Merge remote-tracking branch 'origin/main' into fix/ctgErrCode

Jenkinsfile2

@@ -387,7 +387,7 @@ pipeline {
                     }
                     steps {
                         catchError(buildResult: 'FAILURE', stageResult: 'FAILURE') {
-                            timeout(time: 55, unit: 'MINUTES'){
+                            timeout(time: 75, unit: 'MINUTES'){
                                 pre_test_win()
                                 pre_test_build_win()
                                 run_win_ctest()

README-CN.md

@@ -276,7 +276,7 @@ sudo make install
 安装成功后，可以在应用程序中双击 TDengine 图标启动服务，或者在终端中启动 TDengine 服务：
 
 ```bash
-launchctl start com.tdengine.taosd
+sudo launchctl start com.tdengine.taosd
 ```
 
 用户可以使用 TDengine CLI 来连接 TDengine 服务，在终端中，输入：

README.md

@@ -14,6 +14,12 @@
 [![Build status](https://ci.appveyor.com/api/projects/status/kf3pwh2or5afsgl9/branch/master?svg=true)](https://ci.appveyor.com/project/sangshuduo/tdengine-2n8ge/branch/master)
 [![Coverage Status](https://coveralls.io/repos/github/taosdata/TDengine/badge.svg?branch=develop)](https://coveralls.io/github/taosdata/TDengine?branch=develop)
 [![CII Best Practices](https://bestpractices.coreinfrastructure.org/projects/4201/badge)](https://bestpractices.coreinfrastructure.org/projects/4201)
+<br />
+[![Twitter Follow](https://img.shields.io/twitter/follow/tdenginedb?label=TDengine&style=social)](https://twitter.com/tdenginedb)
+[![YouTube Channel](https://img.shields.io/badge/Subscribe_@tdengine--white?logo=youtube&style=social)](https://www.youtube.com/@tdengine)
+[![Discord Community](https://img.shields.io/badge/Join_Discord--white?logo=discord&style=social)](https://discord.com/invite/VZdSuUg4pS)
+[![LinkedIn](https://img.shields.io/badge/Follow_LinkedIn--white?logo=linkedin&style=social)](https://www.linkedin.com/company/tdengine)
+[![StackOverflow](https://img.shields.io/badge/Ask_StackOverflow--white?logo=stackoverflow&style=social&logoColor=orange)](https://stackoverflow.com/questions/tagged/tdengine)
 
 English | [简体中文](README-CN.md) | [TDengine Cloud](https://cloud.tdengine.com) | [Learn more about TSDB](https://tdengine.com/tsdb/)
 
@@ -280,7 +286,7 @@ Installing from source code will also configure service management for TDengine.
 To start the service after installation, double-click the /applications/TDengine to start the program, or in a terminal, use:
 
 ```bash
-launchctl start com.tdengine.taosd
+sudo launchctl start com.tdengine.taosd
 ```
 
 Then users can use the TDengine CLI to connect the TDengine server. In a terminal, use:

cmake/cmake.version

@@ -2,7 +2,7 @@
 IF (DEFINED VERNUMBER)
   SET(TD_VER_NUMBER ${VERNUMBER})
 ELSE ()
-  SET(TD_VER_NUMBER "3.0.2.4")
+  SET(TD_VER_NUMBER "3.0.2.5")
 ENDIF ()
 
 IF (DEFINED VERCOMPATIBLE)

cmake/taosadapter_CMakeLists.txt.in

@@ -2,7 +2,7 @@
 # taosadapter
 ExternalProject_Add(taosadapter
         GIT_REPOSITORY https://github.com/taosdata/taosadapter.git
-        GIT_TAG 3e08996
+        GIT_TAG db6c843
         SOURCE_DIR "${TD_SOURCE_DIR}/tools/taosadapter"
         BINARY_DIR ""
         #BUILD_IN_SOURCE TRUE

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 c4a567b
+        GIT_TAG 61cbfd2
         SOURCE_DIR "${TD_SOURCE_DIR}/tools/taos-tools"
         BINARY_DIR ""
         #BUILD_IN_SOURCE TRUE

docs/en/12-taos-sql/06-select.md

@@ -12,6 +12,7 @@ SELECT [DISTINCT] select_list
     from_clause
     [WHERE condition]
     [partition_by_clause]
+    [interp_clause]
     [window_clause]
     [group_by_clause]
     [order_by_clasue]
@@ -52,8 +53,11 @@ window_clause: {
   | STATE_WINDOW(col)
   | INTERVAL(interval_val [, interval_offset]) [SLIDING (sliding_val)] [WATERMARK(watermark_val)] [FILL(fill_mod_and_val)]
 
+interp_clause:
+    RANGE(ts_val, ts_val), EVERY(every_val), FILL(fill_mod_and_val)
+
 partition_by_clause:
-    PARTITION BY expr [, expr] ... 
+    PARTITION BY expr [, expr] ...
 
 group_by_clause:
     GROUP BY expr [, expr] ... HAVING condition

docs/en/12-taos-sql/10-function.md

@@ -872,9 +872,9 @@ INTERP(expr)
 - `INTERP` is used to get the value that matches the specified time slice from a column. If no such value exists an interpolation value will be returned based on `FILL` parameter.
 - The input data of `INTERP` is the value of the specified column and a `where` clause can be used to filter the original data. If no `where` condition is specified then all original data is the input.
 - `INTERP` must be used along with `RANGE`, `EVERY`, `FILL` keywords.
-- The output time range of `INTERP` is specified by `RANGE(timestamp1,timestamp2)` parameter, with timestamp1<=timestamp2. timestamp1 is the starting point of the output time range and must be specified. timestamp2 is the ending point of the output time range and must be specified.
+- The output time range of `INTERP` is specified by `RANGE(timestamp1,timestamp2)` parameter, with timestamp1 <= timestamp2. timestamp1 is the starting point of the output time range and must be specified. timestamp2 is the ending point of the output time range and must be specified.
-- The number of rows in the result set of `INTERP` is determined by the parameter `EVERY`. Starting from timestamp1, one interpolation is performed for every time interval specified `EVERY` parameter. The parameter `EVERY` must be an integer, with no quotes, with a time unit of: b(nanosecond), u(microsecond), a(millisecond)), s(second), m(minute), h(hour), d(day), or w(week). For example, `EVERY(500a)` will interpolate every 500 milliseconds.
+- The number of rows in the result set of `INTERP` is determined by the parameter `EVERY(time_unit)`. Starting from timestamp1, one interpolation is performed for every time interval specified `time_unit` parameter. The parameter `time_unit` must be an integer, with no quotes, with a time unit of: a(millisecond)), s(second), m(minute), h(hour), d(day), or w(week). For example, `EVERY(500a)` will interpolate every 500 milliseconds.
-- Interpolation is performed based on `FILL` parameter.
+- Interpolation is performed based on `FILL` parameter. For more information about FILL clause, see [FILL Clause](../distinguished/#fill-clause).
 - `INTERP` can only be used to interpolate in single timeline. So it must be used with `partition by tbname` when it's used on a STable.
 - Pseudo column `_irowts` can be used along with `INTERP` to return the timestamps associated with interpolation points(support after version 3.0.1.4).
 

docs/examples/go/go.mod

@@ -0,0 +1,6 @@
+module goexample
+
+go 1.17
+
+require github.com/taosdata/driver-go/v3 v3.1.0
+

docs/zh/12-taos-sql/06-select.md

@@ -13,6 +13,7 @@ SELECT [DISTINCT] select_list
     from_clause
     [WHERE condition]
     [partition_by_clause]
+    [interp_clause]
     [window_clause]
     [group_by_clause]
     [order_by_clasue]
@@ -53,8 +54,11 @@ window_clause: {
   | STATE_WINDOW(col)
   | INTERVAL(interval_val [, interval_offset]) [SLIDING (sliding_val)] [WATERMARK(watermark_val)] [FILL(fill_mod_and_val)]
 
+interp_clause:
+    RANGE(ts_val, ts_val), EVERY(every_val), FILL(fill_mod_and_val)
+
 partition_by_clause:
-    PARTITION BY expr [, expr] ... 
+    PARTITION BY expr [, expr] ...
 
 group_by_clause:
     GROUP BY expr [, expr] ... HAVING condition

docs/zh/12-taos-sql/10-function.md

@@ -875,9 +875,9 @@ INTERP(expr)
 - INTERP 用于在指定时间断面获取指定列的记录值，如果该时间断面不存在符合条件的行数据，那么会根据 FILL 参数的设定进行插值。
 - INTERP 的输入数据为指定列的数据，可以通过条件语句（where 子句）来对原始列数据进行过滤，如果没有指定过滤条件则输入为全部数据。
 - INTERP 需要同时与 RANGE，EVERY 和 FILL 关键字一起使用。
-- INTERP 的输出时间范围根据 RANGE(timestamp1,timestamp2)字段来指定，需满足 timestamp1<=timestamp2。其中 timestamp1（必选值）为输出时间范围的起始值，即如果 timestamp1 时刻符合插值条件则 timestamp1 为输出的第一条记录，timestamp2（必选值）为输出时间范围的结束值，即输出的最后一条记录的 timestamp 不能大于 timestamp2。
+- INTERP 的输出时间范围根据 RANGE(timestamp1,timestamp2)字段来指定，需满足 timestamp1 <= timestamp2。其中 timestamp1（必选值）为输出时间范围的起始值，即如果 timestamp1 时刻符合插值条件则 timestamp1 为输出的第一条记录，timestamp2（必选值）为输出时间范围的结束值，即输出的最后一条记录的 timestamp 不能大于 timestamp2。
-- INTERP 根据 EVERY 字段来确定输出时间范围内的结果条数，即从 timestamp1 开始每隔固定长度的时间（EVERY 值）进行插值。
+- INTERP 根据 EVERY(time_unit) 字段来确定输出时间范围内的结果条数，即从 timestamp1 开始每隔固定长度的时间（time_unit 值）进行插值，time_unit 可取值时间单位：1a(毫秒)，1s(秒)，1m(分)，1h(小时)，1d(天)，1w(周)。例如 EVERY(500a) 将对于指定数据每500毫秒间隔进行一次插值.
-- INTERP 根据 FILL 字段来决定在每个符合输出条件的时刻如何进行插值。
+- INTERP 根据 FILL 字段来决定在每个符合输出条件的时刻如何进行插值。关于 FILL 子句如何使用请参考 [FILL 子句](../distinguished/#fill-子句)
 - INTERP 只能在一个时间序列内进行插值，因此当作用于超级表时必须跟 partition by tbname 一起使用。
 - INTERP 可以与伪列 _irowts 一起使用，返回插值点所对应的时间戳(3.0.1.4版本以后支持)。
 

examples/JDBC/readme.md

@@ -10,4 +10,4 @@
 |  6   |      taosdemo      | This is an internal tool for testing Our JDBC-JNI, JDBC-RESTful, RESTful interfaces |
 
 
-more detail: https://docs.taosdata.com/reference/connector/java/
+more detail: https://docs.taosdata.com/connector/java/

examples/c/tmq.c

@@ -189,27 +189,46 @@ void tmq_commit_cb_print(tmq_t* tmq, int32_t code, void* param) {
 
 tmq_t* build_consumer() {
   tmq_conf_res_t code;
+  tmq_t* tmq = NULL;
+
   tmq_conf_t*    conf = tmq_conf_new();
   code = tmq_conf_set(conf, "enable.auto.commit", "true");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
   code = tmq_conf_set(conf, "auto.commit.interval.ms", "1000");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
   code = tmq_conf_set(conf, "group.id", "cgrpName");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
   code = tmq_conf_set(conf, "client.id", "user defined name");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
   code = tmq_conf_set(conf, "td.connect.user", "root");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
   code = tmq_conf_set(conf, "td.connect.pass", "taosdata");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
   code = tmq_conf_set(conf, "auto.offset.reset", "earliest");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
   code = tmq_conf_set(conf, "experimental.snapshot.enable", "false");
-  if (TMQ_CONF_OK != code) return NULL;
+  if (TMQ_CONF_OK != code) {
+    goto _end;
+  }
 
   tmq_conf_set_auto_commit_cb(conf, tmq_commit_cb_print, NULL);
+  tmq = tmq_consumer_new(conf, NULL, 0);
 
-  tmq_t* tmq = tmq_consumer_new(conf, NULL, 0);
+  _end:
   tmq_conf_destroy(conf);
   return tmq;
 }

include/common/tcommon.h

@@ -195,7 +195,7 @@ typedef struct SDataBlockInfo {
   uint32_t    capacity;
   SBlockID    id;
   int16_t     hasVarCol;
-  int16_t     dataLoad;   // denote if the data is loaded or not
+  int16_t     dataLoad;  // denote if the data is loaded or not
 
   // TODO: optimize and remove following
   int64_t     version;    // used for stream, and need serialization
@@ -204,8 +204,8 @@ typedef struct SDataBlockInfo {
   STimeWindow calWin;     // used for stream, do not serialize
   TSKEY       watermark;  // used for stream
 
-  char        parTbName[TSDB_TABLE_NAME_LEN];  // used for stream partition
+  char  parTbName[TSDB_TABLE_NAME_LEN];  // used for stream partition
-  STag*       pTag;                            // used for stream partition
+  STag* pTag;                            // used for stream partition
 } SDataBlockInfo;
 
 typedef struct SSDataBlock {
@@ -239,22 +239,22 @@ typedef struct SVarColAttr {
 // pBlockAgg->numOfNull == info.rows, all data are null
 // pBlockAgg->numOfNull == 0, no data are null.
 typedef struct SColumnInfoData {
-  char*         pData;       // the corresponding block data in memory
+  char* pData;  // the corresponding block data in memory
   union {
     char*       nullbitmap;  // bitmap, one bit for each item in the list
     SVarColAttr varmeta;
   };
-  SColumnInfo   info;        // column info
+  SColumnInfo info;     // column info
-  bool          hasNull;     // if current column data has null value.
+  bool        hasNull;  // if current column data has null value.
 } SColumnInfoData;
 
 typedef struct SQueryTableDataCond {
   uint64_t     suid;
-  int32_t      order;    // desc|asc order to iterate the data block
+  int32_t      order;  // desc|asc order to iterate the data block
   int32_t      numOfCols;
   SColumnInfo* colList;
-  int32_t*     pSlotList; // the column output destation slot, and it may be null
+  int32_t*     pSlotList;  // the column output destation slot, and it may be null
-  int32_t      type;     // data block load type:
+  int32_t      type;       // data block load type:
   STimeWindow  twindows;
   int64_t      startVersion;
   int64_t      endVersion;
@@ -300,6 +300,7 @@ typedef struct STableBlockDistInfo {
   int32_t  firstSeekTimeUs;
   uint32_t numOfInmemRows;
   uint32_t numOfSmallBlocks;
+  uint32_t numOfVgroups;
   int32_t  blockRowsHisto[20];
 } STableBlockDistInfo;
 
@@ -367,6 +368,12 @@ typedef struct SSortExecInfo {
   int32_t readBytes;   // read io bytes
 } SSortExecInfo;
 
+typedef struct STUidTagInfo {
+  char*    name;
+  uint64_t uid;
+  void*    pTagVal;
+} STUidTagInfo;
+
 // stream special block column
 
 #define START_TS_COLUMN_INDEX           0

include/common/tglobal.h

@@ -49,6 +49,7 @@ extern int32_t tsTagFilterResCacheSize;
 
 // queue & threads
 extern int32_t tsNumOfRpcThreads;
+extern int32_t tsNumOfRpcSessions;
 extern int32_t tsNumOfCommitThreads;
 extern int32_t tsNumOfTaskQueueThreads;
 extern int32_t tsNumOfMnodeQueryThreads;
@@ -86,9 +87,9 @@ extern int32_t  tsTelemInterval;
 extern char     tsTelemServer[];
 extern uint16_t tsTelemPort;
 extern bool     tsEnableCrashReport;
-extern char*    tsTelemUri;
+extern char    *tsTelemUri;
-extern char*    tsClientCrashReportUri;
+extern char    *tsClientCrashReportUri;
-extern char*    tsSvrCrashReportUri;
+extern char    *tsSvrCrashReportUri;
 
 // query buffer management
 extern int32_t tsQueryBufferSize;  // maximum allowed usage buffer size in MB for each data node during query processing
@@ -159,6 +160,8 @@ extern int32_t tsUptimeInterval;
 extern int32_t tsRpcRetryLimit;
 extern int32_t tsRpcRetryInterval;
 
+extern bool tsDisableStream;
+
 // #define NEEDTO_COMPRESSS_MSG(size) (tsCompressMsgSize != -1 && (size) > tsCompressMsgSize)
 
 int32_t taosCreateLog(const char *logname, int32_t logFileNum, const char *cfgDir, const char **envCmd,

include/common/tmisce.h

@@ -27,13 +27,14 @@ typedef struct SCorEpSet {
   SEpSet  epSet;
 } SCorEpSet;
 
+#define GET_ACTIVE_EP(_eps) (&((_eps)->eps[(_eps)->inUse]))
 int32_t taosGetFqdnPortFromEp(const char* ep, SEp* pEp);
 void    addEpIntoEpSet(SEpSet* pEpSet, const char* fqdn, uint16_t port);
 
-bool isEpsetEqual(const SEpSet* s1, const SEpSet* s2);
+bool    isEpsetEqual(const SEpSet* s1, const SEpSet* s2);
-
+void    epsetAssign(SEpSet* dst, const SEpSet* pSrc);
-void   updateEpSet_s(SCorEpSet* pEpSet, SEpSet* pNewEpSet);
+void    updateEpSet_s(SCorEpSet* pEpSet, SEpSet* pNewEpSet);
-SEpSet getEpSet_s(SCorEpSet* pEpSet);
+SEpSet  getEpSet_s(SCorEpSet* pEpSet);
 
 #ifdef __cplusplus
 }

include/common/tmsg.h

@@ -58,7 +58,7 @@ extern int32_t tMsgDict[];
 #define TMSG_INFO(TYPE)                                                                                                \
   ((TYPE) < TDMT_DND_MAX_MSG || (TYPE) < TDMT_MND_MAX_MSG || (TYPE) < TDMT_VND_MAX_MSG || (TYPE) < TDMT_SCH_MAX_MSG || \
    (TYPE) < TDMT_STREAM_MAX_MSG || (TYPE) < TDMT_MON_MAX_MSG || (TYPE) < TDMT_SYNC_MAX_MSG) ||                         \
-          (TYPE) < TDMT_VND_STREAM_MSG || (TYPE) < TDMT_VND_TMQ_MSG                                                    \
+          (TYPE) < TDMT_VND_STREAM_MSG || (TYPE) < TDMT_VND_TMQ_MSG || (TYPE) < TDMT_VND_TMQ_MAX_MSG                   \
       ? tMsgInfo[tMsgDict[TMSG_SEG_CODE(TYPE)] + TMSG_SEG_SEQ(TYPE)]                                                   \
       : 0
 
@@ -144,12 +144,14 @@ typedef enum _mgmt_table {
 #define TSDB_ALTER_TABLE_UPDATE_OPTIONS      9
 #define TSDB_ALTER_TABLE_UPDATE_COLUMN_NAME  10
 
-#define TSDB_FILL_NONE      0
+#define TSDB_FILL_NONE        0
-#define TSDB_FILL_NULL      1
+#define TSDB_FILL_NULL        1
-#define TSDB_FILL_SET_VALUE 2
+#define TSDB_FILL_NULL_F      2
-#define TSDB_FILL_LINEAR    3
+#define TSDB_FILL_SET_VALUE   3
-#define TSDB_FILL_PREV      4
+#define TSDB_FILL_SET_VALUE_F 4
-#define TSDB_FILL_NEXT      5
+#define TSDB_FILL_LINEAR      5
+#define TSDB_FILL_PREV        6
+#define TSDB_FILL_NEXT        7
 
 #define TSDB_ALTER_USER_PASSWD                 0x1
 #define TSDB_ALTER_USER_SUPERUSER              0x2
@@ -1752,6 +1754,7 @@ typedef struct {
 #define STREAM_FILL_HISTORY_ON        1
 #define STREAM_FILL_HISTORY_OFF       0
 #define STREAM_DEFAULT_FILL_HISTORY   STREAM_FILL_HISTORY_OFF
+#define STREAM_DEFAULT_IGNORE_UPDATE  0
 
 typedef struct {
   char    name[TSDB_STREAM_FNAME_LEN];
@@ -1769,6 +1772,8 @@ typedef struct {
   SArray* pTags;  // array of SField
   // 3.0.20
   int64_t checkpointFreq;  // ms
+  int64_t deleteMark;
+  int8_t  igUpdate;
 } SCMCreateStreamReq;
 
 typedef struct {
@@ -1825,10 +1830,10 @@ typedef struct {
 } SMqConsumerLostMsg, SMqConsumerRecoverMsg, SMqConsumerClearMsg;
 
 typedef struct {
-  int64_t consumerId;
+  int64_t  consumerId;
-  char    cgroup[TSDB_CGROUP_LEN];
+  char     cgroup[TSDB_CGROUP_LEN];
-  char    clientId[256];
+  char     clientId[256];
-  SArray* topicNames;  // SArray<char**>
+  SArray*  topicNames;  // SArray<char**>
 } SCMSubscribeReq;
 
 static FORCE_INLINE int32_t tSerializeSCMSubscribeReq(void** buf, const SCMSubscribeReq* pReq) {

include/common/ttokendef.h

@@ -209,134 +209,136 @@
 #define TK_IGNORE               191
 #define TK_EXPIRED              192
 #define TK_FILL_HISTORY         193
-#define TK_SUBTABLE             194
+#define TK_UPDATE               194
-#define TK_KILL                 195
+#define TK_SUBTABLE             195
-#define TK_CONNECTION           196
+#define TK_KILL                 196
-#define TK_TRANSACTION          197
+#define TK_CONNECTION           197
-#define TK_BALANCE              198
+#define TK_TRANSACTION          198
-#define TK_VGROUP               199
+#define TK_BALANCE              199
-#define TK_MERGE                200
+#define TK_VGROUP               200
-#define TK_REDISTRIBUTE         201
+#define TK_MERGE                201
-#define TK_SPLIT                202
+#define TK_REDISTRIBUTE         202
-#define TK_DELETE               203
+#define TK_SPLIT                203
-#define TK_INSERT               204
+#define TK_DELETE               204
-#define TK_NULL                 205
+#define TK_INSERT               205
-#define TK_NK_QUESTION          206
+#define TK_NULL                 206
-#define TK_NK_ARROW             207
+#define TK_NK_QUESTION          207
-#define TK_ROWTS                208
+#define TK_NK_ARROW             208
-#define TK_QSTART               209
+#define TK_ROWTS                209
-#define TK_QEND                 210
+#define TK_QSTART               210
-#define TK_QDURATION            211
+#define TK_QEND                 211
-#define TK_WSTART               212
+#define TK_QDURATION            212
-#define TK_WEND                 213
+#define TK_WSTART               213
-#define TK_WDURATION            214
+#define TK_WEND                 214
-#define TK_IROWTS               215
+#define TK_WDURATION            215
-#define TK_CAST                 216
+#define TK_IROWTS               216
-#define TK_NOW                  217
+#define TK_CAST                 217
-#define TK_TODAY                218
+#define TK_NOW                  218
-#define TK_TIMEZONE             219
+#define TK_TODAY                219
-#define TK_CLIENT_VERSION       220
+#define TK_TIMEZONE             220
-#define TK_SERVER_VERSION       221
+#define TK_CLIENT_VERSION       221
-#define TK_SERVER_STATUS        222
+#define TK_SERVER_VERSION       222
-#define TK_CURRENT_USER         223
+#define TK_SERVER_STATUS        223
-#define TK_COUNT                224
+#define TK_CURRENT_USER         224
-#define TK_LAST_ROW             225
+#define TK_COUNT                225
-#define TK_CASE                 226
+#define TK_LAST_ROW             226
-#define TK_END                  227
+#define TK_CASE                 227
-#define TK_WHEN                 228
+#define TK_END                  228
-#define TK_THEN                 229
+#define TK_WHEN                 229
-#define TK_ELSE                 230
+#define TK_THEN                 230
-#define TK_BETWEEN              231
+#define TK_ELSE                 231
-#define TK_IS                   232
+#define TK_BETWEEN              232
-#define TK_NK_LT                233
+#define TK_IS                   233
-#define TK_NK_GT                234
+#define TK_NK_LT                234
-#define TK_NK_LE                235
+#define TK_NK_GT                235
-#define TK_NK_GE                236
+#define TK_NK_LE                236
-#define TK_NK_NE                237
+#define TK_NK_GE                237
-#define TK_MATCH                238
+#define TK_NK_NE                238
-#define TK_NMATCH               239
+#define TK_MATCH                239
-#define TK_CONTAINS             240
+#define TK_NMATCH               240
-#define TK_IN                   241
+#define TK_CONTAINS             241
-#define TK_JOIN                 242
+#define TK_IN                   242
-#define TK_INNER                243
+#define TK_JOIN                 243
-#define TK_SELECT               244
+#define TK_INNER                244
-#define TK_DISTINCT             245
+#define TK_SELECT               245
-#define TK_WHERE                246
+#define TK_DISTINCT             246
-#define TK_PARTITION            247
+#define TK_WHERE                247
-#define TK_BY                   248
+#define TK_PARTITION            248
-#define TK_SESSION              249
+#define TK_BY                   249
-#define TK_STATE_WINDOW         250
+#define TK_SESSION              250
-#define TK_SLIDING              251
+#define TK_STATE_WINDOW         251
-#define TK_FILL                 252
+#define TK_SLIDING              252
-#define TK_VALUE                253
+#define TK_FILL                 253
-#define TK_NONE                 254
+#define TK_VALUE                254
-#define TK_PREV                 255
+#define TK_VALUE_F              255
-#define TK_LINEAR               256
+#define TK_NONE                 256
-#define TK_NEXT                 257
+#define TK_PREV                 257
-#define TK_HAVING               258
+#define TK_NULL_F               258
-#define TK_RANGE                259
+#define TK_LINEAR               259
-#define TK_EVERY                260
+#define TK_NEXT                 260
-#define TK_ORDER                261
+#define TK_HAVING               261
-#define TK_SLIMIT               262
+#define TK_RANGE                262
-#define TK_SOFFSET              263
+#define TK_EVERY                263
-#define TK_LIMIT                264
+#define TK_ORDER                264
-#define TK_OFFSET               265
+#define TK_SLIMIT               265
-#define TK_ASC                  266
+#define TK_SOFFSET              266
-#define TK_NULLS                267
+#define TK_LIMIT                267
-#define TK_ABORT                268
+#define TK_OFFSET               268
-#define TK_AFTER                269
+#define TK_ASC                  269
-#define TK_ATTACH               270
+#define TK_NULLS                270
-#define TK_BEFORE               271
+#define TK_ABORT                271
-#define TK_BEGIN                272
+#define TK_AFTER                272
-#define TK_BITAND               273
+#define TK_ATTACH               273
-#define TK_BITNOT               274
+#define TK_BEFORE               274
-#define TK_BITOR                275
+#define TK_BEGIN                275
-#define TK_BLOCKS               276
+#define TK_BITAND               276
-#define TK_CHANGE               277
+#define TK_BITNOT               277
-#define TK_COMMA                278
+#define TK_BITOR                278
-#define TK_COMPACT              279
+#define TK_BLOCKS               279
-#define TK_CONCAT               280
+#define TK_CHANGE               280
-#define TK_CONFLICT             281
+#define TK_COMMA                281
-#define TK_COPY                 282
+#define TK_COMPACT              282
-#define TK_DEFERRED             283
+#define TK_CONCAT               283
-#define TK_DELIMITERS           284
+#define TK_CONFLICT             284
-#define TK_DETACH               285
+#define TK_COPY                 285
-#define TK_DIVIDE               286
+#define TK_DEFERRED             286
-#define TK_DOT                  287
+#define TK_DELIMITERS           287
-#define TK_EACH                 288
+#define TK_DETACH               288
-#define TK_FAIL                 289
+#define TK_DIVIDE               289
-#define TK_FILE                 290
+#define TK_DOT                  290
-#define TK_FOR                  291
+#define TK_EACH                 291
-#define TK_GLOB                 292
+#define TK_FAIL                 292
-#define TK_ID                   293
+#define TK_FILE                 293
-#define TK_IMMEDIATE            294
+#define TK_FOR                  294
-#define TK_IMPORT               295
+#define TK_GLOB                 295
-#define TK_INITIALLY            296
+#define TK_ID                   296
-#define TK_INSTEAD              297
+#define TK_IMMEDIATE            297
-#define TK_ISNULL               298
+#define TK_IMPORT               298
-#define TK_KEY                  299
+#define TK_INITIALLY            299
-#define TK_MODULES              300
+#define TK_INSTEAD              300
-#define TK_NK_BITNOT            301
+#define TK_ISNULL               301
-#define TK_NK_SEMI              302
+#define TK_KEY                  302
-#define TK_NOTNULL              303
+#define TK_MODULES              303
-#define TK_OF                   304
+#define TK_NK_BITNOT            304
-#define TK_PLUS                 305
+#define TK_NK_SEMI              305
-#define TK_PRIVILEGE            306
+#define TK_NOTNULL              306
-#define TK_RAISE                307
+#define TK_OF                   307
-#define TK_REPLACE              308
+#define TK_PLUS                 308
-#define TK_RESTRICT             309
+#define TK_PRIVILEGE            309
-#define TK_ROW                  310
+#define TK_RAISE                310
-#define TK_SEMI                 311
+#define TK_REPLACE              311
-#define TK_STAR                 312
+#define TK_RESTRICT             312
-#define TK_STATEMENT            313
+#define TK_ROW                  313
-#define TK_STRICT               314
+#define TK_SEMI                 314
-#define TK_STRING               315
+#define TK_STAR                 315
-#define TK_TIMES                316
+#define TK_STATEMENT            316
-#define TK_UPDATE               317
+#define TK_STRICT               317
-#define TK_VALUES               318
+#define TK_STRING               318
-#define TK_VARIABLE             319
+#define TK_TIMES                319
-#define TK_VIEW                 320
+#define TK_VALUES               320
-#define TK_WAL                  321
+#define TK_VARIABLE             321
+#define TK_VIEW                 322
+#define TK_WAL                  323
 
 #define TK_NK_SPACE   600
 #define TK_NK_COMMENT 601

include/libs/function/function.h

@@ -132,14 +132,16 @@ typedef struct SqlFunctionCtx {
   SInputColumnInfoData input;
   SResultDataInfo      resDataInfo;
   uint32_t             order;       // data block scanner order: asc|desc
+  uint8_t              isPseudoFunc;// denote current function is pseudo function or not [added for perf reason]
+  uint8_t              isNotNullFunc;// not return null value.
   uint8_t              scanFlag;    // record current running step, default: 0
   int16_t              functionId;  // function id
   char                *pOutput;     // final result output buffer, point to sdata->data
-  int32_t              numOfParams;
   // input parameter, e.g., top(k, 20), the number of results of top query is kept in param
   SFunctParam *param;
   // corresponding output buffer for timestamp of each result, e.g., diff/csum
   SColumnInfoData     *pTsOutput;
+  int32_t              numOfParams;
   int32_t              offset;
   SResultRowEntryInfo *resultInfo;
   SSubsidiaryResInfo   subsidiaries;
@@ -152,7 +154,7 @@ typedef struct SqlFunctionCtx {
   struct SSDataBlock  *pDstBlock;  // used by indefinite rows function to set selectivity
   SSerializeDataHandle saveHandle;
   int32_t              exprIdx;
-  char                 udfName[TSDB_FUNC_NAME_LEN];
+  char                *udfName;
 } SqlFunctionCtx;
 
 typedef struct tExprNode {

include/libs/function/functionMgt.h

@@ -218,6 +218,7 @@ bool fmIsKeepOrderFunc(int32_t funcId);
 bool fmIsCumulativeFunc(int32_t funcId);
 bool fmIsInterpPseudoColumnFunc(int32_t funcId);
 bool fmIsGroupKeyFunc(int32_t funcId);
+bool fmIsBlockDistFunc(int32_t funcId);
 
 void getLastCacheDataType(SDataType* pType);
 

include/libs/nodes/cmdnodes.h

@@ -389,6 +389,7 @@ typedef struct SStreamOptions {
   SNode*    pDeleteMark;
   int8_t    fillHistory;
   int8_t    ignoreExpired;
+  int8_t    ignoreUpdate;
 } SStreamOptions;
 
 typedef struct SCreateStreamStmt {

include/libs/nodes/plannodes.h

@@ -93,6 +93,7 @@ typedef struct SScanLogicNode {
   int64_t       watermark;
   int64_t       deleteMark;
   int8_t        igExpired;
+  int8_t        igCheckUpdate;
   SArray*       pSmaIndexes;
   SNodeList*    pGroupTags;
   bool          groupSort;
@@ -217,6 +218,7 @@ typedef struct SWindowLogicNode {
   int64_t          watermark;
   int64_t          deleteMark;
   int8_t           igExpired;
+  int8_t           igCheckUpdate;
   EWindowAlgorithm windowAlgo;
   EOrder           inputTsOrder;
   EOrder           outputTsOrder;
@@ -357,6 +359,7 @@ typedef struct STableScanPhysiNode {
   int64_t        watermark;
   int8_t         igExpired;
   bool           assignBlockUid;
+  int8_t         igCheckUpdate;
 } STableScanPhysiNode;
 
 typedef STableScanPhysiNode STableSeqScanPhysiNode;

include/libs/nodes/querynodes.h

@@ -226,8 +226,10 @@ typedef struct SIntervalWindowNode {
 typedef enum EFillMode {
   FILL_MODE_NONE = 1,
   FILL_MODE_VALUE,
+  FILL_MODE_VALUE_F,
   FILL_MODE_PREV,
   FILL_MODE_NULL,
+  FILL_MODE_NULL_F,
   FILL_MODE_LINEAR,
   FILL_MODE_NEXT
 } EFillMode;

include/libs/planner/planner.h

@@ -36,6 +36,7 @@ typedef struct SPlanContext {
   int64_t     watermark;
   int64_t     deleteMark;
   int8_t      igExpired;
+  int8_t      igCheckUpdate;
   char*       pMsg;
   int32_t     msgLen;
   const char* pUser;

include/libs/stream/streamState.h

@@ -110,6 +110,7 @@ int32_t streamStateGetParName(SStreamState* pState, int64_t groupId, void** pVal
 
 #if 0
 char* streamStateSessionDump(SStreamState* pState);
+char* streamStateIntervalDump(SStreamState* pState);
 #endif
 
 #ifdef __cplusplus

include/libs/stream/tstream.h

@@ -616,6 +616,7 @@ typedef struct SStreamMeta {
 SStreamMeta* streamMetaOpen(const char* path, void* ahandle, FTaskExpand expandFunc, int32_t vgId);
 void         streamMetaClose(SStreamMeta* streamMeta);
 
+int32_t      streamMetaSaveTask(SStreamMeta* pMeta, SStreamTask* pTask);
 int32_t      streamMetaAddTask(SStreamMeta* pMeta, int64_t ver, SStreamTask* pTask);
 int32_t      streamMetaAddSerializedTask(SStreamMeta* pMeta, int64_t startVer, char* msg, int32_t msgLen);
 SStreamTask* streamMetaGetTask(SStreamMeta* pMeta, int32_t taskId);
@@ -627,7 +628,7 @@ void         streamMetaRemoveTask(SStreamMeta* pMeta, int32_t taskId);
 int32_t streamMetaBegin(SStreamMeta* pMeta);
 int32_t streamMetaCommit(SStreamMeta* pMeta);
 int32_t streamMetaRollBack(SStreamMeta* pMeta);
-int32_t streamLoadTasks(SStreamMeta* pMeta);
+int32_t streamLoadTasks(SStreamMeta* pMeta, int64_t ver);
 
 // checkpoint
 int32_t streamProcessCheckpointSourceReq(SStreamMeta* pMeta, SStreamTask* pTask, SStreamCheckpointSourceReq* pReq);

include/libs/sync/sync.h

@@ -36,7 +36,7 @@ extern "C" {
 #define SYNC_DEL_WAL_MS              (1000 * 60)
 #define SYNC_ADD_QUORUM_COUNT        3
 #define SYNC_MNODE_LOG_RETENTION     10000
-#define SYNC_VNODE_LOG_RETENTION     20
+#define SYNC_VNODE_LOG_RETENTION     (TSDB_SYNC_LOG_BUFFER_RETENTION + 1)
 #define SNAPSHOT_MAX_CLOCK_SKEW_MS   1000 * 10
 #define SNAPSHOT_WAIT_MS             1000 * 30
 

include/libs/transport/trpc.h

@@ -112,7 +112,12 @@ typedef struct SRpcInit {
   // fail fast fp
   RpcFFfp ffp;
 
-  void *parent;
+  int32_t connLimitNum;
+  int32_t connLimitLock;
+
+  int8_t  supportBatch;  // 0: no batch, 1. batch
+  int32_t batchSize;
+  void   *parent;
 } SRpcInit;
 
 typedef struct {

include/libs/wal/wal.h

@@ -191,6 +191,7 @@ int32_t walApplyVer(SWal *, int64_t ver);
 // read
 SWalReader *walOpenReader(SWal *, SWalFilterCond *pCond);
 void        walCloseReader(SWalReader *pRead);
+void        walReadReset(SWalReader *pReader);
 int32_t     walReadVer(SWalReader *pRead, int64_t ver);
 int32_t     walReadSeekVer(SWalReader *pRead, int64_t ver);
 int32_t     walNextValidMsg(SWalReader *pRead);

include/os/osEnv.h

@@ -41,6 +41,7 @@ extern char            tsSSE42Enable;
 extern char            tsAVXEnable;
 extern char            tsAVX2Enable;
 extern char            tsFMAEnable;
+extern char            tsTagFilterCache;
 
 extern char configDir[];
 extern char tsDataDir[];

include/os/osFile.h

@@ -110,6 +110,8 @@ bool taosValidFile(TdFilePtr pFile);
 
 int32_t taosGetErrorFile(TdFilePtr pFile);
 
+int32_t taosCompressFile(char *srcFileName, char *destFileName);
+
 #ifdef __cplusplus
 }
 #endif

include/util/taoserror.h

@@ -67,6 +67,10 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_RPC_TIMEOUT                   TAOS_DEF_ERROR_CODE(0, 0x0019) //
 #define TSDB_CODE_RPC_SOMENODE_NOT_CONNECTED    TAOS_DEF_ERROR_CODE(0, 0x0020) // "Vgroup could not be connected"
 #define TSDB_CODE_RPC_SOMENODE_BROKEN_LINK      TAOS_DEF_ERROR_CODE(0, 0x0021) // 
+#define TSDB_CODE_RPC_MAX_SESSIONS              TAOS_DEF_ERROR_CODE(0, 0x0022) // 
+
+
+
 
 //common & util
 #define TSDB_CODE_OPS_NOT_SUPPORT               TAOS_DEF_ERROR_CODE(0, 0x0100) //
@@ -111,6 +115,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_NOT_FOUND                     TAOS_DEF_ERROR_CODE(0, 0x012A)
 #define TSDB_CODE_NO_DISKSPACE                  TAOS_DEF_ERROR_CODE(0, 0x012B)
 #define TSDB_CODE_TIMEOUT_ERROR                 TAOS_DEF_ERROR_CODE(0, 0x012C)
+#define TSDB_CODE_MSG_ENCODE_ERROR              TAOS_DEF_ERROR_CODE(0, 0x012D)
 
 #define TSDB_CODE_APP_IS_STARTING               TAOS_DEF_ERROR_CODE(0, 0x0130) //
 #define TSDB_CODE_APP_IS_STOPPING               TAOS_DEF_ERROR_CODE(0, 0x0131) //
@@ -355,6 +360,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_MND_STREAM_MUST_BE_DELETED    TAOS_DEF_ERROR_CODE(0, 0x03F3)
 #define TSDB_CODE_MND_STREAM_TASK_DROPPED       TAOS_DEF_ERROR_CODE(0, 0x03F4)
 #define TSDB_CODE_MND_MULTI_REPLICA_SOURCE_DB   TAOS_DEF_ERROR_CODE(0, 0x03F5)
+#define TSDB_CODE_MND_TOO_MANY_STREAMS          TAOS_DEF_ERROR_CODE(0, 0x03F6)
 
 // mnode-sma
 #define TSDB_CODE_MND_SMA_ALREADY_EXIST         TAOS_DEF_ERROR_CODE(0, 0x0480)

include/util/tarray.h

@@ -53,6 +53,7 @@ typedef struct SArray {
  * @return
  */
 SArray* taosArrayInit(size_t size, size_t elemSize);
+SArray* taosArrayInit_s(size_t size, size_t elemSize, size_t initialSize);
 
 /**
  *
@@ -149,14 +150,6 @@ void* taosArrayGetLast(const SArray* pArray);
  */
 size_t taosArrayGetSize(const SArray* pArray);
 
-/**
- * set the size of array
- * @param pArray
- * @param size size of the array
- * @return
- */
-void taosArraySetSize(SArray* pArray, size_t size);
-
 /**
  * insert data into array
  * @param pArray

include/util/tdef.h

@@ -281,8 +281,9 @@ typedef enum ELogicConditionType {
 #define TSDB_DNODE_ROLE_MGMT  1
 #define TSDB_DNODE_ROLE_VNODE 2
 
-#define TSDB_MAX_REPLICA          5
+#define TSDB_MAX_REPLICA               5
-#define TSDB_SYNC_LOG_BUFFER_SIZE 4096
+#define TSDB_SYNC_LOG_BUFFER_SIZE      4096
+#define TSDB_SYNC_LOG_BUFFER_RETENTION (TSDB_SYNC_LOG_BUFFER_SIZE >> 4)
 
 #define TSDB_TBNAME_COLUMN_INDEX     (-1)
 #define TSDB_MULTI_TABLEMETA_MAX_NUM 100000  // maximum batch size allowed to load table meta
@@ -412,7 +413,7 @@ typedef enum ELogicConditionType {
 #ifdef WINDOWS
 #define TSDB_MAX_RPC_THREADS 4  // windows pipe only support 4 connections.
 #else
-#define TSDB_MAX_RPC_THREADS 20
+#define TSDB_MAX_RPC_THREADS 10
 #endif
 
 #define TSDB_QUERY_TYPE_NON_TYPE 0x00u  // none type

include/util/tlog.h

@@ -89,7 +89,7 @@ bool taosAssertRelease(bool condition);
 // Disable all asserts that may compromise the performance.
 #if defined DISABLE_ASSERT
 #define ASSERT(condition)
-#define ASSERTS(condition, ...)
+#define ASSERTS(condition, ...)    (0)
 #else
 #define ASSERTS(condition, ...) taosAssertDebug(condition, __FILE__, __LINE__, __VA_ARGS__)
 #ifdef NDEBUG

include/util/tsimplehash.h

@@ -116,6 +116,7 @@ typedef struct SHNode {
   struct SHNode *next;
   uint32_t       keyLen : 20;
   uint32_t       dataLen : 12;
+  uint32_t       hashVal;
   char           data[];
 } SHNode;
 #pragma pack(pop)

include/util/ttrace.h

@@ -45,11 +45,25 @@ typedef struct STraceId {
 
 #define TRACE_GET_MSGID(traceId) (traceId)->msgId
 
-#define TRACE_TO_STR(traceId, buf)                              \
+//#define TRACE_TO_STR(traceId, buf)                              \
-  do {                                                          \
+//  do {                                                          \
-    int64_t rootId = (traceId) != NULL ? (traceId)->rootId : 0; \
+//    int64_t rootId = (traceId) != NULL ? (traceId)->rootId : 0; \
-    int64_t msgId = (traceId) != NULL ? (traceId)->msgId : 0;   \
+//    int64_t msgId = (traceId) != NULL ? (traceId)->msgId : 0;   \
-    sprintf(buf, "0x%" PRIx64 ":0x%" PRIx64 "", rootId, msgId); \
+//    sprintf(buf, "0x%" PRIx64 ":0x%" PRIx64 "", rootId, msgId); \
+//  } while (0)
+
+#define TRACE_TO_STR(_traceId, _buf)                             \
+  do {                                                            \
+    int64_t rootId = (_traceId) != NULL ? (_traceId)->rootId : 0; \
+    int64_t msgId = (_traceId) != NULL ? (_traceId)->msgId : 0;   \
+    char*   _t = _buf;                                            \
+    _t[0] = '0';                                                  \
+    _t[1] = 'x';                                                  \
+    _t += titoa(rootId, 16, &_t[2]);                              \
+    _t[0] = ':';                                                  \
+    _t[1] = '0';                                                  \
+    _t[2] = 'x';                                                  \
+    _t += titoa(msgId, 16, &_t[3]);                               \
   } while (0)
 
 #ifdef __cplusplus

include/util/tutil.h

@@ -46,6 +46,9 @@ char   *paGetToken(char *src, char **token, int32_t *tokenLen);
 int32_t taosByteArrayToHexStr(char bytes[], int32_t len, char hexstr[]);
 int32_t taosHexStrToByteArray(char hexstr[], char bytes[]);
 
+int32_t tintToHex(uint64_t val, char hex[]);
+int32_t titoa(uint64_t val, size_t radix, char str[]);
+
 char    *taosIpStr(uint32_t ipInt);
 uint32_t ip2uint(const char *const ip_addr);
 void     taosIp2String(uint32_t ip, char *str);

include/util/tworker.h

@@ -26,12 +26,12 @@ extern "C" {
 typedef struct SQWorkerPool SQWorkerPool;
 typedef struct SWWorkerPool SWWorkerPool;
 
-typedef struct SQWorker {
+typedef struct SQueueWorker {
   int32_t  id;      // worker id
   int64_t  pid;     // thread pid
   TdThread thread;  // thread id
   void    *pool;
-} SQWorker;
+} SQueueWorker;
 
 typedef struct SQWorkerPool {
   int32_t       max;  // max number of workers
@@ -39,7 +39,7 @@ typedef struct SQWorkerPool {
   int32_t       num;  // current number of workers
   STaosQset    *qset;
   const char   *name;
-  SQWorker     *workers;
+  SQueueWorker *workers;
   TdThreadMutex mutex;
 } SQWorkerPool;
 

packaging/tools/make_install.bat

@@ -2,12 +2,20 @@
 
 for /F %%a in ('echo prompt $E ^| cmd') do set "ESC=%%a"
 
-goto %1
+if "%1" NEQ "" goto %1
+
 :needAdmin
 
 if exist C:\\TDengine\\data\\dnode\\dnodeCfg.json (
   echo The default data directory C:/TDengine/data contains old data of tdengine 2.x, please clear it before installing!
 )
+
+rem // stop and delete service
+mshta vbscript:createobject("shell.application").shellexecute("%~s0",":stop_delete","","runas",1)(window.close)
+echo This might take a few moment to accomplish deleting service taosd/taosadapter ...
+call :check_svc taosd
+call :check_svc taosadapter
+
 set source_dir=%2
 set source_dir=%source_dir:/=\\%
 set binary_dir=%3
@@ -60,7 +68,6 @@ if exist %binary_dir%\\build\\bin\\taosdump.exe (
 
 copy %binary_dir%\\build\\bin\\taosd.exe %target_dir% > nul
 copy %binary_dir%\\build\\bin\\udfd.exe %target_dir% > nul
-
 if exist %binary_dir%\\build\\bin\\taosadapter.exe (
     copy %binary_dir%\\build\\bin\\taosadapter.exe %target_dir% > nul
 )
@@ -80,22 +87,23 @@ goto :eof
 
 :hasAdmin
 
-sc query "taosd" && sc stop taosd && sc delete taosd
+call :stop_delete
-sc query "taosadapter" && sc stop taosadapter && sc delete taosd
+call :check_svc taosd
+call :check_svc taosadapter
 
 copy /y C:\\TDengine\\driver\\taos.dll C:\\Windows\\System32 > nul
 if exist C:\\TDengine\\driver\\taosws.dll (
     copy /y C:\\TDengine\\driver\\taosws.dll C:\\Windows\\System32 > nul
 )
 
-sc query "taosd" >nul || sc create "taosd" binPath= "C:\\TDengine\\taosd.exe --win_service" start= DEMAND
+rem // create services
-sc query "taosadapter" >nul || sc create "taosadapter" binPath= "C:\\TDengine\\taosadapter.exe" start= DEMAND
+sc create "taosd" binPath= "C:\\TDengine\\taosd.exe --win_service" start= DEMAND
+sc create "taosadapter" binPath= "C:\\TDengine\\taosadapter.exe" start= DEMAND
 
 set "env=HKLM\System\CurrentControlSet\Control\Session Manager\Environment"
 for /f "tokens=2*" %%I in ('reg query "%env%" /v Path ^| findstr /i "\<Path\>"') do (
 
-    rem // make addition persistent through reboots
+    call :append_if_not_exists %%J
-    reg add "%env%" /f /v Path /t REG_EXPAND_SZ /d "%%J;C:\TDengine"
 
     rem // apply change to the current process
     for %%a in ("%%J;C:\TDengine") do path %%~a
@@ -105,3 +113,36 @@ rem // use setx to set a temporary throwaway value to trigger a WM_SETTINGCHANGE
 rem // applies change to new console windows without requiring a reboot
 (setx /m foo bar & reg delete "%env%" /f /v foo) >NUL 2>NUL
 
+goto :end
+
+:append_if_not_exists
+set "_origin_paths=%*"
+set "_paths=%*"
+set "_found=0"
+:loop
+for /f "tokens=1* delims=;" %%x in ("%_paths%") do (
+    if "%%x" EQU "C:\TDengine" (
+      set "_found=1"
+    ) else (
+      set "_paths=%%y"
+      goto :loop
+    )
+)
+if "%_found%" == "0" (
+  rem // make addition persistent through reboots
+  reg add "%env%" /f /v Path /t REG_EXPAND_SZ /d "%_origin_paths%;C:\TDengine"
+)
+exit /B 0
+
+:stop_delete
+sc stop taosd
+sc delete taosd
+sc stop taosadapter
+sc delete taosadapter
+exit /B 0
+
+:check_svc
+sc query %1 >nul 2>nul && goto :check_svc %1
+exit /B 0
+
+:end

packaging/tools/make_install.sh

@@ -531,13 +531,13 @@ function install_taosadapter_service() {
 }
 
 function install_service_on_launchctl() {
-  ${csudouser}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosd.plist > /dev/null 2>&1 || :
+  ${csudo}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosd.plist > /dev/null 2>&1 || :
   ${csudo}cp ${script_dir}/com.taosdata.taosd.plist /Library/LaunchDaemons/com.taosdata.taosd.plist
-  ${csudouser}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosd.plist > /dev/null 2>&1 || :
+  ${csudo}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosd.plist > /dev/null 2>&1 || :
 
-  ${csudouser}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist > /dev/null 2>&1 || :
+  ${csudo}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist > /dev/null 2>&1 || :
   ${csudo}cp ${script_dir}/com.taosdata.taosadapter.plist /Library/LaunchDaemons/com.taosdata.taosadapter.plist
-  ${csudouser}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist > /dev/null 2>&1 || :
+  ${csudo}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist > /dev/null 2>&1 || :
 }
 
 function install_service() {

packaging/tools/post.sh

@@ -512,14 +512,14 @@ function install_service_on_systemd() {
 
 function install_service_on_launchctl() {
   if [ -f ${install_main_dir}/service/com.taosdata.taosd.plist ]; then
-    ${csudouser}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosd.plist > /dev/null 2>&1 || :
+    ${csudo}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosd.plist > /dev/null 2>&1 || :
     ${csudo}cp ${install_main_dir}/service/com.taosdata.taosd.plist /Library/LaunchDaemons/com.taosdata.taosd.plist || :
-    ${csudouser}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosd.plist || :
+    ${csudo}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosd.plist || :
   fi
   if [ -f ${install_main_dir}/service/com.taosdata.taosadapter.plist ]; then
-    ${csudouser}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist > /dev/null 2>&1 || :
+    ${csudo}launchctl unload -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist > /dev/null 2>&1 || :
     ${csudo}cp ${install_main_dir}/service/com.taosdata.taosadapter.plist /Library/LaunchDaemons/com.taosdata.taosadapter.plist || :
-    ${csudouser}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist || :
+    ${csudo}launchctl load -w /Library/LaunchDaemons/com.taosdata.taosadapter.plist || :
   fi
 }
 

source/client/CMakeLists.txt

@@ -62,4 +62,4 @@ target_link_libraries(
 
 if(${BUILD_TEST})
     ADD_SUBDIRECTORY(test)
-endif(${BUILD_TEST})
+endif(${BUILD_TEST})

source/client/inc/clientInt.h

@@ -97,16 +97,14 @@ typedef struct {
 
 typedef struct SQueryExecMetric {
   int64_t start;        // start timestamp, us
-  int64_t syntaxStart;  // start to parse, us
-  int64_t syntaxEnd;    // end to parse, us
   int64_t ctgStart;     // start to parse, us
-  int64_t ctgEnd;       // end to parse, us
+  int64_t execStart;     // start to parse, us
-  int64_t semanticEnd;
+
-  int64_t planEnd;
+  int64_t parseCostUs;   
-  int64_t resultReady;
+  int64_t ctgCostUs;
-  int64_t execEnd;
+  int64_t analyseCostUs;
-  int64_t send;  // start to send to server, us
+  int64_t planCostUs;
-  int64_t rsp;   // receive response from server, us
+  int64_t execCostUs;
 } SQueryExecMetric;
 
 struct SAppInstInfo {

source/client/src/clientEnv.c

@@ -79,22 +79,21 @@ static void deregisterRequest(SRequestObj *pRequest) {
            "current:%d, app current:%d",
            pRequest->self, pTscObj->id, pRequest->requestId, duration / 1000.0, num, currentInst);
 
-  if (QUERY_NODE_VNODE_MODIF_STMT == pRequest->stmtType) {
+  if (pRequest->pQuery && pRequest->pQuery->pRoot) {
-    //    tscPerf("insert duration %" PRId64 "us: syntax:%" PRId64 "us, ctg:%" PRId64 "us, semantic:%" PRId64
+    if (QUERY_NODE_VNODE_MODIF_STMT == pRequest->pQuery->pRoot->type && (0 == ((SVnodeModifOpStmt*)pRequest->pQuery->pRoot)->sqlNodeType)) {
-    //            "us, exec:%" PRId64 "us",
+      tscDebug("insert duration %" PRId64 "us: parseCost:%" PRId64 "us, ctgCost:%" PRId64 "us, analyseCost:%" PRId64
-    //            duration, pRequest->metric.syntaxEnd - pRequest->metric.syntaxStart,
+              "us, planCost:%" PRId64 "us, exec:%" PRId64 "us",
-    //            pRequest->metric.ctgEnd - pRequest->metric.ctgStart, pRequest->metric.semanticEnd -
+              duration, pRequest->metric.parseCostUs, pRequest->metric.ctgCostUs, pRequest->metric.analyseCostUs, 
-    //            pRequest->metric.ctgEnd, pRequest->metric.execEnd - pRequest->metric.semanticEnd);
+              pRequest->metric.planCostUs, pRequest->metric.execCostUs);
-    atomic_add_fetch_64((int64_t *)&pActivity->insertElapsedTime, duration);
+      atomic_add_fetch_64((int64_t *)&pActivity->insertElapsedTime, duration);
-  } else if (QUERY_NODE_SELECT_STMT == pRequest->stmtType) {
+    } else if (QUERY_NODE_SELECT_STMT == pRequest->stmtType) {
-    //    tscPerf("select duration %" PRId64 "us: syntax:%" PRId64 "us, ctg:%" PRId64 "us, semantic:%" PRId64
+      tscDebug("query duration %" PRId64 "us: parseCost:%" PRId64 "us, ctgCost:%" PRId64 "us, analyseCost:%" PRId64
-    //            "us, planner:%" PRId64 "us, exec:%" PRId64 "us, reqId:0x%" PRIx64,
+              "us, planCost:%" PRId64 "us, exec:%" PRId64 "us",
-    //            duration, pRequest->metric.syntaxEnd - pRequest->metric.syntaxStart,
+              duration, pRequest->metric.parseCostUs, pRequest->metric.ctgCostUs, pRequest->metric.analyseCostUs, 
-    //            pRequest->metric.ctgEnd - pRequest->metric.ctgStart, pRequest->metric.semanticEnd -
+              pRequest->metric.planCostUs, pRequest->metric.execCostUs);
-    //            pRequest->metric.ctgEnd, pRequest->metric.planEnd - pRequest->metric.semanticEnd,
-    //            pRequest->metric.resultReady - pRequest->metric.planEnd, pRequest->requestId);
 
-    atomic_add_fetch_64((int64_t *)&pActivity->queryElapsedTime, duration);
+      atomic_add_fetch_64((int64_t *)&pActivity->queryElapsedTime, duration);
+    }
   }
 
   if (duration >= SLOW_QUERY_INTERVAL) {
@@ -362,8 +361,6 @@ void doDestroyRequest(void *p) {
   taosArrayDestroy(pRequest->tableList);
   taosArrayDestroy(pRequest->dbList);
   taosArrayDestroy(pRequest->targetTableList);
-  qDestroyQuery(pRequest->pQuery);
-  nodesDestroyAllocator(pRequest->allocatorRefId);
 
   destroyQueryExecRes(&pRequest->body.resInfo.execRes);
 
@@ -378,6 +375,9 @@ void doDestroyRequest(void *p) {
     taosMemoryFree(pRequest->body.param);
   }
 
+  qDestroyQuery(pRequest->pQuery);
+  nodesDestroyAllocator(pRequest->allocatorRefId);
+
   taosMemoryFreeClear(pRequest->sqlstr);
   taosMemoryFree(pRequest);
   tscTrace("end to destroy request %" PRIx64 " p:%p", reqId, pRequest);

source/client/src/clientHb.c

@@ -347,7 +347,7 @@ int32_t hbBuildQueryDesc(SQueryHbReqBasic *hbBasic, STscObj *pObj) {
       continue;
     }
 
-    if (pRequest->killed) {
+    if (pRequest->killed || 0 == pRequest->body.queryJob) {
       releaseRequest(*rid);
       pIter = taosHashIterate(pObj->pRequests, pIter);
       continue;

source/client/src/clientImpl.c

@@ -138,6 +138,12 @@ STscObj* taos_connect_internal(const char* ip, const char* user, const char* pas
     p->mgmtEp = epSet;
     taosThreadMutexInit(&p->qnodeMutex, NULL);
     p->pTransporter = openTransporter(user, secretEncrypt, tsNumOfCores / 2);
+    if (p->pTransporter == NULL) {
+      taosThreadMutexUnlock(&appInfo.mutex);
+      taosMemoryFreeClear(key);
+      taosMemoryFree(p);
+      return NULL;
+    }
     p->pAppHbMgr = appHbMgrInit(p, key);
     if (NULL == p->pAppHbMgr) {
       destroyAppInst(p);
@@ -940,7 +946,7 @@ void schedulerExecCb(SExecResult* pResult, void* param, int32_t code) {
     removeMeta(pTscObj, pRequest->targetTableList);
   }
 
-  pRequest->metric.execEnd = taosGetTimestampUs();
+  pRequest->metric.execCostUs = taosGetTimestampUs() - pRequest->metric.execStart;
   int32_t code1 = handleQueryExecRsp(pRequest);
   if (pRequest->code == TSDB_CODE_SUCCESS && pRequest->code != code1) {
     pRequest->code = code1;
@@ -1066,11 +1072,10 @@ static int32_t asyncExecSchQuery(SRequestObj* pRequest, SQuery* pQuery, SMetaDat
     pRequest->body.subplanNum = pDag->numOfSubplans;
   }
 
-  pRequest->metric.planEnd = taosGetTimestampUs();
+  pRequest->metric.execStart = taosGetTimestampUs();
-  if (code == TSDB_CODE_SUCCESS) {
+
-    tscDebug("0x%" PRIx64 " create query plan success, elapsed time:%.2f ms, 0x%" PRIx64, pRequest->self,
+  pRequest->metric.planCostUs = pRequest->metric.execStart - st;
-             (pRequest->metric.planEnd - st) / 1000.0, pRequest->requestId);
+
-  }
   if (TSDB_CODE_SUCCESS == code && !pRequest->validateOnly) {
     SArray* pNodeList = NULL;
     if (QUERY_NODE_VNODE_MODIF_STMT != nodeType(pQuery->pRoot)) {
@@ -1118,6 +1123,16 @@ void launchAsyncQuery(SRequestObj* pRequest, SQuery* pQuery, SMetaData* pResultM
     destorySqlCallbackWrapper(pWrapper);
   }
 
+  if (pQuery->pRoot && !pRequest->inRetry) {
+    STscObj*            pTscObj = pRequest->pTscObj;
+    SAppClusterSummary* pActivity = &pTscObj->pAppInfo->summary;
+    if (QUERY_NODE_VNODE_MODIF_STMT == pQuery->pRoot->type && (0 == ((SVnodeModifOpStmt*)pQuery->pRoot)->sqlNodeType)) {
+      atomic_add_fetch_64((int64_t*)&pActivity->numOfInsertsReq, 1);
+    } else if (QUERY_NODE_SELECT_STMT == pQuery->pRoot->type) {
+      atomic_add_fetch_64((int64_t*)&pActivity->numOfQueryReq, 1);
+    }
+  }
+
   switch (pQuery->execMode) {
     case QUERY_EXEC_MODE_LOCAL:
       asyncExecLocalCmd(pRequest, pQuery);
@@ -1387,23 +1402,7 @@ int32_t doProcessMsgFromServer(void* param) {
     SRequestObj* pRequest = (SRequestObj*)taosAcquireRef(clientReqRefPool, pSendInfo->requestObjRefId);
     if (pRequest) {
       assert(pRequest->self == pSendInfo->requestObjRefId);
-
-      pRequest->metric.rsp = taosGetTimestampUs();
       pTscObj = pRequest->pTscObj;
-      /*
-       * There is not response callback function for submit response.
-       * The actual inserted number of points is the first number.
-       */
-      int32_t elapsed = pRequest->metric.rsp - pRequest->metric.start;
-      if (pMsg->code == TSDB_CODE_SUCCESS) {
-        tscDebug("0x%" PRIx64 " rsp msg:%s, code:%s rspLen:%d, elapsed:%d ms, reqId:0x%" PRIx64, pRequest->self,
-                 TMSG_INFO(pMsg->msgType), tstrerror(pMsg->code), pMsg->contLen, elapsed / 1000, pRequest->requestId);
-      } else {
-        tscError("0x%" PRIx64 " rsp msg:%s, code:%s rspLen:%d, elapsed time:%d ms, reqId:0x%" PRIx64, pRequest->self,
-                 TMSG_INFO(pMsg->msgType), tstrerror(pMsg->code), pMsg->contLen, elapsed / 1000, pRequest->requestId);
-      }
-
-      taosReleaseRef(clientReqRefPool, pSendInfo->requestObjRefId);
     }
   }
 
@@ -1423,6 +1422,11 @@ int32_t doProcessMsgFromServer(void* param) {
   }
 
   pSendInfo->fp(pSendInfo->param, &buf, pMsg->code);
+
+  if (pTscObj) {
+    taosReleaseRef(clientReqRefPool, pSendInfo->requestObjRefId);
+  }
+
   rpcFreeCont(pMsg->pCont);
   destroySendMsgInfo(pSendInfo);
 
@@ -1460,6 +1464,7 @@ void processMsgFromServer(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet) {
     tscError("failed to sched msg to tsc, tsc ready to quit");
     rpcFreeCont(pMsg->pCont);
     taosMemoryFree(arg->pEpset);
+    destroySendMsgInfo(pMsg->info.ahandle);
     taosMemoryFree(arg);
   }
 }

source/client/src/clientMain.c

@@ -707,7 +707,8 @@ static void doAsyncQueryFromAnalyse(SMetaData *pResultMeta, void *param, int32_t
   SRequestObj         *pRequest = pWrapper->pRequest;
   SQuery              *pQuery = pRequest->pQuery;
 
-  pRequest->metric.ctgEnd = taosGetTimestampUs();
+  int64_t analyseStart = taosGetTimestampUs();
+  pRequest->metric.ctgCostUs = analyseStart - pRequest->metric.ctgStart;
   qDebug("0x%" PRIx64 " start to semantic analysis, reqId:0x%" PRIx64, pRequest->self, pRequest->requestId);
 
   if (code == TSDB_CODE_SUCCESS) {
@@ -718,7 +719,7 @@ static void doAsyncQueryFromAnalyse(SMetaData *pResultMeta, void *param, int32_t
     }
   }
 
-  pRequest->metric.semanticEnd = taosGetTimestampUs();
+  pRequest->metric.analyseCostUs = taosGetTimestampUs() - analyseStart;
 
   if (code == TSDB_CODE_SUCCESS) {
     if (pQuery->haveResultSet) {
@@ -730,10 +731,6 @@ static void doAsyncQueryFromAnalyse(SMetaData *pResultMeta, void *param, int32_t
     TSWAP(pRequest->tableList, (pQuery)->pTableList);
     TSWAP(pRequest->targetTableList, (pQuery)->pTargetTableList);
 
-    double el = (pRequest->metric.semanticEnd - pRequest->metric.ctgEnd) / 1000.0;
-    tscDebug("0x%" PRIx64 " analysis semantics completed, start async query, elapsed time:%.2f ms, reqId:0x%" PRIx64,
-             pRequest->self, el, pRequest->requestId);
-
     launchAsyncQuery(pRequest, pQuery, pResultMeta, pWrapper);
   } else {
     destorySqlCallbackWrapper(pWrapper);
@@ -798,7 +795,7 @@ static void doAsyncQueryFromParse(SMetaData *pResultMeta, void *param, int32_t c
   SRequestObj         *pRequest = pWrapper->pRequest;
   SQuery              *pQuery = pRequest->pQuery;
 
-  pRequest->metric.ctgEnd = taosGetTimestampUs();
+  pRequest->metric.ctgCostUs += taosGetTimestampUs() - pRequest->metric.ctgStart;
   qDebug("0x%" PRIx64 " start to continue parse, reqId:0x%" PRIx64 ", code:%s", pRequest->self, pRequest->requestId,
          tstrerror(code));
 
@@ -910,7 +907,7 @@ void doAsyncQuery(SRequestObj *pRequest, bool updateMetaForce) {
   }
 
   if (TSDB_CODE_SUCCESS == code) {
-    pRequest->metric.syntaxStart = taosGetTimestampUs();
+    int64_t syntaxStart = taosGetTimestampUs();
 
     pWrapper->pCatalogReq = taosMemoryCalloc(1, sizeof(SCatalogReq));
     if (pWrapper->pCatalogReq == NULL) {
@@ -921,19 +918,11 @@ void doAsyncQuery(SRequestObj *pRequest, bool updateMetaForce) {
       code = qParseSqlSyntax(pWrapper->pParseCtx, &pRequest->pQuery, pWrapper->pCatalogReq);
     }
 
-    pRequest->metric.syntaxEnd = taosGetTimestampUs();
+    pRequest->metric.parseCostUs += taosGetTimestampUs() - syntaxStart;
-  }
-
-  if (TSDB_CODE_SUCCESS == code && !updateMetaForce) {
-    SAppClusterSummary *pActivity = &pTscObj->pAppInfo->summary;
-    if (QUERY_NODE_INSERT_STMT == nodeType(pRequest->pQuery->pRoot)) {
-      atomic_add_fetch_64((int64_t *)&pActivity->numOfInsertsReq, 1);
-    } else if (QUERY_NODE_SELECT_STMT == nodeType(pRequest->pQuery->pRoot)) {
-      atomic_add_fetch_64((int64_t *)&pActivity->numOfQueryReq, 1);
-    }
   }
 
   if (TSDB_CODE_SUCCESS == code) {
+    pRequest->stmtType = pRequest->pQuery->pRoot->type;
     phaseAsyncQuery(pWrapper);
   } else {
     tscError("0x%" PRIx64 " error happens, code:%d - %s, reqId:0x%" PRIx64, pRequest->self, code, tstrerror(code),
@@ -960,7 +949,6 @@ static void fetchCallback(void *pResult, void *param, int32_t code) {
   SRequestObj *pRequest = (SRequestObj *)param;
 
   SReqResultInfo *pResultInfo = &pRequest->body.resInfo;
-  pRequest->metric.resultReady = taosGetTimestampUs();
 
   tscDebug("0x%" PRIx64 " enter scheduler fetch cb, code:%d - %s, reqId:0x%" PRIx64, pRequest->self, code,
            tstrerror(code), pRequest->requestId);

source/client/src/clientRawBlockWrite.c

@@ -25,6 +25,10 @@
 #include "tref.h"
 #include "ttimer.h"
 
+static tb_uid_t processSuid(tb_uid_t suid, char* db){
+  return suid + MurmurHash3_32(db, strlen(db));
+}
+
 static char* buildCreateTableJson(SSchemaWrapper* schemaRow, SSchemaWrapper* schemaTag, char* name, int64_t id,
                                   int8_t t) {
   char*  string = NULL;
@@ -681,7 +685,7 @@ static int32_t taosCreateStb(TAOS* taos, void* meta, int32_t metaLen) {
   pReq.numOfColumns = req.schemaRow.nCols;
   pReq.numOfTags = req.schemaTag.nCols;
   pReq.commentLen = -1;
-  pReq.suid = req.suid;
+  pReq.suid = processSuid(req.suid, pRequest->pDb);
   pReq.source = TD_REQ_FROM_TAOX;
   pReq.igExists = true;
 
@@ -753,7 +757,7 @@ static int32_t taosDropStb(TAOS* taos, void* meta, int32_t metaLen) {
   // build drop stable
   pReq.igNotExists = true;
   pReq.source = TD_REQ_FROM_TAOX;
-  pReq.suid = req.suid;
+  pReq.suid = processSuid(req.suid, pRequest->pDb);
 
   STscObj* pTscObj = pRequest->pTscObj;
   SName    tableName = {0};
@@ -871,6 +875,7 @@ static int32_t taosCreateTable(TAOS* taos, void* meta, int32_t metaLen) {
     if (pCreateReq->type == TSDB_CHILD_TABLE) {
       STableMeta* pTableMeta = NULL;
       SName       sName = {0};
+      pCreateReq->ctb.suid = processSuid(pCreateReq->ctb.suid, pRequest->pDb);
       toName(pTscObj->acctId, pRequest->pDb, pCreateReq->ctb.stbName, &sName);
       code = catalogGetTableMeta(pCatalog, &conn, &sName, &pTableMeta);
       if (code != TSDB_CODE_SUCCESS) {
@@ -1008,6 +1013,7 @@ static int32_t taosDropTable(TAOS* taos, void* meta, int32_t metaLen) {
   for (int32_t iReq = 0; iReq < req.nReqs; iReq++) {
     pDropReq = req.pReqs + iReq;
     pDropReq->igNotExists = true;
+    pDropReq->suid = processSuid(pDropReq->suid, pRequest->pDb);
 
     SVgroupInfo pInfo = {0};
     SName       pName = {0};
@@ -1922,6 +1928,7 @@ static int32_t tmqWriteRawMetaDataImpl(TAOS* taos, void* data, int32_t dataLen)
   SMqTaosxRspObj rspObj = {0};
   SDecoder       decoder = {0};
   STableMeta*    pTableMeta = NULL;
+  void*          schemaContent = NULL;
 
   terrno = TSDB_CODE_SUCCESS;
   SRequestObj* pRequest = (SRequestObj*)createRequest(*(int64_t*)taos, TSDB_SQL_INSERT, 0);
@@ -2008,27 +2015,49 @@ static int32_t tmqWriteRawMetaDataImpl(TAOS* taos, void* data, int32_t dataLen)
       SDecoder      decoderTmp = {0};
       SVCreateTbReq pCreateReq = {0};
 
-      tDecoderInit(&decoderTmp, *dataTmp, *lenTmp);
+      do{
-      if (tDecodeSVCreateTbReq(&decoderTmp, &pCreateReq) < 0) {
+        tDecoderInit(&decoderTmp, *dataTmp, *lenTmp);
-        tDecoderClear(&decoderTmp);
+        if (tDecodeSVCreateTbReq(&decoderTmp, &pCreateReq) < 0) {
-        taosMemoryFreeClear(pCreateReq.comment);
+          code = TSDB_CODE_MSG_DECODE_ERROR;
-        taosArrayDestroy(pCreateReq.ctb.tagName);
+          break;
-        goto end;
+        }
-      }
 
-      ASSERT(pCreateReq.type == TSDB_CHILD_TABLE);
+        if (strcmp(tbName, pCreateReq.name) != 0) {
-      if (strcmp(tbName, pCreateReq.name) == 0) {
+          break;
-        schemaLen = *lenTmp;
+        }
-        schemaData = *dataTmp;
+
+        pCreateReq.ctb.suid = processSuid(pCreateReq.ctb.suid, pRequest->pDb);
+
+        int32_t len = 0;
+        tEncodeSize(tEncodeSVCreateTbReq, &pCreateReq, len, code);
+        if(code != 0) {
+          code = TSDB_CODE_MSG_ENCODE_ERROR;
+          break;
+        }
+        taosMemoryFree(schemaContent);
+        schemaContent = taosMemoryMalloc(len);
+        if(!schemaContent) {
+          code = TSDB_CODE_OUT_OF_MEMORY;
+          break;
+        }
+        SEncoder encoder = {0};
+        tEncoderInit(&encoder, schemaContent, len);
+        code = tEncodeSVCreateTbReq(&encoder, &pCreateReq);
+        if (code != 0) {
+          tEncoderClear(&encoder);
+          code = TSDB_CODE_MSG_ENCODE_ERROR;
+          break;
+        }
+        schemaLen = len;
+        schemaData = schemaContent;
         strcpy(pName.tname, pCreateReq.ctb.stbName);
-        tDecoderClear(&decoderTmp);
+        tEncoderClear(&encoder);
-        taosMemoryFreeClear(pCreateReq.comment);
+      }while(0);
-        taosArrayDestroy(pCreateReq.ctb.tagName);
-        break;
-      }
       tDecoderClear(&decoderTmp);
       taosMemoryFreeClear(pCreateReq.comment);
       taosArrayDestroy(pCreateReq.ctb.tagName);
+      if(code != 0) goto end;
+      if(schemaLen != 0) break;
     }
 
     code = catalogGetTableMeta(pCatalog, &conn, &pName, &pTableMeta);
@@ -2217,6 +2246,7 @@ static int32_t tmqWriteRawMetaDataImpl(TAOS* taos, void* data, int32_t dataLen)
   destroyRequest(pRequest);
   taosHashCleanup(pVgHash);
   taosMemoryFreeClear(pTableMeta);
+  taosMemoryFree(schemaContent);
   return code;
 }
 

source/client/src/clientSml.c

@@ -149,17 +149,8 @@ typedef struct {
   int64_t endTime;
 } SSmlCostInfo;
 
-typedef struct {
-  SRequestObj     *request;
-  tsem_t           sem;
-  int32_t          cnt;
-  int32_t          total;
-  TdThreadSpinlock lock;
-} Params;
-
 typedef struct {
   int64_t id;
-  Params *params;
 
   SMLProtocolType protocol;
   int8_t          precision;
@@ -178,7 +169,6 @@ typedef struct {
   SQuery      *pQuery;
 
   SSmlCostInfo cost;
-  int32_t      affectedRows;
   SSmlMsgBuf   msgBuf;
   SHashObj    *dumplicateKey;  // for dumplicate key
   SArray      *colsContainer;  // for cols parse, if dataFormat == false
@@ -1513,7 +1503,6 @@ static void smlDestroyInfo(SSmlHandle *info) {
   if (!info->dataFormat) {
     taosArrayDestroy(info->colsContainer);
   }
-  destroyRequest(info->pRequest);
 
   cJSON_Delete(info->root);
   taosMemoryFreeClear(info);
@@ -2351,20 +2340,11 @@ static int32_t smlInsertData(SSmlHandle *info) {
   }
   info->cost.insertRpcTime = taosGetTimestampUs();
 
-  // launchQueryImpl(info->pRequest, info->pQuery, false, NULL);
-  //  info->affectedRows = taos_affected_rows(info->pRequest);
-  //  return info->pRequest->code;
-
   SAppClusterSummary *pActivity = &info->taos->pAppInfo->summary;
   atomic_add_fetch_64((int64_t *)&pActivity->numOfInsertsReq, 1);
 
-  SSqlCallbackWrapper *pWrapper = (SSqlCallbackWrapper *)taosMemoryCalloc(1, sizeof(SSqlCallbackWrapper));
+  launchQueryImpl(info->pRequest, info->pQuery, true, NULL);
-  if (pWrapper == NULL) {
+  return info->pRequest->code;
-    return TSDB_CODE_OUT_OF_MEMORY;
-  }
-  pWrapper->pRequest = info->pRequest;
-  launchAsyncQuery(info->pRequest, info->pQuery, NULL, pWrapper);
-  return TSDB_CODE_SUCCESS;
 }
 
 static void smlPrintStatisticInfo(SSmlHandle *info) {
@@ -2498,48 +2478,13 @@ static int32_t isSchemalessDb(STscObj *taos, SRequestObj *request) {
   return TSDB_CODE_SUCCESS;
 }
 
-static void smlInsertCallback(void *param, void *res, int32_t code) {
-  SRequestObj *pRequest = (SRequestObj *)res;
-  SSmlHandle  *info = (SSmlHandle *)param;
-  int32_t      rows = taos_affected_rows(pRequest);
-
-  uDebug("SML:0x%" PRIx64 " result. code:%d, msg:%s", info->id, pRequest->code, pRequest->msgBuf);
-  Params *pParam = info->params;
-  // lock
-  taosThreadSpinLock(&pParam->lock);
-  pParam->cnt++;
-  if (code != TSDB_CODE_SUCCESS) {
-    pParam->request->code = code;
-    pParam->request->body.resInfo.numOfRows += rows;
-  } else {
-    pParam->request->body.resInfo.numOfRows += info->affectedRows;
-  }
-  // unlock
-  taosThreadSpinUnlock(&pParam->lock);
-
-  if (pParam->cnt == pParam->total) {
-    tsem_post(&pParam->sem);
-  }
-  uDebug("SML:0x%" PRIx64 " insert finished, code: %d, rows: %d, total: %d", info->id, code, rows, info->affectedRows);
-  info->cost.endTime = taosGetTimestampUs();
-  info->cost.code = code;
-  smlPrintStatisticInfo(info);
-  smlDestroyInfo(info);
-}
-
 TAOS_RES *taos_schemaless_insert_inner(SRequestObj *request, char *lines[], char *rawLine, char *rawLineEnd,
                                        int numLines, int protocol, int precision, int32_t ttl) {
-  int      batchs = 0;
   STscObj *pTscObj = request->pTscObj;
-
+  SSmlHandle *info = NULL;
   pTscObj->schemalessType = 1;
   SSmlMsgBuf msg = {ERROR_MSG_BUF_DEFAULT_SIZE, request->msgBuf};
 
-  Params params = {0};
-  params.request = request;
-  tsem_init(&params.sem, 0, 0);
-  taosThreadSpinInit(&(params.lock), 0);
-
   if (request->pDb == NULL) {
     request->code = TSDB_CODE_PAR_DB_NOT_SPECIFIED;
     smlBuildInvalidDataMsg(&msg, "Database not specified", NULL);
@@ -2573,65 +2518,24 @@ TAOS_RES *taos_schemaless_insert_inner(SRequestObj *request, char *lines[], char
     goto end;
   }
 
-  batchs = ceil(((double)numLines) / tsSmlBatchSize);
+  info = smlBuildSmlInfo(pTscObj, request, (SMLProtocolType)protocol, precision);
-  params.total = batchs;
+  if (!info) {
-  for (int i = 0; i < batchs; ++i) {
+    request->code = TSDB_CODE_OUT_OF_MEMORY;
-    SRequestObj *req = (SRequestObj *)createRequest(pTscObj->id, TSDB_SQL_INSERT, 0);
+    uError("SML:taos_schemaless_insert error SSmlHandle is null");
-    if (!req) {
+    goto end;
-      request->code = TSDB_CODE_OUT_OF_MEMORY;
-      uError("SML:taos_schemaless_insert error request is null");
-      goto end;
-    }
-    SSmlHandle *info = smlBuildSmlInfo(pTscObj, req, (SMLProtocolType)protocol, precision);
-    if (!info) {
-      request->code = TSDB_CODE_OUT_OF_MEMORY;
-      uError("SML:taos_schemaless_insert error SSmlHandle is null");
-      goto end;
-    }
-
-    info->isRawLine = (rawLine == NULL);
-    info->ttl       = ttl;
-
-    int32_t perBatch = tsSmlBatchSize;
-
-    if (numLines > perBatch) {
-      numLines -= perBatch;
-    } else {
-      perBatch = numLines;
-      numLines = 0;
-    }
-
-    info->params = &params;
-    info->affectedRows = perBatch;
-    info->pRequest->body.queryFp = smlInsertCallback;
-    info->pRequest->body.param = info;
-    int32_t code = smlProcess(info, lines, rawLine, rawLineEnd, perBatch);
-    if (lines) {
-      lines += perBatch;
-    }
-    if (rawLine) {
-      int num = 0;
-      while (rawLine < rawLineEnd) {
-        if (*(rawLine++) == '\n') {
-          num++;
-        }
-        if (num == perBatch) {
-          break;
-        }
-      }
-    }
-    if (code != TSDB_CODE_SUCCESS) {
-      info->pRequest->body.queryFp(info, req, code);
-    }
   }
-  tsem_wait(&params.sem);
+
+  info->isRawLine = (rawLine == NULL);
+  info->ttl       = ttl;
+  request->code = smlProcess(info, lines, rawLine, rawLineEnd, numLines);
 
 end:
-  taosThreadSpinDestroy(&params.lock);
+  uDebug("SML:0x%" PRIx64 " insert finished, code: %d", info->id, request->code);
-  tsem_destroy(&params.sem);
+  info->cost.endTime = taosGetTimestampUs();
-  //  ((STscObj *)taos)->schemalessType = 0;
+  info->cost.code = request->code;
-  pTscObj->schemalessType = 1;
+  smlPrintStatisticInfo(info);
-  uDebug("resultend:%s", request->msgBuf);
+  smlDestroyInfo(info);
+
   return (TAOS_RES *)request;
 }
 

source/client/src/clientTmq.c

@@ -32,15 +32,15 @@
   sem_post(x)
 #endif
 
-int32_t tmqAskEp(tmq_t* tmq, bool async);
+struct SMqMgmt{
-
-typedef struct {
   int8_t  inited;
   tmr_h   timer;
   int32_t rsetId;
-} SMqMgmt;
+};
 
-static SMqMgmt tmqMgmt = {0};
+static TdThreadOnce tmqInit = PTHREAD_ONCE_INIT;   // initialize only once
+volatile int32_t    tmqInitRes = 0;                // initialize rsp code
+static struct SMqMgmt tmqMgmt = {0};
 
 typedef struct {
   int8_t  tmqRspType;
@@ -58,16 +58,14 @@ struct tmq_list_t {
 };
 
 struct tmq_conf_t {
-  char    clientId[256];
+  char           clientId[256];
-  char    groupId[TSDB_CGROUP_LEN];
+  char           groupId[TSDB_CGROUP_LEN];
-  int8_t  autoCommit;
+  int8_t         autoCommit;
-  int8_t  resetOffset;
+  int8_t         resetOffset;
-  int8_t  withTbName;
+  int8_t         withTbName;
-  int8_t  snapEnable;
+  int8_t         snapEnable;
-  int32_t snapBatchSize;
+  int32_t        snapBatchSize;
-
+  bool           hbBgEnable;
-  bool hbBgEnable;
-
   uint16_t       port;
   int32_t        autoCommitInterval;
   char*          ip;
@@ -78,18 +76,17 @@ struct tmq_conf_t {
 };
 
 struct tmq_t {
-  int64_t refId;
+  int64_t  refId;
   // conf
-  char    groupId[TSDB_CGROUP_LEN];
+  char     groupId[TSDB_CGROUP_LEN];
-  char    clientId[256];
+  char     clientId[256];
-  int8_t  withTbName;
+  int8_t   withTbName;
-  int8_t  useSnapshot;
+  int8_t   useSnapshot;
-  int8_t  autoCommit;
+  int8_t   autoCommit;
-  int32_t autoCommitInterval;
+  int32_t  autoCommitInterval;
-  int32_t resetOffsetCfg;
+  int32_t  resetOffsetCfg;
-  int64_t consumerId;
+  uint64_t consumerId;
-
+  bool     hbBgEnable;
-  bool hbBgEnable;
 
   tmq_commit_cb* commitCb;
   void*          commitCbUserParam;
@@ -155,11 +152,9 @@ typedef struct {
 
 typedef struct {
   // subscribe info
-  char topicName[TSDB_TOPIC_FNAME_LEN];
+  char           topicName[TSDB_TOPIC_FNAME_LEN];
-  char db[TSDB_DB_FNAME_LEN];
+  char           db[TSDB_DB_FNAME_LEN];
-
+  SArray*        vgs;  // SArray<SMqClientVg>
-  SArray* vgs;  // SArray<SMqClientVg>
-
   SSchemaWrapper schema;
 } SMqClientTopic;
 
@@ -217,17 +212,26 @@ typedef struct {
 typedef struct {
   SMqCommitCbParamSet* params;
   STqOffset*           pOffset;
+  SMqClientVg*         pMqVg;
   /*char                 topicName[TSDB_TOPIC_FNAME_LEN];*/
   /*int32_t              vgId;*/
 } SMqCommitCbParam;
 
+static int32_t tmqAskEp(tmq_t* tmq, bool async);
+
 tmq_conf_t* tmq_conf_new() {
   tmq_conf_t* conf = taosMemoryCalloc(1, sizeof(tmq_conf_t));
+  if (conf == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return conf;
+  }
+
   conf->withTbName = false;
   conf->autoCommit = true;
   conf->autoCommitInterval = 5000;
   conf->resetOffset = TMQ_CONF__RESET_OFFSET__EARLIEAST;
   conf->hbBgEnable = true;
+
   return conf;
 }
 
@@ -418,7 +422,6 @@ int32_t tmqCommitDone(SMqCommitCbParamSet* pParamSet) {
 
 static void tmqCommitRspCountDown(SMqCommitCbParamSet* pParamSet) {
   int32_t waitingRspNum = atomic_sub_fetch_32(&pParamSet->waitingRspNum, 1);
-  ASSERT(waitingRspNum >= 0);
   if (waitingRspNum == 0) {
     tmqCommitDone(pParamSet);
   }
@@ -436,6 +439,17 @@ int32_t tmqCommitCb(void* param, SDataBuf* pBuf, int32_t code) {
   }
 #endif
 
+  // there may be race condition. fix it
+  if (pBuf->pEpSet != NULL && pParam->pMqVg != NULL) {
+    SMqClientVg* pMqVg = pParam->pMqVg;
+
+    SEp* pEp = GET_ACTIVE_EP(pBuf->pEpSet);
+    SEp* pOld = GET_ACTIVE_EP(&(pMqVg->epSet));
+    uDebug("subKey:%s update the epset vgId:%d, ep:%s:%d, old ep:%s:%d", pParam->pOffset->subKey, pMqVg->vgId,
+           pEp->fqdn, pEp->port, pOld->fqdn, pOld->port);
+    pParam->pMqVg->epSet = *pBuf->pEpSet;
+  }
+
   taosMemoryFree(pParam->pOffset);
   taosMemoryFree(pBuf->pData);
   taosMemoryFree(pBuf->pEpSet);
@@ -444,7 +458,6 @@ int32_t tmqCommitCb(void* param, SDataBuf* pBuf, int32_t code) {
    * pOffset->version);*/
 
   tmqCommitRspCountDown(pParamSet);
-
   return 0;
 }
 
@@ -454,6 +467,7 @@ static int32_t tmqSendCommitReq(tmq_t* tmq, SMqClientVg* pVg, SMqClientTopic* pT
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
+
   pOffset->val = pVg->currentOffset;
 
   int32_t groupLen = strlen(tmq->groupId);
@@ -467,11 +481,13 @@ static int32_t tmqSendCommitReq(tmq_t* tmq, SMqClientVg* pVg, SMqClientTopic* pT
   if (code < 0) {
     return -1;
   }
+
   void* buf = taosMemoryCalloc(1, sizeof(SMsgHead) + len);
   if (buf == NULL) {
     taosMemoryFree(pOffset);
     return -1;
   }
+
   ((SMsgHead*)buf)->vgId = htonl(pVg->vgId);
 
   void* abuf = POINTER_SHIFT(buf, sizeof(SMsgHead));
@@ -488,8 +504,10 @@ static int32_t tmqSendCommitReq(tmq_t* tmq, SMqClientVg* pVg, SMqClientTopic* pT
     taosMemoryFree(buf);
     return -1;
   }
+
   pParam->params = pParamSet;
   pParam->pOffset = pOffset;
+  pParam->pMqVg = pVg;  // there may be an race condition
 
   // build send info
   SMsgSendInfo* pMsgSendInfo = taosMemoryCalloc(1, sizeof(SMsgSendInfo));
@@ -499,16 +517,18 @@ static int32_t tmqSendCommitReq(tmq_t* tmq, SMqClientVg* pVg, SMqClientTopic* pT
     taosMemoryFree(pParam);
     return -1;
   }
+
   pMsgSendInfo->msgInfo = (SDataBuf){
       .pData = buf,
       .len = sizeof(SMsgHead) + len,
       .handle = NULL,
   };
 
-  tscDebug("consumer:%" PRId64 ", commit offset of %s on vgId:%d, offset is %" PRId64, tmq->consumerId, pOffset->subKey,
+  SEp* pEp = &pVg->epSet.eps[pVg->epSet.inUse];
-           pVg->vgId, pOffset->val.version);
+  tscDebug("consumer:0x%" PRIx64 " topic:%s on vgId:%d offset:%" PRId64" prev:%"PRId64", ep:%s:%d", tmq->consumerId, pOffset->subKey,
+           pVg->vgId, pOffset->val.version, pVg->committedOffset.version, pEp->fqdn, pEp->port);
 
-  // TODO: put into cb
+  // TODO: put into cb, the commit offset should be move to the callback function
   pVg->committedOffset = pVg->currentOffset;
 
   pMsgSendInfo->requestId = generateRequestId();
@@ -530,7 +550,7 @@ static int32_t tmqSendCommitReq(tmq_t* tmq, SMqClientVg* pVg, SMqClientTopic* pT
 int32_t tmqCommitMsgImpl(tmq_t* tmq, const TAOS_RES* msg, int8_t async, tmq_commit_cb* userCb, void* userParam) {
   char*   topic;
   int32_t vgId;
-  ASSERT(msg != NULL);
+
   if (TD_RES_TMQ(msg)) {
     SMqRspObj* pRspObj = (SMqRspObj*)msg;
     topic = pRspObj->topic;
@@ -633,24 +653,22 @@ static int32_t tmqCommitConsumerImpl(tmq_t* tmq, int8_t automatic, int8_t async,
   // init as 1 to prevent concurrency issue
   pParamSet->waitingRspNum = 1;
 
-  for (int32_t i = 0; i < taosArrayGetSize(tmq->clientTopics); i++) {
+  int32_t numOfTopics = taosArrayGetSize(tmq->clientTopics);
+  tscDebug("consumer:0x%"PRIx64" start to commit offset for %d topics", tmq->consumerId, numOfTopics);
+
+  for (int32_t i = 0; i < numOfTopics; i++) {
     SMqClientTopic* pTopic = taosArrayGet(tmq->clientTopics, i);
 
-    tscDebug("consumer:%" PRId64 ", begin commit for topic %s, vgNum %d", tmq->consumerId, pTopic->topicName,
+    int32_t numOfVgroups = taosArrayGetSize(pTopic->vgs);
-             (int32_t)taosArrayGetSize(pTopic->vgs));
+    for (int32_t j = 0; j < numOfVgroups; j++) {
-
-    for (int32_t j = 0; j < taosArrayGetSize(pTopic->vgs); j++) {
       SMqClientVg* pVg = taosArrayGet(pTopic->vgs, j);
-
-      tscDebug("consumer:%" PRId64 ", begin commit for topic %s, vgId:%d", tmq->consumerId, pTopic->topicName,
-               pVg->vgId);
-
       if (pVg->currentOffset.type > 0 && !tOffsetEqual(&pVg->currentOffset, &pVg->committedOffset)) {
-        tscDebug("consumer: %" PRId64 ", vg:%d, current %" PRId64 ", committed %" PRId64 "", tmq->consumerId, pVg->vgId,
-                 pVg->currentOffset.version, pVg->committedOffset.version);
         if (tmqSendCommitReq(tmq, pVg, pTopic, pParamSet) < 0) {
           continue;
         }
+      } else {
+        tscDebug("consumer:0x%" PRIx64 " topic:%s vgId:%d, not commit, current:%" PRId64 ", ordinal:%d/%d",
+                 tmq->consumerId, pTopic->topicName, pVg->vgId, pVg->currentOffset.version, j + 1, numOfVgroups);
       }
     }
   }
@@ -786,34 +804,46 @@ OVER:
   taosTmrReset(tmqSendHbReq, 1000, param, tmqMgmt.timer, &tmq->hbLiveTimer);
 }
 
-int32_t tmqHandleAllDelayedTask(tmq_t* tmq) {
+int32_t tmqHandleAllDelayedTask(tmq_t* pTmq) {
   STaosQall* qall = taosAllocateQall();
-  taosReadAllQitems(tmq->delayedTask, qall);
+  taosReadAllQitems(pTmq->delayedTask, qall);
-  while (1) {
-    int8_t* pTaskType = NULL;
-    taosGetQitem(qall, (void**)&pTaskType);
-    if (pTaskType == NULL) break;
 
+  if (qall->numOfItems == 0) {
+    taosFreeQall(qall);
+    return TSDB_CODE_SUCCESS;
+  }
+
+  tscDebug("consumer:0x%"PRIx64" handle delayed %d tasks before poll data", pTmq->consumerId, qall->numOfItems);
+  int8_t* pTaskType = NULL;
+  taosGetQitem(qall, (void**)&pTaskType);
+
+  while (pTaskType != NULL) {
     if (*pTaskType == TMQ_DELAYED_TASK__ASK_EP) {
-      tmqAskEp(tmq, true);
+      tmqAskEp(pTmq, true);
 
       int64_t* pRefId = taosMemoryMalloc(sizeof(int64_t));
-      *pRefId = tmq->refId;
+      *pRefId = pTmq->refId;
 
-      taosTmrReset(tmqAssignAskEpTask, 1000, pRefId, tmqMgmt.timer, &tmq->epTimer);
+      tscDebug("consumer:0x%"PRIx64" next retrieve ep from mnode in 1s", pTmq->consumerId);
+      taosTmrReset(tmqAssignAskEpTask, 1000, pRefId, tmqMgmt.timer, &pTmq->epTimer);
     } else if (*pTaskType == TMQ_DELAYED_TASK__COMMIT) {
-      tmqCommitInner(tmq, NULL, 1, 1, tmq->commitCb, tmq->commitCbUserParam);
+      tmqCommitInner(pTmq, NULL, 1, 1, pTmq->commitCb, pTmq->commitCbUserParam);
 
       int64_t* pRefId = taosMemoryMalloc(sizeof(int64_t));
-      *pRefId = tmq->refId;
+      *pRefId = pTmq->refId;
 
-      taosTmrReset(tmqAssignDelayedCommitTask, tmq->autoCommitInterval, pRefId, tmqMgmt.timer, &tmq->commitTimer);
+      tscDebug("consumer:0x%"PRIx64" next commit to mnode in %.2fs", pTmq->consumerId, pTmq->autoCommitInterval/1000.0);
+      taosTmrReset(tmqAssignDelayedCommitTask, pTmq->autoCommitInterval, pRefId, tmqMgmt.timer, &pTmq->commitTimer);
     } else if (*pTaskType == TMQ_DELAYED_TASK__REPORT) {
+      // do nothing
     } else {
       ASSERT(0);
     }
+
     taosFreeQitem(pTaskType);
+    taosGetQitem(qall, (void**)&pTaskType);
   }
+
   taosFreeQall(qall);
   return 0;
 }
@@ -932,41 +962,45 @@ void tmqFreeImpl(void* handle) {
   taosMemoryFree(tmq);
 }
 
+static void tmqMgmtInit(void) {
+  tmqInitRes = 0;
+  tmqMgmt.timer = taosTmrInit(1000, 100, 360000, "TMQ");
+
+  if (tmqMgmt.timer == NULL) {
+    tmqInitRes = TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  tmqMgmt.rsetId = taosOpenRef(10000, tmqFreeImpl);
+  if (tmqMgmt.rsetId < 0) {
+    tmqInitRes = terrno;
+  }
+}
+
 tmq_t* tmq_consumer_new(tmq_conf_t* conf, char* errstr, int32_t errstrLen) {
-  // init timer
+  taosThreadOnce(&tmqInit, tmqMgmtInit);
-  int8_t inited = atomic_val_compare_exchange_8(&tmqMgmt.inited, 0, 1);
+  if (tmqInitRes != 0) {
-  if (inited == 0) {
+    terrno = tmqInitRes;
-    tmqMgmt.timer = taosTmrInit(1000, 100, 360000, "TMQ");
+    return NULL;
-    if (tmqMgmt.timer == NULL) {
-      atomic_store_8(&tmqMgmt.inited, 0);
-      terrno = TSDB_CODE_OUT_OF_MEMORY;
-      return NULL;
-    }
-    tmqMgmt.rsetId = taosOpenRef(10000, tmqFreeImpl);
   }
 
   tmq_t* pTmq = taosMemoryCalloc(1, sizeof(tmq_t));
   if (pTmq == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
-    tscError("setting up new consumer failed since %s, consumer group %s", terrstr(), conf->groupId);
+    tscError("failed to create consumer, consumer group %s, code:%s", conf->groupId, terrstr());
     return NULL;
   }
 
   const char* user = conf->user == NULL ? TSDB_DEFAULT_USER : conf->user;
   const char* pass = conf->pass == NULL ? TSDB_DEFAULT_PASS : conf->pass;
 
-  ASSERT(user);
-  ASSERT(pass);
-  ASSERT(conf->groupId[0]);
-
   pTmq->clientTopics = taosArrayInit(0, sizeof(SMqClientTopic));
   pTmq->mqueue = taosOpenQueue();
   pTmq->qall = taosAllocateQall();
   pTmq->delayedTask = taosOpenQueue();
 
-  if (pTmq->clientTopics == NULL || pTmq->mqueue == NULL || pTmq->qall == NULL || pTmq->delayedTask == NULL) {
+  if (pTmq->clientTopics == NULL || pTmq->mqueue == NULL || pTmq->qall == NULL || pTmq->delayedTask == NULL || conf->groupId[0] == 0) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
-    tscError("consumer %" PRId64 " setup failed since %s, consumer group %s", pTmq->consumerId, terrstr(),
+    tscError("consumer:0x%" PRIx64 " setup failed since %s, consumer group %s", pTmq->consumerId, terrstr(),
              pTmq->groupId);
     goto FAIL;
   }
@@ -996,7 +1030,7 @@ tmq_t* tmq_consumer_new(tmq_conf_t* conf, char* errstr, int32_t errstrLen) {
 
   // init semaphore
   if (tsem_init(&pTmq->rspSem, 0, 0) != 0) {
-    tscError("consumer %" PRId64 " setup failed since %s, consumer group %s", pTmq->consumerId, terrstr(),
+    tscError("consumer:0x %" PRIx64 " setup failed since %s, consumer group %s", pTmq->consumerId, terrstr(),
              pTmq->groupId);
     goto FAIL;
   }
@@ -1004,7 +1038,7 @@ tmq_t* tmq_consumer_new(tmq_conf_t* conf, char* errstr, int32_t errstrLen) {
   // init connection
   pTmq->pTscObj = taos_connect_internal(conf->ip, user, pass, NULL, NULL, conf->port, CONN_TYPE__TMQ);
   if (pTmq->pTscObj == NULL) {
-    tscError("consumer %" PRId64 " setup failed since %s, consumer group %s", pTmq->consumerId, terrstr(),
+    tscError("consumer:0x %" PRIx64 " setup failed since %s, consumer group %s", pTmq->consumerId, terrstr(),
              pTmq->groupId);
     tsem_destroy(&pTmq->rspSem);
     goto FAIL;
@@ -1022,8 +1056,7 @@ tmq_t* tmq_consumer_new(tmq_conf_t* conf, char* errstr, int32_t errstrLen) {
     pTmq->hbLiveTimer = taosTmrStart(tmqSendHbReq, 1000, pRefId, tmqMgmt.timer);
   }
 
-  tscInfo("consumer %" PRId64 " is setup, consumer group %s", pTmq->consumerId, pTmq->groupId);
+  tscInfo("consumer:0x%" PRIx64 " is setup, consumer groupId %s", pTmq->consumerId, pTmq->groupId);
-
   return pTmq;
 
 FAIL:
@@ -1032,6 +1065,7 @@ FAIL:
   if (pTmq->delayedTask) taosCloseQueue(pTmq->delayedTask);
   if (pTmq->qall) taosFreeQall(pTmq->qall);
   taosMemoryFree(pTmq);
+
   return NULL;
 }
 
@@ -1041,44 +1075,52 @@ int32_t tmq_subscribe(tmq_t* tmq, const tmq_list_t* topic_list) {
   void*           buf = NULL;
   SMsgSendInfo*   sendInfo = NULL;
   SCMSubscribeReq req = {0};
-  int32_t         code = -1;
+  int32_t         code = 0;
 
-  tscDebug("tmq subscribe, consumer: %" PRId64 ", topic num %d", tmq->consumerId, sz);
+  tscDebug("consumer:0x%"PRIx64" subscribe %d topics", tmq->consumerId, sz);
 
   req.consumerId = tmq->consumerId;
   tstrncpy(req.clientId, tmq->clientId, 256);
   tstrncpy(req.cgroup, tmq->groupId, TSDB_CGROUP_LEN);
   req.topicNames = taosArrayInit(sz, sizeof(void*));
-  if (req.topicNames == NULL) goto FAIL;
 
-  tscDebug("tmq subscribe, consumer: %" PRId64 ", topic num %d", tmq->consumerId, sz);
+  if (req.topicNames == NULL) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto FAIL;
+  }
 
   for (int32_t i = 0; i < sz; i++) {
     char* topic = taosArrayGetP(container, i);
 
     SName name = {0};
     tNameSetDbName(&name, tmq->pTscObj->acctId, topic, strlen(topic));
-
     char* topicFName = taosMemoryCalloc(1, TSDB_TOPIC_FNAME_LEN);
     if (topicFName == NULL) {
       goto FAIL;
     }
-    tNameExtractFullName(&name, topicFName);
 
-    tscDebug("subscribe topic: %s", topicFName);
+    tNameExtractFullName(&name, topicFName);
+    tscDebug("consumer:0x%"PRIx64" subscribe topic:%s", tmq->consumerId, topicFName);
 
     taosArrayPush(req.topicNames, &topicFName);
   }
 
   int32_t tlen = tSerializeSCMSubscribeReq(NULL, &req);
+
   buf = taosMemoryMalloc(tlen);
-  if (buf == NULL) goto FAIL;
+  if (buf == NULL) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto FAIL;
+  }
 
   void* abuf = buf;
   tSerializeSCMSubscribeReq(&abuf, &req);
 
   sendInfo = taosMemoryCalloc(1, sizeof(SMsgSendInfo));
-  if (sendInfo == NULL) goto FAIL;
+  if (sendInfo == NULL) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto FAIL;
+  }
 
   SMqSubscribeCbParam param = {
       .rspErr = 0,
@@ -1086,7 +1128,9 @@ int32_t tmq_subscribe(tmq_t* tmq, const tmq_list_t* topic_list) {
       .epoch = tmq->epoch,
   };
 
-  if (tsem_init(&param.rspSem, 0, 0) != 0) goto FAIL;
+  if (tsem_init(&param.rspSem, 0, 0) != 0) {
+    goto FAIL;
+  }
 
   sendInfo->msgInfo = (SDataBuf){
       .pData = buf,
@@ -1112,15 +1156,18 @@ int32_t tmq_subscribe(tmq_t* tmq, const tmq_list_t* topic_list) {
   tsem_wait(&param.rspSem);
   tsem_destroy(&param.rspSem);
 
-  code = param.rspErr;
+  if (param.rspErr != 0) {
-  if (code != 0) goto FAIL;
+    code = param.rspErr;
+    goto FAIL;
+  }
 
   int32_t retryCnt = 0;
   while (TSDB_CODE_MND_CONSUMER_NOT_READY == tmqAskEp(tmq, false)) {
     if (retryCnt++ > 10) {
       goto FAIL;
     }
-    tscDebug("consumer not ready, retry");
+
+    tscDebug("consumer:0x%"PRIx64", mnd not ready for subscribe, retry:%d in 500ms", tmq->consumerId, retryCnt);
     taosMsleep(500);
   }
 
@@ -1138,7 +1185,6 @@ int32_t tmq_subscribe(tmq_t* tmq, const tmq_list_t* topic_list) {
     tmq->commitTimer = taosTmrStart(tmqAssignDelayedCommitTask, tmq->autoCommitInterval, pRefId2, tmqMgmt.timer);
   }
 
-  code = 0;
 FAIL:
   taosArrayDestroyP(req.topicNames, taosMemoryFree);
   taosMemoryFree(buf);
@@ -1233,7 +1279,7 @@ int32_t tmqPollCb(void* param, SDataBuf* pMsg, int32_t code) {
     tDecoderClear(&decoder);
     memcpy(&pRspWrapper->dataRsp, pMsg->pData, sizeof(SMqRspHead));
 
-    tscDebug("consumer:%" PRId64 ", recv poll: vgId:%d, req offset %" PRId64 ", rsp offset %" PRId64 " type %d",
+    tscDebug("consumer:0x%" PRIx64 ", recv poll: vgId:%d, req offset %" PRId64 ", rsp offset %" PRId64 " type %d",
              tmq->consumerId, pVg->vgId, pRspWrapper->dataRsp.reqOffset.version, pRspWrapper->dataRsp.rspOffset.version,
              rspType);
 
@@ -1256,7 +1302,7 @@ int32_t tmqPollCb(void* param, SDataBuf* pMsg, int32_t code) {
   taosMemoryFree(pMsg->pData);
   taosMemoryFree(pMsg->pEpSet);
 
-  tscDebug("consumer:%" PRId64 ", put poll res into mqueue %p", tmq->consumerId, pRspWrapper);
+  tscDebug("consumer:0x%" PRIx64 ", put poll res into mqueue %p", tmq->consumerId, pRspWrapper);
 
   taosWriteQitem(tmq->mqueue, pRspWrapper);
   tsem_post(&tmq->rspSem);
@@ -1273,10 +1319,12 @@ CREATE_MSG_FAIL:
 bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, const SMqAskEpRsp* pRsp) {
   bool set = false;
 
+  int32_t topicNumCur = taosArrayGetSize(tmq->clientTopics);
   int32_t topicNumGet = taosArrayGetSize(pRsp->topics);
+
   char    vgKey[TSDB_TOPIC_FNAME_LEN + 22];
-  tscDebug("consumer:%" PRId64 ", update ep epoch %d to epoch %d, topic num:%d", tmq->consumerId, tmq->epoch, epoch,
+  tscDebug("consumer:0x%" PRIx64" update ep epoch from %d to epoch %d, incoming topics:%d, existed topics:%d",
-           topicNumGet);
+           tmq->consumerId, tmq->epoch, epoch, topicNumGet, topicNumCur);
 
   SArray* newTopics = taosArrayInit(topicNumGet, sizeof(SMqClientTopic));
   if (newTopics == NULL) {
@@ -1288,19 +1336,19 @@ bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, const SMqAskEpRsp* pRsp) {
     taosArrayDestroy(newTopics);
     return false;
   }
-  int32_t topicNumCur = taosArrayGetSize(tmq->clientTopics);
+
   for (int32_t i = 0; i < topicNumCur; i++) {
     // find old topic
     SMqClientTopic* pTopicCur = taosArrayGet(tmq->clientTopics, i);
     if (pTopicCur->vgs) {
       int32_t vgNumCur = taosArrayGetSize(pTopicCur->vgs);
-      tscDebug("consumer:%" PRId64 ", new vg num: %d", tmq->consumerId, vgNumCur);
+      tscDebug("consumer:0x%" PRIx64 ", new vg num: %d", tmq->consumerId, vgNumCur);
       for (int32_t j = 0; j < vgNumCur; j++) {
         SMqClientVg* pVgCur = taosArrayGet(pTopicCur->vgs, j);
         sprintf(vgKey, "%s:%d", pTopicCur->topicName, pVgCur->vgId);
         char buf[80];
         tFormatOffset(buf, 80, &pVgCur->currentOffset);
-        tscDebug("consumer:%" PRId64 ", epoch %d vgId:%d vgKey is %s, offset is %s", tmq->consumerId, epoch,
+        tscDebug("consumer:0x%" PRIx64 ", epoch %d vgId:%d vgKey is %s, offset is %s", tmq->consumerId, epoch,
                  pVgCur->vgId, vgKey, buf);
         taosHashPut(pHash, vgKey, strlen(vgKey), &pVgCur->currentOffset, sizeof(STqOffsetVal));
       }
@@ -1316,7 +1364,7 @@ bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, const SMqAskEpRsp* pRsp) {
     tstrncpy(topic.topicName, pTopicEp->topic, TSDB_TOPIC_FNAME_LEN);
     tstrncpy(topic.db, pTopicEp->db, TSDB_DB_FNAME_LEN);
 
-    tscDebug("consumer:%" PRId64 ", update topic: %s", tmq->consumerId, topic.topicName);
+    tscDebug("consumer:0x%" PRIx64 ", update topic: %s", tmq->consumerId, topic.topicName);
 
     int32_t vgNumGet = taosArrayGetSize(pTopicEp->vgs);
     topic.vgs = taosArrayInit(vgNumGet, sizeof(SMqClientVg));
@@ -1342,6 +1390,8 @@ bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, const SMqAskEpRsp* pRsp) {
     }
     taosArrayPush(newTopics, &topic);
   }
+
+  // destroy current buffered existed topics info
   if (tmq->clientTopics) {
     int32_t sz = taosArrayGetSize(tmq->clientTopics);
     for (int32_t i = 0; i < sz; i++) {
@@ -1349,17 +1399,21 @@ bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, const SMqAskEpRsp* pRsp) {
       if (pTopic->schema.nCols) taosMemoryFreeClear(pTopic->schema.pSchema);
       taosArrayDestroy(pTopic->vgs);
     }
+
     taosArrayDestroy(tmq->clientTopics);
   }
+
   taosHashCleanup(pHash);
   tmq->clientTopics = newTopics;
 
-  if (taosArrayGetSize(tmq->clientTopics) == 0)
+  if (taosArrayGetSize(tmq->clientTopics) == 0) {
     atomic_store_8(&tmq->status, TMQ_CONSUMER_STATUS__NO_TOPIC);
-  else
+  } else {
     atomic_store_8(&tmq->status, TMQ_CONSUMER_STATUS__READY);
+  }
 
   atomic_store_32(&tmq->epoch, epoch);
+  tscDebug("consumer:0x%" PRIx64 " update topic info completed", tmq->consumerId);
   return set;
 }
 
@@ -1382,8 +1436,8 @@ int32_t tmqAskEpCb(void* param, SDataBuf* pMsg, int32_t code) {
 
   pParam->code = code;
   if (code != 0) {
-    tscError("consumer:%" PRId64 ", get topic endpoint error, not ready, wait:%d, code %x", tmq->consumerId,
+    tscError("consumer:0x%" PRIx64 ", get topic endpoint error, async:%d, code:%s", tmq->consumerId,
-             pParam->async, code);
+             pParam->async, tstrerror(code));
     goto END;
   }
 
@@ -1392,11 +1446,15 @@ int32_t tmqAskEpCb(void* param, SDataBuf* pMsg, int32_t code) {
   // Epoch will only increase when received newer epoch ep msg
   SMqRspHead* head = pMsg->pData;
   int32_t     epoch = atomic_load_32(&tmq->epoch);
-  tscDebug("consumer:%" PRId64 ", recv ep, msg epoch %d, current epoch %d", tmq->consumerId, head->epoch, epoch);
   if (head->epoch <= epoch) {
+    tscDebug("consumer:0x%" PRIx64 ", recv ep, msg epoch %d, current epoch %d, no need to update local ep",
+             tmq->consumerId, head->epoch, epoch);
     goto END;
   }
 
+  tscDebug("consumer:0x%" PRIx64 ", recv ep, msg epoch %d, current epoch %d, update local ep", tmq->consumerId,
+           head->epoch, epoch);
+
   if (!async) {
     SMqAskEpRsp rsp;
     tDecodeSMqAskEpRsp(POINTER_SHIFT(pMsg->pData, sizeof(SMqRspHead)), &rsp);
@@ -1411,6 +1469,7 @@ int32_t tmqAskEpCb(void* param, SDataBuf* pMsg, int32_t code) {
       code = -1;
       goto END;
     }
+
     pWrapper->tmqRspType = TMQ_MSG_TYPE__EP_RSP;
     pWrapper->epoch = head->epoch;
     memcpy(&pWrapper->msg, pMsg->pData, sizeof(SMqRspHead));
@@ -1434,16 +1493,17 @@ END:
 }
 
 int32_t tmqAskEp(tmq_t* tmq, bool async) {
-  int32_t code = 0;
+  int32_t code = TSDB_CODE_SUCCESS;
 #if 0
   int8_t  epStatus = atomic_val_compare_exchange_8(&tmq->epStatus, 0, 1);
   if (epStatus == 1) {
     int32_t epSkipCnt = atomic_add_fetch_32(&tmq->epSkipCnt, 1);
-    tscTrace("consumer:%" PRId64 ", skip ask ep cnt %d", tmq->consumerId, epSkipCnt);
+    tscTrace("consumer:0x%" PRIx64 ", skip ask ep cnt %d", tmq->consumerId, epSkipCnt);
     if (epSkipCnt < 5000) return 0;
   }
   atomic_store_32(&tmq->epSkipCnt, 0);
 #endif
+
   SMqAskEpReq req = {0};
   req.consumerId = tmq->consumerId;
   req.epoch = tmq->epoch;
@@ -1451,27 +1511,31 @@ int32_t tmqAskEp(tmq_t* tmq, bool async) {
 
   int32_t tlen = tSerializeSMqAskEpReq(NULL, 0, &req);
   if (tlen < 0) {
-    tscError("tSerializeSMqAskEpReq failed");
+    tscError("consumer:0x%"PRIx64", tSerializeSMqAskEpReq failed", tmq->consumerId);
     return -1;
   }
+
   void* pReq = taosMemoryCalloc(1, tlen);
   if (pReq == NULL) {
-    tscError("failed to malloc askEpReq msg, size:%d", tlen);
+    tscError("consumer:0x%"PRIx64", failed to malloc askEpReq msg, size:%d", tmq->consumerId, tlen);
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
+
   if (tSerializeSMqAskEpReq(pReq, tlen, &req) < 0) {
-    tscError("tSerializeSMqAskEpReq %d failed", tlen);
+    tscError("consumer:0x%"PRIx64", tSerializeSMqAskEpReq %d failed", tmq->consumerId, tlen);
     taosMemoryFree(pReq);
     return -1;
   }
 
   SMqAskEpCbParam* pParam = taosMemoryCalloc(1, sizeof(SMqAskEpCbParam));
   if (pParam == NULL) {
-    tscError("failed to malloc subscribe param");
+    tscError("consumer:0x%"PRIx64", failed to malloc subscribe param", tmq->consumerId);
     taosMemoryFree(pReq);
     /*atomic_store_8(&tmq->epStatus, 0);*/
     return -1;
   }
+
   pParam->refId = tmq->refId;
   pParam->epoch = tmq->epoch;
   pParam->async = async;
@@ -1492,15 +1556,14 @@ int32_t tmqAskEp(tmq_t* tmq, bool async) {
       .handle = NULL,
   };
 
-  sendInfo->requestId = generateRequestId();
+  sendInfo->requestId = tmq->consumerId;
   sendInfo->requestObjRefId = 0;
   sendInfo->param = pParam;
   sendInfo->fp = tmqAskEpCb;
   sendInfo->msgType = TDMT_MND_TMQ_ASK_EP;
 
   SEpSet epSet = getEpSet_s(&tmq->pTscObj->pAppInfo->mgmtEp);
-
+  tscDebug("consumer:0x%" PRIx64 " ask ep from mnode, async:%d, reqId:0x%"PRIx64, tmq->consumerId, async, tmq->consumerId);
-  tscDebug("consumer:%" PRId64 ", ask ep", tmq->consumerId);
 
   int64_t transporterId = 0;
   asyncSendMsgToServer(tmq->pTscObj->pAppInfo->pTransporter, &epSet, &transporterId, sendInfo);
@@ -1510,6 +1573,7 @@ int32_t tmqAskEp(tmq_t* tmq, bool async) {
     code = pParam->code;
     taosMemoryFree(pParam);
   }
+
   return code;
 }
 
@@ -1582,15 +1646,20 @@ SMqTaosxRspObj* tmqBuildTaosxRspFromWrapper(SMqPollRspWrapper* pWrapper) {
   return pRspObj;
 }
 
+// broadcast the poll request to all related vnodes
 int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
-  for (int i = 0; i < taosArrayGetSize(tmq->clientTopics); i++) {
+  int32_t numOfTopics = taosArrayGetSize(tmq->clientTopics);
+  tscDebug("consumer:0x%" PRIx64" start to poll data, numOfTopics:%d", tmq->consumerId, numOfTopics);
+
+  for (int i = 0; i < numOfTopics; i++) {
+
     SMqClientTopic* pTopic = taosArrayGet(tmq->clientTopics, i);
     for (int j = 0; j < taosArrayGetSize(pTopic->vgs); j++) {
       SMqClientVg* pVg = taosArrayGet(pTopic->vgs, j);
       int32_t      vgStatus = atomic_val_compare_exchange_32(&pVg->vgStatus, TMQ_VG_STATUS__IDLE, TMQ_VG_STATUS__WAIT);
-      if (vgStatus != TMQ_VG_STATUS__IDLE) {
+      if (vgStatus == TMQ_VG_STATUS__WAIT) {
         int32_t vgSkipCnt = atomic_add_fetch_32(&pVg->vgSkipCnt, 1);
-        tscTrace("consumer:%" PRId64 ", epoch %d skip vgId:%d skip cnt %d", tmq->consumerId, tmq->epoch, pVg->vgId,
+        tscDebug("consumer:0x%" PRIx64 " epoch %d wait poll-rsp, skip vgId:%d skip cnt %d", tmq->consumerId, tmq->epoch, pVg->vgId,
                  vgSkipCnt);
         continue;
         /*if (vgSkipCnt < 10000) continue;*/
@@ -1598,10 +1667,11 @@ int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
         if (skipCnt < 30000) {
           continue;
         } else {
-        tscDebug("consumer:%" PRId64 ",skip vgId:%d skip too much reset", tmq->consumerId, pVg->vgId);
+        tscDebug("consumer:0x%" PRIx64 ",skip vgId:%d skip too much reset", tmq->consumerId, pVg->vgId);
         }
 #endif
       }
+
       atomic_store_32(&pVg->vgSkipCnt, 0);
 
       SMqPollReq req = {0};
@@ -1612,6 +1682,7 @@ int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
         tsem_post(&tmq->rspSem);
         return -1;
       }
+
       char* msg = taosMemoryCalloc(1, msgSize);
       if (NULL == msg) {
         atomic_store_32(&pVg->vgStatus, TMQ_VG_STATUS__IDLE);
@@ -1633,6 +1704,7 @@ int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
         tsem_post(&tmq->rspSem);
         return -1;
       }
+
       pParam->refId = tmq->refId;
       pParam->epoch = tmq->epoch;
 
@@ -1654,6 +1726,7 @@ int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
           .len = msgSize,
           .handle = NULL,
       };
+
       sendInfo->requestId = req.reqId;
       sendInfo->requestObjRefId = 0;
       sendInfo->param = pParam;
@@ -1661,18 +1734,19 @@ int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
       sendInfo->msgType = TDMT_VND_TMQ_CONSUME;
 
       int64_t transporterId = 0;
-      /*printf("send poll\n");*/
 
       char offsetFormatBuf[80];
       tFormatOffset(offsetFormatBuf, 80, &pVg->currentOffset);
-      tscDebug("consumer:%" PRId64 ", send poll to %s vgId:%d, epoch %d, req offset:%s, reqId:%" PRIu64,
+
+      tscDebug("consumer:0x%" PRIx64 ", send poll to %s vgId:%d, epoch %d, req offset:%s, reqId:0x%" PRIx64,
                tmq->consumerId, pTopic->topicName, pVg->vgId, tmq->epoch, offsetFormatBuf, req.reqId);
-      /*printf("send vgId:%d %" PRId64 "\n", pVg->vgId, pVg->currentOffset);*/
       asyncSendMsgToServer(tmq->pTscObj->pAppInfo->pTransporter, &pVg->epSet, &transporterId, sendInfo);
+
       pVg->pollCnt++;
       tmq->pollCnt++;
     }
   }
+
   return 0;
 }
 
@@ -1697,28 +1771,28 @@ int32_t tmqHandleNoPollRsp(tmq_t* tmq, SMqRspWrapper* rspWrapper, bool* pReset)
 }
 
 void* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout, bool pollIfReset) {
+  tscDebug("consumer:0x%"PRIx64" start to handle the rsp", tmq->consumerId);
+
   while (1) {
     SMqRspWrapper* rspWrapper = NULL;
     taosGetQitem(tmq->qall, (void**)&rspWrapper);
+
     if (rspWrapper == NULL) {
       taosReadAllQitems(tmq->mqueue, tmq->qall);
       taosGetQitem(tmq->qall, (void**)&rspWrapper);
 
       if (rspWrapper == NULL) {
-        /*tscDebug("consumer %" PRId64 " mqueue empty", tmq->consumerId);*/
         return NULL;
       }
     }
 
-    tscDebug("consumer:%" PRId64 " handle rsp %p", tmq->consumerId, rspWrapper);
-
     if (rspWrapper->tmqRspType == TMQ_MSG_TYPE__END_RSP) {
       taosFreeQitem(rspWrapper);
       terrno = TSDB_CODE_TQ_NO_COMMITTED_OFFSET;
       return NULL;
     } else if (rspWrapper->tmqRspType == TMQ_MSG_TYPE__POLL_RSP) {
       SMqPollRspWrapper* pollRspWrapper = (SMqPollRspWrapper*)rspWrapper;
-      tscDebug("consumer %" PRId64 " actual process poll rsp", tmq->consumerId);
+      tscDebug("consumer:0x%" PRIx64 " process poll rsp", tmq->consumerId);
       /*atomic_sub_fetch_32(&tmq->readyRequest, 1);*/
       int32_t consumerEpoch = atomic_load_32(&tmq->epoch);
       if (pollRspWrapper->dataRsp.head.epoch == consumerEpoch) {
@@ -1737,14 +1811,17 @@ void* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout, bool pollIfReset) {
         taosFreeQitem(pollRspWrapper);
         return pRsp;
       } else {
-        tscDebug("msg discard since epoch mismatch: msg epoch %d, consumer epoch %d",
+        tscDebug("consumer:0x%"PRIx64" msg discard since epoch mismatch: msg epoch %d, consumer epoch %d",
-                 pollRspWrapper->dataRsp.head.epoch, consumerEpoch);
+                 tmq->consumerId, pollRspWrapper->dataRsp.head.epoch, consumerEpoch);
         tmqFreeRspWrapper(rspWrapper);
         taosFreeQitem(pollRspWrapper);
       }
     } else if (rspWrapper->tmqRspType == TMQ_MSG_TYPE__POLL_META_RSP) {
       SMqPollRspWrapper* pollRspWrapper = (SMqPollRspWrapper*)rspWrapper;
       int32_t            consumerEpoch = atomic_load_32(&tmq->epoch);
+
+      tscDebug("consumer:0x%" PRIx64 " process meta rsp", tmq->consumerId);
+
       if (pollRspWrapper->metaRsp.head.epoch == consumerEpoch) {
         SMqClientVg* pVg = pollRspWrapper->vgHandle;
         /*printf("vgId:%d, offset %" PRId64 " up to %" PRId64 "\n", pVg->vgId, pVg->currentOffset,
@@ -1756,8 +1833,8 @@ void* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout, bool pollIfReset) {
         taosFreeQitem(pollRspWrapper);
         return pRsp;
       } else {
-        tscDebug("msg discard since epoch mismatch: msg epoch %d, consumer epoch %d",
+        tscDebug("consumer:0x%"PRIx64" msg discard since epoch mismatch: msg epoch %d, consumer epoch %d",
-                 pollRspWrapper->metaRsp.head.epoch, consumerEpoch);
+                 tmq->consumerId, pollRspWrapper->metaRsp.head.epoch, consumerEpoch);
         tmqFreeRspWrapper(rspWrapper);
         taosFreeQitem(pollRspWrapper);
       }
@@ -1787,8 +1864,8 @@ void* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout, bool pollIfReset) {
         taosFreeQitem(pollRspWrapper);
         return pRsp;
       } else {
-        tscDebug("msg discard since epoch mismatch: msg epoch %d, consumer epoch %d",
+        tscDebug("consumer:0x%"PRIx64" msg discard since epoch mismatch: msg epoch %d, consumer epoch %d",
-                 pollRspWrapper->taosxRsp.head.epoch, consumerEpoch);
+                 tmq->consumerId, pollRspWrapper->taosxRsp.head.epoch, consumerEpoch);
         tmqFreeRspWrapper(rspWrapper);
         taosFreeQitem(pollRspWrapper);
       }
@@ -1798,7 +1875,7 @@ void* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout, bool pollIfReset) {
       tmqHandleNoPollRsp(tmq, rspWrapper, &reset);
       taosFreeQitem(rspWrapper);
       if (pollIfReset && reset) {
-        tscDebug("consumer:%" PRId64 ", reset and repoll", tmq->consumerId);
+        tscDebug("consumer:0x%" PRIx64 ", reset and repoll", tmq->consumerId);
         tmqPollImpl(tmq, timeout);
       }
     }
@@ -1809,7 +1886,7 @@ TAOS_RES* tmq_consumer_poll(tmq_t* tmq, int64_t timeout) {
   void*   rspObj;
   int64_t startTime = taosGetTimestampMs();
 
-  tscDebug("consumer:%" PRId64 ", start poll at %" PRId64, tmq->consumerId, startTime);
+  tscDebug("consumer:0x%" PRIx64 " start to poll at %" PRId64, tmq->consumerId, startTime);
 
 #if 0
   tmqHandleAllDelayedTask(tmq);
@@ -1822,7 +1899,7 @@ TAOS_RES* tmq_consumer_poll(tmq_t* tmq, int64_t timeout) {
 
   // in no topic status, delayed task also need to be processed
   if (atomic_load_8(&tmq->status) == TMQ_CONSUMER_STATUS__INIT) {
-    tscDebug("consumer:%" PRId64 ", poll return since consumer status is init", tmq->consumerId);
+    tscDebug("consumer:0x%" PRIx64 " poll return since consumer is init", tmq->consumerId);
     return NULL;
   }
 
@@ -1832,38 +1909,41 @@ TAOS_RES* tmq_consumer_poll(tmq_t* tmq, int64_t timeout) {
       if (retryCnt++ > 10) {
         return NULL;
       }
-      tscDebug("consumer not ready, retry");
+
+      tscDebug("consumer:0x%"PRIx64" not ready, retry:%d/10 in 500ms", tmq->consumerId, retryCnt);
       taosMsleep(500);
     }
   }
 
   while (1) {
     tmqHandleAllDelayedTask(tmq);
+
     if (tmqPollImpl(tmq, timeout) < 0) {
-      tscDebug("consumer:%" PRId64 " return since poll err", tmq->consumerId);
+      tscDebug("consumer:0x%" PRIx64 " return due to poll error", tmq->consumerId);
       /*return NULL;*/
     }
 
     rspObj = tmqHandleAllRsp(tmq, timeout, false);
     if (rspObj) {
-      tscDebug("consumer:%" PRId64 ", return rsp %p", tmq->consumerId, rspObj);
+      tscDebug("consumer:0x%" PRIx64 " return rsp %p", tmq->consumerId, rspObj);
       return (TAOS_RES*)rspObj;
     } else if (terrno == TSDB_CODE_TQ_NO_COMMITTED_OFFSET) {
-      tscDebug("consumer:%" PRId64 ", return null since no committed offset", tmq->consumerId);
+      tscDebug("consumer:0x%" PRIx64 " return null since no committed offset", tmq->consumerId);
       return NULL;
     }
+
     if (timeout != -1) {
       int64_t currentTime = taosGetTimestampMs();
-      int64_t passedTime = currentTime - startTime;
+      int64_t elapsedTime = currentTime - startTime;
-      if (passedTime > timeout) {
+      if (elapsedTime > timeout) {
-        tscDebug("consumer:%" PRId64 ", (epoch %d) timeout, no rsp, start time %" PRId64 ", current time %" PRId64,
+        tscDebug("consumer:0x%" PRIx64 " (epoch %d) timeout, no rsp, start time %" PRId64 ", current time %" PRId64,
                  tmq->consumerId, tmq->epoch, startTime, currentTime);
         return NULL;
       }
-      /*tscInfo("consumer:%" PRId64 ", (epoch %d) wait, start time %" PRId64 ", current time %" PRId64*/
+      /*tscInfo("consumer:0x%" PRIx64 ", (epoch %d) wait, start time %" PRId64 ", current time %" PRId64*/
       /*", left time %" PRId64,*/
-      /*tmq->consumerId, tmq->epoch, startTime, currentTime, (timeout - passedTime));*/
+      /*tmq->consumerId, tmq->epoch, startTime, currentTime, (timeout - elapsedTime));*/
-      tsem_timewait(&tmq->rspSem, (timeout - passedTime));
+      tsem_timewait(&tmq->rspSem, (timeout - elapsedTime));
     } else {
       // use tsem_timewait instead of tsem_wait to avoid unexpected stuck
       tsem_timewait(&tmq->rspSem, 1000);

source/client/test/clientTests.cpp

@@ -112,7 +112,7 @@ void createNewTable(TAOS* pConn, int32_t index) {
   }
   taos_free_result(pRes);
 
-  for(int32_t i = 0; i < 20; i += 20) {
+  for(int32_t i = 0; i < 2000; i += 20) {
     char sql[1024] = {0};
     sprintf(sql,
             "insert into tu%d values(now+%da, %d)(now+%da, %d)(now+%da, %d)(now+%da, %d)"
@@ -162,6 +162,11 @@ void *queryThread(void *arg) {
 }
 
 static int32_t numOfThreads = 1;
+
+void tmq_commit_cb_print(tmq_t *pTmq, int32_t code, void *param) {
+  printf("success, code:%d\n", code);
+}
+
 }  // namespace
 
 int main(int argc, char** argv) {
@@ -176,12 +181,12 @@ int main(int argc, char** argv) {
   return RUN_ALL_TESTS();
 }
 
-TEST(testCase, driverInit_Test) {
+TEST(clientCase, driverInit_Test) {
   // taosInitGlobalCfg();
   //  taos_init();
 }
 
-TEST(testCase, connect_Test) {
+TEST(clientCase, connect_Test) {
   taos_options(TSDB_OPTION_CONFIGDIR, "~/first/cfg");
 
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
@@ -190,8 +195,8 @@ TEST(testCase, connect_Test) {
   }
   taos_close(pConn);
 }
-#if 0
+
-TEST(testCase, create_user_Test) {
+TEST(clientCase, create_user_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -204,7 +209,7 @@ TEST(testCase, create_user_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, create_account_Test) {
+TEST(clientCase, create_account_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -217,7 +222,7 @@ TEST(testCase, create_account_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, drop_account_Test) {
+TEST(clientCase, drop_account_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -230,7 +235,7 @@ TEST(testCase, drop_account_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, show_user_Test) {
+TEST(clientCase, show_user_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -250,7 +255,7 @@ TEST(testCase, show_user_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, drop_user_Test) {
+TEST(clientCase, drop_user_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -263,7 +268,7 @@ TEST(testCase, drop_user_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, show_db_Test) {
+TEST(clientCase, show_db_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -282,7 +287,7 @@ TEST(testCase, show_db_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, create_db_Test) {
+TEST(clientCase, create_db_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -306,7 +311,7 @@ TEST(testCase, create_db_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, create_dnode_Test) {
+TEST(clientCase, create_dnode_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -325,7 +330,7 @@ TEST(testCase, create_dnode_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, drop_dnode_Test) {
+TEST(clientCase, drop_dnode_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -349,7 +354,7 @@ TEST(testCase, drop_dnode_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, use_db_test) {
+TEST(clientCase, use_db_test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -367,7 +372,7 @@ TEST(testCase, use_db_test) {
   taos_close(pConn);
 }
 
-// TEST(testCase, drop_db_test) {
+// TEST(clientCase, drop_db_test) {
 //  TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
 //  assert(pConn != NULL);
 //
@@ -389,7 +394,7 @@ TEST(testCase, use_db_test) {
 //  taos_close(pConn);
 //}
 
-TEST(testCase, create_stable_Test) {
+TEST(clientCase, create_stable_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -428,7 +433,7 @@ TEST(testCase, create_stable_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, create_table_Test) {
+TEST(clientCase, create_table_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -447,7 +452,7 @@ TEST(testCase, create_table_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, create_ctable_Test) {
+TEST(clientCase, create_ctable_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -472,7 +477,7 @@ TEST(testCase, create_ctable_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, show_stable_Test) {
+TEST(clientCase, show_stable_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != nullptr);
 
@@ -497,7 +502,7 @@ TEST(testCase, show_stable_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, show_vgroup_Test) {
+TEST(clientCase, show_vgroup_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -529,7 +534,7 @@ TEST(testCase, show_vgroup_Test) {
   taos_close(pConn);
 }
 
-TEST(testCase, create_multiple_tables) {
+TEST(clientCase, create_multiple_tables) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
@@ -600,7 +605,7 @@ TEST(testCase, create_multiple_tables) {
   taos_close(pConn);
 }
 
-TEST(testCase, show_table_Test) {
+TEST(clientCase, show_table_Test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   assert(pConn != NULL);
 
@@ -634,7 +639,7 @@ TEST(testCase, show_table_Test) {
   taos_close(pConn);
 }
 
-//TEST(testCase, drop_stable_Test) {
+//TEST(clientCase, drop_stable_Test) {
 //  TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
 //  assert(pConn != nullptr);
 //
@@ -659,14 +664,14 @@ TEST(testCase, show_table_Test) {
 //  taos_close(pConn);
 //}
 
-TEST(testCase, generated_request_id_test) {
+TEST(clientCase, generated_request_id_test) {
   SHashObj* phash = taosHashInit(10000, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_ENTRY_LOCK);
 
   for (int32_t i = 0; i < 50000; ++i) {
     uint64_t v = generateRequestId();
     void*    result = taosHashGet(phash, &v, sizeof(v));
     if (result != nullptr) {
-      printf("0x%lx, index:%d\n", v, i);
+//      printf("0x%llx, index:%d\n", v, i);
     }
     assert(result == nullptr);
     taosHashPut(phash, &v, sizeof(v), NULL, 0);
@@ -675,7 +680,7 @@ TEST(testCase, generated_request_id_test) {
   taosHashCleanup(phash);
 }
 
-TEST(testCase, insert_test) {
+TEST(clientCase, insert_test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
@@ -693,7 +698,7 @@ TEST(testCase, insert_test) {
   taos_close(pConn);
 }
 
-TEST(testCase, projection_query_tables) {
+TEST(clientCase, projection_query_tables) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
@@ -725,7 +730,7 @@ TEST(testCase, projection_query_tables) {
   }
   taos_free_result(pRes);
 
-  for (int32_t i = 0; i < 200000; ++i) {
+  for (int32_t i = 0; i < 2; ++i) {
     printf("create table :%d\n", i);
     createNewTable(pConn, i);
   }
@@ -751,7 +756,7 @@ TEST(testCase, projection_query_tables) {
   taos_close(pConn);
 }
 
-TEST(testCase, tsbs_perf_test) {
+TEST(clientCase, tsbs_perf_test) {
   TdThread qid[20] = {0};
 
   for(int32_t i = 0; i < numOfThreads; ++i) {
@@ -760,9 +765,7 @@ TEST(testCase, tsbs_perf_test) {
   getchar();
 }
 
-#endif
+TEST(clientCase, projection_query_stables) {
-
-TEST(testCase, projection_query_stables) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
@@ -790,8 +793,7 @@ TEST(testCase, projection_query_stables) {
   taos_close(pConn);
 }
 
-#if 0
+TEST(clientCase, agg_query_tables) {
-TEST(testCase, agg_query_tables) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
@@ -826,12 +828,12 @@ create table tm1 using m1 tags(2);
 insert into tm0 values('2021-1-1 1:1:1.120', 1) ('2021-1-1 1:1:2.9', 2) tm1 values('2021-1-1 1:1:1.120', 11) ('2021-1-1 1:1:2.99', 22);
 
  */
-TEST(testCase, async_api_test) {
+TEST(clientCase, async_api_test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
   taos_query(pConn, "use abc1");
-#if 0
+
   TAOS_RES* pRes = taos_query(pConn, "insert into tu(ts) values('2022-02-27 12:12:61')");
   if (taos_errno(pRes) != 0) {
     printf("failed, reason:%s\n", taos_errstr(pRes));
@@ -854,14 +856,13 @@ TEST(testCase, async_api_test) {
     printf("%s\n", str);
     memset(str, 0, sizeof(str));
   }
-#endif
 
   taos_query_a(pConn, "select count(*) from tu", queryCallback, pConn);
   getchar();
   taos_close(pConn);
 }
 
-TEST(testCase, update_test) {
+TEST(clientCase, update_test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
@@ -897,6 +898,76 @@ TEST(testCase, update_test) {
   }
 }
 
-#endif
+TEST(clientCase, subscription_test) {
+  TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
+  ASSERT_NE(pConn, nullptr);
+
+  //  TAOS_RES* pRes = taos_query(pConn, "create topic topic_t1 as select * from t1");
+  //  if (taos_errno(pRes) != TSDB_CODE_SUCCESS) {
+  //    printf("failed to create topic, code:%s", taos_errstr(pRes));
+  //    taos_free_result(pRes);
+  //    return;
+  //  }
+
+  tmq_conf_t* conf = tmq_conf_new();
+  tmq_conf_set(conf, "enable.auto.commit", "true");
+  tmq_conf_set(conf, "auto.commit.interval.ms", "1000");
+  tmq_conf_set(conf, "group.id", "consumer_group");
+  tmq_conf_set(conf, "td.connect.user", "root");
+  tmq_conf_set(conf, "td.connect.pass", "taosdata");
+  tmq_conf_set(conf, "auto.offset.reset", "earliest");
+  tmq_conf_set(conf, "experimental.snapshot.enable", "true");
+  tmq_conf_set(conf, "msg.with.table.name", "true");
+  tmq_conf_set_auto_commit_cb(conf, tmq_commit_cb_print, NULL);
+
+  tmq_t* tmq = tmq_consumer_new(conf, NULL, 0);
+  tmq_conf_destroy(conf);
+
+  // 创建订阅 topics 列表
+  tmq_list_t* topicList = tmq_list_new();
+  tmq_list_append(topicList, "topic_t1");
+
+  // 启动订阅
+  tmq_subscribe(tmq, topicList);
+  tmq_list_destroy(topicList);
+
+  TAOS_FIELD* fields = NULL;
+  int32_t     numOfFields = 0;
+  int32_t     precision = 0;
+  int32_t     totalRows = 0;
+  int32_t     msgCnt = 0;
+  int32_t     timeout = 5000;
+
+  while (1) {
+    TAOS_RES* pRes = tmq_consumer_poll(tmq, timeout);
+    if (pRes) {
+      char    buf[1024];
+      int32_t rows = 0;
+
+      const char* topicName = tmq_get_topic_name(pRes);
+      const char* dbName = tmq_get_db_name(pRes);
+      int32_t     vgroupId = tmq_get_vgroup_id(pRes);
+
+      printf("topic: %s\n", topicName);
+      printf("db: %s\n", dbName);
+      printf("vgroup id: %d\n", vgroupId);
+
+      while (1) {
+        TAOS_ROW row = taos_fetch_row(pRes);
+        if (row == NULL) break;
+
+        fields = taos_fetch_fields(pRes);
+        numOfFields = taos_field_count(pRes);
+        precision = taos_result_precision(pRes);
+        rows++;
+        taos_print_row(buf, row, fields, numOfFields);
+        printf("precision: %d, row content: %s\n", precision, buf);
+      }
+    }
+//      return rows;
+  }
+
+  fprintf(stderr, "%d msg consumed, include %d rows\n", msgCnt, totalRows);
+}
 
 #pragma GCC diagnostic pop

source/common/src/tdatablock.c

@@ -1546,7 +1546,10 @@ size_t blockDataGetCapacityInRow(const SSDataBlock* pBlock, size_t pageSize) {
 }
 
 void colDataDestroy(SColumnInfoData* pColData) {
-  if (!pColData) return;
+  if (!pColData) {
+    return;
+  }
+
   if (IS_VAR_DATA_TYPE(pColData->info.type)) {
     taosMemoryFreeClear(pColData->varmeta.offset);
   } else {
@@ -2353,8 +2356,7 @@ const char* blockDecode(SSDataBlock* pBlock, const char* pData) {
   pStart += sizeof(uint64_t);
 
   if (pBlock->pDataBlock == NULL) {
-    pBlock->pDataBlock = taosArrayInit(numOfCols, sizeof(SColumnInfoData));
+    pBlock->pDataBlock = taosArrayInit_s(numOfCols, sizeof(SColumnInfoData), numOfCols);
-    taosArraySetSize(pBlock->pDataBlock, numOfCols);
   }
 
   for (int32_t i = 0; i < numOfCols; ++i) {

source/common/src/tglobal.c

@@ -41,6 +41,7 @@ bool    tsPrintAuth = false;
 
 // queue & threads
 int32_t tsNumOfRpcThreads = 1;
+int32_t tsNumOfRpcSessions = 2000;
 int32_t tsNumOfCommitThreads = 2;
 int32_t tsNumOfTaskQueueThreads = 4;
 int32_t tsNumOfMnodeQueryThreads = 4;
@@ -54,7 +55,6 @@ int32_t tsNumOfQnodeQueryThreads = 4;
 int32_t tsNumOfQnodeFetchThreads = 1;
 int32_t tsNumOfSnodeStreamThreads = 4;
 int32_t tsNumOfSnodeWriteThreads = 1;
-
 // sync raft
 int32_t tsElectInterval = 25 * 1000;
 int32_t tsHeartbeatInterval = 1000;
@@ -76,11 +76,11 @@ bool     tsEnableTelem = true;
 int32_t  tsTelemInterval = 43200;
 char     tsTelemServer[TSDB_FQDN_LEN] = "telemetry.taosdata.com";
 uint16_t tsTelemPort = 80;
-char*    tsTelemUri = "/report";
+char    *tsTelemUri = "/report";
 
-bool     tsEnableCrashReport = true;
+bool  tsEnableCrashReport = true;
-char*    tsClientCrashReportUri = "/ccrashreport";
+char *tsClientCrashReportUri = "/ccrashreport";
-char*    tsSvrCrashReportUri = "/dcrashreport";
+char *tsSvrCrashReportUri = "/dcrashreport";
 
 // schemaless
 char tsSmlTagName[TSDB_COL_NAME_LEN] = "_tag_null";
@@ -140,6 +140,7 @@ int32_t tsMaxMemUsedByInsert = 1024;
 
 float   tsSelectivityRatio = 1.0;
 int32_t tsTagFilterResCacheSize = 1024 * 10;
+char    tsTagFilterCache = 0;
 
 // the maximum allowed query buffer size during query processing for each data node.
 // -1 no limit (default)
@@ -188,6 +189,7 @@ int32_t tsGrantHBInterval = 60;
 int32_t tsUptimeInterval = 300;    // seconds
 char    tsUdfdResFuncs[512] = "";  // udfd resident funcs that teardown when udfd exits
 char    tsUdfdLdLibPath[512] = "";
+bool    tsDisableStream = false;
 
 #ifndef _STORAGE
 int32_t taosSetTfsCfg(SConfig *pCfg) {
@@ -210,9 +212,7 @@ int32_t taosSetTfsCfg(SConfig *pCfg) {
 int32_t taosSetTfsCfg(SConfig *pCfg);
 #endif
 
-struct SConfig *taosGetCfg() {
+struct SConfig *taosGetCfg() { return tsCfg; }
-  return tsCfg;
-}
 
 static int32_t taosLoadCfg(SConfig *pCfg, const char **envCmd, const char *inputCfgDir, const char *envFile,
                            char *apolloUrl) {
@@ -351,6 +351,7 @@ static int32_t taosAddSystemCfg(SConfig *pCfg) {
   if (cfgAddBool(pCfg, "AVX2", tsAVX2Enable, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "FMA", tsFMAEnable, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "SIMD-builtins", tsSIMDBuiltins, 0) != 0) return -1;
+  if (cfgAddBool(pCfg, "tagFilterCache", tsTagFilterCache, 0) != 0) return -1;
 
   if (cfgAddInt64(pCfg, "openMax", tsOpenMax, 0, INT64_MAX, 1) != 0) return -1;
   if (cfgAddInt64(pCfg, "streamMax", tsStreamMax, 0, INT64_MAX, 1) != 0) return -1;
@@ -388,9 +389,12 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddInt32(pCfg, "queryRspPolicy", tsQueryRspPolicy, 0, 1, 0) != 0) return -1;
 
   tsNumOfRpcThreads = tsNumOfCores / 2;
-  tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 1, TSDB_MAX_RPC_THREADS);
+  tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 2, TSDB_MAX_RPC_THREADS);
   if (cfgAddInt32(pCfg, "numOfRpcThreads", tsNumOfRpcThreads, 1, 1024, 0) != 0) return -1;
 
+  tsNumOfRpcSessions = TRANGE(tsNumOfRpcSessions, 100, 10000);
+  if (cfgAddInt32(pCfg, "numOfRpcSessions", tsNumOfRpcSessions, 1, 100000, 0) != 0) return -1;
+
   tsNumOfCommitThreads = tsNumOfCores / 2;
   tsNumOfCommitThreads = TRANGE(tsNumOfCommitThreads, 2, 4);
   if (cfgAddInt32(pCfg, "numOfCommitThreads", tsNumOfCommitThreads, 1, 1024, 0) != 0) return -1;
@@ -467,6 +471,8 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddString(pCfg, "udfdResFuncs", tsUdfdResFuncs, 0) != 0) return -1;
   if (cfgAddString(pCfg, "udfdLdLibPath", tsUdfdLdLibPath, 0) != 0) return -1;
 
+  if (cfgAddBool(pCfg, "disableStream", tsDisableStream, 0) != 0) return -1;
+
   GRANT_CFG_ADD;
   return 0;
 }
@@ -496,11 +502,19 @@ static int32_t taosUpdateServerCfg(SConfig *pCfg) {
   pItem = cfgGetItem(tsCfg, "numOfRpcThreads");
   if (pItem != NULL && pItem->stype == CFG_STYPE_DEFAULT) {
     tsNumOfRpcThreads = numOfCores / 2;
-    tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 1, 4);
+    tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 2, TSDB_MAX_RPC_THREADS);
     pItem->i32 = tsNumOfRpcThreads;
     pItem->stype = stype;
   }
 
+  pItem = cfgGetItem(tsCfg, "numOfRpcSessions");
+  if (pItem != NULL && pItem->stype == CFG_STYPE_DEFAULT) {
+    tsNumOfRpcSessions = 2000;
+    tsNumOfRpcSessions = TRANGE(tsNumOfRpcSessions, 100, 10000);
+    pItem->i32 = tsNumOfRpcSessions;
+    pItem->stype = stype;
+  }
+
   pItem = cfgGetItem(tsCfg, "numOfCommitThreads");
   if (pItem != NULL && pItem->stype == CFG_STYPE_DEFAULT) {
     tsNumOfCommitThreads = numOfCores / 2;
@@ -718,6 +732,7 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsPrintAuth = cfgGetItem(pCfg, "printAuth")->bval;
 
   tsNumOfRpcThreads = cfgGetItem(pCfg, "numOfRpcThreads")->i32;
+  tsNumOfRpcSessions = cfgGetItem(pCfg, "numOfRpcSessions")->i32;
   tsNumOfCommitThreads = cfgGetItem(pCfg, "numOfCommitThreads")->i32;
   tsNumOfMnodeReadThreads = cfgGetItem(pCfg, "numOfMnodeReadThreads")->i32;
   tsNumOfVnodeQueryThreads = cfgGetItem(pCfg, "numOfVnodeQueryThreads")->i32;
@@ -731,6 +746,7 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsRpcQueueMemoryAllowed = cfgGetItem(pCfg, "rpcQueueMemoryAllowed")->i64;
 
   tsSIMDBuiltins = (bool)cfgGetItem(pCfg, "SIMD-builtins")->bval;
+  tsTagFilterCache = (bool)cfgGetItem(pCfg, "tagFilterCache")->bval;
 
   tsEnableMonitor = cfgGetItem(pCfg, "monitor")->bval;
   tsMonitorInterval = cfgGetItem(pCfg, "monitorInterval")->i32;
@@ -767,6 +783,9 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   if (tsQueryBufferSize >= 0) {
     tsQueryBufferSizeBytes = tsQueryBufferSize * 1048576UL;
   }
+
+  tsDisableStream = cfgGetItem(pCfg, "disableStream")->bval;
+
   GRANT_CFG_GET;
   return 0;
 }
@@ -973,6 +992,8 @@ int32_t taosSetCfg(SConfig *pCfg, char *name) {
         tsNumOfTaskQueueThreads = cfgGetItem(pCfg, "numOfTaskQueueThreads")->i32;
       } else if (strcasecmp("numOfRpcThreads", name) == 0) {
         tsNumOfRpcThreads = cfgGetItem(pCfg, "numOfRpcThreads")->i32;
+      } else if (strcasecmp("numOfRpcSessions", name) == 0) {
+        tsNumOfRpcSessions = cfgGetItem(pCfg, "numOfRpcSessions")->i32;
       } else if (strcasecmp("numOfCommitThreads", name) == 0) {
         tsNumOfCommitThreads = cfgGetItem(pCfg, "numOfCommitThreads")->i32;
       } else if (strcasecmp("numOfMnodeReadThreads", name) == 0) {

source/common/src/tmisce.c

@@ -60,6 +60,19 @@ bool isEpsetEqual(const SEpSet* s1, const SEpSet* s2) {
   return true;
 }
 
+void epsetAssign(SEpSet* pDst, const SEpSet* pSrc) {
+  if (pSrc == NULL || pDst == NULL) {
+      return;
+  }
+
+  pDst->inUse = pSrc->inUse;
+  pDst->numOfEps = pSrc->numOfEps;
+  for (int32_t i = 0; i < pSrc->numOfEps; ++i) {
+    pDst->eps[i].port = pSrc->eps[i].port;
+    tstrncpy(pDst->eps[i].fqdn, pSrc->eps[i].fqdn, tListLen(pSrc->eps[i].fqdn));
+  }
+}
+
 void updateEpSet_s(SCorEpSet* pEpSet, SEpSet* pNewEpSet) {
   taosCorBeginWrite(&pEpSet->version);
   pEpSet->epSet = *pNewEpSet;

source/common/src/tmsg.c

@@ -5425,6 +5425,8 @@ int32_t tSerializeSCMCreateStreamReq(void *buf, int32_t bufLen, const SCMCreateS
     if (tEncodeI32(&encoder, pField->bytes) < 0) return -1;
     if (tEncodeCStr(&encoder, pField->name) < 0) return -1;
   }
+  if (tEncodeI64(&encoder, pReq->deleteMark) < 0) return -1;
+  if (tEncodeI8(&encoder, pReq->igUpdate) < 0) return -1;
 
   tEndEncode(&encoder);
 
@@ -5486,6 +5488,9 @@ int32_t tDeserializeSCMCreateStreamReq(void *buf, int32_t bufLen, SCMCreateStrea
     }
   }
 
+  if (tDecodeI64(&decoder, &pReq->deleteMark) < 0) return -1;
+  if (tDecodeI8(&decoder, &pReq->igUpdate) < 0) return -1;
+
   tEndDecode(&decoder);
 
   tDecoderClear(&decoder);

source/dnode/mgmt/mgmt_mnode/src/mmWorker.c

@@ -49,7 +49,7 @@ static void mmProcessRpcMsg(SQueueInfo *pInfo, SRpcMsg *pMsg) {
   pMsg->info.node = pMgmt->pMnode;
 
   const STraceId *trace = &pMsg->info.traceId;
-  dGTrace("msg:%p, get from mnode queue", pMsg);
+  dGTrace("msg:%p, get from mnode queue, type:%s", pMsg, TMSG_INFO(pMsg->msgType));
 
   int32_t code = mndProcessRpcMsg(pMsg);
 

source/dnode/mgmt/node_mgmt/src/dmTransport.c

@@ -280,10 +280,19 @@ int32_t dmInitClient(SDnode *pDnode) {
   rpcInit.retryMaxInterval = tsRedirectMaxPeriod;
   rpcInit.retryMaxTimouet = tsMaxRetryWaitTime;
 
-  rpcInit.failFastInterval = 1000;  // interval threshold(ms)
+  rpcInit.failFastInterval = 5000;  // interval threshold(ms)
   rpcInit.failFastThreshold = 3;    // failed threshold
   rpcInit.ffp = dmFailFastFp;
 
+  int32_t connLimitNum = tsNumOfRpcSessions / (tsNumOfRpcThreads * 3);
+  connLimitNum = TMAX(connLimitNum, 10);
+  connLimitNum = TMIN(connLimitNum, 500);
+
+  rpcInit.connLimitNum = connLimitNum;
+  rpcInit.connLimitLock = 1;
+  rpcInit.supportBatch = 1;
+  rpcInit.batchSize = 8 * 1024;
+
   pTrans->clientRpc = rpcOpen(&rpcInit);
   if (pTrans->clientRpc == NULL) {
     dError("failed to init dnode rpc client");

source/dnode/mnode/impl/inc/mndDef.h

@@ -648,6 +648,7 @@ typedef struct {
   int64_t checkpointFreq;  // ms
   int64_t currentTick;     // do not serialize
   int64_t deleteMark;
+  int8_t  igCheckUpdate;
 } SStreamObj;
 
 int32_t tEncodeSStreamObj(SEncoder* pEncoder, const SStreamObj* pObj);

source/dnode/mnode/impl/inc/mndInt.h

@@ -58,7 +58,7 @@ typedef int32_t (*MndInitFp)(SMnode *pMnode);
 typedef void (*MndCleanupFp)(SMnode *pMnode);
 typedef int32_t (*ShowRetrieveFp)(SRpcMsg *pMsg, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows);
 typedef void (*ShowFreeIterFp)(SMnode *pMnode, void *pIter);
-typedef struct SQWorker SQHandle;
+typedef struct SQueueWorker SQHandle;
 
 typedef struct {
   const char  *name;

source/dnode/mnode/impl/inc/mndTrans.h

@@ -81,7 +81,7 @@ int32_t mndTransPrepare(SMnode *pMnode, STrans *pTrans);
 int32_t mndTransProcessRsp(SRpcMsg *pRsp);
 void    mndTransPullup(SMnode *pMnode);
 int32_t mndKillTrans(SMnode *pMnode, STrans *pTrans);
-void    mndTransExecute(SMnode *pMnode, STrans *pTrans);
+void    mndTransExecute(SMnode *pMnode, STrans *pTrans, bool isLeader);
 int32_t mndSetRpcInfoForDbTrans(SMnode *pMnode, SRpcMsg *pMsg, EOperType oper, const char *dbname);
 
 #ifdef __cplusplus

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

@@ -238,7 +238,9 @@ static int32_t mndProcessMqTimerMsg(SRpcMsg *pMsg) {
   // iterate all consumers, find all modification
   while (1) {
     pIter = sdbFetch(pSdb, SDB_CONSUMER, pIter, (void **)&pConsumer);
-    if (pIter == NULL) break;
+    if (pIter == NULL) {
+      break;
+    }
 
     int32_t hbStatus = atomic_add_fetch_32(&pConsumer->hbStatus, 1);
     int32_t status = atomic_load_32(&pConsumer->status);
@@ -335,7 +337,7 @@ static int32_t mndProcessMqHbReq(SRpcMsg *pMsg) {
   int64_t         consumerId = req.consumerId;
   SMqConsumerObj *pConsumer = mndAcquireConsumer(pMnode, consumerId);
   if (pConsumer == NULL) {
-    mError("consumer %" PRId64 " not exist", consumerId);
+    mError("consumer:0x%"PRIx64 " not exist", consumerId);
     terrno = TSDB_CODE_MND_CONSUMER_NOT_EXIST;
     return -1;
   }
@@ -345,7 +347,7 @@ static int32_t mndProcessMqHbReq(SRpcMsg *pMsg) {
   int32_t status = atomic_load_32(&pConsumer->status);
 
   if (status == MQ_CONSUMER_STATUS__LOST_REBD) {
-    mInfo("try to recover consumer %" PRId64 "", consumerId);
+    mInfo("try to recover consumer:0x%"PRIx64 "", consumerId);
     SMqConsumerRecoverMsg *pRecoverMsg = rpcMallocCont(sizeof(SMqConsumerRecoverMsg));
 
     pRecoverMsg->consumerId = consumerId;
@@ -390,7 +392,7 @@ static int32_t mndProcessAskEpReq(SRpcMsg *pMsg) {
 
 #if 1
   if (status == MQ_CONSUMER_STATUS__LOST_REBD) {
-    mInfo("try to recover consumer %" PRId64 "", consumerId);
+    mInfo("try to recover consumer:0x%"PRIx64 "", consumerId);
     SMqConsumerRecoverMsg *pRecoverMsg = rpcMallocCont(sizeof(SMqConsumerRecoverMsg));
 
     pRecoverMsg->consumerId = consumerId;
@@ -404,14 +406,14 @@ static int32_t mndProcessAskEpReq(SRpcMsg *pMsg) {
 #endif
 
   if (status != MQ_CONSUMER_STATUS__READY) {
-    mInfo("consumer %" PRId64 " not ready, status: %s", consumerId, mndConsumerStatusName(status));
+    mInfo("consumer:0x%"PRIx64 " not ready, status: %s", consumerId, mndConsumerStatusName(status));
     terrno = TSDB_CODE_MND_CONSUMER_NOT_READY;
     return -1;
   }
 
   int32_t serverEpoch = atomic_load_32(&pConsumer->epoch);
 
-  // 2. check epoch, only send ep info when epoches do not match
+  // 2. check epoch, only send ep info when epochs do not match
   if (epoch != serverEpoch) {
     taosRLockLatch(&pConsumer->lock);
     mInfo("process ask ep, consumer:%" PRId64 "(epoch %d), server epoch %d", consumerId, epoch, serverEpoch);
@@ -526,12 +528,14 @@ int32_t mndSetConsumerCommitLogs(SMnode *pMnode, STrans *pTrans, SMqConsumerObj
   return 0;
 }
 
-static int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
+int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
-  SMnode         *pMnode = pMsg->info.node;
+  SMnode *pMnode = pMsg->info.node;
-  char           *msgStr = pMsg->pCont;
+  char   *msgStr = pMsg->pCont;
+
   SCMSubscribeReq subscribe = {0};
   tDeserializeSCMSubscribeReq(msgStr, &subscribe);
-  int64_t         consumerId = subscribe.consumerId;
+
+  uint64_t        consumerId = subscribe.consumerId;
   char           *cgroup = subscribe.cgroup;
   SMqConsumerObj *pConsumerOld = NULL;
   SMqConsumerObj *pConsumerNew = NULL;
@@ -542,21 +546,23 @@ static int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
   taosArrayRemoveDuplicateP(newSub, taosArrayCompareString, taosMemoryFree);
 
   int32_t newTopicNum = taosArrayGetSize(newSub);
+
   // check topic existance
   STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_NOTHING, pMsg, "subscribe");
-  if (pTrans == NULL) goto SUBSCRIBE_OVER;
+  if (pTrans == NULL) {
+    goto _over;
+  }
 
   for (int32_t i = 0; i < newTopicNum; i++) {
     char        *topic = taosArrayGetP(newSub, i);
     SMqTopicObj *pTopic = mndAcquireTopic(pMnode, topic);
-    if (pTopic == NULL) {
+    if (pTopic == NULL) {  // terrno has been set by callee function
-      terrno = TSDB_CODE_MND_TOPIC_NOT_EXIST;
+      goto _over;
-      goto SUBSCRIBE_OVER;
     }
 
     if (mndCheckTopicPrivilege(pMnode, pMsg->info.conn.user, MND_OPER_SUBSCRIBE, pTopic) != 0) {
       mndReleaseTopic(pMnode, pTopic);
-      goto SUBSCRIBE_OVER;
+      goto _over;
     }
 
     mndReleaseTopic(pMnode, pTopic);
@@ -578,8 +584,8 @@ static int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
       taosArrayPush(pConsumerNew->assignedTopics, &newTopicCopy);
     }
 
-    if (mndSetConsumerCommitLogs(pMnode, pTrans, pConsumerNew) != 0) goto SUBSCRIBE_OVER;
+    if (mndSetConsumerCommitLogs(pMnode, pTrans, pConsumerNew) != 0) goto _over;
-    if (mndTransPrepare(pMnode, pTrans) != 0) goto SUBSCRIBE_OVER;
+    if (mndTransPrepare(pMnode, pTrans) != 0) goto _over;
 
   } else {
     /*taosRLockLatch(&pConsumerOld->lock);*/
@@ -591,13 +597,13 @@ static int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
 
     if (status != MQ_CONSUMER_STATUS__READY) {
       terrno = TSDB_CODE_MND_CONSUMER_NOT_READY;
-      goto SUBSCRIBE_OVER;
+      goto _over;
     }
 
     pConsumerNew = tNewSMqConsumerObj(consumerId, cgroup);
     if (pConsumerNew == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
-      goto SUBSCRIBE_OVER;
+      goto _over;
     }
     pConsumerNew->updateType = CONSUMER_UPDATE__MODIFY;
 
@@ -650,16 +656,16 @@ static int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
       /*if (taosArrayGetSize(pConsumerNew->assignedTopics) == 0) {*/
       /*pConsumerNew->updateType = */
       /*}*/
-      goto SUBSCRIBE_OVER;
+      goto _over;
     }
 
-    if (mndSetConsumerCommitLogs(pMnode, pTrans, pConsumerNew) != 0) goto SUBSCRIBE_OVER;
+    if (mndSetConsumerCommitLogs(pMnode, pTrans, pConsumerNew) != 0) goto _over;
-    if (mndTransPrepare(pMnode, pTrans) != 0) goto SUBSCRIBE_OVER;
+    if (mndTransPrepare(pMnode, pTrans) != 0) goto _over;
   }
 
   code = TSDB_CODE_ACTION_IN_PROGRESS;
 
-SUBSCRIBE_OVER:
+_over:
   mndTransDrop(pTrans);
 
   if (pConsumerOld) {
@@ -971,16 +977,19 @@ static int32_t mndRetrieveConsumer(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *
 
   while (numOfRows < rowsCapacity) {
     pShow->pIter = sdbFetch(pSdb, SDB_CONSUMER, pShow->pIter, (void **)&pConsumer);
-    if (pShow->pIter == NULL) break;
+    if (pShow->pIter == NULL) {
+      break;
+    }
+
     if (taosArrayGetSize(pConsumer->assignedTopics) == 0) {
-      mDebug("showing consumer %" PRId64 " no assigned topic, skip", pConsumer->consumerId);
+      mDebug("showing consumer:0x%"PRIx64 " no assigned topic, skip", pConsumer->consumerId);
       sdbRelease(pSdb, pConsumer);
       continue;
     }
 
     taosRLockLatch(&pConsumer->lock);
 
-    mDebug("showing consumer %" PRId64, pConsumer->consumerId);
+    mDebug("showing consumer:0x%"PRIx64, pConsumer->consumerId);
 
     int32_t topicSz = taosArrayGetSize(pConsumer->assignedTopics);
     bool    hasTopic = true;

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

@@ -78,6 +78,7 @@ int32_t tEncodeSStreamObj(SEncoder *pEncoder, const SStreamObj *pObj) {
 
   // 3.0.20
   if (tEncodeI64(pEncoder, pObj->checkpointFreq) < 0) return -1;
+  if (tEncodeI8(pEncoder, pObj->igCheckUpdate) < 0) return -1;
 
   tEndEncode(pEncoder);
   return pEncoder->pos;
@@ -145,6 +146,9 @@ int32_t tDecodeSStreamObj(SDecoder *pDecoder, SStreamObj *pObj, int32_t sver) {
   // 3.0.20
   if (sver >= 2) {
     if (tDecodeI64(pDecoder, &pObj->checkpointFreq) < 0) return -1;
+    if (!tDecodeIsEnd(pDecoder)) {
+      if (tDecodeI8(pDecoder, &pObj->igCheckUpdate) < 0) return -1;
+    }
   }
   tEndDecode(pDecoder);
   return 0;
@@ -409,19 +413,21 @@ void *tDecodeSMqConsumerEp(const void *buf, SMqConsumerEp *pConsumerEp) {
   return (void *)buf;
 }
 
-SMqSubscribeObj *tNewSubscribeObj(const char key[TSDB_SUBSCRIBE_KEY_LEN]) {
+SMqSubscribeObj *tNewSubscribeObj(const char* key) {
-  SMqSubscribeObj *pSubNew = taosMemoryCalloc(1, sizeof(SMqSubscribeObj));
+  SMqSubscribeObj *pSubObj = taosMemoryCalloc(1, sizeof(SMqSubscribeObj));
-  if (pSubNew == NULL) return NULL;
+  if (pSubObj == NULL) {
-  memcpy(pSubNew->key, key, TSDB_SUBSCRIBE_KEY_LEN);
+    return NULL;
-  taosInitRWLatch(&pSubNew->lock);
+  }
-  pSubNew->vgNum = 0;
+
-  pSubNew->consumerHash = taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
+  memcpy(pSubObj->key, key, TSDB_SUBSCRIBE_KEY_LEN);
+  taosInitRWLatch(&pSubObj->lock);
+  pSubObj->vgNum = 0;
+  pSubObj->consumerHash = taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
+
   // TODO set hash free fp
-  /*taosHashSetFreeFp(pSubNew->consumerHash, tDeleteSMqConsumerEp);*/
+  /*taosHashSetFreeFp(pSubObj->consumerHash, tDeleteSMqConsumerEp);*/
-
+  pSubObj->unassignedVgs = taosArrayInit(0, POINTER_BYTES);
-  pSubNew->unassignedVgs = taosArrayInit(0, sizeof(void *));
+  return pSubObj;
-
-  return pSubNew;
 }
 
 SMqSubscribeObj *tCloneSubscribeObj(const SMqSubscribeObj *pSub) {
@@ -489,7 +495,7 @@ int32_t tEncodeSubscribeObj(void **buf, const SMqSubscribeObj *pSub) {
     tlen += tEncodeSMqConsumerEp(buf, pConsumerEp);
     cnt++;
   }
-  if(cnt != sz) return -1;
+  if (cnt != sz) return -1;
   tlen += taosEncodeArray(buf, pSub->unassignedVgs, (FEncode)tEncodeSMqVgEp);
   tlen += taosEncodeString(buf, pSub->dbName);
   return tlen;

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

@@ -880,6 +880,12 @@ static int32_t mndProcessConfigDnodeReq(SRpcMsg *pReq) {
   if (strcasecmp(cfgReq.config, "resetlog") == 0) {
     strcpy(dcfgReq.config, "resetlog");
   } else if (strncasecmp(cfgReq.config, "monitor", 7) == 0) {
+    if (' ' != cfgReq.config[7] && 0 != cfgReq.config[7]) {
+      mError("dnode:%d, failed to config monitor since invalid conf:%s", cfgReq.dnodeId, cfgReq.config);
+      terrno = TSDB_CODE_INVALID_CFG;
+      return -1;
+    }
+
     const char *value = cfgReq.value;
     int32_t     flag = atoi(value);
     if (flag <= 0) {
@@ -900,12 +906,18 @@ static int32_t mndProcessConfigDnodeReq(SRpcMsg *pReq) {
       int32_t     optLen = strlen(optName);
       if (strncasecmp(cfgReq.config, optName, optLen) != 0) continue;
 
+      if (' ' != cfgReq.config[optLen] && 0 != cfgReq.config[optLen]) {
+        mError("dnode:%d, failed to config since invalid conf:%s", cfgReq.dnodeId, cfgReq.config);
+        terrno = TSDB_CODE_INVALID_CFG;
+        return -1;
+      }
+
       const char *value = cfgReq.value;
       int32_t     flag = atoi(value);
       if (flag <= 0) {
         flag = atoi(cfgReq.config + optLen + 1);
       }
-      if (flag <= 0 || flag > 255) {
+      if (flag < 0 || flag > 255) {
         mError("dnode:%d, failed to config %s since value:%d", cfgReq.dnodeId, optName, flag);
         terrno = TSDB_CODE_INVALID_CFG;
         return -1;

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

@@ -523,7 +523,7 @@ int32_t mndSchedInitSubEp(SMnode* pMnode, const SMqTopicObj* pTopic, SMqSubscrib
   SSdb*       pSdb = pMnode->pSdb;
   SVgObj*     pVgroup = NULL;
   SQueryPlan* pPlan = NULL;
-  SSubplan*   plan = NULL;
+  SSubplan*   pSubplan = NULL;
 
   if (pTopic->subType == TOPIC_SUB_TYPE__COLUMN) {
     pPlan = qStringToQueryPlan(pTopic->physicalPlan);
@@ -539,24 +539,27 @@ int32_t mndSchedInitSubEp(SMnode* pMnode, const SMqTopicObj* pTopic, SMqSubscrib
       return -1;
     }
 
-    SNodeListNode* inner = (SNodeListNode*)nodesListGetNode(pPlan->pSubplans, 0);
+    SNodeListNode* pNodeListNode = (SNodeListNode*)nodesListGetNode(pPlan->pSubplans, 0);
 
-    int32_t opNum = LIST_LENGTH(inner->pNodeList);
+    int32_t opNum = LIST_LENGTH(pNodeListNode->pNodeList);
     if (opNum != 1) {
       qDestroyQueryPlan(pPlan);
       terrno = TSDB_CODE_MND_INVALID_TOPIC_QUERY;
       return -1;
     }
-    plan = (SSubplan*)nodesListGetNode(inner->pNodeList, 0);
+
+    pSubplan = (SSubplan*)nodesListGetNode(pNodeListNode->pNodeList, 0);
   }
 
   ASSERT(pSub->unassignedVgs);
-  ASSERT(taosHashGetSize(pSub->consumerHash) == 0);
 
   void* pIter = NULL;
   while (1) {
     pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void**)&pVgroup);
-    if (pIter == NULL) break;
+    if (pIter == NULL) {
+      break;
+    }
+
     if (!mndVgroupInDb(pVgroup, pTopic->dbUid)) {
       sdbRelease(pSdb, pVgroup);
       continue;
@@ -569,15 +572,15 @@ int32_t mndSchedInitSubEp(SMnode* pMnode, const SMqTopicObj* pTopic, SMqSubscrib
     pVgEp->vgId = pVgroup->vgId;
     taosArrayPush(pSub->unassignedVgs, &pVgEp);
 
-    mDebug("init subscription %s, assign vg: %d", pSub->key, pVgEp->vgId);
+    mDebug("init subscription %s for topic:%s assign vgId:%d", pSub->key, pTopic->name, pVgEp->vgId);
 
     if (pTopic->subType == TOPIC_SUB_TYPE__COLUMN) {
       int32_t msgLen;
 
-      plan->execNode.epSet = pVgEp->epSet;
+      pSubplan->execNode.epSet = pVgEp->epSet;
-      plan->execNode.nodeId = pVgEp->vgId;
+      pSubplan->execNode.nodeId = pVgEp->vgId;
 
-      if (qSubPlanToString(plan, &pVgEp->qmsg, &msgLen) < 0) {
+      if (qSubPlanToString(pSubplan, &pVgEp->qmsg, &msgLen) < 0) {
         sdbRelease(pSdb, pVgroup);
         qDestroyQueryPlan(pPlan);
         terrno = TSDB_CODE_QRY_INVALID_INPUT;
@@ -590,11 +593,7 @@ int32_t mndSchedInitSubEp(SMnode* pMnode, const SMqTopicObj* pTopic, SMqSubscrib
     sdbRelease(pSdb, pVgroup);
   }
 
-  ASSERT(pSub->unassignedVgs->size > 0);
+  ASSERT(taosArrayGetSize(pSub->unassignedVgs) > 0);
-
-  ASSERT(taosHashGetSize(pSub->consumerHash) == 0);
-
   qDestroyQueryPlan(pPlan);
-
   return 0;
 }

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

@@ -31,6 +31,8 @@
 #define MND_STREAM_VER_NUMBER   2
 #define MND_STREAM_RESERVE_SIZE 64
 
+#define MND_STREAM_MAX_NUM 10
+
 static int32_t mndStreamActionInsert(SSdb *pSdb, SStreamObj *pStream);
 static int32_t mndStreamActionDelete(SSdb *pSdb, SStreamObj *pStream);
 static int32_t mndStreamActionUpdate(SSdb *pSdb, SStreamObj *pStream, SStreamObj *pNewStream);
@@ -295,6 +297,8 @@ static int32_t mndBuildStreamObjFromCreateReq(SMnode *pMnode, SStreamObj *pObj,
   pObj->triggerParam = pCreate->maxDelay;
   pObj->watermark = pCreate->watermark;
   pObj->fillHistory = pCreate->fillHistory;
+  pObj->deleteMark = pCreate->deleteMark;
+  pObj->igCheckUpdate = pCreate->igUpdate;
 
   memcpy(pObj->sourceDb, pCreate->sourceDB, TSDB_DB_FNAME_LEN);
   SDbObj *pSourceDb = mndAcquireDb(pMnode, pCreate->sourceDB);
@@ -343,6 +347,8 @@ static int32_t mndBuildStreamObjFromCreateReq(SMnode *pMnode, SStreamObj *pObj,
       .triggerType = pObj->trigger == STREAM_TRIGGER_MAX_DELAY ? STREAM_TRIGGER_WINDOW_CLOSE : pObj->trigger,
       .watermark = pObj->watermark,
       .igExpired = pObj->igExpired,
+      .deleteMark = pObj->deleteMark,
+      .igCheckUpdate = pObj->igCheckUpdate,
   };
 
   // using ast and param to build physical plan
@@ -473,9 +479,8 @@ static int32_t mndCreateStbForStream(SMnode *pMnode, STrans *pTrans, const SStre
   tstrncpy(createReq.name, pStream->targetSTbName, TSDB_TABLE_FNAME_LEN);
   createReq.numOfColumns = pStream->outputSchema.nCols;
   createReq.numOfTags = 1;  // group id
-  createReq.pColumns = taosArrayInit(createReq.numOfColumns, sizeof(SField));
+  createReq.pColumns = taosArrayInit_s(createReq.numOfColumns, sizeof(SField), createReq.numOfColumns);
   // build fields
-  taosArraySetSize(createReq.pColumns, createReq.numOfColumns);
   for (int32_t i = 0; i < createReq.numOfColumns; i++) {
     SField *pField = taosArrayGet(createReq.pColumns, i);
     tstrncpy(pField->name, pStream->outputSchema.pSchema[i].name, TSDB_COL_NAME_LEN);
@@ -483,8 +488,8 @@ static int32_t mndCreateStbForStream(SMnode *pMnode, STrans *pTrans, const SStre
     pField->type = pStream->outputSchema.pSchema[i].type;
     pField->bytes = pStream->outputSchema.pSchema[i].bytes;
   }
-  createReq.pTags = taosArrayInit(createReq.numOfTags, sizeof(SField));
+  createReq.pTags = taosArrayInit_s(createReq.numOfTags, sizeof(SField), 1);
-  taosArraySetSize(createReq.pTags, 1);
+
   // build tags
   SField *pField = taosArrayGet(createReq.pTags, 0);
   strcpy(pField->name, "group_id");
@@ -625,6 +630,35 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
+  {
+    int32_t numOfStream = 0;
+
+    SStreamObj *pStream = NULL;
+    void       *pIter = NULL;
+
+    while (1) {
+      pIter = sdbFetch(pMnode->pSdb, SDB_STREAM, pIter, (void **)&pStream);
+      if (pIter == NULL) {
+        if (numOfStream > MND_STREAM_MAX_NUM) {
+          mError("too many streams, no more than 10 for each database");
+          terrno = TSDB_CODE_MND_TOO_MANY_STREAMS;
+          goto _OVER;
+        }
+        break;
+      }
+
+      if (pStream->sourceDbUid == streamObj.sourceDbUid) {
+        ++numOfStream;
+      }
+      sdbRelease(pMnode->pSdb, pStream);
+      if (numOfStream > MND_STREAM_MAX_NUM) {
+        mError("too many streams, no more than 10 for each database");
+        terrno = TSDB_CODE_MND_TOO_MANY_STREAMS;
+        goto _OVER;
+      }
+    }
+  }
+
   pDb = mndAcquireDb(pMnode, streamObj.sourceDb);
   if (pDb->cfg.replications != 1) {
     mError("stream source db must have only 1 replica, but %s has %d", pDb->name, pDb->cfg.replications);

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

@@ -39,12 +39,10 @@ static SSdbRow *mndSubActionDecode(SSdbRaw *pRaw);
 static int32_t  mndSubActionInsert(SSdb *pSdb, SMqSubscribeObj *);
 static int32_t  mndSubActionDelete(SSdb *pSdb, SMqSubscribeObj *);
 static int32_t  mndSubActionUpdate(SSdb *pSdb, SMqSubscribeObj *pOldSub, SMqSubscribeObj *pNewSub);
-
+static int32_t  mndProcessRebalanceReq(SRpcMsg *pMsg);
-static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg);
+static int32_t  mndProcessDropCgroupReq(SRpcMsg *pMsg);
-static int32_t mndProcessDropCgroupReq(SRpcMsg *pMsg);
+static int32_t  mndRetrieveSubscribe(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows);
-
+static void     mndCancelGetNextSubscribe(SMnode *pMnode, void *pIter);
-static int32_t mndRetrieveSubscribe(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows);
-static void    mndCancelGetNextSubscribe(SMnode *pMnode, void *pIter);
 
 static int32_t mndSetSubRedoLogs(SMnode *pMnode, STrans *pTrans, SMqSubscribeObj *pSub) {
   SSdbRaw *pRedoRaw = mndSubActionEncode(pSub);
@@ -85,12 +83,13 @@ int32_t mndInitSubscribe(SMnode *pMnode) {
   return sdbSetTable(pMnode->pSdb, table);
 }
 
-static SMqSubscribeObj *mndCreateSub(SMnode *pMnode, const SMqTopicObj *pTopic, const char *subKey) {
+static SMqSubscribeObj *mndCreateSubscription(SMnode *pMnode, const SMqTopicObj *pTopic, const char *subKey) {
   SMqSubscribeObj *pSub = tNewSubscribeObj(subKey);
   if (pSub == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return NULL;
   }
+
   pSub->dbUid = pTopic->dbUid;
   pSub->stbUid = pTopic->stbUid;
   pSub->subType = pTopic->subType;
@@ -205,7 +204,7 @@ static SMqRebInfo *mndGetOrCreateRebSub(SHashObj *pHash, const char *key) {
 static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqRebOutputObj *pOutput) {
   int32_t     totalVgNum = pOutput->pSub->vgNum;
   const char *sub = pOutput->pSub->key;
-  mInfo("sub:%s, mq rebalance vgNum:%d", sub, pOutput->pSub->vgNum);
+  mInfo("sub:%s mq re-balance %d vgroups", sub, pOutput->pSub->vgNum);
 
   // 1. build temporary hash(vgId -> SMqRebOutputVg) to store modified vg
   SHashObj *pHash = taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), false, HASH_NO_LOCK);
@@ -214,7 +213,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
   int32_t removedNum = taosArrayGetSize(pInput->pRebInfo->removedConsumers);
   int32_t actualRemoved = 0;
   for (int32_t i = 0; i < removedNum; i++) {
-    int64_t consumerId = *(int64_t *)taosArrayGet(pInput->pRebInfo->removedConsumers, i);
+    uint64_t consumerId = *(uint64_t *)taosArrayGet(pInput->pRebInfo->removedConsumers, i);
 
     SMqConsumerEp *pConsumerEp = taosHashGet(pOutput->pSub->consumerHash, &consumerId, sizeof(int64_t));
 
@@ -229,7 +228,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
             .pVgEp = pVgEp,
         };
         taosHashPut(pHash, &pVgEp->vgId, sizeof(int32_t), &outputVg, sizeof(SMqRebOutputVg));
-        mInfo("sub:%s, mq rebalance remove vgId:%d from consumer:%" PRId64, sub, pVgEp->vgId, consumerId);
+        mInfo("sub:%s mq re-balance remove vgId:%d from consumer:%" PRId64, sub, pVgEp->vgId, consumerId);
       }
       taosArrayDestroy(pConsumerEp->vgs);
       taosHashRemove(pOutput->pSub->consumerHash, &consumerId, sizeof(int64_t));
@@ -239,7 +238,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
   }
 
   if (removedNum != actualRemoved) {
-    mError("sub:%s, mq rebalance removedNum:%d not matched with actual:%d", sub, removedNum, actualRemoved);
+    mError("sub:%s mq re-balance removedNum:%d not matched with actual:%d", sub, removedNum, actualRemoved);
   }
 
   // if previously no consumer, there are vgs not assigned
@@ -253,7 +252,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
           .pVgEp = pVgEp,
       };
       taosHashPut(pHash, &pVgEp->vgId, sizeof(int32_t), &rebOutput, sizeof(SMqRebOutputVg));
-      mInfo("sub:%s, mq rebalance remove vgId:%d from unassigned", sub, pVgEp->vgId);
+      mInfo("sub:%s mq re-balance remove vgId:%d from unassigned", sub, pVgEp->vgId);
     }
   }
 
@@ -267,7 +266,8 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
     minVgCnt = totalVgNum / afterRebConsumerNum;
     imbConsumerNum = totalVgNum % afterRebConsumerNum;
   }
-  mInfo("sub:%s, mq rebalance %d consumer after rebalance, at least %d vg each, %d consumer has more vg", sub,
+
+  mInfo("sub:%s mq re-balance %d consumers: at least %d vg each, %d consumer has more vg", sub,
         afterRebConsumerNum, minVgCnt, imbConsumerNum);
 
   // 4. first scan: remove consumer more than wanted, put to remove hash
@@ -275,7 +275,10 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
   void   *pIter = NULL;
   while (1) {
     pIter = taosHashIterate(pOutput->pSub->consumerHash, pIter);
-    if (pIter == NULL) break;
+    if (pIter == NULL) {
+      break;
+    }
+
     SMqConsumerEp *pConsumerEp = (SMqConsumerEp *)pIter;
 
     int32_t consumerVgNum = taosArrayGetSize(pConsumerEp->vgs);
@@ -297,7 +300,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
                 .pVgEp = pVgEp,
             };
             taosHashPut(pHash, &pVgEp->vgId, sizeof(int32_t), &outputVg, sizeof(SMqRebOutputVg));
-            mInfo("sub:%s, mq rebalance remove vgId:%d from consumer:%" PRId64 ",(first scan)", sub, pVgEp->vgId,
+            mInfo("sub:%s mq rebalance remove vgId:%d from consumer:%" PRId64 ",(first scan)", sub, pVgEp->vgId,
                   pConsumerEp->consumerId);
           }
           imbCnt++;
@@ -312,7 +315,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
               .pVgEp = pVgEp,
           };
           taosHashPut(pHash, &pVgEp->vgId, sizeof(int32_t), &outputVg, sizeof(SMqRebOutputVg));
-          mInfo("sub:%s, mq rebalance remove vgId:%d from consumer:%" PRId64 ",(first scan)", sub, pVgEp->vgId,
+          mInfo("sub:%s mq rebalance remove vgId:%d from consumer:%" PRId64 ",(first scan)", sub, pVgEp->vgId,
                 pConsumerEp->consumerId);
         }
       }
@@ -330,7 +333,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
       newConsumerEp.vgs = taosArrayInit(0, sizeof(void *));
       taosHashPut(pOutput->pSub->consumerHash, &consumerId, sizeof(int64_t), &newConsumerEp, sizeof(SMqConsumerEp));
       taosArrayPush(pOutput->newConsumers, &consumerId);
-      mInfo("sub:%s, mq rebalance add new consumer:%" PRId64, sub, consumerId);
+      mInfo("sub:%s mq rebalance add new consumer:%" PRId64, sub, consumerId);
     }
   }
 
@@ -349,7 +352,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
       // iter hash and find one vg
       pRemovedIter = taosHashIterate(pHash, pRemovedIter);
       if (pRemovedIter == NULL) {
-        mError("sub:%s, removed iter is null", sub);
+        mError("sub:%s removed iter is null", sub);
         continue;
       }
 
@@ -402,33 +405,36 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
     SMqRebOutputVg *pRebOutput = NULL;
     while (1) {
       pIter = taosHashIterate(pHash, pIter);
-      if (pIter == NULL) break;
+      if (pIter == NULL) {
+        break;
+      }
+
       pRebOutput = (SMqRebOutputVg *)pIter;
 
       taosArrayPush(pOutput->pSub->unassignedVgs, &pRebOutput->pVgEp);
       taosArrayPush(pOutput->rebVgs, pRebOutput);
-      mInfo("sub:%s, mq rebalance unassign vgId:%d (second scan)", sub, pRebOutput->pVgEp->vgId);
+      mInfo("sub:%s mq re-balance unassign vgId:%d (second scan)", sub, pRebOutput->pVgEp->vgId);
     }
   }
 
   // 8. generate logs
-  mInfo("sub:%s, mq rebalance calculation completed, rebalanced vg", sub);
+  mInfo("sub:%s mq re-balance calculation completed, re-balanced vg", sub);
   for (int32_t i = 0; i < taosArrayGetSize(pOutput->rebVgs); i++) {
     SMqRebOutputVg *pOutputRebVg = taosArrayGet(pOutput->rebVgs, i);
-    mInfo("sub:%s, mq rebalance vgId:%d, moved from consumer:%" PRId64 ", to consumer:%" PRId64, sub,
+    mInfo("sub:%s mq re-balance vgId:%d, moved from consumer:0x%" PRIx64 ", to consumer:0x%" PRIx64, sub,
           pOutputRebVg->pVgEp->vgId, pOutputRebVg->oldConsumerId, pOutputRebVg->newConsumerId);
   }
   {
-    void *pIter = NULL;
+    pIter = NULL;
     while (1) {
       pIter = taosHashIterate(pOutput->pSub->consumerHash, pIter);
       if (pIter == NULL) break;
       SMqConsumerEp *pConsumerEp = (SMqConsumerEp *)pIter;
       int32_t        sz = taosArrayGetSize(pConsumerEp->vgs);
-      mInfo("sub:%s, mq rebalance final cfg: consumer %" PRId64 " has %d vg", sub, pConsumerEp->consumerId, sz);
+      mInfo("sub:%s mq re-balance final cfg: consumer:0x%" PRId64 " has %d vg", sub, pConsumerEp->consumerId, sz);
       for (int32_t i = 0; i < sz; i++) {
         SMqVgEp *pVgEp = taosArrayGetP(pConsumerEp->vgs, i);
-        mInfo("sub:%s, mq rebalance final cfg: vg %d to consumer %" PRId64 "", sub, pVgEp->vgId,
+        mInfo("sub:%s mq re-balance final cfg: vg %d to consumer:0x%" PRId64, sub, pVgEp->vgId,
               pConsumerEp->consumerId);
       }
     }
@@ -552,11 +558,14 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
   SMqDoRebalanceMsg *pReq = pMsg->pCont;
   void              *pIter = NULL;
 
-  mInfo("mq rebalance start");
+  mInfo("mq re-balance start");
 
   while (1) {
     pIter = taosHashIterate(pReq->rebSubHash, pIter);
-    if (pIter == NULL) break;
+    if (pIter == NULL) {
+      break;
+    }
+
     SMqRebInputObj rebInput = {0};
 
     SMqRebOutputObj rebOutput = {0};
@@ -577,12 +586,13 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
       mndSplitSubscribeKey(pRebInfo->key, topic, cgroup, true);
       SMqTopicObj *pTopic = mndAcquireTopic(pMnode, topic);
       if (pTopic == NULL) {
-        mError("mq rebalance %s failed since topic %s not exist, abort", pRebInfo->key, topic);
+        mError("mq re-balance %s ignored since topic %s not exist", pRebInfo->key, topic);
         continue;
       }
+
       taosRLockLatch(&pTopic->lock);
 
-      rebOutput.pSub = mndCreateSub(pMnode, pTopic, pRebInfo->key);
+      rebOutput.pSub = mndCreateSubscription(pMnode, pTopic, pRebInfo->key);
 
       if (rebOutput.pSub == NULL) {
         mError("mq rebalance %s failed create sub since %s, abort", pRebInfo->key, terrstr());
@@ -605,15 +615,16 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
     }
 
     if (mndDoRebalance(pMnode, &rebInput, &rebOutput) < 0) {
-      mError("mq rebalance internal error");
+      mError("mq re-balance internal error");
     }
 
     // if add more consumer to balanced subscribe,
     // possibly no vg is changed
 
     if (mndPersistRebResult(pMnode, pMsg, &rebOutput) < 0) {
-      mError("mq rebalance persist rebalance output error, possibly vnode splitted or dropped");
+      mError("mq re-balance persist re-balance output error, possibly vnode splitted or dropped");
     }
+
     taosArrayDestroy(pRebInfo->lostConsumers);
     taosArrayDestroy(pRebInfo->newConsumers);
     taosArrayDestroy(pRebInfo->removedConsumers);
@@ -627,19 +638,18 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
   }
 
   // reset flag
-  mInfo("mq rebalance completed successfully");
+  mInfo("mq re-balance completed successfully");
   taosHashCleanup(pReq->rebSubHash);
   mndRebEnd();
 
   return 0;
 }
 
-static int32_t mndProcessDropCgroupReq(SRpcMsg *pReq) {
+static int32_t mndProcessDropCgroupReq(SRpcMsg *pMsg) {
-  SMnode         *pMnode = pReq->info.node;
+  SMnode         *pMnode = pMsg->info.node;
-  SSdb           *pSdb = pMnode->pSdb;
   SMDropCgroupReq dropReq = {0};
 
-  if (tDeserializeSMDropCgroupReq(pReq->pCont, pReq->contLen, &dropReq) != 0) {
+  if (tDeserializeSMDropCgroupReq(pMsg->pCont, pMsg->contLen, &dropReq) != 0) {
     terrno = TSDB_CODE_INVALID_MSG;
     return -1;
   }
@@ -663,7 +673,7 @@ static int32_t mndProcessDropCgroupReq(SRpcMsg *pReq) {
     return -1;
   }
 
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_NOTHING, pReq, "drop-cgroup");
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_NOTHING, pMsg, "drop-cgroup");
   if (pTrans == NULL) {
     mError("cgroup: %s on topic:%s, failed to drop since %s", dropReq.cgroup, dropReq.topic, terrstr());
     mndReleaseSubscribe(pMnode, pSub);
@@ -956,7 +966,7 @@ END:
   return code;
 }
 
-static int32_t mndRetrieveSubscribe(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rowsCapacity) {
+int32_t mndRetrieveSubscribe(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rowsCapacity) {
   SMnode          *pMnode = pReq->info.node;
   SSdb            *pSdb = pMnode->pSdb;
   int32_t          numOfRows = 0;
@@ -966,7 +976,9 @@ static int32_t mndRetrieveSubscribe(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock
 
   while (numOfRows < rowsCapacity) {
     pShow->pIter = sdbFetch(pSdb, SDB_SUBSCRIBE, pShow->pIter, (void **)&pSub);
-    if (pShow->pIter == NULL) break;
+    if (pShow->pIter == NULL) {
+      break;
+    }
 
     taosRLockLatch(&pSub->lock);
 
@@ -1075,6 +1087,9 @@ static int32_t mndRetrieveSubscribe(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock
     }
 
 #endif
+
+    pBlock->info.rows = numOfRows;
+
     taosRUnLockLatch(&pSub->lock);
     sdbRelease(pSdb, pSub);
   }
@@ -1085,7 +1100,7 @@ static int32_t mndRetrieveSubscribe(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock
   return numOfRows;
 }
 
-static void mndCancelGetNextSubscribe(SMnode *pMnode, void *pIter) {
+void mndCancelGetNextSubscribe(SMnode *pMnode, void *pIter) {
   SSdb *pSdb = pMnode->pSdb;
   sdbCancelFetch(pSdb, pIter);
 }

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

@@ -85,7 +85,11 @@ int32_t mndProcessWriteMsg(const SSyncFSM *pFsm, SRpcMsg *pMsg, const SFsmCbMeta
         pRaw, pMgmt->transSec, pMgmt->transSeq);
 
   if (pMeta->code == 0) {
-    sdbWriteWithoutFree(pMnode->pSdb, pRaw);
+    int32_t code = sdbWriteWithoutFree(pMnode->pSdb, pRaw);
+    if (code != 0) {
+      mError("trans:%d, failed to write to sdb since %s", transId, terrstr());
+      return 0;
+    }
     sdbSetApplyInfo(pMnode->pSdb, pMeta->index, pMeta->term, pMeta->lastConfigIndex);
   }
 
@@ -110,8 +114,9 @@ int32_t mndProcessWriteMsg(const SSyncFSM *pFsm, SRpcMsg *pMsg, const SFsmCbMeta
     taosThreadMutexUnlock(&pMgmt->lock);
     STrans *pTrans = mndAcquireTrans(pMnode, transId);
     if (pTrans != NULL) {
-      mInfo("trans:%d, execute in mnode which not leader or sync timeout", transId);
+      mInfo("trans:%d, execute in mnode which not leader or sync timeout, createTime:%" PRId64 " saved trans:%d",
-      mndTransExecute(pMnode, pTrans);
+            transId, pTrans->createdTime, pMgmt->transId);
+      mndTransExecute(pMnode, pTrans, false);
       mndReleaseTrans(pMnode, pTrans);
       // sdbWriteFile(pMnode->pSdb, SDB_WRITE_DELTA);
     } else {
@@ -368,7 +373,7 @@ int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) {
   taosThreadMutexLock(&pMgmt->lock);
   pMgmt->errCode = 0;
 
-  if (pMgmt->transId != 0) {
+  if (pMgmt->transId != 0 /* && pMgmt->transId != transId*/) {
     mError("trans:%d, can't be proposed since trans:%d already waiting for confirm", transId, pMgmt->transId);
     taosThreadMutexUnlock(&pMgmt->lock);
     rpcFreeCont(req.pCont);

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

@@ -33,7 +33,7 @@
 
 static int32_t mndTopicActionInsert(SSdb *pSdb, SMqTopicObj *pTopic);
 static int32_t mndTopicActionDelete(SSdb *pSdb, SMqTopicObj *pTopic);
-static int32_t mndTopicActionUpdate(SSdb *pSdb, SMqTopicObj *pTopic, SMqTopicObj *pNewTopic);
+static int32_t mndTopicActionUpdate(SSdb *pSdb, SMqTopicObj *pOldTopic, SMqTopicObj *pNewTopic);
 static int32_t mndProcessCreateTopicReq(SRpcMsg *pReq);
 static int32_t mndProcessDropTopicReq(SRpcMsg *pReq);
 

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

@@ -572,8 +572,20 @@ static void mndTransUpdateActions(SArray *pOldArray, SArray *pNewArray) {
 }
 
 static int32_t mndTransActionUpdate(SSdb *pSdb, STrans *pOld, STrans *pNew) {
-  mTrace("trans:%d, perform update action, old row:%p stage:%s, new row:%p stage:%s", pOld->id, pOld,
+  mTrace("trans:%d, perform update action, old row:%p stage:%s create:%" PRId64 ", new row:%p stage:%s create:%" PRId64,
-         mndTransStr(pOld->stage), pNew, mndTransStr(pNew->stage));
+         pOld->id, pOld, mndTransStr(pOld->stage), pOld->createdTime, pNew, mndTransStr(pNew->stage),
+         pNew->createdTime);
+
+  if (pOld->createdTime != pNew->createdTime) {
+    mError("trans:%d, failed to perform update action since createTime not match, old row:%p stage:%s create:%" PRId64
+           ", new row:%p stage:%s create:%" PRId64,
+           pOld->id, pOld, mndTransStr(pOld->stage), pOld->createdTime, pNew, mndTransStr(pNew->stage),
+           pNew->createdTime);
+    // only occured while sync timeout
+    terrno = TSDB_CODE_MND_TRNAS_SYNC_TIMEOUT;
+    return -1;
+  }
+
   mndTransUpdateActions(pOld->redoActions, pNew->redoActions);
   mndTransUpdateActions(pOld->undoActions, pNew->undoActions);
   mndTransUpdateActions(pOld->commitActions, pNew->commitActions);
@@ -779,16 +791,18 @@ static int32_t mndTransSync(SMnode *pMnode, STrans *pTrans) {
   }
   (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
 
-  mInfo("trans:%d, sync to other mnodes, stage:%s", pTrans->id, mndTransStr(pTrans->stage));
+  mInfo("trans:%d, sync to other mnodes, stage:%s createTime:%" PRId64, pTrans->id, mndTransStr(pTrans->stage),
+        pTrans->createdTime);
   int32_t code = mndSyncPropose(pMnode, pRaw, pTrans->id);
   if (code != 0) {
-    mError("trans:%d, failed to sync, errno:%s code:%s", pTrans->id, terrstr(), tstrerror(code));
+    mError("trans:%d, failed to sync, errno:%s code:%s createTime:%" PRId64 " saved trans:%d", pTrans->id, terrstr(),
+           tstrerror(code), pTrans->createdTime, pMnode->syncMgmt.transId);
     sdbFreeRaw(pRaw);
     return -1;
   }
 
   sdbFreeRaw(pRaw);
-  mInfo("trans:%d, sync finished", pTrans->id);
+  mInfo("trans:%d, sync finished, createTime:%" PRId64, pTrans->id, pTrans->createdTime);
   return 0;
 }
 
@@ -891,7 +905,7 @@ int32_t mndTransPrepare(SMnode *pMnode, STrans *pTrans) {
   pTrans->rpcRsp = NULL;
   pTrans->rpcRspLen = 0;
 
-  mndTransExecute(pMnode, pNew);
+  mndTransExecute(pMnode, pNew, true);
   mndReleaseTrans(pMnode, pNew);
   return 0;
 }
@@ -1054,7 +1068,7 @@ int32_t mndTransProcessRsp(SRpcMsg *pRsp) {
 
   mInfo("trans:%d, %s:%d response is received, code:0x%x, accept:0x%x retry:0x%x", transId, mndTransStr(pAction->stage),
         action, pRsp->code, pAction->acceptableCode, pAction->retryCode);
-  mndTransExecute(pMnode, pTrans);
+  mndTransExecute(pMnode, pTrans, true);
 
 _OVER:
   mndReleaseTrans(pMnode, pTrans);
@@ -1483,15 +1497,17 @@ static bool mndTransPerfromFinishedStage(SMnode *pMnode, STrans *pTrans) {
     mError("trans:%d, failed to write sdb since %s", pTrans->id, terrstr());
   }
 
-  mInfo("trans:%d, execute finished, code:0x%x, failedTimes:%d", pTrans->id, pTrans->code, pTrans->failedTimes);
+  mInfo("trans:%d, execute finished, code:0x%x, failedTimes:%d createTime:%" PRId64, pTrans->id, pTrans->code,
+        pTrans->failedTimes, pTrans->createdTime);
   return continueExec;
 }
 
-void mndTransExecute(SMnode *pMnode, STrans *pTrans) {
+void mndTransExecute(SMnode *pMnode, STrans *pTrans, bool isLeader) {
   bool continueExec = true;
 
   while (continueExec) {
-    mInfo("trans:%d, continue to execute, stage:%s", pTrans->id, mndTransStr(pTrans->stage));
+    mInfo("trans:%d, continue to execute, stage:%s createTime:%" PRId64 " leader:%d", pTrans->id,
+          mndTransStr(pTrans->stage), pTrans->createdTime, isLeader);
     pTrans->lastExecTime = taosGetTimestampMs();
     switch (pTrans->stage) {
       case TRN_STAGE_PREPARE:
@@ -1501,13 +1517,23 @@ void mndTransExecute(SMnode *pMnode, STrans *pTrans) {
         continueExec = mndTransPerformRedoActionStage(pMnode, pTrans);
         break;
       case TRN_STAGE_COMMIT:
-        continueExec = mndTransPerformCommitStage(pMnode, pTrans);
+        if (isLeader) {
+          continueExec = mndTransPerformCommitStage(pMnode, pTrans);
+        } else {
+          mInfo("trans:%d, can not commit since not leader", pTrans->id);
+          continueExec = false;
+        }
         break;
       case TRN_STAGE_COMMIT_ACTION:
         continueExec = mndTransPerformCommitActionStage(pMnode, pTrans);
         break;
       case TRN_STAGE_ROLLBACK:
-        continueExec = mndTransPerformRollbackStage(pMnode, pTrans);
+        if (isLeader) {
+          continueExec = mndTransPerformRollbackStage(pMnode, pTrans);
+        } else {
+          mInfo("trans:%d, can not rollback since not leader", pTrans->id);
+          continueExec = false;
+        }
         break;
       case TRN_STAGE_UNDO_ACTION:
         continueExec = mndTransPerformUndoActionStage(pMnode, pTrans);
@@ -1550,7 +1576,7 @@ int32_t mndKillTrans(SMnode *pMnode, STrans *pTrans) {
     pAction->errCode = 0;
   }
 
-  mndTransExecute(pMnode, pTrans);
+  mndTransExecute(pMnode, pTrans, true);
   return 0;
 }
 
@@ -1608,7 +1634,7 @@ void mndTransPullup(SMnode *pMnode) {
     int32_t *pTransId = taosArrayGet(pArray, i);
     STrans  *pTrans = mndAcquireTrans(pMnode, *pTransId);
     if (pTrans != NULL) {
-      mndTransExecute(pMnode, pTrans);
+      mndTransExecute(pMnode, pTrans, true);
     }
     mndReleaseTrans(pMnode, pTrans);
   }

source/dnode/qnode/inc/qndInt.h

@@ -29,7 +29,7 @@
 extern "C" {
 #endif
 
-typedef struct SQWorker SQHandle;
+typedef struct SQueueWorker SQHandle;
 
 typedef struct SQnode {
   int32_t   qndId;

source/dnode/vnode/inc/vnode.h

@@ -104,8 +104,8 @@ void        metaReaderClear(SMetaReader *pReader);
 int32_t     metaGetTableEntryByUid(SMetaReader *pReader, tb_uid_t uid);
 int32_t     metaGetTableEntryByUidCache(SMetaReader *pReader, tb_uid_t uid);
 int         metaGetTableEntryByName(SMetaReader *pReader, const char *name);
-int32_t     metaGetTableTags(SMeta *pMeta, uint64_t suid, SArray *uidList, SHashObj *tags);
+int32_t     metaGetTableTags(SMeta *pMeta, uint64_t suid, SArray *uidList);
-int32_t     metaGetTableTagsByUids(SMeta *pMeta, int64_t suid, SArray *uidList, SHashObj *tags);
+int32_t     metaGetTableTagsByUids(SMeta *pMeta, int64_t suid, SArray *uidList);
 int32_t     metaReadNext(SMetaReader *pReader);
 const void *metaGetTableTagVal(void *tag, int16_t type, STagVal *tagVal);
 int         metaGetTableNameByUid(void *meta, uint64_t uid, char *tbName);
@@ -178,7 +178,6 @@ int32_t tsdbReaderOpen(SVnode *pVnode, SQueryTableDataCond *pCond, void *pTableL
 
 void         tsdbReaderClose(STsdbReader *pReader);
 bool         tsdbNextDataBlock(STsdbReader *pReader);
-void         tsdbRetrieveDataBlockInfo(const STsdbReader *pReader, int32_t *rows, uint64_t *uid, STimeWindow *pWindow);
 int32_t      tsdbRetrieveDatablockSMA(STsdbReader *pReader, SSDataBlock *pDataBlock, bool *allHave);
 SSDataBlock *tsdbRetrieveDataBlock(STsdbReader *pTsdbReadHandle, SArray *pColumnIdList);
 int32_t      tsdbReaderReset(STsdbReader *pReader, SQueryTableDataCond *pCond);

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

@@ -236,6 +236,7 @@ void tsdbHeadFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SHeadFile *pHeadF,
 void tsdbDataFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SDataFile *pDataF, char fname[]);
 void tsdbSttFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SSttFile *pSttF, char fname[]);
 void tsdbSmaFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SSmaFile *pSmaF, char fname[]);
+
 // SDelFile
 void tsdbDelFileName(STsdb *pTsdb, SDelFile *pFile, char fname[]);
 // tsdbFS.c ==============================================================================================
@@ -321,6 +322,8 @@ struct STsdb {
   STsdbFS        fs;
   SLRUCache     *lruCache;
   TdThreadMutex  lruMutex;
+  SLRUCache     *biCache;
+  TdThreadMutex  biMutex;
 };
 
 struct TSDBKEY {
@@ -644,6 +647,7 @@ typedef struct SSttBlockLoadInfo {
   int16_t   *colIds;
   int32_t    numOfCols;
   bool       sttBlockLoaded;
+  int32_t    numOfStt;
 
   // keep the last access position, this position may be used to reduce the binary times for
   // starting last block data for a new table
@@ -709,7 +713,7 @@ bool    tMergeTreeNext(SMergeTree *pMTree);
 TSDBROW tMergeTreeGetRow(SMergeTree *pMTree);
 void    tMergeTreeClose(SMergeTree *pMTree);
 
-SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols);
+SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols, int32_t numOfStt);
 void               resetLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo);
 void               getLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo, int64_t *blocks, double *el);
 void              *destroyLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo);
@@ -746,6 +750,9 @@ int32_t tsdbCacheGetLastH(SLRUCache *pCache, tb_uid_t uid, SCacheRowsReader *pr,
 int32_t tsdbCacheGetLastrowH(SLRUCache *pCache, tb_uid_t uid, SCacheRowsReader *pr, LRUHandle **h);
 int32_t tsdbCacheRelease(SLRUCache *pCache, LRUHandle *h);
 
+int32_t tsdbCacheGetBlockIdx(SLRUCache *pCache, SDataFReader *pFileReader, LRUHandle **handle);
+int32_t tsdbBICacheRelease(SLRUCache *pCache, LRUHandle *h);
+
 int32_t tsdbCacheDeleteLastrow(SLRUCache *pCache, tb_uid_t uid, TSKEY eKey);
 int32_t tsdbCacheDeleteLast(SLRUCache *pCache, tb_uid_t uid, TSKEY eKey);
 int32_t tsdbCacheDelete(SLRUCache *pCache, tb_uid_t uid, TSKEY eKey);

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

@@ -58,7 +58,7 @@ typedef struct STQ                STQ;
 typedef struct SVState            SVState;
 typedef struct SVStatis           SVStatis;
 typedef struct SVBufPool          SVBufPool;
-typedef struct SQWorker           SQHandle;
+typedef struct SQueueWorker       SQHandle;
 typedef struct STsdbKeepCfg       STsdbKeepCfg;
 typedef struct SMetaSnapReader    SMetaSnapReader;
 typedef struct SMetaSnapWriter    SMetaSnapWriter;
@@ -107,7 +107,7 @@ typedef struct STbUidStore STbUidStore;
 #define META_BEGIN_HEAP_NIL        2
 
 int             metaOpen(SVnode* pVnode, SMeta** ppMeta, int8_t rollback);
-int             metaClose(SMeta* pMeta);
+int             metaClose(SMeta** pMeta);
 int             metaBegin(SMeta* pMeta, int8_t fromSys);
 TXN*            metaGetTxn(SMeta* pMeta);
 int             metaCommit(SMeta* pMeta, TXN* txn);

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

@@ -32,9 +32,9 @@ typedef struct SMetaStbStatsEntry {
 } SMetaStbStatsEntry;
 
 typedef struct STagFilterResEntry {
-  uint64_t suid;    // uid for super table
   SList    list;    // the linked list of md5 digest, extracted from the serialized tag query condition
-  uint32_t qTimes;  // queried times for current super table
+  uint32_t hitTimes;  // queried times for current super table
+  uint32_t accTime;
 } STagFilterResEntry;
 
 struct SMetaCache {
@@ -55,6 +55,7 @@ struct SMetaCache {
   // query cache
   struct STagFilterResCache {
     TdThreadMutex lock;
+    uint32_t   accTimes;
     SHashObj*  pTableEntry;
     SLRUCache* pUidResCache;
   } sTagFilterResCache;
@@ -132,6 +133,7 @@ int32_t metaCacheOpen(SMeta* pMeta) {
     goto _err2;
   }
 
+  pCache->sTagFilterResCache.accTimes = 0;
   pCache->sTagFilterResCache.pTableEntry =
       taosHashInit(1024, taosGetDefaultHashFunction(TSDB_DATA_TYPE_VARCHAR), false, HASH_NO_LOCK);
   if (pCache->sTagFilterResCache.pTableEntry == NULL) {
@@ -159,9 +161,9 @@ void metaCacheClose(SMeta* pMeta) {
     entryCacheClose(pMeta);
     statsCacheClose(pMeta);
 
-    taosHashCleanup(pMeta->pCache->sTagFilterResCache.pTableEntry);
     taosLRUCacheCleanup(pMeta->pCache->sTagFilterResCache.pUidResCache);
     taosThreadMutexDestroy(&pMeta->pCache->sTagFilterResCache.lock);
+    taosHashCleanup(pMeta->pCache->sTagFilterResCache.pTableEntry);
 
     taosMemoryFree(pMeta->pCache);
     pMeta->pCache = NULL;
@@ -424,6 +426,31 @@ int32_t metaStatsCacheGet(SMeta* pMeta, int64_t uid, SMetaStbStats* pInfo) {
   return code;
 }
 
+static int checkAllEntriesInCache(const STagFilterResEntry* pEntry, SArray* pInvalidRes, int32_t keyLen, SLRUCache* pCache, uint64_t suid) {
+  SListIter iter = {0};
+  tdListInitIter((SList*)&(pEntry->list), &iter, TD_LIST_FORWARD);
+
+  SListNode* pNode = NULL;
+  uint64_t buf[3];
+  buf[0] = suid;
+
+  int32_t len = sizeof(uint64_t) * tListLen(buf);
+
+  while ((pNode = tdListNext(&iter)) != NULL) {
+    memcpy(&buf[1], pNode->data, keyLen);
+
+    // check whether it is existed in LRU cache, and remove it from linked list if not.
+    LRUHandle* pRes = taosLRUCacheLookup(pCache, buf, len);
+    if (pRes == NULL) {  // remove the item in the linked list
+      taosArrayPush(pInvalidRes, &pNode);
+    } else {
+      taosLRUCacheRelease(pCache, pRes, false);
+    }
+  }
+
+  return 0;
+}
+
 int32_t metaGetCachedTableUidList(SMeta* pMeta, tb_uid_t suid, const uint8_t* pKey, int32_t keyLen, SArray* pList1,
                                   bool* acquireRes) {
   // generate the composed key for LRU cache
@@ -431,16 +458,18 @@ int32_t metaGetCachedTableUidList(SMeta* pMeta, tb_uid_t suid, const uint8_t* pK
   SHashObj*      pTableMap = pMeta->pCache->sTagFilterResCache.pTableEntry;
   TdThreadMutex* pLock = &pMeta->pCache->sTagFilterResCache.lock;
 
-  uint64_t buf[3] = {0};
+  uint64_t buf[4];
-  uint32_t times = 0;
 
   *acquireRes = 0;
-  buf[0] = suid;
+
-  memcpy(&buf[1], pKey, keyLen);
+  buf[0] = (uint64_t) pTableMap;
+  buf[1] = suid;
+  memcpy(&buf[2], pKey, keyLen);
 
   taosThreadMutexLock(pLock);
+  pMeta->pCache->sTagFilterResCache.accTimes += 1;
 
-  int32_t    len = keyLen + sizeof(uint64_t);
+  int32_t    len = keyLen + sizeof(uint64_t) * 2;
   LRUHandle* pHandle = taosLRUCacheLookup(pCache, buf, len);
   if (pHandle == NULL) {
     taosThreadMutexUnlock(pLock);
@@ -458,48 +487,17 @@ int32_t metaGetCachedTableUidList(SMeta* pMeta, tb_uid_t suid, const uint8_t* pK
   // set the result into the buffer
   taosArrayAddBatch(pList1, p + sizeof(int32_t), size);
 
-  times = atomic_add_fetch_32(&(*pEntry)->qTimes, 1);
+  (*pEntry)->hitTimes += 1;
+
+  uint32_t acc = pMeta->pCache->sTagFilterResCache.accTimes;
+  if ((*pEntry)->hitTimes % 5000 == 0 && (*pEntry)->hitTimes > 0) {
+    metaInfo("cache hit:%d, total acc:%d, rate:%.2f", (*pEntry)->hitTimes, acc, ((double)(*pEntry)->hitTimes) / acc);
+  }
+
   taosLRUCacheRelease(pCache, pHandle, false);
 
   // unlock meta
   taosThreadMutexUnlock(pLock);
-
-  // check if scanning all items are necessary or not
-  if (times >= 5000 && TD_DLIST_NELES(&(*pEntry)->list) > 10) {
-    taosThreadMutexLock(pLock);
-
-    SArray* pInvalidRes = taosArrayInit(64, POINTER_BYTES);
-
-    SListIter iter = {0};
-    tdListInitIter(&(*pEntry)->list, &iter, TD_LIST_FORWARD);
-
-    SListNode* pNode = NULL;
-    while ((pNode = tdListNext(&iter)) != NULL) {
-      memcpy(&buf[1], pNode->data, keyLen);
-
-      // check whether it is existed in LRU cache, and remove it from linked list if not.
-      LRUHandle* pRes = taosLRUCacheLookup(pCache, buf, len);
-      if (pRes == NULL) {  // remove the item in the linked list
-        taosArrayPush(pInvalidRes, &pNode);
-      } else {
-        taosLRUCacheRelease(pCache, pRes, false);
-      }
-    }
-
-    // remove the keys, of which query uid lists have been replaced already.
-    size_t s = taosArrayGetSize(pInvalidRes);
-    for (int32_t i = 0; i < s; ++i) {
-      SListNode** p1 = taosArrayGet(pInvalidRes, i);
-      tdListPopNode(&(*pEntry)->list, *p1);
-      taosMemoryFree(*p1);
-    }
-
-    atomic_store_32(&(*pEntry)->qTimes, 0); // reset the query times
-    taosArrayDestroy(pInvalidRes);
-
-    taosThreadMutexUnlock(pLock);
-  }
-
   return TSDB_CODE_SUCCESS;
 }
 
@@ -507,9 +505,53 @@ static void freePayload(const void* key, size_t keyLen, void* value) {
   if (value == NULL) {
     return;
   }
+
+  const uint64_t* p = key;
+  if (keyLen != sizeof(int64_t) * 4) {
+    metaError("key length is invalid, length:%d, expect:%d", (int32_t)keyLen, (int32_t)sizeof(uint64_t) * 2);
+    return;
+  }
+
+  SHashObj*            pHashObj = (SHashObj*)p[0];
+  STagFilterResEntry** pEntry = taosHashGet(pHashObj, &p[1], sizeof(uint64_t));
+
+  {
+    int64_t st = taosGetTimestampUs();
+
+    SListIter iter = {0};
+    tdListInitIter((SList*)&((*pEntry)->list), &iter, TD_LIST_FORWARD);
+
+    SListNode* pNode = NULL;
+    while ((pNode = tdListNext(&iter)) != NULL) {
+      uint64_t* digest = (uint64_t*)pNode->data;
+      if (digest[0] == p[2] && digest[1] == p[3]) {
+        void* tmp = tdListPopNode(&((*pEntry)->list), pNode);
+        taosMemoryFree(tmp);
+
+        int64_t et = taosGetTimestampUs();
+        metaInfo("clear items in cache, remain cached item:%d, elapsed time:%.2fms", listNEles(&((*pEntry)->list)),
+                 (et - st) / 1000.0);
+        break;
+      }
+    }
+  }
+
   taosMemoryFree(value);
 }
 
+static int32_t addNewEntry(SHashObj* pTableEntry, const void* pKey, int32_t keyLen, uint64_t suid) {
+  STagFilterResEntry* p = taosMemoryMalloc(sizeof(STagFilterResEntry));
+  if (p == NULL) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  p->hitTimes = 0;
+  tdListInit(&p->list, keyLen);
+  taosHashPut(pTableEntry, &suid, sizeof(uint64_t), &p, POINTER_BYTES);
+  tdListAppend(&p->list, pKey);
+  return 0;
+}
+
 // check both the payload size and selectivity ratio
 int32_t metaUidFilterCachePut(SMeta* pMeta, uint64_t suid, const void* pKey, int32_t keyLen, void* pPayload,
                               int32_t payloadLen, double selectivityRatio) {
@@ -533,42 +575,61 @@ int32_t metaUidFilterCachePut(SMeta* pMeta, uint64_t suid, const void* pKey, int
   SHashObj*      pTableEntry = pMeta->pCache->sTagFilterResCache.pTableEntry;
   TdThreadMutex* pLock = &pMeta->pCache->sTagFilterResCache.lock;
 
+  // the format of key:
+  // hash table address(8bytes) + suid(8bytes) + MD5 digest(16bytes)
+
+  uint64_t buf[4] = {0};
+  buf[0] = (uint64_t) pTableEntry;
+  buf[1] = suid;
+  memcpy(&buf[2], pKey, keyLen);
+  ASSERT(keyLen == 16);
+
+  int32_t code = 0;
   taosThreadMutexLock(pLock);
 
   STagFilterResEntry** pEntry = taosHashGet(pTableEntry, &suid, sizeof(uint64_t));
   if (pEntry == NULL) {
-    STagFilterResEntry* p = taosMemoryMalloc(sizeof(STagFilterResEntry));
+    code = addNewEntry(pTableEntry, pKey, keyLen, suid);
-    p->qTimes = 0;
+    if (code != TSDB_CODE_SUCCESS) {
-    tdListInit(&p->list, keyLen);
+      goto _end;
-    taosHashPut(pTableEntry, &suid, sizeof(uint64_t), &p, POINTER_BYTES);
+    }
-    tdListAppend(&p->list, pKey);
   } else {
-    tdListAppend(&(*pEntry)->list, pKey);
+    // check if it exists or not
+    size_t size = listNEles(&(*pEntry)->list);
+    if (size == 0) {
+      tdListAppend(&(*pEntry)->list, pKey);
+    } else {
+      SListNode* pNode = listHead(&(*pEntry)->list);
+      uint64_t*  p = (uint64_t*)pNode->data;
+      if (p[1] == ((uint64_t*)pKey)[1] && p[0] == ((uint64_t*)pKey)[0]) {
+        // we have already found the existed items, no need to added to cache anymore.
+        taosThreadMutexUnlock(pLock);
+        return TSDB_CODE_SUCCESS;
+      } else {  // not equal, append it
+        tdListAppend(&(*pEntry)->list, pKey);
+      }
+    }
   }
 
-  uint64_t buf[3] = {0};
-  buf[0] = suid;
-
-  memcpy(&buf[1], pKey, keyLen);
-  ASSERT(sizeof(uint64_t) + keyLen == 24);
-
   // add to cache.
-  taosLRUCacheInsert(pCache, buf, sizeof(uint64_t) + keyLen, pPayload, payloadLen, freePayload, NULL,
+  taosLRUCacheInsert(pCache, buf, sizeof(uint64_t)*2 + keyLen, pPayload, payloadLen, freePayload, NULL,
                      TAOS_LRU_PRIORITY_LOW);
-
+  _end:
   taosThreadMutexUnlock(pLock);
 
   metaDebug("vgId:%d, suid:%" PRIu64 " list cache added into cache, total:%d, tables:%d", TD_VID(pMeta->pVnode), suid,
             (int32_t)taosLRUCacheGetUsage(pCache), taosHashGetSize(pTableEntry));
 
-  return TSDB_CODE_SUCCESS;
+  return code;
 }
 
 // remove the lru cache that are expired due to the tags value update, or creating, or dropping, of child tables
 int32_t metaUidCacheClear(SMeta* pMeta, uint64_t suid) {
   int32_t  keyLen = sizeof(uint64_t) * 3;
-  uint64_t p[3] = {0};
+  uint64_t p[4] = {0};
-  p[0] = suid;
+
+  p[0] = (uint64_t) pMeta->pCache->sTagFilterResCache.pTableEntry;
+  p[1] = suid;
 
   TdThreadMutex* pLock = &pMeta->pCache->sTagFilterResCache.lock;
 
@@ -584,11 +645,11 @@ int32_t metaUidCacheClear(SMeta* pMeta, uint64_t suid) {
 
   SListNode* pNode = NULL;
   while ((pNode = tdListNext(&iter)) != NULL) {
-    memcpy(&p[1], pNode->data, 16);
+    memcpy(&p[2], pNode->data, 16);
     taosLRUCacheErase(pMeta->pCache->sTagFilterResCache.pUidResCache, p, keyLen);
   }
 
-  (*pEntry)->qTimes = 0;
+  (*pEntry)->hitTimes = 0;
   tdListEmpty(&(*pEntry)->list);
 
   taosThreadMutexUnlock(pLock);

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

@@ -201,7 +201,8 @@ _err:
   return -1;
 }
 
-int metaClose(SMeta *pMeta) {
+int metaClose(SMeta **ppMeta) {
+  SMeta *pMeta = *ppMeta;
   if (pMeta) {
     if (pMeta->pEnv) metaAbort(pMeta);
     if (pMeta->pCache) metaCacheClose(pMeta);
@@ -221,7 +222,8 @@ int metaClose(SMeta *pMeta) {
     if (pMeta->pTbDb) tdbTbClose(pMeta->pTbDb);
     if (pMeta->pEnv) tdbClose(pMeta->pEnv);
     metaDestroyLock(pMeta);
-    taosMemoryFree(pMeta);
+
+    taosMemoryFreeClear(*ppMeta);
   }
 
   return 0;

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

@@ -1371,13 +1371,14 @@ static int32_t metaGetTableTagByUid(SMeta *pMeta, int64_t suid, int64_t uid, voi
 
   return ret;
 }
-int32_t metaGetTableTagsByUids(SMeta *pMeta, int64_t suid, SArray *uidList, SHashObj *tags) {
+
+int32_t metaGetTableTagsByUids(SMeta *pMeta, int64_t suid, SArray *uidList) {
   const int32_t LIMIT = 128;
 
   int32_t isLock = false;
   int32_t sz = uidList ? taosArrayGetSize(uidList) : 0;
   for (int i = 0; i < sz; i++) {
-    tb_uid_t *id = taosArrayGet(uidList, i);
+    STUidTagInfo *p = taosArrayGet(uidList, i);
 
     if (i % LIMIT == 0) {
       if (isLock) metaULock(pMeta);
@@ -1386,51 +1387,72 @@ int32_t metaGetTableTagsByUids(SMeta *pMeta, int64_t suid, SArray *uidList, SHas
       isLock = true;
     }
 
-    if (taosHashGet(tags, id, sizeof(tb_uid_t)) == NULL) {
+//    if (taosHashGet(tags, &p->uid, sizeof(tb_uid_t)) == NULL) {
       void   *val = NULL;
       int32_t len = 0;
-      if (metaGetTableTagByUid(pMeta, suid, *id, &val, &len, false) == 0) {
+      if (metaGetTableTagByUid(pMeta, suid, p->uid, &val, &len, false) == 0) {
-        taosHashPut(tags, id, sizeof(tb_uid_t), val, len);
+        p->pTagVal = taosMemoryMalloc(len);
+        memcpy(p->pTagVal, val, len);
         tdbFree(val);
       } else {
-        metaError("vgId:%d, failed to table IDs, suid: %" PRId64 ", uid: %" PRId64 "", TD_VID(pMeta->pVnode), suid,
+        metaError("vgId:%d, failed to table tags, suid: %" PRId64 ", uid: %" PRId64 "", TD_VID(pMeta->pVnode), suid,
-                  *id);
+                  p->uid);
       }
     }
-  }
+//  }
   if (isLock) metaULock(pMeta);
-
   return 0;
 }
 
-int32_t metaGetTableTags(SMeta *pMeta, uint64_t suid, SArray *uidList, SHashObj *tags) {
+int32_t metaGetTableTags(SMeta *pMeta, uint64_t suid, SArray *pUidTagInfo) {
   SMCtbCursor *pCur = metaOpenCtbCursor(pMeta, suid, 1);
 
-  SHashObj *uHash = NULL;
+  // If len > 0 means there already have uids, and we only want the
-  size_t    len = taosArrayGetSize(uidList);  // len > 0 means there already have uids
+  // tags of the specified tables, of which uid in the uid list. Otherwise, all table tags are retrieved and kept
-  if (len > 0) {
+  // in the hash map, that may require a lot of memory
-    uHash = taosHashInit(32, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
+  SHashObj *pSepecifiedUidMap = NULL;
-    for (int i = 0; i < len; i++) {
+  size_t    numOfElems = taosArrayGetSize(pUidTagInfo);
-      int64_t *uid = taosArrayGet(uidList, i);
+  if (numOfElems > 0) {
-      taosHashPut(uHash, uid, sizeof(int64_t), &i, sizeof(i));
+    pSepecifiedUidMap = taosHashInit(numOfElems / 0.7, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
+    for (int i = 0; i < numOfElems; i++) {
+      STUidTagInfo *pTagInfo = taosArrayGet(pUidTagInfo, i);
+      taosHashPut(pSepecifiedUidMap, &pTagInfo->uid, sizeof(uint64_t), &i, sizeof(int32_t));
     }
   }
-  while (1) {
-    tb_uid_t id = metaCtbCursorNext(pCur);
-    if (id == 0) {
-      break;
-    }
 
-    if (len > 0 && taosHashGet(uHash, &id, sizeof(int64_t)) == NULL) {
+  if (numOfElems == 0) {  // all data needs to be added into the pUidTagInfo list
-      continue;
+    while (1) {
-    } else if (len == 0) {
+      tb_uid_t uid = metaCtbCursorNext(pCur);
-      taosArrayPush(uidList, &id);
+      if (uid == 0) {
-    }
+        break;
+      }
 
-    taosHashPut(tags, &id, sizeof(int64_t), pCur->pVal, pCur->vLen);
+      STUidTagInfo info = {.uid = uid, .pTagVal = pCur->pVal};
+      info.pTagVal = taosMemoryMalloc(pCur->vLen);
+      memcpy(info.pTagVal, pCur->pVal, pCur->vLen);
+      taosArrayPush(pUidTagInfo, &info);
+    }
+  } else {  // only the specified tables need to be added
+    while (1) {
+      tb_uid_t uid = metaCtbCursorNext(pCur);
+      if (uid == 0) {
+        break;
+      }
+
+      int32_t *index = taosHashGet(pSepecifiedUidMap, &uid, sizeof(uint64_t));
+      if (index == NULL) {
+        continue;
+      }
+
+      STUidTagInfo *pTagInfo = taosArrayGet(pUidTagInfo, *index);
+      if (pTagInfo->pTagVal == NULL) {
+        pTagInfo->pTagVal = taosMemoryMalloc(pCur->vLen);
+        memcpy(pTagInfo->pTagVal, pCur->pVal, pCur->vLen);
+      }
+    }
   }
 
-  taosHashCleanup(uHash);
+  taosHashCleanup(pSepecifiedUidMap);
   metaCloseCtbCursor(pCur, 1);
   return TSDB_CODE_SUCCESS;
 }

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

@@ -1343,6 +1343,9 @@ static int metaUpdateTagIdx(SMeta *pMeta, const SMetaEntry *pCtbEntry) {
   int32_t        ret = 0;
   // get super table
   if (tdbTbGet(pMeta->pUidIdx, &pCtbEntry->ctbEntry.suid, sizeof(tb_uid_t), &pData, &nData) != 0) {
+    metaError("vgId:%d, failed to get stable suid for update. version:%" PRId64, TD_VID(pMeta->pVnode),
+              pCtbEntry->version);
+    terrno = TSDB_CODE_TDB_INVALID_TABLE_ID;
     ret = -1;
     goto end;
   }

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

@@ -279,10 +279,7 @@ static void tdDestroyRSmaStat(void *pRSmaStat) {
     // step 1: set rsma trigger stat cancelled
     atomic_store_8(RSMA_TRIGGER_STAT(pStat), TASK_TRIGGER_STAT_CANCELLED);
 
-    // step 2: destroy the rsma info and associated fetch tasks
+    // step 2: wait for all triggered fetch tasks to finish
-    taosHashCleanup(RSMA_INFO_HASH(pStat));
-
-    // step 3: wait for all triggered fetch tasks to finish
     int32_t nLoops = 0;
     while (1) {
       if (T_REF_VAL_GET((SSmaStat *)pStat) == 0) {
@@ -298,9 +295,12 @@ static void tdDestroyRSmaStat(void *pRSmaStat) {
       }
     }
 
-    // step 4:
+    // step 3:
     tdRsmaStopExecutor(pSma);
 
+    // step 4: destroy the rsma info and associated fetch tasks
+    taosHashCleanup(RSMA_INFO_HASH(pStat));
+
     // step 5:
     tdRSmaFSClose(RSMA_FS(pStat));
 

source/dnode/vnode/src/tq/tq.c

@@ -106,7 +106,7 @@ STQ* tqOpen(const char* path, SVnode* pVnode) {
     return NULL;
   }
 
-  if (streamLoadTasks(pTq->pStreamMeta) < 0) {
+  if (streamLoadTasks(pTq->pStreamMeta, walGetCommittedVer(pVnode->pWal)) < 0) {
     return NULL;
   }
 
@@ -816,7 +816,6 @@ int32_t tqProcessSubscribeReq(STQ* pTq, int64_t version, char* msg, int32_t msgL
     // TODO version should be assigned and refed during preprocess
     SWalRef* pRef = walRefCommittedVer(pTq->pVnode->pWal);
     if (pRef == NULL) {
-      ASSERT(0);
       return -1;
     }
     int64_t ver = pRef->refVer;
@@ -837,12 +836,9 @@ int32_t tqProcessSubscribeReq(STQ* pTq, int64_t version, char* msg, int32_t msgL
 
       pHandle->execHandle.task =
           qCreateQueueExecTaskInfo(pHandle->execHandle.execCol.qmsg, &handle, &pHandle->execHandle.numOfCols, NULL);
-      ASSERT(pHandle->execHandle.task);
       void* scanner = NULL;
       qExtractStreamScanner(pHandle->execHandle.task, &scanner);
-      ASSERT(scanner);
       pHandle->execHandle.pExecReader = qExtractReaderFromStreamScanner(scanner);
-      ASSERT(pHandle->execHandle.pExecReader);
     } else if (pHandle->execHandle.subType == TOPIC_SUB_TYPE__DB) {
       pHandle->pWalReader = walOpenReader(pTq->pVnode->pWal, NULL);
       pHandle->execHandle.pExecReader = tqOpenReader(pTq->pVnode);
@@ -875,8 +871,7 @@ int32_t tqProcessSubscribeReq(STQ* pTq, int64_t version, char* msg, int32_t msgL
     taosHashPut(pTq->pHandle, req.subKey, strlen(req.subKey), pHandle, sizeof(STqHandle));
     tqDebug("try to persist handle %s consumer %" PRId64, req.subKey, pHandle->consumerId);
     if (tqMetaSaveHandle(pTq, req.subKey, pHandle) < 0) {
-      // TODO
+      return -1;
-      ASSERT(0);
     }
   } else {
     /*ASSERT(pExec->consumerId == req.oldConsumerId);*/
@@ -886,8 +881,7 @@ int32_t tqProcessSubscribeReq(STQ* pTq, int64_t version, char* msg, int32_t msgL
     atomic_add_fetch_32(&pHandle->epoch, 1);
     taosMemoryFree(req.qmsg);
     if (tqMetaSaveHandle(pTq, req.subKey, pHandle) < 0) {
-      // TODO
+      return -1;
-      ASSERT(0);
     }
     // close handle
   }
@@ -1202,6 +1196,9 @@ int32_t tqProcessTaskRecover2Req(STQ* pTq, int64_t version, char* msg, int32_t m
     return -1;
   }
 
+  atomic_store_8(&pTask->fillHistory, 0);
+  streamMetaSaveTask(pTq->pStreamMeta, pTask);
+
   streamMetaReleaseTask(pTq->pStreamMeta, pTask);
 
   return 0;

source/dnode/vnode/src/tq/tqMeta.c

@@ -71,17 +71,14 @@ int32_t tDecodeSTqHandle(SDecoder* pDecoder, STqHandle* pHandle) {
 
 int32_t tqMetaOpen(STQ* pTq) {
   if (tdbOpen(pTq->path, 16 * 1024, 1, &pTq->pMetaDB, 0) < 0) {
-    ASSERT(0);
     return -1;
   }
 
   if (tdbTbOpen("tq.db", -1, -1, NULL, pTq->pMetaDB, &pTq->pExecStore, 0) < 0) {
-    ASSERT(0);
     return -1;
   }
 
   if (tdbTbOpen("tq.check.db", -1, -1, NULL, pTq->pMetaDB, &pTq->pCheckStore, 0) < 0) {
-    ASSERT(0);
     return -1;
   }
 
@@ -197,40 +194,49 @@ int32_t tqMetaSaveHandle(STQ* pTq, const char* key, const STqHandle* pHandle) {
   int32_t code;
   int32_t vlen;
   tEncodeSize(tEncodeSTqHandle, pHandle, vlen, code);
-  ASSERT(code == 0);
 
   tqDebug("tq save %s(%d) consumer %" PRId64 " vgId:%d", pHandle->subKey, (int32_t)strlen(pHandle->subKey),
           pHandle->consumerId, TD_VID(pTq->pVnode));
 
   void* buf = taosMemoryCalloc(1, vlen);
   if (buf == NULL) {
-    ASSERT(0);
+    return -1;
   }
 
   SEncoder encoder;
   tEncoderInit(&encoder, buf, vlen);
 
   if (tEncodeSTqHandle(&encoder, pHandle) < 0) {
-    ASSERT(0);
+    tEncoderClear(&encoder);
+    taosMemoryFree(buf);
+    return -1;
   }
 
   TXN* txn;
 
   if (tdbBegin(pTq->pMetaDB, &txn, tdbDefaultMalloc, tdbDefaultFree, NULL, TDB_TXN_WRITE | TDB_TXN_READ_UNCOMMITTED) <
       0) {
-    ASSERT(0);
+    tEncoderClear(&encoder);
+    taosMemoryFree(buf);
+    return -1;
   }
 
   if (tdbTbUpsert(pTq->pExecStore, key, (int)strlen(key), buf, vlen, txn) < 0) {
-    ASSERT(0);
+    tEncoderClear(&encoder);
+    taosMemoryFree(buf);
+    return -1;
   }
 
   if (tdbCommit(pTq->pMetaDB, txn) < 0) {
-    ASSERT(0);
+    tEncoderClear(&encoder);
+    taosMemoryFree(buf);
+    return -1;
   }
 
   if (tdbPostCommit(pTq->pMetaDB, txn) < 0) {
-    ASSERT(0);
+    tEncoderClear(&encoder);
+    taosMemoryFree(buf);
+    return -1;
   }
 
   tEncoderClear(&encoder);

source/dnode/vnode/src/tq/tqOffset.c

@@ -101,7 +101,8 @@ STqOffsetStore* tqOffsetOpen(STQ* pTq) {
   }
   char* fname = tqOffsetBuildFName(pStore->pTq->path, 0);
   if (tqOffsetRestoreFromFile(pStore, fname) < 0) {
-    ASSERT(0);
+    taosMemoryFree(fname);
+    return NULL;
   }
   taosMemoryFree(fname);
   return pStore;

source/dnode/vnode/src/tq/tqPush.c

@@ -307,7 +307,7 @@ int tqPushMsg(STQ* pTq, void* msg, int32_t msgLen, tmsg_t msgType, int64_t ver)
     taosWUnLockLatch(&pTq->pushLock);
   }
 
-  if (vnodeIsRoleLeader(pTq->pVnode)) {
+  if (!tsDisableStream && vnodeIsRoleLeader(pTq->pVnode)) {
     if (taosHashGetSize(pTq->pStreamMeta->pTasks) == 0) return 0;
     if (msgType == TDMT_VND_SUBMIT) {
       void* data = taosMemoryMalloc(msgLen);

source/dnode/vnode/src/tsdb/tsdbCache.c

@@ -15,17 +15,51 @@
 
 #include "tsdb.h"
 
+static int32_t tsdbOpenBICache(STsdb *pTsdb) {
+  int32_t    code = 0;
+  SLRUCache *pCache = taosLRUCacheInit(10 * 1024 * 1024, 0, .5);
+  if (pCache == NULL) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto _err;
+  }
+
+  taosLRUCacheSetStrictCapacity(pCache, false);
+
+  taosThreadMutexInit(&pTsdb->biMutex, NULL);
+
+_err:
+  pTsdb->biCache = pCache;
+  return code;
+}
+
+static void tsdbCloseBICache(STsdb *pTsdb) {
+  SLRUCache *pCache = pTsdb->biCache;
+  if (pCache) {
+    taosLRUCacheEraseUnrefEntries(pCache);
+
+    taosLRUCacheCleanup(pCache);
+
+    taosThreadMutexDestroy(&pTsdb->biMutex);
+  }
+}
+
 int32_t tsdbOpenCache(STsdb *pTsdb) {
   int32_t    code = 0;
   SLRUCache *pCache = NULL;
   size_t     cfgCapacity = pTsdb->pVnode->config.cacheLastSize * 1024 * 1024;
 
-  pCache = taosLRUCacheInit(cfgCapacity, -1, .5);
+  pCache = taosLRUCacheInit(cfgCapacity, 1, .5);
   if (pCache == NULL) {
     code = TSDB_CODE_OUT_OF_MEMORY;
     goto _err;
   }
 
+  code = tsdbOpenBICache(pTsdb);
+  if (code != TSDB_CODE_SUCCESS) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto _err;
+  }
+
   taosLRUCacheSetStrictCapacity(pCache, false);
 
   taosThreadMutexInit(&pTsdb->lruMutex, NULL);
@@ -44,6 +78,8 @@ void tsdbCloseCache(STsdb *pTsdb) {
 
     taosThreadMutexDestroy(&pTsdb->lruMutex);
   }
+
+  tsdbCloseBICache(pTsdb);
 }
 
 static void getTableCacheKey(tb_uid_t uid, int cacheType, char *key, int *len) {
@@ -208,6 +244,11 @@ int32_t tsdbCacheInsertLastrow(SLRUCache *pCache, STsdb *pTsdb, tb_uid_t uid, ST
     int16_t nCol = taosArrayGetSize(pLast);
     int16_t iCol = 0;
 
+    if (nCol != pTSchema->numOfCols) {
+      invalidate = true;
+      goto _invalidate;
+    }
+
     SLastCol *tTsVal = (SLastCol *)taosArrayGet(pLast, iCol);
     if (keyTs > tTsVal->ts) {
       STColumn *pTColumn = &pTSchema->columns[0];
@@ -223,6 +264,12 @@ int32_t tsdbCacheInsertLastrow(SLRUCache *pCache, STsdb *pTsdb, tb_uid_t uid, ST
 
         SColVal colVal = {0};
         tTSRowGetVal(row, pTSchema, iCol, &colVal);
+
+        if (colVal.cid != tColVal->cid) {
+          invalidate = true;
+          goto _invalidate;
+        }
+
         if (!COL_VAL_IS_NONE(&colVal)) {
           if (keyTs == tTsVal1->ts && !COL_VAL_IS_NONE(tColVal)) {
             invalidate = true;
@@ -232,7 +279,8 @@ int32_t tsdbCacheInsertLastrow(SLRUCache *pCache, STsdb *pTsdb, tb_uid_t uid, ST
             SLastCol lastCol = {.ts = keyTs, .colVal = colVal};
             if (IS_VAR_DATA_TYPE(colVal.type) && colVal.value.nData > 0) {
               SLastCol *pLastCol = (SLastCol *)taosArrayGet(pLast, iCol);
-              taosMemoryFree(pLastCol->colVal.value.pData);
+              if (pLastCol->colVal.value.nData > 0 && NULL != pLastCol->colVal.value.pData)
+                taosMemoryFree(pLastCol->colVal.value.pData);
 
               lastCol.colVal.value.pData = taosMemoryMalloc(colVal.value.nData);
               if (lastCol.colVal.value.pData == NULL) {
@@ -279,6 +327,11 @@ int32_t tsdbCacheInsertLast(SLRUCache *pCache, tb_uid_t uid, STSRow *row, STsdb
     int16_t nCol = taosArrayGetSize(pLast);
     int16_t iCol = 0;
 
+    if (nCol != pTSchema->numOfCols) {
+      invalidate = true;
+      goto _invalidate;
+    }
+
     SLastCol *tTsVal = (SLastCol *)taosArrayGet(pLast, iCol);
     if (keyTs > tTsVal->ts) {
       STColumn *pTColumn = &pTSchema->columns[0];
@@ -294,6 +347,12 @@ int32_t tsdbCacheInsertLast(SLRUCache *pCache, tb_uid_t uid, STSRow *row, STsdb
 
         SColVal colVal = {0};
         tTSRowGetVal(row, pTSchema, iCol, &colVal);
+
+        if (colVal.cid != tColVal->cid) {
+          invalidate = true;
+          goto _invalidate;
+        }
+
         if (COL_VAL_IS_VALUE(&colVal)) {
           if (keyTs == tTsVal1->ts && COL_VAL_IS_VALUE(tColVal)) {
             invalidate = true;
@@ -303,7 +362,8 @@ int32_t tsdbCacheInsertLast(SLRUCache *pCache, tb_uid_t uid, STSRow *row, STsdb
             SLastCol lastCol = {.ts = keyTs, .colVal = colVal};
             if (IS_VAR_DATA_TYPE(colVal.type) && colVal.value.nData > 0) {
               SLastCol *pLastCol = (SLastCol *)taosArrayGet(pLast, iCol);
-              taosMemoryFree(pLastCol->colVal.value.pData);
+              if (pLastCol->colVal.value.nData > 0 && NULL != pLastCol->colVal.value.pData)
+                taosMemoryFree(pLastCol->colVal.value.pData);
 
               lastCol.colVal.value.pData = taosMemoryMalloc(colVal.value.nData);
               if (lastCol.colVal.value.pData == NULL) {
@@ -584,6 +644,7 @@ typedef struct SFSNextRowIter {
   SArray            *aDFileSet;
   SDataFReader     **pDataFReader;
   SArray            *aBlockIdx;
+  LRUHandle         *aBlockIdxHandle;
   SBlockIdx         *pBlockIdx;
   SMapData           blockMap;
   int32_t            nBlock;
@@ -637,6 +698,7 @@ static int32_t getNextRowFromFS(void *iter, TSDBROW **ppRow) {
       }
 
       // tMapDataReset(&state->blockIdxMap);
+      /*
       if (!state->aBlockIdx) {
         state->aBlockIdx = taosArrayInit(0, sizeof(SBlockIdx));
       } else {
@@ -644,6 +706,13 @@ static int32_t getNextRowFromFS(void *iter, TSDBROW **ppRow) {
       }
       code = tsdbReadBlockIdx(*state->pDataFReader, state->aBlockIdx);
       if (code) goto _err;
+      */
+      int32_t code =
+          tsdbCacheGetBlockIdx((*state->pDataFReader)->pTsdb->biCache, *state->pDataFReader, &state->aBlockIdxHandle);
+      if (code != TSDB_CODE_SUCCESS || state->aBlockIdxHandle == NULL) {
+        goto _err;
+      }
+      state->aBlockIdx = (SArray *)taosLRUCacheValue((*state->pDataFReader)->pTsdb->biCache, state->aBlockIdxHandle);
 
       /* if (state->pBlockIdx) { */
       /* } */
@@ -712,7 +781,10 @@ static int32_t getNextRowFromFS(void *iter, TSDBROW **ppRow) {
             // resetLastBlockLoadInfo(state->pLoadInfo);
 
             if (state->aBlockIdx) {
-              taosArrayDestroy(state->aBlockIdx);
+              // taosArrayDestroy(state->aBlockIdx);
+              tsdbBICacheRelease((*state->pDataFReader)->pTsdb->biCache, state->aBlockIdxHandle);
+
+              state->aBlockIdxHandle = NULL;
               state->aBlockIdx = NULL;
             }
 
@@ -735,7 +807,10 @@ _err:
     resetLastBlockLoadInfo(state->pLoadInfo);
     }*/
   if (state->aBlockIdx) {
-    taosArrayDestroy(state->aBlockIdx);
+    // taosArrayDestroy(state->aBlockIdx);
+    tsdbBICacheRelease((*state->pDataFReader)->pTsdb->biCache, state->aBlockIdxHandle);
+
+    state->aBlockIdxHandle = NULL;
     state->aBlockIdx = NULL;
   }
   if (state->pBlockData) {
@@ -761,7 +836,10 @@ int32_t clearNextRowFromFS(void *iter) {
     state->pDataFReader = NULL;
     }*/
   if (state->aBlockIdx) {
-    taosArrayDestroy(state->aBlockIdx);
+    // taosArrayDestroy(state->aBlockIdx);
+    tsdbBICacheRelease((*state->pDataFReader)->pTsdb->biCache, state->aBlockIdxHandle);
+
+    state->aBlockIdxHandle = NULL;
     state->aBlockIdx = NULL;
   }
   if (state->pBlockData) {
@@ -1475,3 +1553,84 @@ size_t tsdbCacheGetUsage(SVnode *pVnode) {
 
   return usage;
 }
+
+static void getBICacheKey(int32_t fid, int64_t commitID, char *key, int *len) {
+  struct {
+    int32_t fid;
+    int64_t commitID;
+  } biKey = {0};
+
+  biKey.fid = fid;
+  biKey.commitID = commitID;
+
+  *len = sizeof(biKey);
+  memcpy(key, &biKey, *len);
+}
+
+static int32_t tsdbCacheLoadBlockIdx(SDataFReader *pFileReader, SArray **aBlockIdx) {
+  SArray *pArray = taosArrayInit(8, sizeof(SBlockIdx));
+  int32_t code = tsdbReadBlockIdx(pFileReader, pArray);
+
+  if (code != TSDB_CODE_SUCCESS) {
+    taosArrayDestroy(pArray);
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    return code;
+  }
+
+  *aBlockIdx = pArray;
+
+  return code;
+}
+
+static void deleteBICache(const void *key, size_t keyLen, void *value) {
+  SArray *pArray = (SArray *)value;
+
+  taosArrayDestroy(pArray);
+}
+
+int32_t tsdbCacheGetBlockIdx(SLRUCache *pCache, SDataFReader *pFileReader, LRUHandle **handle) {
+  int32_t code = 0;
+  char    key[128] = {0};
+  int     keyLen = 0;
+
+  getBICacheKey(pFileReader->pSet->fid, pFileReader->pSet->pHeadF->commitID, key, &keyLen);
+  LRUHandle *h = taosLRUCacheLookup(pCache, key, keyLen);
+  if (!h) {
+    STsdb *pTsdb = pFileReader->pTsdb;
+    taosThreadMutexLock(&pTsdb->biMutex);
+
+    h = taosLRUCacheLookup(pCache, key, keyLen);
+    if (!h) {
+      SArray *pArray = NULL;
+      code = tsdbCacheLoadBlockIdx(pFileReader, &pArray);
+      //  if table's empty or error, return code of -1
+      if (code != TSDB_CODE_SUCCESS || pArray == NULL) {
+        taosThreadMutexUnlock(&pTsdb->biMutex);
+
+        *handle = NULL;
+        return 0;
+      }
+
+      size_t              charge = pArray->capacity * pArray->elemSize + sizeof(*pArray);
+      _taos_lru_deleter_t deleter = deleteBICache;
+      LRUStatus status = taosLRUCacheInsert(pCache, key, keyLen, pArray, charge, deleter, &h, TAOS_LRU_PRIORITY_LOW);
+      if (status != TAOS_LRU_STATUS_OK) {
+        code = -1;
+      }
+    }
+
+    taosThreadMutexUnlock(&pTsdb->biMutex);
+  }
+
+  *handle = h;
+
+  return code;
+}
+
+int32_t tsdbBICacheRelease(SLRUCache *pCache, LRUHandle *h) {
+  int32_t code = 0;
+
+  taosLRUCacheRelease(pCache, h, false);
+
+  return code;
+}

source/dnode/vnode/src/tsdb/tsdbCacheRead.c

@@ -41,6 +41,13 @@ static int32_t saveOneRow(SArray* pRow, SSDataBlock* pBlock, SCacheRowsReader* p
         int32_t   slotId = slotIds[i];
         SLastCol* pColVal = (SLastCol*)taosArrayGet(pRow, slotId);
 
+        // add check for null value, caused by the modification of table schema (new column added).
+        if (pColVal == NULL) {
+          p->ts = 0;
+          p->isNull = true;
+          continue;
+        }
+
         p->ts = pColVal->ts;
         p->isNull = !COL_VAL_IS_VALUE(&pColVal->colVal);
         allNullRow = p->isNull & allNullRow;
@@ -99,6 +106,38 @@ static int32_t saveOneRow(SArray* pRow, SSDataBlock* pBlock, SCacheRowsReader* p
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t setTableSchema(SCacheRowsReader* p, uint64_t suid, const char* idstr) {
+  int32_t numOfTables = p->numOfTables;
+
+  if (suid != 0) {
+    p->pSchema = metaGetTbTSchema(p->pVnode->pMeta, suid, -1, 1);
+    if (p->pSchema == NULL) {
+      taosMemoryFree(p);
+      tsdbWarn("stable:%" PRIu64 " has been dropped, failed to retrieve cached rows, %s", suid, idstr);
+      return TSDB_CODE_PAR_TABLE_NOT_EXIST;
+    }
+  } else {
+    for (int32_t i = 0; i < numOfTables; ++i) {
+      uint64_t uid = p->pTableList[i].uid;
+      p->pSchema = metaGetTbTSchema(p->pVnode->pMeta, uid, -1, 1);
+      if (p->pSchema != NULL) {
+        break;
+      }
+
+      tsdbWarn("table:%" PRIu64 " has been dropped, failed to retrieve cached rows, %s", uid, idstr);
+    }
+
+    // all queried tables have been dropped already, return immediately.
+    if (p->pSchema == NULL) {
+      taosMemoryFree(p);
+      tsdbWarn("all queried tables has been dropped, try next group, %s", idstr);
+      return TSDB_CODE_PAR_TABLE_NOT_EXIST;
+    }
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+
 int32_t tsdbCacherowsReaderOpen(void* pVnode, int32_t type, void* pTableIdList, int32_t numOfTables, int32_t numOfCols,
                                 uint64_t suid, void** pReader, const char* idstr) {
   *pReader = NULL;
@@ -117,11 +156,15 @@ int32_t tsdbCacherowsReaderOpen(void* pVnode, int32_t type, void* pTableIdList,
     return TSDB_CODE_SUCCESS;
   }
 
-  STableKeyInfo* pKeyInfo = &((STableKeyInfo*)pTableIdList)[0];
-  p->pSchema = metaGetTbTSchema(p->pVnode->pMeta, pKeyInfo->uid, -1, 1);
   p->pTableList = pTableIdList;
   p->numOfTables = numOfTables;
 
+  int32_t code = setTableSchema(p, suid, idstr);
+  if (code != TSDB_CODE_SUCCESS) {
+    tsdbCacherowsReaderClose(p);
+    return code;
+  }
+
   p->transferBuf = taosMemoryCalloc(p->pSchema->numOfCols, POINTER_BYTES);
   if (p->transferBuf == NULL) {
     tsdbCacherowsReaderClose(p);
@@ -138,7 +181,8 @@ int32_t tsdbCacherowsReaderOpen(void* pVnode, int32_t type, void* pTableIdList,
     }
   }
 
-  p->pLoadInfo = tCreateLastBlockLoadInfo(p->pSchema, NULL, 0);
+  int32_t numOfStt = ((SVnode*)pVnode)->config.sttTrigger;
+  p->pLoadInfo = tCreateLastBlockLoadInfo(p->pSchema, NULL, 0, numOfStt);
   if (p->pLoadInfo == NULL) {
     tsdbCacherowsReaderClose(p);
     return TSDB_CODE_OUT_OF_MEMORY;

source/dnode/vnode/src/tsdb/tsdbFile.c

@@ -92,31 +92,63 @@ static int32_t tGetSmaFile(uint8_t *p, SSmaFile *pSmaFile) {
 }
 
 // EXPOSED APIS ==================================================
+static char* getFileNamePrefix(STsdb *pTsdb, SDiskID did, int32_t fid, uint64_t commitId, char fname[]) {
+  const char* p1 = tfsGetDiskPath(pTsdb->pVnode->pTfs, did);
+  int32_t len = strlen(p1);
+
+  char* p = memcpy(fname, p1, len);
+  p += len;
+
+  *(p++) = TD_DIRSEP[0];
+  len = strlen(pTsdb->path);
+
+  memcpy(p, pTsdb->path, len);
+  p += len;
+
+  *(p++) = TD_DIRSEP[0];
+  *(p++) = 'v';
+
+  p += titoa(TD_VID(pTsdb->pVnode), 10, p);
+  *(p++) = 'f';
+
+  p += titoa(fid, 10, p);
+
+  memcpy(p, "ver", 3);
+  p += 3;
+
+  p += titoa(commitId, 10, p);
+  return p;
+}
+
 void tsdbHeadFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SHeadFile *pHeadF, char fname[]) {
-  snprintf(fname, TSDB_FILENAME_LEN - 1, "%s%s%s%sv%df%dver%" PRId64 "%s", tfsGetDiskPath(pTsdb->pVnode->pTfs, did),
+  char* p = getFileNamePrefix(pTsdb, did, fid, pHeadF->commitID, fname);
-           TD_DIRSEP, pTsdb->path, TD_DIRSEP, TD_VID(pTsdb->pVnode), fid, pHeadF->commitID, ".head");
+  memcpy(p, ".head", 5);
+  p[5] = 0;
 }
 
 void tsdbDataFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SDataFile *pDataF, char fname[]) {
-  snprintf(fname, TSDB_FILENAME_LEN - 1, "%s%s%s%sv%df%dver%" PRId64 "%s", tfsGetDiskPath(pTsdb->pVnode->pTfs, did),
+  char* p = getFileNamePrefix(pTsdb, did, fid, pDataF->commitID, fname);
-           TD_DIRSEP, pTsdb->path, TD_DIRSEP, TD_VID(pTsdb->pVnode), fid, pDataF->commitID, ".data");
+  memcpy(p, ".data", 5);
+  p[5] = 0;
 }
 
 void tsdbSttFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SSttFile *pSttF, char fname[]) {
-  snprintf(fname, TSDB_FILENAME_LEN - 1, "%s%s%s%sv%df%dver%" PRId64 "%s", tfsGetDiskPath(pTsdb->pVnode->pTfs, did),
+  char* p = getFileNamePrefix(pTsdb, did, fid, pSttF->commitID, fname);
-           TD_DIRSEP, pTsdb->path, TD_DIRSEP, TD_VID(pTsdb->pVnode), fid, pSttF->commitID, ".stt");
+  memcpy(p, ".stt", 4);
+  p[4] = 0;
 }
 
 void tsdbSmaFileName(STsdb *pTsdb, SDiskID did, int32_t fid, SSmaFile *pSmaF, char fname[]) {
-  snprintf(fname, TSDB_FILENAME_LEN - 1, "%s%s%s%sv%df%dver%" PRId64 "%s", tfsGetDiskPath(pTsdb->pVnode->pTfs, did),
+  char* p = getFileNamePrefix(pTsdb, did, fid, pSmaF->commitID, fname);
-           TD_DIRSEP, pTsdb->path, TD_DIRSEP, TD_VID(pTsdb->pVnode), fid, pSmaF->commitID, ".sma");
+  memcpy(p, ".sma", 4);
+  p[4] = 0;
 }
 
 bool tsdbDelFileIsSame(SDelFile *pDelFile1, SDelFile *pDelFile2) { return pDelFile1->commitID == pDelFile2->commitID; }
 
 int32_t tsdbDFileRollback(STsdb *pTsdb, SDFileSet *pSet, EDataFileT ftype) {
   int32_t   code = 0;
-  int64_t   size;
+  int64_t   size = 0;
   int64_t   n;
   TdFilePtr pFD;
   char      fname[TSDB_FILENAME_LEN];
@@ -135,7 +167,7 @@ int32_t tsdbDFileRollback(STsdb *pTsdb, SDFileSet *pSet, EDataFileT ftype) {
       tPutSmaFile(hdr, pSet->pSmaF);
       break;
     default:
-      ASSERT(0);
+      goto _err;  // make the coverity scan happy
   }
 
   taosCalcChecksumAppend(0, hdr, TSDB_FHDR_SIZE);

source/dnode/vnode/src/tsdb/tsdbMergeTree.c

@@ -31,14 +31,16 @@ struct SLDataIter {
   SSttBlockLoadInfo *pBlockLoadInfo;
 };
 
-SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols) {
+SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols, int32_t numOfSttTrigger) {
-  SSttBlockLoadInfo *pLoadInfo = taosMemoryCalloc(TSDB_MAX_STT_TRIGGER, sizeof(SSttBlockLoadInfo));
+  SSttBlockLoadInfo *pLoadInfo = taosMemoryCalloc(numOfSttTrigger, sizeof(SSttBlockLoadInfo));
   if (pLoadInfo == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return NULL;
   }
 
-  for (int32_t i = 0; i < TSDB_MAX_STT_TRIGGER; ++i) {
+  pLoadInfo->numOfStt = numOfSttTrigger;
+
+  for (int32_t i = 0; i < numOfSttTrigger; ++i) {
     pLoadInfo[i].blockIndex[0] = -1;
     pLoadInfo[i].blockIndex[1] = -1;
     pLoadInfo[i].currentLoadBlockIndex = 1;
@@ -63,7 +65,7 @@ SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList,
 }
 
 void resetLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo) {
-  for (int32_t i = 0; i < TSDB_MAX_STT_TRIGGER; ++i) {
+  for (int32_t i = 0; i < pLoadInfo->numOfStt; ++i) {
     pLoadInfo[i].currentLoadBlockIndex = 1;
     pLoadInfo[i].blockIndex[0] = -1;
     pLoadInfo[i].blockIndex[1] = -1;
@@ -77,14 +79,14 @@ void resetLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo) {
 }
 
 void getLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo, int64_t *blocks, double *el) {
-  for (int32_t i = 0; i < TSDB_MAX_STT_TRIGGER; ++i) {
+  for (int32_t i = 0; i < pLoadInfo->numOfStt; ++i) {
     *el += pLoadInfo[i].elapsedTime;
     *blocks += pLoadInfo[i].loadBlocks;
   }
 }
 
 void *destroyLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo) {
-  for (int32_t i = 0; i < TSDB_MAX_STT_TRIGGER; ++i) {
+  for (int32_t i = 0; i < pLoadInfo->numOfStt; ++i) {
     pLoadInfo[i].currentLoadBlockIndex = 1;
     pLoadInfo[i].blockIndex[0] = -1;
     pLoadInfo[i].blockIndex[1] = -1;
@@ -515,7 +517,7 @@ bool tLDataIterNextRow(SLDataIter *pIter, const char *idStr) {
   pIter->rInfo.row = tsdbRowFromBlockData(pBlockData, pIter->iRow);
 
 _exit:
-  return (terrno == TSDB_CODE_SUCCESS) && (pIter->pSttBlk != NULL);
+  return (terrno == TSDB_CODE_SUCCESS) && (pIter->pSttBlk != NULL) && (pBlockData != NULL);
 }
 
 SRowInfo *tLDataIterGet(SLDataIter *pIter) { return &pIter->rInfo; }

source/dnode/vnode/src/tsdb/tsdbReaderWriter.c

@@ -47,15 +47,21 @@ static int32_t tsdbOpenFile(const char *path, int32_t szPage, int32_t flag, STsd
     taosMemoryFree(pFD);
     goto _exit;
   }
-  if (taosStatFile(path, &pFD->szFile, NULL) < 0) {
+
-    code = TAOS_SYSTEM_ERROR(errno);
+  // not check file size when reading data files.
-    taosMemoryFree(pFD->pBuf);
+  if (flag != TD_FILE_READ) {
-    taosCloseFile(&pFD->pFD);
+    if (taosStatFile(path, &pFD->szFile, NULL) < 0) {
-    taosMemoryFree(pFD);
+      code = TAOS_SYSTEM_ERROR(errno);
-    goto _exit;
+      taosMemoryFree(pFD->pBuf);
+      taosCloseFile(&pFD->pFD);
+      taosMemoryFree(pFD);
+      goto _exit;
+    }
+
+    ASSERT(pFD->szFile % szPage == 0);
+    pFD->szFile = pFD->szFile / szPage;
   }
-  ASSERT(pFD->szFile % szPage == 0);
+
-  pFD->szFile = pFD->szFile / szPage;
   *ppFD = pFD;
 
 _exit:
@@ -103,7 +109,7 @@ _exit:
 static int32_t tsdbReadFilePage(STsdbFD *pFD, int64_t pgno) {
   int32_t code = 0;
 
-  ASSERT(pgno <= pFD->szFile);
+  // ASSERT(pgno <= pFD->szFile);
 
   // seek
   int64_t offset = PAGE_OFFSET(pgno, pFD->szPage);
@@ -175,7 +181,7 @@ static int32_t tsdbReadFile(STsdbFD *pFD, int64_t offset, uint8_t *pBuf, int64_t
   int32_t szPgCont = PAGE_CONTENT_SIZE(pFD->szPage);
   int64_t bOffset = fOffset % pFD->szPage;
 
-  ASSERT(pgno && pgno <= pFD->szFile);
+  // ASSERT(pgno && pgno <= pFD->szFile);
   ASSERT(bOffset < szPgCont);
 
   while (n < size) {

source/dnode/vnode/src/tsdb/tsdbUtil.c

@@ -929,8 +929,9 @@ int32_t tRowMergerGetRow(SRowMerger *pMerger, STSRow **ppRow) {
   return code;
 }
 
+/*
 // delete skyline ======================================================
-static int32_t tsdbMergeSkyline(SArray *aSkyline1, SArray *aSkyline2, SArray *aSkyline) {
+static int32_t tsdbMergeSkyline2(SArray *aSkyline1, SArray *aSkyline2, SArray *aSkyline) {
   int32_t  code = 0;
   int32_t  i1 = 0;
   int32_t  n1 = taosArrayGetSize(aSkyline1);
@@ -996,7 +997,139 @@ static int32_t tsdbMergeSkyline(SArray *aSkyline1, SArray *aSkyline2, SArray *aS
 _exit:
   return code;
 }
+*/
+
+// delete skyline ======================================================
+static int32_t tsdbMergeSkyline(SArray *pSkyline1, SArray *pSkyline2, SArray *pSkyline) {
+  int32_t  code = 0;
+  int32_t  i1 = 0;
+  int32_t  n1 = taosArrayGetSize(pSkyline1);
+  int32_t  i2 = 0;
+  int32_t  n2 = taosArrayGetSize(pSkyline2);
+  TSDBKEY *pKey1;
+  TSDBKEY *pKey2;
+  int64_t  version1 = 0;
+  int64_t  version2 = 0;
+
+  ASSERT(n1 > 0 && n2 > 0);
+
+  taosArrayClear(pSkyline);
+  TSDBKEY **pItem = TARRAY_GET_ELEM(pSkyline, 0);
+
+  while (i1 < n1 && i2 < n2) {
+    pKey1 = (TSDBKEY *)taosArrayGetP(pSkyline1, i1);
+    pKey2 = (TSDBKEY *)taosArrayGetP(pSkyline2, i2);
+
+    if (pKey1->ts < pKey2->ts) {
+      version1 = pKey1->version;
+      *pItem = pKey1;
+      i1++;
+    } else if (pKey1->ts > pKey2->ts) {
+      version2 = pKey2->version;
+      *pItem = pKey2;
+      i2++;
+    } else {
+      version1 = pKey1->version;
+      version2 = pKey2->version;
+      *pItem = pKey1;
+      i1++;
+      i2++;
+    }
+
+    (*pItem)->version = TMAX(version1, version2);
+    pItem++;
+  }
+
+  while (i1 < n1) {
+    pKey1 = (TSDBKEY *)taosArrayGetP(pSkyline1, i1);
+    *pItem = pKey1;
+    pItem++;
+    i1++;
+  }
+
+  while (i2 < n2) {
+    pKey2 = (TSDBKEY *)taosArrayGetP(pSkyline2, i2);
+    *pItem = pKey2;
+    pItem++;
+    i2++;
+  }
+
+  pSkyline->size = TARRAY_ELEM_IDX(pSkyline, pItem);
+  return code;
+}
+
+
+int32_t tsdbBuildDeleteSkylineImpl(SArray *aSkyline, int32_t sidx, int32_t eidx, SArray *pSkyline) {
+  int32_t   code = 0;
+  SDelData *pDelData;
+  int32_t   midx;
+
+  taosArrayClear(pSkyline);
+  if (sidx == eidx) {    
+    TSDBKEY *pItem1 = taosArrayGet(aSkyline, sidx * 2);
+    TSDBKEY *pItem2 = taosArrayGet(aSkyline, sidx * 2 + 1);
+    taosArrayPush(pSkyline, &pItem1);
+    taosArrayPush(pSkyline, &pItem2);
+  } else {
+    SArray *pSkyline1 = NULL;
+    SArray *pSkyline2 = NULL;
+    midx = (sidx + eidx) / 2;
+
+    pSkyline1 = taosArrayInit((midx - sidx + 1) * 2, POINTER_BYTES);
+    pSkyline2 = taosArrayInit((eidx - midx) * 2, POINTER_BYTES);
+    if (pSkyline1 == NULL || pSkyline1 == NULL) {
+      code = TSDB_CODE_OUT_OF_MEMORY;
+      goto _clear;
+    }
+
+    code = tsdbBuildDeleteSkylineImpl(aSkyline, sidx, midx, pSkyline1);
+    if (code) goto _clear;
+
+    code = tsdbBuildDeleteSkylineImpl(aSkyline, midx + 1, eidx, pSkyline2);
+    if (code) goto _clear;
+
+    code = tsdbMergeSkyline(pSkyline1, pSkyline2, pSkyline);
+
+  _clear:
+    taosArrayDestroy(pSkyline1);
+    taosArrayDestroy(pSkyline2);
+  }
+
+  return code;
+}
+
+
 int32_t tsdbBuildDeleteSkyline(SArray *aDelData, int32_t sidx, int32_t eidx, SArray *aSkyline) {
+  SDelData *pDelData;
+  int32_t code = 0;
+  int32_t dataNum = eidx - sidx + 1;
+  SArray *aTmpSkyline = taosArrayInit(dataNum * 2, sizeof(TSDBKEY));
+  SArray *pSkyline = taosArrayInit(dataNum * 2, POINTER_BYTES);
+  
+  for (int32_t i = sidx; i <= eidx; ++i) {
+    pDelData = (SDelData *)taosArrayGet(aDelData, i);
+    taosArrayPush(aTmpSkyline, &(TSDBKEY){.ts = pDelData->sKey, .version = pDelData->version});
+    taosArrayPush(aTmpSkyline, &(TSDBKEY){.ts = pDelData->eKey, .version = 0});
+  }
+
+  code = tsdbBuildDeleteSkylineImpl(aTmpSkyline, sidx, eidx, pSkyline);
+  if (code) goto _clear;
+
+  int32_t skylineNum = taosArrayGetSize(pSkyline);
+  for (int32_t i = 0; i < skylineNum; ++i) {
+     TSDBKEY *p = taosArrayGetP(pSkyline, i);
+     taosArrayPush(aSkyline, p);
+  }
+
+_clear:
+  taosArrayDestroy(aTmpSkyline);
+  taosArrayDestroy(pSkyline);
+
+  return code;
+}
+
+/*
+int32_t tsdbBuildDeleteSkyline2(SArray *aDelData, int32_t sidx, int32_t eidx, SArray *aSkyline) {
   int32_t   code = 0;
   SDelData *pDelData;
   int32_t   midx;
@@ -1033,6 +1166,7 @@ int32_t tsdbBuildDeleteSkyline(SArray *aDelData, int32_t sidx, int32_t eidx, SAr
 
   return code;
 }
+*/
 
 // SBlockData ======================================================
 int32_t tBlockDataCreate(SBlockData *pBlockData) {

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

@@ -21,12 +21,34 @@
 static int vnodeEncodeInfo(const SVnodeInfo *pInfo, char **ppData);
 static int vnodeCommitImpl(SCommitInfo *pInfo);
 
+#define WAIT_TIME_MILI_SEC 50
+
 int vnodeBegin(SVnode *pVnode) {
   // alloc buffer pool
+  int32_t nTry = 0;
+
   taosThreadMutexLock(&pVnode->mutex);
 
   while (pVnode->pPool == NULL) {
-    taosThreadCondWait(&pVnode->poolNotEmpty, &pVnode->mutex);
+    vInfo("vgId:%d no free buffer pool on %d try, wait %d ms...", TD_VID(pVnode), ++nTry, WAIT_TIME_MILI_SEC);
+
+    struct timeval  tv;
+    struct timespec ts;
+    taosGetTimeOfDay(&tv);
+    ts.tv_nsec = tv.tv_usec * 1000 + WAIT_TIME_MILI_SEC * 1000000;
+    if (ts.tv_nsec > 999999999l) {
+      ts.tv_sec = tv.tv_sec + 1;
+      ts.tv_nsec -= 1000000000l;
+    } else {
+      ts.tv_sec = tv.tv_sec;
+    }
+
+    int32_t rc = taosThreadCondTimedWait(&pVnode->poolNotEmpty, &pVnode->mutex, &ts);
+    if (rc && rc != ETIMEDOUT) {
+      terrno = TAOS_SYSTEM_ERROR(rc);
+      taosThreadMutexUnlock(&pVnode->mutex);
+      return -1;
+    }
   }
 
   pVnode->inUse = pVnode->pPool;
@@ -70,7 +92,7 @@ int vnodeShouldCommit(SVnode *pVnode) {
   }
 
   SVCommitSched *pSched = &pVnode->commitSched;
-  int64_t nowMs = taosGetMonoTimestampMs();
+  int64_t        nowMs = taosGetMonoTimestampMs();
 
   return (((pVnode->inUse->size > pVnode->inUse->node.size) && (pSched->commitMs + SYNC_VND_COMMIT_MIN_MS < nowMs)) ||
           (pVnode->inUse->size > 0 && pSched->commitMs + pSched->maxWaitMs < nowMs));

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

@@ -239,7 +239,7 @@ _err:
   if (pVnode->pWal) walClose(pVnode->pWal);
   if (pVnode->pTsdb) tsdbClose(&pVnode->pTsdb);
   if (pVnode->pSma) smaClose(pVnode->pSma);
-  if (pVnode->pMeta) metaClose(pVnode->pMeta);
+  if (pVnode->pMeta) metaClose(&pVnode->pMeta);
   if (pVnode->pPool) vnodeCloseBufPool(pVnode);
 
   tsem_destroy(&(pVnode->canCommit));
@@ -263,7 +263,7 @@ void vnodeClose(SVnode *pVnode) {
     tqClose(pVnode->pTq);
     if (pVnode->pTsdb) tsdbClose(&pVnode->pTsdb);
     smaClose(pVnode->pSma);
-    metaClose(pVnode->pMeta);
+    if (pVnode->pMeta) metaClose(&pVnode->pMeta);
     vnodeCloseBufPool(pVnode);
     tsem_post(&pVnode->canCommit);
 

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

@@ -426,7 +426,13 @@ int32_t vnodeSnapWrite(SVSnapWriter *pWriter, uint8_t *pData, uint32_t nData) {
   SVnode       *pVnode = pWriter->pVnode;
 
   ASSERT(pHdr->size + sizeof(SSnapDataHdr) == nData);
-  ASSERT(pHdr->index == pWriter->index + 1);
+
+  if (pHdr->index != pWriter->index + 1) {
+    vError("vgId:%d, unexpected vnode snapshot msg. index:%" PRId64 ", expected index:%" PRId64, TD_VID(pVnode),
+           pHdr->index, pWriter->index + 1);
+    return -1;
+  }
+
   pWriter->index = pHdr->index;
 
   vDebug("vgId:%d, vnode snapshot write data, index:%" PRId64 " type:%d blockLen:%d", TD_VID(pVnode), pHdr->index,

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

@@ -312,10 +312,13 @@ int32_t vnodeProcessWriteMsg(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
 
   walApplyVer(pVnode->pWal, version);
 
+  /*vInfo("vgId:%d, push msg begin", pVnode->config.vgId);*/
   if (tqPushMsg(pVnode->pTq, pMsg->pCont, pMsg->contLen, pMsg->msgType, version) < 0) {
+    /*vInfo("vgId:%d, push msg end", pVnode->config.vgId);*/
     vError("vgId:%d, failed to push msg to TQ since %s", TD_VID(pVnode), tstrerror(terrno));
     return -1;
   }
+  /*vInfo("vgId:%d, push msg end", pVnode->config.vgId);*/
 
   // commit if need
   if (needCommit) {
@@ -348,7 +351,6 @@ int32_t vnodePreprocessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg) {
 
 int32_t vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg) {
   vTrace("message in vnode query queue is processing");
-  // if ((pMsg->msgType == TDMT_SCH_QUERY) && !vnodeIsLeader(pVnode)) {
   if ((pMsg->msgType == TDMT_SCH_QUERY) && !syncIsReadyForRead(pVnode->sync)) {
     vnodeRedirectRpcMsg(pVnode, pMsg, terrno);
     return 0;
@@ -372,7 +374,6 @@ int32_t vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo) {
   if ((pMsg->msgType == TDMT_SCH_FETCH || pMsg->msgType == TDMT_VND_TABLE_META || pMsg->msgType == TDMT_VND_TABLE_CFG ||
        pMsg->msgType == TDMT_VND_BATCH_META) &&
       !syncIsReadyForRead(pVnode->sync)) {
-    //      !vnodeIsLeader(pVnode)) {
     vnodeRedirectRpcMsg(pVnode, pMsg, terrno);
     return 0;
   }
@@ -1019,7 +1020,7 @@ _exit:
   atomic_add_fetch_64(&pVnode->statis.nBatchInsert, statis.nBatchInsert);
   atomic_add_fetch_64(&pVnode->statis.nBatchInsertSuccess, statis.nBatchInsertSuccess);
 
-  vDebug("vgId:%d, submit success, index:%" PRId64, pVnode->config.vgId, version);
+  vDebug("vgId:%d %s done, index:%" PRId64, TD_VID(pVnode), __func__, version);
   return 0;
 }
 
@@ -1185,7 +1186,7 @@ static int32_t vnodeProcessBatchDeleteReq(SVnode *pVnode, int64_t version, void
   tDecodeSBatchDeleteReq(&decoder, &deleteReq);
 
   SMetaReader mr = {0};
-  metaReaderInit(&mr, pVnode->pMeta, 0);
+  metaReaderInit(&mr, pVnode->pMeta, META_READER_NOLOCK);
 
   int32_t sz = taosArrayGetSize(deleteReq.deleteReqs);
   for (int32_t i = 0; i < sz; i++) {

source/dnode/vnode/test/tsdbSmaTest.cpp

@@ -283,7 +283,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   metaRemoveSmaFromDb(pMeta, indexUid2);
 
   tDestroyTSma(&tSma);
-  metaClose(pMeta);
+  metaClose(&pMeta);
 }
 #endif
 
@@ -577,9 +577,9 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   tDestroyTSma(&tSma);
   tfsClose(pTsdb->pTfs);
   tsdbClose(pTsdb);
-  metaClose(pMeta);
+  metaClose(&pMeta);
 }
 
 #endif
 
-#pragma GCC diagnostic pop
+#pragma GCC diagnostic pop

source/libs/catalog/inc/catalogInt.h

@@ -300,7 +300,7 @@ typedef struct SCtgSubRes {
   ctgSubTaskCbFp fp;
 } SCtgSubRes;
 
-typedef struct SCtgTask {
+struct SCtgTask {
   CTG_TASK_TYPE   type;
   int32_t         taskId;
   SCtgJob*        pJob;
@@ -313,7 +313,7 @@ typedef struct SCtgTask {
   SRWLatch        lock;
   SArray*         pParents;
   SCtgSubRes      subRes;
-} SCtgTask;
+};
 
 typedef struct SCtgTaskReq {
   SCtgTask* pTask;

source/libs/catalog/src/ctgAsync.c

@@ -1712,9 +1712,7 @@ int32_t ctgLaunchGetTbMetasTask(SCtgTask* pTask) {
     return TSDB_CODE_SUCCESS;
   }
 
-  pTask->msgCtxs = taosArrayInit(pCtx->fetchNum, sizeof(SCtgMsgCtx));
+  pTask->msgCtxs = taosArrayInit_s(pCtx->fetchNum, sizeof(SCtgMsgCtx), pCtx->fetchNum);
-  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);
@@ -1849,7 +1847,10 @@ int32_t ctgLaunchGetTbHashsTask(SCtgTask* pTask) {
       ctgAddFetch(&pCtx->pFetchs, i, -1, &fetchIdx, baseResIdx, 0);
 
       baseResIdx += taosArrayGetSize(pReq->pTables);
-      taosArraySetSize(pCtx->pResList, baseResIdx);
+      int32_t inc = baseResIdx - taosArrayGetSize(pCtx->pResList);
+      for(int32_t j = 0; j < inc; ++j) {
+        taosArrayPush(pCtx->pResList, &(SMetaRes){0});
+      }
     }
   }
 
@@ -1861,8 +1862,7 @@ int32_t ctgLaunchGetTbHashsTask(SCtgTask* pTask) {
     return TSDB_CODE_SUCCESS;
   }
 
-  pTask->msgCtxs = taosArrayInit(pCtx->fetchNum, sizeof(SCtgMsgCtx));
+  pTask->msgCtxs = taosArrayInit_s(pCtx->fetchNum, sizeof(SCtgMsgCtx), pCtx->fetchNum);
-  taosArraySetSize(pTask->msgCtxs, pCtx->fetchNum);
 
   for (int32_t i = 0; i < pCtx->fetchNum; ++i) {
     SCtgFetch*  pFetch = taosArrayGet(pCtx->pFetchs, i);

source/libs/catalog/src/ctgCache.c

@@ -2480,20 +2480,20 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
     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);
+      taosArrayPush(ctx->pResList, &(SMetaData){0});
     }
 
     return TSDB_CODE_SUCCESS;
   }
 
   for (int32_t i = 0; i < tbNum; ++i) {
-    SName *pName = taosArrayGet(pList, i);
+    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);
+      taosArrayPush(ctx->pResList, &(SMetaRes){0});
 
       continue;
     }
@@ -2503,7 +2503,7 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
       CTG_UNLOCK(CTG_READ, &pCache->metaLock);
       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);
+      taosArrayPush(ctx->pResList, &(SMetaRes){0});
 
       continue;
     }
@@ -2576,7 +2576,7 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
     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);
+      taosArrayPush(ctx->pResList, &(SMetaRes){0});
 
       taosMemoryFreeClear(pTableMeta);
       continue;
@@ -2588,7 +2588,7 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
       taosHashRelease(dbCache->stbCache, stName);
 
       ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
-      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+      taosArrayPush(ctx->pResList, &(SMetaRes){0});
 
       taosMemoryFreeClear(pTableMeta);
       continue;
@@ -2603,7 +2603,7 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
       taosHashRelease(dbCache->tbCache, pCache);
 
       ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
-      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+      taosArrayPush(ctx->pResList, &(SMetaRes){0});
 
       taosMemoryFreeClear(pTableMeta);
 
@@ -2619,7 +2619,7 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
                nctx.tbInfo.suid);
 
       ctgAddFetch(&ctx->pFetchs, dbIdx, i, fetchIdx, baseResIdx + i, flag);
-      taosArraySetSize(ctx->pResList, taosArrayGetSize(ctx->pResList) + 1);
+      taosArrayPush(ctx->pResList, &(SMetaRes){0});
 
       taosMemoryFreeClear(pTableMeta);
 

source/libs/executor/inc/executil.h

@@ -44,6 +44,8 @@
 typedef struct SGroupResInfo {
   int32_t index;
   SArray* pRows;  // SArray<SResKeyPos>
+  char*   pBuf;
+  bool    freeItem;
 } SGroupResInfo;
 
 typedef struct SResultRow {
@@ -115,10 +117,6 @@ struct SResultRowEntryInfo* getResultEntryInfo(const SResultRow* pRow, int32_t i
 
 static FORCE_INLINE SResultRow* getResultRowByPos(SDiskbasedBuf* pBuf, SResultRowPosition* pos, bool forUpdate) {
   SFilePage* bufPage = (SFilePage*)getBufPage(pBuf, pos->pageId);
-  if (NULL == bufPage) {
-    return NULL;
-  }
-  
   if (forUpdate) {
     setBufPageDirty(bufPage, true);
   }

source/libs/executor/inc/executorimpl.h

@@ -474,6 +474,8 @@ typedef struct SStreamScanInfo {
   int32_t blockRecoverContiCnt;
   int32_t blockRecoverTotCnt;
 
+  int8_t igCheckUpdate;
+  int8_t igExpired;
 } SStreamScanInfo;
 
 typedef struct {