepFree) {
   if (pInfo) {
     for (int32_t i = 0; i < TSDB_RETENTION_L2; ++i) {
       SRSmaInfoItem *pItem = &pInfo->items[i];
-      if (pItem->taskInfo) {
+
       if (isDeepFree && pItem->tmrId) {
         smaDebug("vgId:%d, stop fetch timer %p for table %" PRIi64 " level %d", SMA_VID(pSma), pInfo->suid,
                  pItem->tmrId, i + 1);
         taosTmrStopA(&pItem->tmrId);
       }
-        tdFreeQTaskInfo(&pItem->taskInfo, SMA_VID(pSma), i + 1);
+
+      if (isDeepFree && pInfo->taskInfo[i]) {
+        tdFreeQTaskInfo(&pInfo->taskInfo[i], SMA_VID(pSma), i + 1);
       } else {
         smaDebug("vgId:%d, table %" PRIi64 " no need to destroy rsma info level %d since empty taskInfo", SMA_VID(pSma),
                  pInfo->suid, i + 1);
       }
+
+      if (pInfo->iTaskInfo[i]) {
+        tdFreeQTaskInfo(&pInfo->iTaskInfo[i], SMA_VID(pSma), i + 1);
+      } else {
+        smaDebug("vgId:%d, table %" PRIi64 " no need to destroy rsma info level %d since empty iTaskInfo",
+                 SMA_VID(pSma), pInfo->suid, i + 1);
+      }
     }
     if (isDeepFree) {
       taosMemoryFreeClear(pInfo->pTSchema);
@@ -155,7 +166,12 @@ static int32_t tdUpdateTbUidListImpl(SSma *pSma, tb_uid_t *suid, SArray *tbUids)
     return TSDB_CODE_FAILED;
   }
 
-  pRSmaInfo = tdGetRSmaInfoBySuid(pSma, *suid);
+  if (!taosArrayGetSize(tbUids)) {
+    smaDebug("vgId:%d, no need to update tbUidList for suid:%" PRIi64 " since Empty tbUids", SMA_VID(pSma), *suid);
+    return TSDB_CODE_SUCCESS;
+  }
+
+  pRSmaInfo = tdAcquireRSmaInfoBySuid(pSma, *suid);
 
   if (!pRSmaInfo) {
     smaError("vgId:%d, failed to get rsma info for uid:%" PRIi64, SMA_VID(pSma), *suid);
@@ -163,26 +179,21 @@ static int32_t tdUpdateTbUidListImpl(SSma *pSma, tb_uid_t *suid, SArray *tbUids)
     return TSDB_CODE_FAILED;
   }
 
-  if (pRSmaInfo->items[0].taskInfo) {
-    if ((qUpdateQualifiedTableId(pRSmaInfo->items[0].taskInfo, tbUids, true) < 0)) {
-      smaError("vgId:%d, update tbUidList failed for uid:%" PRIi64 " since %s", SMA_VID(pSma), *suid, terrstr());
+  for (int32_t i = 0; i < TSDB_RETENTION_L2; ++i) {
+    if (pRSmaInfo->taskInfo[i]) {
+      if ((qUpdateQualifiedTableId(pRSmaInfo->taskInfo[i], tbUids, true) < 0)) {
+        tdReleaseRSmaInfo(pSma, pRSmaInfo);
+        smaError("vgId:%d, update tbUidList failed for uid:%" PRIi64 " level %d since %s", SMA_VID(pSma), *suid, i,
+                 terrstr());
         return TSDB_CODE_FAILED;
       } else {
-      smaDebug("vgId:%d, update tbUidList succeed for qTaskInfo:%p with suid:%" PRIi64 ", uid:%" PRIi64, SMA_VID(pSma),
-               pRSmaInfo->items[0].taskInfo, *suid, *(int64_t *)taosArrayGet(tbUids, 0));
-    }
+        smaDebug("vgId:%d, update tbUidList succeed for qTaskInfo:%p with suid:%" PRIi64 " uid:%" PRIi64 " level %d",
+                 SMA_VID(pSma), pRSmaInfo->taskInfo[0], *suid, *(int64_t *)taosArrayGet(tbUids, 0), i);
       }
-
-  if (pRSmaInfo->items[1].taskInfo) {
-    if ((qUpdateQualifiedTableId(pRSmaInfo->items[1].taskInfo, tbUids, true) < 0)) {
-      smaError("vgId:%d, update tbUidList failed for uid:%" PRIi64 " since %s", SMA_VID(pSma), *suid, terrstr());
-      return TSDB_CODE_FAILED;
-    } else {
-      smaDebug("vgId:%d, update tbUidList succeed for qTaskInfo:%p with suid:%" PRIi64 ", uid:%" PRIi64, SMA_VID(pSma),
-               pRSmaInfo->items[1].taskInfo, *suid, *(int64_t *)taosArrayGet(tbUids, 0));
     }
   }
 
+  tdReleaseRSmaInfo(pSma, pRSmaInfo);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -267,13 +278,12 @@ static int32_t tdSetRSmaInfoItemParams(SSma *pSma, SRSmaParam *param, SRSmaStat
         .initTqReader = 1,
     };
 
-    SRSmaInfoItem *pItem = &(pRSmaInfo->items[idx]);
-    pItem->refId = RSMA_REF_ID(pStat);
-    pItem->taskInfo = qCreateStreamExecTaskInfo(param->qmsg[idx], &handle);
-    if (!pItem->taskInfo) {
+    pRSmaInfo->taskInfo[idx] = qCreateStreamExecTaskInfo(param->qmsg[idx], &handle);
+    if (!pRSmaInfo->taskInfo[idx]) {
       terrno = TSDB_CODE_RSMA_QTASKINFO_CREATE;
       return TSDB_CODE_FAILED;
     }
+    SRSmaInfoItem *pItem = &(pRSmaInfo->items[idx]);
     pItem->triggerStat = TASK_TRIGGER_STAT_INACTIVE;
     if (param->maxdelay[idx] < TSDB_MIN_ROLLUP_MAX_DELAY) {
       int64_t msInterval =
@@ -342,6 +352,7 @@ int32_t tdProcessRSmaCreateImpl(SSma *pSma, SRSmaParam *param, int64_t suid, con
   }
   pRSmaInfo->pTSchema = pTSchema;
   pRSmaInfo->suid = suid;
+  pRSmaInfo->refId = RSMA_REF_ID(pStat);
   T_REF_INIT_VAL(pRSmaInfo, 1);
 
   if (tdSetRSmaInfoItemParams(pSma, param, pStat, pRSmaInfo, 0) < 0) {
@@ -411,7 +422,7 @@ int32_t tdProcessRSmaDrop(SSma *pSma, SVDropStbReq *pReq) {
   SSmaStat  *pStat = SMA_ENV_STAT(pSmaEnv);
   SRSmaStat *pRSmaStat = SMA_RSMA_STAT(pStat);
 
-  SRSmaInfo *pRSmaInfo = tdGetRSmaInfoBySuid(pSma, pReq->suid);
+  SRSmaInfo *pRSmaInfo = tdAcquireRSmaInfoBySuid(pSma, pReq->suid);
 
   if (!pRSmaInfo) {
     smaWarn("vgId:%d, drop rsma for stable %s %" PRIi64 " failed no rsma in hash", TD_VID(pVnode), pReq->name,
@@ -423,8 +434,10 @@ int32_t tdProcessRSmaDrop(SSma *pSma, SVDropStbReq *pReq) {
   RSMA_INFO_SET_DEL(pRSmaInfo);
   tdUnRefRSmaInfo(pSma, pRSmaInfo);
 
-  // save to file
+  tdReleaseRSmaInfo(pSma, pRSmaInfo);
 
+  // save to file
+  // TODO
   smaDebug("vgId:%d, drop rsma for table %" PRIi64 " succeed", TD_VID(pVnode), pReq->suid);
   return TSDB_CODE_SUCCESS;
 }
@@ -567,8 +580,32 @@ static void tdDestroySDataBlockArray(SArray *pArray) {
   taosArrayDestroy(pArray);
 }
 
-static int32_t tdRSmaFetchAndSubmitResult(SRSmaInfoItem *pItem, STSchema *pTSchema, int64_t suid, SRSmaStat *pStat,
-                                          int8_t blkType) {
+/**
+ * @brief retention of rsma1/rsma2
+ *
+ * @param pSma
+ * @param now
+ * @return int32_t
+ */
+int32_t smaDoRetention(SSma *pSma, int64_t now) {
+  int32_t code = TSDB_CODE_SUCCESS;
+  if (VND_IS_RSMA(pSma->pVnode)) {
+    return code;
+  }
+
+  for (int32_t i = 0; i < TSDB_RETENTION_L2; ++i) {
+    if (pSma->pRSmaTsdb[i]) {
+      code = tsdbDoRetention(pSma->pRSmaTsdb[i], now);
+      if (code) goto _end;
+    }
+  }
+
+_end:
+  return code;
+}
+
+static int32_t tdRSmaFetchAndSubmitResult(qTaskInfo_t taskInfo, SRSmaInfoItem *pItem, STSchema *pTSchema, int64_t suid,
+                                          SRSmaStat *pStat, int8_t blkType) {
   SArray *pResult = NULL;
   SSma   *pSma = pStat->pSma;
 
@@ -576,7 +613,7 @@ static int32_t tdRSmaFetchAndSubmitResult(SRSmaInfoItem *pItem, STSchema *pTSche
     SSDataBlock *output = NULL;
     uint64_t     ts;
 
-    int32_t code = qExecTask(pItem->taskInfo, &output, &ts);
+    int32_t code = qExecTask(taskInfo, &output, &ts);
     if (code < 0) {
       smaError("vgId:%d, qExecTask for rsma table %" PRIi64 " level %" PRIi8 " failed since %s", SMA_VID(pSma), suid,
                pItem->level, terrstr(code));
@@ -637,29 +674,32 @@ _err:
   return TSDB_CODE_FAILED;
 }
 
-static int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, SRSmaInfoItem *pItem,
-                                 STSchema *pTSchema, tb_uid_t suid, int8_t level) {
-  if (!pItem || !pItem->taskInfo) {
+static int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, SRSmaInfo *pInfo, tb_uid_t suid,
+                                 int8_t level) {
+  int32_t idx = level - 1;
+  if (!pInfo || !RSMA_INFO_QTASK(pInfo, idx)) {
     smaDebug("vgId:%d, no qTaskInfo to execute rsma %" PRIi8 " task for suid:%" PRIu64, SMA_VID(pSma), level, suid);
     return TSDB_CODE_SUCCESS;
   }
-  if (!pTSchema) {
+  if (!pInfo->pTSchema) {
     smaWarn("vgId:%d, no schema to execute rsma %" PRIi8 " task for suid:%" PRIu64, SMA_VID(pSma), level, suid);
     return TSDB_CODE_FAILED;
   }
 
   smaDebug("vgId:%d, execute rsma %" PRIi8 " task for qTaskInfo:%p suid:%" PRIu64, SMA_VID(pSma), level,
-           pItem->taskInfo, suid);
+           RSMA_INFO_QTASK(pInfo, idx), suid);
 
-  if (qSetMultiStreamInput(pItem->taskInfo, pMsg, 1, inputType) < 0) {  // INPUT__DATA_SUBMIT
+  if (qSetMultiStreamInput(RSMA_INFO_QTASK(pInfo, idx), pMsg, 1, inputType) < 0) {  // INPUT__DATA_SUBMIT
     smaError("vgId:%d, rsma %" PRIi8 " qSetStreamInput failed since %s", SMA_VID(pSma), level, tstrerror(terrno));
     return TSDB_CODE_FAILED;
   }
 
   SSmaEnv       *pEnv = SMA_RSMA_ENV(pSma);
   SRSmaStat     *pStat = SMA_RSMA_STAT(pEnv->pStat);
+  SRSmaInfoItem *pItem = RSMA_INFO_ITEM(pInfo, idx);
 
-  tdRSmaFetchAndSubmitResult(pItem, pTSchema, suid, pStat, STREAM_INPUT__DATA_SUBMIT);
+  tdRSmaFetchAndSubmitResult(RSMA_INFO_QTASK(pInfo, idx), pItem, pInfo->pTSchema, suid, pStat,
+                             STREAM_INPUT__DATA_SUBMIT);
   atomic_store_8(&pItem->triggerStat, TASK_TRIGGER_STAT_ACTIVE);
 
   if (smaMgmt.tmrHandle) {
@@ -678,7 +718,7 @@ static int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType
  * @param suid
  * @return SRSmaInfo*
  */
-static SRSmaInfo *tdGetRSmaInfoBySuid(SSma *pSma, int64_t suid) {
+static SRSmaInfo *tdAcquireRSmaInfoBySuid(SSma *pSma, int64_t suid) {
   SSmaEnv   *pEnv = SMA_RSMA_ENV(pSma);
   SRSmaStat *pStat = NULL;
   SRSmaInfo *pRSmaInfo = NULL;
@@ -692,94 +732,86 @@ static SRSmaInfo *tdGetRSmaInfoBySuid(SSma *pSma, int64_t suid) {
     return NULL;
   }
 
+  taosRLockLatch(SMA_ENV_LOCK(pEnv));
   pRSmaInfo = taosHashGet(RSMA_INFO_HASH(pStat), &suid, sizeof(tb_uid_t));
   if (pRSmaInfo && (pRSmaInfo = *(SRSmaInfo **)pRSmaInfo)) {
+    if (RSMA_INFO_IS_DEL(pRSmaInfo)) {
+      taosRUnLockLatch(SMA_ENV_LOCK(pEnv));
+      return NULL;
+    }
+    tdRefRSmaInfo(pSma, pRSmaInfo);
+    taosRUnLockLatch(SMA_ENV_LOCK(pEnv));
     return pRSmaInfo;
   }
 
-  if (RSMA_COMMIT_STAT(pStat) == 0) {
+  if (RSMA_COMMIT_STAT(pStat) == 0) {  // return NULL if not in committing stat
+    taosRUnLockLatch(SMA_ENV_LOCK(pEnv));
     return NULL;
   }
+  taosRUnLockLatch(SMA_ENV_LOCK(pEnv));
 
   // clone the SRSmaInfo from iRsmaInfoHash to rsmaInfoHash if in committing stat
   SRSmaInfo *pCowRSmaInfo = NULL;
   // lock
   taosWLockLatch(SMA_ENV_LOCK(pEnv));
-  if (!taosHashGet(RSMA_INFO_HASH(pStat), &suid, sizeof(tb_uid_t))) {  // 2-phase lock
+  if (!(pCowRSmaInfo = taosHashGet(RSMA_INFO_HASH(pStat), &suid, sizeof(tb_uid_t)))) {  // 2-phase lock
     void *iRSmaInfo = taosHashGet(RSMA_IMU_INFO_HASH(pStat), &suid, sizeof(tb_uid_t));
     if (iRSmaInfo) {
       SRSmaInfo *pIRSmaInfo = *(SRSmaInfo **)iRSmaInfo;
-      if (pIRSmaInfo) {
-        if (tdCloneRSmaInfo(pSma, pCowRSmaInfo, pIRSmaInfo) < 0) {
+      if (pIRSmaInfo && !RSMA_INFO_IS_DEL(pIRSmaInfo)) {
+        if (tdCloneRSmaInfo(pSma, &pCowRSmaInfo, pIRSmaInfo) < 0) {
+          // unlock
           taosWUnLockLatch(SMA_ENV_LOCK(pEnv));
           smaError("vgId:%d, clone rsma info failed for suid:%" PRIu64 " since %s", SMA_VID(pSma), suid, terrstr());
           return NULL;
         }
         smaDebug("vgId:%d, clone rsma info succeed for suid:%" PRIu64, SMA_VID(pSma), suid);
         if (taosHashPut(RSMA_INFO_HASH(pStat), &suid, sizeof(tb_uid_t), &pCowRSmaInfo, sizeof(pCowRSmaInfo)) < 0) {
+          // unlock
           taosWUnLockLatch(SMA_ENV_LOCK(pEnv));
+          smaError("vgId:%d, clone rsma info failed for suid:%" PRIu64 " since %s", SMA_VID(pSma), suid, terrstr());
           return NULL;
         }
       }
     }
+  } else {
+    pCowRSmaInfo = *(SRSmaInfo **)pCowRSmaInfo;
+    ASSERT(!pCowRSmaInfo);
+  }
+
+  if(pCowRSmaInfo) {
+    tdRefRSmaInfo(pSma, pCowRSmaInfo);
   }
   // unlock
   taosWUnLockLatch(SMA_ENV_LOCK(pEnv));
   return pCowRSmaInfo;
 }
 
-/**
- * @brief During the drop procedure, only need to delete the object in rsmaInfoHash.
- *
- * @param pSma
- * @param suid
- * @return SRSmaInfo*
- */
-void tdRemoveRSmaInfoBySuid(SSma *pSma, int64_t suid) {
-  SSmaEnv   *pEnv = SMA_RSMA_ENV(pSma);
-  SRSmaStat *pStat = NULL;
-  SRSmaInfo *pRSmaInfo = NULL;
-
-  if (!pEnv) {
-    return;
-  }
-
-  pStat = (SRSmaStat *)SMA_ENV_STAT(pEnv);
-  if (!pStat || !RSMA_INFO_HASH(pStat)) {
-    return;
-  }
-
-  pRSmaInfo = taosHashGet(RSMA_INFO_HASH(pStat), &suid, sizeof(tb_uid_t));
-  if (pRSmaInfo) {
-    if ((pRSmaInfo = *(SRSmaInfo **)pRSmaInfo)) {
-      tdFreeRSmaInfo(pSma, pRSmaInfo, true);
-    }
-    taosHashRemove(RSMA_INFO_HASH(pStat), &suid, sizeof(tb_uid_t));
-    smaDebug("vgId:%d, remove from infoHash for table:%" PRIu64 " succeed", SMA_VID(pSma), suid);
+static FORCE_INLINE void tdReleaseRSmaInfo(SSma *pSma, SRSmaInfo *pInfo) {
+  if (pInfo) {
+    tdUnRefRSmaInfo(pSma, pInfo);
   }
 }
 
 static int32_t tdExecuteRSma(SSma *pSma, const void *pMsg, int32_t inputType, tb_uid_t suid) {
-  SRSmaInfo *pRSmaInfo = tdGetRSmaInfoBySuid(pSma, suid);
+  SRSmaInfo *pRSmaInfo = tdAcquireRSmaInfoBySuid(pSma, suid);
   if (!pRSmaInfo) {
     smaDebug("vgId:%d, execute rsma, no rsma info for suid:%" PRIu64, SMA_VID(pSma), suid);
     return TSDB_CODE_SUCCESS;
   }
 
-  if (!pRSmaInfo->items[0].taskInfo) {
+  if (!RSMA_INFO_QTASK(pRSmaInfo, 0)) {
+    tdReleaseRSmaInfo(pSma, pRSmaInfo);
     smaDebug("vgId:%d, execute rsma, no rsma qTaskInfo for suid:%" PRIu64, SMA_VID(pSma), suid);
     return TSDB_CODE_SUCCESS;
   }
 
   if (inputType == STREAM_INPUT__DATA_SUBMIT) {
-    tdRefRSmaInfo(pSma, pRSmaInfo);
-
-    tdExecuteRSmaImpl(pSma, pMsg, inputType, &pRSmaInfo->items[0], pRSmaInfo->pTSchema, suid, TSDB_RETENTION_L1);
-    tdExecuteRSmaImpl(pSma, pMsg, inputType, &pRSmaInfo->items[1], pRSmaInfo->pTSchema, suid, TSDB_RETENTION_L2);
-
-    tdUnRefRSmaInfo(pSma, pRSmaInfo);
+    tdExecuteRSmaImpl(pSma, pMsg, inputType, pRSmaInfo, suid, TSDB_RETENTION_L1);
+    tdExecuteRSmaImpl(pSma, pMsg, inputType, pRSmaInfo, suid, TSDB_RETENTION_L2);
   }
 
+  tdReleaseRSmaInfo(pSma, pRSmaInfo);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -990,26 +1022,28 @@ static int32_t tdRSmaQTaskInfoItemRestore(SSma *pSma, const SRSmaQTaskInfoItem *
   SRSmaInfo *pRSmaInfo = NULL;
   void      *qTaskInfo = NULL;
 
-  pRSmaInfo = tdGetRSmaInfoBySuid(pSma, pItem->suid);
+  pRSmaInfo = tdAcquireRSmaInfoBySuid(pSma, pItem->suid);
   if (!pRSmaInfo) {
     smaDebug("vgId:%d, no restore as no rsma info for table:%" PRIu64, SMA_VID(pSma), pItem->suid);
     return TSDB_CODE_SUCCESS;
   }
 
   if (pItem->type == TSDB_RETENTION_L1) {
-    qTaskInfo = pRSmaInfo->items[0].taskInfo;
+    qTaskInfo = RSMA_INFO_QTASK(pRSmaInfo, 0);
   } else if (pItem->type == TSDB_RETENTION_L2) {
-    qTaskInfo = pRSmaInfo->items[1].taskInfo;
+    qTaskInfo = RSMA_INFO_QTASK(pRSmaInfo, 1);
   } else {
     ASSERT(0);
   }
 
   if (!qTaskInfo) {
+    tdReleaseRSmaInfo(pSma, pRSmaInfo);
     smaDebug("vgId:%d, no restore as NULL rsma qTaskInfo for table:%" PRIu64, SMA_VID(pSma), pItem->suid);
     return TSDB_CODE_SUCCESS;
   }
 
   if (qDeserializeTaskStatus(qTaskInfo, pItem->qTaskInfo, pItem->len) < 0) {
+    tdReleaseRSmaInfo(pSma, pRSmaInfo);
     smaError("vgId:%d, restore rsma task failed for table:%" PRIi64 " level %d since %s", SMA_VID(pSma), pItem->suid,
              pItem->type, terrstr());
     return TSDB_CODE_FAILED;
@@ -1017,6 +1051,7 @@ static int32_t tdRSmaQTaskInfoItemRestore(SSma *pSma, const SRSmaQTaskInfoItem *
   smaDebug("vgId:%d, restore rsma task success for table:%" PRIi64 " level %d", SMA_VID(pSma), pItem->suid,
            pItem->type);
 
+  tdReleaseRSmaInfo(pSma, pRSmaInfo);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -1195,8 +1230,14 @@ int32_t tdRSmaPersistExecImpl(SRSmaStat *pRSmaStat, SHashObj *pInfoHash) {
 
   while (infoHash) {
     SRSmaInfo *pRSmaInfo = *(SRSmaInfo **)infoHash;
+
+    if (RSMA_INFO_IS_DEL(pRSmaInfo)) {
+      infoHash = taosHashIterate(pInfoHash, infoHash);
+      continue;
+    }
+
     for (int32_t i = 0; i < TSDB_RETENTION_L2; ++i) {
-      qTaskInfo_t taskInfo = pRSmaInfo->items[i].taskInfo;
+      qTaskInfo_t taskInfo = RSMA_INFO_QTASK(pRSmaInfo, i);
       if (!taskInfo) {
         smaDebug("vgId:%d, rsma, table %" PRIi64 " level %d qTaskInfo is NULL", vid, pRSmaInfo->suid, i + 1);
         continue;
@@ -1290,11 +1331,17 @@ _err:
 static void tdRSmaFetchTrigger(void *param, void *tmrId) {
   SRSmaInfoItem *pItem = param;
   SSma          *pSma = NULL;
-  SRSmaStat     *pStat = (SRSmaStat *)tdAcquireSmaRef(smaMgmt.rsetId, pItem->refId, __func__, __LINE__);
+  SRSmaInfo     *pRSmaInfo = tdGetRSmaInfoByItem(pItem);
+
+  if (RSMA_INFO_IS_DEL(pRSmaInfo)) {
+    return;
+  }
+
+  SRSmaStat *pStat = (SRSmaStat *)tdAcquireSmaRef(smaMgmt.rsetId, pRSmaInfo->refId);
 
   if (!pStat) {
     smaDebug("rsma fetch task not start since already destroyed, rsetId rsetId:%" PRIi64 " refId:%d)", smaMgmt.rsetId,
-             pItem->refId);
+             pRSmaInfo->refId);
     return;
   }
 
@@ -1305,10 +1352,10 @@ static void tdRSmaFetchTrigger(void *param, void *tmrId) {
   switch (rsmaTriggerStat) {
     case TASK_TRIGGER_STAT_PAUSED:
     case TASK_TRIGGER_STAT_CANCELLED: {
-      tdReleaseSmaRef(smaMgmt.rsetId, pItem->refId, __func__, __LINE__);
+      tdReleaseSmaRef(smaMgmt.rsetId, pRSmaInfo->refId);
       smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data since stat is %" PRIi8 ", rsetId rsetId:%" PRIi64
                " refId:%d",
-               SMA_VID(pSma), pItem->level, rsmaTriggerStat, smaMgmt.rsetId, pItem->refId);
+               SMA_VID(pSma), pItem->level, rsmaTriggerStat, smaMgmt.rsetId, pRSmaInfo->refId);
       if (rsmaTriggerStat == TASK_TRIGGER_STAT_PAUSED) {
         taosTmrReset(tdRSmaFetchTrigger, pItem->maxDelay > 5000 ? 5000 : pItem->maxDelay, pItem, smaMgmt.tmrHandle,
                      &pItem->tmrId);
@@ -1319,11 +1366,6 @@ static void tdRSmaFetchTrigger(void *param, void *tmrId) {
       break;
   }
 
-  SRSmaInfo *pRSmaInfo = tdGetRSmaInfoByItem(pItem);
-  if (RSMA_INFO_IS_DEL(pRSmaInfo)) {
-    goto _end;
-  }
-
   int8_t fetchTriggerStat =
       atomic_val_compare_exchange_8(&pItem->triggerStat, TASK_TRIGGER_STAT_ACTIVE, TASK_TRIGGER_STAT_INACTIVE);
   switch (fetchTriggerStat) {
@@ -1332,16 +1374,14 @@ static void tdRSmaFetchTrigger(void *param, void *tmrId) {
                pItem->level, pRSmaInfo->suid);
 
       // sync procedure => async process
-      tdRefRSmaInfo(pSma, pRSmaInfo);
 
       SSDataBlock dataBlock = {.info.type = STREAM_GET_ALL};
-      qSetMultiStreamInput(pItem->taskInfo, &dataBlock, 1, STREAM_INPUT__DATA_BLOCK);
-      tdRSmaFetchAndSubmitResult(pItem, pRSmaInfo->pTSchema, pRSmaInfo->suid, pStat, STREAM_INPUT__DATA_BLOCK);
-      tdCleanupStreamInputDataBlock(pItem->taskInfo);
+      qTaskInfo_t taskInfo = pRSmaInfo->taskInfo[pItem->level - 1];
+      qSetMultiStreamInput(taskInfo, &dataBlock, 1, STREAM_INPUT__DATA_BLOCK);
+      tdRSmaFetchAndSubmitResult(taskInfo, pItem, pRSmaInfo->pTSchema, pRSmaInfo->suid, pStat,
+                                 STREAM_INPUT__DATA_BLOCK);
+      tdCleanupStreamInputDataBlock(taskInfo);
 
-      tdUnRefRSmaInfo(pSma, pRSmaInfo);
-      // atomic_store_8(&pItem->triggerStat, TASK_TRIGGER_STAT_ACTIVE);
-      // taosTmrReset(tdRSmaFetchTrigger, 5000, pItem, smaMgmt.tmrHandle, &pItem->tmrId);
     } break;
     case TASK_TRIGGER_STAT_PAUSED: {
       smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is paused",
@@ -1362,22 +1402,5 @@ static void tdRSmaFetchTrigger(void *param, void *tmrId) {
   }
 
 _end:
-  tdReleaseSmaRef(smaMgmt.rsetId, pItem->refId, __func__, __LINE__);
-}
-
-int32_t smaDoRetention(SSma *pSma, int64_t now) {
-  int32_t code = TSDB_CODE_SUCCESS;
-  if (VND_IS_RSMA(pSma->pVnode)) {
-    return code;
-  }
-
-  for (int32_t i = 0; i < TSDB_RETENTION_L2; ++i) {
-    if (pSma->pRSmaTsdb[i]) {
-      code = tsdbDoRetention(pSma->pRSmaTsdb[i], now);
-      if (code) goto _end;
-    }
-  }
-
-_end:
-  return code;
+  tdReleaseSmaRef(smaMgmt.rsetId, pRSmaInfo->refId);
 }

source/dnode/vnode/src/sma/smaTimeRange.c

@@ -116,8 +116,10 @@ int32_t tdProcessTSmaCreateImpl(SSma *pSma, int64_t version, const char *pMsg) {
     }
 
     // create stable to save tsma result in dstVgId
+    SName stbFullName = {0};
+    tNameFromString(&stbFullName, pCfg->dstTbName, T_NAME_ACCT | T_NAME_DB | T_NAME_TABLE);
     SVCreateStbReq pReq = {0};
-    pReq.name = pCfg->dstTbName;
+    pReq.name = (char*)tNameGetTableName(&stbFullName);
     pReq.suid = pCfg->dstTbUid;
     pReq.schemaRow = pCfg->schemaRow;
     pReq.schemaTag = pCfg->schemaTag;
@@ -125,6 +127,12 @@ int32_t tdProcessTSmaCreateImpl(SSma *pSma, int64_t version, const char *pMsg) {
     if (metaCreateSTable(SMA_META(pSma), version, &pReq) < 0) {
       return -1;
     }
+
+    smaDebug("vgId:%d, success to create sma index %s %" PRIi64 " on stb:%" PRIi64 ", dstSuid:%" PRIi64
+             " dstTb:%s dstVg:%d",
+             SMA_VID(pSma), pCfg->indexName, pCfg->indexUid, pCfg->tableUid, pCfg->dstTbUid, pReq.name, pCfg->dstVgId);
+  } else {
+    ASSERT(0);
   }
 
   return 0;

source/dnode/vnode/src/sma/smaUtil.c

@@ -287,22 +287,22 @@ int32_t tdRemoveTFile(STFile *pTFile) {
 }
 
 // smaXXXUtil ================
-void *tdAcquireSmaRef(int32_t rsetId, int64_t refId, const char *tags, int32_t ln) {
+void *tdAcquireSmaRef(int32_t rsetId, int64_t refId) {
   void *pResult = taosAcquireRef(rsetId, refId);
   if (!pResult) {
-    smaWarn("%s:%d taosAcquireRef for rsetId:%" PRIi64 " refId:%d failed since %s", tags, ln, rsetId, refId, terrstr());
+    smaWarn("rsma acquire ref for rsetId:%" PRIi64 " refId:%d failed since %s", rsetId, refId, terrstr());
   } else {
-    smaDebug("%s:%d taosAcquireRef for rsetId:%" PRIi64 " refId:%d success", tags, ln, rsetId, refId);
+    smaDebug("rsma acquire ref for rsetId:%" PRIi64 " refId:%d success", rsetId, refId);
   }
   return pResult;
 }
 
-int32_t tdReleaseSmaRef(int32_t rsetId, int64_t refId, const char *tags, int32_t ln) {
+int32_t tdReleaseSmaRef(int32_t rsetId, int64_t refId) {
   if (taosReleaseRef(rsetId, refId) < 0) {
-    smaWarn("%s:%d taosReleaseRef for rsetId:%" PRIi64 " refId:%d failed since %s", tags, ln, rsetId, refId, terrstr());
+    smaWarn("rsma release ref for rsetId:%" PRIi64 " refId:%d failed since %s", rsetId, refId, terrstr());
     return TSDB_CODE_FAILED;
   }
-  smaDebug("%s:%d taosReleaseRef for rsetId:%" PRIi64 " refId:%d success", tags, ln, rsetId, refId);
+  smaDebug("rsma release ref for rsetId:%" PRIi64 " refId:%d success", rsetId, refId);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -313,7 +313,7 @@ static int32_t tdCloneQTaskInfo(SSma *pSma, qTaskInfo_t dstTaskInfo, qTaskInfo_t
   char   *pOutput = NULL;
   int32_t len = 0;
 
-  if (qSerializeTaskStatus(srcTaskInfo, &pOutput, &len) < 0) {
+  if ((terrno = qSerializeTaskStatus(srcTaskInfo, &pOutput, &len)) < 0) {
     smaError("vgId:%d, rsma clone, table %" PRIi64 " serialize qTaskInfo failed since %s", TD_VID(pVnode), suid,
              terrstr());
     goto _err;
@@ -337,13 +337,15 @@ static int32_t tdCloneQTaskInfo(SSma *pSma, qTaskInfo_t dstTaskInfo, qTaskInfo_t
     goto _err;
   }
 
-  smaError("vgId:%d, rsma clone, restore rsma task for table:%" PRIi64 " succeed", TD_VID(pVnode), suid);
+  smaDebug("vgId:%d, rsma clone, restore rsma task for table:%" PRIi64 " succeed", TD_VID(pVnode), suid);
 
   taosMemoryFreeClear(pOutput);
   return TSDB_CODE_SUCCESS;
 _err:
   taosMemoryFreeClear(pOutput);
   tdFreeQTaskInfo(dstTaskInfo, TD_VID(pVnode), idx + 1);
+  smaError("vgId:%d, rsma clone, restore rsma task for table:%" PRIi64 " failed since %s", TD_VID(pVnode), suid,
+           terrstr());
   return TSDB_CODE_FAILED;
 }
 
@@ -355,23 +357,14 @@ _err:
  * @param pSrc
  * @return int32_t
  */
-int32_t tdCloneRSmaInfo(SSma *pSma, SRSmaInfo *pDest, SRSmaInfo *pSrc) {
+int32_t tdCloneRSmaInfo(SSma *pSma, SRSmaInfo **pDest, SRSmaInfo *pSrc) {
   SVnode     *pVnode = pSma->pVnode;
   SRSmaParam *param = NULL;
   if (!pSrc) {
+    *pDest = NULL;
     return TSDB_CODE_SUCCESS;
   }
 
-  if (!pDest) {
-    pDest = taosMemoryCalloc(1, sizeof(SRSmaInfo));
-    if (!pDest) {
-      terrno = TSDB_CODE_OUT_OF_MEMORY;
-      return TSDB_CODE_FAILED;
-    }
-  }
-
-  memcpy(pDest, pSrc, sizeof(SRSmaInfo));
-
   SMetaReader mr = {0};
   metaReaderInit(&mr, SMA_META(pSma), 0);
   smaDebug("vgId:%d, rsma clone, suid is %" PRIi64, TD_VID(pVnode), pSrc->suid);
@@ -384,21 +377,22 @@ int32_t tdCloneRSmaInfo(SSma *pSma, SRSmaInfo *pDest, SRSmaInfo *pSrc) {
   ASSERT(mr.me.uid == pSrc->suid);
   if (TABLE_IS_ROLLUP(mr.me.flags)) {
     param = &mr.me.stbEntry.rsmaParam;
-    for (int i = 0; i < TSDB_RETENTION_L2; ++i) {
-      SRSmaInfoItem *pItem = &pSrc->items[i];
-      if (pItem->taskInfo) {
-        tdCloneQTaskInfo(pSma, pDest->items[i].taskInfo, pItem->taskInfo, param, pSrc->suid, i);
+    for (int32_t i = 0; i < TSDB_RETENTION_L2; ++i) {
+      if (tdCloneQTaskInfo(pSma, pSrc->iTaskInfo[i], pSrc->taskInfo[i], param, pSrc->suid, i) < 0) {
+        goto _err;
       }
     }
     smaDebug("vgId:%d, rsma clone env success for %" PRIi64, TD_VID(pVnode), pSrc->suid);
   }
 
   metaReaderClear(&mr);
+
+  *pDest = pSrc;  // pointer copy
+
   return TSDB_CODE_SUCCESS;
 _err:
+  *pDest = NULL;
   metaReaderClear(&mr);
-  tdFreeRSmaInfo(pSma, pDest, false);
+  smaError("vgId:%d, rsma clone env failed for %" PRIi64 " since %s", TD_VID(pVnode), pSrc->suid, terrstr());
   return TSDB_CODE_FAILED;
 }
\ No newline at end of file
-
-// ...
\ No newline at end of file

source/dnode/vnode/src/vnd/vnodeCommit.c

@@ -223,6 +223,8 @@ int vnodeCommit(SVnode *pVnode) {
   vnodeBufPoolUnRef(pVnode->inUse);
   pVnode->inUse = NULL;
 
+  pVnode->state.commitTerm = pVnode->state.applyTerm;
+
   // save info
   info.config = pVnode->config;
   info.state.committed = pVnode->state.applied;

source/dnode/vnode/src/vnd/vnodeQuery.c

@@ -273,6 +273,9 @@ int32_t vnodeGetBatchMeta(SVnode *pVnode, SRpcMsg *pMsg) {
   }
 
   for (int32_t i = 0; i < msgNum; ++i) {
+    req.msgIdx = ntohl(*(int32_t *)((char *)pMsg->pCont + offset));
+    offset += sizeof(req.msgIdx);
+
     req.msgType = ntohl(*(int32_t *)((char *)pMsg->pCont + offset));
     offset += sizeof(req.msgType);
 
@@ -301,6 +304,7 @@ int32_t vnodeGetBatchMeta(SVnode *pVnode, SRpcMsg *pMsg) {
         break;
     }
 
+    rsp.msgIdx = req.msgIdx;
     rsp.reqType = reqMsg.msgType;
     rsp.msgLen = reqMsg.contLen;
     rsp.rspCode = reqMsg.code;
@@ -327,6 +331,8 @@ int32_t vnodeGetBatchMeta(SVnode *pVnode, SRpcMsg *pMsg) {
 
     *(int32_t *)((char *)pRsp + offset) = htonl(p->reqType);
     offset += sizeof(p->reqType);
+    *(int32_t *)((char *)pRsp + offset) = htonl(p->msgIdx);
+    offset += sizeof(p->msgIdx);
     *(int32_t *)((char *)pRsp + offset) = htonl(p->msgLen);
     offset += sizeof(p->msgLen);
     *(int32_t *)((char *)pRsp + offset) = htonl(p->rspCode);

source/dnode/vnode/src/vnd/vnodeSvr.c

@@ -351,7 +351,6 @@ int32_t vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo) {
 // TODO: remove the function
 void smaHandleRes(void *pVnode, int64_t smaId, const SArray *data) {
   // TODO
-
   blockDebugShowDataBlocks(data, __func__);
   tdProcessTSmaInsert(((SVnode *)pVnode)->pSma, smaId, (const char *)data);
 }
@@ -852,6 +851,7 @@ static int32_t vnodeProcessSubmitReq(SVnode *pVnode, int64_t version, void *pReq
       if (metaCreateTable(pVnode->pMeta, version, &createTbReq) < 0) {
         if (terrno != TSDB_CODE_TDB_TABLE_ALREADY_EXIST) {
           submitBlkRsp.code = terrno;
+          pRsp->code = terrno;
           tDecoderClear(&decoder);
           taosArrayDestroy(createTbReq.ctb.tagName);
           goto _exit;

source/libs/catalog/inc/catalogInt.h

@@ -32,6 +32,7 @@ extern "C" {
 #define CTG_DEFAULT_RENT_SLOT_SIZE 10
 #define CTG_DEFAULT_MAX_RETRY_TIMES 3
 #define CTG_DEFAULT_BATCH_NUM 64
+#define CTG_DEFAULT_FETCH_NUM 8
 
 #define CTG_RENT_SLOT_SECOND 1.5
 
@@ -80,6 +81,8 @@ typedef enum {
   CTG_TASK_GET_UDF,
   CTG_TASK_GET_USER,
   CTG_TASK_GET_SVR_VER,
+  CTG_TASK_GET_TB_META_BATCH,
+  CTG_TASK_GET_TB_HASH_BATCH,
 } CTG_TASK_TYPE;
 
 typedef enum {
@@ -110,6 +113,23 @@ typedef struct SCtgTbMetaCtx {
   int32_t flag;
 } SCtgTbMetaCtx;
 
+typedef struct SCtgFetch {
+  int32_t dbIdx;
+  int32_t tbIdx;
+  int32_t fetchIdx;
+  int32_t resIdx;
+  int32_t flag;
+  SCtgTbCacheInfo tbInfo;
+  int32_t vgId;
+} SCtgFetch;
+
+typedef struct SCtgTbMetasCtx {
+  int32_t fetchNum;
+  SArray* pNames;
+  SArray* pResList;
+  SArray* pFetchs;
+} SCtgTbMetasCtx;
+
 typedef struct SCtgTbIndexCtx {
   SName* pName;
 } SCtgTbIndexCtx;
@@ -137,6 +157,14 @@ typedef struct SCtgTbHashCtx {
   SName* pName;
 } SCtgTbHashCtx;
 
+typedef struct SCtgTbHashsCtx {
+  int32_t fetchNum;
+  SArray* pNames;
+  SArray* pResList;
+  SArray* pFetchs;
+} SCtgTbHashsCtx;
+
+
 typedef struct SCtgIndexCtx {
   char indexFName[TSDB_INDEX_FNAME_LEN];
 } SCtgIndexCtx;
@@ -211,6 +239,7 @@ typedef struct SCtgBatch {
   SRequestConnInfo conn;
   char    dbFName[TSDB_DB_FNAME_LEN];
   SArray* pTaskIds;
+  SArray* pMsgIdxs;
 } SCtgBatch;
 
 typedef struct SCtgJob {
@@ -218,6 +247,7 @@ typedef struct SCtgJob {
   int32_t           batchId;
   SHashObj*         pBatchs;
   SArray*           pTasks;
+  int32_t           subTaskNum;
   int32_t           taskDone;
   SMetaData         jobRes;
   int32_t           jobResCode;
@@ -258,6 +288,7 @@ typedef struct SCtgTaskCallbackParam {
   SArray*                 taskId;
   int32_t                 reqType;
   int32_t                 batchId;
+  SArray*                 msgIdx;
 } SCtgTaskCallbackParam;
 
 
@@ -276,6 +307,7 @@ typedef struct SCtgTask {
   int32_t         taskId;
   SCtgJob*        pJob;
   void*           taskCtx;
+  SArray*         msgCtxs;
   SCtgMsgCtx      msgCtx;
   int32_t         code;
   void*           res;
@@ -286,9 +318,14 @@ typedef struct SCtgTask {
   SHashObj*       pBatchs;
 } SCtgTask;
 
+typedef struct SCtgTaskReq {
+  SCtgTask* pTask;
+  int32_t   msgIdx;
+} SCtgTaskReq;
+
 typedef int32_t (*ctgInitTaskFp)(SCtgJob*, int32_t, void*);
 typedef int32_t (*ctgLanchTaskFp)(SCtgTask*);
-typedef int32_t (*ctgHandleTaskMsgRspFp)(SCtgTask*, int32_t, const SDataBuf *, int32_t);
+typedef int32_t (*ctgHandleTaskMsgRspFp)(SCtgTaskReq*, int32_t, const SDataBuf *, int32_t);
 typedef int32_t (*ctgDumpTaskResFp)(SCtgTask*);
 typedef int32_t (*ctgCloneTaskResFp)(SCtgTask*, void**);
 typedef int32_t (*ctgCompTaskFp)(SCtgTask*, void*, bool*);
@@ -487,6 +524,8 @@ typedef struct SCtgOperation {
 #define CTG_FLAG_MAKE_STB(_isStb) (((_isStb) == 1) ? CTG_FLAG_STB : ((_isStb) == 0 ? CTG_FLAG_NOT_STB : CTG_FLAG_UNKNOWN_STB))
 #define CTG_FLAG_MATCH_STB(_flag, tbType) (CTG_FLAG_IS_UNKNOWN_STB(_flag) || (CTG_FLAG_IS_STB(_flag) && (tbType) == TSDB_SUPER_TABLE) || (CTG_FLAG_IS_NOT_STB(_flag) && (tbType) != TSDB_SUPER_TABLE))
 
+#define CTG_GET_TASK_MSGCTX(_task, _id) (((CTG_TASK_GET_TB_META_BATCH == (_task)->type) || (CTG_TASK_GET_TB_HASH_BATCH == (_task)->type)) ? taosArrayGet((_task)->msgCtxs, (_id)) : &(_task)->msgCtx)
+
 #define CTG_META_SIZE(pMeta) (sizeof(STableMeta) + ((pMeta)->tableInfo.numOfTags + (pMeta)->tableInfo.numOfColumns) * sizeof(SSchema))
 
 #define CTG_TABLE_NOT_EXIST(code) (code == CTG_ERR_CODE_TABLE_NOT_EXIST) 
@@ -586,6 +625,7 @@ int32_t ctgdShowCacheInfo(void);
 
 int32_t ctgRemoveTbMetaFromCache(SCatalog* pCtg, SName* pTableName, bool syncReq);
 int32_t ctgGetTbMetaFromCache(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgTbMetaCtx* ctx, STableMeta** pTableMeta);
+int32_t ctgGetTbMetasFromCache(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgTbMetasCtx* ctx, int32_t dbIdx, int32_t *fetchIdx, int32_t baseResIdx, SArray* pList);
 
 int32_t ctgOpUpdateVgroup(SCtgCacheOperation *action);
 int32_t ctgOpUpdateTbMeta(SCtgCacheOperation *action);
@@ -631,7 +671,7 @@ int32_t ctgGetTbHashVgroupFromCache(SCatalog *pCtg, const SName *pTableName, SVg
 
 
 int32_t ctgProcessRspMsg(void* out, int32_t reqType, char* msg, int32_t msgSize, int32_t rspCode, char* target);
-int32_t ctgGetDBVgInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SBuildUseDBInput *input, SUseDbOutput *out, SCtgTask* pTask);
+int32_t ctgGetDBVgInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SBuildUseDBInput *input, SUseDbOutput *out, SCtgTaskReq* tReq);
 int32_t ctgGetQnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SArray *out, SCtgTask* pTask);
 int32_t ctgGetDnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SArray **out, SCtgTask* pTask);
 int32_t ctgGetDBCfgFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const char *dbFName, SDbCfgInfo *out, SCtgTask* pTask);
@@ -639,9 +679,9 @@ int32_t ctgGetIndexInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const
 int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SName *name, STableIndex* out, SCtgTask* pTask);
 int32_t ctgGetUdfInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const char *funcName, SFuncInfo *out, SCtgTask* pTask);
 int32_t ctgGetUserDbAuthFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const char *user, SGetUserAuthRsp *out, SCtgTask* pTask);
-int32_t ctgGetTbMetaFromMnodeImpl(SCatalog* pCtg, SRequestConnInfo *pConn, char *dbFName, char* tbName, STableMetaOutput* out, SCtgTask* pTask);
-int32_t ctgGetTbMetaFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const SName* pTableName, STableMetaOutput* out, SCtgTask* pTask);
-int32_t ctgGetTbMetaFromVnode(SCatalog* pCtg, SRequestConnInfo *pConn, const SName* pTableName, SVgroupInfo *vgroupInfo, STableMetaOutput* out, SCtgTask* pTask);
+int32_t ctgGetTbMetaFromMnodeImpl(SCatalog* pCtg, SRequestConnInfo *pConn, char *dbFName, char* tbName, STableMetaOutput* out, SCtgTaskReq* tReq);
+int32_t ctgGetTbMetaFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const SName* pTableName, STableMetaOutput* out, SCtgTaskReq* tReq);
+int32_t ctgGetTbMetaFromVnode(SCatalog* pCtg, SRequestConnInfo *pConn, const SName* pTableName, SVgroupInfo *vgroupInfo, STableMetaOutput* out, SCtgTaskReq* tReq);
 int32_t ctgGetTableCfgFromVnode(SCatalog* pCtg, SRequestConnInfo *pConn, const SName* pTableName, SVgroupInfo *vgroupInfo, STableCfg **out, SCtgTask* pTask);
 int32_t ctgGetTableCfgFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const SName* pTableName, STableCfg **out, SCtgTask* pTask);
 int32_t ctgGetSvrVerFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, char **out, SCtgTask* pTask);
@@ -664,6 +704,7 @@ void    ctgFreeJob(void* job);
 void    ctgFreeHandleImpl(SCatalog* pCtg);
 void    ctgFreeVgInfo(SDBVgInfo *vgInfo);
 int32_t ctgGetVgInfoFromHashValue(SCatalog *pCtg, SDBVgInfo *dbInfo, const SName *pTableName, SVgroupInfo *pVgroup);
+int32_t ctgGetVgInfosFromHashValue(SCatalog *pCtg, SCtgTaskReq* tReq, SDBVgInfo *dbInfo, SCtgTbHashsCtx *pCtx, char* dbFName, SArray* pNames, bool update);
 void    ctgResetTbMetaTask(SCtgTask* pTask);
 void    ctgFreeDbCache(SCtgDBCache *dbCache);
 int32_t ctgStbVersionSortCompare(const void* key1, const void* key2);
@@ -672,8 +713,11 @@ int32_t ctgStbVersionSearchCompare(const void* key1, const void* key2);
 int32_t ctgDbVgVersionSearchCompare(const void* key1, const void* key2);
 void    ctgFreeSTableMetaOutput(STableMetaOutput* pOutput);
 int32_t ctgUpdateMsgCtx(SCtgMsgCtx* pCtx, int32_t reqType, void* out, char* target);
+int32_t ctgAddMsgCtx(SArray* pCtxs, int32_t reqType, void* out, char* target);
 char *  ctgTaskTypeStr(CTG_TASK_TYPE type);
 int32_t ctgUpdateSendTargetInfo(SMsgSendInfo *pMsgSendInfo, int32_t msgType, char* dbFName, int32_t vgId);
+int32_t ctgGetTablesReqNum(SArray *pList);
+int32_t ctgAddFetch(SArray** pFetchs, int32_t dbIdx, int32_t tbIdx, int32_t *fetchIdx, int32_t resIdx, int32_t flag);
 int32_t ctgCloneTableIndex(SArray* pIndex, SArray** pRes);
 void    ctgFreeSTableIndex(void *info);
 void    ctgClearSubTaskRes(SCtgSubRes *pRes);
@@ -682,6 +726,7 @@ void    ctgClearHandle(SCatalog* pCtg);
 void    ctgFreeTbCacheImpl(SCtgTbCache *pCache);
 int32_t ctgRemoveTbMeta(SCatalog* pCtg, SName* pTableName);
 int32_t ctgGetTbHashVgroup(SCatalog *pCtg, SRequestConnInfo *pConn, const SName *pTableName, SVgroupInfo *pVgroup);
+SName*  ctgGetFetchName(SArray* pNames, SCtgFetch* pFetch);
 
 
 extern SCatalogMgmt gCtgMgmt;

source/libs/catalog/src/ctgAsync.c

@@ -50,6 +50,31 @@ int32_t ctgInitGetTbMetaTask(SCtgJob *pJob, int32_t taskIdx, void* param) {
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t ctgInitGetTbMetasTask(SCtgJob *pJob, int32_t taskIdx, void* param) {
+  SName *name = (SName*)param;
+  SCtgTask task = {0};
+
+  task.type = CTG_TASK_GET_TB_META_BATCH;
+  task.taskId = taskIdx;
+  task.pJob = pJob;
+
+  task.taskCtx = taosMemoryCalloc(1, sizeof(SCtgTbMetasCtx));
+  if (NULL == task.taskCtx) {
+    CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+  }
+
+  SCtgTbMetasCtx* ctx = task.taskCtx;
+  ctx->pNames = param;
+  ctx->pResList = taosArrayInit(pJob->tbMetaNum, sizeof(SMetaRes));
+
+  taosArrayPush(pJob->pTasks, &task);
+
+  qDebug("QID:0x%" PRIx64 " the %dth task type %s initialized, dbNum:%d, tbNum:%d", 
+         pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), taosArrayGetSize(ctx->pNames), pJob->tbMetaNum);
+
+  return TSDB_CODE_SUCCESS;
+}
+
 int32_t ctgInitGetDbVgTask(SCtgJob *pJob, int32_t taskIdx, void* param) {
   char *dbFName = (char*)param;
   SCtgTask task = {0};
@@ -153,6 +178,32 @@ int32_t ctgInitGetTbHashTask(SCtgJob *pJob, int32_t taskIdx, void* param) {
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t ctgInitGetTbHashsTask(SCtgJob *pJob, int32_t taskIdx, void* param) {
+  SName *name = (SName*)param;
+  SCtgTask task = {0};
+
+  task.type = CTG_TASK_GET_TB_HASH_BATCH;
+  task.taskId = taskIdx;
+  task.pJob = pJob;
+
+  task.taskCtx = taosMemoryCalloc(1, sizeof(SCtgTbHashsCtx));
+  if (NULL == task.taskCtx) {
+    CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+  }
+
+  SCtgTbHashsCtx* ctx = task.taskCtx;
+  ctx->pNames = param;
+  ctx->pResList = taosArrayInit(pJob->tbHashNum, sizeof(SMetaRes));
+
+  taosArrayPush(pJob->pTasks, &task);
+
+  qDebug("QID:0x%" PRIx64 " the %dth task type %s initialized, dbNum:%d, tbNum:%d", 
+         pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), taosArrayGetSize(ctx->pNames), pJob->tbHashNum);
+
+  return TSDB_CODE_SUCCESS;
+}
+
+
 int32_t ctgInitGetQnodeTask(SCtgJob *pJob, int32_t taskIdx, void* param) {
   SCtgTask task = {0};
 
@@ -328,10 +379,12 @@ int32_t ctgInitGetTbCfgTask(SCtgJob *pJob, int32_t taskIdx, void* param) {
 }
 
 
-
 int32_t ctgHandleForceUpdate(SCatalog* pCtg, int32_t taskNum, SCtgJob *pJob, const SCatalogReq* pReq) {
   SHashObj* pDb = taosHashInit(taskNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
-  if (NULL == pDb) {
+  SHashObj* pTb = taosHashInit(taskNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
+  if (NULL == pDb || NULL == pTb) {
+    taosHashCleanup(pDb);
+    taosHashCleanup(pTb);
     CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
   }
   
@@ -350,18 +403,26 @@ int32_t ctgHandleForceUpdate(SCatalog* pCtg, int32_t taskNum, SCtgJob *pJob, con
     taosHashPut(pDb, dbFName, strlen(dbFName), dbFName, TSDB_DB_FNAME_LEN);
   }
 
-  for (int32_t i = 0; i < pJob->tbMetaNum; ++i) {
-    SName* name = taosArrayGet(pReq->pTableMeta, i);
-    char dbFName[TSDB_DB_FNAME_LEN];
-    tNameGetFullDbName(name, dbFName);
-    taosHashPut(pDb, dbFName, strlen(dbFName), dbFName, TSDB_DB_FNAME_LEN);
+  int32_t dbNum = taosArrayGetSize(pReq->pTableMeta);
+  for (int32_t i = 0; i < dbNum; ++i) {
+    STablesReq* p = taosArrayGet(pReq->pTableMeta, i);
+    taosHashPut(pDb, p->dbFName, strlen(p->dbFName), p->dbFName, TSDB_DB_FNAME_LEN);
+    int32_t tbNum = taosArrayGetSize(p->pTables);
+    for (int32_t m = 0; m < tbNum; ++m) {
+      SName* name = taosArrayGet(p->pTables, m);
+      taosHashPut(pTb, name, sizeof(SName), name, sizeof(SName));
+    }
   }
 
-  for (int32_t i = 0; i < pJob->tbHashNum; ++i) {
-    SName* name = taosArrayGet(pReq->pTableHash, i);
-    char dbFName[TSDB_DB_FNAME_LEN];
-    tNameGetFullDbName(name, dbFName);
-    taosHashPut(pDb, dbFName, strlen(dbFName), dbFName, TSDB_DB_FNAME_LEN);
+  dbNum = taosArrayGetSize(pReq->pTableHash);
+  for (int32_t i = 0; i < dbNum; ++i) {
+    STablesReq* p = taosArrayGet(pReq->pTableHash, i);
+    taosHashPut(pDb, p->dbFName, strlen(p->dbFName), p->dbFName, TSDB_DB_FNAME_LEN);
+    int32_t tbNum = taosArrayGetSize(p->pTables);
+    for (int32_t m = 0; m < tbNum; ++m) {
+      SName* name = taosArrayGet(p->pTables, m);
+      taosHashPut(pTb, name, sizeof(SName), name, sizeof(SName));
+    }
   }
 
   for (int32_t i = 0; i < pJob->tbCfgNum; ++i) {
@@ -380,16 +441,6 @@ int32_t ctgHandleForceUpdate(SCatalog* pCtg, int32_t taskNum, SCtgJob *pJob, con
   taosHashCleanup(pDb);
 
   // REFRESH TABLE META
-  SHashObj* pTb = taosHashInit(taskNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
-  for (int32_t i = 0; i < pJob->tbMetaNum; ++i) {
-    SName* name = taosArrayGet(pReq->pTableMeta, i);
-    taosHashPut(pTb, name, sizeof(SName), name, sizeof(SName));
-  }
-  
-  for (int32_t i = 0; i < pJob->tbHashNum; ++i) {
-    SName* name = taosArrayGet(pReq->pTableHash, i);
-    taosHashPut(pTb, name, sizeof(SName), name, sizeof(SName));
-  }
 
   for (int32_t i = 0; i < pJob->tbCfgNum; ++i) {
     SName* name = taosArrayGet(pReq->pTableCfg, i);
@@ -429,9 +480,9 @@ int32_t ctgInitTask(SCtgJob *pJob, CTG_TASK_TYPE type, void* param, int32_t *tas
 
 int32_t ctgInitJob(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgJob** job, const SCatalogReq* pReq, catalogCallback fp, void* param) {
   int32_t code = 0;
-  int32_t tbMetaNum = (int32_t)taosArrayGetSize(pReq->pTableMeta);
+  int32_t tbMetaNum = (int32_t)ctgGetTablesReqNum(pReq->pTableMeta);
   int32_t dbVgNum = (int32_t)taosArrayGetSize(pReq->pDbVgroup);
-  int32_t tbHashNum = (int32_t)taosArrayGetSize(pReq->pTableHash);
+  int32_t tbHashNum = (int32_t)ctgGetTablesReqNum(pReq->pTableHash);
   int32_t udfNum = (int32_t)taosArrayGetSize(pReq->pUdf);
   int32_t qnodeNum = pReq->qNodeRequired ? 1 : 0;
   int32_t dnodeNum = pReq->dNodeRequired ? 1 : 0;
@@ -453,6 +504,7 @@ int32_t ctgInitJob(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgJob** job, const
 
   SCtgJob *pJob = *job;
 
+  pJob->subTaskNum = taskNum;
   pJob->queryId = pConn->requestId;
   pJob->userFp = fp;
   pJob->pCtg    = pCtg;
@@ -506,15 +558,27 @@ int32_t ctgInitJob(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgJob** job, const
     CTG_ERR_JRET(ctgInitTask(pJob, CTG_TASK_GET_DB_INFO, dbFName, NULL));
   }
 
+#if 0
   for (int32_t i = 0; i < tbMetaNum; ++i) {
     SName* name = taosArrayGet(pReq->pTableMeta, i);
     CTG_ERR_JRET(ctgInitTask(pJob, CTG_TASK_GET_TB_META, name, NULL));
   }
+#else
+  if (tbMetaNum > 0) {
+    CTG_ERR_JRET(ctgInitTask(pJob, CTG_TASK_GET_TB_META_BATCH, pReq->pTableMeta, NULL));
+  }
+#endif
 
+#if 0
   for (int32_t i = 0; i < tbHashNum; ++i) {
     SName* name = taosArrayGet(pReq->pTableHash, i);
     CTG_ERR_JRET(ctgInitTask(pJob, CTG_TASK_GET_TB_HASH, name, NULL));
   }
+#else
+  if (tbHashNum > 0) {
+    CTG_ERR_JRET(ctgInitTask(pJob, CTG_TASK_GET_TB_HASH_BATCH, pReq->pTableHash, NULL));
+  }
+#endif
 
   for (int32_t i = 0; i < tbIndexNum; ++i) {
     SName* name = taosArrayGet(pReq->pTableIndex, i);
@@ -586,6 +650,15 @@ int32_t ctgDumpTbMetaRes(SCtgTask* pTask) {
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t ctgDumpTbMetasRes(SCtgTask* pTask) {
+  SCtgJob* pJob = pTask->pJob;
+
+  pJob->jobRes.pTableMeta = pTask->res;
+
+  return TSDB_CODE_SUCCESS;
+}
+
+
 int32_t ctgDumpDbVgRes(SCtgTask* pTask) {
   SCtgJob* pJob = pTask->pJob;
   if (NULL == pJob->jobRes.pDbVgroup) {
@@ -616,6 +689,14 @@ int32_t ctgDumpTbHashRes(SCtgTask* pTask) {
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t ctgDumpTbHashsRes(SCtgTask* pTask) {
+  SCtgJob* pJob = pTask->pJob;
+
+  pJob->jobRes.pTableHash = pTask->res;
+
+  return TSDB_CODE_SUCCESS;
+}
+
 int32_t ctgDumpTbIndexRes(SCtgTask* pTask) {
   SCtgJob* pJob = pTask->pJob;
   if (NULL == pJob->jobRes.pTableIndex) {
@@ -844,46 +925,51 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetTbMetaRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetTbMetaRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
   SCtgDBCache *dbCache = NULL;
-  SCtgTbMetaCtx* ctx = (SCtgTbMetaCtx*)pTask->taskCtx;
+  SCtgTask* pTask = tReq->pTask;
   SCatalog* pCtg = pTask->pJob->pCtg; 
   SRequestConnInfo* pConn = &pTask->pJob->conn;
+  SCtgMsgCtx* pMsgCtx = CTG_GET_TASK_MSGCTX(pTask, tReq->msgIdx);
+  SCtgTbMetaCtx* ctx = (SCtgTbMetaCtx*)pTask->taskCtx;
+  SName* pName = ctx->pName;
+  int32_t flag = ctx->flag;
+  int32_t* vgId = &ctx->vgId;
 
-  CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
+  CTG_ERR_JRET(ctgProcessRspMsg(pMsgCtx->out, reqType, pMsg->pData, pMsg->len, rspCode, pMsgCtx->target));
 
   switch (reqType) {
     case TDMT_MND_USE_DB: {
-      SUseDbOutput* pOut = (SUseDbOutput*)pTask->msgCtx.out;
+      SUseDbOutput* pOut = (SUseDbOutput*)pMsgCtx->out;
 
       SVgroupInfo vgInfo = {0};
-      CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, pOut->dbVgroup, ctx->pName, &vgInfo));
+      CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, pOut->dbVgroup, pName, &vgInfo));
 
-      ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(ctx->pName), ctx->flag);
+      ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(pName), flag);
 
-      ctx->vgId = vgInfo.vgId;
-      CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, ctx->pName, &vgInfo, NULL, pTask));
+      *vgId = vgInfo.vgId;
+      CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, pName, &vgInfo, NULL, tReq));
 
       return TSDB_CODE_SUCCESS;
     }
     case TDMT_MND_TABLE_META: {
-      STableMetaOutput* pOut = (STableMetaOutput*)pTask->msgCtx.out;
+      STableMetaOutput* pOut = (STableMetaOutput*)pMsgCtx->out;
       
       if (CTG_IS_META_NULL(pOut->metaType)) {
-        if (CTG_FLAG_IS_STB(ctx->flag)) {
+        if (CTG_FLAG_IS_STB(flag)) {
           char dbFName[TSDB_DB_FNAME_LEN] = {0};
-          tNameGetFullDbName(ctx->pName, dbFName);
+          tNameGetFullDbName(pName, dbFName);
           
           CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, dbFName, &dbCache));
           if (NULL != dbCache) {
             SVgroupInfo vgInfo = {0};
-            CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, ctx->pName, &vgInfo));
+            CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, pName, &vgInfo));
           
-            ctgDebug("will refresh tbmeta, supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(ctx->pName), ctx->flag);
+            ctgDebug("will refresh tbmeta, supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(pName), flag);
 
-            ctx->vgId = vgInfo.vgId;          
-            CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, ctx->pName, &vgInfo, NULL, pTask));
+            *vgId = vgInfo.vgId;          
+            CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, pName, &vgInfo, NULL, tReq));
 
             ctgReleaseVgInfoToCache(pCtg, dbCache);
           } else {
@@ -892,58 +978,231 @@ int32_t ctgHandleGetTbMetaRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *
             tstrncpy(input.db, dbFName, tListLen(input.db));
             input.vgVersion = CTG_DEFAULT_INVALID_VERSION;
           
-            CTG_ERR_JRET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, pTask));
+            CTG_ERR_JRET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, tReq));
           }
 
           return TSDB_CODE_SUCCESS;
         }
         
-        ctgError("no tbmeta got, tbName:%s", tNameGetTableName(ctx->pName));
-        ctgRemoveTbMetaFromCache(pCtg, ctx->pName, false);
+        ctgError("no tbmeta got, tbName:%s", tNameGetTableName(pName));
+        ctgRemoveTbMetaFromCache(pCtg, pName, false);
         
         CTG_ERR_JRET(CTG_ERR_CODE_TABLE_NOT_EXIST);
       }
 
-      if (pTask->msgCtx.lastOut) {
-        TSWAP(pTask->msgCtx.out, pTask->msgCtx.lastOut);
-        STableMetaOutput* pLastOut = (STableMetaOutput*)pTask->msgCtx.out;
+      if (pMsgCtx->lastOut) {
+        TSWAP(pMsgCtx->out, pMsgCtx->lastOut);
+        STableMetaOutput* pLastOut = (STableMetaOutput*)pMsgCtx->out;
         TSWAP(pLastOut->tbMeta, pOut->tbMeta);
       }
       
       break;
     }
     case TDMT_VND_TABLE_META: {
-      STableMetaOutput* pOut = (STableMetaOutput*)pTask->msgCtx.out;
+      STableMetaOutput* pOut = (STableMetaOutput*)pMsgCtx->out;
       
       if (CTG_IS_META_NULL(pOut->metaType)) {
-        ctgError("no tbmeta got, tbNmae:%s", tNameGetTableName(ctx->pName));
-        ctgRemoveTbMetaFromCache(pCtg, ctx->pName, false);
+        ctgError("no tbmeta got, tbNmae:%s", tNameGetTableName(pName));
+        ctgRemoveTbMetaFromCache(pCtg, pName, false);
         CTG_ERR_JRET(CTG_ERR_CODE_TABLE_NOT_EXIST);
       }
 
-      if (CTG_FLAG_IS_STB(ctx->flag)) {
+      if (CTG_FLAG_IS_STB(flag)) {
         break;
       }
       
       if (CTG_IS_META_TABLE(pOut->metaType) && TSDB_SUPER_TABLE == pOut->tbMeta->tableType) {
-        ctgDebug("will continue to refresh tbmeta since got stb, tbName:%s", tNameGetTableName(ctx->pName));
+        ctgDebug("will continue to refresh tbmeta since got stb, tbName:%s", tNameGetTableName(pName));
       
         taosMemoryFreeClear(pOut->tbMeta);
         
-        CTG_RET(ctgGetTbMetaFromMnode(pCtg, pConn, ctx->pName, NULL, pTask));
+        CTG_RET(ctgGetTbMetaFromMnode(pCtg, pConn, pName, NULL, tReq));
       } else if (CTG_IS_META_BOTH(pOut->metaType)) {
         int32_t exist = 0;
-        if (!CTG_FLAG_IS_FORCE_UPDATE(ctx->flag)) {
-          CTG_ERR_JRET(ctgTbMetaExistInCache(pCtg, pOut->dbFName, pOut->tbName, &exist));
+        if (!CTG_FLAG_IS_FORCE_UPDATE(flag)) {
+          SName stbName = *pName;
+          strcpy(stbName.tname, pOut->tbName);
+          SCtgTbMetaCtx stbCtx = {0};
+          stbCtx.flag = flag;
+          stbCtx.pName = &stbName;
+
+          taosMemoryFreeClear(pOut->tbMeta);          
+          CTG_ERR_JRET(ctgReadTbMetaFromCache(pCtg, &stbCtx, &pOut->tbMeta));
+          if (pOut->tbMeta) {
+            exist = 1;
+          }
         }
       
         if (0 == exist) {
-          TSWAP(pTask->msgCtx.lastOut, pTask->msgCtx.out);
-          CTG_RET(ctgGetTbMetaFromMnodeImpl(pCtg, pConn, pOut->dbFName, pOut->tbName, NULL, pTask));
+          TSWAP(pMsgCtx->lastOut, pMsgCtx->out);
+          CTG_RET(ctgGetTbMetaFromMnodeImpl(pCtg, pConn, pOut->dbFName, pOut->tbName, NULL, tReq));
+        }
+      }
+      break;
+    }
+    default:
+      ctgError("invalid reqType %d", reqType);
+      CTG_ERR_JRET(TSDB_CODE_INVALID_MSG);
+      break;
+  }
+
+  STableMetaOutput* pOut = (STableMetaOutput*)pMsgCtx->out;
+
+  ctgUpdateTbMetaToCache(pCtg, pOut, false);
+
+  if (CTG_IS_META_BOTH(pOut->metaType)) {
+    memcpy(pOut->tbMeta, &pOut->ctbMeta, sizeof(pOut->ctbMeta));
+  }
+  
+/*
+  else if (CTG_IS_META_CTABLE(pOut->metaType)) {
+    SName stbName = *pName;
+    strcpy(stbName.tname, pOut->tbName);
+    SCtgTbMetaCtx stbCtx = {0};
+    stbCtx.flag = flag;
+    stbCtx.pName = &stbName;
+    
+    CTG_ERR_JRET(ctgReadTbMetaFromCache(pCtg, &stbCtx, &pOut->tbMeta));
+    if (NULL == pOut->tbMeta) {
+      ctgDebug("stb no longer exist, stbName:%s", stbName.tname);
+      CTG_ERR_JRET(ctgRelaunchGetTbMetaTask(pTask));
+
+      return TSDB_CODE_SUCCESS;
+    }
+
+    memcpy(pOut->tbMeta, &pOut->ctbMeta, sizeof(pOut->ctbMeta));
+  }
+*/
+
+  TSWAP(pTask->res, pOut->tbMeta);
+
+_return:
+
+  if (dbCache) {
+    ctgReleaseVgInfoToCache(pCtg, dbCache);
+  }
+
+  if (pTask->res) {
+    ctgHandleTaskEnd(pTask, code);
+  }
+  
+  CTG_RET(code);
+}
+
+
+int32_t ctgHandleGetTbMetasRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+  int32_t code = 0;
+  SCtgDBCache *dbCache = NULL;
+  SCtgTask* pTask = tReq->pTask;
+  SCatalog* pCtg = pTask->pJob->pCtg; 
+  SRequestConnInfo* pConn = &pTask->pJob->conn;
+  SCtgMsgCtx* pMsgCtx = CTG_GET_TASK_MSGCTX(pTask, tReq->msgIdx);
+  SCtgTbMetasCtx* ctx = (SCtgTbMetasCtx*)pTask->taskCtx;
+  SCtgFetch* pFetch = taosArrayGet(ctx->pFetchs, tReq->msgIdx);
+  SName* pName = ctgGetFetchName(ctx->pNames, pFetch);
+  int32_t flag = pFetch->flag;
+  int32_t* vgId = &pFetch->vgId;
+
+  CTG_ERR_JRET(ctgProcessRspMsg(pMsgCtx->out, reqType, pMsg->pData, pMsg->len, rspCode, pMsgCtx->target));
+
+  switch (reqType) {
+    case TDMT_MND_USE_DB: {
+      SUseDbOutput* pOut = (SUseDbOutput*)pMsgCtx->out;
+
+      SVgroupInfo vgInfo = {0};
+      CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, pOut->dbVgroup, pName, &vgInfo));
+
+      ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(pName), flag);
+
+      *vgId = vgInfo.vgId;
+      CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, pName, &vgInfo, NULL, tReq));
+
+      return TSDB_CODE_SUCCESS;
+    }
+    case TDMT_MND_TABLE_META: {
+      STableMetaOutput* pOut = (STableMetaOutput*)pMsgCtx->out;
+      
+      if (CTG_IS_META_NULL(pOut->metaType)) {
+        if (CTG_FLAG_IS_STB(flag)) {
+          char dbFName[TSDB_DB_FNAME_LEN] = {0};
+          tNameGetFullDbName(pName, dbFName);
+          
+          CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, dbFName, &dbCache));
+          if (NULL != dbCache) {
+            SVgroupInfo vgInfo = {0};
+            CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, pName, &vgInfo));
+          
+            ctgDebug("will refresh tbmeta, supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(pName), flag);
+
+            *vgId = vgInfo.vgId;          
+            CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, pName, &vgInfo, NULL, tReq));
+
+            ctgReleaseVgInfoToCache(pCtg, dbCache);
           } else {
+            SBuildUseDBInput input = {0};
+          
+            tstrncpy(input.db, dbFName, tListLen(input.db));
+            input.vgVersion = CTG_DEFAULT_INVALID_VERSION;
+          
+            CTG_ERR_JRET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, tReq));
+          }
+
+          return TSDB_CODE_SUCCESS;
+        }
+        
+        ctgError("no tbmeta got, tbName:%s", tNameGetTableName(pName));
+        ctgRemoveTbMetaFromCache(pCtg, pName, false);
+        
+        CTG_ERR_JRET(CTG_ERR_CODE_TABLE_NOT_EXIST);
+      }
+
+      if (pMsgCtx->lastOut) {
+        TSWAP(pMsgCtx->out, pMsgCtx->lastOut);
+        STableMetaOutput* pLastOut = (STableMetaOutput*)pMsgCtx->out;
+        TSWAP(pLastOut->tbMeta, pOut->tbMeta);
+      }
+      
+      break;
+    }
+    case TDMT_VND_TABLE_META: {
+      STableMetaOutput* pOut = (STableMetaOutput*)pMsgCtx->out;
+      
+      if (CTG_IS_META_NULL(pOut->metaType)) {
+        ctgError("no tbmeta got, tbNmae:%s", tNameGetTableName(pName));
+        ctgRemoveTbMetaFromCache(pCtg, pName, false);
+        CTG_ERR_JRET(CTG_ERR_CODE_TABLE_NOT_EXIST);
+      }
+
+      if (CTG_FLAG_IS_STB(flag)) {
+        break;
+      }
+      
+      if (CTG_IS_META_TABLE(pOut->metaType) && TSDB_SUPER_TABLE == pOut->tbMeta->tableType) {
+        ctgDebug("will continue to refresh tbmeta since got stb, tbName:%s", tNameGetTableName(pName));
+      
+        taosMemoryFreeClear(pOut->tbMeta);
+        
+        CTG_RET(ctgGetTbMetaFromMnode(pCtg, pConn, pName, NULL, tReq));
+      } else if (CTG_IS_META_BOTH(pOut->metaType)) {
+        int32_t exist = 0;
+        if (!CTG_FLAG_IS_FORCE_UPDATE(flag)) {
+          SName stbName = *pName;
+          strcpy(stbName.tname, pOut->tbName);
+          SCtgTbMetaCtx stbCtx = {0};
+          stbCtx.flag = flag;
+          stbCtx.pName = &stbName;
+
           taosMemoryFreeClear(pOut->tbMeta);          
+          CTG_ERR_JRET(ctgReadTbMetaFromCache(pCtg, &stbCtx, &pOut->tbMeta));
+          if (pOut->tbMeta) {
+            ctgDebug("use cached stb meta, tbName:%s", tNameGetTableName(pName));
+            exist = 1;
+          }
+        }
       
-          SET_META_TYPE_CTABLE(pOut->metaType); 
+        if (0 == exist) {
+          TSWAP(pMsgCtx->lastOut, pMsgCtx->out);
+          CTG_RET(ctgGetTbMetaFromMnodeImpl(pCtg, pConn, pOut->dbFName, pOut->tbName, NULL, tReq));
         }
       }
       break;
@@ -954,17 +1213,20 @@ int32_t ctgHandleGetTbMetaRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *
       break;
   }
 
-  STableMetaOutput* pOut = (STableMetaOutput*)pTask->msgCtx.out;
+  STableMetaOutput* pOut = (STableMetaOutput*)pMsgCtx->out;
 
   ctgUpdateTbMetaToCache(pCtg, pOut, false);
 
   if (CTG_IS_META_BOTH(pOut->metaType)) {
     memcpy(pOut->tbMeta, &pOut->ctbMeta, sizeof(pOut->ctbMeta));
-  } else if (CTG_IS_META_CTABLE(pOut->metaType)) {
-    SName stbName = *ctx->pName;
+  }
+  
+/*
+  else if (CTG_IS_META_CTABLE(pOut->metaType)) {
+    SName stbName = *pName;
     strcpy(stbName.tname, pOut->tbName);
     SCtgTbMetaCtx stbCtx = {0};
-    stbCtx.flag = ctx->flag;
+    stbCtx.flag = flag;
     stbCtx.pName = &stbName;
     
     CTG_ERR_JRET(ctgReadTbMetaFromCache(pCtg, &stbCtx, &pOut->tbMeta));
@@ -977,8 +1239,15 @@ int32_t ctgHandleGetTbMetaRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *
 
     memcpy(pOut->tbMeta, &pOut->ctbMeta, sizeof(pOut->ctbMeta));
   }
+*/
 
-  TSWAP(pTask->res, pOut->tbMeta);
+  SMetaRes* pRes = taosArrayGet(ctx->pResList, pFetch->resIdx);
+  pRes->code = 0;
+  pRes->pRes = pOut->tbMeta;
+  pOut->tbMeta = NULL;
+  if (0 == atomic_sub_fetch_32(&ctx->fetchNum, 1)) {
+    TSWAP(pTask->res, ctx->pResList);
+  }
 
 _return:
 
@@ -986,13 +1255,26 @@ _return:
     ctgReleaseVgInfoToCache(pCtg, dbCache);
   }
 
+  if (code) {
+    SMetaRes* pRes = taosArrayGet(ctx->pResList, pFetch->resIdx);
+    pRes->code = code;
+    pRes->pRes = NULL;
+    if (0 == atomic_sub_fetch_32(&ctx->fetchNum, 1)) {
+      TSWAP(pTask->res, ctx->pResList);
+    }
+  }
+
+  if (pTask->res) {
     ctgHandleTaskEnd(pTask, code);
+  }
   
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetDbVgRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+
+int32_t ctgHandleGetDbVgRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   SCtgDbVgCtx* ctx = (SCtgDbVgCtx*)pTask->taskCtx;
   SCatalog* pCtg = pTask->pJob->pCtg; 
 
@@ -1024,8 +1306,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetTbHashRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetTbHashRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   SCtgTbHashCtx* ctx = (SCtgTbHashCtx*)pTask->taskCtx;
   SCatalog* pCtg = pTask->pJob->pCtg; 
 
@@ -1061,8 +1344,65 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetTbIndexRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetTbHashsRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
+  SCtgTbHashsCtx* ctx = (SCtgTbHashsCtx*)pTask->taskCtx;
+  SCatalog* pCtg = pTask->pJob->pCtg; 
+  SCtgMsgCtx* pMsgCtx = CTG_GET_TASK_MSGCTX(pTask, tReq->msgIdx);
+  SCtgFetch* pFetch = taosArrayGet(ctx->pFetchs, tReq->msgIdx);
+
+  CTG_ERR_JRET(ctgProcessRspMsg(pMsgCtx->out, reqType, pMsg->pData, pMsg->len, rspCode, pMsgCtx->target));
+
+  switch (reqType) {
+    case TDMT_MND_USE_DB: {
+      SUseDbOutput* pOut = (SUseDbOutput*)pMsgCtx->out;
+
+      STablesReq* pReq = taosArrayGet(ctx->pNames, pFetch->dbIdx);
+      CTG_ERR_JRET(ctgGetVgInfosFromHashValue(pCtg, tReq, pOut->dbVgroup, ctx, pMsgCtx->target, pReq->pTables, true));
+     
+      CTG_ERR_JRET(ctgUpdateVgroupEnqueue(pCtg, pMsgCtx->target, pOut->dbId, pOut->dbVgroup, false));
+      pOut->dbVgroup = NULL;
+
+      break;
+    }
+    default:
+      ctgError("invalid reqType %d", reqType);
+      CTG_ERR_JRET(TSDB_CODE_INVALID_MSG);
+      break;
+  }
+
+  if (0 == atomic_sub_fetch_32(&ctx->fetchNum, 1)) {
+    TSWAP(pTask->res, ctx->pResList);
+  }
+
+_return:
+
+  if (code) {
+    STablesReq* pReq = taosArrayGet(ctx->pNames, pFetch->dbIdx);
+    int32_t num = taosArrayGetSize(pReq->pTables);
+    for (int32_t i = 0; i < num; ++i) {
+      SMetaRes *pRes = taosArrayGet(ctx->pResList, pFetch->resIdx + i);
+      pRes->code = code;
+      pRes->pRes = NULL;
+    }
+    
+    if (0 == atomic_sub_fetch_32(&ctx->fetchNum, 1)) {
+      TSWAP(pTask->res, ctx->pResList);
+    }
+  }
+
+  if (pTask->res) {
+    ctgHandleTaskEnd(pTask, code);
+  }
+
+  CTG_RET(code);
+}
+
+
+int32_t ctgHandleGetTbIndexRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+  int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
   STableIndex* pOut = (STableIndex*)pTask->msgCtx.out;
@@ -1083,8 +1423,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetTbCfgRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetTbCfgRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   CTG_ERR_JRET(ctgProcessRspMsg(&pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
   TSWAP(pTask->res, pTask->msgCtx.out);
@@ -1096,8 +1437,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetDbCfgRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetDbCfgRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
   TSWAP(pTask->res, pTask->msgCtx.out);
@@ -1109,13 +1451,14 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetDbInfoRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetDbInfoRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   CTG_RET(TSDB_CODE_APP_ERROR);
 }
 
 
-int32_t ctgHandleGetQnodeRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetQnodeRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
   TSWAP(pTask->res, pTask->msgCtx.out);
@@ -1127,8 +1470,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetDnodeRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetDnodeRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   CTG_ERR_JRET(ctgProcessRspMsg(&pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
   TSWAP(pTask->res, pTask->msgCtx.out);
@@ -1140,8 +1484,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetIndexRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetIndexRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
   TSWAP(pTask->res, pTask->msgCtx.out);
@@ -1153,8 +1498,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetUdfRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetUdfRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
   TSWAP(pTask->res, pTask->msgCtx.out);
@@ -1166,8 +1512,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetUserRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetUserRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
   SCtgUserCtx* ctx = (SCtgUserCtx*)pTask->taskCtx;
   SCatalog* pCtg = pTask->pJob->pCtg; 
   bool pass = false;
@@ -1210,8 +1557,9 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgHandleGetSvrVerRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
+int32_t ctgHandleGetSvrVerRsp(SCtgTaskReq* tReq, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
   int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
 
   CTG_ERR_JRET(ctgProcessRspMsg(&pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
 
@@ -1224,45 +1572,45 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgAsyncRefreshTbMeta(SCtgTask *pTask) {
+int32_t ctgAsyncRefreshTbMeta(SCtgTaskReq *tReq, int32_t flag, SName* pName, int32_t* vgId) {
+  SCtgTask* pTask = tReq->pTask;
   SCatalog* pCtg = pTask->pJob->pCtg; 
   SRequestConnInfo* pConn = &pTask->pJob->conn;
   int32_t code = 0;
-  SCtgTbMetaCtx* ctx = (SCtgTbMetaCtx*)pTask->taskCtx;
 
-  if (CTG_FLAG_IS_SYS_DB(ctx->flag)) {
-    ctgDebug("will refresh sys db tbmeta, tbName:%s", tNameGetTableName(ctx->pName));
+  if (CTG_FLAG_IS_SYS_DB(flag)) {
+    ctgDebug("will refresh sys db tbmeta, tbName:%s", tNameGetTableName(pName));
 
-    CTG_RET(ctgGetTbMetaFromMnodeImpl(pCtg, pConn, (char *)ctx->pName->dbname, (char *)ctx->pName->tname, NULL, pTask));
+    CTG_RET(ctgGetTbMetaFromMnodeImpl(pCtg, pConn, (char *)pName->dbname, (char *)pName->tname, NULL, tReq));
   }
 
-  if (CTG_FLAG_IS_STB(ctx->flag)) {
-    ctgDebug("will refresh tbmeta, supposed to be stb, tbName:%s", tNameGetTableName(ctx->pName));
+  if (CTG_FLAG_IS_STB(flag)) {
+    ctgDebug("will refresh tbmeta, supposed to be stb, tbName:%s", tNameGetTableName(pName));
 
     // if get from mnode failed, will not try vnode
-    CTG_RET(ctgGetTbMetaFromMnode(pCtg, pConn, ctx->pName, NULL, pTask));
+    CTG_RET(ctgGetTbMetaFromMnode(pCtg, pConn, pName, NULL, tReq));
   }
 
   SCtgDBCache *dbCache = NULL;
   char dbFName[TSDB_DB_FNAME_LEN] = {0};
-  tNameGetFullDbName(ctx->pName, dbFName);
+  tNameGetFullDbName(pName, dbFName);
   
   CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, dbFName, &dbCache));
   if (dbCache) {
     SVgroupInfo vgInfo = {0};
-    CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, ctx->pName, &vgInfo));
+    CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, pName, &vgInfo));
 
-    ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(ctx->pName), ctx->flag);
+    ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(pName), flag);
 
-    ctx->vgId = vgInfo.vgId;
-    CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, ctx->pName, &vgInfo, NULL, pTask));
+    *vgId = vgInfo.vgId;
+    CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, pName, &vgInfo, NULL, tReq));
   } else {
     SBuildUseDBInput input = {0};
 
     tstrncpy(input.db, dbFName, tListLen(input.db));
     input.vgVersion = CTG_DEFAULT_INVALID_VERSION;
 
-    CTG_ERR_JRET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, pTask));
+    CTG_ERR_JRET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, tReq));
   }
 
 _return:
@@ -1284,7 +1632,50 @@ int32_t ctgLaunchGetTbMetaTask(SCtgTask *pTask) {
     return TSDB_CODE_SUCCESS;
   }
 
-  CTG_ERR_RET(ctgAsyncRefreshTbMeta(pTask));
+  SCtgTbMetaCtx* pCtx = (SCtgTbMetaCtx*)pTask->taskCtx;
+  SCtgTaskReq tReq;
+  tReq.pTask = pTask;
+  tReq.msgIdx = -1;    
+  CTG_ERR_RET(ctgAsyncRefreshTbMeta(&tReq, pCtx->flag, pCtx->pName, &pCtx->vgId));
+
+  return TSDB_CODE_SUCCESS;
+}
+
+int32_t ctgLaunchGetTbMetasTask(SCtgTask *pTask) {
+  SCatalog* pCtg = pTask->pJob->pCtg; 
+  SRequestConnInfo* pConn = &pTask->pJob->conn;
+  SCtgTbMetasCtx* pCtx = (SCtgTbMetasCtx*)pTask->taskCtx;
+
+  int32_t dbNum = taosArrayGetSize(pCtx->pNames);
+  int32_t fetchIdx = 0;
+  int32_t baseResIdx = 0;
+  for (int32_t i = 0; i < dbNum; ++i) {
+    STablesReq* pReq = taosArrayGet(pCtx->pNames, i);
+    ctgDebug("start to check tb metas in db %s, tbNum %d", pReq->dbFName, taosArrayGetSize(pReq->pTables));
+    CTG_ERR_RET(ctgGetTbMetasFromCache(pCtg, pConn, pCtx, i, &fetchIdx, baseResIdx, pReq->pTables));
+    baseResIdx += taosArrayGetSize(pReq->pTables);
+  }
+  
+  pCtx->fetchNum = taosArrayGetSize(pCtx->pFetchs);
+  if (pCtx->fetchNum <= 0) {
+    TSWAP(pTask->res, pCtx->pResList);
+    
+    CTG_ERR_RET(ctgHandleTaskEnd(pTask, 0));
+    return TSDB_CODE_SUCCESS;
+  }
+  
+  pTask->msgCtxs = taosArrayInit(pCtx->fetchNum, sizeof(SCtgMsgCtx));
+  taosArraySetSize(pTask->msgCtxs, pCtx->fetchNum);
+  
+  for (int32_t i = 0; i < pCtx->fetchNum; ++i) {
+    SCtgFetch* pFetch = taosArrayGet(pCtx->pFetchs, i);
+    SName* pName = ctgGetFetchName(pCtx->pNames, pFetch);
+
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = pFetch->fetchIdx;    
+    CTG_ERR_RET(ctgAsyncRefreshTbMeta(&tReq, pFetch->flag, pName, &pFetch->vgId));
+  }
   
   return TSDB_CODE_SUCCESS;
 }
@@ -1310,7 +1701,10 @@ int32_t ctgLaunchGetDbVgTask(SCtgTask *pTask) {
     tstrncpy(input.db, pCtx->dbFName, tListLen(input.db));
     input.vgVersion = CTG_DEFAULT_INVALID_VERSION;
 
-    CTG_ERR_RET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, pTask));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;
+    CTG_ERR_RET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, &tReq));
   }
 
 _return:
@@ -1347,7 +1741,10 @@ int32_t ctgLaunchGetTbHashTask(SCtgTask *pTask) {
     tstrncpy(input.db, pCtx->dbFName, tListLen(input.db));
     input.vgVersion = CTG_DEFAULT_INVALID_VERSION;
 
-    CTG_ERR_RET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, pTask));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_ERR_RET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, &tReq));
   }
 
 _return:
@@ -1359,6 +1756,75 @@ _return:
   CTG_RET(code);
 }
 
+int32_t ctgLaunchGetTbHashsTask(SCtgTask *pTask) {
+  SCatalog* pCtg = pTask->pJob->pCtg; 
+  SRequestConnInfo* pConn = &pTask->pJob->conn;
+  SCtgTbHashsCtx* pCtx = (SCtgTbHashsCtx*)pTask->taskCtx;
+  SCtgDBCache *dbCache = NULL;
+  int32_t dbNum = taosArrayGetSize(pCtx->pNames);
+  int32_t fetchIdx = 0;
+  int32_t baseResIdx = 0;
+  int32_t code = 0;
+  
+  for (int32_t i = 0; i < dbNum; ++i) {
+    STablesReq* pReq = taosArrayGet(pCtx->pNames, i);
+
+    CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, pReq->dbFName, &dbCache));
+
+    if (NULL != dbCache) {
+      SCtgTaskReq tReq;
+      tReq.pTask = pTask;
+      tReq.msgIdx = -1;    
+      CTG_ERR_JRET(ctgGetVgInfosFromHashValue(pCtg, &tReq, dbCache->vgCache.vgInfo, pCtx, pReq->dbFName, pReq->pTables, false));
+
+      ctgReleaseVgInfoToCache(pCtg, dbCache);
+      dbCache = NULL;
+
+      baseResIdx += taosArrayGetSize(pReq->pTables);
+    } else {
+      ctgAddFetch(&pCtx->pFetchs, i, -1, &fetchIdx, baseResIdx, 0);
+
+      baseResIdx += taosArrayGetSize(pReq->pTables);
+      taosArraySetSize(pCtx->pResList, baseResIdx);      
+    }
+  }
+
+  pCtx->fetchNum = taosArrayGetSize(pCtx->pFetchs);
+  if (pCtx->fetchNum <= 0) {
+    TSWAP(pTask->res, pCtx->pResList);
+    
+    CTG_ERR_RET(ctgHandleTaskEnd(pTask, 0));
+    return TSDB_CODE_SUCCESS;
+  }
+  
+  pTask->msgCtxs = taosArrayInit(pCtx->fetchNum, sizeof(SCtgMsgCtx));
+  taosArraySetSize(pTask->msgCtxs, pCtx->fetchNum);
+  
+  for (int32_t i = 0; i < pCtx->fetchNum; ++i) {
+    SCtgFetch* pFetch = taosArrayGet(pCtx->pFetchs, i);
+    STablesReq* pReq = taosArrayGet(pCtx->pNames, pFetch->dbIdx);
+
+    SBuildUseDBInput input = {0};
+    strcpy(input.db, pReq->dbFName);
+    
+    input.vgVersion = CTG_DEFAULT_INVALID_VERSION;
+
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = pFetch->fetchIdx;
+    CTG_ERR_RET(ctgGetDBVgInfoFromMnode(pCtg, pConn, &input, NULL, &tReq));
+  }
+
+_return:
+
+  if (dbCache) {
+    ctgReleaseVgInfoToCache(pCtg, dbCache);
+  }
+  
+  return code;
+}
+
+
 int32_t ctgLaunchGetTbIndexTask(SCtgTask *pTask) {
   int32_t code = 0;
   SCatalog* pCtg = pTask->pJob->pCtg; 
@@ -1610,6 +2076,8 @@ SCtgAsyncFps gCtgAsyncFps[] = {
   {ctgInitGetUdfTask,      ctgLaunchGetUdfTask,      ctgHandleGetUdfRsp,     ctgDumpUdfRes,     NULL,               NULL},
   {ctgInitGetUserTask,     ctgLaunchGetUserTask,     ctgHandleGetUserRsp,    ctgDumpUserRes,    NULL,               NULL},
   {ctgInitGetSvrVerTask,   ctgLaunchGetSvrVerTask,   ctgHandleGetSvrVerRsp,  ctgDumpSvrVer,     NULL,               NULL},
+  {ctgInitGetTbMetasTask,  ctgLaunchGetTbMetasTask,  ctgHandleGetTbMetasRsp, ctgDumpTbMetasRes, NULL,               NULL},
+  {ctgInitGetTbHashsTask,  ctgLaunchGetTbHashsTask,  ctgHandleGetTbHashsRsp, ctgDumpTbHashsRes, NULL,               NULL},
 };
 
 int32_t ctgMakeAsyncRes(SCtgJob *pJob) {

source/libs/catalog/src/ctgCache.c

@@ -242,7 +242,6 @@ int32_t ctgAcquireTbMetaFromCache(SCatalog* pCtg, char *dbFName, char* tbName, S
     goto _return;
   }
   
-  int32_t sz = 0;
   pCache = taosHashAcquire(dbCache->tbCache, tbName, strlen(tbName));
   if (NULL == pCache) {
     ctgDebug("tb %s not in cache, dbFName:%s", tbName, dbFName);
@@ -282,7 +281,6 @@ int32_t ctgAcquireStbMetaFromCache(SCatalog* pCtg, char *dbFName, uint64_t suid,
     goto _return;
   }
   
-  int32_t sz = 0;
   char* stName = taosHashAcquire(dbCache->stbCache, &suid, sizeof(suid));
   if (NULL == stName) {
     ctgDebug("stb 0x%" PRIx64 " not in cache, dbFName:%s", suid, dbFName);
@@ -2152,6 +2150,254 @@ int32_t ctgGetTbMetaFromCache(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgTbMet
   return TSDB_CODE_SUCCESS;
 }
 
+#if 0
+int32_t ctgGetTbMetaBFromCache(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgTbMetasCtx* ctx, SArray** pResList) {
+  int32_t tbNum = taosArrayGetSize(ctx->pNames);
+  SName* fName = taosArrayGet(ctx->pNames, 0);
+  int32_t fIdx = 0;
+  
+  for (int32_t i = 0; i < tbNum; ++i) {
+    SName* pName = taosArrayGet(ctx->pNames, i);
+    SCtgTbMetaCtx nctx = {0};
+    nctx.flag = CTG_FLAG_UNKNOWN_STB;
+    nctx.pName = pName;
+    
+    if (IS_SYS_DBNAME(pName->dbname)) {
+      CTG_FLAG_SET_SYS_DB(nctx.flag);
+    }
+
+    STableMeta *pTableMeta = NULL;
+    CTG_ERR_RET(ctgReadTbMetaFromCache(pCtg, &nctx, &pTableMeta));
+    SMetaRes res = {0};
+    
+    if (pTableMeta) {
+      if (CTG_FLAG_MATCH_STB(nctx.flag, pTableMeta->tableType) &&
+          ((!CTG_FLAG_IS_FORCE_UPDATE(nctx.flag)) || (CTG_FLAG_IS_SYS_DB(nctx.flag)))) {
+        res.pRes = pTableMeta;
+      } else {
+        taosMemoryFreeClear(pTableMeta);
+      }
+    }
+
+    if (NULL == res.pRes) {
+      if (NULL == ctx->pFetchs) {
+        ctx->pFetchs = taosArrayInit(tbNum, sizeof(SCtgFetch));
+      }
+      
+      if (CTG_FLAG_IS_UNKNOWN_STB(nctx.flag)) {
+        CTG_FLAG_SET_STB(nctx.flag, nctx.tbInfo.tbType);
+      }
+
+      SCtgFetch fetch = {0};
+      fetch.tbIdx = i;
+      fetch.fetchIdx = fIdx++;
+      fetch.flag = nctx.flag;
+
+      taosArrayPush(ctx->pFetchs, &fetch);
+    }
+    
+    taosArrayPush(ctx->pResList, &res);
+  }
+
+  if (NULL == ctx->pFetchs) {
+    TSWAP(*pResList, ctx->pResList);
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+#endif
+
+int32_t ctgGetTbMetasFromCache(SCatalog* pCtg, SRequestConnInfo *pConn, SCtgTbMetasCtx* ctx, int32_t dbIdx, int32_t *fetchIdx, int32_t baseResIdx, SArray* pList) {
+  int32_t tbNum = taosArrayGetSize(pList);
+  SName* pName = taosArrayGet(pList, 0);
+  char dbFName[TSDB_DB_FNAME_LEN] = {0};
+  int32_t flag = CTG_FLAG_UNKNOWN_STB;
+  uint64_t lastSuid = 0;
+  STableMeta* lastTableMeta = NULL;
+  
+  if (IS_SYS_DBNAME(pName->dbname)) {
+    CTG_FLAG_SET_SYS_DB(flag);
+    strcpy(dbFName, pName->dbname);
+  } else {
+    tNameGetFullDbName(pName, dbFName);
+  }
+
+  SCtgDBCache *dbCache = NULL;
+  SCtgTbCache* pCache = NULL;
+  ctgAcquireDBCache(pCtg, dbFName, &dbCache);
+  
+  if (NULL == dbCache) {
+    ctgDebug("db %s not in cache", dbFName);
+    for (int32_t i = 0; i < tbNum; ++i) {
+      ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
+      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+    }
+
+    return TSDB_CODE_SUCCESS;
+  }
+
+  for (int32_t i = 0; i < tbNum; ++i) {
+    SName* pName = taosArrayGet(pList, i);
+
+    pCache = taosHashAcquire(dbCache->tbCache, pName->tname, strlen(pName->tname));
+    if (NULL == pCache) {
+      ctgDebug("tb %s not in cache, dbFName:%s", pName->tname, dbFName);
+      ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
+      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+      
+      continue;
+    }
+
+    CTG_LOCK(CTG_READ, &pCache->metaLock);
+    if (NULL == pCache->pMeta) {
+      ctgDebug("tb %s meta not in cache, dbFName:%s", pName->tname, dbFName);
+      ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
+      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+      
+      continue;
+    }
+
+    STableMeta* tbMeta = pCache->pMeta;
+
+    SCtgTbMetaCtx nctx = {0};
+    nctx.flag = flag;
+    nctx.tbInfo.inCache = true;
+    nctx.tbInfo.dbId = dbCache->dbId;
+    nctx.tbInfo.suid = tbMeta->suid;
+    nctx.tbInfo.tbType = tbMeta->tableType;
+
+    SMetaRes res = {0};
+    STableMeta* pTableMeta = NULL;
+    if (tbMeta->tableType != TSDB_CHILD_TABLE) {
+      int32_t metaSize = CTG_META_SIZE(tbMeta);
+      pTableMeta = taosMemoryCalloc(1, metaSize);
+      if (NULL == pTableMeta) {
+        ctgReleaseTbMetaToCache(pCtg, dbCache, pCache);
+        CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+      }
+    
+      memcpy(pTableMeta, tbMeta, metaSize);
+      
+      CTG_UNLOCK(CTG_READ, &pCache->metaLock);
+      taosHashRelease(dbCache->tbCache, pCache); 
+      
+      ctgDebug("Got tb %s meta from cache, type:%d, dbFName:%s", pName->tname, tbMeta->tableType, dbFName);
+      
+      res.pRes = pTableMeta;
+      taosArrayPush(ctx->pResList, &res);
+
+      continue;
+    }
+    
+    // PROCESS FOR CHILD TABLE
+
+    if (lastSuid && tbMeta->suid == lastSuid && lastTableMeta) {
+      cloneTableMeta(lastTableMeta, &pTableMeta);
+      memcpy(pTableMeta, tbMeta, sizeof(SCTableMeta));
+
+      CTG_UNLOCK(CTG_READ, &pCache->metaLock);
+      taosHashRelease(dbCache->tbCache, pCache); 
+      
+      ctgDebug("Got tb %s meta from cache, type:%d, dbFName:%s", pName->tname, tbMeta->tableType, dbFName);
+      
+      res.pRes = pTableMeta;
+      taosArrayPush(ctx->pResList, &res);
+      
+      continue;
+    }
+    
+    int32_t metaSize = sizeof(SCTableMeta);
+    pTableMeta = taosMemoryCalloc(1, metaSize);
+    if (NULL == pTableMeta) {
+      ctgReleaseTbMetaToCache(pCtg, dbCache, pCache);    
+      CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+    }
+    
+    memcpy(pTableMeta, tbMeta, metaSize);
+    
+    CTG_UNLOCK(CTG_READ, &pCache->metaLock);
+    taosHashRelease(dbCache->tbCache, pCache); 
+    
+    ctgDebug("Got ctb %s meta from cache, will continue to get its stb meta, type:%d, dbFName:%s", 
+             pName->tname, nctx.tbInfo.tbType, dbFName);
+    
+    char* stName = taosHashAcquire(dbCache->stbCache, &pTableMeta->suid, sizeof(pTableMeta->suid));
+    if (NULL == stName) {
+      ctgDebug("stb 0x%" PRIx64 " not in cache, dbFName:%s", pTableMeta->suid, dbFName);
+      ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
+      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+
+      taosMemoryFreeClear(pTableMeta);
+      continue;
+    }
+
+    pCache = taosHashAcquire(dbCache->tbCache, stName, strlen(stName));
+    if (NULL == pCache) {
+      ctgDebug("stb 0x%" PRIx64 " name %s not in cache, dbFName:%s", pTableMeta->suid, stName, dbFName);
+      taosHashRelease(dbCache->stbCache, stName);
+      
+      ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
+      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+
+      taosMemoryFreeClear(pTableMeta);      
+      continue;
+    }
+
+    taosHashRelease(dbCache->stbCache, stName);
+
+    CTG_LOCK(CTG_READ, &pCache->metaLock);
+    if (NULL == pCache->pMeta) {
+      ctgDebug("stb 0x%" PRIx64 " meta not in cache, dbFName:%s", pTableMeta->suid, dbFName);
+      CTG_UNLOCK(CTG_READ, &pCache->metaLock);
+      taosHashRelease(dbCache->tbCache, pCache); 
+      
+      ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
+      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+
+      taosMemoryFreeClear(pTableMeta);
+
+      continue;
+    }
+    
+    STableMeta* stbMeta = pCache->pMeta;
+    if (stbMeta->suid != nctx.tbInfo.suid) {    
+      CTG_UNLOCK(CTG_READ, &pCache->metaLock);
+      taosHashRelease(dbCache->tbCache, pCache); 
+      
+      ctgError("stb suid 0x%" PRIx64 " in stbCache mis-match, expected suid 0x%"PRIx64 , stbMeta->suid, nctx.tbInfo.suid);
+      
+      ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
+      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+
+      taosMemoryFreeClear(pTableMeta);
+
+      continue;
+    }
+    
+    metaSize = CTG_META_SIZE(stbMeta);
+    pTableMeta = taosMemoryRealloc(pTableMeta, metaSize);
+    if (NULL == pTableMeta) {    
+      ctgReleaseTbMetaToCache(pCtg, dbCache, pCache);
+      CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+    }
+    
+    memcpy(&pTableMeta->sversion, &stbMeta->sversion, metaSize - sizeof(SCTableMeta));
+    
+    CTG_UNLOCK(CTG_READ, &pCache->metaLock);
+    taosHashRelease(dbCache->tbCache, pCache); 
+    
+    res.pRes = pTableMeta;
+    taosArrayPush(ctx->pResList, &res);
+
+    lastSuid = pTableMeta->suid;
+    lastTableMeta = pTableMeta;
+  }
+
+  ctgReleaseDBCache(pCtg, dbCache);
+  
+  return TSDB_CODE_SUCCESS;
+}
+
 
 int32_t ctgRemoveTbMetaFromCache(SCatalog* pCtg, SName* pTableName, bool syncReq) {
   int32_t       code = 0;

source/libs/catalog/src/ctgRemote.c

@@ -22,9 +22,8 @@
 
 int32_t ctgHandleBatchRsp(SCtgJob* pJob, SCtgTaskCallbackParam* cbParam, SDataBuf* pMsg, int32_t rspCode) {
   int32_t   code = 0;
-  SArray*   pTaskId = cbParam->taskId;
   SCatalog* pCtg = pJob->pCtg;
-  int32_t   taskNum = taosArrayGetSize(pTaskId);
+  int32_t   taskNum = taosArrayGetSize(cbParam->taskId);
   SDataBuf  taskMsg = *pMsg;
   int32_t   offset = 0;
   int32_t msgNum = (TSDB_CODE_SUCCESS == rspCode && pMsg->pData && (pMsg->len > 0)) ? ntohl(*(int32_t*)pMsg->pData) : 0;
@@ -42,11 +41,14 @@ int32_t ctgHandleBatchRsp(SCtgJob* pJob, SCtgTaskCallbackParam* cbParam, SDataBu
   }
 
   for (int32_t i = 0; i < taskNum; ++i) {
-    int32_t*  taskId = taosArrayGet(pTaskId, i);
+    int32_t*  taskId = taosArrayGet(cbParam->taskId, i);
+    int32_t*  msgIdx = taosArrayGet(cbParam->msgIdx, i);
     SCtgTask* pTask = taosArrayGet(pJob->pTasks, *taskId);
     if (msgNum > 0) {
       rsp.reqType = ntohl(*(int32_t*)((char*)pMsg->pData + offset));
       offset += sizeof(rsp.reqType);
+      rsp.msgIdx = ntohl(*(int32_t*)((char*)pMsg->pData + offset));
+      offset += sizeof(rsp.msgIdx);
       rsp.msgLen = ntohl(*(int32_t*)((char*)pMsg->pData + offset));
       offset += sizeof(rsp.msgLen);
       rsp.rspCode = ntohl(*(int32_t*)((char*)pMsg->pData + offset));
@@ -57,7 +59,10 @@ int32_t ctgHandleBatchRsp(SCtgJob* pJob, SCtgTaskCallbackParam* cbParam, SDataBu
       taskMsg.msgType = rsp.reqType;
       taskMsg.pData = rsp.msg;
       taskMsg.len = rsp.msgLen;
+
+      ASSERT(rsp.msgIdx == *msgIdx);
     } else {
+      rsp.msgIdx = *msgIdx;
       rsp.reqType = -1;
       taskMsg.msgType = -1;
       taskMsg.pData = NULL;
@@ -66,10 +71,13 @@ int32_t ctgHandleBatchRsp(SCtgJob* pJob, SCtgTaskCallbackParam* cbParam, SDataBu
 
     pTask->pBatchs = pBatchs;
     
-    ctgDebug("QID:0x%" PRIx64 " ctg task %d start to handle rsp %s", pJob->queryId, pTask->taskId,
-             TMSG_INFO(taskMsg.msgType + 1));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = rsp.msgIdx;    
+
+    ctgDebug("QID:0x%" PRIx64 " ctg task %d idx %d start to handle rsp %s", pJob->queryId, pTask->taskId, rsp.msgIdx, TMSG_INFO(taskMsg.msgType + 1));
 
-    (*gCtgAsyncFps[pTask->type].handleRspFp)(pTask, rsp.reqType, &taskMsg, (rsp.rspCode ? rsp.rspCode : rspCode));
+    (*gCtgAsyncFps[pTask->type].handleRspFp)(&tReq, rsp.reqType, &taskMsg, (rsp.rspCode ? rsp.rspCode : rspCode));
   }
 
   CTG_ERR_JRET(ctgLaunchBatchs(pJob->pCtg, pJob, pBatchs));
@@ -338,7 +346,10 @@ int32_t ctgHandleMsgCallback(void* param, SDataBuf* pMsg, int32_t rspCode) {
     pTask->pBatchs = pBatchs;
 #endif
 
-    CTG_ERR_JRET((*gCtgAsyncFps[pTask->type].handleRspFp)(pTask, cbParam->reqType, pMsg, rspCode));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;
+    CTG_ERR_JRET((*gCtgAsyncFps[pTask->type].handleRspFp)(&tReq, cbParam->reqType, pMsg, rspCode));
 
 #if CTG_BATCH_FETCH
     CTG_ERR_JRET(ctgLaunchBatchs(pJob->pCtg, pJob, pBatchs));
@@ -356,7 +367,7 @@ _return:
   CTG_API_LEAVE(code);
 }
 
-int32_t ctgMakeMsgSendInfo(SCtgJob* pJob, SArray* pTaskId, int32_t batchId, int32_t msgType,
+int32_t ctgMakeMsgSendInfo(SCtgJob* pJob, SArray* pTaskId, int32_t batchId, SArray* pMsgIdx, int32_t msgType,
                            SMsgSendInfo** pMsgSendInfo) {
   int32_t       code = 0;
   SMsgSendInfo* msgSendInfo = taosMemoryCalloc(1, sizeof(SMsgSendInfo));
@@ -376,6 +387,7 @@ int32_t ctgMakeMsgSendInfo(SCtgJob* pJob, SArray* pTaskId, int32_t batchId, int3
   param->refId = pJob->refId;
   param->taskId = pTaskId;
   param->batchId = batchId;
+  param->msgIdx = pMsgIdx;
 
   msgSendInfo->param = param;
   msgSendInfo->paramFreeFp = ctgFreeMsgSendParam;
@@ -394,10 +406,10 @@ _return:
 }
 
 int32_t ctgAsyncSendMsg(SCatalog* pCtg, SRequestConnInfo* pConn, SCtgJob* pJob, SArray* pTaskId, int32_t batchId,
-                        char* dbFName, int32_t vgId, int32_t msgType, void* msg, uint32_t msgSize) {
+                        SArray* pMsgIdx, char* dbFName, int32_t vgId, int32_t msgType, void* msg, uint32_t msgSize) {
   int32_t       code = 0;
   SMsgSendInfo* pMsgSendInfo = NULL;
-  CTG_ERR_JRET(ctgMakeMsgSendInfo(pJob, pTaskId, batchId, msgType, &pMsgSendInfo));
+  CTG_ERR_JRET(ctgMakeMsgSendInfo(pJob, pTaskId, batchId, pMsgIdx, msgType, &pMsgSendInfo));
 
   ctgUpdateSendTargetInfo(pMsgSendInfo, msgType, dbFName, vgId);
 
@@ -428,25 +440,27 @@ _return:
   CTG_RET(code);
 }
 
-int32_t ctgAddBatch(SCatalog* pCtg, int32_t vgId, SRequestConnInfo* pConn, SCtgTask* pTask, int32_t msgType, void* msg,
+int32_t ctgAddBatch(SCatalog* pCtg, int32_t vgId, SRequestConnInfo* pConn, SCtgTaskReq* tReq, int32_t msgType, void* msg,
                     uint32_t msgSize) {
   int32_t    code = 0;
+  SCtgTask*  pTask = tReq->pTask;
   SHashObj*  pBatchs = pTask->pBatchs;
   SCtgJob*   pJob = pTask->pJob;
   SCtgBatch* pBatch = taosHashGet(pBatchs, &vgId, sizeof(vgId));
-  int32_t    taskNum = taosArrayGetSize(pTask->pJob->pTasks);
   SCtgBatch newBatch = {0};
   SBatchMsg req = {0};
   
   if (NULL == pBatch) {
-    newBatch.pMsgs = taosArrayInit(taskNum, sizeof(SBatchMsg));
-    newBatch.pTaskIds = taosArrayInit(taskNum, sizeof(int32_t));
-    if (NULL == newBatch.pMsgs || NULL == newBatch.pTaskIds) {
+    newBatch.pMsgs = taosArrayInit(pJob->subTaskNum, sizeof(SBatchMsg));
+    newBatch.pTaskIds = taosArrayInit(pJob->subTaskNum, sizeof(int32_t));
+    newBatch.pMsgIdxs = taosArrayInit(pJob->subTaskNum, sizeof(int32_t));
+    if (NULL == newBatch.pMsgs || NULL == newBatch.pTaskIds || NULL == newBatch.pMsgIdxs) {
       CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY);
     }
 
     newBatch.conn = *pConn;
 
+    req.msgIdx = tReq->msgIdx;
     req.msgType = msgType;
     req.msgLen = msgSize;
     req.msg = msg;
@@ -456,19 +470,31 @@ int32_t ctgAddBatch(SCatalog* pCtg, int32_t vgId, SRequestConnInfo* pConn, SCtgT
     if (NULL == taosArrayPush(newBatch.pTaskIds, &pTask->taskId)) {
       CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY);
     }
+    if (NULL == taosArrayPush(newBatch.pMsgIdxs, &req.msgIdx)) {
+      CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY);
+    }
     newBatch.msgSize = sizeof(SBatchReq) + sizeof(req) + msgSize - POINTER_BYTES;
 
     if (vgId > 0) {
+      SName* pName = NULL;
       if (TDMT_VND_TABLE_CFG == msgType) {
         SCtgTbCfgCtx* ctx = (SCtgTbCfgCtx*)pTask->taskCtx;
-        tNameGetFullDbName(ctx->pName, newBatch.dbFName);
+        pName = ctx->pName;
       } else if (TDMT_VND_TABLE_META == msgType) {
+        if (CTG_TASK_GET_TB_META_BATCH == pTask->type) {
+          SCtgTbMetasCtx* ctx = (SCtgTbMetasCtx*)pTask->taskCtx;
+          SCtgFetch* fetch = taosArrayGet(ctx->pFetchs, tReq->msgIdx);
+          pName = ctgGetFetchName(ctx->pNames, fetch);
+        } else {
           SCtgTbMetaCtx* ctx = (SCtgTbMetaCtx*)pTask->taskCtx;
-        tNameGetFullDbName(ctx->pName, newBatch.dbFName);
+          pName = ctx->pName;
+        }
       } else {
         ctgError("invalid vnode msgType %d", msgType);
         CTG_ERR_JRET(TSDB_CODE_APP_ERROR);
       }
+
+      tNameGetFullDbName(pName, newBatch.dbFName);
     }
 
     newBatch.msgType = (vgId > 0) ? TDMT_VND_BATCH_META : TDMT_MND_BATCH_META;
@@ -484,6 +510,7 @@ int32_t ctgAddBatch(SCatalog* pCtg, int32_t vgId, SRequestConnInfo* pConn, SCtgT
     return TSDB_CODE_SUCCESS;
   }
 
+  req.msgIdx = tReq->msgIdx;
   req.msgType = msgType;
   req.msgLen = msgSize;
   req.msg = msg;
@@ -493,19 +520,32 @@ int32_t ctgAddBatch(SCatalog* pCtg, int32_t vgId, SRequestConnInfo* pConn, SCtgT
   if (NULL == taosArrayPush(pBatch->pTaskIds, &pTask->taskId)) {
     CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY);
   }
+  if (NULL == taosArrayPush(pBatch->pMsgIdxs, &req.msgIdx)) {
+    CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY);
+  }
+  
   pBatch->msgSize += sizeof(req) + msgSize - POINTER_BYTES;
 
   if (vgId > 0) {
+    SName* pName = NULL;
     if (TDMT_VND_TABLE_CFG == msgType) {
       SCtgTbCfgCtx* ctx = (SCtgTbCfgCtx*)pTask->taskCtx;
-      tNameGetFullDbName(ctx->pName, newBatch.dbFName);
+      pName = ctx->pName;
     } else if (TDMT_VND_TABLE_META == msgType) {
+      if (CTG_TASK_GET_TB_META_BATCH == pTask->type) {
+        SCtgTbMetasCtx* ctx = (SCtgTbMetasCtx*)pTask->taskCtx;
+        SCtgFetch* fetch = taosArrayGet(ctx->pFetchs, tReq->msgIdx);
+        pName = ctgGetFetchName(ctx->pNames, fetch);
+      } else {
         SCtgTbMetaCtx* ctx = (SCtgTbMetaCtx*)pTask->taskCtx;
-      tNameGetFullDbName(ctx->pName, newBatch.dbFName);
+        pName = ctx->pName;
+      }
     } else {
       ctgError("invalid vnode msgType %d", msgType);
       CTG_ERR_JRET(TSDB_CODE_APP_ERROR);
     }
+
+    tNameGetFullDbName(pName, newBatch.dbFName);    
   }
 
   ctgDebug("task %d %s req added to batch %d, target vgId %d", pTask->taskId, TMSG_INFO(msgType), pBatch->batchId,
@@ -537,6 +577,8 @@ int32_t ctgBuildBatchReqMsg(SCtgBatch* pBatch, int32_t vgId, void** msg) {
 
   for (int32_t i = 0; i < num; ++i) {
     SBatchMsg* pReq = taosArrayGet(pBatch->pMsgs, i);
+    *(int32_t*)((char*)(*msg) + offset) = htonl(pReq->msgIdx);
+    offset += sizeof(pReq->msgIdx);    
     *(int32_t*)((char*)(*msg) + offset) = htonl(pReq->msgType);
     offset += sizeof(pReq->msgType);
     *(int32_t*)((char*)(*msg) + offset) = htonl(pReq->msgLen);
@@ -564,8 +606,8 @@ int32_t ctgLaunchBatchs(SCatalog* pCtg, SCtgJob* pJob, SHashObj* pBatchs) {
     ctgDebug("QID:0x%" PRIx64 " ctg start to launch batch %d", pJob->queryId, pBatch->batchId);
 
     CTG_ERR_JRET(ctgBuildBatchReqMsg(pBatch, *vgId, &msg));
-    code = ctgAsyncSendMsg(pCtg, &pBatch->conn, pJob, pBatch->pTaskIds, pBatch->batchId, pBatch->dbFName, *vgId,
-                           pBatch->msgType, msg, pBatch->msgSize);
+    code = ctgAsyncSendMsg(pCtg, &pBatch->conn, pJob, pBatch->pTaskIds, pBatch->batchId, pBatch->pMsgIdxs, 
+                           pBatch->dbFName, *vgId, pBatch->msgType, msg, pBatch->msgSize);
     pBatch->pTaskIds = NULL;
     CTG_ERR_JRET(code);
 
@@ -603,10 +645,14 @@ int32_t ctgGetQnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SArray
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, NULL));
+
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, pOut, NULL));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -614,7 +660,7 @@ int32_t ctgGetQnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SArray
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -649,10 +695,13 @@ int32_t ctgGetDnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SArray
   }
 
   if (pTask) {
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, NULL, NULL));
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, NULL, NULL));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -660,7 +709,7 @@ int32_t ctgGetDnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SArray
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -681,10 +730,11 @@ int32_t ctgGetDnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SArray
 }
 
 int32_t ctgGetDBVgInfoFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SBuildUseDBInput* input, SUseDbOutput* out,
-                                SCtgTask* pTask) {
+                                SCtgTaskReq* tReq) {
   char*   msg = NULL;
   int32_t msgLen = 0;
   int32_t reqType = TDMT_MND_USE_DB;
+  SCtgTask* pTask = tReq ? tReq->pTask : NULL;
   void* (*mallocFp)(int32_t) = pTask ? taosMemoryMalloc : rpcMallocCont;
 
   ctgDebug("try to get db vgInfo from mnode, dbFName:%s", input->db);
@@ -700,10 +750,11 @@ int32_t ctgGetDBVgInfoFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SBuildU
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, input->db));
+
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, tReq->msgIdx), reqType, pOut, input->db));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -711,7 +762,7 @@ int32_t ctgGetDBVgInfoFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SBuildU
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -751,10 +802,14 @@ int32_t ctgGetDBCfgFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const char
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, (char*)dbFName));
+
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, pOut, (char*)dbFName));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -762,7 +817,7 @@ int32_t ctgGetDBCfgFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const char
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -802,10 +857,14 @@ int32_t ctgGetIndexInfoFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, (char*)indexName));
+
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, pOut, (char*)indexName));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -813,7 +872,7 @@ int32_t ctgGetIndexInfoFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -856,10 +915,13 @@ int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SName* n
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
     
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, (char*)tbFName));
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, pOut, (char*)tbFName));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -867,7 +929,7 @@ int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, SName* n
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -907,10 +969,14 @@ int32_t ctgGetUdfInfoFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const ch
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, (char*)funcName));
+
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, pOut, (char*)funcName));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -918,7 +984,7 @@ int32_t ctgGetUdfInfoFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const ch
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -958,10 +1024,14 @@ int32_t ctgGetUserDbAuthFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, (char*)user));
+
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, pOut, (char*)user));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -969,7 +1039,7 @@ int32_t ctgGetUserDbAuthFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -990,7 +1060,8 @@ int32_t ctgGetUserDbAuthFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const
 }
 
 int32_t ctgGetTbMetaFromMnodeImpl(SCatalog* pCtg, SRequestConnInfo* pConn, char* dbFName, char* tbName,
-                                  STableMetaOutput* out, SCtgTask* pTask) {
+                                  STableMetaOutput* out, SCtgTaskReq* tReq) {
+  SCtgTask *pTask = tReq ? tReq->pTask : NULL;                                  
   SBuildTableInput bInput = {.vgId = 0, .dbFName = dbFName, .tbName = tbName};
   char*            msg = NULL;
   SEpSet*          pVnodeEpSet = NULL;
@@ -1013,9 +1084,11 @@ int32_t ctgGetTbMetaFromMnodeImpl(SCatalog* pCtg, SRequestConnInfo* pConn, char*
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, tbFName));
+
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, tReq->msgIdx), reqType, pOut, tbFName));
+    
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -1023,7 +1096,7 @@ int32_t ctgGetTbMetaFromMnodeImpl(SCatalog* pCtg, SRequestConnInfo* pConn, char*
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -1044,15 +1117,16 @@ int32_t ctgGetTbMetaFromMnodeImpl(SCatalog* pCtg, SRequestConnInfo* pConn, char*
 }
 
 int32_t ctgGetTbMetaFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const SName* pTableName, STableMetaOutput* out,
-                              SCtgTask* pTask) {
+                              SCtgTaskReq* tReq) {
   char dbFName[TSDB_DB_FNAME_LEN];
   tNameGetFullDbName(pTableName, dbFName);
 
-  return ctgGetTbMetaFromMnodeImpl(pCtg, pConn, dbFName, (char*)pTableName->tname, out, pTask);
+  return ctgGetTbMetaFromMnodeImpl(pCtg, pConn, dbFName, (char*)pTableName->tname, out, tReq);
 }
 
 int32_t ctgGetTbMetaFromVnode(SCatalog* pCtg, SRequestConnInfo* pConn, const SName* pTableName, SVgroupInfo* vgroupInfo,
-                              STableMetaOutput* out, SCtgTask* pTask) {
+                              STableMetaOutput* out, SCtgTaskReq* tReq) {
+  SCtgTask *pTask = tReq ? tReq->pTask : NULL;
   char dbFName[TSDB_DB_FNAME_LEN];
   tNameGetFullDbName(pTableName, dbFName);
   int32_t reqType = TDMT_VND_TABLE_META;
@@ -1080,15 +1154,16 @@ int32_t ctgGetTbMetaFromVnode(SCatalog* pCtg, SRequestConnInfo* pConn, const SNa
     if (NULL == pOut) {
       CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
     }
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, tbFName));
 
     SRequestConnInfo vConn = {.pTrans = pConn->pTrans,
                               .requestId = pConn->requestId,
                               .requestObjRefId = pConn->requestObjRefId,
                               .mgmtEps = vgroupInfo->epSet};
 
+  CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, tReq->msgIdx), reqType, pOut, tbFName));
+
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, vgroupInfo->vgId, &vConn, pTask, reqType, msg, msgLen));
+    CTG_RET(ctgAddBatch(pCtg, vgroupInfo->vgId, &vConn, tReq, reqType, msg, msgLen));
 #else
     SCtgTbMetaCtx* ctx = (SCtgTbMetaCtx*)pTask->taskCtx;
     char           dbFName[TSDB_DB_FNAME_LEN];
@@ -1099,7 +1174,7 @@ int32_t ctgGetTbMetaFromVnode(SCatalog* pCtg, SRequestConnInfo* pConn, const SNa
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, &vConn, pTask->pJob, pTaskId, -1, dbFName, ctx->vgId, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, &vConn, pTask->pJob, pTaskId, -1, NULL, dbFName, ctx->vgId, reqType, msg, msgLen));
 #endif
   }
 
@@ -1142,14 +1217,17 @@ int32_t ctgGetTableCfgFromVnode(SCatalog* pCtg, SRequestConnInfo* pConn, const S
   }
 
   if (pTask) {
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, NULL, (char*)tbFName));
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, NULL, (char*)tbFName));
 
     SRequestConnInfo vConn = {.pTrans = pConn->pTrans,
                               .requestId = pConn->requestId,
                               .requestObjRefId = pConn->requestObjRefId,
                               .mgmtEps = vgroupInfo->epSet};
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, vgroupInfo->vgId, &vConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, vgroupInfo->vgId, &vConn, &tReq, reqType, msg, msgLen));
 #else
     SCtgTbCfgCtx* ctx = (SCtgTbCfgCtx*)pTask->taskCtx;
     char          dbFName[TSDB_DB_FNAME_LEN];
@@ -1160,7 +1238,7 @@ int32_t ctgGetTableCfgFromVnode(SCatalog* pCtg, SRequestConnInfo* pConn, const S
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, &vConn, pTask->pJob, pTaskId, -1, dbFName, ctx->pVgInfo->vgId, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, &vConn, pTask->pJob, pTaskId, -1, NULL, dbFName, ctx->pVgInfo->vgId, reqType, msg, msgLen));
 #endif
   }
 
@@ -1201,9 +1279,13 @@ int32_t ctgGetTableCfgFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const S
   }
 
   if (pTask) {
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, NULL, (char*)tbFName));
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, NULL, (char*)tbFName));
+
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -1211,7 +1293,7 @@ int32_t ctgGetTableCfgFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, const S
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 
@@ -1246,10 +1328,13 @@ int32_t ctgGetSvrVerFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, char** ou
   }
 
   if (pTask) {
-    CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, NULL, NULL));
+    CTG_ERR_RET(ctgUpdateMsgCtx(CTG_GET_TASK_MSGCTX(pTask, -1), reqType, NULL, NULL));
 
 #if CTG_BATCH_FETCH
-    CTG_RET(ctgAddBatch(pCtg, 0, pConn, pTask, reqType, msg, msgLen));
+    SCtgTaskReq tReq;
+    tReq.pTask = pTask;
+    tReq.msgIdx = -1;    
+    CTG_RET(ctgAddBatch(pCtg, 0, pConn, &tReq, reqType, msg, msgLen));
 #else
     SArray* pTaskId = taosArrayInit(1, sizeof(int32_t));
     if (NULL == pTaskId) {
@@ -1257,7 +1342,7 @@ int32_t ctgGetSvrVerFromMnode(SCatalog* pCtg, SRequestConnInfo* pConn, char** ou
     }
     taosArrayPush(pTaskId, &pTask->taskId);
 
-    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, 0, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask->pJob, pTaskId, -1, NULL, NULL, 0, reqType, msg, msgLen));
 #endif
   }
 

source/libs/catalog/src/ctgUtil.c

@@ -26,6 +26,7 @@ void ctgFreeMsgSendParam(void* param) {
 
   SCtgTaskCallbackParam* pParam = (SCtgTaskCallbackParam*)param;
   taosArrayDestroy(pParam->taskId);
+  taosArrayDestroy(pParam->msgIdx);
 
   taosMemoryFree(param);
 }
@@ -88,6 +89,10 @@ char *ctgTaskTypeStr(CTG_TASK_TYPE type) {
       return "[get user]";
     case CTG_TASK_GET_SVR_VER:
       return "[get svr ver]";
+    case CTG_TASK_GET_TB_META_BATCH:
+      return "[bget table meta]";
+    case CTG_TASK_GET_TB_HASH_BATCH:
+      return "[bget table hash]";
     default:
       return "unknown";
   }
@@ -460,6 +465,25 @@ void ctgResetTbMetaTask(SCtgTask* pTask) {
   taosMemoryFreeClear(pTask->res);
 }
 
+void ctgFreeBatchMeta(void* meta) {
+  if (NULL == meta) {
+    return;
+  }
+  
+  SMetaRes* pRes = (SMetaRes*)meta;
+  taosMemoryFreeClear(pRes->pRes);
+}
+
+void ctgFreeBatchHash(void* hash) {
+  if (NULL == hash) {
+    return;
+  }
+  
+  SMetaRes* pRes = (SMetaRes*)hash;
+  taosMemoryFreeClear(pRes->pRes);
+}
+
+
 void ctgFreeTaskRes(CTG_TASK_TYPE type, void **pRes) {
   switch (type) {
     case CTG_TASK_GET_QNODE:
@@ -500,6 +524,24 @@ void ctgFreeTaskRes(CTG_TASK_TYPE type, void **pRes) {
       taosMemoryFreeClear(*pRes);
       break;
     }
+    case CTG_TASK_GET_TB_META_BATCH: {
+      SArray* pArray = (SArray*)*pRes;
+      int32_t num = taosArrayGetSize(pArray);
+      for (int32_t i = 0; i < num; ++i) {
+        ctgFreeBatchMeta(taosArrayGet(pArray, i));
+      }
+      *pRes = NULL; // no need to free it
+      break;
+    }
+    case CTG_TASK_GET_TB_HASH_BATCH: {
+      SArray* pArray = (SArray*)*pRes;
+      int32_t num = taosArrayGetSize(pArray);
+      for (int32_t i = 0; i < num; ++i) {
+        ctgFreeBatchHash(taosArrayGet(pArray, i));
+      }
+      *pRes = NULL; // no need to free it
+      break;
+    }    
     default:
       qError("invalid task type %d", type);
       break;
@@ -554,6 +596,16 @@ void ctgFreeSubTaskRes(CTG_TASK_TYPE type, void **pRes) {
       taosMemoryFreeClear(*pRes);
       break;
     }
+    case CTG_TASK_GET_TB_META_BATCH: {
+      taosArrayDestroyEx(*pRes, ctgFreeBatchMeta);
+      *pRes = NULL;
+      break;
+    }
+    case CTG_TASK_GET_TB_HASH_BATCH: {
+      taosArrayDestroyEx(*pRes, ctgFreeBatchHash);
+      *pRes = NULL;
+      break;
+    }
     default:
       qError("invalid task type %d", type);
       break;
@@ -583,12 +635,38 @@ void ctgFreeTaskCtx(SCtgTask* pTask) {
       taosMemoryFreeClear(pTask->taskCtx);
       break;
     }
+    case CTG_TASK_GET_TB_META_BATCH: {
+      SCtgTbMetasCtx* taskCtx = (SCtgTbMetasCtx*)pTask->taskCtx;
+      taosArrayDestroyEx(taskCtx->pResList, ctgFreeBatchMeta);
+      taosArrayDestroy(taskCtx->pFetchs);
+      // NO NEED TO FREE pNames
+
+      taosArrayDestroyEx(pTask->msgCtxs, (FDelete)ctgFreeMsgCtx);
+      
+      if (pTask->msgCtx.lastOut) {
+        ctgFreeSTableMetaOutput((STableMetaOutput*)pTask->msgCtx.lastOut);
+        pTask->msgCtx.lastOut = NULL;
+      }
+      taosMemoryFreeClear(pTask->taskCtx);
+      break;
+    }
     case CTG_TASK_GET_TB_HASH: {
       SCtgTbHashCtx* taskCtx = (SCtgTbHashCtx*)pTask->taskCtx;
       taosMemoryFreeClear(taskCtx->pName);
       taosMemoryFreeClear(pTask->taskCtx);      
       break;
     }
+    case CTG_TASK_GET_TB_HASH_BATCH: {
+      SCtgTbHashsCtx* taskCtx = (SCtgTbHashsCtx*)pTask->taskCtx;
+      taosArrayDestroyEx(taskCtx->pResList, ctgFreeBatchHash);
+      taosArrayDestroy(taskCtx->pFetchs);
+      // NO NEED TO FREE pNames
+
+      taosArrayDestroyEx(pTask->msgCtxs, (FDelete)ctgFreeMsgCtx);
+      
+      taosMemoryFreeClear(pTask->taskCtx);
+      break;
+    }    
     case CTG_TASK_GET_TB_INDEX: {
       SCtgTbIndexCtx* taskCtx = (SCtgTbIndexCtx*)pTask->taskCtx;
       taosMemoryFreeClear(taskCtx->pName);
@@ -679,6 +757,23 @@ int32_t ctgUpdateMsgCtx(SCtgMsgCtx* pCtx, int32_t reqType, void* out, char* targ
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t ctgAddMsgCtx(SArray* pCtxs, int32_t reqType, void* out, char* target) {
+  SCtgMsgCtx ctx = {0};
+  
+  ctx.reqType = reqType;
+  ctx.out = out;
+  if (target) {
+    ctx.target = strdup(target);
+    if (NULL == ctx.target) {
+      CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+    }
+  }
+
+  taosArrayPush(pCtxs, &ctx);
+
+  return TSDB_CODE_SUCCESS;
+}
+
 
 int32_t ctgGetHashFunction(int8_t hashMethod, tableNameHashFp *fp) {
   switch (hashMethod) {
@@ -780,6 +875,104 @@ int32_t ctgGetVgInfoFromHashValue(SCatalog *pCtg, SDBVgInfo *dbInfo, const SName
   CTG_RET(code);
 }
 
+int32_t ctgGetVgInfosFromHashValue(SCatalog *pCtg, SCtgTaskReq* tReq, SDBVgInfo *dbInfo, SCtgTbHashsCtx *pCtx, char* dbFName, SArray* pNames, bool update) {
+  int32_t code = 0;
+  SCtgTask* pTask = tReq->pTask;
+  SMetaRes res = {0};
+  int32_t vgNum = taosHashGetSize(dbInfo->vgHash);
+  if (vgNum <= 0) {
+    ctgError("db vgroup cache invalid, db:%s, vgroup number:%d", dbFName, vgNum);
+    CTG_ERR_RET(TSDB_CODE_TSC_DB_NOT_SELECTED);
+  }
+
+  tableNameHashFp fp = NULL;
+  SVgroupInfo *vgInfo = NULL;
+
+  CTG_ERR_RET(ctgGetHashFunction(dbInfo->hashMethod, &fp));
+
+  int32_t tbNum = taosArrayGetSize(pNames);
+
+  if (1 == vgNum) {
+    void *pIter = taosHashIterate(dbInfo->vgHash, NULL);
+    for (int32_t i = 0; i < tbNum; ++i) {
+      vgInfo = taosMemoryMalloc(sizeof(SVgroupInfo));
+      if (NULL == vgInfo) {
+        CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+      }
+
+      *vgInfo = *(SVgroupInfo*)pIter;
+
+      ctgDebug("Got tb hash vgroup, vgId:%d, epNum %d, current %s port %d", vgInfo->vgId, vgInfo->epSet.numOfEps,
+        vgInfo->epSet.eps[vgInfo->epSet.inUse].fqdn, vgInfo->epSet.eps[vgInfo->epSet.inUse].port);
+
+      if (update) {
+        SCtgFetch* pFetch = taosArrayGet(pCtx->pFetchs, tReq->msgIdx);
+        SMetaRes *pRes = taosArrayGet(pCtx->pResList, pFetch->resIdx + i);
+        pRes->pRes = vgInfo;
+      } else {
+        res.pRes = vgInfo;
+        taosArrayPush(pCtx->pResList, &res);
+      }
+    }
+
+    return TSDB_CODE_SUCCESS;
+  }
+
+  char tbFullName[TSDB_TABLE_FNAME_LEN];
+  sprintf(tbFullName, "%s.", dbFName);
+  int32_t offset = strlen(tbFullName);
+  SName* pName = NULL;
+  int32_t tbNameLen = 0;
+  
+  for (int32_t i = 0; i < tbNum; ++i) {
+    pName = taosArrayGet(pNames, i);
+
+    tbNameLen = offset + strlen(pName->tname);
+    strcpy(tbFullName + offset, pName->tname);
+
+    uint32_t hashValue = (*fp)(tbFullName, (uint32_t)tbNameLen);
+
+    void *pIter = taosHashIterate(dbInfo->vgHash, NULL);
+    while (pIter) {
+      vgInfo = pIter;
+      if (hashValue >= vgInfo->hashBegin && hashValue <= vgInfo->hashEnd) {
+        taosHashCancelIterate(dbInfo->vgHash, pIter);
+        break;
+      }
+      
+      pIter = taosHashIterate(dbInfo->vgHash, pIter);
+      vgInfo = NULL;
+    }
+
+    if (NULL == vgInfo) {
+      ctgError("no hash range found for hash value [%u], db:%s, numOfVgId:%d", hashValue, dbFName, taosHashGetSize(dbInfo->vgHash));
+      CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR);
+    }
+
+    SVgroupInfo* pNewVg = taosMemoryMalloc(sizeof(SVgroupInfo));
+    if (NULL == pNewVg) {
+      CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+    }
+
+    *pNewVg = *vgInfo;
+
+    ctgDebug("Got tb %s hash vgroup, vgId:%d, epNum %d, current %s port %d", tbFullName, vgInfo->vgId, vgInfo->epSet.numOfEps,
+             vgInfo->epSet.eps[vgInfo->epSet.inUse].fqdn, vgInfo->epSet.eps[vgInfo->epSet.inUse].port);
+
+    if (update) {
+      SCtgFetch* pFetch = taosArrayGet(pCtx->pFetchs, tReq->msgIdx);
+      SMetaRes *pRes = taosArrayGet(pCtx->pResList, pFetch->resIdx + i);
+      pRes->pRes = pNewVg;
+    } else {
+      res.pRes = pNewVg;
+      taosArrayPush(pCtx->pResList, &res);
+    }    
+  }
+
+  CTG_RET(code);
+}
+
+
 int32_t ctgStbVersionSearchCompare(const void* key1, const void* key2) {
   if (*(uint64_t *)key1 < ((SSTableVersion*)key2)->suid) {
     return -1;
@@ -921,4 +1114,41 @@ int32_t ctgUpdateSendTargetInfo(SMsgSendInfo *pMsgSendInfo, int32_t msgType, cha
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t ctgGetTablesReqNum(SArray *pList) {
+  if (NULL == pList) {
+    return 0;
+  }
+
+  int32_t total = 0;
+  int32_t n = taosArrayGetSize(pList);
+  for (int32_t i = 0; i < n; ++i) {
+    STablesReq *pReq = taosArrayGet(pList, i);
+    total += taosArrayGetSize(pReq->pTables);
+  }
+
+  return total;
+}
+
+int32_t ctgAddFetch(SArray** pFetchs, int32_t dbIdx, int32_t tbIdx, int32_t *fetchIdx, int32_t resIdx, int32_t flag) {
+  if (NULL == (*pFetchs)) {
+    *pFetchs = taosArrayInit(CTG_DEFAULT_FETCH_NUM, sizeof(SCtgFetch));
+  }
+  
+  SCtgFetch fetch = {0};
+  fetch.dbIdx = dbIdx;
+  fetch.tbIdx = tbIdx;
+  fetch.fetchIdx = (*fetchIdx)++;
+  fetch.resIdx = resIdx;
+  fetch.flag = flag;
+  
+  taosArrayPush(*pFetchs, &fetch);
+
+  return TSDB_CODE_SUCCESS;
+}
+
+SName* ctgGetFetchName(SArray* pNames, SCtgFetch* pFetch) {
+  STablesReq* pReq = (STablesReq*)taosArrayGet(pNames, pFetch->dbIdx);
+  return (SName*)taosArrayGet(pReq->pTables, pFetch->tbIdx);
+}
+
 

source/libs/executor/inc/executorimpl.h

@@ -737,6 +737,9 @@ typedef struct STimeSliceOperatorInfo {
   SInterval      interval;
   int64_t        current;
   SArray*        pPrevRow;      // SArray<SGroupValue>
+  SArray*        pNextRow;      // SArray<SGroupValue>
+  bool           isPrevRowSet;
+  bool           isNextRowSet;
   int32_t        fillType;      // fill type
   SColumn        tsCol;         // primary timestamp column
   SExprSupp      scalarSup;     // scalar calculation
@@ -987,7 +990,7 @@ int32_t decodeOperator(SOperatorInfo* ops, const char* data, int32_t length);
 
 void    setTaskStatus(SExecTaskInfo* pTaskInfo, int8_t status);
 int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId,
-                               const char* sql, EOPTR_EXEC_MODEL model);
+                               char* sql, EOPTR_EXEC_MODEL model);
 int32_t createDataSinkParam(SDataSinkNode *pNode, void **pParam, qTaskInfo_t* pTaskInfo, SReadHandle* readHandle);
 int32_t getOperatorExplainExecInfo(SOperatorInfo* operatorInfo, SArray* pExecInfoList);
 

source/libs/executor/src/executor.c

@@ -341,7 +341,7 @@ int32_t qGetQueryTableSchemaVersion(qTaskInfo_t tinfo, char* dbName, char* table
 }
 
 int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, SSubplan* pSubplan,
-                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, const char* sql, EOPTR_EXEC_MODEL model) {
+                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, char* sql, EOPTR_EXEC_MODEL model) {
   assert(pSubplan != NULL);
   SExecTaskInfo** pTask = (SExecTaskInfo**)pTaskInfo;
 

source/libs/executor/src/executorimpl.c

@@ -1437,7 +1437,8 @@ static void setExecutionContext(SOperatorInfo* pOperator, int32_t numOfOutput, u
   pAggInfo->groupId = groupId;
 }
 
-static void doUpdateNumOfRows(SResultRow* pRow, int32_t numOfExprs, const int32_t* rowCellOffset) {
+static void doUpdateNumOfRows(SqlFunctionCtx* pCtx, SResultRow* pRow, int32_t numOfExprs, const int32_t* rowCellOffset) {
+  bool returnNotNull = false;
   for (int32_t j = 0; j < numOfExprs; ++j) {
     struct SResultRowEntryInfo* pResInfo = getResultEntryInfo(pRow, j, rowCellOffset);
     if (!isRowEntryInitialized(pResInfo)) {
@@ -1447,6 +1448,15 @@ static void doUpdateNumOfRows(SResultRow* pRow, int32_t numOfExprs, const int32_
     if (pRow->numOfRows < pResInfo->numOfRes) {
       pRow->numOfRows = pResInfo->numOfRes;
     }
+
+    if (fmIsNotNullOutputFunc(pCtx[j].functionId)) {
+      returnNotNull = true;
+    }
+  }
+  // if all expr skips all blocks, e.g. all null inputs for max function, output one row in final result.
+  //  except for first/last, which require not null output, output no rows
+  if (pRow->numOfRows == 0 && !returnNotNull) {
+    pRow->numOfRows = 1;
   }
 }
 
@@ -1458,7 +1468,7 @@ int32_t finalizeResultRowIntoResultDataBlock(SDiskbasedBuf* pBuf, SResultRowPosi
   SFilePage*  page = getBufPage(pBuf, resultRowPosition->pageId);
   SResultRow* pRow = (SResultRow*)((char*)page + resultRowPosition->offset);
 
-  doUpdateNumOfRows(pRow, numOfExprs, rowCellOffset);
+  doUpdateNumOfRows(pCtx, pRow, numOfExprs, rowCellOffset);
   if (pRow->numOfRows == 0) {
     releaseBufPage(pBuf, page);
     return 0;
@@ -1514,7 +1524,7 @@ int32_t doCopyToSDataBlock(SExecTaskInfo* pTaskInfo, SSDataBlock* pBlock, SExprI
 
     SResultRow* pRow = (SResultRow*)((char*)page + pPos->pos.offset);
 
-    doUpdateNumOfRows(pRow, numOfExprs, rowCellOffset);
+    doUpdateNumOfRows(pCtx, pRow, numOfExprs, rowCellOffset);
     if (pRow->numOfRows == 0) {
       pGroupResInfo->index += 1;
       releaseBufPage(pBuf, page);
@@ -4492,7 +4502,7 @@ int32_t createDataSinkParam(SDataSinkNode* pNode, void** pParam, qTaskInfo_t* pT
 }
 
 int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId,
-                               const char* sql, EOPTR_EXEC_MODEL model) {
+                               char* sql, EOPTR_EXEC_MODEL model) {
   uint64_t queryId = pPlan->id.queryId;
 
   int32_t code = TSDB_CODE_SUCCESS;
@@ -4503,6 +4513,7 @@ int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SRead
   }
 
   (*pTaskInfo)->sql = sql;
+  sql = NULL;
   (*pTaskInfo)->pSubplan = pPlan;
   (*pTaskInfo)->pRoot = createOperatorTree(pPlan->pNode, *pTaskInfo, pHandle, &(*pTaskInfo)->tableqinfoList,
                                            pPlan->pTagCond, pPlan->pTagIndexCond, pPlan->user);
@@ -4515,6 +4526,7 @@ int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SRead
   return code;
 
 _complete:
+  taosMemoryFree(sql);
   doDestroyTask(*pTaskInfo);
   terrno = code;
   return code;

source/libs/executor/src/timewindowoperator.c

@@ -2065,10 +2065,30 @@ static void doKeepPrevRows(STimeSliceOperatorInfo* pSliceInfo, const SSDataBlock
       memcpy(pkey->pData, val, pkey->bytes);
     }
   }
+
+  pSliceInfo->isPrevRowSet = true;
+}
+
+static void doKeepNextRows(STimeSliceOperatorInfo* pSliceInfo, const SSDataBlock* pBlock, int32_t rowIndex) {
+  int32_t numOfCols = taosArrayGetSize(pBlock->pDataBlock);
+  for (int32_t i = 0; i < numOfCols; ++i) {
+    SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, i);
+
+    // null data should not be kept since it can not be used to perform interpolation
+    if (!colDataIsNull_s(pColInfoData, i)) {
+      SGroupKeys* pkey = taosArrayGet(pSliceInfo->pNextRow, i);
+
+      pkey->isNull = false;
+      char* val = colDataGetData(pColInfoData, rowIndex);
+      memcpy(pkey->pData, val, pkey->bytes);
+    }
+  }
+
+  pSliceInfo->isNextRowSet = true;
 }
 
 static void genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp* pExprSup, SSDataBlock* pBlock,
-                                   int32_t rowIndex, SSDataBlock* pResBlock) {
+                                   SSDataBlock* pResBlock) {
   int32_t rows = pResBlock->info.rows;
 
   // todo set the correct primary timestamp column
@@ -2144,6 +2164,10 @@ static void genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp
         break;
       }
       case TSDB_FILL_PREV: {
+        if (!pSliceInfo->isPrevRowSet) {
+          break;
+        }
+
         SGroupKeys* pkey = taosArrayGet(pSliceInfo->pPrevRow, srcSlot);
         colDataAppend(pDst, rows, pkey->pData, false);
         pResBlock->info.rows += 1;
@@ -2151,8 +2175,12 @@ static void genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp
       }
 
       case TSDB_FILL_NEXT: {
-        char* p = colDataGetData(pSrc, rowIndex);
-        colDataAppend(pDst, rows, p, colDataIsNull_s(pSrc, rowIndex));
+        if (!pSliceInfo->isNextRowSet) {
+          break;
+        }
+
+        SGroupKeys* pkey = taosArrayGet(pSliceInfo->pNextRow, srcSlot);
+        colDataAppend(pDst, rows, pkey->pData, false);
         pResBlock->info.rows += 1;
         break;
       }
@@ -2186,6 +2214,35 @@ static int32_t initPrevRowsKeeper(STimeSliceOperatorInfo* pInfo, SSDataBlock* pB
     taosArrayPush(pInfo->pPrevRow, &key);
   }
 
+  pInfo->isPrevRowSet = false;
+
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t initNextRowsKeeper(STimeSliceOperatorInfo* pInfo, SSDataBlock* pBlock) {
+  if (pInfo->pNextRow != NULL) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  pInfo->pNextRow = taosArrayInit(4, sizeof(SGroupKeys));
+  if (pInfo->pNextRow == NULL) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  int32_t numOfCols = taosArrayGetSize(pBlock->pDataBlock);
+  for (int32_t i = 0; i < numOfCols; ++i) {
+    SColumnInfoData* pColInfo = taosArrayGet(pBlock->pDataBlock, i);
+
+    SGroupKeys key = {0};
+    key.bytes = pColInfo->info.bytes;
+    key.type = pColInfo->info.type;
+    key.isNull = false;
+    key.pData = taosMemoryCalloc(1, pColInfo->info.bytes);
+    taosArrayPush(pInfo->pNextRow, &key);
+  }
+
+  pInfo->isNextRowSet = false;
+
   return TSDB_CODE_SUCCESS;
 }
 
@@ -2215,14 +2272,19 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
 
   blockDataCleanup(pResBlock);
 
-  //int32_t numOfRows = 0;
   while (1) {
     SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream);
     if (pBlock == NULL) {
       break;
     }
 
-    int32_t code = initPrevRowsKeeper(pSliceInfo, pBlock);
+    int32_t code;
+    code = initPrevRowsKeeper(pSliceInfo, pBlock);
+    if (code != TSDB_CODE_SUCCESS) {
+      longjmp(pTaskInfo->env, code);
+    }
+
+    code = initNextRowsKeeper(pSliceInfo, pBlock);
     if (code != TSDB_CODE_SUCCESS) {
       longjmp(pTaskInfo->env, code);
     }
@@ -2244,7 +2306,6 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
           SColumnInfoData* pDst = taosArrayGet(pResBlock->pDataBlock, dstSlot);
 
           char* v = colDataGetData(pSrc, i);
-          //colDataAppend(pDst, numOfRows, v, false);
           colDataAppend(pDst, pResBlock->info.rows, v, false);
         }
 
@@ -2262,11 +2323,16 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
           break;
         }
       } else if (ts < pSliceInfo->current) {
+        // in case interpolation window starts and ends between two datapoints, fill(prev) need to interpolate
+        doKeepPrevRows(pSliceInfo, pBlock, i);
+
         if (i < pBlock->info.rows - 1) {
+          // in case interpolation window starts and ends between two datapoints, fill(next) need to interpolate
+          doKeepNextRows(pSliceInfo, pBlock, i + 1);
           int64_t nextTs = *(int64_t*)colDataGetData(pTsCol, i + 1);
           if (nextTs > pSliceInfo->current) {
             while (pSliceInfo->current < nextTs && pSliceInfo->current <= pSliceInfo->win.ekey) {
-              genInterpolationResult(pSliceInfo, &pOperator->exprSupp, pBlock, i, pResBlock);
+              genInterpolationResult(pSliceInfo, &pOperator->exprSupp, pBlock, pResBlock);
               pSliceInfo->current =
                   taosTimeAdd(pSliceInfo->current, pInterval->interval, pInterval->intervalUnit, pInterval->precision);
               if (pResBlock->info.rows >= pResBlock->info.capacity) {
@@ -2285,8 +2351,11 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
           doKeepPrevRows(pSliceInfo, pBlock, i);
         }
       } else {  // ts > pSliceInfo->current
+        // in case interpolation window starts and ends between two datapoints, fill(next) need to interpolate
+        doKeepNextRows(pSliceInfo, pBlock, i);
+
         while (pSliceInfo->current < ts && pSliceInfo->current <= pSliceInfo->win.ekey) {
-          genInterpolationResult(pSliceInfo, &pOperator->exprSupp, pBlock, i, pResBlock);
+          genInterpolationResult(pSliceInfo, &pOperator->exprSupp, pBlock, pResBlock);
           pSliceInfo->current =
               taosTimeAdd(pSliceInfo->current, pInterval->interval, pInterval->intervalUnit, pInterval->precision);
           if (pResBlock->info.rows >= pResBlock->info.capacity) {
@@ -2326,9 +2395,10 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
       }
     }
 
-    //check if need to interpolate after ts range
-    while (pSliceInfo->current <= pSliceInfo->win.ekey) {
-      genInterpolationResult(pSliceInfo, &pOperator->exprSupp, pBlock, pBlock->info.rows - 1, pResBlock);
+    // check if need to interpolate after ts range
+    // except for fill(next)
+    while (pSliceInfo->current <= pSliceInfo->win.ekey && pSliceInfo->fillType != TSDB_FILL_NEXT) {
+      genInterpolationResult(pSliceInfo, &pOperator->exprSupp, pBlock, pResBlock);
       pSliceInfo->current =
           taosTimeAdd(pSliceInfo->current, pInterval->interval, pInterval->intervalUnit, pInterval->precision);
       if (pResBlock->info.rows >= pResBlock->info.capacity) {

source/libs/function/src/builtinsimpl.c

@@ -64,6 +64,9 @@ typedef struct SMinmaxResInfo {
   bool      assign;  // assign the first value or not
   int64_t   v;
   STuplePos tuplePos;
+
+  STuplePos nullTuplePos;
+  bool      nullTupleSaved;
 } SMinmaxResInfo;
 
 typedef struct STopBotResItem {
@@ -75,6 +78,10 @@ typedef struct STopBotResItem {
 typedef struct STopBotRes {
   int32_t         maxSize;
   int16_t         type;
+
+  STuplePos       nullTuplePos;
+  bool            nullTupleSaved;
+
   STopBotResItem* pItems;
 } STopBotRes;
 
@@ -221,6 +228,10 @@ typedef struct SSampleInfo {
   int32_t    numSampled;
   uint8_t    colType;
   int16_t    colBytes;
+
+  STuplePos  nullTuplePos;
+  bool       nullTupleSaved;
+
   char*      data;
   STuplePos* tuplePos;
 } SSampleInfo;
@@ -1134,6 +1145,9 @@ bool minmaxFunctionSetup(SqlFunctionCtx* pCtx, SResultRowEntryInfo* pResultInfo)
   SMinmaxResInfo* buf = GET_ROWCELL_INTERBUF(pResultInfo);
   buf->assign = false;
   buf->tuplePos.pageId = -1;
+
+  buf->nullTupleSaved = false;
+  buf->nullTuplePos.pageId = -1;
   return true;
 }
 
@@ -1575,6 +1589,10 @@ int32_t doMinMaxHelper(SqlFunctionCtx* pCtx, int32_t isMinFunc) {
   }
 
 _min_max_over:
+  if (numOfElems == 0 && pCtx->subsidiaries.num > 0 && !pBuf->nullTupleSaved ) {
+    doSaveTupleData(pCtx, pInput->startRowIndex, pCtx->pSrcBlock, &pBuf->nullTuplePos);
+    pBuf->nullTupleSaved = true;
+  }
   return numOfElems;
 }
 
@@ -1615,7 +1633,7 @@ int32_t minmaxFunctionFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   if (pEntryInfo->numOfRes > 0) {
     setSelectivityValue(pCtx, pBlock, &pRes->tuplePos, currentRow);
   } else {
-    setNullSelectivityValue(pCtx, pBlock, currentRow);
+    setSelectivityValue(pCtx, pBlock, &pRes->nullTuplePos, currentRow);
   }
 
   return pEntryInfo->numOfRes;
@@ -3366,6 +3384,8 @@ bool topBotFunctionSetup(SqlFunctionCtx* pCtx, SResultRowEntryInfo* pResInfo) {
 
   pRes->maxSize = pCtx->param[1].param.i;
 
+  pRes->nullTupleSaved = false;
+  pRes->nullTuplePos.pageId = -1;
   return true;
 }
 
@@ -3403,6 +3423,10 @@ int32_t topFunction(SqlFunctionCtx* pCtx) {
     doAddIntoResult(pCtx, data, i, pCtx->pSrcBlock, pRes->type, pInput->uid, pResInfo, true);
   }
 
+  if (numOfElems == 0 && pCtx->subsidiaries.num > 0 && !pRes->nullTupleSaved) {
+    doSaveTupleData(pCtx, pInput->startRowIndex, pCtx->pSrcBlock, &pRes->nullTuplePos);
+    pRes->nullTupleSaved = true;
+  }
   return TSDB_CODE_SUCCESS;
 }
 
@@ -3427,6 +3451,11 @@ int32_t bottomFunction(SqlFunctionCtx* pCtx) {
     doAddIntoResult(pCtx, data, i, pCtx->pSrcBlock, pRes->type, pInput->uid, pResInfo, false);
   }
 
+  if (numOfElems == 0 && pCtx->subsidiaries.num > 0 && !pRes->nullTupleSaved) {
+    doSaveTupleData(pCtx, pInput->startRowIndex, pCtx->pSrcBlock, &pRes->nullTuplePos);
+    pRes->nullTupleSaved = true;
+  }
+
   return TSDB_CODE_SUCCESS;
 }
 
@@ -3625,6 +3654,11 @@ int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
 
   // todo assign the tag value and the corresponding row data
   int32_t currentRow = pBlock->info.rows;
+  if (pEntryInfo->numOfRes <= 0) {
+    colDataAppendNULL(pCol, currentRow);
+    setSelectivityValue(pCtx, pBlock, &pRes->nullTuplePos, currentRow);
+    return pEntryInfo->numOfRes;
+  }
   for (int32_t i = 0; i < pEntryInfo->numOfRes; ++i) {
     STopBotResItem* pItem = &pRes->pItems[i];
     if (type == TSDB_DATA_TYPE_FLOAT) {
@@ -4897,7 +4931,8 @@ bool sampleFunctionSetup(SqlFunctionCtx* pCtx, SResultRowEntryInfo* pResultInfo)
   pInfo->numSampled = 0;
   pInfo->colType = pCtx->resDataInfo.type;
   pInfo->colBytes = pCtx->resDataInfo.bytes;
-
+  pInfo->nullTuplePos.pageId = -1;
+  pInfo->nullTupleSaved = false;
   pInfo->data = (char*)pInfo + sizeof(SSampleInfo);
   pInfo->tuplePos = (STuplePos*)((char*)pInfo + sizeof(SSampleInfo) + pInfo->samples * pInfo->colBytes);
 
@@ -4943,6 +4978,11 @@ int32_t sampleFunction(SqlFunctionCtx* pCtx) {
     doReservoirSample(pCtx, pInfo, data, i);
   }
 
+  if (pInfo->numSampled == 0 && pCtx->subsidiaries.num > 0 && !pInfo->nullTupleSaved) {
+    doSaveTupleData(pCtx, pInput->startRowIndex, pCtx->pSrcBlock, &pInfo->nullTuplePos);
+    pInfo->nullTupleSaved = true;
+  }
+
   SET_VAL(pResInfo, pInfo->numSampled, pInfo->numSampled);
   return TSDB_CODE_SUCCESS;
 }
@@ -4957,6 +4997,11 @@ int32_t sampleFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   SColumnInfoData* pCol = taosArrayGet(pBlock->pDataBlock, slotId);
 
   int32_t currentRow = pBlock->info.rows;
+  if (pInfo->numSampled == 0) {
+    colDataAppendNULL(pCol, currentRow);
+    setSelectivityValue(pCtx, pBlock, &pInfo->nullTuplePos, currentRow);
+    return pInfo->numSampled;
+  }
   for (int32_t i = 0; i < pInfo->numSampled; ++i) {
     colDataAppend(pCol, currentRow + i, pInfo->data + i * pInfo->colBytes, false);
     setSelectivityValue(pCtx, pBlock, &pInfo->tuplePos[i], currentRow + i);

source/libs/function/src/functionMgt.c

@@ -221,6 +221,18 @@ bool fmIsLastRowFunc(int32_t funcId) {
   return FUNCTION_TYPE_LAST_ROW == funcMgtBuiltins[funcId].type;
 }
 
+bool fmIsNotNullOutputFunc(int32_t funcId) {
+  if (funcId < 0 || funcId >= funcMgtBuiltinsNum) {
+    return false;
+  }
+  return FUNCTION_TYPE_LAST == funcMgtBuiltins[funcId].type ||
+         FUNCTION_TYPE_LAST_PARTIAL == funcMgtBuiltins[funcId].type ||
+         FUNCTION_TYPE_LAST_MERGE == funcMgtBuiltins[funcId].type ||
+         FUNCTION_TYPE_FIRST == funcMgtBuiltins[funcId].type ||
+         FUNCTION_TYPE_FIRST_PARTIAL == funcMgtBuiltins[funcId].type ||
+         FUNCTION_TYPE_FIRST_MERGE == funcMgtBuiltins[funcId].type;
+}
+
 bool fmIsSelectValueFunc(int32_t funcId) {
   if (funcId < 0 || funcId >= funcMgtBuiltinsNum) {
     return false;

source/libs/parser/inc/parUtil.h

@@ -39,6 +39,11 @@ typedef struct SMsgBuf {
   char*   buf;
 } SMsgBuf;
 
+typedef struct SParseTablesMetaReq {
+  char      dbFName[TSDB_DB_FNAME_LEN];
+  SHashObj* pTables;
+} SParseTablesMetaReq;
+
 typedef struct SParseMetaCache {
   SHashObj* pTableMeta;    // key is tbFName, element is STableMeta*
   SHashObj* pDbVgroup;     // key is dbFName, element is SArray<SVgroupInfo>*
@@ -95,7 +100,7 @@ int32_t getUdfInfoFromCache(SParseMetaCache* pMetaCache, const char* pFunc, SFun
 int32_t getTableIndexFromCache(SParseMetaCache* pMetaCache, const SName* pName, SArray** pIndexes);
 int32_t getTableCfgFromCache(SParseMetaCache* pMetaCache, const SName* pName, STableCfg** pOutput);
 int32_t getDnodeListFromCache(SParseMetaCache* pMetaCache, SArray** pDnodes);
-void    destoryParseMetaCache(SParseMetaCache* pMetaCache);
+void    destoryParseMetaCache(SParseMetaCache* pMetaCache, bool request);
 
 #ifdef __cplusplus
 }

source/libs/parser/src/parUtil.c

@@ -474,6 +474,24 @@ static int32_t buildDbReq(SHashObj* pDbsHash, SArray** pDbs) {
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t buildTableReqFromDb(SHashObj* pDbsHash, SArray** pDbs) {
+  if (NULL != pDbsHash) {
+    *pDbs = taosArrayInit(taosHashGetSize(pDbsHash), sizeof(STablesReq));
+    if (NULL == *pDbs) {
+      return TSDB_CODE_OUT_OF_MEMORY;
+    }
+    SParseTablesMetaReq* p = taosHashIterate(pDbsHash, NULL);
+    while (NULL != p) {
+      STablesReq req = {0};
+      strcpy(req.dbFName, p->dbFName);
+      buildTableReq(p->pTables, &req.pTables);
+      taosArrayPush(*pDbs, &req);
+      p = taosHashIterate(pDbsHash, p);
+    }
+  }
+  return TSDB_CODE_SUCCESS;
+}
+
 static int32_t buildUserAuthReq(SHashObj* pUserAuthHash, SArray** pUserAuth) {
   if (NULL != pUserAuthHash) {
     *pUserAuth = taosArrayInit(taosHashGetSize(pUserAuthHash), sizeof(SUserAuthInfo));
@@ -513,12 +531,12 @@ static int32_t buildUdfReq(SHashObj* pUdfHash, SArray** pUdf) {
 }
 
 int32_t buildCatalogReq(const SParseMetaCache* pMetaCache, SCatalogReq* pCatalogReq) {
-  int32_t code = buildTableReq(pMetaCache->pTableMeta, &pCatalogReq->pTableMeta);
+  int32_t code = buildTableReqFromDb(pMetaCache->pTableMeta, &pCatalogReq->pTableMeta);
   if (TSDB_CODE_SUCCESS == code) {
     code = buildDbReq(pMetaCache->pDbVgroup, &pCatalogReq->pDbVgroup);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = buildTableReq(pMetaCache->pTableVgroup, &pCatalogReq->pTableHash);
+    code = buildTableReqFromDb(pMetaCache->pTableVgroup, &pCatalogReq->pTableHash);
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = buildDbReq(pMetaCache->pDbCfg, &pCatalogReq->pDbCfg);
@@ -587,6 +605,24 @@ static int32_t putDbDataToCache(const SArray* pDbReq, const SArray* pDbData, SHa
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t putDbTableDataToCache(const SArray* pDbReq, const SArray* pTableData, SHashObj** pTable) {
+  int32_t ndbs = taosArrayGetSize(pDbReq);
+  int32_t tableNo = 0;
+  for (int32_t i = 0; i < ndbs; ++i) {
+    STablesReq* pReq = taosArrayGet(pDbReq, i);
+    int32_t     ntables = taosArrayGetSize(pReq->pTables);
+    for (int32_t j = 0; j < ntables; ++j) {
+      char fullName[TSDB_TABLE_FNAME_LEN];
+      tNameExtractFullName(taosArrayGet(pReq->pTables, j), fullName);
+      if (TSDB_CODE_SUCCESS != putMetaDataToHash(fullName, strlen(fullName), pTableData, tableNo, pTable)) {
+        return TSDB_CODE_OUT_OF_MEMORY;
+      }
+      ++tableNo;
+    }
+  }
+  return TSDB_CODE_SUCCESS;
+}
+
 static int32_t putUserAuthToCache(const SArray* pUserAuthReq, const SArray* pUserAuthData, SHashObj** pUserAuth) {
   int32_t nvgs = taosArrayGetSize(pUserAuthReq);
   for (int32_t i = 0; i < nvgs; ++i) {
@@ -612,12 +648,12 @@ static int32_t putUdfToCache(const SArray* pUdfReq, const SArray* pUdfData, SHas
 }
 
 int32_t putMetaDataToCache(const SCatalogReq* pCatalogReq, const SMetaData* pMetaData, SParseMetaCache* pMetaCache) {
-  int32_t code = putTableDataToCache(pCatalogReq->pTableMeta, pMetaData->pTableMeta, &pMetaCache->pTableMeta);
+  int32_t code = putDbTableDataToCache(pCatalogReq->pTableMeta, pMetaData->pTableMeta, &pMetaCache->pTableMeta);
   if (TSDB_CODE_SUCCESS == code) {
     code = putDbDataToCache(pCatalogReq->pDbVgroup, pMetaData->pDbVgroup, &pMetaCache->pDbVgroup);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = putTableDataToCache(pCatalogReq->pTableHash, pMetaData->pTableHash, &pMetaCache->pTableVgroup);
+    code = putDbTableDataToCache(pCatalogReq->pTableHash, pMetaData->pTableHash, &pMetaCache->pTableVgroup);
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = putDbDataToCache(pCatalogReq->pDbCfg, pMetaData->pDbCfg, &pMetaCache->pDbCfg);
@@ -657,14 +693,38 @@ static int32_t reserveTableReqInCache(int32_t acctId, const char* pDb, const cha
   return reserveTableReqInCacheImpl(fullName, len, pTables);
 }
 
+static int32_t reserveTableReqInDbCacheImpl(int32_t acctId, const char* pDb, const char* pTable, SHashObj* pDbs) {
+  SParseTablesMetaReq req = {0};
+  int32_t             len = snprintf(req.dbFName, sizeof(req.dbFName), "%d.%s", acctId, pDb);
+  int32_t             code = reserveTableReqInCache(acctId, pDb, pTable, &req.pTables);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = taosHashPut(pDbs, req.dbFName, len, &req, sizeof(SParseTablesMetaReq));
+  }
+  return code;
+}
+
+static int32_t reserveTableReqInDbCache(int32_t acctId, const char* pDb, const char* pTable, SHashObj** pDbs) {
+  if (NULL == *pDbs) {
+    *pDbs = taosHashInit(4, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_NO_LOCK);
+    if (NULL == *pDbs) {
+      return TSDB_CODE_OUT_OF_MEMORY;
+    }
+  }
+  char                 fullName[TSDB_DB_FNAME_LEN];
+  int32_t              len = snprintf(fullName, sizeof(fullName), "%d.%s", acctId, pDb);
+  SParseTablesMetaReq* pReq = taosHashGet(*pDbs, fullName, len);
+  if (NULL == pReq) {
+    return reserveTableReqInDbCacheImpl(acctId, pDb, pTable, *pDbs);
+  }
+  return reserveTableReqInCache(acctId, pDb, pTable, &pReq->pTables);
+}
+
 int32_t reserveTableMetaInCache(int32_t acctId, const char* pDb, const char* pTable, SParseMetaCache* pMetaCache) {
-  return reserveTableReqInCache(acctId, pDb, pTable, &pMetaCache->pTableMeta);
+  return reserveTableReqInDbCache(acctId, pDb, pTable, &pMetaCache->pTableMeta);
 }
 
 int32_t reserveTableMetaInCacheExt(const SName* pName, SParseMetaCache* pMetaCache) {
-  char fullName[TSDB_TABLE_FNAME_LEN];
-  tNameExtractFullName(pName, fullName);
-  return reserveTableReqInCacheImpl(fullName, strlen(fullName), &pMetaCache->pTableMeta);
+  return reserveTableReqInDbCache(pName->acctId, pName->dbname, pName->tname, &pMetaCache->pTableMeta);
 }
 
 int32_t getTableMetaFromCache(SParseMetaCache* pMetaCache, const SName* pName, STableMeta** pMeta) {
@@ -711,13 +771,11 @@ int32_t getDbVgInfoFromCache(SParseMetaCache* pMetaCache, const char* pDbFName,
 }
 
 int32_t reserveTableVgroupInCache(int32_t acctId, const char* pDb, const char* pTable, SParseMetaCache* pMetaCache) {
-  return reserveTableReqInCache(acctId, pDb, pTable, &pMetaCache->pTableVgroup);
+  return reserveTableReqInDbCache(acctId, pDb, pTable, &pMetaCache->pTableVgroup);
 }
 
 int32_t reserveTableVgroupInCacheExt(const SName* pName, SParseMetaCache* pMetaCache) {
-  char fullName[TSDB_TABLE_FNAME_LEN];
-  tNameExtractFullName(pName, fullName);
-  return reserveTableReqInCacheImpl(fullName, strlen(fullName), &pMetaCache->pTableVgroup);
+  return reserveTableReqInDbCache(pName->acctId, pName->dbname, pName->tname, &pMetaCache->pTableVgroup);
 }
 
 int32_t getTableVgroupFromCache(SParseMetaCache* pMetaCache, const SName* pName, SVgroupInfo* pVgroup) {
@@ -919,10 +977,24 @@ int32_t getDnodeListFromCache(SParseMetaCache* pMetaCache, SArray** pDnodes) {
   return TSDB_CODE_SUCCESS;
 }
 
-void destoryParseMetaCache(SParseMetaCache* pMetaCache) {
+void destoryParseTablesMetaReqHash(SHashObj* pHash) {
+  SParseTablesMetaReq* p = taosHashIterate(pHash, NULL);
+  while (NULL != p) {
+    taosHashCleanup(p->pTables);
+    p = taosHashIterate(pHash, p);
+  }
+  taosHashCleanup(pHash);
+}
+
+void destoryParseMetaCache(SParseMetaCache* pMetaCache, bool request) {
+  if (request) {
+    destoryParseTablesMetaReqHash(pMetaCache->pTableMeta);
+    destoryParseTablesMetaReqHash(pMetaCache->pTableVgroup);
+  } else {
     taosHashCleanup(pMetaCache->pTableMeta);
-  taosHashCleanup(pMetaCache->pDbVgroup);
     taosHashCleanup(pMetaCache->pTableVgroup);
+  }
+  taosHashCleanup(pMetaCache->pDbVgroup);
   taosHashCleanup(pMetaCache->pDbCfg);
   taosHashCleanup(pMetaCache->pDbInfo);
   taosHashCleanup(pMetaCache->pUserAuth);

source/libs/parser/src/parser.c

@@ -187,7 +187,7 @@ int32_t qParseSqlSyntax(SParseContext* pCxt, SQuery** pQuery, struct SCatalogReq
   if (TSDB_CODE_SUCCESS == code) {
     code = buildCatalogReq(&metaCache, pCatalogReq);
   }
-  destoryParseMetaCache(&metaCache);
+  destoryParseMetaCache(&metaCache, true);
   terrno = code;
   return code;
 }
@@ -203,7 +203,7 @@ int32_t qAnalyseSqlSemantic(SParseContext* pCxt, const struct SCatalogReq* pCata
       code = analyseSemantic(pCxt, pQuery, &metaCache);
     }
   }
-  destoryParseMetaCache(&metaCache);
+  destoryParseMetaCache(&metaCache, false);
   terrno = code;
   return code;
 }

source/libs/parser/test/mockCatalogService.cpp

@@ -472,28 +472,36 @@ class MockCatalogServiceImpl {
 
   int32_t getAllTableMeta(SArray* pTableMetaReq, SArray** pTableMetaData) const {
     if (NULL != pTableMetaReq) {
-      int32_t ntables = taosArrayGetSize(pTableMetaReq);
-      *pTableMetaData = taosArrayInit(ntables, sizeof(SMetaRes));
-      for (int32_t i = 0; i < ntables; ++i) {
+      int32_t ndbs = taosArrayGetSize(pTableMetaReq);
+      *pTableMetaData = taosArrayInit(ndbs, sizeof(SMetaRes));
+      for (int32_t i = 0; i < ndbs; ++i) {
+        STablesReq* pReq = (STablesReq*)taosArrayGet(pTableMetaReq, i);
+        int32_t     ntables = taosArrayGetSize(pReq->pTables);
+        for (int32_t j = 0; j < ntables; ++j) {
           SMetaRes res = {0};
-        res.code = catalogGetTableMeta((const SName*)taosArrayGet(pTableMetaReq, i), (STableMeta**)&res.pRes);
+          res.code = catalogGetTableMeta((const SName*)taosArrayGet(pReq->pTables, j), (STableMeta**)&res.pRes);
           taosArrayPush(*pTableMetaData, &res);
         }
       }
+    }
     return TSDB_CODE_SUCCESS;
   }
 
   int32_t getAllTableVgroup(SArray* pTableVgroupReq, SArray** pTableVgroupData) const {
     if (NULL != pTableVgroupReq) {
-      int32_t ntables = taosArrayGetSize(pTableVgroupReq);
-      *pTableVgroupData = taosArrayInit(ntables, sizeof(SMetaRes));
-      for (int32_t i = 0; i < ntables; ++i) {
+      int32_t ndbs = taosArrayGetSize(pTableVgroupReq);
+      *pTableVgroupData = taosArrayInit(ndbs, sizeof(SMetaRes));
+      for (int32_t i = 0; i < ndbs; ++i) {
+        STablesReq* pReq = (STablesReq*)taosArrayGet(pTableVgroupReq, i);
+        int32_t     ntables = taosArrayGetSize(pReq->pTables);
+        for (int32_t j = 0; j < ntables; ++j) {
           SMetaRes res = {0};
           res.pRes = taosMemoryCalloc(1, sizeof(SVgroupInfo));
-        res.code = catalogGetTableHashVgroup((const SName*)taosArrayGet(pTableVgroupReq, i), (SVgroupInfo*)res.pRes);
+          res.code = catalogGetTableHashVgroup((const SName*)taosArrayGet(pReq->pTables, j), (SVgroupInfo*)res.pRes);
           taosArrayPush(*pTableVgroupData, &res);
         }
       }
+    }
     return TSDB_CODE_SUCCESS;
   }
 
@@ -677,12 +685,17 @@ int32_t MockCatalogService::catalogGetAllMeta(const SCatalogReq* pCatalogReq, SM
   return impl_->catalogGetAllMeta(pCatalogReq, pMetaData);
 }
 
+void MockCatalogService::destoryTablesReq(void* p) {
+  STablesReq* pRes = (STablesReq*)p;
+  taosArrayDestroy(pRes->pTables);
+}
+
 void MockCatalogService::destoryCatalogReq(SCatalogReq* pReq) {
   taosArrayDestroy(pReq->pDbVgroup);
   taosArrayDestroy(pReq->pDbCfg);
   taosArrayDestroy(pReq->pDbInfo);
-  taosArrayDestroy(pReq->pTableMeta);
-  taosArrayDestroy(pReq->pTableHash);
+  taosArrayDestroyEx(pReq->pTableMeta, destoryTablesReq);
+  taosArrayDestroyEx(pReq->pTableHash, destoryTablesReq);
   taosArrayDestroy(pReq->pUdf);
   taosArrayDestroy(pReq->pIndex);
   taosArrayDestroy(pReq->pUser);

source/libs/parser/test/mockCatalogService.h

@@ -50,6 +50,7 @@ struct MockTableMeta {
 class MockCatalogServiceImpl;
 class MockCatalogService {
  public:
+  static void destoryTablesReq(void* p);
   static void destoryCatalogReq(SCatalogReq* pReq);
   static void destoryMetaRes(void* p);
   static void destoryMetaArrayRes(void* p);

source/libs/parser/test/parInsertTest.cpp

@@ -13,6 +13,8 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <functional>
+
 #include <gtest/gtest.h>
 
 #include "mockCatalogService.h"
@@ -20,6 +22,7 @@
 #include "parInt.h"
 
 using namespace std;
+using namespace std::placeholders;
 using namespace testing;
 
 namespace {
@@ -63,7 +66,9 @@ class InsertTest : public Test {
 
   int32_t runAsync() {
     cxt_.async = true;
-    unique_ptr<SParseMetaCache, void (*)(SParseMetaCache*)> metaCache(new SParseMetaCache(), _destoryParseMetaCache);
+    bool                                                           request = true;
+    unique_ptr<SParseMetaCache, function<void(SParseMetaCache*)> > metaCache(
+        new SParseMetaCache(), std::bind(_destoryParseMetaCache, _1, cref(request)));
     code_ = parseInsertSyntax(&cxt_, &res_, metaCache.get());
     if (code_ != TSDB_CODE_SUCCESS) {
       cout << "parseInsertSyntax code:" << toString(code_) << ", msg:" << errMagBuf_ << endl;
@@ -81,6 +86,8 @@ class InsertTest : public Test {
     unique_ptr<SMetaData, void (*)(SMetaData*)> metaData(new SMetaData(), MockCatalogService::destoryMetaData);
     g_mockCatalogService->catalogGetAllMeta(catalogReq.get(), metaData.get());
 
+    metaCache.reset(new SParseMetaCache());
+    request = false;
     code_ = putMetaDataToCache(catalogReq.get(), metaData.get(), metaCache.get());
     if (code_ != TSDB_CODE_SUCCESS) {
       cout << "putMetaDataToCache code:" << toString(code_) << ", msg:" << errMagBuf_ << endl;
@@ -144,8 +151,8 @@ class InsertTest : public Test {
   static const int max_err_len = 1024;
   static const int max_sql_len = 1024 * 1024;
 
-  static void _destoryParseMetaCache(SParseMetaCache* pMetaCache) {
-    destoryParseMetaCache(pMetaCache);
+  static void _destoryParseMetaCache(SParseMetaCache* pMetaCache, bool request) {
+    destoryParseMetaCache(pMetaCache, request);
     delete pMetaCache;
   }
 

source/libs/parser/test/parTestUtil.cpp

@@ -24,6 +24,7 @@
 #include "parInt.h"
 
 using namespace std;
+using namespace std::placeholders;
 using namespace testing;
 
 namespace ParserTest {
@@ -118,8 +119,8 @@ class ParserTestBaseImpl {
     TEST_INTERFACE_ASYNC_API
   };
 
-  static void _destoryParseMetaCache(SParseMetaCache* pMetaCache) {
-    destoryParseMetaCache(pMetaCache);
+  static void _destoryParseMetaCache(SParseMetaCache* pMetaCache, bool request) {
+    destoryParseMetaCache(pMetaCache, request);
     delete pMetaCache;
   }
 
@@ -340,7 +341,9 @@ class ParserTestBaseImpl {
       doParse(&cxt, query.get());
       SQuery* pQuery = *(query.get());
 
-      unique_ptr<SParseMetaCache, void (*)(SParseMetaCache*)> metaCache(new SParseMetaCache(), _destoryParseMetaCache);
+      bool                                                           request = true;
+      unique_ptr<SParseMetaCache, function<void(SParseMetaCache*)> > metaCache(
+          new SParseMetaCache(), bind(_destoryParseMetaCache, _1, cref(request)));
       doCollectMetaKey(&cxt, pQuery, metaCache.get());
 
       unique_ptr<SCatalogReq, void (*)(SCatalogReq*)> catalogReq(new SCatalogReq(),
@@ -353,6 +356,8 @@ class ParserTestBaseImpl {
           unique_ptr<SMetaData, void (*)(SMetaData*)> metaData(new SMetaData(), MockCatalogService::destoryMetaData);
           doGetAllMeta(catalogReq.get(), metaData.get());
 
+          metaCache.reset(new SParseMetaCache());
+          request = false;
           doPutMetaDataToCache(catalogReq.get(), metaData.get(), metaCache.get());
 
           doAuthenticate(&cxt, pQuery, metaCache.get());

source/libs/qcom/src/queryUtil.c

@@ -462,3 +462,5 @@ int32_t cloneDbVgInfo(SDBVgInfo* pSrc, SDBVgInfo** pDst) {
 
   return TSDB_CODE_SUCCESS;
 }
+
+

source/libs/qworker/inc/qwMsg.h

@@ -25,7 +25,7 @@ extern "C" {
 
 int32_t qwAbortPrerocessQuery(QW_FPARAMS_DEF);
 int32_t qwPreprocessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg);
-int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, const char* sql);
+int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, char* sql);
 int32_t qwProcessCQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg);
 int32_t qwProcessReady(QW_FPARAMS_DEF, SQWMsg *qwMsg);
 int32_t qwProcessFetch(QW_FPARAMS_DEF, SQWMsg *qwMsg);

source/libs/qworker/src/qwMsg.c

@@ -387,10 +387,13 @@ int32_t qWorkerProcessQueryMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, int
   
   char * sql = strndup(msg->msg, msg->sqlLen);
   QW_SCH_TASK_DLOG("processQuery start, node:%p, type:%s, handle:%p, SQL:%s", node, TMSG_INFO(pMsg->msgType), pMsg->info.handle, sql);
-  QW_ERR_RET(qwProcessQuery(QW_FPARAMS(), &qwMsg, sql));
-  QW_SCH_TASK_DLOG("processQuery end, node:%p", node);
+  QW_ERR_JRET(qwProcessQuery(QW_FPARAMS(), &qwMsg, sql));
 
-  return TSDB_CODE_SUCCESS;
+_return:
+
+  QW_SCH_TASK_DLOG("processQuery end, node:%p, code:%d", node, code);
+
+  return code;
 }
 
 int32_t qWorkerProcessCQueryMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, int64_t ts) {

source/libs/qworker/src/qworker.c

@@ -508,7 +508,7 @@ _return:
 }
 
 
-int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, const char* sql) {
+int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, char* sql) {
   int32_t        code = 0;
   bool           queryRsped = false;
   SSubplan      *plan = NULL;
@@ -536,6 +536,7 @@ int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, const char* sql) {
   }
 
   code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle, sql, OPTR_EXEC_MODEL_BATCH);
+  sql = NULL;
   if (code) {
     QW_TASK_ELOG("qCreateExecTask failed, code:%x - %s", code, tstrerror(code));
     QW_ERR_JRET(code);
@@ -561,6 +562,8 @@ int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, const char* sql) {
 
 _return:
 
+  taosMemoryFree(sql);
+  
   input.code = code;
   input.msgType = qwMsg->msgType;
   code = qwHandlePostPhaseEvents(QW_FPARAMS(), QW_PHASE_POST_QUERY, &input, NULL);

source/libs/scheduler/src/schStatus.c

@@ -64,7 +64,7 @@ _return:
 int32_t schHandleOpBeginEvent(int64_t jobId, SSchJob** job, SCH_OP_TYPE type, SSchedulerReq* pReq) {
   SSchJob *pJob = schAcquireJob(jobId);
   if (NULL == pJob) {
-    qError("Acquire sch job failed, may be dropped, jobId:0x%" PRIx64, jobId);
+    qWarn("Acquire sch job failed, may be dropped, jobId:0x%" PRIx64, jobId);
     SCH_ERR_RET(TSDB_CODE_SCH_STATUS_ERROR);
   }
 

source/libs/scheduler/src/schTask.c

@@ -537,6 +537,7 @@ int32_t schMoveTaskToExecList(SSchJob *pJob, SSchTask *pTask, bool *moved) {
 int32_t schTaskCheckSetRetry(SSchJob *pJob, SSchTask *pTask, int32_t errCode, bool *needRetry) {
   if (TSDB_CODE_SCH_TIMEOUT_ERROR == errCode) {
     pTask->maxExecTimes++;
+    pTask->maxRetryTimes++;
     if (pTask->timeoutUsec < SCH_MAX_TASK_TIMEOUT_USEC) {
       pTask->timeoutUsec *= 2;
       if (pTask->timeoutUsec > SCH_MAX_TASK_TIMEOUT_USEC) {

source/libs/scheduler/src/scheduler.c

@@ -150,7 +150,7 @@ void schedulerFreeJob(int64_t* jobId, int32_t errCode) {
 
   SSchJob *pJob = schAcquireJob(*jobId);
   if (NULL == pJob) {
-    qError("Acquire sch job failed, may be dropped, jobId:0x%" PRIx64, *jobId);
+    qWarn("Acquire sch job failed, may be dropped, jobId:0x%" PRIx64, *jobId);
     return;
   }
 

source/libs/transport/inc/transComm.h

@@ -95,8 +95,9 @@ typedef void* queue[2];
 #define QUEUE_DATA(e, type, field) ((type*)((void*)((char*)(e)-offsetof(type, field))))
 
 #define TRANS_RETRY_COUNT_LIMIT 100   // retry count limit
-#define TRANS_RETRY_INTERVAL    15   // ms retry interval
-#define TRANS_CONN_TIMEOUT      3    // connect timeout
+#define TRANS_RETRY_INTERVAL    15    // retry interval (ms)
+#define TRANS_CONN_TIMEOUT      3     // connect timeout (s)
+#define TRANS_READ_TIMEOUT      3000  // read timeout  (ms)
 
 typedef SRpcMsg      STransMsg;
 typedef SRpcCtx      STransCtx;

source/libs/transport/inc/transportInt.h

@@ -53,6 +53,7 @@ typedef struct {
 
   void (*cfp)(void* parent, SRpcMsg*, SEpSet*);
   bool (*retry)(int32_t code, tmsg_t msgType);
+  bool (*startTimer)(int32_t code, tmsg_t msgType);
   int index;
 
   void*         parent;

source/libs/transport/src/trans.c

@@ -48,6 +48,7 @@ void* rpcOpen(const SRpcInit* pInit) {
   // register callback handle
   pRpc->cfp = pInit->cfp;
   pRpc->retry = pInit->rfp;
+  pRpc->startTimer = pInit->tfp;
 
   if (pInit->connType == TAOS_CONN_SERVER) {
     pRpc->numOfThreads = pInit->numOfThreads > TSDB_MAX_RPC_THREADS ? TSDB_MAX_RPC_THREADS : pInit->numOfThreads;

source/libs/transport/src/transCli.c

@@ -24,6 +24,7 @@ typedef struct SCliConn {
   uv_connect_t connReq;
   uv_stream_t* stream;
   queue        wreqQueue;
+  uv_timer_t*  timer;
 
   void* hostThrd;
 
@@ -65,12 +66,13 @@ typedef struct SCliThrd {
   int64_t       pid;     // pid
   uv_loop_t*    loop;
   SAsyncPool*   asyncPool;
-  uv_idle_t*    idle;
   uv_prepare_t* prepare;
-  uv_timer_t    timer;
   void*         pool;  // conn pool
 
+  SArray* timerList;
+
   // msg queue
+
   queue         msg;
   TdThreadMutex msgMtx;
   SDelayQueue*  delayQueue;
@@ -108,6 +110,8 @@ static int sockDebugInfo(struct sockaddr* sockname, char* dst) {
   sprintf(dst, "%s:%d", buf, ntohs(addr.sin_port));
   return r;
 }
+// register timer for read
+static void cliReadTimeoutCb(uv_timer_t* handle);
 // register timer in each thread to clear expire conn
 // static void cliTimeoutCb(uv_timer_t* handle);
 // alloc buf for recv
@@ -330,6 +334,16 @@ void cliHandleResp(SCliConn* conn) {
   SCliThrd* pThrd = conn->hostThrd;
   STrans*   pTransInst = pThrd->pTransInst;
 
+  if (conn->timer) {
+    if (uv_is_active((uv_handle_t*)conn->timer)) {
+      tDebug("%s conn %p stop timer", CONN_GET_INST_LABEL(conn), conn);
+      uv_timer_stop(conn->timer);
+    }
+    conn->timer->data = NULL;
+    taosArrayPush(pThrd->timerList, &conn->timer);
+    conn->timer = NULL;
+  }
+
   STransMsgHead* pHead = NULL;
   transDumpFromBuffer(&conn->readBuf, (char**)&pHead);
   pHead->code = htonl(pHead->code);
@@ -409,7 +423,7 @@ void cliHandleResp(SCliConn* conn) {
   uv_read_start((uv_stream_t*)conn->stream, cliAllocRecvBufferCb, cliRecvCb);
 }
 
-void cliHandleExcept(SCliConn* pConn) {
+void cliHandleExceptImpl(SCliConn* pConn, int32_t code) {
   if (transQueueEmpty(&pConn->cliMsgs)) {
     if (pConn->broken == true && CONN_NO_PERSIST_BY_APP(pConn)) {
       tTrace("%s conn %p handle except, persist:0", CONN_GET_INST_LABEL(pConn), pConn);
@@ -428,7 +442,7 @@ void cliHandleExcept(SCliConn* pConn) {
     STransConnCtx* pCtx = pMsg ? pMsg->ctx : NULL;
 
     STransMsg transMsg = {0};
-    transMsg.code = pConn->broken ? TSDB_CODE_RPC_BROKEN_LINK : TSDB_CODE_RPC_NETWORK_UNAVAIL;
+    transMsg.code = code == -1 ? (pConn->broken ? TSDB_CODE_RPC_BROKEN_LINK : TSDB_CODE_RPC_NETWORK_UNAVAIL) : code;
     transMsg.msgType = pMsg ? pMsg->msg.msgType + 1 : 0;
     transMsg.info.ahandle = NULL;
 
@@ -459,31 +473,17 @@ void cliHandleExcept(SCliConn* pConn) {
   } while (!transQueueEmpty(&pConn->cliMsgs));
   transUnrefCliHandle(pConn);
 }
+void cliHandleExcept(SCliConn* conn) {
+  tTrace("%s conn %p except ref:%d", CONN_GET_INST_LABEL(conn), conn, T_REF_VAL_GET(conn));
+  cliHandleExceptImpl(conn, -1);
+}
 
-// void cliTimeoutCb(uv_timer_t* handle) {
-//   SCliThrd* pThrd = handle->data;
-//   STrans*   pTransInst = pThrd->pTransInst;
-//   int64_t   currentTime = pThrd->nextTimeout;
-//   tTrace("%s conn timeout, try to remove expire conn from conn pool", pTransInst->label);
-//
-//   SConnList* p = taosHashIterate((SHashObj*)pThrd->pool, NULL);
-//   while (p != NULL) {
-//     while (!QUEUE_IS_EMPTY(&p->conn)) {
-//       queue*    h = QUEUE_HEAD(&p->conn);
-//       SCliConn* c = QUEUE_DATA(h, SCliConn, q);
-//       if (c->expireTime < currentTime) {
-//         QUEUE_REMOVE(h);
-//         transUnrefCliHandle(c);
-//       } else {
-//         break;
-//       }
-//     }
-//     p = taosHashIterate((SHashObj*)pThrd->pool, p);
-//   }
-//
-//   pThrd->nextTimeout = taosGetTimestampMs() + CONN_PERSIST_TIME(pTransInst->idleTime);
-//   uv_timer_start(handle, cliTimeoutCb, CONN_PERSIST_TIME(pTransInst->idleTime) / 2, 0);
-// }
+void cliReadTimeoutCb(uv_timer_t* handle) {
+  // set up timeout cb
+  SCliConn* conn = handle->data;
+  tTrace("%s conn %p timeout, ref:%d", CONN_GET_INST_LABEL(conn), conn, T_REF_VAL_GET(conn));
+  cliHandleExceptImpl(conn, TSDB_CODE_RPC_TIMEOUT);
+}
 
 void* createConnPool(int size) {
   // thread local, no lock
@@ -654,13 +654,23 @@ static SCliConn* cliCreateConn(SCliThrd* pThrd) {
   return conn;
 }
 static void cliDestroyConn(SCliConn* conn, bool clear) {
+  SCliThrd* pThrd = conn->hostThrd;
   tTrace("%s conn %p remove from conn pool", CONN_GET_INST_LABEL(conn), conn);
   QUEUE_REMOVE(&conn->q);
   QUEUE_INIT(&conn->q);
   transRemoveExHandle(transGetRefMgt(), conn->refId);
-
   conn->refId = -1;
-  if (conn->task != NULL) transDQCancel(((SCliThrd*)conn->hostThrd)->timeoutQueue, conn->task);
+
+  if (conn->task != NULL) {
+    transDQCancel(pThrd->timeoutQueue, conn->task);
+    conn->task = NULL;
+  }
+  if (conn->timer != NULL) {
+    uv_timer_stop(conn->timer);
+    taosArrayPush(pThrd->timerList, &conn->timer);
+    conn->timer->data = NULL;
+    conn->timer = NULL;
+  }
 
   if (clear) {
     if (!uv_is_closing((uv_handle_t*)conn->stream)) {
@@ -673,8 +683,15 @@ static void cliDestroy(uv_handle_t* handle) {
   if (uv_handle_get_type(handle) != UV_TCP || handle->data == NULL) {
     return;
   }
-
   SCliConn* conn = handle->data;
+  SCliThrd* pThrd = conn->hostThrd;
+  if (conn->timer != NULL) {
+    uv_timer_stop(conn->timer);
+    taosArrayPush(pThrd->timerList, &conn->timer);
+    conn->timer->data = NULL;
+    conn->timer = NULL;
+  }
+
   transRemoveExHandle(transGetRefMgt(), conn->refId);
   taosMemoryFree(conn->ip);
   conn->stream->data = NULL;
@@ -764,6 +781,19 @@ void cliSend(SCliConn* pConn) {
     CONN_SET_PERSIST_BY_APP(pConn);
   }
 
+  if (pTransInst->startTimer != NULL && pTransInst->startTimer(0, pMsg->msgType)) {
+    uv_timer_t* timer = taosArrayGetSize(pThrd->timerList) > 0 ? *(uv_timer_t**)taosArrayPop(pThrd->timerList) : NULL;
+    if (timer == NULL) {
+      tDebug("no avaiable timer, create");
+      timer = taosMemoryCalloc(1, sizeof(uv_timer_t));
+      uv_timer_init(pThrd->loop, timer);
+    }
+    timer->data = pConn;
+    pConn->timer = timer;
+
+    tGTrace("%s conn %p start timer for msg:%s", CONN_GET_INST_LABEL(pConn), pConn, TMSG_INFO(pMsg->msgType));
+    uv_timer_start((uv_timer_t*)pConn->timer, cliReadTimeoutCb, TRANS_READ_TIMEOUT, 0);
+  }
   uv_write_t* req = transReqQueuePush(&pConn->wreqQueue);
   uv_write(req, (uv_stream_t*)pConn->stream, &wb, 1, cliSendCb);
   return;
@@ -781,8 +811,8 @@ void cliConnCb(uv_connect_t* req, int status) {
   }
   // int addrlen = sizeof(pConn->addr);
   struct sockaddr peername, sockname;
-  int             addrlen = sizeof(peername);
 
+  int addrlen = sizeof(peername);
   uv_tcp_getpeername((uv_tcp_t*)pConn->stream, &peername, &addrlen);
   transGetSockDebugInfo(&peername, pConn->dst);
 
@@ -806,7 +836,6 @@ static void cliHandleQuit(SCliMsg* pMsg, SCliThrd* pThrd) {
   tDebug("cli work thread %p start to quit", pThrd);
   destroyCmsg(pMsg);
   destroyConnPool(pThrd->pool);
-  uv_timer_stop(&pThrd->timer);
   uv_walk(pThrd->loop, cliWalkCb, NULL);
 }
 static void cliHandleRelease(SCliMsg* pMsg, SCliThrd* pThrd) {
@@ -879,8 +908,8 @@ void cliMayCvtFqdnToIp(SEpSet* pEpSet, SCvtAddr* pCvtAddr) {
   }
 }
 void cliHandleReq(SCliMsg* pMsg, SCliThrd* pThrd) {
-  STransConnCtx* pCtx = pMsg->ctx;
   STrans*        pTransInst = pThrd->pTransInst;
+  STransConnCtx* pCtx = pMsg->ctx;
 
   cliMayCvtFqdnToIp(&pCtx->epSet, &pThrd->cvtAddr);
   if (!EPSET_IS_VALID(&pCtx->epSet)) {
@@ -966,36 +995,6 @@ static void cliAsyncCb(uv_async_t* handle) {
   }
   if (pThrd->stopMsg != NULL) cliHandleQuit(pThrd->stopMsg, pThrd);
 }
-static void cliIdleCb(uv_idle_t* handle) {
-  SCliThrd* thrd = handle->data;
-  tTrace("do idle work");
-
-  SAsyncPool* pool = thrd->asyncPool;
-  for (int i = 0; i < pool->nAsync; i++) {
-    uv_async_t* async = &(pool->asyncs[i]);
-    SAsyncItem* item = async->data;
-
-    queue wq;
-    taosThreadMutexLock(&item->mtx);
-    QUEUE_MOVE(&item->qmsg, &wq);
-    taosThreadMutexUnlock(&item->mtx);
-
-    int count = 0;
-    while (!QUEUE_IS_EMPTY(&wq)) {
-      queue* h = QUEUE_HEAD(&wq);
-      QUEUE_REMOVE(h);
-
-      SCliMsg* pMsg = QUEUE_DATA(h, SCliMsg, q);
-      if (pMsg == NULL) {
-        continue;
-      }
-      (*cliAsyncHandle[pMsg->type])(pMsg, thrd);
-      count++;
-    }
-  }
-  tTrace("prepare work end");
-  if (thrd->stopMsg != NULL) cliHandleQuit(thrd->stopMsg, thrd);
-}
 static void cliPrepareCb(uv_prepare_t* handle) {
   SCliThrd* thrd = handle->data;
   tTrace("prepare work start");
@@ -1085,19 +1084,20 @@ static SCliThrd* createThrdObj() {
   uv_loop_init(pThrd->loop);
 
   pThrd->asyncPool = transAsyncPoolCreate(pThrd->loop, 5, pThrd, cliAsyncCb);
-  uv_timer_init(pThrd->loop, &pThrd->timer);
-  pThrd->timer.data = pThrd;
-
-  // pThrd->idle = taosMemoryCalloc(1, sizeof(uv_idle_t));
-  // uv_idle_init(pThrd->loop, pThrd->idle);
-  // pThrd->idle->data = pThrd;
-  // uv_idle_start(pThrd->idle, cliIdleCb);
 
   pThrd->prepare = taosMemoryCalloc(1, sizeof(uv_prepare_t));
   uv_prepare_init(pThrd->loop, pThrd->prepare);
   pThrd->prepare->data = pThrd;
   uv_prepare_start(pThrd->prepare, cliPrepareCb);
 
+  int32_t timerSize = 512;
+  pThrd->timerList = taosArrayInit(timerSize, sizeof(void*));
+  for (int i = 0; i < timerSize; i++) {
+    uv_timer_t* timer = taosMemoryCalloc(1, sizeof(uv_timer_t));
+    uv_timer_init(pThrd->loop, timer);
+    taosArrayPush(pThrd->timerList, &timer);
+  }
+
   pThrd->pool = createConnPool(4);
   transDQCreate(pThrd->loop, &pThrd->delayQueue);
 
@@ -1120,7 +1120,12 @@ static void destroyThrdObj(SCliThrd* pThrd) {
   transDQDestroy(pThrd->delayQueue, destroyCmsg);
   transDQDestroy(pThrd->timeoutQueue, NULL);
 
-  taosMemoryFree(pThrd->idle);
+  for (int i = 0; i < taosArrayGetSize(pThrd->timerList); i++) {
+    uv_timer_t* timer = taosArrayGetP(pThrd->timerList, i);
+    taosMemoryFree(timer);
+  }
+  taosArrayDestroy(pThrd->timerList);
+
   taosMemoryFree(pThrd->prepare);
   taosMemoryFree(pThrd->loop);
   taosMemoryFree(pThrd);

source/os/src/osSocket.c

@@ -333,7 +333,7 @@ int32_t taosWriteMsg(TdSocketPtr pSocket, void *buf, int32_t nbytes) {
     return -1;
   }
   int32_t nleft, nwritten;
-  char *  ptr = (char *)buf;
+  char   *ptr = (char *)buf;
 
   nleft = nbytes;
 
@@ -362,7 +362,7 @@ int32_t taosReadMsg(TdSocketPtr pSocket, void *buf, int32_t nbytes) {
     return -1;
   }
   int32_t nleft, nread;
-  char *  ptr = (char *)buf;
+  char   *ptr = (char *)buf;
 
   nleft = nbytes;
 
@@ -912,7 +912,7 @@ uint32_t taosGetIpv4FromFqdn(const char *fqdn) {
 
   int32_t ret = getaddrinfo(fqdn, NULL, &hints, &result);
   if (result) {
-    struct sockaddr *   sa = result->ai_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;

source/util/src/terror.c

@@ -52,6 +52,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_RPC_NETWORK_UNAVAIL,          "Unable to establish c
 TAOS_DEFINE_ERROR(TSDB_CODE_RPC_FQDN_ERROR,               "Unable to resolve FQDN")
 TAOS_DEFINE_ERROR(TSDB_CODE_RPC_PORT_EADDRINUSE,          "Port already in use")
 TAOS_DEFINE_ERROR(TSDB_CODE_RPC_BROKEN_LINK,              "Conn is broken")
+TAOS_DEFINE_ERROR(TSDB_CODE_RPC_TIMEOUT,                  "Conn read timeout")
 
 //common & util
 TAOS_DEFINE_ERROR(TSDB_CODE_TIME_UNSYNCED,                "Client and server's time is not synchronized")

tests/docs-examples-test/go.sh

+#!/bin/bash
+
+set -e
+
+taosd >>/dev/null 2>&1 &
+taosadapter >>/dev/null 2>&1 &
+
+cd ../../docs/examples/go
+
+go mod tidy
+
+go run ./connect/afconn/main.go
+go run ./connect/cgoexample/main.go
+go run ./connect/restexample/main.go
+
+taos -s "drop database if exists test"
+go run ./insert/json/main.go
+taos -s "drop database if exists test"
+go run ./insert/line/main.go
+taos -s "drop database if exists power"
+go run ./insert/sql/main.go
+taos -s "drop database if exists power"
+go run ./insert/stmt/main.go
+taos -s "drop database if exists test"
+go run ./insert/telnet/main.go
+
+go run ./query/sync/main.go
+
+taos -s "drop topic if exists example_tmq_topic"
+taos -s "drop database if exists example_tmq"
+go run ./sub/main.go

tests/parallel_test/run_case.sh

@@ -49,11 +49,13 @@ if [ $ent -eq 0 ]; then
     export PATH=$PATH:/home/TDengine/debug/build/bin
     export LD_LIBRARY_PATH=/home/TDengine/debug/build/lib
     ln -s /home/TDengine/debug/build/lib/libtaos.so /usr/lib/libtaos.so 2>/dev/null
+    ln -s /home/TDengine/debug/build/lib/libtaos.so /usr/lib/libtaos.so.1 2>/dev/null
     CONTAINER_TESTDIR=/home/TDengine
 else
     export PATH=$PATH:/home/TDinternal/debug/build/bin
     export LD_LIBRARY_PATH=/home/TDinternal/debug/build/lib
     ln -s /home/TDinternal/debug/build/lib/libtaos.so /usr/lib/libtaos.so 2>/dev/null
+    ln -s /home/TDinternal/debug/build/lib/libtaos.so /usr/lib/libtaos.so.1 2>/dev/null
     CONTAINER_TESTDIR=/home/TDinternal/community
 fi
 mkdir -p /var/lib/taos/subscribe

tests/parallel_test/run_container.sh

@@ -100,6 +100,7 @@ docker run \
     -v "$TMP_DIR/thread_volume/$thread_no/sim:${SIM_DIR}" \
     -v ${TMP_DIR}/thread_volume/$thread_no/coredump:$coredump_dir \
     -v $WORKDIR/taos-connector-python/taos:/usr/local/lib/python3.8/site-packages/taos:ro \
+    -v $WORKDIR/taos-connector-python/taosrest:/usr/local/lib/python3.8/site-packages/taosrest:ro \
     --rm --ulimit core=-1 taos_test:v1.0 $CONTAINER_TESTDIR/tests/parallel_test/run_case.sh -d "$exec_dir" -c "$cmd" $extra_param
 ret=$?
 exit $ret

tests/pytest/util/sql.py

@@ -84,10 +84,10 @@ class TDSql:
             self.queryResult = None
             tdLog.info("sql:%s, expect error occured" % (sql))
 
-    def query(self, sql, row_tag=None,queyTimes=10):
+    def query(self, sql, row_tag=None,queryTimes=10):
         self.sql = sql
         i=1
-        while i <= queyTimes:
+        while i <= queryTimes:
             try:
                 self.cursor.execute(sql)
                 self.queryResult = self.cursor.fetchall()
@@ -97,26 +97,15 @@ class TDSql:
                     return self.queryResult
                 return self.queryRows
             except Exception as e:
-                i+=1
                 tdLog.notice("Try to query again, query times: %d "%i)
-                time.sleep(1)
-                pass
-        else:
-            try:
-                tdLog.notice("Try the last query ")
-                self.cursor.execute(sql)
-                self.queryResult = self.cursor.fetchall()
-                self.queryRows = len(self.queryResult)
-                self.queryCols = len(self.cursor.description)
-                if row_tag:
-                    return self.queryResult
-                return self.queryRows
-            except Exception as e:
+                if i == queryTimes:
                     caller = inspect.getframeinfo(inspect.stack()[1][0])
                     args = (caller.filename, caller.lineno, sql, repr(e))
                     tdLog.notice("%s(%d) failed: sql:%s, %s" % args)
-                traceback.print_exc()
                     raise Exception(repr(e))   
+                i+=1
+                time.sleep(1)
+                pass
 
 
     def is_err_sql(self, sql):
@@ -305,28 +294,23 @@ class TDSql:
                 time.sleep(1)
                 continue
 
-    def execute(self, sql,queyTimes=10):
+    def execute(self, sql,queryTimes=10):
         self.sql = sql
         i=1
-        while i <= queyTimes:
+        while i <= queryTimes:
             try:
                 self.affectedRows = self.cursor.execute(sql)
                 return self.affectedRows
             except Exception as e:
-                i+=1
                 tdLog.notice("Try to execute sql again, query times: %d "%i)
-                time.sleep(1)
-                pass
-        else:
-            try:
-                tdLog.notice("Try the last execute sql ")
-                self.affectedRows = self.cursor.execute(sql)
-                return self.affectedRows
-            except Exception as e:
+                if i == queryTimes:
                     caller = inspect.getframeinfo(inspect.stack()[1][0])
                     args = (caller.filename, caller.lineno, sql, repr(e))
                     tdLog.notice("%s(%d) failed: sql:%s, %s" % args)
                     raise Exception(repr(e))                    
+                i+=1
+                time.sleep(1)
+                pass
 
     def checkAffectedRows(self, expectAffectedRows):
         if self.affectedRows != expectAffectedRows:

tests/script/jenkins/basic.txt

@@ -179,6 +179,7 @@
 ./test.sh -f tsim/query/scalarFunction.sim
 ./test.sh -f tsim/query/scalarNull.sim
 ./test.sh -f tsim/query/session.sim
+./test.sh -f tsim/query/udf.sim
 
 # ---- qnode
 ./test.sh -f tsim/qnode/basic1.sim

tests/script/sh/compile_udf.sh

@@ -3,8 +3,8 @@ set +e
 rm -rf /tmp/udf/libbitand.so /tmp/udf/libsqrsum.so
 mkdir -p /tmp/udf
 echo "compile udf bit_and and sqr_sum"
-gcc -fPIC -shared sh/bit_and.c -o /tmp/udf/libbitand.so
-gcc -fPIC -shared sh/sqr_sum.c -o /tmp/udf/libsqrsum.so
+gcc -fPIC -shared sh/bit_and.c -I../../include/libs/function/ -I../../include/client -I../../include/util   -o /tmp/udf/libbitand.so
+gcc -fPIC -shared sh/sqr_sum.c -I../../include/libs/function/ -I../../include/client -I../../include/util   -o /tmp/udf/libsqrsum.so
 echo "debug show /tmp/udf/*.so"
 ls /tmp/udf/*.so
 

tests/script/tsim/parser/function.sim

@@ -327,8 +327,8 @@ endi
 
 print =================>td-2610
 sql select stddev(k) from tm2 where ts='2020-12-29 18:46:19.109'
-if $rows != 0 then
-  print expect 0, actual:$rows
+if $rows != 1 then
+  print expect 1, actual:$rows
   return -1
 endi
 sql select twa(k) from tm2 where ts='2020-12-29 18:46:19.109'
@@ -406,7 +406,7 @@ if $data00 != 1.378704626 then
 endi
 
 sql select stddev(c) from m1
-if $rows != 0 then
+if $rows != 1 then
   return -1
 endi
 

tests/script/tsim/sync/3Replica1VgElect.sim

@@ -48,7 +48,7 @@ endi
 $replica = 3
 $vgroups = 1
 
-print ============= create database
+print ============= create database db
 sql create database db replica $replica vgroups $vgroups
 
 $loop_cnt = 0
@@ -135,7 +135,7 @@ endw
 $totalTblNum = $tbNum * 2
 sleep 1000
 sql show tables
-print ====> expect $totalTblNum and infinsert $rows  in fact 
+print ====> expect $totalTblNum and insert $rows  in fact 
 if $rows != $totalTblNum then 
   return -1
 endi

tests/script/tsim/sync/3Replica5VgElect.sim

@@ -202,7 +202,7 @@ else
 endi
 
 vg_ready:
-print ====>  create stable/child table
+print ====>  create stable stb  /child table ctb and general table ntb
 sql create table stb (ts timestamp, c1 int, c2 float, c3 binary(10)) tags (t1 int)
 
 sql show stables
@@ -225,7 +225,7 @@ endw
 $totalTblNum = $tbNum * 2
 sleep 1000
 sql show tables
-print ====> expect $totalTblNum and infinsert $rows  in fact 
+print ====> expect $totalTblNum and insert $rows  in fact 
 if $rows != $totalTblNum then 
   return -1
 endi
@@ -237,6 +237,7 @@ sql show vgroups
 $dnodeId = $data[0][3]
 $dnodeId = dnode . $dnodeId
 
+print ====> switch_leader_to_offine_loop
 switch_leader_to_offine_loop:
 
 print $dnodeId
@@ -271,22 +272,7 @@ system sh/exec.sh -n $dnodeId -s start
 $switch_loop_cnt = $switch_loop_cnt + 1
 print $switch_loop_cnt
 
-if $switch_loop_cnt == 1 then
-  sql show vgroups
-  $dnodeId = $data[0][3]
-  $dnodeId = dnode . $dnodeId
-  goto switch_leader_to_offine_loop
-elif $switch_loop_cnt == 2 then
-  sql show vgroups
-  $dnodeId = $data[0][3]
-  $dnodeId = dnode . $dnodeId
-  goto switch_leader_to_offine_loop
-elif $switch_loop_cnt == 3 then
-  sql show vgroups
-  $dnodeId = $data[0][3]
-  $dnodeId = dnode . $dnodeId
-  goto switch_leader_to_offine_loop
-elif $switch_loop_cnt == 4 then
+if $switch_loop_cnt <= 4 then
   sql show vgroups
   $dnodeId = $data[0][3]  
   $dnodeId = dnode . $dnodeId

tests/system-test/1-insert/performanceInsert.json

 {
     "filetype": "insert",
     "cfgdir": "/etc/taos/",
-    "host": "test216",
+    "host": "localhost",
     "port": 6030,
     "user": "root",
     "password": "taosdata",
@@ -10,14 +10,14 @@
     "result_file": "./insert_res.txt",
     "confirm_parameter_prompt": "no",
     "insert_interval": 0,
-    "interlace_rows": 1000,
+    "interlace_rows": 0,
     "num_of_records_per_req": 100000,
     "databases": [
         {
             "dbinfo": {
                 "name": "db",
                 "drop": "yes",
-                "vgroups": 24
+                "vgroups": 8
             },
             "super_tables": [
                 {
@@ -29,8 +29,8 @@
                     "batch_create_tbl_num": 50000,
                     "data_source": "rand",
                     "insert_mode": "taosc",
-                    "insert_rows": 5,
-                    "interlace_rows": 100000,
+                    "insert_rows": 1000,
+                    "interlace_rows": 0,
                     "insert_interval": 0,
                     "max_sql_len": 10000000,
                     "disorder_ratio": 0,

tests/system-test/2-query/avg.py

@@ -295,7 +295,7 @@ class TDTestCase:
         tdSql.checkData(0, 0, 4.500000000)
 
         tdSql.query(f" select avg(c1) from {dbname}.stb1 where c1 is null ")
-        tdSql.checkRows(0)
+        tdSql.checkRows(1)
 
 
     def avg_func_filter(self, dbname="db"):

tests/system-test/2-query/distribute_agg_apercentile.py

@@ -86,7 +86,7 @@ class TDTestCase:
     def distribute_agg_query(self, dbname="testdb"):
         # basic filter
         tdSql.query(f"select apercentile(c1 , 20) from {dbname}.stb1 where c1 is null")
-        tdSql.checkRows(0)
+        tdSql.checkRows(1)
 
         tdSql.query(f"select apercentile(c1 , 20) from {dbname}.stb1 where t1=1")
         tdSql.checkData(0,0,2.800000000)

tests/system-test/2-query/distribute_agg_max.py

@@ -168,7 +168,7 @@ class TDTestCase:
     def distribute_agg_query(self, dbname="testdb"):
         # basic filter
         tdSql.query(f"select max(c1) from {dbname}.stb1 where c1 is null")
-        tdSql.checkRows(0)
+        tdSql.checkRows(1)
 
         tdSql.query(f"select max(c1) from {dbname}.stb1 where t1=1")
         tdSql.checkData(0,0,10)

tests/system-test/2-query/distribute_agg_min.py

@@ -167,7 +167,7 @@ class TDTestCase:
     def distribute_agg_query(self, dbname="testdb"):
         # basic filter
         tdSql.query(f"select min(c1) from {dbname}.stb1 where c1 is null")
-        tdSql.checkRows(0)
+        tdSql.checkRows(1)
 
         tdSql.query(f"select min(c1) from {dbname}.stb1 where t1=1")
         tdSql.checkData(0,0,2)

tests/system-test/2-query/sample.py

@@ -743,7 +743,7 @@ class TDTestCase:
         # filter data
 
         tdSql.query(" select sample(c1, 20 ) from t1 where c1 is null ")
-        tdSql.checkRows(0)
+        tdSql.checkRows(1)
 
         tdSql.query(" select sample(c1, 20 ) from t1 where c1 =6 ")
         tdSql.checkRows(1)

tests/system-test/2-query/tsbsQuery.py

@@ -4,6 +4,8 @@ import sys
 import datetime
 import inspect
 import random
+from util.dnodes import TDDnode
+from util.dnodes import tdDnodes
 
 from util.log import *
 from util.sql import *
@@ -30,10 +32,10 @@ class TDTestCase:
             for i in range(ctbNum):
                 tagValue = 'beijing'
                 if (i % 10 == 0):
-                    sql += " %s%d using %s (name,fleet,driver,device_version)  tags('truck_%d', 'South%d','Trish%d','v2.%d')"%(ctbPrefix,i,stbName,i,i,i,i)
+                    sql += " %s%d using %s (name,fleet,driver,device_version,load_capacity,fuel_capacity,nominal_fuel_consumption)  tags('truck_%d', 'South%d','Trish%d','v2.%d', 1500+%d*20, 150+%d*2, 5+%d)"%(ctbPrefix,i,stbName,i,i,i,i,(1500+i*20),(150+i*2),(5+i))
                 else:
                     model = 'H-%d'%i
-                    sql += " %s%d using %s tags('truck_%d', 'South%d','Trish%d','%s','v2.%d')"%(ctbPrefix,i,stbName,i,i,i,model,i)
+                    sql += " %s%d using %s tags('truck_%d', 'South%d','Trish%d','%s','v2.%d', %d, %d,%d)"%(ctbPrefix,i,stbName,i,i,i,model,i,(1500+i*20),(150+i*2),(5+i))
                 if (i > 0) and (i%1000 == 0):
                     tsql.execute(sql)
                     sql = pre_create
@@ -55,10 +57,10 @@ class TDTestCase:
                 sql += " %s%d values "%(ctbPrefix,i)
                 for j in range(rowsPerTbl):
                     if(ctbPrefix=="rct"):
-                        sql += f"({startTs+j*60000}, {80+j}, {90+j}, {85+j}, {30+j*10}, {1.2*j}, {221+j*2}, {20+j*0.2}, {1500+j*20}, {150+j*2},{5+j}) "
+                        sql += f"({startTs+j*60000}, {80+j}, {90+j}, {85+j}, {30+j*10}, {1.2*j}, {221+j*2}, {20+j*0.2}) "
                     elif ( ctbPrefix=="dct"):
                         status= random.randint(0,1)
-                        sql += f"( {startTs+j*60000}, {1+j*0.1},{1400+j*15},  {status},{1500+j*20}, {150+j*2},{5+j} ) "   
+                        sql += f"( {startTs+j*60000}, {1+j*0.1},{1400+j*15},  {status} ) "   
                     # tdLog.debug("1insert sql:%s"%sql)
                     if (j > 0) and ((j%batchNum == 0) or (j == rowsPerTbl - 1)):
                         # tdLog.debug("2insert sql:%s"%sql)
@@ -79,22 +81,22 @@ class TDTestCase:
         stabname2="diagnostics"
         ctbnamePre1="rct"        
         ctbnamePre2="dct"
-        ctbNums=40
+        ctbNums=50
         self.ctbNums=ctbNums
-        rowNUms=200
+        rowNUms=5000
         ts=1451606400000
         tdSql.execute(f"create database {dbname};")
         tdSql.execute(f"use {dbname} ")
         tdSql.execute(f'''
-        create table {stabname1} (ts timestamp,latitude double,longitude double,elevation double,velocity double,heading double,grade double,fuel_consumption double,load_capacity double,fuel_capacity double,nominal_fuel_consumption double) tags (name binary(30),fleet binary(30),driver binary(30),model binary(30),device_version binary(30));
+        create table {stabname1} (ts timestamp,latitude double,longitude double,elevation double,velocity double,heading double,grade double,fuel_consumption double) tags (name binary(30),fleet binary(30),driver binary(30),model binary(30),device_version binary(30),load_capacity double,fuel_capacity double,nominal_fuel_consumption double);
         ''')
         tdSql.execute(f'''
-        create table {stabname2} (ts timestamp,fuel_state double,current_load double,status bigint,load_capacity double,fuel_capacity double,nominal_fuel_consumption double) tags (name binary(30),fleet binary(30),driver binary(30),model binary(30),device_version binary(30)) ;
+        create table {stabname2} (ts timestamp,fuel_state double,current_load double,status bigint) tags (name binary(30),fleet binary(30),driver binary(30),model binary(30),device_version binary(30),load_capacity double,fuel_capacity double,nominal_fuel_consumption double) ;
         ''')
         self.create_ctable(tsql=tdSql,dbName=dbname,stbName=stabname1,ctbPrefix=ctbnamePre1,ctbNum=ctbNums)
         self.create_ctable(tsql=tdSql,dbName=dbname,stbName=stabname2,ctbPrefix=ctbnamePre2,ctbNum=ctbNums)
-        self.insertData(tsql=tdSql,dbName=dbname,stbName=stabname1,ctbPrefix=ctbnamePre1,ctbNum=ctbNums,rowsPerTbl=rowNUms,startTs=ts,batchNum=1000)
-        self.insertData(tsql=tdSql,dbName=dbname,stbName=stabname2,ctbPrefix=ctbnamePre2,ctbNum=ctbNums,rowsPerTbl=rowNUms,startTs=ts,batchNum=1000)
+        self.insertData(tsql=tdSql,dbName=dbname,stbName=stabname1,ctbPrefix=ctbnamePre1,ctbNum=ctbNums,rowsPerTbl=rowNUms,startTs=ts,batchNum=10000)
+        self.insertData(tsql=tdSql,dbName=dbname,stbName=stabname2,ctbPrefix=ctbnamePre2,ctbNum=ctbNums,rowsPerTbl=rowNUms,startTs=ts,batchNum=10000)
         # for i in range(ctbNum):
         #     if i %10 == 0 :
         #         # tdLog.debug(f"create table rct{i} using readings (name,fleet,driver,model,device_version) tags ('truck_{i}','South{i}','Trish{i}', NULL,'v2.3')")
@@ -131,7 +133,7 @@ class TDTestCase:
     #         tdLog.info("avg value check pass , it work as expected ,sql is \"%s\"   "%check_query )
 
 
-    def tsbsIotQuery(self):
+    def tsbsIotQuery(self,insertinto=True):
         
         tdSql.execute("use db_tsbs")
         
@@ -143,6 +145,7 @@ class TDTestCase:
         
         
         # test insert into 
+        if insertinto == True :
             tdSql.execute("create table testsnode (ts timestamp, c1 float,c2 binary(30),c3 binary(30),c4 binary(30)) ;")
             tdSql.query("insert into testsnode SELECT ts,avg(velocity) as mean_velocity,name,driver,fleet FROM readings WHERE ts > 1451606400000 AND ts <= 1451606460000 partition BY name,driver,fleet,ts interval(10m);")
             
@@ -182,8 +185,7 @@ class TDTestCase:
 
         # # 6. avg-daily-driving-session        
         # #taosc core dumped
-        # tdSql.execute("create table random_measure2_1 (ts timestamp,ela float, name binary(40))")
-        # tdSql.query("SELECT ts,diff(mv) AS difka  FROM (SELECT ts,name,floor(avg(velocity)/10)/floor(avg(velocity)/10) AS mv FROM readings   WHERE name!='' AND ts > '2016-01-01T00:00:00Z' AND ts < '2016-01-05T00:00:01Z'   partition by name,ts interval(10m) fill(value,0))  GROUP BY name,ts;")
+        tdSql.query(" SELECT _wstart as ts,name,floor(avg(velocity)/5) AS mv FROM readings   WHERE name is not null  AND ts > '2016-01-01T00:00:00Z' AND ts < '2016-01-05T00:00:01Z'   partition by name interval(10m) fill(value,0);")
         # tdSql.query("select name,diff(mv) AS difka  FROM (SELECT  ts,name,mv  FROM (SELECT _wstart as ts,name,floor(avg(velocity)/10)/floor(avg(velocity)/10) AS mv FROM readings   WHERE name!='' AND ts > '2016-01-01T00:00:00Z' AND ts < '2016-01-05T00:00:01Z'   partition by name interval(10m) fill(value,0)))  group  BY name ;")
         # tdSql.query("SELECT _wstart,name,floor(avg(velocity)/10)/floor(avg(velocity)/10) AS mv FROM readings   WHERE name!='' AND ts > '2016-01-01T00:00:00Z' AND ts < '2016-01-05T00:00:01Z'   partition by name interval(10m) fill(value,0)")
 
@@ -234,7 +236,9 @@ class TDTestCase:
         tdLog.printNoPrefix("==========step1:create database and table,insert data  ==============")
         self.prepareData()
         self.tsbsIotQuery()
-
+        tdDnodes.stop(1)
+        tdDnodes.start(1)
+        self.tsbsIotQuery(False)
 
     def stop(self):
         tdSql.close()

tests/system-test/fulltest.sh

@@ -219,7 +219,7 @@ python3 test.py -f 6-cluster/vnode/4dnode1mnode_basic_replica3_insertdatas_query
 # python3 test.py -f 6-cluster/vnode/4dnode1mnode_basic_replica3_querydatas_stop_follower.py -N 4 -M 1
 # python3 test.py -f 6-cluster/vnode/4dnode1mnode_basic_replica3_querydatas_stop_leader_force_stop.py -N 4 -M 1
 # python3 test.py -f 6-cluster/vnode/4dnode1mnode_basic_replica3_querydatas_stop_leader.py -N 4 -M 1
-python3 test.py -f 6-cluster/vnode/4dnode1mnode_basic_replica3_vgroups.py  -N 4 -M 1
+# python3 test.py -f 6-cluster/vnode/4dnode1mnode_basic_replica3_vgroups.py  -N 4 -M 1
 # python3 test.py -f 6-cluster/vnode/4dnode1mnode_basic_replica3_vgroups_stopOne.py -N 4 -M 1