From 7930260feda84ff59acb6459b504ccba88e08c8a Mon Sep 17 00:00:00 2001
From: Cary Xu <klxu@taosdata.com>
Date: Sat, 11 Jun 2022 16:44:01 +0800
Subject: [PATCH 01/40] other: add debug logs when load block info

---
 source/dnode/vnode/src/tsdb/tsdbReadImpl.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/source/dnode/vnode/src/tsdb/tsdbReadImpl.c b/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
index 1c2514d46f..7e3dc5b98f 100644
--- a/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
+++ b/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
@@ -233,6 +233,17 @@ int tsdbLoadBlockInfo(SReadH *pReadh, void *pTarget) {
     tsdbError("vgId:%d, SBlockInfo part in file %s is corrupted since wrong checksum, offset:%u len :%u",
               TSDB_READ_REPO_ID(pReadh), TSDB_FILE_FULL_NAME(pHeadf), pBlkIdx->offset, pBlkIdx->len);
     return -1;
+  } else {
+    SBlockInfo *pBlkInfo = pReadh->pBlkInfo;
+    SBlock *pBlk = NULL;
+    int         nBlks = (pBlkIdx->len - sizeof(SBlockInfo)) / sizeof(SBlock);
+    for (int n = 0; n < nBlks; ++n) {
+      pBlk = &pBlkInfo->blocks[n];
+      if (pBlk->numOfSubBlocks == 1) {
+        tsdbDebug("prop:vgId:%d, file %s , offset:%u len :%u has %d rows", TSDB_READ_REPO_ID(pReadh),
+                  TSDB_FILE_FULL_NAME(pHeadf), pBlkIdx->offset, pBlkIdx->len, pBlk->numOfRows);
+      }
+    }
   }
 
   // ASSERT(pBlkIdx->tid == pReadh->pBlkInfo->tid && pBlkIdx->uid == pReadh->pBlkInfo->uid);

From 9b8a7d11df194c003006ea000ab5c4bac781cff7 Mon Sep 17 00:00:00 2001
From: Cary Xu <klxu@taosdata.com>
Date: Sat, 11 Jun 2022 18:50:28 +0800
Subject: [PATCH 02/40] other: add logs when read blk info

---
 source/dnode/vnode/src/tsdb/tsdbReadImpl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/source/dnode/vnode/src/tsdb/tsdbReadImpl.c b/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
index 7e3dc5b98f..e4c5ef8fc3 100644
--- a/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
+++ b/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
@@ -239,7 +239,7 @@ int tsdbLoadBlockInfo(SReadH *pReadh, void *pTarget) {
     int         nBlks = (pBlkIdx->len - sizeof(SBlockInfo)) / sizeof(SBlock);
     for (int n = 0; n < nBlks; ++n) {
       pBlk = &pBlkInfo->blocks[n];
-      if (pBlk->numOfSubBlocks == 1) {
+      if (pBlk->numOfSubBlocks >= 1) {
         tsdbDebug("prop:vgId:%d, file %s , offset:%u len :%u has %d rows", TSDB_READ_REPO_ID(pReadh),
                   TSDB_FILE_FULL_NAME(pHeadf), pBlkIdx->offset, pBlkIdx->len, pBlk->numOfRows);
       }

From f2de1af904765db5a6e8e08eca2485ba02f703a2 Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Wed, 15 Jun 2022 23:00:31 +0800
Subject: [PATCH 03/40] enh(query): partition by support arithmetic expr, and
 do some internal refactor.

---
 include/libs/function/function.h              |   4 -
 source/libs/executor/inc/executil.h           |   1 -
 source/libs/executor/inc/executorimpl.h       |  26 ++--
 source/libs/executor/src/executil.c           |  11 --
 source/libs/executor/src/executorimpl.c       | 133 +-----------------
 source/libs/executor/src/groupoperator.c      |  49 +++++--
 source/libs/executor/src/scanoperator.c       |  11 +-
 source/libs/executor/src/timewindowoperator.c |   5 -
 source/libs/function/inc/texpr.h              |  67 ---------
 source/libs/function/src/texpr.c              |   8 --
 10 files changed, 59 insertions(+), 256 deletions(-)
 delete mode 100644 source/libs/function/inc/texpr.h

diff --git a/include/libs/function/function.h b/include/libs/function/function.h
index b359fa5d6a..54762a0f17 100644
--- a/include/libs/function/function.h
+++ b/include/libs/function/function.h
@@ -150,12 +150,8 @@ typedef struct SqlFunctionCtx {
 } SqlFunctionCtx;
 
 enum {
-  TEXPR_NODE_DUMMY     = 0x0,
   TEXPR_BINARYEXPR_NODE= 0x1,
   TEXPR_UNARYEXPR_NODE = 0x2,
-  TEXPR_FUNCTION_NODE  = 0x3,
-  TEXPR_COL_NODE       = 0x4,
-  TEXPR_VALUE_NODE     = 0x8,
 };
 
 typedef struct tExprNode {
diff --git a/source/libs/executor/inc/executil.h b/source/libs/executor/inc/executil.h
index ca25b7aab1..a70cff5552 100644
--- a/source/libs/executor/inc/executil.h
+++ b/source/libs/executor/inc/executil.h
@@ -55,7 +55,6 @@ typedef struct SResultRow {
   uint32_t      numOfRows;   // number of rows of current time window
   STimeWindow   win;
   struct SResultRowEntryInfo pEntryInfo[];  // For each result column, there is a resultInfo
-//  char         *key;       // start key of current result row
 } SResultRow;
 
 typedef struct SResultRowPosition {
diff --git a/source/libs/executor/inc/executorimpl.h b/source/libs/executor/inc/executorimpl.h
index e3e85bf2b8..49316ff746 100644
--- a/source/libs/executor/inc/executorimpl.h
+++ b/source/libs/executor/inc/executorimpl.h
@@ -540,6 +540,13 @@ typedef struct SFillOperatorInfo {
   bool              multigroupResult;
 } SFillOperatorInfo;
 
+typedef struct SScalarSupp {
+  SExprInfo*      pScalarExprInfo;
+  int32_t         numOfScalarExpr;  // the number of scalar expression in group operator
+  SqlFunctionCtx* pScalarFuncCtx;
+  int32_t*        rowCellInfoOffset;  // offset value for each row result cell info
+} SScalarSupp;
+
 typedef struct SGroupbyOperatorInfo {
   // SOptrBasicInfo should be first, SAggSupporter should be second for stream encode
   SOptrBasicInfo  binfo;
@@ -552,10 +559,7 @@ typedef struct SGroupbyOperatorInfo {
   char*           keyBuf;       // group by keys for hash
   int32_t         groupKeyLen;  // total group by column width
   SGroupResInfo   groupResInfo;
-  SExprInfo*      pScalarExprInfo;
-  int32_t         numOfScalarExpr;  // the number of scalar expression in group operator
-  SqlFunctionCtx* pScalarFuncCtx;
-  int32_t*        rowCellInfoOffset;  // offset value for each row result cell info
+  SScalarSupp     scalarSup;
 } SGroupbyOperatorInfo;
 
 typedef struct SDataGroupInfo {
@@ -578,6 +582,7 @@ typedef struct SPartitionOperatorInfo {
   int32_t*       columnOffset;  // start position for each column data
   void*          pGroupIter;  // group iterator
   int32_t        pageIndex;   // page index of current group
+  SScalarSupp    scalarSupp;
 } SPartitionOperatorInfo;
 
 typedef struct SWindowRowsSup {
@@ -755,6 +760,8 @@ void    getAlignQueryTimeWindow(SInterval* pInterval, int32_t precision, int64_t
 int32_t getTableScanInfo(SOperatorInfo* pOperator, int32_t *order, int32_t* scanFlag);
 int32_t getBufferPgSize(int32_t rowSize, uint32_t* defaultPgsz, uint32_t* defaultBufsz);
 
+SArray* extractPartitionColInfo(SNodeList* pNodeList);
+
 void    doSetOperatorCompleted(SOperatorInfo* pOperator);
 void    doFilter(const SNode* pFilterNode, SSDataBlock* pBlock);
 SqlFunctionCtx* createSqlFunctionCtx(SExprInfo* pExprInfo, int32_t numOfOutput, int32_t** rowCellInfoOffset);
@@ -834,20 +841,17 @@ SOperatorInfo* createFillOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExp
 SOperatorInfo* createStatewindowOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExpr, int32_t numOfCols,
                                              SSDataBlock* pResBlock, STimeWindowAggSupp *pTwAggSupp, int32_t tsSlotId, SColumn* pStateKeyCol, SExecTaskInfo* pTaskInfo);
 
-SOperatorInfo* createPartitionOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
-                                           SSDataBlock* pResultBlock, SArray* pGroupColList, SExecTaskInfo* pTaskInfo);
+SOperatorInfo* createPartitionOperatorInfo(SOperatorInfo* downstream, SPartitionPhysiNode* pPartNode, SExecTaskInfo* pTaskInfo);
 
 SOperatorInfo* createTimeSliceOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
                                            SSDataBlock* pResultBlock, const SNodeListNode* pValNode, SExecTaskInfo* pTaskInfo);
 
 SOperatorInfo* createMergeJoinOperatorInfo(SOperatorInfo** pDownstream, int32_t numOfDownstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResBlock, SNode* pOnCondition, SExecTaskInfo* pTaskInfo);
 
-SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream,
-    SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResBlock, int64_t gap,
-    int32_t tsSlotId, STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo);
+SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
+    SSDataBlock* pResBlock, int64_t gap, int32_t tsSlotId, STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo);
 
-SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream,
-    SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo);
+SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo);
 
 #if 0
 SOperatorInfo* createTableSeqScanOperatorInfo(void* pTsdbReadHandle, STaskRuntimeEnv* pRuntimeEnv);
diff --git a/source/libs/executor/src/executil.c b/source/libs/executor/src/executil.c
index 6ba1cf859e..01ed30c189 100644
--- a/source/libs/executor/src/executil.c
+++ b/source/libs/executor/src/executil.c
@@ -28,17 +28,6 @@ typedef struct SCompSupporter {
   int32_t           order;
 } SCompSupporter;
 
-int32_t getOutputInterResultBufSize(STaskAttr* pQueryAttr) {
-  int32_t size = 0;
-
-  for (int32_t i = 0; i < pQueryAttr->numOfOutput; ++i) {
-//    size += pQueryAttr->pExpr1[i].base.interBytes;
-  }
-
-  assert(size >= 0);
-  return size;
-}
-
 int32_t initResultRowInfo(SResultRowInfo *pResultRowInfo, int32_t size) {
   pResultRowInfo->size       = 0;
   pResultRowInfo->cur.pageId = -1;
diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c
index de109a24ca..89692c0061 100644
--- a/source/libs/executor/src/executorimpl.c
+++ b/source/libs/executor/src/executorimpl.c
@@ -40,9 +40,6 @@
 #define IS_MAIN_SCAN(runtime)          ((runtime)->scanFlag == MAIN_SCAN)
 #define SET_REVERSE_SCAN_FLAG(runtime) ((runtime)->scanFlag = REVERSE_SCAN)
 
-#define SDATA_BLOCK_INITIALIZER \
-  (SDataBlockInfo) { {0}, 0 }
-
 #define GET_FORWARD_DIRECTION_FACTOR(ord) (((ord) == TSDB_ORDER_ASC) ? QUERY_ASC_FORWARD_STEP : QUERY_DESC_FORWARD_STEP)
 
 #if 0
@@ -95,8 +92,6 @@ static void setBlockStatisInfo(SqlFunctionCtx* pCtx, SExprInfo* pExpr, SSDataBlo
 
 static void destroyTableQueryInfoImpl(STableQueryInfo* pTableQueryInfo);
 
-static SColumnInfo* extractColumnFilterInfo(SExprInfo* pExpr, int32_t numOfOutput, int32_t* numOfFilterCols);
-
 static void releaseQueryBuf(size_t numOfTables);
 
 static void destroySFillOperatorInfo(void* param, int32_t numOfOutput);
@@ -454,75 +449,6 @@ static bool chkWindowOutputBufByKey(STaskRuntimeEnv* pRuntimeEnv, SResultRowInfo
   return chkResultRowFromKey(pRuntimeEnv, pResultRowInfo, (char*)&win->skey, TSDB_KEYSIZE, masterscan, groupId);
 }
 
-static void doUpdateResultRowIndex(SResultRowInfo* pResultRowInfo, TSKEY lastKey, bool ascQuery,
-                                   bool timeWindowInterpo) {
-  int64_t skey = TSKEY_INITIAL_VAL;
-#if 0
-  int32_t i = 0;
-  for (i = pResultRowInfo->size - 1; i >= 0; --i) {
-    SResultRow* pResult = pResultRowInfo->pResult[i];
-    if (pResult->closed) {
-      break;
-    }
-
-    // new closed result rows
-    if (timeWindowInterpo) {
-      if (pResult->endInterp &&
-          ((pResult->win.skey <= lastKey && ascQuery) || (pResult->win.skey >= lastKey && !ascQuery))) {
-        if (i > 0) {  // the first time window, the startInterp is false.
-          assert(pResult->startInterp);
-        }
-
-        closeResultRow(pResultRowInfo, i);
-      } else {
-        skey = pResult->win.skey;
-      }
-    } else {
-      if ((pResult->win.ekey <= lastKey && ascQuery) || (pResult->win.skey >= lastKey && !ascQuery)) {
-        closeResultRow(pResultRowInfo, i);
-      } else {
-        skey = pResult->win.skey;
-      }
-    }
-  }
-
-  // all result rows are closed, set the last one to be the skey
-  if (skey == TSKEY_INITIAL_VAL) {
-    if (pResultRowInfo->size == 0) {
-      //      assert(pResultRowInfo->current == NULL);
-      assert(pResultRowInfo->curPos == -1);
-      pResultRowInfo->curPos = -1;
-    } else {
-      pResultRowInfo->curPos = pResultRowInfo->size - 1;
-    }
-  } else {
-    for (i = pResultRowInfo->size - 1; i >= 0; --i) {
-      SResultRow* pResult = pResultRowInfo->pResult[i];
-      if (pResult->closed) {
-        break;
-      }
-    }
-
-    if (i == pResultRowInfo->size - 1) {
-      pResultRowInfo->curPos = i;
-    } else {
-      pResultRowInfo->curPos = i + 1;  // current not closed result object
-    }
-  }
-#endif
-}
-//
-// static void updateResultRowInfoActiveIndex(SResultRowInfo* pResultRowInfo, const STimeWindow* pWin, TSKEY lastKey,
-//                                           bool ascQuery, bool interp) {
-//  if ((lastKey > pWin->ekey && ascQuery) || (lastKey < pWin->ekey && (!ascQuery))) {
-//    closeAllResultRows(pResultRowInfo);
-//    pResultRowInfo->curPos = pResultRowInfo->size - 1;
-//  } else {
-//    int32_t step = ascQuery ? 1 : -1;
-//    doUpdateResultRowIndex(pResultRowInfo, lastKey - step, ascQuery, interp);
-//  }
-//}
-
 //  query_range_start, query_range_end, window_duration, window_start, window_end
 void initExecTimeWindowInfo(SColumnInfoData* pColData, STimeWindow* pQueryWindow) {
   pColData->info.type = TSDB_DATA_TYPE_TIMESTAMP;
@@ -810,20 +736,6 @@ int32_t projectApplyFunctions(SExprInfo* pExpr, SSDataBlock* pResult, SSDataBloc
       // _rowts/_c0, not tbname column
       if (fmIsPseudoColumnFunc(pfCtx->functionId) && (!fmIsScanPseudoColumnFunc(pfCtx->functionId))) {
         // do nothing
-      } else if (fmIsIndefiniteRowsFunc(pfCtx->functionId)) {
-        SResultRowEntryInfo* pResInfo = GET_RES_INFO(&pCtx[k]);
-        pfCtx->fpSet.init(&pCtx[k], pResInfo);
-
-        pfCtx->pOutput = taosArrayGet(pResult->pDataBlock, outputSlotId);
-        pfCtx->offset = createNewColModel ? 0 : pResult->info.rows;  // set the start offset
-
-        // set the timestamp(_rowts) output buffer
-        if (taosArrayGetSize(pPseudoList) > 0) {
-          int32_t* outputColIndex = taosArrayGet(pPseudoList, 0);
-          pfCtx->pTsOutput = (SColumnInfoData*)pCtx[*outputColIndex].pOutput;
-        }
-
-        numOfRows = pfCtx->fpSet.process(pfCtx);
       } else {
         SArray* pBlockList = taosArrayInit(4, POINTER_BYTES);
         taosArrayPush(pBlockList, &pSrcBlock);
@@ -886,7 +798,6 @@ int32_t setGroupResultOutputBuf(SOptrBasicInfo* binfo, int32_t numOfCols, char*
       doSetResultOutBufByKey(pBuf, pResultRowInfo, (char*)pData, bytes, true, groupId, pTaskInfo, false, pAggSup);
   assert(pResultRow != NULL);
 
-  setResultRowKey(pResultRow, pData, type);
   setResultRowInitCtx(pResultRow, pCtx, numOfCols, binfo->rowCellInfoOffset);
   return TSDB_CODE_SUCCESS;
 }
@@ -908,21 +819,6 @@ bool functionNeedToExecute(SqlFunctionCtx* pCtx) {
     return false;
   }
 
-  //  if (functionId == FUNCTION_FIRST_DST || functionId == FUNCTION_FIRST) {
-  //    //    return QUERY_IS_ASC_QUERY(pQueryAttr);
-  //  }
-  //
-  //  // denote the order type
-  //  if ((functionId == FUNCTION_LAST_DST || functionId == FUNCTION_LAST)) {
-  //    //    return pCtx->param[0].i == pQueryAttr->order.order;
-  //  }
-
-  // in the reverse table scan, only the following functions need to be executed
-  //  if (IS_REVERSE_SCAN(pRuntimeEnv) ||
-  //      (pRuntimeEnv->scanFlag == REPEAT_SCAN && functionId != FUNCTION_STDDEV && functionId != FUNCTION_PERCT)) {
-  //    return false;
-  //  }
-
   return true;
 }
 
@@ -3934,27 +3830,6 @@ void initResultSizeInfo(SOperatorInfo* pOperator, int32_t numOfRows) {
   }
 }
 
-// static STableQueryInfo* initTableQueryInfo(const STableListInfo* pTableListInfo) {
-//  int32_t size = taosArrayGetSize(pTableListInfo->pTableList);
-//  if (size == 0) {
-//    return NULL;
-//  }
-//
-//  STableQueryInfo* pTableQueryInfo = taosMemoryCalloc(size, sizeof(STableQueryInfo));
-//  if (pTableQueryInfo == NULL) {
-//    return NULL;
-//  }
-//
-//  for (int32_t j = 0; j < size; ++j) {
-//    STableKeyInfo*   pk = taosArrayGet(pTableListInfo->pTableList, j);
-//    STableQueryInfo* pTQueryInfo = &pTableQueryInfo[j];
-//    pTQueryInfo->lastKey = pk->lastKey;
-//  }
-//
-//  pTableQueryInfo->lastKey = 0;
-//  return pTableQueryInfo;
-//}
-
 SOperatorInfo* createAggregateOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
                                            SSDataBlock* pResultBlock, SExprInfo* pScalarExprInfo,
                                            int32_t numOfScalarExpr, SExecTaskInfo* pTaskInfo) {
@@ -4526,7 +4401,6 @@ static int32_t getTableList(void* metaHandle, int32_t tableType, uint64_t tableU
 static SArray* extractColumnInfo(SNodeList* pNodeList);
 
 static SArray* createSortInfo(SNodeList* pNodeList);
-static SArray* extractPartitionColInfo(SNodeList* pNodeList);
 
 int32_t extractTableSchemaVersion(SReadHandle* pHandle, uint64_t uid, SExecTaskInfo* pTaskInfo) {
   SMetaReader mr = {0};
@@ -4861,12 +4735,7 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
                                                pTaskInfo);
 
   } else if (QUERY_NODE_PHYSICAL_PLAN_PARTITION == type) {
-    SPartitionPhysiNode* pPartNode = (SPartitionPhysiNode*)pPhyNode;
-    SArray*              pColList = extractPartitionColInfo(pPartNode->pPartitionKeys);
-    SSDataBlock*         pResBlock = createResDataBlock(pPhyNode->pOutputDataBlockDesc);
-
-    SExprInfo* pExprInfo = createExprInfo(pPartNode->pTargets, NULL, &num);
-    pOptr = createPartitionOperatorInfo(ops[0], pExprInfo, num, pResBlock, pColList, pTaskInfo);
+    pOptr = createPartitionOperatorInfo(ops[0], (SPartitionPhysiNode*)pPhyNode, pTaskInfo);
   } else if (QUERY_NODE_PHYSICAL_PLAN_MERGE_STATE == type) {
     SStateWinodwPhysiNode* pStateNode = (SStateWinodwPhysiNode*)pPhyNode;
 
diff --git a/source/libs/executor/src/groupoperator.c b/source/libs/executor/src/groupoperator.c
index 688c576d03..23226a0134 100644
--- a/source/libs/executor/src/groupoperator.c
+++ b/source/libs/executor/src/groupoperator.c
@@ -320,8 +320,8 @@ static SSDataBlock* hashGroupbyAggregate(SOperatorInfo* pOperator) {
     setInputDataBlock(pOperator, pInfo->binfo.pCtx, pBlock, order, scanFlag, true);
 
     // there is an scalar expression that needs to be calculated right before apply the group aggregation.
-    if (pInfo->pScalarExprInfo != NULL) {
-      pTaskInfo->code = projectApplyFunctions(pInfo->pScalarExprInfo, pBlock, pBlock, pInfo->pScalarFuncCtx, pInfo->numOfScalarExpr, NULL);
+    if (pInfo->scalarSup.pScalarExprInfo != NULL) {
+      pTaskInfo->code = projectApplyFunctions(pInfo->scalarSup.pScalarExprInfo, pBlock, pBlock, pInfo->scalarSup.pScalarFuncCtx, pInfo->scalarSup.numOfScalarExpr, NULL);
       if (pTaskInfo->code != TSDB_CODE_SUCCESS) {
         longjmp(pTaskInfo->env, pTaskInfo->code);
       }
@@ -385,9 +385,9 @@ SOperatorInfo* createGroupOperatorInfo(SOperatorInfo* downstream, SExprInfo* pEx
   pInfo->pGroupCols      = pGroupColList;
   pInfo->pCondition      = pCondition;
 
-  pInfo->pScalarExprInfo = pScalarExprInfo;
-  pInfo->numOfScalarExpr = numOfScalarExpr;
-  pInfo->pScalarFuncCtx  = createSqlFunctionCtx(pScalarExprInfo, numOfScalarExpr, &pInfo->rowCellInfoOffset);
+  pInfo->scalarSup.pScalarExprInfo = pScalarExprInfo;
+  pInfo->scalarSup.numOfScalarExpr = numOfScalarExpr;
+  pInfo->scalarSup.pScalarFuncCtx  = createSqlFunctionCtx(pScalarExprInfo, numOfScalarExpr, &pInfo->scalarSup.rowCellInfoOffset);
 
   int32_t code = initGroupOptrInfo(&pInfo->pGroupColVals, &pInfo->groupKeyLen, &pInfo->keyBuf, pGroupColList);
   if (code != TSDB_CODE_SUCCESS) {
@@ -624,7 +624,9 @@ static SSDataBlock* hashPartition(SOperatorInfo* pOperator) {
     return NULL;
   }
 
-  SGroupbyOperatorInfo* pInfo = pOperator->info;
+  SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
+
+  SPartitionOperatorInfo* pInfo = pOperator->info;
   SSDataBlock* pRes = pInfo->binfo.pRes;
 
   if (pOperator->status == OP_RES_TO_RETURN) {
@@ -641,6 +643,14 @@ static SSDataBlock* hashPartition(SOperatorInfo* pOperator) {
       break;
     }
 
+    // there is an scalar expression that needs to be calculated right before apply the group aggregation.
+    if (pInfo->scalarSupp.pScalarExprInfo != NULL) {
+      pTaskInfo->code = projectApplyFunctions(pInfo->scalarSupp.pScalarExprInfo, pBlock, pBlock, pInfo->scalarSupp.pScalarFuncCtx, pInfo->scalarSupp.numOfScalarExpr, NULL);
+      if (pTaskInfo->code != TSDB_CODE_SUCCESS) {
+        longjmp(pTaskInfo->env, pTaskInfo->code);
+      }
+    }
+
     doHashPartition(pOperator, pBlock);
   }
 
@@ -665,15 +675,26 @@ static void destroyPartitionOperatorInfo(void* param, int32_t numOfOutput) {
   taosMemoryFree(pInfo->columnOffset);
 }
 
-SOperatorInfo* createPartitionOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
-                                           SSDataBlock* pResultBlock, SArray* pGroupColList, SExecTaskInfo* pTaskInfo) {
+SOperatorInfo* createPartitionOperatorInfo(SOperatorInfo* downstream, SPartitionPhysiNode* pPartNode, SExecTaskInfo* pTaskInfo) {
   SPartitionOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SPartitionOperatorInfo));
   SOperatorInfo*        pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
   if (pInfo == NULL || pOperator == NULL) {
     goto _error;
   }
 
-  pInfo->pGroupCols = pGroupColList;
+  SSDataBlock* pResBlock = createResDataBlock(pPartNode->node.pOutputDataBlockDesc);
+
+  int32_t numOfCols = 0;
+  SExprInfo* pExprInfo = createExprInfo(pPartNode->pTargets, NULL, &numOfCols);
+
+  pInfo->pGroupCols = extractPartitionColInfo(pPartNode->pPartitionKeys);
+
+  if (pPartNode->pExprs != NULL) {
+    pInfo->scalarSupp.numOfScalarExpr = 0;
+    pInfo->scalarSupp.pScalarExprInfo = createExprInfo(pPartNode->pExprs, NULL, &pInfo->scalarSupp.numOfScalarExpr);
+    pInfo->scalarSupp.pScalarFuncCtx = createSqlFunctionCtx(
+        pInfo->scalarSupp.pScalarExprInfo, pInfo->scalarSupp.numOfScalarExpr, &pInfo->scalarSupp.rowCellInfoOffset);
+  }
 
   _hash_fn_t hashFn = taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY);
   pInfo->pGroupSet = taosHashInit(100, hashFn, false, HASH_NO_LOCK);
@@ -683,16 +704,16 @@ SOperatorInfo* createPartitionOperatorInfo(SOperatorInfo* downstream, SExprInfo*
 
   uint32_t defaultPgsz  = 0;
   uint32_t defaultBufsz = 0;
-  getBufferPgSize(pResultBlock->info.rowSize, &defaultPgsz, &defaultBufsz);
+  getBufferPgSize(pResBlock->info.rowSize, &defaultPgsz, &defaultBufsz);
 
   int32_t code = createDiskbasedBuf(&pInfo->pBuf, defaultPgsz, defaultBufsz, pTaskInfo->id.str, TD_TMP_DIR_PATH);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }
 
-  pInfo->rowCapacity = blockDataGetCapacityInRow(pResultBlock, getBufPageSize(pInfo->pBuf));
-  pInfo->columnOffset = setupColumnOffset(pResultBlock, pInfo->rowCapacity);
-  code = initGroupOptrInfo(&pInfo->pGroupColVals, &pInfo->groupKeyLen, &pInfo->keyBuf, pGroupColList);
+  pInfo->rowCapacity = blockDataGetCapacityInRow(pResBlock, getBufPageSize(pInfo->pBuf));
+  pInfo->columnOffset = setupColumnOffset(pResBlock, pInfo->rowCapacity);
+  code = initGroupOptrInfo(&pInfo->pGroupColVals, &pInfo->groupKeyLen, &pInfo->keyBuf, pInfo->pGroupCols);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }
@@ -701,7 +722,7 @@ SOperatorInfo* createPartitionOperatorInfo(SOperatorInfo* downstream, SExprInfo*
   pOperator->blocking     = true;
   pOperator->status       = OP_NOT_OPENED;
   pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_PARTITION;
-  pInfo->binfo.pRes       = pResultBlock;
+  pInfo->binfo.pRes       = pResBlock;
   pOperator->numOfExprs   = numOfCols;
   pOperator->pExpr        = pExprInfo;
   pOperator->info         = pInfo;
diff --git a/source/libs/executor/src/scanoperator.c b/source/libs/executor/src/scanoperator.c
index d30e4ef6db..62c2240098 100644
--- a/source/libs/executor/src/scanoperator.c
+++ b/source/libs/executor/src/scanoperator.c
@@ -1157,6 +1157,8 @@ static SSDataBlock* doFilterResult(SSysTableScanInfo* pInfo) {
     return pInfo->pRes->info.rows == 0 ? NULL : pInfo->pRes;
   }
 
+  doFilter(pInfo->pCondition, pInfo->pRes);
+#if 0
   SFilterInfo* filter = NULL;
 
   int32_t code = filterInitFromNode(pInfo->pCondition, &filter, 0);
@@ -1202,6 +1204,7 @@ static SSDataBlock* doFilterResult(SSysTableScanInfo* pInfo) {
 
   px->info.rows = numOfRow;
   pInfo->pRes = px;
+#endif
 
   return pInfo->pRes->info.rows == 0 ? NULL : pInfo->pRes;
 }
@@ -1380,6 +1383,8 @@ static SSDataBlock* doSysTableScan(SOperatorInfo* pOperator) {
       relocateColumnData(pInfo->pRes, pInfo->scanCols, p->pDataBlock);
       doFilterResult(pInfo);
 
+      blockDataDestroy(p);
+
       pInfo->loadInfo.totalRows += pInfo->pRes->info.rows;
       return (pInfo->pRes->info.rows == 0) ? NULL : pInfo->pRes;
     }
@@ -1469,10 +1474,10 @@ int32_t buildSysDbTableInfo(const SSysTableScanInfo* pInfo, int32_t capacity) {
   p->info.rows = buildDbTableInfoBlock(p, pSysDbTableMeta, size, TSDB_PERFORMANCE_SCHEMA_DB);
 
   relocateColumnData(pInfo->pRes, pInfo->scanCols, p->pDataBlock);
-  //  blockDataDestroy(p);  todo handle memory leak
-
   pInfo->pRes->info.rows = p->info.rows;
-  return p->info.rows;
+  blockDataDestroy(p);
+
+  return pInfo->pRes->info.rows;
 }
 
 int32_t buildDbTableInfoBlock(const SSDataBlock* p, const SSysTableMeta* pSysDbTableMeta, size_t size,
diff --git a/source/libs/executor/src/timewindowoperator.c b/source/libs/executor/src/timewindowoperator.c
index 4141c85366..4673123779 100644
--- a/source/libs/executor/src/timewindowoperator.c
+++ b/source/libs/executor/src/timewindowoperator.c
@@ -333,11 +333,6 @@ void doTimeWindowInterpolation(SIntervalAggOperatorInfo* pInfo, int32_t numOfExp
       continue;
     }
 
-    //    if (functionId != FUNCTION_TWA && functionId != FUNCTION_INTERP) {
-    //      pCtx[k].start.key = INT64_MIN;
-    //      continue;
-    //    }
-
     SFunctParam*     pParam = &pCtx[k].param[0];
     SColumnInfoData* pColInfo = taosArrayGet(pDataBlock, pParam->pCol->slotId);
 
diff --git a/source/libs/function/inc/texpr.h b/source/libs/function/inc/texpr.h
deleted file mode 100644
index 68aedcb5a2..0000000000
--- a/source/libs/function/inc/texpr.h
+++ /dev/null
@@ -1,67 +0,0 @@
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef _TD_COMMON_EXPR_H_
-#define _TD_COMMON_EXPR_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#include "os.h"
-
-#include "tmsg.h"
-#include "taosdef.h"
-#include "tskiplist.h"
-#include "function.h"
-
-struct tExprNode;
-struct SSchema;
-
-#define QUERY_COND_REL_PREFIX_IN "IN|"
-#define QUERY_COND_REL_PREFIX_LIKE "LIKE|"
-#define QUERY_COND_REL_PREFIX_MATCH "MATCH|"
-#define QUERY_COND_REL_PREFIX_NMATCH "NMATCH|"
-
-#define QUERY_COND_REL_PREFIX_IN_LEN   3
-#define QUERY_COND_REL_PREFIX_LIKE_LEN 5
-#define QUERY_COND_REL_PREFIX_MATCH_LEN 6
-#define QUERY_COND_REL_PREFIX_NMATCH_LEN 7
-
-typedef bool (*__result_filter_fn_t)(const void *, void *);
-typedef void (*__do_filter_suppl_fn_t)(void *, void *);
-
-/**
- * this structure is used to filter data in tags, so the offset of filtered tag column in tagdata string is required
- */
-typedef struct tQueryInfo {
-  uint8_t       optr;     // expression operator
-  SSchema       sch;      // schema of tags
-  char*         q;
-  __compar_fn_t compare;  // filter function
-  bool          indexed;  // indexed columns
-} tQueryInfo;
-
-typedef struct SExprTraverseSupp {
-  __result_filter_fn_t   nodeFilterFn;
-  __do_filter_suppl_fn_t setupInfoFn;
-  void                  *pExtInfo;
-} SExprTraverseSupp;
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif  /*_TD_COMMON_EXPR_H_*/
diff --git a/source/libs/function/src/texpr.c b/source/libs/function/src/texpr.c
index f04b4f17f9..039e0a9dfc 100644
--- a/source/libs/function/src/texpr.c
+++ b/source/libs/function/src/texpr.c
@@ -17,15 +17,7 @@
 #include "os.h"
 
 #include "texception.h"
-#include "taosdef.h"
 #include "tmsg.h"
-#include "tarray.h"
-#include "tbuffer.h"
-#include "tcompare.h"
-#include "thash.h"
-#include "texpr.h"
-#include "tvariant.h"
-#include "tdef.h"
 
 static void doExprTreeDestroy(tExprNode **pExpr, void (*fp)(void *));
 

From 803249837a00f12207a455fbc228209cf2372360 Mon Sep 17 00:00:00 2001
From: Liu Jicong <liujicong@qq.com>
Date: Thu, 16 Jun 2022 10:29:30 +0800
Subject: [PATCH 04/40] feat(stream): drop stream

---
 include/common/tmsg.h                   |  6 ++---
 source/dnode/mnode/impl/src/mndStream.c | 31 ++++++++++++++-----------
 2 files changed, 21 insertions(+), 16 deletions(-)

diff --git a/include/common/tmsg.h b/include/common/tmsg.h
index 4894bef8ff..7c0e8206e1 100644
--- a/include/common/tmsg.h
+++ b/include/common/tmsg.h
@@ -2008,7 +2008,7 @@ typedef struct {
   char     sql[TSDB_SHOW_SQL_LEN];
   uint64_t queryId;
   int64_t  useconds;
-  int64_t  stime;            // timestamp precision ms
+  int64_t  stime;  // timestamp precision ms
   int64_t  reqRid;
   int32_t  pid;
   bool     stableQuery;
@@ -2257,8 +2257,8 @@ typedef struct {
 } SMqVDeleteRsp;
 
 typedef struct {
-  char    name[TSDB_STREAM_FNAME_LEN];
-  int64_t streamId;
+  char   name[TSDB_STREAM_FNAME_LEN];
+  int8_t igNotExists;
 } SMDropStreamTaskReq;
 
 typedef struct {
diff --git a/source/dnode/mnode/impl/src/mndStream.c b/source/dnode/mnode/impl/src/mndStream.c
index 24e41435e9..491835fc15 100644
--- a/source/dnode/mnode/impl/src/mndStream.c
+++ b/source/dnode/mnode/impl/src/mndStream.c
@@ -278,8 +278,8 @@ int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast
 }
 
 static int32_t mndCreateStbForStream(SMnode *pMnode, STrans *pTrans, const SStreamObj *pStream, const char *user) {
-  SStbObj  *pStb = NULL;
-  SDbObj   *pDb = NULL;
+  SStbObj *pStb = NULL;
+  SDbObj  *pDb = NULL;
 
   SMCreateStbReq createReq = {0};
   tstrncpy(createReq.name, pStream->targetSTbName, TSDB_TABLE_FNAME_LEN);
@@ -321,7 +321,6 @@ static int32_t mndCreateStbForStream(SMnode *pMnode, STrans *pTrans, const SStre
     goto _OVER;
   }
 
-
   if (mndCheckDbAuth(pMnode, user, MND_OPER_WRITE_DB, pDb) != 0) {
     goto _OVER;
   }
@@ -520,13 +519,19 @@ static int32_t mndProcessDropStreamReq(SRpcMsg *pReq) {
   /*SDbObj     *pDb = NULL;*/
   /*SUserObj   *pUser = NULL;*/
 
-  SMDropStreamTaskReq dropStreamReq = *(SMDropStreamTaskReq *)pReq->pCont;
+  SMDropStreamTaskReq dropReq = *(SMDropStreamTaskReq *)pReq->pCont;
 
-  pStream = mndAcquireStream(pMnode, dropStreamReq.name);
+  pStream = mndAcquireStream(pMnode, dropReq.name);
 
   if (pStream == NULL) {
-    terrno = TSDB_CODE_MND_STREAM_NOT_EXIST;
-    return -1;
+    if (dropReq.igNotExists) {
+      mDebug("stream:%s, not exist, ignore not exist is set", dropReq.name);
+      code = 0;
+      goto DROP_STREAM_OVER;
+    } else {
+      terrno = TSDB_CODE_MND_STREAM_NOT_EXIST;
+      return -1;
+    }
   }
 
 #if 0
@@ -539,19 +544,19 @@ static int32_t mndProcessDropStreamReq(SRpcMsg *pReq) {
 
   STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_NOTHING, pReq);
   if (pTrans == NULL) {
-    mError("stream:%s, failed to drop since %s", dropStreamReq.name, terrstr());
-    return -1;
+    mError("stream:%s, failed to drop since %s", dropReq.name, terrstr());
+    return code;
   }
-  mDebug("trans:%d, used to drop stream:%s", pTrans->id, dropStreamReq.name);
+  mDebug("trans:%d, used to drop stream:%s", pTrans->id, dropReq.name);
 
   // drop all tasks
   if (mndDropStreamTasks(pMnode, pTrans, pStream) < 0) {
-    mError("stream:%s, failed to drop task since %s", dropStreamReq.name, terrstr());
-    return -1;
+    mError("stream:%s, failed to drop task since %s", dropReq.name, terrstr());
+    return code;
   }
 
 DROP_STREAM_OVER:
-  return 0;
+  return code;
 }
 
 static int32_t mndGetNumOfStreams(SMnode *pMnode, char *dbName, int32_t *pNumOfStreams) {

From 63884b554bf9104ffc6bb7e13d0efd5483b3b1be Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 10:51:29 +0800
Subject: [PATCH 05/40] fix(query): set correct sort buffer page size.

---
 source/common/src/tdatablock.c          |  1 +
 source/libs/executor/inc/tsort.h        |  7 +++++++
 source/libs/executor/src/sortoperator.c | 27 +++++++++++++++----------
 source/libs/executor/src/tsort.c        | 26 +++++++++++++++---------
 4 files changed, 41 insertions(+), 20 deletions(-)

diff --git a/source/common/src/tdatablock.c b/source/common/src/tdatablock.c
index 5a2aaed74e..6a4ddacf60 100644
--- a/source/common/src/tdatablock.c
+++ b/source/common/src/tdatablock.c
@@ -1226,6 +1226,7 @@ SSDataBlock* createOneDataBlock(const SSDataBlock* pDataBlock, bool copyData) {
     SColumnInfoData  colInfo = {0};
     SColumnInfoData* p = taosArrayGet(pDataBlock->pDataBlock, i);
     colInfo.info = p->info;
+    colInfo.hasNull = true;
     taosArrayPush(pBlock->pDataBlock, &colInfo);
   }
 
diff --git a/source/libs/executor/inc/tsort.h b/source/libs/executor/inc/tsort.h
index fd3581e2bf..e731c55a7d 100644
--- a/source/libs/executor/inc/tsort.h
+++ b/source/libs/executor/inc/tsort.h
@@ -151,6 +151,13 @@ SSDataBlock* tsortGetSortedDataBlock(const SSortHandle* pSortHandle);
  */
 SSortExecInfo tsortGetSortExecInfo(SSortHandle* pHandle);
 
+/**
+ * get proper sort buffer pages according to the row size
+ * @param rowSize
+ * @return
+ */
+int32_t getProperSortPageSize(size_t rowSize);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/source/libs/executor/src/sortoperator.c b/source/libs/executor/src/sortoperator.c
index 1008a5263c..98a1aacbb6 100644
--- a/source/libs/executor/src/sortoperator.c
+++ b/source/libs/executor/src/sortoperator.c
@@ -420,24 +420,29 @@ SOperatorInfo* createMultiwaySortMergeOperatorInfo(SOperatorInfo** downStreams,
     goto _error;
   }
 
-  pInfo->binfo.pRes = pResBlock;
 
   initResultSizeInfo(pOperator, 1024);
 
-  pInfo->pSortInfo = pSortInfo;
+  pInfo->binfo.pRes    = pResBlock;
+  pInfo->pSortInfo     = pSortInfo;
   pInfo->pColMatchInfo = pColMatchColInfo;
-  pInfo->pInputBlock = pInputBlock;
-  pOperator->name = "MultiwaySortMerge";
+  pInfo->pInputBlock   = pInputBlock;
+  pOperator->name      = "MultiwaySortMerge";
   pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE;
-  pOperator->blocking = false;
-  pOperator->status = OP_NOT_OPENED;
-  pOperator->info = pInfo;
-
-  pInfo->bufPageSize = rowSize < 1024 ? 1024 : rowSize * 2;
-  pInfo->sortBufSize = pInfo->bufPageSize * 16;
-  pInfo->hasGroupId = false;
+  pOperator->blocking  = false;
+  pOperator->status    = OP_NOT_OPENED;
+  pOperator->info      = pInfo;
+  pInfo->hasGroupId    = false;
   pInfo->prefetchedTuple = NULL;
   pOperator->pTaskInfo = pTaskInfo;
+
+  pInfo->bufPageSize   = getProperSortPageSize(rowSize);
+
+  uint32_t numOfSources = taosArrayGetSize(pSortInfo);
+  numOfSources = MAX(2, numOfSources);
+
+  pInfo->sortBufSize    = numOfSources * pInfo->bufPageSize;
+
   pOperator->fpSet =
       createOperatorFpSet(doOpenMultiwaySortMergeOperator, doMultiwaySortMerge, NULL, NULL,
                           destroyMultiwaySortMergeOperatorInfo, NULL, NULL, getMultiwaySortMergeExplainExecInfo);
diff --git a/source/libs/executor/src/tsort.c b/source/libs/executor/src/tsort.c
index 1473bd81bb..daaba74d7c 100644
--- a/source/libs/executor/src/tsort.c
+++ b/source/libs/executor/src/tsort.c
@@ -532,6 +532,19 @@ static int32_t doInternalMergeSort(SSortHandle* pHandle) {
   return 0;
 }
 
+int32_t getProperSortPageSize(size_t rowSize) {
+  uint32_t defaultPageSize = 4096;
+
+  uint32_t pgSize = 0;
+  if (rowSize * 4 > defaultPageSize) {
+    pgSize = rowSize * 4;
+  } else {
+    pgSize = defaultPageSize;
+  }
+
+  return pgSize;
+}
+
 static int32_t createInitialSources(SSortHandle* pHandle) {
   size_t sortBufSize = pHandle->numOfPages * pHandle->pageSize;
 
@@ -557,14 +570,9 @@ static int32_t createInitialSources(SSortHandle* pHandle) {
 
       if (!hasGroupId) {
         // calculate the buffer pages according to the total available buffers.
-        int32_t rowSize = blockDataGetRowSize(pBlock);
-        if (rowSize * 4 > 4096) {
-          pHandle->pageSize = rowSize * 4;
-        } else {
-          pHandle->pageSize = 4096;
-        }
-        
-        // todo!!
+        pHandle->pageSize = getProperSortPageSize(blockDataGetRowSize(pBlock));
+
+        // todo, number of pages are set according to the total available sort buffer
         pHandle->numOfPages = 1024;
         sortBufSize = pHandle->numOfPages * pHandle->pageSize;
 
@@ -577,7 +585,7 @@ static int32_t createInitialSources(SSortHandle* pHandle) {
         if (pHandle->beforeFp != NULL) {
           pHandle->beforeFp(pBlock, pHandle->param);
         }
-        // todo relocate the columns
+
         int32_t code = blockDataMerge(pHandle->pDataBlock, pBlock);
         if (code != 0) {
           return code;

From f89fcece482499638cf6d4623111ca1902f164e7 Mon Sep 17 00:00:00 2001
From: wade zhang <95411902+gccgdb1234@users.noreply.github.com>
Date: Thu, 16 Jun 2022 10:46:20 +0800
Subject: [PATCH 06/40] Update index.mdx

---
 docs/en/07-develop/02-model/index.mdx | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/en/07-develop/02-model/index.mdx b/docs/en/07-develop/02-model/index.mdx
index 86853aaaa3..e0378cc77c 100644
--- a/docs/en/07-develop/02-model/index.mdx
+++ b/docs/en/07-develop/02-model/index.mdx
@@ -37,7 +37,7 @@ USE power;
 
 ## Create STable
 
-In a time-series application, there may be multiple kinds of data collection points. For example, in the electrical power system there are meters, transformers, bus bars, switches, etc. For easy and efficient aggregation of multiple tables, one STable needs to be created for each kind of data collection point. For example, for the meters in [table 1](/tdinternal/arch#model_table1), the SQL statement below can be used to create the super table.
+In a time-series application, there may be multiple kinds of data collection points. For example, in the electrical power system there are meters, transformers, bus bars, switches, etc. For easy and efficient aggregation of multiple tables, one STable needs to be created for each kind of data collection point. For example, for the meters in [table 1](/concept/#model_table1), the SQL statement below can be used to create the super table.
 
 ```sql
 CREATE STable meters (ts timestamp, current float, voltage int, phase float) TAGS (location binary(64), groupId int);

From 58051169d2102ad1b9c43f62673d40e34d7aab63 Mon Sep 17 00:00:00 2001
From: Minghao Li <castermode@gmail.com>
Date: Thu, 16 Jun 2022 10:54:13 +0800
Subject: [PATCH 07/40] refactor(sync): add trace log

---
 source/libs/sync/src/syncAppendEntriesReply.c |  8 +--
 source/libs/sync/src/syncSnapshot.c           | 70 ++++++++++---------
 2 files changed, 40 insertions(+), 38 deletions(-)

diff --git a/source/libs/sync/src/syncAppendEntriesReply.c b/source/libs/sync/src/syncAppendEntriesReply.c
index 0e116e13ee..fbe3319675 100644
--- a/source/libs/sync/src/syncAppendEntriesReply.c
+++ b/source/libs/sync/src/syncAppendEntriesReply.c
@@ -191,17 +191,17 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
           char* s = snapshotSender2Str(pSender);
           sDebug(
               "vgId:%d sync event snapshot send to %s:%d start sender first time, lastApplyIndex:%ld lastApplyTerm:%lu "
-              "lastConfigIndex:%ld"
+              "lastConfigIndex:%ld privateTerm:%lu "
               "sender:%s",
               ths->vgId, host, port, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
-              pSender->snapshot.lastConfigIndex, s);
+              pSender->snapshot.lastConfigIndex, pSender->privateTerm, s);
           taosMemoryFree(s);
         } else {
           sDebug(
               "vgId:%d sync event snapshot send to %s:%d start sender first time, lastApplyIndex:%ld "
-              "lastApplyTerm:%lu lastConfigIndex:%ld",
+              "lastApplyTerm:%lu lastConfigIndex:%ld privateTerm:%lu",
               ths->vgId, host, port, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
-              pSender->snapshot.lastConfigIndex);
+              pSender->snapshot.lastConfigIndex, pSender->privateTerm);
         }
       }
 
diff --git a/source/libs/sync/src/syncSnapshot.c b/source/libs/sync/src/syncSnapshot.c
index afebfa798e..c92920a936 100644
--- a/source/libs/sync/src/syncSnapshot.c
+++ b/source/libs/sync/src/syncSnapshot.c
@@ -142,17 +142,17 @@ void snapshotSenderStart(SSyncSnapshotSender *pSender) {
     char *msgStr = syncSnapshotSend2Str(pMsg);
     sDebug(
         "vgId:%d sync event snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
-        "lastConfigIndex:%ld send "
+        "lastConfigIndex:%ld privateTerm:%lu send "
         "msg:%s",
         pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, msgStr);
+        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm, msgStr);
     taosMemoryFree(msgStr);
   } else {
     sDebug(
         "vgId:%d sync event snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
-        "lastConfigIndex:%ld",
+        "lastConfigIndex:%ld privateTerm:%lu",
         pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex);
+        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm);
   }
 
   syncSnapshotSendDestroy(pMsg);
@@ -280,24 +280,24 @@ int32_t snapshotSend(SSyncSnapshotSender *pSender) {
       char *msgStr = syncSnapshotSend2Str(pMsg);
       sDebug(
           "vgId:%d sync event snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
-          "lastConfigIndex:%ld send "
+          "lastConfigIndex:%ld privateTerm:%lu send "
           "msg:%s",
           pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-          pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, msgStr);
+          pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm, msgStr);
       taosMemoryFree(msgStr);
     } else {
       sDebug(
           "vgId:%d sync event snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
-          "lastConfigIndex:%ld",
+          "lastConfigIndex:%ld privateTerm:%lu",
           pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-          pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex);
+          pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm);
     }
   } else {
     sDebug(
         "vgId:%d sync event snapshot send to %s:%d sending seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
-        "lastConfigIndex:%ld",
+        "lastConfigIndex:%ld privateTerm:%lu",
         pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex);
+        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm);
   }
 
   syncSnapshotSendDestroy(pMsg);
@@ -328,12 +328,12 @@ int32_t snapshotReSend(SSyncSnapshotSender *pSender) {
 
     if (gRaftDetailLog) {
       char *msgStr = syncSnapshotSend2Str(pMsg);
-      sDebug("vgId:%d sync event snapshot send to %s:%d resend seq:%d ack:%d send msg:%s", pSender->pSyncNode->vgId,
-             host, port, pSender->seq, pSender->ack, msgStr);
+      sDebug("vgId:%d sync event snapshot send to %s:%d resend seq:%d ack:%d privateTerm:%lu send msg:%s",
+             pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->privateTerm, msgStr);
       taosMemoryFree(msgStr);
     } else {
-      sDebug("vgId:%d sync event snapshot send to %s:%d resend seq:%d ack:%d", pSender->pSyncNode->vgId, host, port,
-             pSender->seq, pSender->ack);
+      sDebug("vgId:%d sync event snapshot send to %s:%d resend seq:%d ack:%d privateTerm:%lu", pSender->pSyncNode->vgId,
+             host, port, pSender->seq, pSender->ack, pSender->privateTerm);
     }
 
     syncSnapshotSendDestroy(pMsg);
@@ -567,15 +567,16 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld, recv msg:%s",
+              "lastConfigIndex:%ld, privateTerm:%lu, recv msg:%s",
               pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
-              msgStr);
+              pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld",
-              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex);
+              "lastConfigIndex:%ld privateTerm:%lu",
+              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
+              pReceiver->privateTerm);
         }
 
       } else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_END) {
@@ -638,17 +639,17 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d finish, update log begin index:%ld, "
               "snapshot.lastApplyIndex:%ld, "
-              "snapshot.lastApplyTerm:%lu, snapshot.lastConfigIndex:%ld, raft log:%s",
+              "snapshot.lastApplyTerm:%lu, snapshot.lastConfigIndex:%ld, privateTerm:%lu, raft log:%s",
               pSyncNode->vgId, host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm,
-              snapshot.lastConfigIndex, logSimpleStr);
+              snapshot.lastConfigIndex, pReceiver->privateTerm, logSimpleStr);
           taosMemoryFree(logSimpleStr);
         } else {
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d finish, update log begin index:%ld, "
               "snapshot.lastApplyIndex:%ld, "
-              "snapshot.lastApplyTerm:%lu, snapshot.lastConfigIndex:%ld",
+              "snapshot.lastApplyTerm:%lu, snapshot.lastConfigIndex:%ld, privateTerm:%lu",
               pSyncNode->vgId, host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm,
-              snapshot.lastConfigIndex);
+              snapshot.lastConfigIndex, pReceiver->privateTerm);
         }
 
         pReceiver->pWriter = NULL;
@@ -660,16 +661,16 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld, recv msg:%s",
+              "lastConfigIndex:%ld, privateTerm:%lu, recv msg:%s",
               pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex, msgStr);
+              pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld",
+              "lastConfigIndex:%ld, privateTerm:%lu",
               pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex);
+              pMsg->lastConfigIndex, pReceiver->privateTerm);
         }
 
       } else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_FORCE_CLOSE) {
@@ -685,17 +686,17 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld, recv "
+              "lastConfigIndex:%ld, privateTerm:%lu, recv "
               "msg:%s",
               pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex, msgStr);
+              pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld",
+              "lastConfigIndex:%ld, privateTerm:%lu",
               pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex);
+              pMsg->lastConfigIndex, pReceiver->privateTerm);
         }
 
       } else if (pMsg->seq > SYNC_SNAPSHOT_SEQ_BEGIN && pMsg->seq < SYNC_SNAPSHOT_SEQ_END) {
@@ -716,15 +717,16 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld, recv msg:%s",
+              "lastConfigIndex:%ld, privateTerm:%lu, recv msg:%s",
               pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
-              msgStr);
+              pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
               "vgId:%d sync event snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu, "
-              "lastConfigIndex:%ld",
-              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex);
+              "lastConfigIndex:%ld, privateTerm:%lu",
+              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
+              pReceiver->privateTerm);
         }
 
       } else {

From 61e3250ee76560cef9d24f871453e34116b28a7c Mon Sep 17 00:00:00 2001
From: gccgdb1234 <wxzhang@taosdata.com>
Date: Thu, 16 Jun 2022 11:17:36 +0800
Subject: [PATCH 08/40] docs: correct TAOS CLI to TDengine CLI

---
 docs/en/05-get-started/index.md                      | 4 ++--
 docs/en/07-develop/03-insert-data/01-sql-writing.mdx | 2 +-
 docs/zh/05-get-started/index.md                      | 4 ++--
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/docs/en/05-get-started/index.md b/docs/en/05-get-started/index.md
index 56958ef3ec..9c5a853820 100644
--- a/docs/en/05-get-started/index.md
+++ b/docs/en/05-get-started/index.md
@@ -1,6 +1,6 @@
 ---
 title: Get Started
-description: 'Install TDengine from Docker image, apt-get or package, and run TAOS CLI and taosBenchmark to experience the features'
+description: 'Install TDengine from Docker image, apt-get or package, and run TDengine CLI and taosBenchmark to experience the features'
 ---
 
 import Tabs from "@theme/Tabs";
@@ -120,7 +120,7 @@ select * from t;
 Query OK, 2 row(s) in set (0.003128s)
 ```
 
-Besides executing SQL commands, system administrators can check running status, add/drop user accounts and manage the running instances. TAOS CLI with client driver can be installed and run on either Linux or Windows machines. For more details on CLI, please [check here](../reference/taos-shell/).
+Besides executing SQL commands, system administrators can check running status, add/drop user accounts and manage the running instances. TDengine CLI with client driver can be installed and run on either Linux or Windows machines. For more details on CLI, please [check here](../reference/taos-shell/).
 
 ## Experience the blazing fast speed
 
diff --git a/docs/en/07-develop/03-insert-data/01-sql-writing.mdx b/docs/en/07-develop/03-insert-data/01-sql-writing.mdx
index 397b1a14fd..d8c4453f40 100644
--- a/docs/en/07-develop/03-insert-data/01-sql-writing.mdx
+++ b/docs/en/07-develop/03-insert-data/01-sql-writing.mdx
@@ -22,7 +22,7 @@ import CStmt from "./_c_stmt.mdx";
 
 ## Introduction
 
-Application programs can execute `INSERT` statement through connectors to insert rows. The TAOS CLI can also be used to manually insert data.
+Application programs can execute `INSERT` statement through connectors to insert rows. The TDengine CLI can also be used to manually insert data.
 
 ### Insert Single Row
 
diff --git a/docs/zh/05-get-started/index.md b/docs/zh/05-get-started/index.md
index 878d7f0202..818027174e 100644
--- a/docs/zh/05-get-started/index.md
+++ b/docs/zh/05-get-started/index.md
@@ -1,6 +1,6 @@
 ---
 title: 立即开始
-description: '从 Docker，安装包或使用 apt-get 快速安装 TDengine, 通过命令行程序TAOS CLI和工具 taosdemo 快速体验 TDengine 功能'
+description: '从 Docker，安装包或使用 apt-get 快速安装 TDengine, 通过命令行程序TDengine CLI和工具 taosdemo 快速体验 TDengine 功能'
 ---
 
 import Tabs from "@theme/Tabs";
@@ -122,7 +122,7 @@ select * from t;
 Query OK, 2 row(s) in set (0.003128s)
 ```
 
-除执行 SQL 语句外，系统管理员还可以从 TDengine CLI 进行检查系统运行状态、添加删除用户账号等操作。TAOS CLI 连同应用驱动也可以独立安装在 Linux 或 Windows 机器上运行，更多细节请参考 [这里](../reference/taos-shell/)
+除执行 SQL 语句外，系统管理员还可以从 TDengine CLI 进行检查系统运行状态、添加删除用户账号等操作。TDengine CLI 连同应用驱动也可以独立安装在 Linux 或 Windows 机器上运行，更多细节请参考 [这里](../reference/taos-shell/)
 
 ## 使用 taosBenchmark 体验写入速度
 

From 2360f072a84dba8086dda533e8297d2821b5b8b9 Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 11:43:11 +0800
Subject: [PATCH 09/40] fix(query): add indef process procedure.

---
 include/libs/executor/executor.h        |  9 +--------
 source/libs/executor/inc/executorimpl.h |  4 ++--
 source/libs/executor/src/executor.c     |  2 +-
 source/libs/executor/src/executorMain.c |  4 ++--
 source/libs/executor/src/executorimpl.c | 17 ++++++++++++++++-
 source/libs/qworker/inc/qwMsg.h         |  2 +-
 source/libs/qworker/src/qwMsg.c         |  4 +---
 source/libs/qworker/src/qworker.c       |  6 +++---
 8 files changed, 27 insertions(+), 21 deletions(-)

diff --git a/include/libs/executor/executor.h b/include/libs/executor/executor.h
index 3f05c27453..fdedb947d7 100644
--- a/include/libs/executor/executor.h
+++ b/include/libs/executor/executor.h
@@ -95,7 +95,7 @@ int32_t qUpdateQualifiedTableId(qTaskInfo_t tinfo, const SArray* tableIdList, bo
  * @return
  */
 int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, struct SSubplan* pPlan,
-                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, EOPTR_EXEC_MODEL model);
+                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, const char* sql, EOPTR_EXEC_MODEL model);
 
 /**
  *
@@ -159,13 +159,6 @@ int64_t qGetQueriedTableUid(qTaskInfo_t tinfo);
  */
 int32_t qGetQualifiedTableIdList(void* pTableList, const char* tagCond, int32_t tagCondLen, SArray* pTableIdList);
 
-/**
- * Update the table id list of a given query.
- * @param uid   child table uid
- * @param type  operation type: ADD|DROP
- * @return
- */
-int32_t qUpdateQueriedTableIdList(qTaskInfo_t tinfo, int64_t uid, int32_t type);
 
 void qProcessFetchRsp(void* parent, struct SRpcMsg* pMsg, struct SEpSet* pEpSet);
 
diff --git a/source/libs/executor/inc/executorimpl.h b/source/libs/executor/inc/executorimpl.h
index 758c031028..b435446513 100644
--- a/source/libs/executor/inc/executorimpl.h
+++ b/source/libs/executor/inc/executorimpl.h
@@ -189,7 +189,7 @@ typedef struct SExecTaskInfo {
   } schemaVer;
 
   STableListInfo   tableqinfoList;       // this is a table list
-  char*            sql;                  // query sql string
+  const char*      sql;                  // query sql string
   jmp_buf          env;                  // jump to this position when error happens.
   EOPTR_EXEC_MODEL execModel;            // operator execution model [batch model|stream model]
   struct SOperatorInfo* pRoot;
@@ -894,7 +894,7 @@ int32_t decodeOperator(SOperatorInfo* ops, char* data, int32_t length);
 
 void    setTaskStatus(SExecTaskInfo* pTaskInfo, int8_t status);
 int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId,
-                               EOPTR_EXEC_MODEL model);
+                               const char* sql, EOPTR_EXEC_MODEL model);
 int32_t createDataSinkParam(SDataSinkNode *pNode, void **pParam, qTaskInfo_t* pTaskInfo);                               
 int32_t getOperatorExplainExecInfo(SOperatorInfo* operatorInfo, SExplainExecInfo** pRes, int32_t* capacity,
                                    int32_t* resNum);
diff --git a/source/libs/executor/src/executor.c b/source/libs/executor/src/executor.c
index bebfc75f17..b1d076e8f5 100644
--- a/source/libs/executor/src/executor.c
+++ b/source/libs/executor/src/executor.c
@@ -121,7 +121,7 @@ qTaskInfo_t qCreateStreamExecTaskInfo(void* msg, void* streamReadHandle) {
   }
 
   qTaskInfo_t pTaskInfo = NULL;
-  code = qCreateExecTask(streamReadHandle, 0, 0, plan, &pTaskInfo, NULL, OPTR_EXEC_MODEL_STREAM);
+  code = qCreateExecTask(streamReadHandle, 0, 0, plan, &pTaskInfo, NULL, NULL, OPTR_EXEC_MODEL_STREAM);
   if (code != TSDB_CODE_SUCCESS) {
     // TODO: destroy SSubplan & pTaskInfo
     terrno = code;
diff --git a/source/libs/executor/src/executorMain.c b/source/libs/executor/src/executorMain.c
index b1125c8e80..00158d7024 100644
--- a/source/libs/executor/src/executorMain.c
+++ b/source/libs/executor/src/executorMain.c
@@ -31,13 +31,13 @@ static void initRefPool() {
 }
 
 int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, SSubplan* pSubplan,
-                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, EOPTR_EXEC_MODEL model) {
+                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, const char* sql, EOPTR_EXEC_MODEL model) {
   assert(readHandle != NULL && pSubplan != NULL);
   SExecTaskInfo** pTask = (SExecTaskInfo**)pTaskInfo;
 
   taosThreadOnce(&initPoolOnce, initRefPool);
 
-  int32_t code = createExecTaskInfoImpl(pSubplan, pTask, readHandle, taskId, model);
+  int32_t code = createExecTaskInfoImpl(pSubplan, pTask, readHandle, taskId, sql, model);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }
diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c
index d066284c56..9ce22431e1 100644
--- a/source/libs/executor/src/executorimpl.c
+++ b/source/libs/executor/src/executorimpl.c
@@ -736,6 +736,20 @@ int32_t projectApplyFunctions(SExprInfo* pExpr, SSDataBlock* pResult, SSDataBloc
       // _rowts/_c0, not tbname column
       if (fmIsPseudoColumnFunc(pfCtx->functionId) && (!fmIsScanPseudoColumnFunc(pfCtx->functionId))) {
         // do nothing
+      } else if (fmIsIndefiniteRowsFunc(pfCtx->functionId)) {
+        SResultRowEntryInfo* pResInfo = GET_RES_INFO(&pCtx[k]);
+        pfCtx->fpSet.init(&pCtx[k], pResInfo);
+
+        pfCtx->pOutput = taosArrayGet(pResult->pDataBlock, outputSlotId);
+        pfCtx->offset = createNewColModel ? 0 : pResult->info.rows;  // set the start offset
+
+        // set the timestamp(_rowts) output buffer
+        if (taosArrayGetSize(pPseudoList) > 0) {
+          int32_t* outputColIndex = taosArrayGet(pPseudoList, 0);
+          pfCtx->pTsOutput = (SColumnInfoData*)pCtx[*outputColIndex].pOutput;
+        }
+
+        numOfRows = pfCtx->fpSet.process(pfCtx);
       } else {
         SArray* pBlockList = taosArrayInit(4, POINTER_BYTES);
         taosArrayPush(pBlockList, &pSrcBlock);
@@ -5183,7 +5197,7 @@ int32_t createDataSinkParam(SDataSinkNode* pNode, void** pParam, qTaskInfo_t* pT
 }
 
 int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId,
-                               EOPTR_EXEC_MODEL model) {
+                               const char* sql, EOPTR_EXEC_MODEL model) {
   uint64_t queryId = pPlan->id.queryId;
 
   int32_t code = TSDB_CODE_SUCCESS;
@@ -5193,6 +5207,7 @@ int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SRead
     goto _complete;
   }
 
+  (*pTaskInfo)->sql = sql;
   (*pTaskInfo)->pRoot = createOperatorTree(pPlan->pNode, *pTaskInfo, pHandle, queryId, taskId,
                                            &(*pTaskInfo)->tableqinfoList, pPlan->pTagCond);
   if (NULL == (*pTaskInfo)->pRoot) {
diff --git a/source/libs/qworker/inc/qwMsg.h b/source/libs/qworker/inc/qwMsg.h
index 29861d87ac..ecff861f50 100644
--- a/source/libs/qworker/inc/qwMsg.h
+++ b/source/libs/qworker/inc/qwMsg.h
@@ -24,7 +24,7 @@ extern "C" {
 #include "dataSinkMgt.h"
 
 int32_t qwPrerocessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg);
-int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, int8_t taskType, int8_t explain);
+int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, int8_t taskType, int8_t explain, const char* sql);
 int32_t qwProcessCQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg);
 int32_t qwProcessReady(QW_FPARAMS_DEF, SQWMsg *qwMsg);
 int32_t qwProcessFetch(QW_FPARAMS_DEF, SQWMsg *qwMsg);
diff --git a/source/libs/qworker/src/qwMsg.c b/source/libs/qworker/src/qwMsg.c
index 848a0420ca..5635ec8fc6 100644
--- a/source/libs/qworker/src/qwMsg.c
+++ b/source/libs/qworker/src/qwMsg.c
@@ -308,10 +308,8 @@ int32_t qWorkerProcessQueryMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, int
   SQWMsg qwMsg = {.node = node, .msg = msg->msg + msg->sqlLen, .msgLen = msg->phyLen, .connInfo = pMsg->info};
   char * sql = strndup(msg->msg, msg->sqlLen);
   QW_SCH_TASK_DLOG("processQuery start, node:%p, handle:%p, sql:%s", node, pMsg->info.handle, sql);
-  taosMemoryFreeClear(sql);
-
-  QW_ERR_RET(qwProcessQuery(QW_FPARAMS(), &qwMsg, msg->taskType, msg->explain));
 
+  QW_ERR_RET(qwProcessQuery(QW_FPARAMS(), &qwMsg, msg->taskType, msg->explain, sql));
   QW_SCH_TASK_DLOG("processQuery end, node:%p", node);
 
   return TSDB_CODE_SUCCESS;
diff --git a/source/libs/qworker/src/qworker.c b/source/libs/qworker/src/qworker.c
index 451607e7d0..57ebb89ed2 100644
--- a/source/libs/qworker/src/qworker.c
+++ b/source/libs/qworker/src/qworker.c
@@ -510,7 +510,7 @@ _return:
 }
 
 
-int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, int8_t taskType, int8_t explain) {
+int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, int8_t taskType, int8_t explain, const char* sql) {
   int32_t        code = 0;
   bool           queryRsped = false;
   SSubplan      *plan = NULL;
@@ -537,7 +537,7 @@ int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, int8_t taskType, int8_t ex
 
   ctx->plan = plan;
 
-  code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle, OPTR_EXEC_MODEL_BATCH);
+  code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle, sql, OPTR_EXEC_MODEL_BATCH);
   if (code) {
     QW_TASK_ELOG("qCreateExecTask failed, code:%x - %s", code, tstrerror(code));
     QW_ERR_JRET(code);
@@ -938,7 +938,7 @@ int32_t qwProcessDelete(QW_FPARAMS_DEF, SQWMsg *qwMsg, SRpcMsg *pRsp, SDeleteRes
 
   ctx.plan = plan;
 
-  code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle, OPTR_EXEC_MODEL_BATCH);
+  code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle, NULL, OPTR_EXEC_MODEL_BATCH);
   if (code) {
     QW_TASK_ELOG("qCreateExecTask failed, code:%x - %s", code, tstrerror(code));
     QW_ERR_JRET(code);

From 102fdf2916aadfe0f438806fd7c188d3580a4537 Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 11:52:58 +0800
Subject: [PATCH 10/40] fix(query): fix the syntax error in unit test.

---
 source/libs/executor/test/executorTests.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/source/libs/executor/test/executorTests.cpp b/source/libs/executor/test/executorTests.cpp
index 880ac6c78e..dba2fb897a 100644
--- a/source/libs/executor/test/executorTests.cpp
+++ b/source/libs/executor/test/executorTests.cpp
@@ -945,7 +945,7 @@ TEST(testCase, build_executor_tree_Test) {
   int32_t          code = qStringToSubplan(msg, &plan);
   ASSERT_EQ(code, 0);
 
-  code = qCreateExecTask(&handle, 2, 1, plan, (void**)&pTaskInfo, &sinkHandle, OPTR_EXEC_MODEL_BATCH);
+  code = qCreateExecTask(&handle, 2, 1, plan, (void**)&pTaskInfo, &sinkHandle, NULL, OPTR_EXEC_MODEL_BATCH);
   ASSERT_EQ(code, 0);
 }
 #if 0

From ed532ed7acb5e0e4916b09e2e1e9e60308790380 Mon Sep 17 00:00:00 2001
From: Shengliang Guan <slguan@taosdata.com>
Date: Thu, 16 Jun 2022 12:01:13 +0800
Subject: [PATCH 11/40] refactor: adjust logs

---
 source/dnode/mnode/impl/src/mndDnode.c | 1 +
 source/dnode/mnode/impl/src/mndMain.c  | 1 +
 source/dnode/mnode/impl/src/mndMnode.c | 4 ++++
 3 files changed, 6 insertions(+)

diff --git a/source/dnode/mnode/impl/src/mndDnode.c b/source/dnode/mnode/impl/src/mndDnode.c
index 53cbb9b669..6d37143fc9 100644
--- a/source/dnode/mnode/impl/src/mndDnode.c
+++ b/source/dnode/mnode/impl/src/mndDnode.c
@@ -370,6 +370,7 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
   SMnodeObj *pObj = mndAcquireMnode(pMnode, pDnode->id);
   if (pObj != NULL) {
     if (pObj->state != statusReq.mload.syncState) {
+      mInfo("dnode:%d, mnode syncstate from %s to %s", pObj->id, syncStr(pObj->state), syncStr(statusReq.mload.syncState));
       pObj->state = statusReq.mload.syncState;
       pObj->stateStartTime = taosGetTimestampMs();
     }
diff --git a/source/dnode/mnode/impl/src/mndMain.c b/source/dnode/mnode/impl/src/mndMain.c
index 00118eac1e..e141e7dfb3 100644
--- a/source/dnode/mnode/impl/src/mndMain.c
+++ b/source/dnode/mnode/impl/src/mndMain.c
@@ -699,6 +699,7 @@ int32_t mndGetMonitorInfo(SMnode *pMnode, SMonClusterInfo *pClusterInfo, SMonVgr
 
 int32_t mndGetLoad(SMnode *pMnode, SMnodeLoad *pLoad) {
   pLoad->syncState = syncGetMyRole(pMnode->syncMgmt.sync);
+  mTrace("mnode current syncstate is %s", syncStr(pLoad->syncState));
   return 0;
 }
 
diff --git a/source/dnode/mnode/impl/src/mndMnode.c b/source/dnode/mnode/impl/src/mndMnode.c
index 7ec490d52f..e82c479cdd 100644
--- a/source/dnode/mnode/impl/src/mndMnode.c
+++ b/source/dnode/mnode/impl/src/mndMnode.c
@@ -667,6 +667,10 @@ static int32_t mndRetrieveMnodes(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pB
     }
     if (pObj->pDnode && mndIsDnodeOnline(pObj->pDnode, curMs)) {
       roles = syncStr(pObj->state);
+      if (pObj->state == TAOS_SYNC_STATE_LEADER && pObj->id != pMnode->selfDnodeId) {
+        roles = syncStr(TAOS_SYNC_STATE_ERROR);
+        mError("mnode:%d, is leader too", pObj->id);
+      }
     }
     char b2[12 + VARSTR_HEADER_SIZE] = {0};
     STR_WITH_MAXSIZE_TO_VARSTR(b2, roles, pShow->pMeta->pSchemas[cols].bytes);

From 75e0eaa53fed96fa0a766475dcb57ebdf1c7032c Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 13:53:02 +0800
Subject: [PATCH 12/40] fix(query): add null pointer check.

---
 source/libs/executor/src/tsort.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/source/libs/executor/src/tsort.c b/source/libs/executor/src/tsort.c
index daaba74d7c..5bcd58f8db 100644
--- a/source/libs/executor/src/tsort.c
+++ b/source/libs/executor/src/tsort.c
@@ -204,7 +204,12 @@ static int32_t sortComparInit(SMsortComparParam* cmpParam, SArray* pSources, int
   if (pHandle->type == SORT_SINGLESOURCE_SORT) {
     for (int32_t i = 0; i < cmpParam->numOfSources; ++i) {
       SSortSource* pSource = cmpParam->pSources[i];
-      SPageInfo*          pPgInfo = *(SPageInfo**)taosArrayGet(pSource->pageIdList, pSource->pageIndex);
+
+      if (taosArrayGetSize(pSource->pageIdList) == 0) {
+        return TSDB_CODE_SUCCESS;
+      }
+
+      SPageInfo* pPgInfo = *(SPageInfo**)taosArrayGet(pSource->pageIdList, pSource->pageIndex);
 
       void* pPage = getBufPage(pHandle->pBuf, getPageId(pPgInfo));
       code = blockDataFromBuf(pSource->src.pBlock, pPage);

From 0671daa70b4db0b689df34543edd588e5a2249ed Mon Sep 17 00:00:00 2001
From: Shuduo Sang <sangshuduo@gmail.com>
Date: Thu, 16 Jun 2022 14:07:15 +0800
Subject: [PATCH 13/40] chore: update taos-tools (#13889)

for 3.0
---
 tools/taos-tools | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/taos-tools b/tools/taos-tools
index 3d5aa76f8c..0a81480420 160000
--- a/tools/taos-tools
+++ b/tools/taos-tools
@@ -1 +1 @@
-Subproject commit 3d5aa76f8c718dcffa100b45e4cbf313d499c356
+Subproject commit 0a81480420d6601bbdb57770ee64e40f24c4ea83

From 2ba3b7538c50b0d7267d0c056b962e0d7f90aebe Mon Sep 17 00:00:00 2001
From: Shengliang Guan <slguan@taosdata.com>
Date: Thu, 16 Jun 2022 14:09:14 +0800
Subject: [PATCH 14/40] test: redistribute vgroup

---
 ...redistribute_vgroup_replica3_v1_leader.sim | 166 +++++-------------
 1 file changed, 48 insertions(+), 118 deletions(-)

diff --git a/tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim b/tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
index 90f7b1ad1f..7b52b51306 100644
--- a/tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
+++ b/tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
@@ -4,6 +4,11 @@ system sh/deploy.sh -n dnode2 -i 2
 system sh/deploy.sh -n dnode3 -i 3
 system sh/deploy.sh -n dnode4 -i 4
 system sh/deploy.sh -n dnode5 -i 5
+system sh/cfg.sh -n dnode1 -c transPullupInterval -v 1
+system sh/cfg.sh -n dnode2 -c transPullupInterval -v 1
+system sh/cfg.sh -n dnode3 -c transPullupInterval -v 1
+system sh/cfg.sh -n dnode4 -c transPullupInterval -v 1
+system sh/cfg.sh -n dnode5 -c transPullupInterval -v 1
 system sh/cfg.sh -n dnode1 -c supportVnodes -v 0
 system sh/exec.sh -n dnode1 -s start
 system sh/exec.sh -n dnode2 -s start
@@ -128,9 +133,9 @@ if $rows != 1 then
 endi
 if $data(2)[4] == leader then
   $leaderExist = 1
-  $leaderVnode = 4
-  $follower1 = 2
-  $follower2 = 3
+  $leaderVnode = 2
+  $follower1 = 3
+  $follower2 = 4
 endi
 if $data(2)[6] == leader then
   $leaderExist = 1
@@ -140,9 +145,9 @@ if $data(2)[6] == leader then
 endi
 if $data(2)[8] == leader then
   $leaderExist = 1
-  $leaderVnode = 2
-  $follower1 = 3
-  $follower2 = 4
+  $leaderVnode = 4
+  $follower1 = 2
+  $follower2 = 3
 endi
 if  $leaderExist != 1 then
   goto step3
@@ -160,110 +165,22 @@ if $rows != 1 then
   return -1
 endi
 
-<<<<<<< HEAD
-print =============== step32: move follower2
-print redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
-sql redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
-sql show d1.vgroups
-print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
-
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
-=======
 print =============== step32: move leader
 print redistribute vgroup 2 dnode $follower1 dnode $follower2 dnode 5 
 sql redistribute vgroup 2 dnode $follower1 dnode $follower2 dnode 5 
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
 
->>>>>>> origin/3.0
 sql show d1.tables
 if $rows != 1 then
   return -1
 endi
 
-<<<<<<< HEAD
+return 
+
 print =============== step33: move follower1
-print redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
-sql redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
-sql show d1.vgroups
-print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
-
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
-=======
-print =============== step33: move follower2
-print redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
-sql redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
-sql show d1.vgroups
-print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
-
->>>>>>> origin/3.0
-sql show d1.tables
-if $rows != 1 then
-  return -1
-endi
-
-<<<<<<< HEAD
-print =============== step34: move follower2
-print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
-sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
-sql show d1.vgroups
-print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
-
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
-=======
-print =============== step34: move follower1
-print redistribute vgroup 2 dnode $follower2 dnode 5 dnode $leaderVnode
-sql redistribute vgroup 2 dnode $follower2 dnode 5 dnode $leaderVnode
-sql show d1.vgroups
-print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
-
->>>>>>> origin/3.0
-sql show d1.tables
-if $rows != 1 then
-  return -1
-endi
-
-<<<<<<< HEAD
-print =============== step35: move follower1
-print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
-sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-sql show d1.tables
-if $rows != 1 then
-  return -1
-endi
-=======
-sql show d1.vgroups
-print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
-=======
-print =============== step35: move 5
-print redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode $follower2
-sql redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode $follower2
-sql show d1.vgroups
-print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
->>>>>>> origin/3.0
-
-sql show d1.tables
-if $rows != 1 then
-  return -1
-endi
-
-<<<<<<< HEAD
-print =============== step36: move follower2
 print redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
 sql redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
-=======
-print =============== step36: move follower1
-print redistribute vgroup 2 dnode $follower1 dnode $follower2 dnode 5 
-sql redistribute vgroup 2 dnode $follower1 dnode $follower2 dnode 5 
->>>>>>> origin/3.0
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
 
@@ -272,11 +189,40 @@ if $rows != 1 then
   return -1
 endi
 
-<<<<<<< HEAD
-print =============== step37: move follower1
-=======
+print =============== step34: move follower2
+print redistribute vgroup 2 dnode $follower1 dnode 5 dnode $leaderVnode
+sql redistribute vgroup 2 dnode $follower1 dnode 5 dnode $leaderVnode
+sql show d1.vgroups
+print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
+
+sql show d1.tables
+if $rows != 1 then
+  return -1
+endi
+
+print =============== step35: move leader
+print redistribute vgroup 2 dnode $follower2 dnode 5 dnode $follower1
+sql redistribute vgroup 2 dnode $follower2 dnode 5 dnode $follower1
+sql show d1.vgroups
+print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
+
+sql show d1.tables
+if $rows != 1 then
+  return -1
+endi
+
+print =============== step36: move follower1
+print redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
+sql redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5 
+sql show d1.vgroups
+print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
+
+sql show d1.tables
+if $rows != 1 then
+  return -1
+endi
+
 print =============== step37: move follower2
->>>>>>> origin/3.0
 print redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
 sql redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
 sql show d1.vgroups
@@ -287,17 +233,9 @@ if $rows != 1 then
   return -1
 endi
 
-<<<<<<< HEAD
-<<<<<<< HEAD:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
-=======
-print =============== step38: move follower2
-print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
-sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
-=======
 print =============== step38: move leader
 print redistribute vgroup 2 dnode $follower1 dnode 5 dnode $follower2
 sql redistribute vgroup 2 dnode $follower1 dnode 5 dnode $follower2
->>>>>>> origin/3.0
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
 
@@ -305,18 +243,10 @@ sql show d1.tables
 if $rows != 1 then
   return -1
 endi
-<<<<<<< HEAD
->>>>>>> origin/3.0:tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
 
 print =============== step39: move follower1
-print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
-sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
-=======
-
-print =============== step39: move 5
-print redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode $follower1
-sql redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode $follower1
->>>>>>> origin/3.0
+print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
+sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
 sql show d1.vgroups
 print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
 

From 77b03ac821c8c03d8f583d2e52203dd7021e605a Mon Sep 17 00:00:00 2001
From: Minghao Li <castermode@gmail.com>
Date: Thu, 16 Jun 2022 14:40:42 +0800
Subject: [PATCH 15/40] refactor(sync): add restore finish when become leader
 again

---
 include/libs/sync/sync.h               |   2 +-
 source/dnode/mnode/impl/src/mndSync.c  |  12 +--
 source/libs/sync/inc/syncInt.h         |   3 +
 source/libs/sync/src/syncMain.c        | 129 +++++++++++++++++++------
 source/libs/sync/src/syncRaftLog.c     |   7 +-
 source/libs/sync/src/syncRespMgr.c     |  15 +--
 source/libs/sync/src/syncSnapshot.c    |   7 ++
 source/libs/sync/test/syncTestTool.cpp |   8 +-
 8 files changed, 134 insertions(+), 49 deletions(-)

diff --git a/include/libs/sync/sync.h b/include/libs/sync/sync.h
index 2d21b3531a..e694203563 100644
--- a/include/libs/sync/sync.h
+++ b/include/libs/sync/sync.h
@@ -24,7 +24,7 @@ extern "C" {
 #include "tdef.h"
 #include "tmsgcb.h"
 
-#define SYNC_INDEX_BEGIN   0
+#define SYNC_INDEX_BEGIN 0
 #define SYNC_INDEX_INVALID -1
 
 typedef uint64_t SyncNodeId;
diff --git a/source/dnode/mnode/impl/src/mndSync.c b/source/dnode/mnode/impl/src/mndSync.c
index 63708aef8a..3a243e36d3 100644
--- a/source/dnode/mnode/impl/src/mndSync.c
+++ b/source/dnode/mnode/impl/src/mndSync.c
@@ -221,17 +221,17 @@ void mndCleanupSync(SMnode *pMnode) {
 
 int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) {
   SSyncMgmt *pMgmt = &pMnode->syncMgmt;
-  SRpcMsg    rsp = {.code = TDMT_MND_APPLY_MSG, .contLen = sdbGetRawTotalSize(pRaw)};
-  rsp.pCont = rpcMallocCont(rsp.contLen);
-  if (rsp.pCont == NULL) return -1;
-  memcpy(rsp.pCont, pRaw, rsp.contLen);
+  SRpcMsg    req = {.msgType = TDMT_MND_APPLY_MSG, .contLen = sdbGetRawTotalSize(pRaw)};
+  req.pCont = rpcMallocCont(req.contLen);
+  if (req.pCont == NULL) return -1;
+  memcpy(req.pCont, pRaw, req.contLen);
 
   pMgmt->errCode = 0;
   pMgmt->transId = transId;
   mTrace("trans:%d, will be proposed", pMgmt->transId);
 
   const bool isWeak = false;
-  int32_t    code = syncPropose(pMgmt->sync, &rsp, isWeak);
+  int32_t    code = syncPropose(pMgmt->sync, &req, isWeak);
   if (code == 0) {
     tsem_wait(&pMgmt->syncSem);
   } else if (code == -1 && terrno == TSDB_CODE_SYN_NOT_LEADER) {
@@ -242,7 +242,7 @@ int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) {
     terrno = TSDB_CODE_APP_ERROR;
   }
 
-  rpcFreeCont(rsp.pCont);
+  rpcFreeCont(req.pCont);
   if (code != 0) {
     mError("trans:%d, failed to propose, code:0x%x", pMgmt->transId, code);
     return code;
diff --git a/source/libs/sync/inc/syncInt.h b/source/libs/sync/inc/syncInt.h
index 8a44dcd9bc..b00c7cbda1 100644
--- a/source/libs/sync/inc/syncInt.h
+++ b/source/libs/sync/inc/syncInt.h
@@ -195,6 +195,7 @@ int32_t syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, S
 cJSON*  syncNode2Json(const SSyncNode* pSyncNode);
 char*   syncNode2Str(const SSyncNode* pSyncNode);
 char*   syncNode2SimpleStr(const SSyncNode* pSyncNode);
+bool    syncNodeInConfig(SSyncNode* pSyncNode, const SSyncCfg* config);
 void    syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex lastConfigChangeIndex, bool* isDrop);
 
 SSyncNode* syncNodeAcquire(int64_t rid);
@@ -230,6 +231,8 @@ int32_t   syncNodeGetPreIndexTerm(SSyncNode* pSyncNode, SyncIndex index, SyncInd
 
 int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, uint64_t flag);
 
+int32_t syncNodeUpdateNewConfigIndex(SSyncNode* ths, SSyncCfg* pNewCfg);
+
 bool                 syncNodeInRaftGroup(SSyncNode* ths, SRaftId* pRaftId);
 SSyncSnapshotSender* syncNodeGetSnapshotSender(SSyncNode* ths, SRaftId* pDestId);
 
diff --git a/source/libs/sync/src/syncMain.c b/source/libs/sync/src/syncMain.c
index a0d8487b63..5e3c1adf32 100644
--- a/source/libs/sync/src/syncMain.c
+++ b/source/libs/sync/src/syncMain.c
@@ -185,7 +185,9 @@ int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg
   ASSERT(rid == pSyncNode->rid);
 
   int32_t ret = 0;
-  bool    IamInNew = false;
+  bool    IamInNew = syncNodeInConfig(pSyncNode, pNewCfg);
+
+#if 0
   for (int i = 0; i < pNewCfg->replicaNum; ++i) {
     if (strcmp((pNewCfg->nodeInfo)[i].nodeFqdn, pSyncNode->myNodeInfo.nodeFqdn) == 0 &&
         (pNewCfg->nodeInfo)[i].nodePort == pSyncNode->myNodeInfo.nodePort) {
@@ -201,6 +203,7 @@ int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg
         }
     */
   }
+#endif
 
   if (!IamInNew) {
     taosReleaseRef(tsNodeRefId, pSyncNode->rid);
@@ -228,7 +231,9 @@ int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg) {
   }
   ASSERT(rid == pSyncNode->rid);
 
-  bool IamInNew = false;
+  bool IamInNew = syncNodeInConfig(pSyncNode, pNewCfg);
+
+#if 0 
   for (int i = 0; i < pNewCfg->replicaNum; ++i) {
     if (strcmp((pNewCfg->nodeInfo)[i].nodeFqdn, pSyncNode->myNodeInfo.nodeFqdn) == 0 &&
         (pNewCfg->nodeInfo)[i].nodePort == pSyncNode->myNodeInfo.nodePort) {
@@ -247,6 +252,7 @@ int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg) {
         }
       */
   }
+#endif
 
   if (!IamInNew) {
     sError("sync reconfig error, not in new config");
@@ -556,7 +562,7 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
     return -1;
   }
   assert(rid == pSyncNode->rid);
-  sDebug("vgId:%d sync event propose msgType:%s", pSyncNode->vgId, TMSG_INFO(pMsg->msgType));
+  sDebug("vgId:%d sync event propose msgType:%s,%d", pSyncNode->vgId, TMSG_INFO(pMsg->msgType), pMsg->msgType);
   ret = syncNodePropose(pSyncNode, pMsg, isWeak);
 
   taosReleaseRef(tsNodeRefId, pSyncNode->rid);
@@ -565,7 +571,7 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
 
 int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak) {
   int32_t ret = 0;
-  sDebug("vgId:%d sync event propose msgType:%s", pSyncNode->vgId, TMSG_INFO(pMsg->msgType));
+  sDebug("vgId:%d sync event propose msgType:%s,%d", pSyncNode->vgId, TMSG_INFO(pMsg->msgType), pMsg->msgType);
 
   if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
     SRespStub stub;
@@ -1256,9 +1262,36 @@ char* syncNode2SimpleStr(const SSyncNode* pSyncNode) {
   return s;
 }
 
-void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex lastConfigChangeIndex, bool* isDrop) {
+bool syncNodeInConfig(SSyncNode* pSyncNode, const SSyncCfg* config) {
+  bool b1 = false;
+  bool b2 = false;
+
+  for (int i = 0; i < config->replicaNum; ++i) {
+    if (strcmp((config->nodeInfo)[i].nodeFqdn, pSyncNode->myNodeInfo.nodeFqdn) == 0 &&
+        (config->nodeInfo)[i].nodePort == pSyncNode->myNodeInfo.nodePort) {
+      b1 = true;
+      break;
+    }
+  }
+
+  for (int i = 0; i < config->replicaNum; ++i) {
+    SRaftId raftId;
+    raftId.addr = syncUtilAddr2U64((config->nodeInfo)[i].nodeFqdn, (config->nodeInfo)[i].nodePort);
+    raftId.vgId = pSyncNode->vgId;
+
+    if (syncUtilSameId(&raftId, &(pSyncNode->myRaftId))) {
+      b2 = true;
+      break;
+    }
+  }
+
+  ASSERT(b1 == b2);
+  return b1;
+}
+
+void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex lastConfigChangeIndex, bool* isDrop) {
   SSyncCfg oldConfig = pSyncNode->pRaftCfg->cfg;
-  pSyncNode->pRaftCfg->cfg = *newConfig;
+  pSyncNode->pRaftCfg->cfg = *pNewConfig;
   pSyncNode->pRaftCfg->lastConfigIndex = lastConfigChangeIndex;
 
   int32_t ret = 0;
@@ -1270,7 +1303,8 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l
   SSyncSnapshotSender* oldSenders[TSDB_MAX_REPLICA];
   for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
     oldSenders[i] = (pSyncNode->senders)[i];
-    sDebug("vgId:%d sync event save senders %d, %p", pSyncNode->vgId, i, oldSenders[i]);
+    sDebug("vgId:%d sync event save senders %d, %p, privateTerm:%lu", pSyncNode->vgId, i, oldSenders[i],
+           oldSenders[i]->privateTerm);
     if (gRaftDetailLog) {
       ;
     }
@@ -1322,8 +1356,8 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l
         char     host[128];
         uint16_t port;
         syncUtilU642Addr((pSyncNode->replicasId)[i].addr, host, sizeof(host), &port);
-        sDebug("vgId:%d sync event reset sender for %lu, newIndex:%d, %s:%d, %p", pSyncNode->vgId,
-               (pSyncNode->replicasId)[i].addr, i, host, port, oldSenders[j]);
+        sDebug("vgId:%d sync event reset sender for %lu, newIndex:%d, %s:%d, %p, privateTerm:%lu", pSyncNode->vgId,
+               (pSyncNode->replicasId)[i].addr, i, host, port, oldSenders[j], oldSenders[j]->privateTerm);
         (pSyncNode->senders)[i] = oldSenders[j];
         oldSenders[j] = NULL;
         reset = true;
@@ -1341,7 +1375,8 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l
   for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
     if ((pSyncNode->senders)[i] == NULL) {
       (pSyncNode->senders)[i] = snapshotSenderCreate(pSyncNode, i);
-      sDebug("vgId:%d sync event create new sender %p replicaIndex:%d", pSyncNode->vgId, (pSyncNode->senders)[i], i);
+      sDebug("vgId:%d sync event create new sender %p replicaIndex:%d, privateTerm:%lu", pSyncNode->vgId,
+             (pSyncNode->senders)[i], i, (pSyncNode->senders)[i]->privateTerm);
     }
   }
 
@@ -1354,8 +1389,10 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l
     }
   }
 
-  bool IamInOld = false;
-  bool IamInNew = false;
+  bool IamInOld = syncNodeInConfig(pSyncNode, &oldConfig);
+  bool IamInNew = syncNodeInConfig(pSyncNode, pNewConfig);
+
+#if 0 
   for (int i = 0; i < oldConfig.replicaNum; ++i) {
     if (strcmp((oldConfig.nodeInfo)[i].nodeFqdn, pSyncNode->myNodeInfo.nodeFqdn) == 0 &&
         (oldConfig.nodeInfo)[i].nodePort == pSyncNode->myNodeInfo.nodePort) {
@@ -1371,6 +1408,7 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l
       break;
     }
   }
+#endif
 
   *isDrop = true;
   if (IamInOld && !IamInNew) {
@@ -1379,6 +1417,10 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l
     *isDrop = false;
   }
 
+  // may be add me to a new raft group
+  if (IamInOld && IamInNew && oldConfig.replicaNum == 1) {
+  }
+
   if (IamInNew) {
     pSyncNode->pRaftCfg->isStandBy = 0;  // change isStandBy to normal
   }
@@ -1404,14 +1446,17 @@ void syncNodeRelease(SSyncNode* pNode) { taosReleaseRef(tsNodeRefId, pNode->rid)
 void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term) {
   if (term > pSyncNode->pRaftStore->currentTerm) {
     raftStoreSetTerm(pSyncNode->pRaftStore, term);
-    syncNodeBecomeFollower(pSyncNode, "update term");
+    char tmpBuf[64];
+    snprintf(tmpBuf, sizeof(tmpBuf), "update term to %lu", term);
+    syncNodeBecomeFollower(pSyncNode, tmpBuf);
     raftStoreClearVote(pSyncNode->pRaftStore);
   }
 }
 
 void syncNodeBecomeFollower(SSyncNode* pSyncNode, const char* debugStr) {
-  sDebug("vgId:%d sync event become follower, isStandBy:%d, replicaNum:%d, %s", pSyncNode->vgId,
-         pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum, debugStr);
+  sDebug("vgId:%d sync event become follower, isStandBy:%d, currentTerm:%lu, replicaNum:%d, restoreFinish:%d, %s",
+         pSyncNode->vgId, pSyncNode->pRaftCfg->isStandBy, pSyncNode->pRaftStore->currentTerm, pSyncNode->replicaNum,
+         pSyncNode->restoreFinish, debugStr);
 
   // maybe clear leader cache
   if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
@@ -1445,8 +1490,12 @@ void syncNodeBecomeFollower(SSyncNode* pSyncNode, const char* debugStr) {
 //     /\ UNCHANGED <<messages, currentTerm, votedFor, candidateVars, logVars>>
 //
 void syncNodeBecomeLeader(SSyncNode* pSyncNode, const char* debugStr) {
-  sDebug("vgId:%d sync event become leader, isStandBy:%d, replicaNum:%d %s", pSyncNode->vgId,
-         pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum, debugStr);
+  // reset restoreFinish
+  pSyncNode->restoreFinish = false;
+
+  sDebug("vgId:%d sync event become leader, isStandBy:%d, currentTerm:%lu, replicaNum:%d, restoreFinish:%d, %s",
+         pSyncNode->vgId, pSyncNode->pRaftCfg->isStandBy, pSyncNode->pRaftStore->currentTerm, pSyncNode->replicaNum,
+         pSyncNode->restoreFinish, debugStr);
 
   // state change
   pSyncNode->state = TAOS_SYNC_STATE_LEADER;
@@ -2028,11 +2077,11 @@ static int32_t syncDoLeaderTransfer(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftE
            pSyncLeaderTransfer->newLeaderId.addr);
   */
 
-  sDebug("vgId:%d sync event, begin leader transfer", ths->vgId);
+  sDebug("vgId:%d sync event begin leader transfer", ths->vgId);
 
   if (strcmp(pSyncLeaderTransfer->newNodeInfo.nodeFqdn, ths->myNodeInfo.nodeFqdn) == 0 &&
       pSyncLeaderTransfer->newNodeInfo.nodePort == ths->myNodeInfo.nodePort) {
-    sDebug("vgId:%d sync event, maybe leader transfer to %s:%d %lu", ths->vgId,
+    sDebug("vgId:%d sync event maybe leader transfer to %s:%d %lu", ths->vgId,
            pSyncLeaderTransfer->newNodeInfo.nodeFqdn, pSyncLeaderTransfer->newNodeInfo.nodePort,
            pSyncLeaderTransfer->newLeaderId.addr);
 
@@ -2067,6 +2116,21 @@ static int32_t syncDoLeaderTransfer(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftE
   return 0;
 }
 
+int32_t syncNodeUpdateNewConfigIndex(SSyncNode* ths, SSyncCfg* pNewCfg) {
+  for (int i = 0; i < pNewCfg->replicaNum; ++i) {
+    SRaftId raftId;
+    raftId.addr = syncUtilAddr2U64((pNewCfg->nodeInfo)[i].nodeFqdn, (pNewCfg->nodeInfo)[i].nodePort);
+    raftId.vgId = ths->vgId;
+
+    if (syncUtilSameId(&(ths->myRaftId), &raftId)) {
+      pNewCfg->myIndex = i;
+      return 0;
+    }
+  }
+
+  return -1;
+}
+
 static int32_t syncNodeConfigChange(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* pEntry) {
   SSyncCfg oldSyncCfg = ths->pRaftCfg->cfg;
 
@@ -2075,19 +2139,23 @@ static int32_t syncNodeConfigChange(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftE
   ASSERT(ret == 0);
 
   // update new config myIndex
-  bool IamInNew = false;
-  for (int i = 0; i < newSyncCfg.replicaNum; ++i) {
-    if (strcmp(ths->myNodeInfo.nodeFqdn, (newSyncCfg.nodeInfo)[i].nodeFqdn) == 0 &&
-        ths->myNodeInfo.nodePort == (newSyncCfg.nodeInfo)[i].nodePort) {
-      newSyncCfg.myIndex = i;
-      IamInNew = true;
-      break;
-    }
-  }
+  syncNodeUpdateNewConfigIndex(ths, &newSyncCfg);
+
+  bool IamInNew = syncNodeInConfig(ths, &newSyncCfg);
+
+  /*
+   for (int i = 0; i < newSyncCfg.replicaNum; ++i) {
+     if (strcmp(ths->myNodeInfo.nodeFqdn, (newSyncCfg.nodeInfo)[i].nodeFqdn) == 0 &&
+         ths->myNodeInfo.nodePort == (newSyncCfg.nodeInfo)[i].nodePort) {
+       newSyncCfg.myIndex = i;
+       IamInNew = true;
+       break;
+     }
+   }
+ */
 
   bool isDrop;
 
-  // if (IamInNew || (!IamInNew && ths->state != TAOS_SYNC_STATE_LEADER)) {
   if (IamInNew) {
     syncNodeUpdateConfig(ths, &newSyncCfg, pEntry->index, &isDrop);
 
@@ -2186,7 +2254,8 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
               ths->pFsm->FpRestoreFinishCb(ths->pFsm);
             }
             ths->restoreFinish = true;
-            sDebug("vgId:%d sync event restore finish, index:%ld", ths->vgId, pEntry->index);
+            sDebug("vgId:%d sync event restore finish, %s, index:%ld", ths->vgId, syncUtilState2String(ths->state),
+                   pEntry->index);
           }
         }
 
diff --git a/source/libs/sync/src/syncRaftLog.c b/source/libs/sync/src/syncRaftLog.c
index 996cd12e4a..79ce84abe9 100644
--- a/source/libs/sync/src/syncRaftLog.c
+++ b/source/libs/sync/src/syncRaftLog.c
@@ -162,9 +162,10 @@ static int32_t raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntr
 
   walFsync(pWal, true);
 
-  sDebug("vgId:%d sync event write index:%ld, %s, isStandBy:%d, msgType:%s, originalRpcType:%s", pData->pSyncNode->vgId,
-         pEntry->index, syncUtilState2String(pData->pSyncNode->state), pData->pSyncNode->pRaftCfg->isStandBy,
-         TMSG_INFO(pEntry->msgType), TMSG_INFO(pEntry->originalRpcType));
+  sDebug("vgId:%d sync event write index:%ld, %s, isStandBy:%d, msgType:%s,%d, originalRpcType:%s,%d",
+         pData->pSyncNode->vgId, pEntry->index, syncUtilState2String(pData->pSyncNode->state),
+         pData->pSyncNode->pRaftCfg->isStandBy, TMSG_INFO(pEntry->msgType), pEntry->msgType,
+         TMSG_INFO(pEntry->originalRpcType), pEntry->originalRpcType);
 
   return code;
 }
diff --git a/source/libs/sync/src/syncRespMgr.c b/source/libs/sync/src/syncRespMgr.c
index 5087cacd02..adc67b5d8d 100644
--- a/source/libs/sync/src/syncRespMgr.c
+++ b/source/libs/sync/src/syncRespMgr.c
@@ -45,8 +45,9 @@ int64_t syncRespMgrAdd(SSyncRespMgr *pObj, SRespStub *pStub) {
   taosHashPut(pObj->pRespHash, &keyCode, sizeof(keyCode), pStub, sizeof(SRespStub));
 
   SSyncNode *pSyncNode = pObj->data;
-  sDebug("vgId:%d sync event resp mgr add, type:%s seq:%lu handle:%p", pSyncNode->vgId,
-         TMSG_INFO(pStub->rpcMsg.msgType), keyCode, pStub->rpcMsg.info.handle);
+  sDebug("vgId:%d sync event resp mgr add, msgType:%s,%d seq:%lu handle:%p ahandle:%p", pSyncNode->vgId,
+         TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType, keyCode, pStub->rpcMsg.info.handle,
+         pStub->rpcMsg.info.ahandle);
 
   taosThreadMutexUnlock(&(pObj->mutex));
   return keyCode;
@@ -69,8 +70,9 @@ int32_t syncRespMgrGet(SSyncRespMgr *pObj, uint64_t index, SRespStub *pStub) {
     memcpy(pStub, pTmp, sizeof(SRespStub));
 
     SSyncNode *pSyncNode = pObj->data;
-    sDebug("vgId:%d sync event resp mgr get, type:%s seq:%lu handle:%p", pSyncNode->vgId,
-           TMSG_INFO(pStub->rpcMsg.msgType), index, pStub->rpcMsg.info.handle);
+    sDebug("vgId:%d sync event resp mgr get, msgType:%s,%d seq:%lu handle:%p ahandle:%p", pSyncNode->vgId,
+           TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType, index, pStub->rpcMsg.info.handle,
+           pStub->rpcMsg.info.ahandle);
 
     taosThreadMutexUnlock(&(pObj->mutex));
     return 1;  // get one object
@@ -87,8 +89,9 @@ int32_t syncRespMgrGetAndDel(SSyncRespMgr *pObj, uint64_t index, SRespStub *pStu
     memcpy(pStub, pTmp, sizeof(SRespStub));
 
     SSyncNode *pSyncNode = pObj->data;
-    sDebug("vgId:%d sync event resp mgr get and del, type:%s seq:%lu handle:%p", pSyncNode->vgId,
-           TMSG_INFO(pStub->rpcMsg.msgType), index, pStub->rpcMsg.info.handle);
+    sDebug("vgId:%d sync event resp mgr get and del, msgType:%s,%d seq:%lu handle:%p ahandle:%p", pSyncNode->vgId,
+           TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType, index, pStub->rpcMsg.info.handle,
+           pStub->rpcMsg.info.ahandle);
 
     taosHashRemove(pObj->pRespHash, &index, sizeof(index));
     taosThreadMutexUnlock(&(pObj->mutex));
diff --git a/source/libs/sync/src/syncSnapshot.c b/source/libs/sync/src/syncSnapshot.c
index c92920a936..545ea953f8 100644
--- a/source/libs/sync/src/syncSnapshot.c
+++ b/source/libs/sync/src/syncSnapshot.c
@@ -591,6 +591,12 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
         if (pMsg->lastConfigIndex >= SYNC_INDEX_BEGIN) {
           int32_t oldReplicaNum = pSyncNode->replicaNum;
 
+          // update new config myIndex
+          SSyncCfg newSyncCfg = pMsg->lastConfig;
+          syncNodeUpdateNewConfigIndex(pSyncNode, &newSyncCfg);
+          bool IamInNew = syncNodeInConfig(pSyncNode, &newSyncCfg);
+
+#if 0
           // update new config myIndex
           bool     IamInNew = false;
           SSyncCfg newSyncCfg = pMsg->lastConfig;
@@ -602,6 +608,7 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
               break;
             }
           }
+#endif
 
           bool isDrop;
           if (IamInNew) {
diff --git a/source/libs/sync/test/syncTestTool.cpp b/source/libs/sync/test/syncTestTool.cpp
index 21049454f4..1eac372bd4 100644
--- a/source/libs/sync/test/syncTestTool.cpp
+++ b/source/libs/sync/test/syncTestTool.cpp
@@ -297,7 +297,7 @@ void usage(char* exe) {
 SRpcMsg* createRpcMsg(int i, int count, int myIndex) {
   SRpcMsg* pMsg = (SRpcMsg*)taosMemoryMalloc(sizeof(SRpcMsg));
   memset(pMsg, 0, sizeof(SRpcMsg));
-  pMsg->msgType = 9999;
+  pMsg->msgType = TDMT_VND_SUBMIT;
   pMsg->contLen = 256;
   pMsg->pCont = rpcMallocCont(pMsg->contLen);
   snprintf((char*)(pMsg->pCont), pMsg->contLen, "value-myIndex:%u-%d-%d-%ld", myIndex, i, count, taosGetTimestampMs());
@@ -384,8 +384,10 @@ int main(int argc, char** argv) {
 
     leaderTransferWait++;
     if (leaderTransferWait == 7) {
-      sTrace("begin leader transfer ...");
-      int32_t ret = syncLeaderTransfer(rid);
+      if (leaderTransfer) {
+        sTrace("begin leader transfer ...");
+        int32_t ret = syncLeaderTransfer(rid);
+      }
     }
 
     if (alreadySend < writeRecordNum) {

From 54921550097c78bd3e769b5797897a4aa720cd8d Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 14:49:37 +0800
Subject: [PATCH 16/40] fix(query): fix syntax error on windows platform.

---
 source/libs/executor/src/sortoperator.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/source/libs/executor/src/sortoperator.c b/source/libs/executor/src/sortoperator.c
index 98a1aacbb6..2f2080d5fe 100644
--- a/source/libs/executor/src/sortoperator.c
+++ b/source/libs/executor/src/sortoperator.c
@@ -439,7 +439,7 @@ SOperatorInfo* createMultiwaySortMergeOperatorInfo(SOperatorInfo** downStreams,
   pInfo->bufPageSize   = getProperSortPageSize(rowSize);
 
   uint32_t numOfSources = taosArrayGetSize(pSortInfo);
-  numOfSources = MAX(2, numOfSources);
+  numOfSources = TMAX(2, numOfSources);
 
   pInfo->sortBufSize    = numOfSources * pInfo->bufPageSize;
 

From d41efb2cfb284be46997466601b929b0fc4527db Mon Sep 17 00:00:00 2001
From: dapan1121 <wpan@taosdata.com>
Date: Thu, 16 Jun 2022 15:02:57 +0800
Subject: [PATCH 17/40] fix rc issue

---
 source/libs/scheduler/src/schJob.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/source/libs/scheduler/src/schJob.c b/source/libs/scheduler/src/schJob.c
index 3e3f3639d7..31536f413d 100644
--- a/source/libs/scheduler/src/schJob.c
+++ b/source/libs/scheduler/src/schJob.c
@@ -1488,13 +1488,15 @@ int32_t schExecJobImpl(SSchedulerReq *pReq, int64_t *job, SQueryResult* pRes, bo
   qDebug("QID:0x%" PRIx64 " job refId 0x%"PRIx64 " started", pReq->pDag->queryId, pJob->refId);
   *job = pJob->refId;
 
+  if (!sync) {
+    pJob->userCb = SCH_EXEC_CB; 
+  }
+
   SCH_ERR_JRET(schLaunchJob(pJob));
 
   if (sync) {
     SCH_JOB_DLOG("will wait for rsp now, job status:%s", SCH_GET_JOB_STATUS_STR(pJob));
     tsem_wait(&pJob->rspSem);
-  } else {
-    pJob->userCb = SCH_EXEC_CB; 
   }
 
   SCH_JOB_DLOG("job exec done, job status:%s, jobId:0x%"PRIx64, SCH_GET_JOB_STATUS_STR(pJob), pJob->refId);

From 7aa087a48d4c5ae48e597f27c35f06b9964e5d44 Mon Sep 17 00:00:00 2001
From: Shengliang Guan <slguan@taosdata.com>
Date: Thu, 16 Jun 2022 15:20:46 +0800
Subject: [PATCH 18/40] feat: redistribute vgroup

---
 include/util/taoserror.h               |   8 +-
 source/dnode/mnode/impl/src/mndMain.c  |   2 +-
 source/dnode/mnode/impl/src/mndSync.c  |   8 --
 source/dnode/vnode/src/vnd/vnodeSvr.c  | 102 ---------------
 source/dnode/vnode/src/vnd/vnodeSync.c | 166 +++++++++++++++++++++----
 source/libs/sync/src/syncMain.c        |   2 +-
 source/util/src/terror.c               |   2 +-
 7 files changed, 152 insertions(+), 138 deletions(-)

diff --git a/include/util/taoserror.h b/include/util/taoserror.h
index 4f93ce9e4a..fcbe5cd878 100644
--- a/include/util/taoserror.h
+++ b/include/util/taoserror.h
@@ -410,10 +410,10 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_SYN_INVALID_CHECKSUM          TAOS_DEF_ERROR_CODE(0, 0x0908)
 #define TSDB_CODE_SYN_INVALID_MSGLEN            TAOS_DEF_ERROR_CODE(0, 0x0909)
 #define TSDB_CODE_SYN_INVALID_MSGTYPE           TAOS_DEF_ERROR_CODE(0, 0x090A)
-
-#define TSDB_CODE_SYN_NOT_LEADER                TAOS_DEF_ERROR_CODE(0, 0x0910)
-#define TSDB_CODE_SYN_ONE_REPLICA               TAOS_DEF_ERROR_CODE(0, 0x0911)
-#define TSDB_CODE_SYN_NOT_IN_NEW_CONFIG         TAOS_DEF_ERROR_CODE(0, 0x0912)
+#define TSDB_CODE_SYN_IS_LEADER                 TAOS_DEF_ERROR_CODE(0, 0x090B)
+#define TSDB_CODE_SYN_NOT_LEADER                TAOS_DEF_ERROR_CODE(0, 0x090C)
+#define TSDB_CODE_SYN_ONE_REPLICA               TAOS_DEF_ERROR_CODE(0, 0x090D)
+#define TSDB_CODE_SYN_NOT_IN_NEW_CONFIG         TAOS_DEF_ERROR_CODE(0, 0x090E)
 #define TSDB_CODE_SYN_INTERNAL_ERROR            TAOS_DEF_ERROR_CODE(0, 0x09FF)
 
 // tq
diff --git a/source/dnode/mnode/impl/src/mndMain.c b/source/dnode/mnode/impl/src/mndMain.c
index e141e7dfb3..f277a5169d 100644
--- a/source/dnode/mnode/impl/src/mndMain.c
+++ b/source/dnode/mnode/impl/src/mndMain.c
@@ -397,7 +397,7 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
 
   char  logBuf[512] = {0};
   char *syncNodeStr = sync2SimpleStr(pMgmt->sync);
-  snprintf(logBuf, sizeof(logBuf), "==vnodeProcessSyncReq== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
+  snprintf(logBuf, sizeof(logBuf), "==mndProcessSyncMsg== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
   static int64_t mndTick = 0;
   if (++mndTick % 10 == 1) {
     mTrace("sync trace msg:%s, %s", TMSG_INFO(pMsg->msgType), syncNodeStr);
diff --git a/source/dnode/mnode/impl/src/mndSync.c b/source/dnode/mnode/impl/src/mndSync.c
index 63708aef8a..e7b4489ed9 100644
--- a/source/dnode/mnode/impl/src/mndSync.c
+++ b/source/dnode/mnode/impl/src/mndSync.c
@@ -256,14 +256,6 @@ void mndSyncStart(SMnode *pMnode) {
   syncSetMsgCb(pMgmt->sync, &pMnode->msgCb);
   syncStart(pMgmt->sync);
   mDebug("mnode sync started, id:%" PRId64 " standby:%d", pMgmt->sync, pMgmt->standby);
-
-  /*
-    if (pMgmt->standby) {
-      syncStartStandBy(pMgmt->sync);
-    } else {
-      syncStart(pMgmt->sync);
-    }
-  */
 }
 
 void mndSyncStop(SMnode *pMnode) {}
diff --git a/source/dnode/vnode/src/vnd/vnodeSvr.c b/source/dnode/vnode/src/vnd/vnodeSvr.c
index c981f6cb99..d170931e7e 100644
--- a/source/dnode/vnode/src/vnd/vnodeSvr.c
+++ b/source/dnode/vnode/src/vnd/vnodeSvr.c
@@ -300,108 +300,6 @@ void vnodeUpdateMetaRsp(SVnode *pVnode, STableMetaRsp *pMetaRsp) {
   pMetaRsp->precision = pVnode->config.tsdbCfg.precision;
 }
 
-int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
-  int32_t ret = 0;
-
-  if (syncEnvIsStart()) {
-    SSyncNode *pSyncNode = syncNodeAcquire(pVnode->sync);
-    assert(pSyncNode != NULL);
-
-    ESyncState state = syncGetMyRole(pVnode->sync);
-    SyncTerm   currentTerm = syncGetMyTerm(pVnode->sync);
-
-    SMsgHead *pHead = pMsg->pCont;
-
-    char  logBuf[512] = {0};
-    char *syncNodeStr = sync2SimpleStr(pVnode->sync);
-    snprintf(logBuf, sizeof(logBuf), "==vnodeProcessSyncReq== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
-    static int64_t vndTick = 0;
-    if (++vndTick % 10 == 1) {
-      vTrace("sync trace msg:%s, %s", TMSG_INFO(pMsg->msgType), syncNodeStr);
-    }
-    syncRpcMsgLog2(logBuf, pMsg);
-    taosMemoryFree(syncNodeStr);
-
-    SRpcMsg *pRpcMsg = pMsg;
-
-    if (pRpcMsg->msgType == TDMT_SYNC_TIMEOUT) {
-      SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
-      syncTimeoutDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_PING) {
-      SyncPing *pSyncMsg = syncPingFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnPingCb(pSyncNode, pSyncMsg);
-      syncPingDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_PING_REPLY) {
-      SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
-      syncPingReplyDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
-      SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
-      syncClientRequestDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
-      SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
-      syncRequestVoteDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
-      SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
-      syncRequestVoteReplyDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
-      SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
-      syncAppendEntriesDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
-      SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
-
-      ret = syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
-      syncAppendEntriesReplyDestroy(pSyncMsg);
-
-    } else if (pRpcMsg->msgType == TDMT_SYNC_SET_VNODE_STANDBY) {
-      ret = syncSetStandby(pVnode->sync);
-      vInfo("vgId:%d, set standby result:0x%x rid:%" PRId64, pVnode->config.vgId, ret, pVnode->sync);
-      SRpcMsg rsp = {.code = ret, .info = pMsg->info};
-      tmsgSendRsp(&rsp);
-    } else {
-      vError("==vnodeProcessSyncReq== error msg type:%d", pRpcMsg->msgType);
-      ret = -1;
-    }
-
-    syncNodeRelease(pSyncNode);
-  } else {
-    vError("==vnodeProcessSyncReq== error syncEnv stop");
-    ret = -1;
-  }
-
-  if (ret != 0) {
-    terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
-  }
-  return ret;
-}
-
 static int32_t vnodeProcessCreateStbReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp) {
   SVCreateStbReq req = {0};
   SDecoder       coder;
diff --git a/source/dnode/vnode/src/vnd/vnodeSync.c b/source/dnode/vnode/src/vnd/vnodeSync.c
index 25661d2e69..8662c8369b 100644
--- a/source/dnode/vnode/src/vnd/vnodeSync.c
+++ b/source/dnode/vnode/src/vnd/vnodeSync.c
@@ -50,6 +50,33 @@ static inline void vnodePostBlockMsg(SVnode *pVnode, tmsg_t type) {
   }
 }
 
+static int32_t vnodeSetStandBy(SVnode *pVnode) {
+  vInfo("vgId:%d, start to set standby", TD_VID(pVnode));
+
+  if (syncSetStandby(pVnode->sync) == 0) {
+    vInfo("vgId:%d, set standby success", TD_VID(pVnode));
+    return 0;
+  } else if (terrno != TSDB_CODE_SYN_IS_LEADER) {
+    vError("vgId:%d, failed to set standby since %s", TD_VID(pVnode), terrstr());
+    return -1;
+  }
+
+  if (syncLeaderTransfer(pVnode->sync) != 0) {
+    vError("vgId:%d, failed to transfer leader since:%s", TD_VID(pVnode), terrstr());
+    return -1;
+  } else {
+    vInfo("vgId:%d, transfer leader success", TD_VID(pVnode));
+  }
+
+  if (syncSetStandby(pVnode->sync) == 0) {
+    vInfo("vgId:%d, set standby success", TD_VID(pVnode));
+    return 0;
+  } else {
+    vError("vgId:%d, failed to set standby since %s", TD_VID(pVnode), terrstr());
+    return -1;
+  }
+}
+
 static int32_t vnodeProcessAlterReplicaReq(SVnode *pVnode, SRpcMsg *pMsg) {
   SAlterVnodeReq req = {0};
   if (tDeserializeSAlterVnodeReq((char *)pMsg->pCont + sizeof(SMsgHead), pMsg->contLen - sizeof(SMsgHead), &req) != 0) {
@@ -74,14 +101,19 @@ static int32_t vnodeProcessAlterReplicaReq(SVnode *pVnode, SRpcMsg *pMsg) {
 
   int32_t code = syncPropose(pVnode->sync, &rpcMsg, false);
   if (code != 0) {
-    vDebug("vgId:%d, failed to propose reconfig msg since %s", TD_VID(pVnode), terrstr());
-    if (syncLeaderTransfer(pVnode->sync) != 0) {
-      vError("vgId:%d, failed to transfer leader since %s", TD_VID(pVnode), terrstr());
-    } else {
-      vDebug("vgId:%d, transfer leader success, propose reconfig config again", TD_VID(pVnode));
+    if (terrno != 0) code = terrno;
+
+    vInfo("vgId:%d, failed to propose reconfig msg since %s", TD_VID(pVnode), terrstr());
+    if (terrno == TSDB_CODE_SYN_IS_LEADER) {
+      if (syncLeaderTransfer(pVnode->sync) != 0) {
+        vError("vgId:%d, failed to transfer leader since %s", TD_VID(pVnode), terrstr());
+      } else {
+        vInfo("vgId:%d, transfer leader success", TD_VID(pVnode));
+      }
     }
   }
 
+  terrno = code;
   return code;
 }
 
@@ -108,7 +140,6 @@ void vnodeProposeMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs) {
 
     if (code == 0) {
       vnodeAccumBlockMsg(pVnode, pMsg->msgType);
-
     } else if (code == -1 && terrno == TSDB_CODE_SYN_NOT_LEADER) {
       SEpSet newEpSet = {0};
       syncGetEpSet(pVnode->sync, &newEpSet);
@@ -170,6 +201,108 @@ void vnodeApplyMsg(SQueueInfo *pInfo, STaosQall *qall, int32_t numOfMsgs) {
   }
 }
 
+int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
+  int32_t ret = 0;
+
+  if (syncEnvIsStart()) {
+    SSyncNode *pSyncNode = syncNodeAcquire(pVnode->sync);
+    assert(pSyncNode != NULL);
+
+    ESyncState state = syncGetMyRole(pVnode->sync);
+    SyncTerm   currentTerm = syncGetMyTerm(pVnode->sync);
+
+    SMsgHead *pHead = pMsg->pCont;
+
+    char  logBuf[512] = {0};
+    char *syncNodeStr = sync2SimpleStr(pVnode->sync);
+    snprintf(logBuf, sizeof(logBuf), "==vnodeProcessSyncReq== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
+    static int64_t vndTick = 0;
+    if (++vndTick % 10 == 1) {
+      vTrace("sync trace msg:%s, %s", TMSG_INFO(pMsg->msgType), syncNodeStr);
+    }
+    syncRpcMsgLog2(logBuf, pMsg);
+    taosMemoryFree(syncNodeStr);
+
+    SRpcMsg *pRpcMsg = pMsg;
+
+    if (pRpcMsg->msgType == TDMT_SYNC_TIMEOUT) {
+      SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
+      syncTimeoutDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_PING) {
+      SyncPing *pSyncMsg = syncPingFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnPingCb(pSyncNode, pSyncMsg);
+      syncPingDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_PING_REPLY) {
+      SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
+      syncPingReplyDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
+      SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
+      syncClientRequestDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
+      SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
+      syncRequestVoteDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
+      SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
+      syncRequestVoteReplyDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
+      SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
+      syncAppendEntriesDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
+      SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pRpcMsg);
+      assert(pSyncMsg != NULL);
+
+      ret = syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
+      syncAppendEntriesReplyDestroy(pSyncMsg);
+
+    } else if (pRpcMsg->msgType == TDMT_SYNC_SET_VNODE_STANDBY) {
+      ret = vnodeSetStandBy(pVnode);
+      if (ret != 0 && terrno != 0) ret = terrno;
+      SRpcMsg rsp = {.code = ret, .info = pMsg->info};
+      tmsgSendRsp(&rsp);
+    } else {
+      vError("==vnodeProcessSyncReq== error msg type:%d", pRpcMsg->msgType);
+      ret = -1;
+    }
+
+    syncNodeRelease(pSyncNode);
+  } else {
+    vError("==vnodeProcessSyncReq== error syncEnv stop");
+    ret = -1;
+  }
+
+  if (ret != 0) {
+    terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
+  }
+  return ret;
+}
+
 static int32_t vnodeSyncEqMsg(const SMsgCb *msgcb, SRpcMsg *pMsg) {
   int32_t code = tmsgPutToQueue(msgcb, SYNC_QUEUE, pMsg);
   if (code != 0) {
@@ -258,17 +391,17 @@ static void vnodeSyncRollBackMsg(SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta
   syncRpcMsgLog2(logBuf, (SRpcMsg *)pMsg);
 }
 
-int32_t vnodeSnapshotStartRead(struct SSyncFSM *pFsm, void **ppReader) { return 0; }
+static int32_t vnodeSnapshotStartRead(struct SSyncFSM *pFsm, void **ppReader) { return 0; }
 
-int32_t vnodeSnapshotStopRead(struct SSyncFSM *pFsm, void *pReader) { return 0; }
+static int32_t vnodeSnapshotStopRead(struct SSyncFSM *pFsm, void *pReader) { return 0; }
 
-int32_t vnodeSnapshotDoRead(struct SSyncFSM *pFsm, void *pReader, void **ppBuf, int32_t *len) { return 0; }
+static int32_t vnodeSnapshotDoRead(struct SSyncFSM *pFsm, void *pReader, void **ppBuf, int32_t *len) { return 0; }
 
-int32_t vnodeSnapshotStartWrite(struct SSyncFSM *pFsm, void **ppWriter) { return 0; }
+static int32_t vnodeSnapshotStartWrite(struct SSyncFSM *pFsm, void **ppWriter) { return 0; }
 
-int32_t vnodeSnapshotStopWrite(struct SSyncFSM *pFsm, void *pWriter, bool isApply) { return 0; }
+static int32_t vnodeSnapshotStopWrite(struct SSyncFSM *pFsm, void *pWriter, bool isApply) { return 0; }
 
-int32_t vnodeSnapshotDoWrite(struct SSyncFSM *pFsm, void *pWriter, void *pBuf, int32_t len) { return 0; }
+static int32_t vnodeSnapshotDoWrite(struct SSyncFSM *pFsm, void *pWriter, void *pBuf, int32_t len) { return 0; }
 
 static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
   SSyncFSM *pFsm = taosMemoryCalloc(1, sizeof(SSyncFSM));
@@ -279,7 +412,6 @@ static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
   pFsm->FpGetSnapshot = vnodeSyncGetSnapshot;
   pFsm->FpRestoreFinishCb = NULL;
   pFsm->FpReConfigCb = vnodeSyncReconfig;
-
   pFsm->FpSnapshotStartRead = vnodeSnapshotStartRead;
   pFsm->FpSnapshotStopRead = vnodeSnapshotStopRead;
   pFsm->FpSnapshotDoRead = vnodeSnapshotDoRead;
@@ -292,7 +424,6 @@ static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
 
 int32_t vnodeSyncOpen(SVnode *pVnode, char *path) {
   SSyncInfo syncInfo = {
-      .isStandBy = false,
       .snapshotEnable = false,
       .vgId = pVnode->config.vgId,
       .isStandBy = pVnode->config.standby,
@@ -321,13 +452,6 @@ int32_t vnodeSyncOpen(SVnode *pVnode, char *path) {
 void vnodeSyncStart(SVnode *pVnode) {
   syncSetMsgCb(pVnode->sync, &pVnode->msgCb);
   syncStart(pVnode->sync);
-  /*
-  if (pVnode->config.standby) {
-    syncStartStandBy(pVnode->sync);
-  } else {
-    syncStart(pVnode->sync);
-  }
-  */
 }
 
 void vnodeSyncClose(SVnode *pVnode) { syncStop(pVnode->sync); }
diff --git a/source/libs/sync/src/syncMain.c b/source/libs/sync/src/syncMain.c
index 85bdf18ab4..03a4f7b2b3 100644
--- a/source/libs/sync/src/syncMain.c
+++ b/source/libs/sync/src/syncMain.c
@@ -156,7 +156,7 @@ int32_t syncSetStandby(int64_t rid) {
 
   if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
     taosReleaseRef(tsNodeRefId, pSyncNode->rid);
-    terrno = TSDB_CODE_SYN_INTERNAL_ERROR;
+    terrno = TSDB_CODE_SYN_IS_LEADER;
     sError("failed to set standby since it is leader, rid:%" PRId64, rid);
     return -1;
   }
diff --git a/source/util/src/terror.c b/source/util/src/terror.c
index fe040df6fc..596aab0e09 100644
--- a/source/util/src/terror.c
+++ b/source/util/src/terror.c
@@ -412,7 +412,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_SYN_MISMATCHED_SIGNATURE,     "Mismatched signature"
 TAOS_DEFINE_ERROR(TSDB_CODE_SYN_INVALID_CHECKSUM,         "Invalid msg checksum")
 TAOS_DEFINE_ERROR(TSDB_CODE_SYN_INVALID_MSGLEN,           "Invalid msg length")
 TAOS_DEFINE_ERROR(TSDB_CODE_SYN_INVALID_MSGTYPE,          "Invalid msg type")
-
+TAOS_DEFINE_ERROR(TSDB_CODE_SYN_IS_LEADER,                "Sync is leader")
 TAOS_DEFINE_ERROR(TSDB_CODE_SYN_NOT_LEADER,               "Sync not leader")
 TAOS_DEFINE_ERROR(TSDB_CODE_SYN_ONE_REPLICA,              "Sync one replica")
 TAOS_DEFINE_ERROR(TSDB_CODE_SYN_NOT_IN_NEW_CONFIG,        "Sync not in new config")

From 855f1842c2760e038c8f3cda43b1ef1455e7a12c Mon Sep 17 00:00:00 2001
From: tangfangzhi <fztang@taosdata.com>
Date: Thu, 16 Jun 2022 15:28:52 +0800
Subject: [PATCH 19/40] ci: allow to run specified cases in windows full test

---
 tests/system-test/test-all.bat | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/tests/system-test/test-all.bat b/tests/system-test/test-all.bat
index 0929b1fc6e..275cbeebbb 100644
--- a/tests/system-test/test-all.bat
+++ b/tests/system-test/test-all.bat
@@ -5,7 +5,12 @@ set /a a=0
 if %1 == full (
     echo Windows Taosd Full Test
     set /a exitNum=0
-    for /F "usebackq tokens=*" %%i in (fulltest.bat) do (
+    del /Q /F failed.txt
+    set caseFile="fulltest.bat"
+    if not "%2" == "" (
+        set caseFile="%2"
+    )
+    for /F "usebackq tokens=*" %%i in (!caseFile!) do (
         for /f "tokens=1* delims= " %%a in ("%%i") do if not "%%a" == "@REM" (
             set /a a+=1
             echo !a! Processing %%i
@@ -13,7 +18,7 @@ if %1 == full (
             set time1=!_timeTemp!
             echo Start at !time!
             call %%i ARG1 > result_!a!.txt 2>error_!a!.txt
-            if errorlevel 1 ( call :colorEcho 0c "failed" &echo. && set /a exitNum=8 ) else ( call :colorEcho 0a "Success" &echo. ) 
+            if errorlevel 1 ( call :colorEcho 0c "failed" &echo. && set /a exitNum=8 && echo %%i >>failed.txt ) else ( call :colorEcho 0a "Success" &echo. )
         )
     )
     exit !exitNum!
@@ -77,4 +82,4 @@ for %%a in (%tt%) do (
    set /a index=index+1
 )
 set /a _timeTemp=(%hh%*60+%mm%)*60+%ss%
-goto :eof
\ No newline at end of file
+goto :eof

From dbef3dfc55d19e1d3e182ef61f23629fe7261eff Mon Sep 17 00:00:00 2001
From: Cary Xu <klxu@taosdata.com>
Date: Thu, 16 Jun 2022 15:34:33 +0800
Subject: [PATCH 20/40] fix: not submit for rsma without rollup

---
 source/common/src/tdatablock.c         | 42 ++++++++++++++++----------
 source/dnode/vnode/src/inc/vnodeInt.h  |  1 +
 source/dnode/vnode/src/sma/smaRollup.c | 14 +++++++--
 3 files changed, 38 insertions(+), 19 deletions(-)

diff --git a/source/common/src/tdatablock.c b/source/common/src/tdatablock.c
index 5a2aaed74e..3ec1b98b16 100644
--- a/source/common/src/tdatablock.c
+++ b/source/common/src/tdatablock.c
@@ -1583,6 +1583,11 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
     int32_t      rowSize = pDataBlock->info.rowSize;
     int64_t      groupId = pDataBlock->info.groupId;
 
+    if (colNum <= 1) {
+      // invalid if only with TS col
+      continue;
+    }
+
     if (rb.nCols != colNum) {
       tdSRowSetTpInfo(&rb, colNum, pTSchema->flen);
     }
@@ -1679,23 +1684,28 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
     msgLen += pSubmitBlk->dataLen;
   }
 
-  (*pReq)->length = msgLen;
+  if (numOfBlks > 0) {
+    (*pReq)->length = msgLen;
 
-  (*pReq)->header.vgId = htonl(vgId);
-  (*pReq)->header.contLen = htonl(msgLen);
-  (*pReq)->length = (*pReq)->header.contLen;
-  (*pReq)->numOfBlocks = htonl(numOfBlks);
-  SSubmitBlk* blk = (SSubmitBlk*)((*pReq) + 1);
-  while (numOfBlks--) {
-    int32_t dataLen = blk->dataLen;
-    blk->uid = htobe64(blk->uid);
-    blk->suid = htobe64(blk->suid);
-    blk->padding = htonl(blk->padding);
-    blk->sversion = htonl(blk->sversion);
-    blk->dataLen = htonl(blk->dataLen);
-    blk->schemaLen = htonl(blk->schemaLen);
-    blk->numOfRows = htons(blk->numOfRows);
-    blk = (SSubmitBlk*)(blk->data + dataLen);
+    (*pReq)->header.vgId = htonl(vgId);
+    (*pReq)->header.contLen = htonl(msgLen);
+    (*pReq)->length = (*pReq)->header.contLen;
+    (*pReq)->numOfBlocks = htonl(numOfBlks);
+    SSubmitBlk* blk = (SSubmitBlk*)((*pReq) + 1);
+    while (numOfBlks--) {
+      int32_t dataLen = blk->dataLen;
+      blk->uid = htobe64(blk->uid);
+      blk->suid = htobe64(blk->suid);
+      blk->padding = htonl(blk->padding);
+      blk->sversion = htonl(blk->sversion);
+      blk->dataLen = htonl(blk->dataLen);
+      blk->schemaLen = htonl(blk->schemaLen);
+      blk->numOfRows = htons(blk->numOfRows);
+      blk = (SSubmitBlk*)(blk->data + dataLen);
+    }
+  } else {
+    // no valid rows
+    taosMemoryFreeClear(*pReq);
   }
 
   return TSDB_CODE_SUCCESS;
diff --git a/source/dnode/vnode/src/inc/vnodeInt.h b/source/dnode/vnode/src/inc/vnodeInt.h
index 2fd0815181..553c6a40ab 100644
--- a/source/dnode/vnode/src/inc/vnodeInt.h
+++ b/source/dnode/vnode/src/inc/vnodeInt.h
@@ -260,6 +260,7 @@ struct SSma {
 
 #define SMA_CFG(s)        (&(s)->pVnode->config)
 #define SMA_TSDB_CFG(s)   (&(s)->pVnode->config.tsdbCfg)
+#define SMA_RETENTION(s)  ((SRetention *)&(s)->pVnode->config.tsdbCfg.retentions)
 #define SMA_LOCKED(s)     ((s)->locked)
 #define SMA_META(s)       ((s)->pVnode->pMeta)
 #define SMA_VID(s)        TD_VID((s)->pVnode)
diff --git a/source/dnode/vnode/src/sma/smaRollup.c b/source/dnode/vnode/src/sma/smaRollup.c
index b2dcce8f4c..0c372dfa70 100644
--- a/source/dnode/vnode/src/sma/smaRollup.c
+++ b/source/dnode/vnode/src/sma/smaRollup.c
@@ -400,22 +400,24 @@ static FORCE_INLINE int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int3
   }
 
   if (taosArrayGetSize(pResult) > 0) {
-#if 1
+#if 0
     char flag[10] = {0};
     snprintf(flag, 10, "level %" PRIi8, level);
     blockDebugShowData(pResult, flag);
 #endif
     STsdb      *sinkTsdb = (level == TSDB_RETENTION_L1 ? pSma->pRSmaTsdb1 : pSma->pRSmaTsdb2);
     SSubmitReq *pReq = NULL;
-    if (buildSubmitReqFromDataBlock(&pReq, pResult, pTSchema, SMA_VID(pSma), suid) != 0) {
+    if (buildSubmitReqFromDataBlock(&pReq, pResult, pTSchema, SMA_VID(pSma), suid) < 0) {
       taosArrayDestroy(pResult);
       return TSDB_CODE_FAILED;
     }
-    if (tdProcessSubmitReq(sinkTsdb, INT64_MAX, pReq) != 0) {
+    
+    if (pReq && tdProcessSubmitReq(sinkTsdb, INT64_MAX, pReq) < 0) {
       taosArrayDestroy(pResult);
       taosMemoryFreeClear(pReq);
       return TSDB_CODE_FAILED;
     }
+
     taosMemoryFreeClear(pReq);
   } else {
     smaDebug("vgId:%d, no rsma % " PRIi8 " data generated since %s", SMA_VID(pSma), level, tstrerror(terrno));
@@ -469,6 +471,12 @@ int32_t tdProcessRSmaSubmit(SSma *pSma, void *pMsg, int32_t inputType) {
     return TSDB_CODE_SUCCESS;
   }
 
+  SRetention *pRetention = SMA_RETENTION(pSma);
+  if (!RETENTION_VALID(pRetention + 1)) {
+    // return directly if retention level 1 is invalid
+    return TSDB_CODE_SUCCESS;
+  }
+
   if (inputType == STREAM_DATA_TYPE_SUBMIT_BLOCK) {
     STbUidStore uidStore = {0};
     tdFetchSubmitReqSuids(pMsg, &uidStore);

From 9a9200d28b93ae286f19b530688c3d28dfce94e6 Mon Sep 17 00:00:00 2001
From: Minghao Li <castermode@gmail.com>
Date: Thu, 16 Jun 2022 15:38:40 +0800
Subject: [PATCH 21/40] refactor(sync): add restore finish when become leader
 again

---
 source/libs/sync/src/syncAppendEntries.c      |   7 +-
 source/libs/sync/src/syncAppendEntriesReply.c |  13 +-
 source/libs/sync/src/syncCommit.c             |   5 +-
 source/libs/sync/src/syncMain.c               |  70 +++++-----
 source/libs/sync/src/syncRaftLog.c            |  16 ++-
 source/libs/sync/src/syncRespMgr.c            |  19 +--
 source/libs/sync/src/syncSnapshot.c           | 131 ++++++++++--------
 7 files changed, 147 insertions(+), 114 deletions(-)

diff --git a/source/libs/sync/src/syncAppendEntries.c b/source/libs/sync/src/syncAppendEntries.c
index 08a4081ad3..89b4761212 100644
--- a/source/libs/sync/src/syncAppendEntries.c
+++ b/source/libs/sync/src/syncAppendEntries.c
@@ -713,7 +713,8 @@ static int32_t syncNodeMakeLogSame(SSyncNode* ths, SyncAppendEntries* pMsg) {
   // delete confict entries
   code = ths->pLogStore->syncLogTruncate(ths->pLogStore, delBegin);
   ASSERT(code == 0);
-  sDebug("vgId:%d sync event log truncate, from %ld to %ld", ths->vgId, delBegin, delEnd);
+  sDebug("vgId:%d sync event currentTerm:%lu log truncate, from %ld to %ld", ths->vgId, ths->pRaftStore->currentTerm,
+         delBegin, delEnd);
   logStoreSimpleLog2("after syncNodeMakeLogSame", ths->pLogStore);
 
   return code;
@@ -994,8 +995,8 @@ int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMs
             SyncIndex commitEnd = snapshot.lastApplyIndex;
             ths->commitIndex = snapshot.lastApplyIndex;
 
-            sDebug("vgId:%d sync event commit by snapshot from index:%ld to index:%ld, %s", ths->vgId, commitBegin,
-                   commitEnd, syncUtilState2String(ths->state));
+            sDebug("vgId:%d sync event currentTerm:%lu commit by snapshot from index:%ld to index:%ld, %s", ths->vgId,
+                   ths->pRaftStore->currentTerm, commitBegin, commitEnd, syncUtilState2String(ths->state));
           }
 
           SyncIndex beginIndex = ths->commitIndex + 1;
diff --git a/source/libs/sync/src/syncAppendEntriesReply.c b/source/libs/sync/src/syncAppendEntriesReply.c
index fbe3319675..5290d7d28e 100644
--- a/source/libs/sync/src/syncAppendEntriesReply.c
+++ b/source/libs/sync/src/syncAppendEntriesReply.c
@@ -190,18 +190,19 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
         if (gRaftDetailLog) {
           char* s = snapshotSender2Str(pSender);
           sDebug(
-              "vgId:%d sync event snapshot send to %s:%d start sender first time, lastApplyIndex:%ld lastApplyTerm:%lu "
+              "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d start sender first time, lastApplyIndex:%ld "
+              "lastApplyTerm:%lu "
               "lastConfigIndex:%ld privateTerm:%lu "
               "sender:%s",
-              ths->vgId, host, port, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
-              pSender->snapshot.lastConfigIndex, pSender->privateTerm, s);
+              ths->vgId, ths->pRaftStore->currentTerm, host, port, pSender->snapshot.lastApplyIndex,
+              pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm, s);
           taosMemoryFree(s);
         } else {
           sDebug(
-              "vgId:%d sync event snapshot send to %s:%d start sender first time, lastApplyIndex:%ld "
+              "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d start sender first time, lastApplyIndex:%ld "
               "lastApplyTerm:%lu lastConfigIndex:%ld privateTerm:%lu",
-              ths->vgId, host, port, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
-              pSender->snapshot.lastConfigIndex, pSender->privateTerm);
+              ths->vgId, ths->pRaftStore->currentTerm, host, port, pSender->snapshot.lastApplyIndex,
+              pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm);
         }
       }
 
diff --git a/source/libs/sync/src/syncCommit.c b/source/libs/sync/src/syncCommit.c
index 3424fac5e7..efdb5b24ce 100644
--- a/source/libs/sync/src/syncCommit.c
+++ b/source/libs/sync/src/syncCommit.c
@@ -56,8 +56,9 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
     SyncIndex commitEnd = snapshot.lastApplyIndex;
     pSyncNode->commitIndex = snapshot.lastApplyIndex;
 
-    sDebug("vgId:%d sync event commit by snapshot from index:%ld to index:%ld, %s", pSyncNode->vgId,
-           pSyncNode->commitIndex, snapshot.lastApplyIndex, syncUtilState2String(pSyncNode->state));
+    sDebug("vgId:%d sync event currentTerm:%lu commit by snapshot from index:%ld to index:%ld, %s", pSyncNode->vgId,
+           pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex, snapshot.lastApplyIndex,
+           syncUtilState2String(pSyncNode->state));
   }
 
   // update commit index
diff --git a/source/libs/sync/src/syncMain.c b/source/libs/sync/src/syncMain.c
index 5e3c1adf32..05a41daa9e 100644
--- a/source/libs/sync/src/syncMain.c
+++ b/source/libs/sync/src/syncMain.c
@@ -562,7 +562,8 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
     return -1;
   }
   assert(rid == pSyncNode->rid);
-  sDebug("vgId:%d sync event propose msgType:%s,%d", pSyncNode->vgId, TMSG_INFO(pMsg->msgType), pMsg->msgType);
+  sDebug("vgId:%d sync event currentTerm:%lu propose msgType:%s,%d", pSyncNode->vgId,
+         pSyncNode->pRaftStore->currentTerm, TMSG_INFO(pMsg->msgType), pMsg->msgType);
   ret = syncNodePropose(pSyncNode, pMsg, isWeak);
 
   taosReleaseRef(tsNodeRefId, pSyncNode->rid);
@@ -571,7 +572,8 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
 
 int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak) {
   int32_t ret = 0;
-  sDebug("vgId:%d sync event propose msgType:%s,%d", pSyncNode->vgId, TMSG_INFO(pMsg->msgType), pMsg->msgType);
+  sDebug("vgId:%d sync event currentTerm:%lu propose msgType:%s,%d", pSyncNode->vgId,
+         pSyncNode->pRaftStore->currentTerm, TMSG_INFO(pMsg->msgType), pMsg->msgType);
 
   if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
     SRespStub stub;
@@ -604,8 +606,6 @@ int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak)
 SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
   SSyncInfo* pSyncInfo = (SSyncInfo*)pOldSyncInfo;
 
-  sDebug("vgId:%d sync event sync open", pSyncInfo->vgId);
-
   SSyncNode* pSyncNode = (SSyncNode*)taosMemoryMalloc(sizeof(SSyncNode));
   assert(pSyncNode != NULL);
   memset(pSyncNode, 0, sizeof(SSyncNode));
@@ -819,6 +819,8 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
   // snapshot meta
   // pSyncNode->sMeta.lastConfigIndex = -1;
 
+  sDebug("vgId:%d sync event currentTerm:%lu sync open", pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm);
+
   return pSyncNode;
 }
 
@@ -864,7 +866,7 @@ void syncNodeStartStandBy(SSyncNode* pSyncNode) {
 }
 
 void syncNodeClose(SSyncNode* pSyncNode) {
-  sDebug("vgId:%d sync event sync close", pSyncNode->vgId);
+  sDebug("vgId:%d sync event currentTerm:%lu sync close", pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm);
 
   int32_t ret;
   assert(pSyncNode != NULL);
@@ -1303,8 +1305,8 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex
   SSyncSnapshotSender* oldSenders[TSDB_MAX_REPLICA];
   for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
     oldSenders[i] = (pSyncNode->senders)[i];
-    sDebug("vgId:%d sync event save senders %d, %p, privateTerm:%lu", pSyncNode->vgId, i, oldSenders[i],
-           oldSenders[i]->privateTerm);
+    sDebug("vgId:%d sync event currentTerm:%lu save senders %d, %p, privateTerm:%lu", pSyncNode->vgId,
+           pSyncNode->pRaftStore->currentTerm, i, oldSenders[i], oldSenders[i]->privateTerm);
     if (gRaftDetailLog) {
       ;
     }
@@ -1356,8 +1358,9 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex
         char     host[128];
         uint16_t port;
         syncUtilU642Addr((pSyncNode->replicasId)[i].addr, host, sizeof(host), &port);
-        sDebug("vgId:%d sync event reset sender for %lu, newIndex:%d, %s:%d, %p, privateTerm:%lu", pSyncNode->vgId,
-               (pSyncNode->replicasId)[i].addr, i, host, port, oldSenders[j], oldSenders[j]->privateTerm);
+        sDebug("vgId:%d sync event currentTerm:%lu reset sender for %lu, newIndex:%d, %s:%d, %p, privateTerm:%lu",
+               pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, (pSyncNode->replicasId)[i].addr, i, host, port,
+               oldSenders[j], oldSenders[j]->privateTerm);
         (pSyncNode->senders)[i] = oldSenders[j];
         oldSenders[j] = NULL;
         reset = true;
@@ -1365,8 +1368,9 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex
         // reset replicaIndex
         int32_t oldreplicaIndex = (pSyncNode->senders)[i]->replicaIndex;
         (pSyncNode->senders)[i]->replicaIndex = i;
-        sDebug("vgId:%d sync event udpate replicaIndex from %d to %d, %s:%d, %p, reset:%d", pSyncNode->vgId,
-               oldreplicaIndex, i, host, port, (pSyncNode->senders)[i], reset);
+        sDebug("vgId:%d sync event currentTerm:%lu udpate replicaIndex from %d to %d, %s:%d, %p, reset:%d",
+               pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, oldreplicaIndex, i, host, port,
+               (pSyncNode->senders)[i], reset);
       }
     }
   }
@@ -1375,8 +1379,9 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex
   for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
     if ((pSyncNode->senders)[i] == NULL) {
       (pSyncNode->senders)[i] = snapshotSenderCreate(pSyncNode, i);
-      sDebug("vgId:%d sync event create new sender %p replicaIndex:%d, privateTerm:%lu", pSyncNode->vgId,
-             (pSyncNode->senders)[i], i, (pSyncNode->senders)[i]->privateTerm);
+      sDebug("vgId:%d sync event currentTerm:%lu create new sender %p replicaIndex:%d, privateTerm:%lu",
+             pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, (pSyncNode->senders)[i], i,
+             (pSyncNode->senders)[i]->privateTerm);
     }
   }
 
@@ -1384,7 +1389,8 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex
   for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
     if (oldSenders[i] != NULL) {
       snapshotSenderDestroy(oldSenders[i]);
-      sDebug("vgId:%d sync event delete old sender %p replicaIndex:%d", pSyncNode->vgId, oldSenders[i], i);
+      sDebug("vgId:%d sync event currentTerm:%lu delete old sender %p replicaIndex:%d", pSyncNode->vgId,
+             pSyncNode->pRaftStore->currentTerm, oldSenders[i], i);
       oldSenders[i] = NULL;
     }
   }
@@ -1454,9 +1460,11 @@ void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term) {
 }
 
 void syncNodeBecomeFollower(SSyncNode* pSyncNode, const char* debugStr) {
-  sDebug("vgId:%d sync event become follower, isStandBy:%d, currentTerm:%lu, replicaNum:%d, restoreFinish:%d, %s",
-         pSyncNode->vgId, pSyncNode->pRaftCfg->isStandBy, pSyncNode->pRaftStore->currentTerm, pSyncNode->replicaNum,
-         pSyncNode->restoreFinish, debugStr);
+  sDebug(
+      "vgId:%d sync event currentTerm:%lu become follower, isStandBy:%d, replicaNum:%d, "
+      "restoreFinish:%d, %s",
+      pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum,
+      pSyncNode->restoreFinish, debugStr);
 
   // maybe clear leader cache
   if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
@@ -1493,8 +1501,8 @@ void syncNodeBecomeLeader(SSyncNode* pSyncNode, const char* debugStr) {
   // reset restoreFinish
   pSyncNode->restoreFinish = false;
 
-  sDebug("vgId:%d sync event become leader, isStandBy:%d, currentTerm:%lu, replicaNum:%d, restoreFinish:%d, %s",
-         pSyncNode->vgId, pSyncNode->pRaftCfg->isStandBy, pSyncNode->pRaftStore->currentTerm, pSyncNode->replicaNum,
+  sDebug("vgId:%d sync event currentTerm:%lu become leader, isStandBy:%d, replicaNum:%d, restoreFinish:%d, %s",
+         pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum,
          pSyncNode->restoreFinish, debugStr);
 
   // state change
@@ -2069,21 +2077,13 @@ const char* syncStr(ESyncState state) {
 static int32_t syncDoLeaderTransfer(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* pEntry) {
   SyncLeaderTransfer* pSyncLeaderTransfer = syncLeaderTransferFromRpcMsg2(pRpcMsg);
 
-  /*
-    char     host[128];
-    uint16_t port;
-    syncUtilU642Addr(pSyncLeaderTransfer->newLeaderId.addr, host, sizeof(host), &port);
-    sDebug("vgId:%d sync event, maybe leader transfer to %s:%d %lu", ths->vgId, host, port,
-           pSyncLeaderTransfer->newLeaderId.addr);
-  */
-
-  sDebug("vgId:%d sync event begin leader transfer", ths->vgId);
+  sDebug("vgId:%d sync event currentTerm:%lu begin leader transfer", ths->vgId, ths->pRaftStore->currentTerm);
 
   if (strcmp(pSyncLeaderTransfer->newNodeInfo.nodeFqdn, ths->myNodeInfo.nodeFqdn) == 0 &&
       pSyncLeaderTransfer->newNodeInfo.nodePort == ths->myNodeInfo.nodePort) {
-    sDebug("vgId:%d sync event maybe leader transfer to %s:%d %lu", ths->vgId,
-           pSyncLeaderTransfer->newNodeInfo.nodeFqdn, pSyncLeaderTransfer->newNodeInfo.nodePort,
-           pSyncLeaderTransfer->newLeaderId.addr);
+    sDebug("vgId:%d sync event currentTerm:%lu maybe leader transfer to %s:%d %lu", ths->vgId,
+           ths->pRaftStore->currentTerm, pSyncLeaderTransfer->newNodeInfo.nodeFqdn,
+           pSyncLeaderTransfer->newNodeInfo.nodePort, pSyncLeaderTransfer->newLeaderId.addr);
 
     // reset elect timer now!
     int32_t electMS = 1;
@@ -2205,8 +2205,8 @@ static int32_t syncNodeConfigChange(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftE
 int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, uint64_t flag) {
   int32_t    code = 0;
   ESyncState state = flag;
-  sDebug("vgId:%d sync event commit by wal from index:%" PRId64 " to index:%" PRId64 ", %s", ths->vgId, beginIndex,
-         endIndex, syncUtilState2String(state));
+  sDebug("vgId:%d sync event currentTerm:%lu commit by wal from index:%" PRId64 " to index:%" PRId64 ", %s", ths->vgId,
+         ths->pRaftStore->currentTerm, beginIndex, endIndex, syncUtilState2String(state));
 
   // execute fsm
   if (ths->pFsm != NULL) {
@@ -2254,8 +2254,8 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
               ths->pFsm->FpRestoreFinishCb(ths->pFsm);
             }
             ths->restoreFinish = true;
-            sDebug("vgId:%d sync event restore finish, %s, index:%ld", ths->vgId, syncUtilState2String(ths->state),
-                   pEntry->index);
+            sDebug("vgId:%d sync event currentTerm:%lu restore finish, %s, index:%ld", ths->vgId,
+                   ths->pRaftStore->currentTerm, syncUtilState2String(ths->state), pEntry->index);
           }
         }
 
diff --git a/source/libs/sync/src/syncRaftLog.c b/source/libs/sync/src/syncRaftLog.c
index 79ce84abe9..31ac44fa81 100644
--- a/source/libs/sync/src/syncRaftLog.c
+++ b/source/libs/sync/src/syncRaftLog.c
@@ -15,6 +15,7 @@
 
 #include "syncRaftLog.h"
 #include "syncRaftCfg.h"
+#include "syncRaftStore.h"
 #include "wal.h"
 
 // refactor, log[0 .. n] ==> log[m .. n]
@@ -162,10 +163,10 @@ static int32_t raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntr
 
   walFsync(pWal, true);
 
-  sDebug("vgId:%d sync event write index:%ld, %s, isStandBy:%d, msgType:%s,%d, originalRpcType:%s,%d",
-         pData->pSyncNode->vgId, pEntry->index, syncUtilState2String(pData->pSyncNode->state),
-         pData->pSyncNode->pRaftCfg->isStandBy, TMSG_INFO(pEntry->msgType), pEntry->msgType,
-         TMSG_INFO(pEntry->originalRpcType), pEntry->originalRpcType);
+  sDebug("vgId:%d sync event currentTerm:%lu write index:%ld, %s, isStandBy:%d, msgType:%s,%d, originalRpcType:%s,%d",
+         pData->pSyncNode->vgId, pData->pSyncNode->pRaftStore->currentTerm, pEntry->index,
+         syncUtilState2String(pData->pSyncNode->state), pData->pSyncNode->pRaftCfg->isStandBy,
+         TMSG_INFO(pEntry->msgType), pEntry->msgType, TMSG_INFO(pEntry->originalRpcType), pEntry->originalRpcType);
 
   return code;
 }
@@ -321,7 +322,12 @@ int32_t logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
 
   walFsync(pWal, true);
 
-  sDebug("sync event old write wal: %ld", pEntry->index);
+  sDebug(
+      "vgId:%d sync event currentTerm:%lu old write index:%ld, %s, isStandBy:%d, msgType:%s,%d, originalRpcType:%s,%d",
+      pData->pSyncNode->vgId, pData->pSyncNode->pRaftStore->currentTerm, pEntry->index,
+      syncUtilState2String(pData->pSyncNode->state), pData->pSyncNode->pRaftCfg->isStandBy, TMSG_INFO(pEntry->msgType),
+      pEntry->msgType, TMSG_INFO(pEntry->originalRpcType), pEntry->originalRpcType);
+
   return code;
 }
 
diff --git a/source/libs/sync/src/syncRespMgr.c b/source/libs/sync/src/syncRespMgr.c
index adc67b5d8d..5b127793d6 100644
--- a/source/libs/sync/src/syncRespMgr.c
+++ b/source/libs/sync/src/syncRespMgr.c
@@ -14,6 +14,7 @@
  */
 
 #include "syncRespMgr.h"
+#include "syncRaftStore.h"
 
 SSyncRespMgr *syncRespMgrCreate(void *data, int64_t ttl) {
   SSyncRespMgr *pObj = (SSyncRespMgr *)taosMemoryMalloc(sizeof(SSyncRespMgr));
@@ -45,9 +46,9 @@ int64_t syncRespMgrAdd(SSyncRespMgr *pObj, SRespStub *pStub) {
   taosHashPut(pObj->pRespHash, &keyCode, sizeof(keyCode), pStub, sizeof(SRespStub));
 
   SSyncNode *pSyncNode = pObj->data;
-  sDebug("vgId:%d sync event resp mgr add, msgType:%s,%d seq:%lu handle:%p ahandle:%p", pSyncNode->vgId,
-         TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType, keyCode, pStub->rpcMsg.info.handle,
-         pStub->rpcMsg.info.ahandle);
+  sDebug("vgId:%d sync event currentTerm:%lu resp mgr add, msgType:%s,%d seq:%lu handle:%p ahandle:%p", pSyncNode->vgId,
+         pSyncNode->pRaftStore->currentTerm, TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType, keyCode,
+         pStub->rpcMsg.info.handle, pStub->rpcMsg.info.ahandle);
 
   taosThreadMutexUnlock(&(pObj->mutex));
   return keyCode;
@@ -70,9 +71,9 @@ int32_t syncRespMgrGet(SSyncRespMgr *pObj, uint64_t index, SRespStub *pStub) {
     memcpy(pStub, pTmp, sizeof(SRespStub));
 
     SSyncNode *pSyncNode = pObj->data;
-    sDebug("vgId:%d sync event resp mgr get, msgType:%s,%d seq:%lu handle:%p ahandle:%p", pSyncNode->vgId,
-           TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType, index, pStub->rpcMsg.info.handle,
-           pStub->rpcMsg.info.ahandle);
+    sDebug("vgId:%d sync event currentTerm:%lu resp mgr get, msgType:%s,%d seq:%lu handle:%p ahandle:%p",
+           pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType,
+           index, pStub->rpcMsg.info.handle, pStub->rpcMsg.info.ahandle);
 
     taosThreadMutexUnlock(&(pObj->mutex));
     return 1;  // get one object
@@ -89,9 +90,9 @@ int32_t syncRespMgrGetAndDel(SSyncRespMgr *pObj, uint64_t index, SRespStub *pStu
     memcpy(pStub, pTmp, sizeof(SRespStub));
 
     SSyncNode *pSyncNode = pObj->data;
-    sDebug("vgId:%d sync event resp mgr get and del, msgType:%s,%d seq:%lu handle:%p ahandle:%p", pSyncNode->vgId,
-           TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType, index, pStub->rpcMsg.info.handle,
-           pStub->rpcMsg.info.ahandle);
+    sDebug("vgId:%d sync event currentTerm:%lu resp mgr get and del, msgType:%s,%d seq:%lu handle:%p ahandle:%p",
+           pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, TMSG_INFO(pStub->rpcMsg.msgType), pStub->rpcMsg.msgType,
+           index, pStub->rpcMsg.info.handle, pStub->rpcMsg.info.ahandle);
 
     taosHashRemove(pObj->pRespHash, &index, sizeof(index));
     taosThreadMutexUnlock(&(pObj->mutex));
diff --git a/source/libs/sync/src/syncSnapshot.c b/source/libs/sync/src/syncSnapshot.c
index 545ea953f8..d3785060d4 100644
--- a/source/libs/sync/src/syncSnapshot.c
+++ b/source/libs/sync/src/syncSnapshot.c
@@ -141,18 +141,22 @@ void snapshotSenderStart(SSyncSnapshotSender *pSender) {
   if (gRaftDetailLog) {
     char *msgStr = syncSnapshotSend2Str(pMsg);
     sDebug(
-        "vgId:%d sync event snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
+        "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld "
+        "lastApplyTerm:%lu "
         "lastConfigIndex:%ld privateTerm:%lu send "
         "msg:%s",
-        pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm, msgStr);
+        pSender->pSyncNode->vgId, pSender->pSyncNode->pRaftStore->currentTerm, host, port, pSender->seq, pSender->ack,
+        pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex,
+        pSender->privateTerm, msgStr);
     taosMemoryFree(msgStr);
   } else {
     sDebug(
-        "vgId:%d sync event snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
+        "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld "
+        "lastApplyTerm:%lu "
         "lastConfigIndex:%ld privateTerm:%lu",
-        pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm);
+        pSender->pSyncNode->vgId, pSender->pSyncNode->pRaftStore->currentTerm, host, port, pSender->seq, pSender->ack,
+        pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex,
+        pSender->privateTerm);
   }
 
   syncSnapshotSendDestroy(pMsg);
@@ -279,25 +283,31 @@ int32_t snapshotSend(SSyncSnapshotSender *pSender) {
     if (gRaftDetailLog) {
       char *msgStr = syncSnapshotSend2Str(pMsg);
       sDebug(
-          "vgId:%d sync event snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
+          "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld "
+          "lastApplyTerm:%lu "
           "lastConfigIndex:%ld privateTerm:%lu send "
           "msg:%s",
-          pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-          pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm, msgStr);
+          pSender->pSyncNode->vgId, pSender->pSyncNode->pRaftStore->currentTerm, host, port, pSender->seq, pSender->ack,
+          pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex,
+          pSender->privateTerm, msgStr);
       taosMemoryFree(msgStr);
     } else {
       sDebug(
-          "vgId:%d sync event snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
+          "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld "
+          "lastApplyTerm:%lu "
           "lastConfigIndex:%ld privateTerm:%lu",
-          pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-          pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm);
+          pSender->pSyncNode->vgId, pSender->pSyncNode->pRaftStore->currentTerm, host, port, pSender->seq, pSender->ack,
+          pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex,
+          pSender->privateTerm);
     }
   } else {
     sDebug(
-        "vgId:%d sync event snapshot send to %s:%d sending seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu "
+        "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d sending seq:%d ack:%d lastApplyIndex:%ld "
+        "lastApplyTerm:%lu "
         "lastConfigIndex:%ld privateTerm:%lu",
-        pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
-        pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex, pSender->privateTerm);
+        pSender->pSyncNode->vgId, pSender->pSyncNode->pRaftStore->currentTerm, host, port, pSender->seq, pSender->ack,
+        pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, pSender->snapshot.lastConfigIndex,
+        pSender->privateTerm);
   }
 
   syncSnapshotSendDestroy(pMsg);
@@ -328,12 +338,15 @@ int32_t snapshotReSend(SSyncSnapshotSender *pSender) {
 
     if (gRaftDetailLog) {
       char *msgStr = syncSnapshotSend2Str(pMsg);
-      sDebug("vgId:%d sync event snapshot send to %s:%d resend seq:%d ack:%d privateTerm:%lu send msg:%s",
-             pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->privateTerm, msgStr);
+      sDebug(
+          "vgId:%d sync event currentTerm:%lu snapshot send to %s:%d resend seq:%d ack:%d privateTerm:%lu send msg:%s",
+          pSender->pSyncNode->vgId, pSender->pSyncNode->pRaftStore->currentTerm, host, port, pSender->seq, pSender->ack,
+          pSender->privateTerm, msgStr);
       taosMemoryFree(msgStr);
     } else {
-      sDebug("vgId:%d sync event snapshot send to %s:%d resend seq:%d ack:%d privateTerm:%lu", pSender->pSyncNode->vgId,
-             host, port, pSender->seq, pSender->ack, pSender->privateTerm);
+      sDebug("vgId:%d sync event currentTerm:%lu snapshot send to %s:%d resend seq:%d ack:%d privateTerm:%lu",
+             pSender->pSyncNode->vgId, pSender->pSyncNode->pRaftStore->currentTerm, host, port, pSender->seq,
+             pSender->ack, pSender->privateTerm);
     }
 
     syncSnapshotSendDestroy(pMsg);
@@ -485,7 +498,7 @@ void snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply) {
   pReceiver->start = false;
 
   if (apply) {
-    ++(pReceiver->privateTerm);
+    //    ++(pReceiver->privateTerm);
   }
 
   if (gRaftDetailLog) {
@@ -566,17 +579,17 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
         if (gRaftDetailLog) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, "
               "lastConfigIndex:%ld, privateTerm:%lu, recv msg:%s",
-              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
-              pReceiver->privateTerm, msgStr);
+              pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack, pMsg->lastIndex,
+              pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, "
               "lastConfigIndex:%ld privateTerm:%lu",
-              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
-              pReceiver->privateTerm);
+              pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack, pMsg->lastIndex,
+              pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm);
         }
 
       } else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_END) {
@@ -612,14 +625,19 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
 
           bool isDrop;
           if (IamInNew) {
-            sDebug("vgId:%d sync event update config by snapshot, lastIndex:%ld, lastTerm:%lu, lastConfigIndex:%ld ",
-                   pSyncNode->vgId, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex);
+            sDebug(
+                "vgId:%d sync event currentTerm:%lu update config by snapshot, lastIndex:%ld, lastTerm:%lu, "
+                "lastConfigIndex:%ld ",
+                pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pMsg->lastIndex, pMsg->lastTerm,
+                pMsg->lastConfigIndex);
             syncNodeUpdateConfig(pSyncNode, &newSyncCfg, pMsg->lastConfigIndex, &isDrop);
           } else {
             sDebug(
-                "vgId:%d sync event do not update config by snapshot, I am not in newCfg, lastIndex:%ld, lastTerm:%lu, "
+                "vgId:%d sync event currentTerm:%lu do not update config by snapshot, I am not in newCfg, "
+                "lastIndex:%ld, lastTerm:%lu, "
                 "lastConfigIndex:%ld ",
-                pSyncNode->vgId, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex);
+                pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pMsg->lastIndex, pMsg->lastTerm,
+                pMsg->lastConfigIndex);
           }
 
           // change isStandBy to normal
@@ -644,19 +662,20 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
         if (gRaftDetailLog) {
           char *logSimpleStr = logStoreSimple2Str(pSyncNode->pLogStore);
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d finish, update log begin index:%ld, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d finish, update log begin index:%ld, "
               "snapshot.lastApplyIndex:%ld, "
               "snapshot.lastApplyTerm:%lu, snapshot.lastConfigIndex:%ld, privateTerm:%lu, raft log:%s",
-              pSyncNode->vgId, host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm,
-              snapshot.lastConfigIndex, pReceiver->privateTerm, logSimpleStr);
+              pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, host, port, pMsg->lastIndex + 1,
+              snapshot.lastApplyIndex, snapshot.lastApplyTerm, snapshot.lastConfigIndex, pReceiver->privateTerm,
+              logSimpleStr);
           taosMemoryFree(logSimpleStr);
         } else {
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d finish, update log begin index:%ld, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d finish, update log begin index:%ld, "
               "snapshot.lastApplyIndex:%ld, "
               "snapshot.lastApplyTerm:%lu, snapshot.lastConfigIndex:%ld, privateTerm:%lu",
-              pSyncNode->vgId, host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm,
-              snapshot.lastConfigIndex, pReceiver->privateTerm);
+              pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, host, port, pMsg->lastIndex + 1,
+              snapshot.lastApplyIndex, snapshot.lastApplyTerm, snapshot.lastConfigIndex, pReceiver->privateTerm);
         }
 
         pReceiver->pWriter = NULL;
@@ -667,17 +686,17 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
         if (gRaftDetailLog) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, "
               "lastConfigIndex:%ld, privateTerm:%lu, recv msg:%s",
-              pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
+              pReceiver->pSyncNode->vgId, pReceiver->pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack,
+              pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, "
               "lastConfigIndex:%ld, privateTerm:%lu",
-              pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex, pReceiver->privateTerm);
+              pReceiver->pSyncNode->vgId, pReceiver->pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack,
+              pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm);
         }
 
       } else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_FORCE_CLOSE) {
@@ -692,18 +711,20 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
         if (gRaftDetailLog) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, "
+              "lastTerm:%lu, "
               "lastConfigIndex:%ld, privateTerm:%lu, recv "
               "msg:%s",
-              pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
+              pReceiver->pSyncNode->vgId, pReceiver->pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack,
+              pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, "
+              "lastTerm:%lu, "
               "lastConfigIndex:%ld, privateTerm:%lu",
-              pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm,
-              pMsg->lastConfigIndex, pReceiver->privateTerm);
+              pReceiver->pSyncNode->vgId, pReceiver->pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack,
+              pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm);
         }
 
       } else if (pMsg->seq > SYNC_SNAPSHOT_SEQ_BEGIN && pMsg->seq < SYNC_SNAPSHOT_SEQ_END) {
@@ -723,17 +744,19 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
         if (gRaftDetailLog) {
           char *msgStr = syncSnapshotSend2Str(pMsg);
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, "
+              "lastTerm:%lu, "
               "lastConfigIndex:%ld, privateTerm:%lu, recv msg:%s",
-              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
-              pReceiver->privateTerm, msgStr);
+              pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack, pMsg->lastIndex,
+              pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm, msgStr);
           taosMemoryFree(msgStr);
         } else {
           sDebug(
-              "vgId:%d sync event snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu, "
+              "vgId:%d sync event currentTerm:%lu snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, "
+              "lastTerm:%lu, "
               "lastConfigIndex:%ld, privateTerm:%lu",
-              pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex,
-              pReceiver->privateTerm);
+              pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, host, port, pReceiver->ack, pMsg->lastIndex,
+              pMsg->lastTerm, pMsg->lastConfigIndex, pReceiver->privateTerm);
         }
 
       } else {

From 528d7d7248cbafb734aec558368cacbcaa832551 Mon Sep 17 00:00:00 2001
From: afwerar <1296468573@qq.com>
Date: Thu, 16 Jun 2022 15:46:51 +0800
Subject: [PATCH 22/40] os: add dll check

---
 include/os/osSysinfo.h             |  1 +
 source/client/src/clientHb.c       |  4 +++-
 source/libs/catalog/src/ctgCache.c |  4 +++-
 source/os/src/osSysinfo.c          | 16 ++++++++++++++++
 source/util/src/tcache.c           |  6 +++++-
 5 files changed, 28 insertions(+), 3 deletions(-)

diff --git a/include/os/osSysinfo.h b/include/os/osSysinfo.h
index c009bcf350..4ec2e2884e 100644
--- a/include/os/osSysinfo.h
+++ b/include/os/osSysinfo.h
@@ -68,6 +68,7 @@ typedef struct {
 } SysNameInfo;
 
 SysNameInfo taosGetSysNameInfo();
+bool taosCheckCurrentInDll();
 
 #ifdef __cplusplus
 }
diff --git a/source/client/src/clientHb.c b/source/client/src/clientHb.c
index 5db27ae438..c450e9351f 100644
--- a/source/client/src/clientHb.c
+++ b/source/client/src/clientHb.c
@@ -588,7 +588,9 @@ void hbThreadFuncUnexpectedStopped(void) {
 static void *hbThreadFunc(void *param) {
   setThreadName("hb");
 #ifdef WINDOWS
-  atexit(hbThreadFuncUnexpectedStopped);
+  if (taosCheckCurrentInDll()) {
+    atexit(hbThreadFuncUnexpectedStopped);
+  }
 #endif
   while (1) {
     int8_t threadStop = atomic_val_compare_exchange_8(&clientHbMgr.threadStop, 1, 2);
diff --git a/source/libs/catalog/src/ctgCache.c b/source/libs/catalog/src/ctgCache.c
index 277516686b..0948c01270 100644
--- a/source/libs/catalog/src/ctgCache.c
+++ b/source/libs/catalog/src/ctgCache.c
@@ -1938,7 +1938,9 @@ void ctgCleanupCacheQueue(void) {
 void* ctgUpdateThreadFunc(void* param) {
   setThreadName("catalog");
 #ifdef WINDOWS
-  atexit(ctgUpdateThreadUnexpectedStopped);
+  if (taosCheckCurrentInDll()) {
+    atexit(ctgUpdateThreadUnexpectedStopped);
+  }
 #endif
   qInfo("catalog update thread started");
 
diff --git a/source/os/src/osSysinfo.c b/source/os/src/osSysinfo.c
index ace870853f..4981f7dc26 100644
--- a/source/os/src/osSysinfo.c
+++ b/source/os/src/osSysinfo.c
@@ -945,3 +945,19 @@ SysNameInfo taosGetSysNameInfo() {
   return info;
 #endif
 }
+
+
+bool taosCheckCurrentInDll() {
+#ifdef WINDOWS
+  MEMORY_BASIC_INFORMATION mbi;
+  char path[PATH_MAX] = {0};
+  GetModuleFileName(((VirtualQuery(taosCheckCurrentInDll,&mbi,sizeof(mbi)) != 0) ? (HMODULE)mbi.AllocationBase : NULL), path, PATH_MAX);
+  int strLastIndex = strlen(path);
+  if ((path[strLastIndex-3] == 'd' || path[strLastIndex-3] == 'D') && (path[strLastIndex-2] == 'l' || path[strLastIndex-2] == 'L') && (path[strLastIndex-1] == 'l' || path[strLastIndex-1] == 'L')) {
+    return true;
+  }
+  return false;
+#else
+  return false;
+#endif
+}
\ No newline at end of file
diff --git a/source/util/src/tcache.c b/source/util/src/tcache.c
index 10a5475555..1a716752a9 100644
--- a/source/util/src/tcache.c
+++ b/source/util/src/tcache.c
@@ -829,7 +829,11 @@ void *taosCacheTimedRefresh(void *handle) {
 
   const int32_t SLEEP_DURATION = 500;  // 500 ms
   int64_t       count = 0;
-  atexit(taosCacheRefreshWorkerUnexpectedStopped);
+#ifdef WINDOWS
+  if (taosCheckCurrentInDll()) {
+    atexit(taosCacheRefreshWorkerUnexpectedStopped);
+  }
+#endif
 
   while (1) {
     taosMsleep(SLEEP_DURATION);

From 14b6523bd34981a1cd7b86dbddc711e278101a32 Mon Sep 17 00:00:00 2001
From: Shengliang Guan <slguan@taosdata.com>
Date: Thu, 16 Jun 2022 15:56:25 +0800
Subject: [PATCH 23/40] refactor: show rpc response code

---
 source/libs/transport/src/transCli.c |  4 ++--
 source/libs/transport/src/transSvr.c | 13 +++++++------
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/source/libs/transport/src/transCli.c b/source/libs/transport/src/transCli.c
index fbf4d6aa9f..3d570e0860 100644
--- a/source/libs/transport/src/transCli.c
+++ b/source/libs/transport/src/transCli.c
@@ -324,9 +324,9 @@ void cliHandleResp(SCliConn* conn) {
     tDebug("%s cli conn %p ref by app", CONN_GET_INST_LABEL(conn), conn);
   }
 
-  tDebug("%s cli conn %p %s received from %s:%d, local info: %s:%d, msg size: %d", pTransInst->label, conn,
+  tDebug("%s cli conn %p %s received from %s:%d, local info:%s:%d, msg size:%d code:0x%x", pTransInst->label, conn,
          TMSG_INFO(pHead->msgType), taosInetNtoa(conn->addr.sin_addr), ntohs(conn->addr.sin_port),
-         taosInetNtoa(conn->localAddr.sin_addr), ntohs(conn->localAddr.sin_port), transMsg.contLen);
+         taosInetNtoa(conn->localAddr.sin_addr), ntohs(conn->localAddr.sin_port), transMsg.contLen, transMsg.code);
 
   if (pCtx == NULL && CONN_NO_PERSIST_BY_APP(conn)) {
     tTrace("except, server continue send while cli ignore it");
diff --git a/source/libs/transport/src/transSvr.c b/source/libs/transport/src/transSvr.c
index 020435d076..b6e3fd2676 100644
--- a/source/libs/transport/src/transSvr.c
+++ b/source/libs/transport/src/transSvr.c
@@ -285,13 +285,14 @@ static void uvHandleReq(SSvrConn* pConn) {
   }
   if (pConn->status == ConnNormal && pHead->noResp == 0) {
     transRefSrvHandle(pConn);
-    tDebug("server conn %p %s received from %s:%d, local info: %s:%d, msg size: %d", pConn, TMSG_INFO(transMsg.msgType),
-           taosInetNtoa(pConn->addr.sin_addr), ntohs(pConn->addr.sin_port), taosInetNtoa(pConn->localAddr.sin_addr),
-           ntohs(pConn->localAddr.sin_port), transMsg.contLen);
-  } else {
-    tDebug("server conn %p %s received from %s:%d, local info: %s:%d, msg size: %d, resp:%d ", pConn,
+    tDebug("server conn %p %s received from %s:%d, local info:%s:%d, msg size:%d code:0x%x", pConn,
            TMSG_INFO(transMsg.msgType), taosInetNtoa(pConn->addr.sin_addr), ntohs(pConn->addr.sin_port),
-           taosInetNtoa(pConn->localAddr.sin_addr), ntohs(pConn->localAddr.sin_port), transMsg.contLen, pHead->noResp);
+           taosInetNtoa(pConn->localAddr.sin_addr), ntohs(pConn->localAddr.sin_port), transMsg.contLen, transMsg.code);
+  } else {
+    tDebug("server conn %p %s received from %s:%d, local info:%s:%d, msg size:%d, resp:%d code:0x%x", pConn,
+           TMSG_INFO(transMsg.msgType), taosInetNtoa(pConn->addr.sin_addr), ntohs(pConn->addr.sin_port),
+           taosInetNtoa(pConn->localAddr.sin_addr), ntohs(pConn->localAddr.sin_port), transMsg.contLen, pHead->noResp,
+           transMsg.code);
     // no ref here
   }
 

From 928f45709dfce0946ce9dc76b7162ba7cd5718df Mon Sep 17 00:00:00 2001
From: Liu Jicong <liujicong@qq.com>
Date: Thu, 16 Jun 2022 16:01:22 +0800
Subject: [PATCH 24/40] feat(stream): drop stream

---
 include/common/tmsg.h                      |   4 +-
 include/util/tdef.h                        |   2 +-
 source/dnode/mnode/impl/inc/mndDef.h       |  51 +++--
 source/dnode/mnode/impl/inc/mndStb.h       |   6 +-
 source/dnode/mnode/impl/inc/mndStream.h    |   2 +
 source/dnode/mnode/impl/src/mndDef.c       |  39 +---
 source/dnode/mnode/impl/src/mndScheduler.c |  12 +-
 source/dnode/mnode/impl/src/mndSma.c       |   4 +-
 source/dnode/mnode/impl/src/mndStb.c       |  31 ++-
 source/dnode/mnode/impl/src/mndStream.c    | 244 +++++++++++++++++++--
 10 files changed, 295 insertions(+), 100 deletions(-)

diff --git a/include/common/tmsg.h b/include/common/tmsg.h
index 7c0e8206e1..01d99ac705 100644
--- a/include/common/tmsg.h
+++ b/include/common/tmsg.h
@@ -1519,7 +1519,7 @@ typedef struct {
 #define STREAM_TRIGGER_MAX_DELAY    3
 
 typedef struct {
-  char    name[TSDB_TABLE_FNAME_LEN];
+  char    name[TSDB_STREAM_FNAME_LEN];
   char    sourceDB[TSDB_DB_FNAME_LEN];
   char    targetStbFullName[TSDB_TABLE_FNAME_LEN];
   int8_t  igExists;
@@ -1539,7 +1539,7 @@ int32_t tDeserializeSCMCreateStreamReq(void* buf, int32_t bufLen, SCMCreateStrea
 void    tFreeSCMCreateStreamReq(SCMCreateStreamReq* pReq);
 
 typedef struct {
-  char    name[TSDB_TOPIC_FNAME_LEN];
+  char    name[TSDB_STREAM_FNAME_LEN];
   int64_t streamId;
   char*   sql;
   char*   executorMsg;
diff --git a/include/util/tdef.h b/include/util/tdef.h
index 4798a9375c..a48e823bad 100644
--- a/include/util/tdef.h
+++ b/include/util/tdef.h
@@ -210,7 +210,7 @@ typedef enum ELogicConditionType {
 #define TSDB_TYPE_STR_MAX_LEN    32
 #define TSDB_TABLE_FNAME_LEN     (TSDB_DB_FNAME_LEN + TSDB_TABLE_NAME_LEN + TSDB_NAME_DELIMITER_LEN)
 #define TSDB_TOPIC_FNAME_LEN     (TSDB_ACCT_ID_LEN + TSDB_TABLE_NAME_LEN + TSDB_NAME_DELIMITER_LEN)
-#define TSDB_STREAM_FNAME_LEN    TSDB_TABLE_FNAME_LEN
+#define TSDB_STREAM_FNAME_LEN    (TSDB_ACCT_ID_LEN + TSDB_TABLE_NAME_LEN + TSDB_NAME_DELIMITER_LEN)
 #define TSDB_SUBSCRIBE_KEY_LEN   (TSDB_CGROUP_LEN + TSDB_TOPIC_FNAME_LEN + 2)
 #define TSDB_PARTITION_KEY_LEN   (TSDB_SUBSCRIBE_KEY_LEN + 20)
 #define TSDB_COL_NAME_LEN        65
diff --git a/source/dnode/mnode/impl/inc/mndDef.h b/source/dnode/mnode/impl/inc/mndDef.h
index 35cbf178ec..db18f87830 100644
--- a/source/dnode/mnode/impl/inc/mndDef.h
+++ b/source/dnode/mnode/impl/inc/mndDef.h
@@ -539,27 +539,38 @@ typedef struct {
 } SMqRebOutputObj;
 
 typedef struct {
-  char           name[TSDB_STREAM_FNAME_LEN];
-  char           sourceDb[TSDB_DB_FNAME_LEN];
-  char           targetDb[TSDB_DB_FNAME_LEN];
-  char           targetSTbName[TSDB_TABLE_FNAME_LEN];
-  int64_t        targetStbUid;
-  int64_t        createTime;
-  int64_t        updateTime;
-  int64_t        uid;
-  int64_t        dbUid;
-  int32_t        version;
-  int32_t        vgNum;
-  SRWLatch       lock;
-  int8_t         status;
-  int8_t         createdBy;      // STREAM_CREATED_BY__USER or SMA
-  int32_t        fixedSinkVgId;  // 0 for shuffle
-  SVgObj         fixedSinkVg;
-  int64_t        smaId;  // 0 for unused
-  int8_t         trigger;
-  int64_t        triggerParam;
-  int64_t        watermark;
+  char name[TSDB_STREAM_FNAME_LEN];
+  // ctl
+  SRWLatch lock;
+  // create info
+  int64_t createTime;
+  int64_t updateTime;
+  int32_t version;
+  // TODO remove it
+  int64_t smaId;  // 0 for unused
+  // info
+  int64_t uid;
+  int8_t  status;
+  // TODO remove it
+  int32_t vgNum;
+  // config
+  int8_t  dropPolicy;
+  int8_t  trigger;
+  int64_t triggerParam;
+  int64_t watermark;
+  // source and target
+  int64_t sourceDbUid;
+  int64_t targetDbUid;
+  char    sourceDb[TSDB_DB_FNAME_LEN];
+  char    targetDb[TSDB_DB_FNAME_LEN];
+  char    targetSTbName[TSDB_TABLE_FNAME_LEN];
+  int64_t targetStbUid;
+  // assigned when scheduling
+  int32_t fixedSinkVgId;  // 0 for shuffle
+  SVgObj  fixedSinkVg;
+  // transformation
   char*          sql;
+  char*          ast;
   char*          physicalPlan;
   SArray*        tasks;  // SArray<SArray<SStreamTask>>
   SSchemaWrapper outputSchema;
diff --git a/source/dnode/mnode/impl/inc/mndStb.h b/source/dnode/mnode/impl/inc/mndStb.h
index d3de4e3b3f..44a7fdadde 100644
--- a/source/dnode/mnode/impl/inc/mndStb.h
+++ b/source/dnode/mnode/impl/inc/mndStb.h
@@ -27,8 +27,7 @@ void     mndCleanupStb(SMnode *pMnode);
 SStbObj *mndAcquireStb(SMnode *pMnode, char *stbName);
 void     mndReleaseStb(SMnode *pMnode, SStbObj *pStb);
 SSdbRaw *mndStbActionEncode(SStbObj *pStb);
-int32_t  mndValidateStbInfo(SMnode *pMnode, SSTableVersion *pStbs, int32_t numOfStbs, void **ppRsp,
-                            int32_t *pRspLen);
+int32_t  mndValidateStbInfo(SMnode *pMnode, SSTableVersion *pStbs, int32_t numOfStbs, void **ppRsp, int32_t *pRspLen);
 int32_t  mndGetNumOfStbs(SMnode *pMnode, char *dbName, int32_t *pNumOfStbs);
 
 int32_t mndCheckCreateStbReq(SMCreateStbReq *pCreate);
@@ -36,6 +35,9 @@ SDbObj *mndAcquireDbByStb(SMnode *pMnode, const char *stbName);
 int32_t mndBuildStbFromReq(SMnode *pMnode, SStbObj *pDst, SMCreateStbReq *pCreate, SDbObj *pDb);
 int32_t mndAddStbToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SStbObj *pStb);
 
+void mndExtractDbNameFromStbFullName(const char *stbFullName, char *dst);
+void mndExtractTbNameFromStbFullName(const char *stbFullName, char *dst, int32_t dstSize);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/source/dnode/mnode/impl/inc/mndStream.h b/source/dnode/mnode/impl/inc/mndStream.h
index b5d22cb7a5..65017027dc 100644
--- a/source/dnode/mnode/impl/inc/mndStream.h
+++ b/source/dnode/mnode/impl/inc/mndStream.h
@@ -33,6 +33,8 @@ SSdbRow *mndStreamActionDecode(SSdbRaw *pRaw);
 
 int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast, STrans *pTrans);
 
+int32_t mndDropStreamByDb(SMnode *pMnode, STrans *pTrans, SDbObj *pDb);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/source/dnode/mnode/impl/src/mndDef.c b/source/dnode/mnode/impl/src/mndDef.c
index 6eac91bb43..b9bd1e4298 100644
--- a/source/dnode/mnode/impl/src/mndDef.c
+++ b/source/dnode/mnode/impl/src/mndDef.c
@@ -19,7 +19,6 @@
 
 int32_t tEncodeSStreamObj(SEncoder *pEncoder, const SStreamObj *pObj) {
   int32_t sz = 0;
-  /*int32_t outputNameSz = 0;*/
   if (tEncodeCStr(pEncoder, pObj->name) < 0) return -1;
   if (tEncodeCStr(pEncoder, pObj->sourceDb) < 0) return -1;
   if (tEncodeCStr(pEncoder, pObj->targetDb) < 0) return -1;
@@ -28,10 +27,12 @@ int32_t tEncodeSStreamObj(SEncoder *pEncoder, const SStreamObj *pObj) {
   if (tEncodeI64(pEncoder, pObj->createTime) < 0) return -1;
   if (tEncodeI64(pEncoder, pObj->updateTime) < 0) return -1;
   if (tEncodeI64(pEncoder, pObj->uid) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->dbUid) < 0) return -1;
+  if (tEncodeI64(pEncoder, pObj->sourceDbUid) < 0) return -1;
+  if (tEncodeI64(pEncoder, pObj->targetDbUid) < 0) return -1;
   if (tEncodeI32(pEncoder, pObj->version) < 0) return -1;
+  if (tEncodeI32(pEncoder, pObj->vgNum) < 0) return -1;
+  if (tEncodeI8(pEncoder, pObj->dropPolicy) < 0) return -1;
   if (tEncodeI8(pEncoder, pObj->status) < 0) return -1;
-  if (tEncodeI8(pEncoder, pObj->createdBy) < 0) return -1;
   if (tEncodeI8(pEncoder, pObj->trigger) < 0) return -1;
   if (tEncodeI64(pEncoder, pObj->triggerParam) < 0) return -1;
   if (tEncodeI64(pEncoder, pObj->watermark) < 0) return -1;
@@ -57,17 +58,6 @@ int32_t tEncodeSStreamObj(SEncoder *pEncoder, const SStreamObj *pObj) {
   }
 
   if (tEncodeSSchemaWrapper(pEncoder, &pObj->outputSchema) < 0) return -1;
-
-#if 0
-  if (pObj->ColAlias != NULL) {
-    outputNameSz = taosArrayGetSize(pObj->ColAlias);
-  }
-  if (tEncodeI32(pEncoder, outputNameSz) < 0) return -1;
-  for (int32_t i = 0; i < outputNameSz; i++) {
-    char *name = taosArrayGetP(pObj->ColAlias, i);
-    if (tEncodeCStr(pEncoder, name) < 0) return -1;
-  }
-#endif
   return pEncoder->pos;
 }
 
@@ -80,10 +70,12 @@ int32_t tDecodeSStreamObj(SDecoder *pDecoder, SStreamObj *pObj) {
   if (tDecodeI64(pDecoder, &pObj->createTime) < 0) return -1;
   if (tDecodeI64(pDecoder, &pObj->updateTime) < 0) return -1;
   if (tDecodeI64(pDecoder, &pObj->uid) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->dbUid) < 0) return -1;
+  if (tDecodeI64(pDecoder, &pObj->sourceDbUid) < 0) return -1;
+  if (tDecodeI64(pDecoder, &pObj->targetDbUid) < 0) return -1;
   if (tDecodeI32(pDecoder, &pObj->version) < 0) return -1;
+  if (tDecodeI32(pDecoder, &pObj->vgNum) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pObj->dropPolicy) < 0) return -1;
   if (tDecodeI8(pDecoder, &pObj->status) < 0) return -1;
-  if (tDecodeI8(pDecoder, &pObj->createdBy) < 0) return -1;
   if (tDecodeI8(pDecoder, &pObj->trigger) < 0) return -1;
   if (tDecodeI64(pDecoder, &pObj->triggerParam) < 0) return -1;
   if (tDecodeI64(pDecoder, &pObj->watermark) < 0) return -1;
@@ -112,21 +104,6 @@ int32_t tDecodeSStreamObj(SDecoder *pDecoder, SStreamObj *pObj) {
   }
 
   if (tDecodeSSchemaWrapper(pDecoder, &pObj->outputSchema) < 0) return -1;
-#if 0
-  int32_t outputNameSz;
-  if (tDecodeI32(pDecoder, &outputNameSz) < 0) return -1;
-  if (outputNameSz != 0) {
-    pObj->ColAlias = taosArrayInit(outputNameSz, sizeof(void *));
-    if (pObj->ColAlias == NULL) {
-      return -1;
-    }
-  }
-  for (int32_t i = 0; i < outputNameSz; i++) {
-    char *name;
-    if (tDecodeCStrAlloc(pDecoder, &name) < 0) return -1;
-    taosArrayPush(pObj->ColAlias, &name);
-  }
-#endif
   return 0;
 }
 
diff --git a/source/dnode/mnode/impl/src/mndScheduler.c b/source/dnode/mnode/impl/src/mndScheduler.c
index ffddab3a91..c53e97cd53 100644
--- a/source/dnode/mnode/impl/src/mndScheduler.c
+++ b/source/dnode/mnode/impl/src/mndScheduler.c
@@ -131,7 +131,7 @@ int32_t mndPersistTaskDeployReq(STrans* pTrans, SStreamTask* pTask, const SEpSet
 int32_t mndAddSinkToTask(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream, SStreamTask* pTask) {
   pTask->dispatchType = TASK_DISPATCH__NONE;
   // sink
-  if (pStream->createdBy == STREAM_CREATED_BY__SMA) {
+  if (pStream->smaId != 0) {
     pTask->sinkType = TASK_SINK__SMA;
     pTask->smaSink.smaId = pStream->smaId;
   } else {
@@ -275,7 +275,7 @@ int32_t mndAddShuffleSinkTasksToStream(SMnode* pMnode, STrans* pTrans, SStreamOb
     pTask->execType = TASK_EXEC__NONE;
 
     // sink
-    if (pStream->createdBy == STREAM_CREATED_BY__SMA) {
+    if (pStream->smaId != 0) {
       pTask->sinkType = TASK_SINK__SMA;
       pTask->smaSink.smaId = pStream->smaId;
     } else {
@@ -321,7 +321,7 @@ int32_t mndAddFixedSinkTaskToStream(SMnode* pMnode, STrans* pTrans, SStreamObj*
   pTask->execType = TASK_EXEC__NONE;
 
   // sink
-  if (pStream->createdBy == STREAM_CREATED_BY__SMA) {
+  if (pStream->smaId != 0) {
     pTask->sinkType = TASK_SINK__SMA;
     pTask->smaSink.smaId = pStream->smaId;
   } else {
@@ -399,7 +399,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
 
       // exec
       pFinalTask->execType = TASK_EXEC__PIPE;
-      SVgObj* pVgroup = mndSchedFetchOneVg(pMnode, pStream->dbUid);
+      SVgObj* pVgroup = mndSchedFetchOneVg(pMnode, pStream->sourceDbUid);
       if (mndAssignTaskToVg(pMnode, pTrans, pFinalTask, plan, pVgroup) < 0) {
         sdbRelease(pSdb, pVgroup);
         qDestroyQueryPlan(pPlan);
@@ -420,7 +420,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
       SVgObj* pVgroup;
       pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void**)&pVgroup);
       if (pIter == NULL) break;
-      if (pVgroup->dbUid != pStream->dbUid) {
+      if (pVgroup->dbUid != pStream->sourceDbUid) {
         sdbRelease(pSdb, pVgroup);
         continue;
       }
@@ -463,7 +463,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
       SVgObj* pVgroup;
       pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void**)&pVgroup);
       if (pIter == NULL) break;
-      if (pVgroup->dbUid != pStream->dbUid) {
+      if (pVgroup->dbUid != pStream->sourceDbUid) {
         sdbRelease(pSdb, pVgroup);
         continue;
       }
diff --git a/source/dnode/mnode/impl/src/mndSma.c b/source/dnode/mnode/impl/src/mndSma.c
index 0f6bb74d9a..54fd0fe867 100644
--- a/source/dnode/mnode/impl/src/mndSma.c
+++ b/source/dnode/mnode/impl/src/mndSma.c
@@ -520,6 +520,7 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   memcpy(smaObj.db, pDb->name, TSDB_DB_FNAME_LEN);
   smaObj.createdTime = taosGetTimestampMs();
   smaObj.uid = mndGenerateUid(pCreate->name, TSDB_TABLE_FNAME_LEN);
+  ASSERT(smaObj.uid != 0);
   char resultTbName[TSDB_TABLE_FNAME_LEN + 16] = {0};
   snprintf(resultTbName, TSDB_TABLE_FNAME_LEN + 16, "td.tsma.rst.tb.%s", pCreate->name);
   memcpy(smaObj.dstTbName, resultTbName, TSDB_TABLE_FNAME_LEN);
@@ -556,10 +557,9 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   streamObj.createTime = taosGetTimestampMs();
   streamObj.updateTime = streamObj.createTime;
   streamObj.uid = mndGenerateUid(pCreate->name, strlen(pCreate->name));
-  streamObj.dbUid = pDb->uid;
+  streamObj.sourceDbUid = pDb->uid;
   streamObj.version = 1;
   streamObj.sql = pCreate->sql;
-  streamObj.createdBy = STREAM_CREATED_BY__SMA;
   streamObj.smaId = smaObj.uid;
   streamObj.watermark = 0;
   streamObj.trigger = STREAM_TRIGGER_AT_ONCE;
diff --git a/source/dnode/mnode/impl/src/mndStb.c b/source/dnode/mnode/impl/src/mndStb.c
index d4a38dd17b..b83aa34a3a 100644
--- a/source/dnode/mnode/impl/src/mndStb.c
+++ b/source/dnode/mnode/impl/src/mndStb.c
@@ -28,7 +28,6 @@
 #include "mndTrans.h"
 #include "mndUser.h"
 #include "mndVgroup.h"
-#include "mndSma.h"
 #include "tname.h"
 
 #define STB_VER_NUMBER   1
@@ -1271,7 +1270,8 @@ static int32_t mndBuildStbSchemaImp(SDbObj *pDb, SStbObj *pStb, const char *tbNa
   return 0;
 }
 
-static int32_t mndBuildStbSchema(SMnode *pMnode, const char *dbFName, const char *tbName, STableMetaRsp *pRsp, int32_t *smaVer) {
+static int32_t mndBuildStbSchema(SMnode *pMnode, const char *dbFName, const char *tbName, STableMetaRsp *pRsp,
+                                 int32_t *smaVer) {
   char tbFName[TSDB_TABLE_FNAME_LEN] = {0};
   snprintf(tbFName, sizeof(tbFName), "%s.%s", dbFName, tbName);
 
@@ -1672,7 +1672,7 @@ int32_t mndValidateStbInfo(SMnode *pMnode, SSTableVersion *pStbVersions, int32_t
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
-  
+
   for (int32_t i = 0; i < numOfStbs; ++i) {
     SSTableVersion *pStbVersion = &pStbVersions[i];
     pStbVersion->suid = be64toh(pStbVersion->suid);
@@ -1681,7 +1681,7 @@ int32_t mndValidateStbInfo(SMnode *pMnode, SSTableVersion *pStbVersions, int32_t
     pStbVersion->smaVer = ntohl(pStbVersion->smaVer);
 
     STableMetaRsp metaRsp = {0};
-    int32_t smaVer = 0;
+    int32_t       smaVer = 0;
     mDebug("stb:%s.%s, start to retrieve meta", pStbVersion->dbFName, pStbVersion->stbName);
     if (mndBuildStbSchema(pMnode, pStbVersion->dbFName, pStbVersion->stbName, &metaRsp, &smaVer) != 0) {
       metaRsp.numOfColumns = -1;
@@ -1697,15 +1697,15 @@ int32_t mndValidateStbInfo(SMnode *pMnode, SSTableVersion *pStbVersions, int32_t
     }
 
     if (pStbVersion->smaVer && pStbVersion->smaVer != smaVer) {
-      bool exist = false;
-      char tbFName[TSDB_TABLE_FNAME_LEN];
+      bool           exist = false;
+      char           tbFName[TSDB_TABLE_FNAME_LEN];
       STableIndexRsp indexRsp = {0};
       indexRsp.pIndex = taosArrayInit(10, sizeof(STableIndexInfo));
       if (NULL == indexRsp.pIndex) {
         terrno = TSDB_CODE_OUT_OF_MEMORY;
         return -1;
       }
-      
+
       sprintf(tbFName, "%s.%s", pStbVersion->dbFName, pStbVersion->stbName);
       int32_t code = mndGetTableSma(pMnode, tbFName, &indexRsp, &exist);
       if (code || !exist) {
@@ -1769,16 +1769,23 @@ int32_t mndGetNumOfStbs(SMnode *pMnode, char *dbName, int32_t *pNumOfStbs) {
   return 0;
 }
 
-static void mndExtractTableName(char *tableId, char *name) {
+void mndExtractDbNameFromStbFullName(const char *stbFullName, char *dst) {
+  SName name = {0};
+  tNameFromString(&name, stbFullName, T_NAME_ACCT | T_NAME_DB | T_NAME_TABLE);
+
+  tNameGetFullDbName(&name, dst);
+}
+
+void mndExtractTbNameFromStbFullName(const char *stbFullName, char *dst, int32_t dstSize) {
   int32_t pos = -1;
   int32_t num = 0;
-  for (pos = 0; tableId[pos] != 0; ++pos) {
-    if (tableId[pos] == TS_PATH_DELIMITER[0]) num++;
+  for (pos = 0; stbFullName[pos] != 0; ++pos) {
+    if (stbFullName[pos] == TS_PATH_DELIMITER[0]) num++;
     if (num == 2) break;
   }
 
   if (num == 2) {
-    strcpy(name, tableId + pos + 1);
+    tstrncpy(dst, stbFullName + pos + 1, dstSize);
   }
 }
 
@@ -1808,7 +1815,7 @@ static int32_t mndRetrieveStb(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBloc
 
     SName name = {0};
     char  stbName[TSDB_TABLE_NAME_LEN + VARSTR_HEADER_SIZE] = {0};
-    mndExtractTableName(pStb->name, &stbName[VARSTR_HEADER_SIZE]);
+    mndExtractTbNameFromStbFullName(pStb->name, &stbName[VARSTR_HEADER_SIZE], TSDB_TABLE_NAME_LEN);
     varDataSetLen(stbName, strlen(&stbName[VARSTR_HEADER_SIZE]));
 
     SColumnInfoData *pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
diff --git a/source/dnode/mnode/impl/src/mndStream.c b/source/dnode/mnode/impl/src/mndStream.c
index 491835fc15..de0b25b1dc 100644
--- a/source/dnode/mnode/impl/src/mndStream.c
+++ b/source/dnode/mnode/impl/src/mndStream.c
@@ -196,7 +196,8 @@ void mndReleaseStream(SMnode *pMnode, SStreamObj *pStream) {
 }
 
 static int32_t mndCheckCreateStreamReq(SCMCreateStreamReq *pCreate) {
-  if (pCreate->name[0] == 0 || pCreate->sql == NULL || pCreate->sql[0] == 0) {
+  if (pCreate->name[0] == 0 || pCreate->sql == NULL || pCreate->sql[0] == 0 || pCreate->sourceDB[0] == 0 ||
+      pCreate->targetStbFullName[0] == 0) {
     terrno = TSDB_CODE_MND_INVALID_STREAM_OPTION;
     return -1;
   }
@@ -232,6 +233,114 @@ static int32_t mndStreamGetPlanString(const char *ast, int8_t triggerType, int64
   return code;
 }
 
+static int32_t mndBuildStreamObjFromCreateReq(SMnode *pMnode, SStreamObj *pObj, SCMCreateStreamReq *pCreate) {
+  SNode *pAst = NULL;
+
+  mDebug("stream:%s to create", pCreate->name);
+  memcpy(pObj->name, pCreate->name, TSDB_STREAM_FNAME_LEN);
+  pObj->createTime = taosGetTimestampMs();
+  pObj->updateTime = pObj->createTime;
+  pObj->version = 1;
+  pObj->smaId = 0;
+
+  pObj->uid = mndGenerateUid(pObj->name, strlen(pObj->name));
+  pObj->status = 0;
+
+  // TODO
+  pObj->dropPolicy = 0;
+  pObj->trigger = pCreate->triggerType;
+  pObj->triggerParam = pCreate->maxDelay;
+  pObj->watermark = pCreate->watermark;
+
+  memcpy(pObj->sourceDb, pCreate->sourceDB, TSDB_DB_FNAME_LEN);
+  SDbObj *pSourceDb = mndAcquireDb(pMnode, pCreate->sourceDB);
+  if (pSourceDb == NULL) {
+    /*ASSERT(0);*/
+    mDebug("stream:%s failed to create, source db %s not exist", pCreate->name, pObj->sourceDb);
+    terrno = TSDB_CODE_MND_DB_NOT_EXIST;
+    return -1;
+  }
+  pObj->sourceDbUid = pSourceDb->uid;
+
+  memcpy(pObj->targetSTbName, pCreate->targetStbFullName, TSDB_TABLE_FNAME_LEN);
+
+  SDbObj *pTargetDb = mndAcquireDbByStb(pMnode, pObj->targetSTbName);
+  if (pTargetDb == NULL) {
+    mDebug("stream:%s failed to create, target db %s not exist", pCreate->name, pObj->targetDb);
+    terrno = TSDB_CODE_MND_DB_NOT_EXIST;
+    return -1;
+  }
+  tstrncpy(pObj->targetDb, pTargetDb->name, TSDB_DB_FNAME_LEN);
+
+  pObj->targetStbUid = mndGenerateUid(pObj->targetSTbName, TSDB_TABLE_FNAME_LEN);
+  pObj->targetDbUid = pTargetDb->uid;
+
+  pObj->sql = pCreate->sql;
+  pObj->ast = pCreate->ast;
+
+  pCreate->sql = NULL;
+  pCreate->ast = NULL;
+
+  // deserialize ast
+  if (nodesStringToNode(pObj->ast, &pAst) < 0) {
+    /*ASSERT(0);*/
+    goto FAIL;
+  }
+
+  // extract output schema from ast
+  if (qExtractResultSchema(pAst, (int32_t *)&pObj->outputSchema.nCols, &pObj->outputSchema.pSchema) != 0) {
+    /*ASSERT(0);*/
+    goto FAIL;
+  }
+
+  SQueryPlan  *pPlan = NULL;
+  SPlanContext cxt = {
+      .pAstRoot = pAst,
+      .topicQuery = false,
+      .streamQuery = true,
+      .triggerType = pObj->trigger == STREAM_TRIGGER_MAX_DELAY ? STREAM_TRIGGER_WINDOW_CLOSE : pObj->trigger,
+      .watermark = pObj->watermark,
+  };
+
+  // using ast and param to build physical plan
+  if (qCreateQueryPlan(&cxt, &pPlan, NULL) < 0) {
+    /*ASSERT(0);*/
+    goto FAIL;
+  }
+
+  // save physcial plan
+  if (nodesNodeToString((SNode *)pPlan, false, &pObj->physicalPlan, NULL) != 0) {
+    /*ASSERT(0);*/
+    goto FAIL;
+  }
+
+FAIL:
+  if (pAst != NULL) nodesDestroyNode(pAst);
+  return 0;
+}
+
+int32_t mndPersistStream(SMnode *pMnode, STrans *pTrans, SStreamObj *pStream) {
+  SSdbRaw *pCommitRaw = mndStreamActionEncode(pStream);
+  if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) {
+    mError("trans:%d, failed to append commit log since %s", pTrans->id, terrstr());
+    mndTransDrop(pTrans);
+    return -1;
+  }
+  sdbSetRawStatus(pCommitRaw, SDB_STATUS_READY);
+  return 0;
+}
+
+int32_t mndPersistDropStreamLog(SMnode *pMnode, STrans *pTrans, SStreamObj *pStream) {
+  SSdbRaw *pCommitRaw = mndStreamActionEncode(pStream);
+  if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) {
+    mError("trans:%d, failed to append commit log since %s", pTrans->id, terrstr());
+    mndTransDrop(pTrans);
+    return -1;
+  }
+  sdbSetRawStatus(pCommitRaw, SDB_STATUS_DROPPED);
+  return 0;
+}
+
 int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast, STrans *pTrans) {
   SNode *pAst = NULL;
 
@@ -356,24 +465,24 @@ static int32_t mndPersistTaskDropReq(STrans *pTrans, SStreamTask *pTask) {
   ASSERT(pTask->nodeId != 0);
 
   // vnode
-  if (pTask->nodeId > 0) {
-    SVDropStreamTaskReq *pReq = taosMemoryCalloc(1, sizeof(SVDropStreamTaskReq));
-    if (pReq == NULL) {
-      terrno = TSDB_CODE_OUT_OF_MEMORY;
-      return -1;
-    }
-    pReq->head.vgId = htonl(pTask->nodeId);
-    pReq->taskId = pTask->taskId;
-    STransAction action = {0};
-    memcpy(&action.epSet, &pTask->epSet, sizeof(SEpSet));
-    action.pCont = pReq;
-    action.contLen = sizeof(SVDropStreamTaskReq);
-    action.msgType = TDMT_VND_STREAM_TASK_DROP;
-    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-      taosMemoryFree(pReq);
-      return -1;
-    }
+  /*if (pTask->nodeId > 0) {*/
+  SVDropStreamTaskReq *pReq = taosMemoryCalloc(1, sizeof(SVDropStreamTaskReq));
+  if (pReq == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return -1;
   }
+  pReq->head.vgId = htonl(pTask->nodeId);
+  pReq->taskId = pTask->taskId;
+  STransAction action = {0};
+  memcpy(&action.epSet, &pTask->epSet, sizeof(SEpSet));
+  action.pCont = pReq;
+  action.contLen = sizeof(SVDropStreamTaskReq);
+  action.msgType = TDMT_VND_STREAM_TASK_DROP;
+  if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+    taosMemoryFree(pReq);
+    return -1;
+  }
+  /*}*/
 
   return 0;
 }
@@ -403,10 +512,9 @@ static int32_t mndCreateStream(SMnode *pMnode, SRpcMsg *pReq, SCMCreateStreamReq
   streamObj.updateTime = streamObj.createTime;
   streamObj.uid = mndGenerateUid(pCreate->name, strlen(pCreate->name));
   streamObj.targetStbUid = mndGenerateUid(pCreate->targetStbFullName, TSDB_TABLE_FNAME_LEN);
-  streamObj.dbUid = pDb->uid;
+  streamObj.sourceDbUid = pDb->uid;
   streamObj.version = 1;
   streamObj.sql = pCreate->sql;
-  streamObj.createdBy = STREAM_CREATED_BY__USER;
   // TODO
   streamObj.fixedSinkVgId = 0;
   streamObj.smaId = 0;
@@ -486,7 +594,8 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  // TODO check auth
+  // TODO check read auth for source and write auth for target
+#if 0
   pDb = mndAcquireDb(pMnode, createStreamReq.sourceDB);
   if (pDb == NULL) {
     terrno = TSDB_CODE_MND_DB_NOT_SELECTED;
@@ -496,9 +605,52 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
   if (mndCheckDbAuth(pMnode, pReq->conn.user, MND_OPER_WRITE_DB, pDb) != 0) {
     goto _OVER;
   }
+#endif
 
-  code = mndCreateStream(pMnode, pReq, &createStreamReq, pDb);
-  if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;
+  // build stream obj from request
+  //
+  // schedule stream task for stream obj
+  //
+  SStreamObj streamObj = {0};
+  if (mndBuildStreamObjFromCreateReq(pMnode, &streamObj, &createStreamReq) < 0) {
+    ASSERT(0);
+    mError("stream:%s, failed to create since %s", createStreamReq.name, terrstr());
+    goto _OVER;
+  }
+
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB_INSIDE, pReq);
+  if (pTrans == NULL) {
+    mError("stream:%s, failed to create since %s", createStreamReq.name, terrstr());
+    goto _OVER;
+  }
+  mndTransSetDbName(pTrans, streamObj.sourceDb);
+  // TODO
+  /*mndTransSetDbName(pTrans, streamObj.targetDb);*/
+  mDebug("trans:%d, used to create stream:%s", pTrans->id, createStreamReq.name);
+
+  if (mndCreateStbForStream(pMnode, pTrans, &streamObj, pReq->conn.user) < 0) {
+    mError("trans:%d, failed to create stb for stream %s since %s", pTrans->id, createStreamReq.name, terrstr());
+    mndTransDrop(pTrans);
+    goto _OVER;
+  }
+
+  if (mndScheduleStream(pMnode, pTrans, &streamObj) < 0) {
+    mError("stream:%s, failed to schedule since %s", createStreamReq.name, terrstr());
+    mndTransDrop(pTrans);
+    goto _OVER;
+  }
+
+  if (mndTransPrepare(pMnode, pTrans) != 0) {
+    mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
+    mndTransDrop(pTrans);
+    goto _OVER;
+  }
+
+  mndTransDrop(pTrans);
+
+  /*code = mndCreateStream(pMnode, pReq, &createStreamReq, pDb);*/
+  /*if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;*/
+  code = TSDB_CODE_ACTION_IN_PROGRESS;
 
 _OVER:
   if (code != 0 && code != TSDB_CODE_ACTION_IN_PROGRESS) {
@@ -555,10 +707,54 @@ static int32_t mndProcessDropStreamReq(SRpcMsg *pReq) {
     return code;
   }
 
+  // drop  stream
+  if (mndPersistDropStreamLog(pMnode, pTrans, pStream) < 0) {
+    sdbRelease(pMnode->pSdb, pStream);
+    return -1;
+  }
+
 DROP_STREAM_OVER:
   return code;
 }
 
+int32_t mndDropStreamByDb(SMnode *pMnode, STrans *pTrans, SDbObj *pDb) {
+  SSdb *pSdb = pMnode->pSdb;
+
+  void       *pIter = NULL;
+  SStreamObj *pStream = NULL;
+  while (1) {
+    pIter = sdbFetch(pSdb, SDB_STREAM, pIter, (void **)&pStream);
+    if (pIter == NULL) break;
+
+    if (pStream->sourceDbUid == pDb->uid || pStream->targetDbUid == pDb->uid) {
+      if (pStream->sourceDbUid != pStream->targetDbUid) {
+        sdbRelease(pSdb, pStream);
+        return -1;
+      } else {
+        // TODO drop all task on snode
+        if (mndPersistDropStreamLog(pMnode, pTrans, pStream) < 0) {
+          sdbRelease(pSdb, pStream);
+          return -1;
+        }
+      }
+    } else {
+      sdbRelease(pSdb, pStream);
+      continue;
+    }
+
+#if 0
+    if (mndSetDropOffsetStreamLogs(pMnode, pTrans, pStream) < 0) {
+      sdbRelease(pSdb, pStream);
+      goto END;
+    }
+#endif
+
+    sdbRelease(pSdb, pStream);
+  }
+
+  return 0;
+}
+
 static int32_t mndGetNumOfStreams(SMnode *pMnode, char *dbName, int32_t *pNumOfStreams) {
   SSdb   *pSdb = pMnode->pSdb;
   SDbObj *pDb = mndAcquireDb(pMnode, dbName);
@@ -574,7 +770,7 @@ static int32_t mndGetNumOfStreams(SMnode *pMnode, char *dbName, int32_t *pNumOfS
     pIter = sdbFetch(pSdb, SDB_STREAM, pIter, (void **)&pStream);
     if (pIter == NULL) break;
 
-    if (pStream->dbUid == pDb->uid) {
+    if (pStream->sourceDbUid == pDb->uid) {
       numOfStreams++;
     }
 

From 2f81a41f98ed0a49d7329eb561753971b6a4faba Mon Sep 17 00:00:00 2001
From: Minghao Li <castermode@gmail.com>
Date: Thu, 16 Jun 2022 16:16:40 +0800
Subject: [PATCH 25/40] refactor(sync): add syncIsReady

---
 include/libs/sync/sync.h        |  1 +
 source/libs/sync/src/syncMain.c | 11 +++++++++++
 2 files changed, 12 insertions(+)

diff --git a/include/libs/sync/sync.h b/include/libs/sync/sync.h
index e694203563..fc9d419b1f 100644
--- a/include/libs/sync/sync.h
+++ b/include/libs/sync/sync.h
@@ -182,6 +182,7 @@ void        syncStart(int64_t rid);
 void        syncStop(int64_t rid);
 int32_t     syncSetStandby(int64_t rid);
 ESyncState  syncGetMyRole(int64_t rid);
+bool        syncIsReady(int64_t rid);
 const char* syncGetMyRoleStr(int64_t rid);
 SyncTerm    syncGetMyTerm(int64_t rid);
 void        syncGetEpSet(int64_t rid, SEpSet* pEpSet);
diff --git a/source/libs/sync/src/syncMain.c b/source/libs/sync/src/syncMain.c
index 05a41daa9e..cc2c82fa5f 100644
--- a/source/libs/sync/src/syncMain.c
+++ b/source/libs/sync/src/syncMain.c
@@ -381,6 +381,17 @@ ESyncState syncGetMyRole(int64_t rid) {
   return state;
 }
 
+bool syncIsReady(int64_t rid) {
+  SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
+  if (pSyncNode == NULL) {
+    return false;
+  }
+  assert(rid == pSyncNode->rid);
+  bool b = (pSyncNode->state == TAOS_SYNC_STATE_LEADER) && pSyncNode->restoreFinish;
+  taosReleaseRef(tsNodeRefId, pSyncNode->rid);
+  return b;
+}
+
 bool syncIsRestoreFinish(int64_t rid) {
   SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
   if (pSyncNode == NULL) {

From 87300b2ab047d2fecdffb0c4b542c8379b3af8d4 Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 16:21:01 +0800
Subject: [PATCH 26/40] fix(query): set correct error code.

---
 source/libs/executor/src/executorimpl.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c
index 9ce22431e1..bcd7b7f86f 100644
--- a/source/libs/executor/src/executorimpl.c
+++ b/source/libs/executor/src/executorimpl.c
@@ -4532,12 +4532,14 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
       tsdbReaderT pDataReader =
           doCreateDataReader(pTableScanNode, pHandle, pTableListInfo, (uint64_t)queryId, taskId, pTagCond);
       if (pDataReader == NULL && terrno != 0) {
+        pTaskInfo->code = terrno;
         return NULL;
       }
 
       int32_t code = extractTableSchemaVersion(pHandle, pTableScanNode->scan.uid, pTaskInfo);
       if (code) {
         tsdbCleanupReadHandle(pDataReader);
+        pTaskInfo->code = terrno;
         return NULL;
       }
 
@@ -4546,6 +4548,7 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
       taosArrayDestroy(groupKeys);
       if (code) {
         tsdbCleanupReadHandle(pDataReader);
+        pTaskInfo->code = terrno;
         return NULL;
       }
 

From 32d8e60ce249dc3cc571736bf2c7eba0327a2613 Mon Sep 17 00:00:00 2001
From: Liu Jicong <liujicong@qq.com>
Date: Thu, 16 Jun 2022 16:22:58 +0800
Subject: [PATCH 27/40] refactor(sma): create

---
 source/dnode/mnode/impl/inc/mndDef.h       |  8 +--
 source/dnode/mnode/impl/inc/mndScheduler.h |  2 +
 source/dnode/mnode/impl/inc/mndStream.h    |  2 +-
 source/dnode/mnode/impl/src/mndDef.c       | 79 ++++++++++++----------
 source/dnode/mnode/impl/src/mndScheduler.c |  2 -
 source/dnode/mnode/impl/src/mndSma.c       | 38 ++++++++++-
 source/dnode/mnode/impl/src/mndStream.c    |  8 ++-
 7 files changed, 93 insertions(+), 46 deletions(-)

diff --git a/source/dnode/mnode/impl/inc/mndDef.h b/source/dnode/mnode/impl/inc/mndDef.h
index db18f87830..4daeeaa9bf 100644
--- a/source/dnode/mnode/impl/inc/mndDef.h
+++ b/source/dnode/mnode/impl/inc/mndDef.h
@@ -546,13 +546,10 @@ typedef struct {
   int64_t createTime;
   int64_t updateTime;
   int32_t version;
-  // TODO remove it
   int64_t smaId;  // 0 for unused
   // info
   int64_t uid;
   int8_t  status;
-  // TODO remove it
-  int32_t vgNum;
   // config
   int8_t  dropPolicy;
   int8_t  trigger;
@@ -565,9 +562,10 @@ typedef struct {
   char    targetDb[TSDB_DB_FNAME_LEN];
   char    targetSTbName[TSDB_TABLE_FNAME_LEN];
   int64_t targetStbUid;
-  // assigned when scheduling
   int32_t fixedSinkVgId;  // 0 for shuffle
-  SVgObj  fixedSinkVg;
+  // fixedSinkVg is not applicable for encode and decode
+  SVgObj fixedSinkVg;
+
   // transformation
   char*          sql;
   char*          ast;
diff --git a/source/dnode/mnode/impl/inc/mndScheduler.h b/source/dnode/mnode/impl/inc/mndScheduler.h
index 80fe472c56..8e816d2dd6 100644
--- a/source/dnode/mnode/impl/inc/mndScheduler.h
+++ b/source/dnode/mnode/impl/inc/mndScheduler.h
@@ -32,6 +32,8 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream);
 int32_t mndConvertRsmaTask(char** pDst, int32_t* pDstLen, const char* ast, int64_t uid, int8_t triggerType,
                            int64_t watermark, double filesFactor);
 
+int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/source/dnode/mnode/impl/inc/mndStream.h b/source/dnode/mnode/impl/inc/mndStream.h
index 65017027dc..6616e84488 100644
--- a/source/dnode/mnode/impl/inc/mndStream.h
+++ b/source/dnode/mnode/impl/inc/mndStream.h
@@ -31,7 +31,7 @@ void        mndReleaseStream(SMnode *pMnode, SStreamObj *pStream);
 SSdbRaw *mndStreamActionEncode(SStreamObj *pStream);
 SSdbRow *mndStreamActionDecode(SSdbRaw *pRaw);
 
-int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast, STrans *pTrans);
+int32_t mndPersistStream(SMnode *pMnode, STrans *pTrans, SStreamObj *pStream);
 
 int32_t mndDropStreamByDb(SMnode *pMnode, STrans *pTrans, SDbObj *pDb);
 
diff --git a/source/dnode/mnode/impl/src/mndDef.c b/source/dnode/mnode/impl/src/mndDef.c
index b9bd1e4298..20ba71992e 100644
--- a/source/dnode/mnode/impl/src/mndDef.c
+++ b/source/dnode/mnode/impl/src/mndDef.c
@@ -18,35 +18,35 @@
 #include "mndConsumer.h"
 
 int32_t tEncodeSStreamObj(SEncoder *pEncoder, const SStreamObj *pObj) {
-  int32_t sz = 0;
   if (tEncodeCStr(pEncoder, pObj->name) < 0) return -1;
+
+  if (tEncodeI64(pEncoder, pObj->createTime) < 0) return -1;
+  if (tEncodeI64(pEncoder, pObj->updateTime) < 0) return -1;
+  if (tEncodeI32(pEncoder, pObj->version) < 0) return -1;
+  if (tEncodeI64(pEncoder, pObj->smaId) < 0) return -1;
+
+  if (tEncodeI64(pEncoder, pObj->uid) < 0) return -1;
+  if (tEncodeI8(pEncoder, pObj->status) < 0) return -1;
+
+  if (tEncodeI8(pEncoder, pObj->dropPolicy) < 0) return -1;
+  if (tEncodeI8(pEncoder, pObj->trigger) < 0) return -1;
+  if (tEncodeI64(pEncoder, pObj->triggerParam) < 0) return -1;
+  if (tEncodeI64(pEncoder, pObj->watermark) < 0) return -1;
+
+  if (tEncodeI64(pEncoder, pObj->sourceDbUid) < 0) return -1;
+  if (tEncodeI64(pEncoder, pObj->targetDbUid) < 0) return -1;
   if (tEncodeCStr(pEncoder, pObj->sourceDb) < 0) return -1;
   if (tEncodeCStr(pEncoder, pObj->targetDb) < 0) return -1;
   if (tEncodeCStr(pEncoder, pObj->targetSTbName) < 0) return -1;
   if (tEncodeI64(pEncoder, pObj->targetStbUid) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->createTime) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->updateTime) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->uid) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->sourceDbUid) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->targetDbUid) < 0) return -1;
-  if (tEncodeI32(pEncoder, pObj->version) < 0) return -1;
-  if (tEncodeI32(pEncoder, pObj->vgNum) < 0) return -1;
-  if (tEncodeI8(pEncoder, pObj->dropPolicy) < 0) return -1;
-  if (tEncodeI8(pEncoder, pObj->status) < 0) return -1;
-  if (tEncodeI8(pEncoder, pObj->trigger) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->triggerParam) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->watermark) < 0) return -1;
   if (tEncodeI32(pEncoder, pObj->fixedSinkVgId) < 0) return -1;
-  if (tEncodeI64(pEncoder, pObj->smaId) < 0) return -1;
-  if (tEncodeCStr(pEncoder, pObj->sql) < 0) return -1;
-  /*if (tEncodeCStr(pEncoder, pObj->logicalPlan) < 0) return -1;*/
-  if (tEncodeCStr(pEncoder, pObj->physicalPlan) < 0) return -1;
-  // TODO encode tasks
-  if (pObj->tasks) {
-    sz = taosArrayGetSize(pObj->tasks);
-  }
-  if (tEncodeI32(pEncoder, sz) < 0) return -1;
 
+  if (tEncodeCStr(pEncoder, pObj->sql) < 0) return -1;
+  if (tEncodeCStr(pEncoder, pObj->ast) < 0) return -1;
+  if (tEncodeCStr(pEncoder, pObj->physicalPlan) < 0) return -1;
+
+  int32_t sz = taosArrayGetSize(pObj->tasks);
+  if (tEncodeI32(pEncoder, sz) < 0) return -1;
   for (int32_t i = 0; i < sz; i++) {
     SArray *pArray = taosArrayGetP(pObj->tasks, i);
     int32_t innerSz = taosArrayGetSize(pArray);
@@ -58,32 +58,38 @@ int32_t tEncodeSStreamObj(SEncoder *pEncoder, const SStreamObj *pObj) {
   }
 
   if (tEncodeSSchemaWrapper(pEncoder, &pObj->outputSchema) < 0) return -1;
+
   return pEncoder->pos;
 }
 
 int32_t tDecodeSStreamObj(SDecoder *pDecoder, SStreamObj *pObj) {
   if (tDecodeCStrTo(pDecoder, pObj->name) < 0) return -1;
+
+  if (tDecodeI64(pDecoder, &pObj->createTime) < 0) return -1;
+  if (tDecodeI64(pDecoder, &pObj->updateTime) < 0) return -1;
+  if (tDecodeI32(pDecoder, &pObj->version) < 0) return -1;
+  if (tDecodeI64(pDecoder, &pObj->smaId) < 0) return -1;
+
+  if (tDecodeI64(pDecoder, &pObj->uid) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pObj->status) < 0) return -1;
+
+  if (tDecodeI8(pDecoder, &pObj->dropPolicy) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pObj->trigger) < 0) return -1;
+  if (tDecodeI64(pDecoder, &pObj->triggerParam) < 0) return -1;
+  if (tDecodeI64(pDecoder, &pObj->watermark) < 0) return -1;
+
+  if (tDecodeI64(pDecoder, &pObj->sourceDbUid) < 0) return -1;
+  if (tDecodeI64(pDecoder, &pObj->targetDbUid) < 0) return -1;
   if (tDecodeCStrTo(pDecoder, pObj->sourceDb) < 0) return -1;
   if (tDecodeCStrTo(pDecoder, pObj->targetDb) < 0) return -1;
   if (tDecodeCStrTo(pDecoder, pObj->targetSTbName) < 0) return -1;
   if (tDecodeI64(pDecoder, &pObj->targetStbUid) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->createTime) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->updateTime) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->uid) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->sourceDbUid) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->targetDbUid) < 0) return -1;
-  if (tDecodeI32(pDecoder, &pObj->version) < 0) return -1;
-  if (tDecodeI32(pDecoder, &pObj->vgNum) < 0) return -1;
-  if (tDecodeI8(pDecoder, &pObj->dropPolicy) < 0) return -1;
-  if (tDecodeI8(pDecoder, &pObj->status) < 0) return -1;
-  if (tDecodeI8(pDecoder, &pObj->trigger) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->triggerParam) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->watermark) < 0) return -1;
   if (tDecodeI32(pDecoder, &pObj->fixedSinkVgId) < 0) return -1;
-  if (tDecodeI64(pDecoder, &pObj->smaId) < 0) return -1;
+
   if (tDecodeCStrAlloc(pDecoder, &pObj->sql) < 0) return -1;
-  /*if (tDecodeCStrAlloc(pDecoder, &pObj->logicalPlan) < 0) return -1;*/
+  if (tDecodeCStrAlloc(pDecoder, &pObj->ast) < 0) return -1;
   if (tDecodeCStrAlloc(pDecoder, &pObj->physicalPlan) < 0) return -1;
+
   pObj->tasks = NULL;
   int32_t sz;
   if (tDecodeI32(pDecoder, &sz) < 0) return -1;
@@ -104,6 +110,7 @@ int32_t tDecodeSStreamObj(SDecoder *pDecoder, SStreamObj *pObj) {
   }
 
   if (tDecodeSSchemaWrapper(pDecoder, &pObj->outputSchema) < 0) return -1;
+
   return 0;
 }
 
diff --git a/source/dnode/mnode/impl/src/mndScheduler.c b/source/dnode/mnode/impl/src/mndScheduler.c
index c53e97cd53..6f8fc748c2 100644
--- a/source/dnode/mnode/impl/src/mndScheduler.c
+++ b/source/dnode/mnode/impl/src/mndScheduler.c
@@ -346,8 +346,6 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
     terrno = TSDB_CODE_QRY_INVALID_INPUT;
     return -1;
   }
-  ASSERT(pStream->vgNum == 0);
-
   int32_t totLevel = LIST_LENGTH(pPlan->pSubplans);
   ASSERT(totLevel <= 2);
   pStream->tasks = taosArrayInit(totLevel, sizeof(void*));
diff --git a/source/dnode/mnode/impl/src/mndSma.c b/source/dnode/mnode/impl/src/mndSma.c
index 54fd0fe867..3eb5b03efd 100644
--- a/source/dnode/mnode/impl/src/mndSma.c
+++ b/source/dnode/mnode/impl/src/mndSma.c
@@ -20,6 +20,7 @@
 #include "mndDnode.h"
 #include "mndInfoSchema.h"
 #include "mndMnode.h"
+#include "mndScheduler.h"
 #include "mndShow.h"
 #include "mndStb.h"
 #include "mndStream.h"
@@ -563,6 +564,7 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   streamObj.smaId = smaObj.uid;
   streamObj.watermark = 0;
   streamObj.trigger = STREAM_TRIGGER_AT_ONCE;
+  streamObj.ast = strdup(smaObj.ast);
 
   if (mndAllocSmaVgroup(pMnode, pDb, &streamObj.fixedSinkVg) != 0) {
     mError("sma:%s, failed to create since %s", smaObj.name, terrstr());
@@ -571,6 +573,39 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   smaObj.dstVgId = streamObj.fixedSinkVg.vgId;
   streamObj.fixedSinkVgId = smaObj.dstVgId;
 
+  SNode *pAst = NULL;
+  if (nodesStringToNode(streamObj.ast, &pAst) < 0) {
+    ASSERT(0);
+    return -1;
+  }
+
+  // extract output schema from ast
+  if (qExtractResultSchema(pAst, (int32_t *)&streamObj.outputSchema.nCols, &streamObj.outputSchema.pSchema) != 0) {
+    ASSERT(0);
+    return -1;
+  }
+
+  SQueryPlan  *pPlan = NULL;
+  SPlanContext cxt = {
+      .pAstRoot = pAst,
+      .topicQuery = false,
+      .streamQuery = true,
+      .triggerType = streamObj.trigger,
+      .watermark = streamObj.watermark,
+  };
+
+  if (qCreateQueryPlan(&cxt, &pPlan, NULL) < 0) {
+    ASSERT(0);
+    return -1;
+  }
+
+  // save physcial plan
+  if (nodesNodeToString((SNode *)pPlan, false, &streamObj.physicalPlan, NULL) != 0) {
+    ASSERT(0);
+    return -1;
+  }
+  if (pAst != NULL) nodesDestroyNode(pAst);
+
   int32_t code = -1;
   STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_DB, pReq);
   if (pTrans == NULL) goto _OVER;
@@ -585,7 +620,8 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   if (mndSetUpdateSmaStbCommitLogs(pMnode, pTrans, pStb) != 0) goto _OVER;
   // if (mndSetCreateSmaRedoActions(pMnode, pTrans, pDb, &smaObj) != 0) goto _OVER;
   if (mndSetCreateSmaVgroupRedoActions(pMnode, pTrans, pDb, &streamObj.fixedSinkVg, &smaObj) != 0) goto _OVER;
-  if (mndAddStreamToTrans(pMnode, &streamObj, pCreate->ast, pTrans) != 0) goto _OVER;
+  if (mndScheduleStream(pMnode, pTrans, &streamObj) != 0) goto _OVER;
+  if (mndPersistStream(pMnode, pTrans, &streamObj) != 0) goto _OVER;
   if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
 
   code = 0;
diff --git a/source/dnode/mnode/impl/src/mndStream.c b/source/dnode/mnode/impl/src/mndStream.c
index de0b25b1dc..c1df9dd68c 100644
--- a/source/dnode/mnode/impl/src/mndStream.c
+++ b/source/dnode/mnode/impl/src/mndStream.c
@@ -640,6 +640,12 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
+  if (mndPersistStream(pMnode, pTrans, &streamObj) < 0) {
+    mError("stream:%s, failed to schedule since %s", createStreamReq.name, terrstr());
+    mndTransDrop(pTrans);
+    goto _OVER;
+  }
+
   if (mndTransPrepare(pMnode, pTrans) != 0) {
     mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
     mndTransDrop(pTrans);
@@ -707,7 +713,7 @@ static int32_t mndProcessDropStreamReq(SRpcMsg *pReq) {
     return code;
   }
 
-  // drop  stream
+  // drop stream
   if (mndPersistDropStreamLog(pMnode, pTrans, pStream) < 0) {
     sdbRelease(pMnode->pSdb, pStream);
     return -1;

From 44e733645cf8fd75b5fc291e960d39b152405062 Mon Sep 17 00:00:00 2001
From: Liu Jicong <liujicong@qq.com>
Date: Thu, 16 Jun 2022 16:35:55 +0800
Subject: [PATCH 28/40] ci: remove unstable case

---
 source/dnode/mnode/impl/inc/mndStream.h |  3 +--
 source/dnode/mnode/impl/src/mndStream.c | 28 +++++++++++++------------
 tests/system-test/fulltest.sh           |  2 +-
 3 files changed, 17 insertions(+), 16 deletions(-)

diff --git a/source/dnode/mnode/impl/inc/mndStream.h b/source/dnode/mnode/impl/inc/mndStream.h
index 6616e84488..69385c3a46 100644
--- a/source/dnode/mnode/impl/inc/mndStream.h
+++ b/source/dnode/mnode/impl/inc/mndStream.h
@@ -31,9 +31,8 @@ void        mndReleaseStream(SMnode *pMnode, SStreamObj *pStream);
 SSdbRaw *mndStreamActionEncode(SStreamObj *pStream);
 SSdbRow *mndStreamActionDecode(SSdbRaw *pRaw);
 
-int32_t mndPersistStream(SMnode *pMnode, STrans *pTrans, SStreamObj *pStream);
-
 int32_t mndDropStreamByDb(SMnode *pMnode, STrans *pTrans, SDbObj *pDb);
+int32_t mndPersistStream(SMnode *pMnode, STrans *pTrans, SStreamObj *pStream);
 
 #ifdef __cplusplus
 }
diff --git a/source/dnode/mnode/impl/src/mndStream.c b/source/dnode/mnode/impl/src/mndStream.c
index c1df9dd68c..0d5bc9b021 100644
--- a/source/dnode/mnode/impl/src/mndStream.c
+++ b/source/dnode/mnode/impl/src/mndStream.c
@@ -607,10 +607,18 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
   }
 #endif
 
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB_INSIDE, pReq);
+  if (pTrans == NULL) {
+    mError("stream:%s, failed to create since %s", createStreamReq.name, terrstr());
+    goto _OVER;
+  }
+
+  mndTransSetDbName(pTrans, createStreamReq.sourceDB);
+  // TODO
+  /*mndTransSetDbName(pTrans, streamObj.targetDb);*/
+  mDebug("trans:%d, used to create stream:%s", pTrans->id, createStreamReq.name);
+
   // build stream obj from request
-  //
-  // schedule stream task for stream obj
-  //
   SStreamObj streamObj = {0};
   if (mndBuildStreamObjFromCreateReq(pMnode, &streamObj, &createStreamReq) < 0) {
     ASSERT(0);
@@ -618,34 +626,28 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB_INSIDE, pReq);
-  if (pTrans == NULL) {
-    mError("stream:%s, failed to create since %s", createStreamReq.name, terrstr());
-    goto _OVER;
-  }
-  mndTransSetDbName(pTrans, streamObj.sourceDb);
-  // TODO
-  /*mndTransSetDbName(pTrans, streamObj.targetDb);*/
-  mDebug("trans:%d, used to create stream:%s", pTrans->id, createStreamReq.name);
-
+  // create stb for stream
   if (mndCreateStbForStream(pMnode, pTrans, &streamObj, pReq->conn.user) < 0) {
     mError("trans:%d, failed to create stb for stream %s since %s", pTrans->id, createStreamReq.name, terrstr());
     mndTransDrop(pTrans);
     goto _OVER;
   }
 
+  // schedule stream task for stream obj
   if (mndScheduleStream(pMnode, pTrans, &streamObj) < 0) {
     mError("stream:%s, failed to schedule since %s", createStreamReq.name, terrstr());
     mndTransDrop(pTrans);
     goto _OVER;
   }
 
+  // add stream to trans
   if (mndPersistStream(pMnode, pTrans, &streamObj) < 0) {
     mError("stream:%s, failed to schedule since %s", createStreamReq.name, terrstr());
     mndTransDrop(pTrans);
     goto _OVER;
   }
 
+  // execute creation
   if (mndTransPrepare(pMnode, pTrans) != 0) {
     mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
     mndTransDrop(pTrans);
diff --git a/tests/system-test/fulltest.sh b/tests/system-test/fulltest.sh
index 552ed26f34..b9f7e0f2eb 100755
--- a/tests/system-test/fulltest.sh
+++ b/tests/system-test/fulltest.sh
@@ -101,7 +101,7 @@ python3 ./test.py -f 2-query/tail.py
 
 python3 ./test.py -f 6-cluster/5dnode1mnode.py 
 python3 ./test.py -f 6-cluster/5dnode2mnode.py 
-python3 ./test.py -f 6-cluster/5dnode3mnodeStop.py
+#python3 ./test.py -f 6-cluster/5dnode3mnodeStop.py
 python3 ./test.py -f 6-cluster/5dnode3mnodeDrop.py
 # BUG python3 ./test.py -f 6-cluster/5dnode3mnodeStopInsert.py 
 

From 07288e26b2ffeb04aadb943cc67f28b64c2113ce Mon Sep 17 00:00:00 2001
From: Liu Jicong <liujicong@qq.com>
Date: Thu, 16 Jun 2022 17:07:31 +0800
Subject: [PATCH 29/40] refactor(stream): change msg name

---
 include/common/tmsg.h                   | 4 ++--
 source/dnode/mnode/impl/src/mndStream.c | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/include/common/tmsg.h b/include/common/tmsg.h
index 01d99ac705..ab0203adf8 100644
--- a/include/common/tmsg.h
+++ b/include/common/tmsg.h
@@ -2259,11 +2259,11 @@ typedef struct {
 typedef struct {
   char   name[TSDB_STREAM_FNAME_LEN];
   int8_t igNotExists;
-} SMDropStreamTaskReq;
+} SMDropStreamReq;
 
 typedef struct {
   int8_t reserved;
-} SMDropStreamTaskRsp;
+} SMDropStreamRsp;
 
 typedef struct {
   SMsgHead head;
diff --git a/source/dnode/mnode/impl/src/mndStream.c b/source/dnode/mnode/impl/src/mndStream.c
index 0d5bc9b021..3a42c694c8 100644
--- a/source/dnode/mnode/impl/src/mndStream.c
+++ b/source/dnode/mnode/impl/src/mndStream.c
@@ -679,7 +679,7 @@ static int32_t mndProcessDropStreamReq(SRpcMsg *pReq) {
   /*SDbObj     *pDb = NULL;*/
   /*SUserObj   *pUser = NULL;*/
 
-  SMDropStreamTaskReq dropReq = *(SMDropStreamTaskReq *)pReq->pCont;
+  SMDropStreamReq dropReq = *(SMDropStreamReq *)pReq->pCont;
 
   pStream = mndAcquireStream(pMnode, dropReq.name);
 

From bd518f3f71557f86b9d5b5273603ca6d14e066a3 Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 17:39:33 +0800
Subject: [PATCH 30/40] fix(query): remove group info during sort.

---
 source/libs/executor/inc/executorimpl.h |  8 ---
 source/libs/executor/inc/tsort.h        |  6 ---
 source/libs/executor/src/executorimpl.c | 31 +++--------
 source/libs/executor/src/scanoperator.c | 14 +----
 source/libs/executor/src/sortoperator.c | 70 +++++--------------------
 source/libs/executor/src/tsort.c        | 57 ++++++--------------
 6 files changed, 37 insertions(+), 149 deletions(-)

diff --git a/source/libs/executor/inc/executorimpl.h b/source/libs/executor/inc/executorimpl.h
index b435446513..034e2893df 100644
--- a/source/libs/executor/inc/executorimpl.h
+++ b/source/libs/executor/inc/executorimpl.h
@@ -696,10 +696,6 @@ typedef struct SSortedMergeOperatorInfo {
   int32_t          numOfResPerPage;
   char**           groupVal;
   SArray          *groupInfo;
-
-  bool hasGroupId;
-  uint64_t groupId;
-  STupleHandle* prefetchedTuple;
 } SSortedMergeOperatorInfo;
 
 typedef struct SSortOperatorInfo {
@@ -712,10 +708,6 @@ typedef struct SSortOperatorInfo {
 
   int64_t      startTs;       // sort start time
   uint64_t     sortElapsed;   // sort elapsed time, time to flush to disk not included.
-
-  STupleHandle *prefetchedTuple;
-  bool  hasGroupId;
-  uint64_t groupId;
 } SSortOperatorInfo;
 
 typedef struct STagFilterOperatorInfo {
diff --git a/source/libs/executor/inc/tsort.h b/source/libs/executor/inc/tsort.h
index e731c55a7d..86ee841cc2 100644
--- a/source/libs/executor/inc/tsort.h
+++ b/source/libs/executor/inc/tsort.h
@@ -130,12 +130,6 @@ bool tsortIsNullVal(STupleHandle* pVHandle, int32_t colId);
  */
 void* tsortGetValue(STupleHandle* pVHandle, int32_t colId);
 
-/**
- *
- * @param pVHandle
- * @return
- */
-uint64_t tsortGetGroupId(STupleHandle* pVHandle);
 /**
  *
  * @param pSortHandle
diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c
index bcd7b7f86f..daee71b666 100644
--- a/source/libs/executor/src/executorimpl.c
+++ b/source/libs/executor/src/executorimpl.c
@@ -3030,31 +3030,12 @@ SSDataBlock* getSortedMergeBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlo
   blockDataEnsureCapacity(p, capacity);
 
   while (1) {
-    STupleHandle* pTupleHandle = NULL;
-    if (pInfo->prefetchedTuple == NULL) {
-      pTupleHandle = tsortNextTuple(pHandle);
-    } else {
-      pTupleHandle = pInfo->prefetchedTuple;
-      pInfo->groupId = tsortGetGroupId(pTupleHandle);
-      pInfo->prefetchedTuple = NULL;
-    }
-
+    STupleHandle* pTupleHandle = tsortNextTuple(pHandle);
     if (pTupleHandle == NULL) {
       break;
     }
 
-    uint64_t tupleGroupId = tsortGetGroupId(pTupleHandle);
-    if (!pInfo->hasGroupId) {
-      pInfo->groupId = tupleGroupId;
-      pInfo->hasGroupId = true;
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else if (pInfo->groupId == tupleGroupId) {
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else {
-      pInfo->prefetchedTuple = pTupleHandle;
-      break;
-    }
-
+    appendOneRowToDataBlock(p, pTupleHandle);
     if (p->info.rows >= capacity) {
       break;
     }
@@ -3073,7 +3054,6 @@ SSDataBlock* getSortedMergeBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlo
 
     pDataBlock->info.rows = p->info.rows;
     pDataBlock->info.capacity = p->info.rows;
-    pDataBlock->info.groupId = pInfo->groupId;
   }
 
   blockDataDestroy(p);
@@ -3339,7 +3319,7 @@ static SSDataBlock* getAggregateResult(SOperatorInfo* pOperator) {
     doSetOperatorCompleted(pOperator);
   }
 
-  size_t rows = blockDataGetNumOfRows(pInfo->pRes);  // pInfo->pRes : NULL;
+  size_t rows = blockDataGetNumOfRows(pInfo->pRes);
   pOperator->resultInfo.totalRows += rows;
 
   return (rows == 0) ? NULL : pInfo->pRes;
@@ -4920,7 +4900,10 @@ SArray* extractColumnInfo(SNodeList* pNodeList) {
 }
 
 SArray* extractPartitionColInfo(SNodeList* pNodeList) {
-  if(!pNodeList) return NULL;
+  if(!pNodeList) {
+      return NULL;
+  }
+
   size_t  numOfCols = LIST_LENGTH(pNodeList);
   SArray* pList = taosArrayInit(numOfCols, sizeof(SColumn));
   if (pList == NULL) {
diff --git a/source/libs/executor/src/scanoperator.c b/source/libs/executor/src/scanoperator.c
index 1ffb1529d7..ddebba1c9e 100644
--- a/source/libs/executor/src/scanoperator.c
+++ b/source/libs/executor/src/scanoperator.c
@@ -2175,25 +2175,13 @@ SSDataBlock* getSortedTableMergeScanBlockData(SSortHandle* pHandle, int32_t capa
       pTupleHandle = tsortNextTuple(pHandle);
     } else {
       pTupleHandle = pInfo->prefetchedTuple;
-      pInfo->groupId = tsortGetGroupId(pTupleHandle);
-      pInfo->prefetchedTuple = NULL;
     }
 
     if (pTupleHandle == NULL) {
       break;
     }
 
-    uint64_t tupleGroupId = tsortGetGroupId(pTupleHandle);
-    if (!pInfo->hasGroupId) {
-      pInfo->groupId = tupleGroupId;
-      pInfo->hasGroupId = true;
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else if (pInfo->groupId == tupleGroupId) {
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else {
-      pInfo->prefetchedTuple = pTupleHandle;
-      break;
-    }
+    appendOneRowToDataBlock(p, pTupleHandle);
 
     if (p->info.rows >= capacity) {
       break;
diff --git a/source/libs/executor/src/sortoperator.c b/source/libs/executor/src/sortoperator.c
index 2f2080d5fe..81899b68cd 100644
--- a/source/libs/executor/src/sortoperator.c
+++ b/source/libs/executor/src/sortoperator.c
@@ -40,15 +40,13 @@ SOperatorInfo* createSortOperatorInfo(SOperatorInfo* downstream, SSDataBlock* pR
 
   initResultSizeInfo(pOperator, 1024);
 
-  pInfo->pSortInfo = pSortInfo;
-  pInfo->pColMatchInfo = pColMatchColInfo;
-  pInfo->hasGroupId = false;
-  pInfo->prefetchedTuple = NULL;
-  pOperator->name = "SortOperator";
+  pInfo->pSortInfo        = pSortInfo;
+  pInfo->pColMatchInfo    = pColMatchColInfo;
+  pOperator->name         = "SortOperator";
   pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_SORT;
-  pOperator->blocking = true;
-  pOperator->status = OP_NOT_OPENED;
-  pOperator->info = pInfo;
+  pOperator->blocking     = true;
+  pOperator->status       = OP_NOT_OPENED;
+  pOperator->info         = pInfo;
 
   // lazy evaluation for the following parameter since the input datablock is not known till now.
   //  pInfo->bufPageSize  = rowSize < 1024 ? 1024 * 2 : rowSize * 2;  // there are headers, so pageSize = rowSize +
@@ -97,31 +95,12 @@ SSDataBlock* getSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, i
   blockDataEnsureCapacity(p, capacity);
 
   while (1) {
-    STupleHandle* pTupleHandle = NULL;
-    if (pInfo->prefetchedTuple == NULL) {
-      pTupleHandle = tsortNextTuple(pHandle);
-    } else {
-      pTupleHandle = pInfo->prefetchedTuple;
-      pInfo->groupId = tsortGetGroupId(pTupleHandle);
-      pInfo->prefetchedTuple = NULL;
-    }
-
+    STupleHandle* pTupleHandle = tsortNextTuple(pHandle);
     if (pTupleHandle == NULL) {
       break;
     }
 
-    uint64_t tupleGroupId = tsortGetGroupId(pTupleHandle);
-    if (!pInfo->hasGroupId) {
-      pInfo->groupId = tupleGroupId;
-      pInfo->hasGroupId = true;
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else if (pInfo->groupId == tupleGroupId) {
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else {
-      pInfo->prefetchedTuple = pTupleHandle;
-      break;
-    }
-
+    appendOneRowToDataBlock(p, pTupleHandle);
     if (p->info.rows >= capacity) {
       break;
     }
@@ -140,7 +119,6 @@ SSDataBlock* getSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, i
 
     pDataBlock->info.rows = p->info.rows;
     pDataBlock->info.capacity = p->info.rows;
-    pDataBlock->info.groupId = pInfo->groupId;
   }
 
   blockDataDestroy(p);
@@ -255,10 +233,7 @@ typedef struct SMultiwaySortMergeOperatorInfo {
 
   SSDataBlock* pInputBlock;
   int64_t      startTs;  // sort start time
-
-  bool          hasGroupId;
-  uint64_t      groupId;
-  STupleHandle* prefetchedTuple;
+  uint64_t     groupId;
 } SMultiwaySortMergeOperatorInfo;
 
 int32_t doOpenMultiwaySortMergeOperator(SOperatorInfo* pOperator) {
@@ -312,31 +287,12 @@ SSDataBlock* getMultiwaySortedBlockData(SSortHandle* pHandle, SSDataBlock* pData
   blockDataEnsureCapacity(p, capacity);
 
   while (1) {
-    STupleHandle* pTupleHandle = NULL;
-    if (pInfo->prefetchedTuple == NULL) {
-      pTupleHandle = tsortNextTuple(pHandle);
-    } else {
-      pTupleHandle = pInfo->prefetchedTuple;
-      pInfo->groupId = tsortGetGroupId(pTupleHandle);
-      pInfo->prefetchedTuple = NULL;
-    }
-
+    STupleHandle* pTupleHandle = tsortNextTuple(pHandle);
     if (pTupleHandle == NULL) {
       break;
     }
 
-    uint64_t tupleGroupId = tsortGetGroupId(pTupleHandle);
-    if (!pInfo->hasGroupId) {
-      pInfo->groupId = tupleGroupId;
-      pInfo->hasGroupId = true;
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else if (pInfo->groupId == tupleGroupId) {
-      appendOneRowToDataBlock(p, pTupleHandle);
-    } else {
-      pInfo->prefetchedTuple = pTupleHandle;
-      break;
-    }
-
+    appendOneRowToDataBlock(p, pTupleHandle);
     if (p->info.rows >= capacity) {
       break;
     }
@@ -432,14 +388,12 @@ SOperatorInfo* createMultiwaySortMergeOperatorInfo(SOperatorInfo** downStreams,
   pOperator->blocking  = false;
   pOperator->status    = OP_NOT_OPENED;
   pOperator->info      = pInfo;
-  pInfo->hasGroupId    = false;
-  pInfo->prefetchedTuple = NULL;
   pOperator->pTaskInfo = pTaskInfo;
 
   pInfo->bufPageSize   = getProperSortPageSize(rowSize);
 
   uint32_t numOfSources = taosArrayGetSize(pSortInfo);
-  numOfSources = TMAX(2, numOfSources);
+  numOfSources = TMAX(4, numOfSources);
 
   pInfo->sortBufSize    = numOfSources * pInfo->bufPageSize;
 
diff --git a/source/libs/executor/src/tsort.c b/source/libs/executor/src/tsort.c
index 5bcd58f8db..1502387360 100644
--- a/source/libs/executor/src/tsort.c
+++ b/source/libs/executor/src/tsort.c
@@ -557,59 +557,40 @@ static int32_t createInitialSources(SSortHandle* pHandle) {
     SSortSource* source = taosArrayGetP(pHandle->pOrderedSource, 0);
     taosArrayClear(pHandle->pOrderedSource);
 
-    bool hasGroupId = false;
-    SSDataBlock* prefetchedDataBlock = NULL;
-
     while (1) {
-      SSDataBlock* pBlock = NULL;
-      if (prefetchedDataBlock == NULL) {
-        pBlock = pHandle->fetchfp(source->param);
-      } else {
-        pBlock = prefetchedDataBlock;
-        prefetchedDataBlock = NULL;
-      }
-
+      SSDataBlock* pBlock = pHandle->fetchfp(source->param);
       if (pBlock == NULL) {
         break;
       }
 
-      if (!hasGroupId) {
-        // calculate the buffer pages according to the total available buffers.
+      if (pHandle->pDataBlock == NULL) {
         pHandle->pageSize = getProperSortPageSize(blockDataGetRowSize(pBlock));
 
         // todo, number of pages are set according to the total available sort buffer
         pHandle->numOfPages = 1024;
         sortBufSize = pHandle->numOfPages * pHandle->pageSize;
-
-        hasGroupId = true;
         pHandle->pDataBlock = createOneDataBlock(pBlock, false);
       }
 
-      if (pHandle->pDataBlock->info.groupId == pBlock->info.groupId) {
-        // perform the scalar function calculation before apply the sort
-        if (pHandle->beforeFp != NULL) {
-          pHandle->beforeFp(pBlock, pHandle->param);
-        }
+      if (pHandle->beforeFp != NULL) {
+        pHandle->beforeFp(pBlock, pHandle->param);
+      }
 
-        int32_t code = blockDataMerge(pHandle->pDataBlock, pBlock);
-        if (code != 0) {
-          return code;
-        }
+      int32_t code = blockDataMerge(pHandle->pDataBlock, pBlock);
+      if (code != 0) {
+        return code;
+      }
 
-        size_t size = blockDataGetSize(pHandle->pDataBlock);
-        if (size > sortBufSize) {
-          // Perform the in-memory sort and then flush data in the buffer into disk.
-          int64_t p = taosGetTimestampUs();
-          blockDataSort(pHandle->pDataBlock, pHandle->pSortInfo);
+      size_t size = blockDataGetSize(pHandle->pDataBlock);
+      if (size > sortBufSize) {
+        // Perform the in-memory sort and then flush data in the buffer into disk.
+        int64_t p = taosGetTimestampUs();
+        blockDataSort(pHandle->pDataBlock, pHandle->pSortInfo);
 
-          int64_t el = taosGetTimestampUs() - p;
-          pHandle->sortElapsed += el;
+        int64_t el = taosGetTimestampUs() - p;
+        pHandle->sortElapsed += el;
 
-          doAddToBuf(pHandle->pDataBlock, pHandle);
-        }
-      } else {
-        prefetchedDataBlock = pBlock;
-        pHandle->pDataBlock = createOneDataBlock(pBlock, false);
+        doAddToBuf(pHandle->pDataBlock, pHandle);
       }
     }
 
@@ -758,10 +739,6 @@ void* tsortGetValue(STupleHandle* pVHandle, int32_t colIndex) {
   }
 }
 
-uint64_t tsortGetGroupId(STupleHandle* pVHandle) {
-  return pVHandle->pBlock->info.groupId;
-}
-
 SSortExecInfo tsortGetSortExecInfo(SSortHandle* pHandle) {
   SSortExecInfo info = {0};
 

From 7b98430906e03208282be22ac084585172ac0a01 Mon Sep 17 00:00:00 2001
From: 54liuyao <54liuyao@163.com>
Date: Thu, 16 Jun 2022 17:26:52 +0800
Subject: [PATCH 31/40] feat(stream): stream partition by

---
 source/libs/executor/src/timewindowoperator.c | 170 ++++++++++--------
 .../tsim/stream/distributeInterval0.sim       |  37 +++-
 tests/script/tsim/stream/partitionby.sim      |  62 ++++++-
 3 files changed, 195 insertions(+), 74 deletions(-)

diff --git a/source/libs/executor/src/timewindowoperator.c b/source/libs/executor/src/timewindowoperator.c
index f8972a02a1..d7ae823522 100644
--- a/source/libs/executor/src/timewindowoperator.c
+++ b/source/libs/executor/src/timewindowoperator.c
@@ -1257,6 +1257,10 @@ static int32_t getAllIntervalWindow(SHashObj* pHashMap, SArray* resWins) {
   return TSDB_CODE_SUCCESS;
 }
 
+bool isCloseWindow(STimeWindow *pWin, STimeWindowAggSupp* pSup) {
+  return pWin->ekey < pSup->maxTs - pSup->waterMark;
+}
+
 static int32_t closeIntervalWindow(SHashObj* pHashMap, STimeWindowAggSupp* pSup, SInterval* pInterval,
                                    SArray* closeWins) {
   void*  pIte = NULL;
@@ -1269,7 +1273,7 @@ static int32_t closeIntervalWindow(SHashObj* pHashMap, STimeWindowAggSupp* pSup,
     SResultRowInfo dumyInfo;
     dumyInfo.cur.pageId = -1;
     STimeWindow win = getActiveTimeWindow(NULL, &dumyInfo, ts, pInterval, pInterval->precision, NULL);
-    if (win.ekey < pSup->maxTs - pSup->waterMark) {
+    if (isCloseWindow(&win, pSup)) {
       char keyBuf[GET_RES_WINDOW_KEY_LEN(sizeof(TSKEY))];
       SET_RES_WINDOW_KEY(keyBuf, &ts, sizeof(TSKEY), groupId);
       taosHashRemove(pHashMap, keyBuf, keyLen);
@@ -2036,59 +2040,6 @@ _error:
   return NULL;
 }
 
-static void doHashInterval(SOperatorInfo* pOperatorInfo, SSDataBlock* pSDataBlock, int32_t tableGroupId,
-                           SArray* pUpdated) {
-  SStreamFinalIntervalOperatorInfo* pInfo = (SStreamFinalIntervalOperatorInfo*)pOperatorInfo->info;
-  SResultRowInfo*                   pResultRowInfo = &(pInfo->binfo.resultRowInfo);
-  SExecTaskInfo*                    pTaskInfo = pOperatorInfo->pTaskInfo;
-  int32_t                           numOfOutput = pOperatorInfo->numOfExprs;
-  int32_t                           step = 1;
-  bool                              ascScan = true;
-  TSKEY*                            tsCols = NULL;
-  SResultRow*                       pResult = NULL;
-  int32_t                           forwardRows = 0;
-
-  if (pSDataBlock->pDataBlock != NULL) {
-    SColumnInfoData* pColDataInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->primaryTsIndex);
-    tsCols = (int64_t*)pColDataInfo->pData;
-  } else {
-    return;
-  }
-
-  int32_t     startPos = ascScan ? 0 : (pSDataBlock->info.rows - 1);
-  TSKEY       ts = getStartTsKey(&pSDataBlock->info.window, tsCols);
-  STimeWindow nextWin = getActiveTimeWindow(pInfo->aggSup.pResultBuf, pResultRowInfo, ts, &pInfo->interval,
-                                            pInfo->interval.precision, NULL);
-  while (1) {
-    int32_t code = setTimeWindowOutputBuf(pResultRowInfo, &nextWin, true, &pResult, tableGroupId, pInfo->binfo.pCtx,
-                                          numOfOutput, pInfo->binfo.rowCellInfoOffset, &pInfo->aggSup, pTaskInfo);
-    if (code != TSDB_CODE_SUCCESS || pResult == NULL) {
-      longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
-    }
-    SResKeyPos* pos = taosMemoryMalloc(sizeof(SResKeyPos) + sizeof(uint64_t));
-    pos->groupId = tableGroupId;
-    pos->pos = (SResultRowPosition){.pageId = pResult->pageId, .offset = pResult->offset};
-    *(int64_t*)pos->key = pResult->win.skey;
-    forwardRows = getNumOfRowsInTimeWindow(&pSDataBlock->info, tsCols, startPos, nextWin.ekey, binarySearchForKey, NULL,
-                                           TSDB_ORDER_ASC);
-    if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE && pUpdated) {
-      saveResultRow(pResult, tableGroupId, pUpdated);
-    }
-    // window start(end) key interpolation
-    // doWindowBorderInterpolation(pInfo, pSDataBlock, numOfOutput, pInfo->binfo.pCtx, pResult, &nextWin, startPos,
-    // forwardRows);
-    updateTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &nextWin, true);
-    doApplyFunctions(pTaskInfo, pInfo->binfo.pCtx, &nextWin, &pInfo->twAggSup.timeWindowData, startPos, forwardRows,
-                     tsCols, pSDataBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
-    int32_t prevEndPos = (forwardRows - 1) * step + startPos;
-    ASSERT(pSDataBlock->info.window.skey > 0 && pSDataBlock->info.window.ekey > 0);
-    startPos = getNextQualifiedWindow(&pInfo->interval, &nextWin, &pSDataBlock->info, tsCols, prevEndPos, pInfo->order);
-    if (startPos < 0) {
-      break;
-    }
-  }
-}
-
 bool isFinalInterval(SStreamFinalIntervalOperatorInfo* pInfo) { return pInfo->pChildren != NULL; }
 
 void compactFunctions(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx, int32_t numOfOutput,
@@ -2130,6 +2081,74 @@ static void rebuildIntervalWindow(SStreamFinalIntervalOperatorInfo* pInfo, SArra
   }
 }
 
+bool isDeletedWindow(STimeWindow* pWin, uint64_t groupId, SAggSupporter* pSup) {
+  SET_RES_WINDOW_KEY(pSup->keyBuf, &pWin->skey, sizeof(int64_t), groupId);
+  SResultRowPosition* p1 = (SResultRowPosition*)taosHashGet(pSup->pResultRowHashTable,
+      pSup->keyBuf, GET_RES_WINDOW_KEY_LEN(sizeof(int64_t)));
+  return p1 == NULL;
+}
+
+static void doHashInterval(SOperatorInfo* pOperatorInfo, SSDataBlock* pSDataBlock, uint64_t tableGroupId,
+                           SArray* pUpdated) {
+  SStreamFinalIntervalOperatorInfo* pInfo = (SStreamFinalIntervalOperatorInfo*)pOperatorInfo->info;
+  SResultRowInfo*                   pResultRowInfo = &(pInfo->binfo.resultRowInfo);
+  SExecTaskInfo*                    pTaskInfo = pOperatorInfo->pTaskInfo;
+  int32_t                           numOfOutput = pOperatorInfo->numOfExprs;
+  int32_t                           step = 1;
+  bool                              ascScan = true;
+  TSKEY*                            tsCols = NULL;
+  SResultRow*                       pResult = NULL;
+  int32_t                           forwardRows = 0;
+
+  if (pSDataBlock->pDataBlock != NULL) {
+    SColumnInfoData* pColDataInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->primaryTsIndex);
+    tsCols = (int64_t*)pColDataInfo->pData;
+  } else {
+    return;
+  }
+
+  int32_t     startPos = ascScan ? 0 : (pSDataBlock->info.rows - 1);
+  TSKEY       ts = getStartTsKey(&pSDataBlock->info.window, tsCols);
+  STimeWindow nextWin = getActiveTimeWindow(pInfo->aggSup.pResultBuf, pResultRowInfo, ts, &pInfo->interval,
+                                            pInfo->interval.precision, NULL);
+  while (1) {
+    if (isFinalInterval(pInfo) && isCloseWindow(&nextWin, &pInfo->twAggSup) &&
+        isDeletedWindow(&nextWin, tableGroupId, &pInfo->aggSup)) {
+      SArray* pUpWins = taosArrayInit(8, sizeof(STimeWindow));
+      taosArrayPush(pUpWins, &nextWin);
+      rebuildIntervalWindow(pInfo, pUpWins, pInfo->binfo.pRes->info.groupId,
+          pOperatorInfo->numOfExprs, pOperatorInfo->pTaskInfo);
+      taosArrayDestroy(pUpWins);
+    }
+    int32_t code = setTimeWindowOutputBuf(pResultRowInfo, &nextWin, true, &pResult, tableGroupId, pInfo->binfo.pCtx,
+                                          numOfOutput, pInfo->binfo.rowCellInfoOffset, &pInfo->aggSup, pTaskInfo);
+    if (code != TSDB_CODE_SUCCESS || pResult == NULL) {
+      longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
+    }
+    SResKeyPos* pos = taosMemoryMalloc(sizeof(SResKeyPos) + sizeof(uint64_t));
+    pos->groupId = tableGroupId;
+    pos->pos = (SResultRowPosition){.pageId = pResult->pageId, .offset = pResult->offset};
+    *(int64_t*)pos->key = pResult->win.skey;
+    forwardRows = getNumOfRowsInTimeWindow(&pSDataBlock->info, tsCols, startPos, nextWin.ekey, binarySearchForKey, NULL,
+                                           TSDB_ORDER_ASC);
+    if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE && pUpdated) {
+      saveResultRow(pResult, tableGroupId, pUpdated);
+    }
+    // window start(end) key interpolation
+    // doWindowBorderInterpolation(pInfo, pSDataBlock, numOfOutput, pInfo->binfo.pCtx, pResult, &nextWin, startPos,
+    // forwardRows);
+    updateTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &nextWin, true);
+    doApplyFunctions(pTaskInfo, pInfo->binfo.pCtx, &nextWin, &pInfo->twAggSup.timeWindowData, startPos, forwardRows,
+                     tsCols, pSDataBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
+    int32_t prevEndPos = (forwardRows - 1) * step + startPos;
+    ASSERT(pSDataBlock->info.window.skey > 0 && pSDataBlock->info.window.ekey > 0);
+    startPos = getNextQualifiedWindow(&pInfo->interval, &nextWin, &pSDataBlock->info, tsCols, prevEndPos, pInfo->order);
+    if (startPos < 0) {
+      break;
+    }
+  }
+}
+
 static void clearStreamIntervalOperator(SStreamFinalIntervalOperatorInfo* pInfo) {
   taosHashClear(pInfo->aggSup.pResultRowHashTable);
   clearDiskbasedBuf(pInfo->aggSup.pResultBuf);
@@ -2169,6 +2188,7 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
   SStreamFinalIntervalOperatorInfo* pInfo = pOperator->info;
   SOperatorInfo*                    downstream = pOperator->pDownstream[0];
   SArray*                           pUpdated = taosArrayInit(4, POINTER_BYTES);
+  TSKEY                             maxTs = INT64_MIN;
 
   if (pOperator->status == OP_EXEC_DONE) {
     return NULL;
@@ -2222,6 +2242,7 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
     }
 
     setInputDataBlock(pOperator, pInfo->binfo.pCtx, pBlock, pInfo->order, MAIN_SCAN, true);
+    doHashInterval(pOperator, pBlock, pBlock->info.groupId, pUpdated);
     if (isFinalInterval(pInfo)) {
       int32_t chIndex = getChildIndex(pBlock);
       int32_t size = taosArrayGetSize(pInfo->pChildren);
@@ -2238,10 +2259,10 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
       setInputDataBlock(pChildOp, pChInfo->binfo.pCtx, pBlock, pChInfo->order, MAIN_SCAN, true);
       doHashInterval(pChildOp, pBlock, pBlock->info.groupId, NULL);
     }
-    doHashInterval(pOperator, pBlock, pBlock->info.groupId, pUpdated);
-    pInfo->twAggSup.maxTs = TMAX(pInfo->twAggSup.maxTs, pBlock->info.window.ekey);
+    maxTs = TMAX(maxTs, pBlock->info.window.ekey);
   }
 
+  pInfo->twAggSup.maxTs = TMAX(pInfo->twAggSup.maxTs, maxTs);
   if (isFinalInterval(pInfo)) {
     closeIntervalWindow(pInfo->aggSup.pResultRowHashTable, &pInfo->twAggSup, &pInfo->interval, pUpdated);
   }
@@ -2564,7 +2585,7 @@ int32_t updateSessionWindowInfo(SResultWindowInfo* pWinInfo, TSKEY* pTs, int32_t
 }
 
 static int32_t setWindowOutputBuf(SResultWindowInfo* pWinInfo, SResultRow** pResult, SqlFunctionCtx* pCtx,
-                                  int32_t groupId, int32_t numOfOutput, int32_t* rowCellInfoOffset,
+                                  uint64_t groupId, int32_t numOfOutput, int32_t* rowCellInfoOffset,
                                   SStreamAggSupporter* pAggSup, SExecTaskInfo* pTaskInfo) {
   assert(pWinInfo->win.skey <= pWinInfo->win.ekey);
   // too many time window in query
@@ -2642,7 +2663,7 @@ int32_t getNumCompactWindow(SArray* pWinInfos, int32_t startIndex, int64_t gap)
   return size - startIndex - 1;
 }
 
-void compactTimeWindow(SStreamSessionAggOperatorInfo* pInfo, int32_t startIndex, int32_t num, int32_t groupId,
+void compactTimeWindow(SStreamSessionAggOperatorInfo* pInfo, int32_t startIndex, int32_t num, uint64_t groupId,
                        int32_t numOfOutput, SExecTaskInfo* pTaskInfo, SHashObj* pStUpdated, SHashObj* pStDeleted) {
   SResultWindowInfo* pCurWin = taosArrayGet(pInfo->streamAggSup.pCurWins, startIndex);
   SResultRow*        pCurResult = NULL;
@@ -2667,13 +2688,18 @@ void compactTimeWindow(SStreamSessionAggOperatorInfo* pInfo, int32_t startIndex,
   }
 }
 
+typedef struct SWinRes {
+  TSKEY    ts;
+  uint64_t groupId;
+} SWinRes;
+
 static void doStreamSessionAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBlock, SHashObj* pStUpdated,
                                    SHashObj* pStDeleted) {
   SExecTaskInfo*                 pTaskInfo = pOperator->pTaskInfo;
   SStreamSessionAggOperatorInfo* pInfo = pOperator->info;
   bool                           masterScan = true;
   int32_t                        numOfOutput = pOperator->numOfExprs;
-  int64_t                        groupId = pSDataBlock->info.groupId;
+  uint64_t                       groupId = pSDataBlock->info.groupId;
   int64_t                        gap = pInfo->gap;
   int64_t                        code = TSDB_CODE_SUCCESS;
 
@@ -2693,7 +2719,7 @@ static void doStreamSessionAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSData
   SStreamAggSupporter* pAggSup = &pInfo->streamAggSup;
   for (int32_t i = 0; i < pSDataBlock->info.rows;) {
     int32_t            winIndex = 0;
-    SResultWindowInfo* pCurWin = getSessionTimeWindow(pAggSup, tsCols[i], pSDataBlock->info.groupId, gap, &winIndex);
+    SResultWindowInfo* pCurWin = getSessionTimeWindow(pAggSup, tsCols[i], groupId, gap, &winIndex);
     winRows = updateSessionWindowInfo(pCurWin, tsCols, pSDataBlock->info.rows, i, pInfo->gap, pStDeleted);
     code = doOneWindowAgg(pInfo, pSDataBlock, pCurWin, &pResult, i, winRows, numOfOutput, pTaskInfo);
     if (code != TSDB_CODE_SUCCESS || pResult == NULL) {
@@ -2709,7 +2735,8 @@ static void doStreamSessionAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSData
     }
     pCurWin->isClosed = false;
     if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE) {
-      code = taosHashPut(pStUpdated, &pCurWin->pos, sizeof(SResultRowPosition), &(pCurWin->win.skey), sizeof(TSKEY));
+      SWinRes value = {.ts = pCurWin->win.skey, .groupId = groupId};
+      code = taosHashPut(pStUpdated, &pCurWin->pos, sizeof(SResultRowPosition), &value, sizeof(SWinRes));
       if (code != TSDB_CODE_SUCCESS) {
         longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
       }
@@ -2736,7 +2763,7 @@ static void doClearSessionWindows(SStreamAggSupporter* pAggSup, SOptrBasicInfo*
   }
 }
 
-static int32_t copyUpdateResult(SHashObj* pStUpdated, SArray* pUpdated, int32_t groupId) {
+static int32_t copyUpdateResult(SHashObj* pStUpdated, SArray* pUpdated) {
   void*  pData = NULL;
   size_t keyLen = 0;
   while ((pData = taosHashIterate(pStUpdated, pData)) != NULL) {
@@ -2746,9 +2773,9 @@ static int32_t copyUpdateResult(SHashObj* pStUpdated, SArray* pUpdated, int32_t
     if (pos == NULL) {
       return TSDB_CODE_QRY_OUT_OF_MEMORY;
     }
-    pos->groupId = groupId;
+    pos->groupId = ((SWinRes*)pData)->groupId;
     pos->pos = *(SResultRowPosition*)key;
-    *(int64_t*)pos->key = *(uint64_t*)pData;
+    *(int64_t*)pos->key = ((SWinRes*)pData)->ts;
     taosArrayPush(pUpdated, &pos);
   }
   return TSDB_CODE_SUCCESS;
@@ -2815,7 +2842,9 @@ int32_t closeSessionWindow(SHashObj* pHashMap, STimeWindowAggSupp* pTwSup, SArra
     __get_win_info_ fn) {
   // Todo(liuyao) save window to tdb
   void **pIte = NULL;
+  size_t keyLen = 0;
   while ((pIte = taosHashIterate(pHashMap, pIte)) != NULL) {
+    uint64_t* pGroupId = taosHashGetKey(pIte, &keyLen);
     SArray *pWins = (SArray *) (*pIte);
     int32_t size = taosArrayGetSize(pWins);
     for (int32_t i = 0; i < size; i++) {
@@ -2825,7 +2854,7 @@ int32_t closeSessionWindow(SHashObj* pHashMap, STimeWindowAggSupp* pTwSup, SArra
         if (!pSeWin->isClosed) {
           pSeWin->isClosed = true;
           if (pTwSup->calTrigger == STREAM_TRIGGER_WINDOW_CLOSE) {
-            int32_t code = saveResult(pSeWin->win.skey, pSeWin->pos.pageId, pSeWin->pos.offset, 0, pClosed);
+            int32_t code = saveResult(pSeWin->win.skey, pSeWin->pos.pageId, pSeWin->pos.offset, *pGroupId, pClosed);
             pSeWin->isOutput = true;
           }
         }
@@ -2892,7 +2921,7 @@ static SSDataBlock* doStreamSessionAgg(SOperatorInfo* pOperator) {
         SStreamSessionAggOperatorInfo* pChildInfo = pChildOp->info;
         doClearSessionWindows(&pChildInfo->streamAggSup, &pChildInfo->binfo, pBlock, 0, pChildOp->numOfExprs,
                               pChildInfo->gap, NULL);
-        rebuildTimeWindow(pInfo, pWins, pInfo->binfo.pRes->info.groupId, pOperator->numOfExprs, pOperator->pTaskInfo);
+        rebuildTimeWindow(pInfo, pWins, pBlock->info.groupId, pOperator->numOfExprs, pOperator->pTaskInfo);
       }
       taosArrayDestroy(pWins);
       continue;
@@ -2916,7 +2945,7 @@ static SSDataBlock* doStreamSessionAgg(SOperatorInfo* pOperator) {
 
   closeSessionWindow(pInfo->streamAggSup.pResultRows, &pInfo->twAggSup, pUpdated,
                      getSessionWinInfo);
-  copyUpdateResult(pStUpdated, pUpdated, pBInfo->pRes->info.groupId);
+  copyUpdateResult(pStUpdated, pUpdated);
   taosHashCleanup(pStUpdated);
 
   finalizeUpdatedResult(pOperator->numOfExprs, pInfo->streamAggSup.pResultBuf, pUpdated,
@@ -3216,8 +3245,9 @@ static void doStreamStateAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBl
     }
     pCurWin->winInfo.isClosed = false;
     if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE) {
-      code = taosHashPut(pSeUpdated, &pCurWin->winInfo.pos, sizeof(SResultRowPosition), &(pCurWin->winInfo.win.skey),
-                         sizeof(TSKEY));
+      SWinRes value = {.ts = pCurWin->winInfo.win.skey, .groupId = groupId};
+      code = taosHashPut(pSeUpdated, &pCurWin->winInfo.pos, sizeof(SResultRowPosition),
+          &value, sizeof(SWinRes));
       if (code != TSDB_CODE_SUCCESS) {
         longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
       }
@@ -3274,7 +3304,7 @@ static SSDataBlock* doStreamStateAgg(SOperatorInfo* pOperator) {
 
   closeSessionWindow(pInfo->streamAggSup.pResultRows, &pInfo->twAggSup, pUpdated,
                      getStateWinInfo);
-  copyUpdateResult(pSeUpdated, pUpdated, pBInfo->pRes->info.groupId);
+  copyUpdateResult(pSeUpdated, pUpdated);
   taosHashCleanup(pSeUpdated);
 
   finalizeUpdatedResult(pOperator->numOfExprs, pInfo->streamAggSup.pResultBuf, pUpdated,
diff --git a/tests/script/tsim/stream/distributeInterval0.sim b/tests/script/tsim/stream/distributeInterval0.sim
index f4f3e04f0a..b720272116 100644
--- a/tests/script/tsim/stream/distributeInterval0.sim
+++ b/tests/script/tsim/stream/distributeInterval0.sim
@@ -173,4 +173,39 @@ endi
 
 sql select  _wstartts, count(*) c1, count(d) c2 , sum(a) c3 , max(b)  c4, min(c) c5, avg(d) from st interval(10s);
 
-system sh/exec.sh -n dnode1 -s stop -x SIGINT
\ No newline at end of file
+
+sql create database test1  vgroups 1;
+sql use test1;
+sql create stable st(ts timestamp, a int, b int , c int) tags(ta int,tb int,tc int);
+sql create table ts1 using st tags(1,1,1);
+sql create table ts2 using st tags(2,2,2);
+sql create stream stream_t2 trigger at_once  into streamtST1 as select  _wstartts, count(*) c1, count(a) c2 , sum(a) c3 ,  max(b)  c5, min(c) c6 from st interval(10s) ;
+
+sql insert into ts1 values(1648791211000,1,2,3);
+sql insert into ts1 values(1648791222001,2,2,3);
+sql insert into ts2 values(1648791211000,1,2,3);
+sql insert into ts2 values(1648791222001,2,2,3);
+
+$loop_count = 0
+loop2:
+sql select * from streamtST1;
+
+sleep 300
+$loop_count = $loop_count + 1
+if $loop_count == 10 then
+  return -1
+endi
+
+# row 0
+if $data01 != 2 then
+  print =====data01=$data01
+  goto loop2
+endi
+
+#rows 1
+if $data11 != 2 then
+  print =====data11=$data11
+  goto loop2
+endi
+
+system sh/stop_dnodes.sh
\ No newline at end of file
diff --git a/tests/script/tsim/stream/partitionby.sim b/tests/script/tsim/stream/partitionby.sim
index df1e096551..b84a01eb4a 100644
--- a/tests/script/tsim/stream/partitionby.sim
+++ b/tests/script/tsim/stream/partitionby.sim
@@ -34,6 +34,7 @@ print =====rows=$rows
 goto loop0
 endi
 
+print =====loop0
 
 sql create database test1  vgroups 1;
 sql use test1;
@@ -51,7 +52,7 @@ sql insert into ts2 values(1648791211000,1,2,3);
 
 $loop_count = 0
 
-loop0:
+loop1:
 sleep 300
 sql select * from streamt;
 
@@ -62,7 +63,62 @@ endi
 
 if $rows != 2 then
 print =====rows=$rows
-goto loop0
+goto loop1
 endi
 
-system sh/exec.sh -n dnode1 -s stop -x SIGINT
\ No newline at end of file
+print =====loop1
+
+sql create database test2  vgroups 1;
+sql use test2;
+sql create stable st(ts timestamp,a int,b int,c int,id int) tags(ta int,tb int,tc int);
+sql create table ts1 using st tags(1,1,1);
+sql create table ts2 using st tags(2,2,2);
+
+sql create stream stream_t2 trigger at_once  into streamtST as select  _wstartts, count(*) c1, count(a) c2 , sum(a) c3 ,  max(b)  c5, min(c) c6, max(id) c7 from st partition by ta interval(10s) ;
+sql insert into ts1 values(1648791211000,1,2,3,1);
+sql insert into ts1 values(1648791222001,2,2,3,2);
+sql insert into ts2 values(1648791211000,1,2,3,3);
+sql insert into ts2 values(1648791222001,2,2,3,4);
+
+sql insert into ts2 values(1648791222002,2,2,3,5);
+sql insert into ts2 values(1648791222002,2,2,3,6);
+
+sql insert into ts1 values(1648791211000,1,2,3,1);
+sql insert into ts1 values(1648791222001,2,2,3,2);
+sql insert into ts2 values(1648791211000,1,2,3,3);
+sql insert into ts2 values(1648791222001,2,2,3,4);
+
+$loop_count = 0
+
+loop2:
+sleep 300
+sql select * from streamtST;
+
+$loop_count = $loop_count + 1
+if $loop_count == 10 then
+  return -1
+endi
+
+if $data01 != 1 then
+print =====data01=$data01
+goto loop2
+endi
+
+if $data02 != 1 then
+print =====data02=$data02
+goto loop2
+endi
+
+if $data03 != 1 then
+print =====data03=$data03
+goto loop2
+endi
+
+if $data04 != 2 then
+print =====data04=$data04
+goto loop2
+endi
+
+print =====loop2
+
+system sh/stop_dnodes.sh
\ No newline at end of file

From ab47de1dae88097fc546a4e57f5dbe81a4e11b80 Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 18:30:22 +0800
Subject: [PATCH 32/40] fix(query): adjust group limitation.

---
 source/libs/executor/src/executorimpl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c
index fb79c53384..66849e49b8 100644
--- a/source/libs/executor/src/executorimpl.c
+++ b/source/libs/executor/src/executorimpl.c
@@ -3508,7 +3508,7 @@ static int32_t handleLimitOffset(SOperatorInfo* pOperator, SSDataBlock* pBlock)
   if (pProjectInfo->limit.limit > 0 && pProjectInfo->curOutput + pRes->info.rows >= pProjectInfo->limit.limit) {
     pRes->info.rows = (int32_t)(pProjectInfo->limit.limit - pProjectInfo->curOutput);
 
-    if (pProjectInfo->slimit.limit == -1 || pProjectInfo->slimit.limit <= pProjectInfo->curGroupOutput) {
+    if (pProjectInfo->slimit.limit != -1 && pProjectInfo->slimit.limit <= pProjectInfo->curGroupOutput) {
       pOperator->status = OP_EXEC_DONE;
     }
 

From c1de4df95bde0ec91dd59203d6403042fad52f20 Mon Sep 17 00:00:00 2001
From: Minglei Jin <mljin@taosdata.com>
Date: Thu, 16 Jun 2022 18:33:45 +0800
Subject: [PATCH 33/40] fix: fix page capacity calculation

---
 source/libs/tdb/src/db/tdbPage.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/source/libs/tdb/src/db/tdbPage.c b/source/libs/tdb/src/db/tdbPage.c
index 92272fb438..78470b6256 100644
--- a/source/libs/tdb/src/db/tdbPage.c
+++ b/source/libs/tdb/src/db/tdbPage.c
@@ -246,14 +246,17 @@ void tdbPageCopy(SPage *pFromPage, SPage *pToPage) {
 
 int tdbPageCapacity(int pageSize, int amHdrSize) {
   int szPageHdr;
+  int minCellIndexSize; // at least one cell in cell index
 
   if (pageSize < 65536) {
     szPageHdr = pageMethods.szPageHdr;
+    minCellIndexSize = pageMethods.szOffset;
   } else {
     szPageHdr = pageLargeMethods.szPageHdr;
+    minCellIndexSize = pageLargeMethods.szOffset;
   }
 
-  return pageSize - szPageHdr - amHdrSize;
+  return pageSize - szPageHdr - amHdrSize - sizeof(SPageFtr) - minCellIndexSize;
 }
 
 static int tdbPageAllocate(SPage *pPage, int szCell, SCell **ppCell) {
@@ -599,4 +602,4 @@ SPageMethods pageLargeMethods = {
     setLPageCellOffset,    // setCellOffset
     getLPageFreeCellInfo,  // getFreeCellInfo
     setLPageFreeCellInfo   // setFreeCellInfo
-};
\ No newline at end of file
+};

From 67540f2504464927ff1f684f97481138bfc19baf Mon Sep 17 00:00:00 2001
From: shenglian zhou <slzhou@taosdata.com>
Date: Thu, 16 Jun 2022 19:49:42 +0800
Subject: [PATCH 34/40] fix: table merge scan sort buf size

---
 source/libs/executor/src/scanoperator.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/source/libs/executor/src/scanoperator.c b/source/libs/executor/src/scanoperator.c
index ddebba1c9e..a8aad8fe68 100644
--- a/source/libs/executor/src/scanoperator.c
+++ b/source/libs/executor/src/scanoperator.c
@@ -2310,14 +2310,14 @@ SOperatorInfo* createTableMergeScanOperatorInfo(STableScanPhysiNode* pTableScanN
   pInfo->pSortInfo = generateSortByTsInfo(pInfo->cond.order);
   pInfo->pSortInputBlock = createOneDataBlock(pInfo->pResBlock, false);
   int32_t rowSize = pInfo->pResBlock->info.rowSize;
-  pInfo->bufPageSize = rowSize < 1024 ? 1024 : rowSize * 2;
+  int32_t blockMetaSize = (int32_t)blockDataGetSerialMetaSize(pInfo->pResBlock->info.numOfCols);
+  pInfo->bufPageSize = (rowSize * 2 + blockMetaSize) < 1024 ? 1024 : (rowSize * 2 + blockMetaSize);
   pInfo->sortBufSize = pInfo->bufPageSize * 16;
   pInfo->hasGroupId = false;
   pInfo->prefetchedTuple = NULL;
 
   pOperator->name = "TableMergeScanOperator";
-  // TODO : change it
-  pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN;
+  pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TABLE_MERGE_SCAN;
   pOperator->blocking = false;
   pOperator->status = OP_NOT_OPENED;
   pOperator->info = pInfo;

From 98cbbc01975f78132c0ac90c0d0c1be6c73ef507 Mon Sep 17 00:00:00 2001
From: afwerar <1296468573@qq.com>
Date: Thu, 16 Jun 2022 19:53:39 +0800
Subject: [PATCH 35/40] test: add auto exit case

---
 tests/system-test/test.py | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/tests/system-test/test.py b/tests/system-test/test.py
index 5a68f548da..f3e39fc9ce 100644
--- a/tests/system-test/test.py
+++ b/tests/system-test/test.py
@@ -13,6 +13,7 @@
 # pip install src/connector/python/
 
 # -*- coding: utf-8 -*-
+from ast import If
 import sys
 import getopt
 import subprocess
@@ -33,8 +34,15 @@ import taos
 
 def checkRunTimeError():
     import win32gui
+    timeCount = 0
     while 1:
         time.sleep(1)
+        timeCount = timeCount + 1
+        if (timeCount>900):
+            os.system("TASKKILL /F /IM taosd.exe")
+            os.system("TASKKILL /F /IM taos.exe")
+            os.system("TASKKILL /F /IM tmq_sim.exe")
+            quit(0)
         hwnd = win32gui.FindWindow(None, "Microsoft Visual C++ Runtime Library")
         if hwnd:
             os.system("TASKKILL /F /IM taosd.exe")

From c858b6952853677f3e2a25d5532b63a5114f7df6 Mon Sep 17 00:00:00 2001
From: afwerar <1296468573@qq.com>
Date: Thu, 16 Jun 2022 19:56:51 +0800
Subject: [PATCH 36/40] test: add auto exit case

---
 tests/system-test/test.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/tests/system-test/test.py b/tests/system-test/test.py
index f3e39fc9ce..85f63a2868 100644
--- a/tests/system-test/test.py
+++ b/tests/system-test/test.py
@@ -13,7 +13,6 @@
 # pip install src/connector/python/
 
 # -*- coding: utf-8 -*-
-from ast import If
 import sys
 import getopt
 import subprocess

From 49d92079ca1005546eca0b87daba8fbdaee3e509 Mon Sep 17 00:00:00 2001
From: afwerar <1296468573@qq.com>
Date: Thu, 16 Jun 2022 19:58:24 +0800
Subject: [PATCH 37/40] test: add auto exit case

---
 tests/system-test/test.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tests/system-test/test.py b/tests/system-test/test.py
index 85f63a2868..55bfd8c945 100644
--- a/tests/system-test/test.py
+++ b/tests/system-test/test.py
@@ -41,6 +41,7 @@ def checkRunTimeError():
             os.system("TASKKILL /F /IM taosd.exe")
             os.system("TASKKILL /F /IM taos.exe")
             os.system("TASKKILL /F /IM tmq_sim.exe")
+            os.system("TASKKILL /F /IM mintty.exe")
             quit(0)
         hwnd = win32gui.FindWindow(None, "Microsoft Visual C++ Runtime Library")
         if hwnd:

From bdf24b6ab06c991db5198835d18fff3720f793f1 Mon Sep 17 00:00:00 2001
From: Cary Xu <klxu@taosdata.com>
Date: Thu, 16 Jun 2022 20:14:48 +0800
Subject: [PATCH 38/40] fix: uncomment the codes of dropping stable

---
 source/dnode/vnode/src/vnd/vnodeSvr.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/source/dnode/vnode/src/vnd/vnodeSvr.c b/source/dnode/vnode/src/vnd/vnodeSvr.c
index d170931e7e..da1a126a76 100644
--- a/source/dnode/vnode/src/vnd/vnodeSvr.c
+++ b/source/dnode/vnode/src/vnd/vnodeSvr.c
@@ -466,10 +466,10 @@ static int32_t vnodeProcessDropStbReq(SVnode *pVnode, int64_t version, void *pRe
   }
 
   // process request
-  // if (metaDropSTable(pVnode->pMeta, version, &req) < 0) {
-  //   rcode = terrno;
-  //   goto _exit;
-  // }
+  if (metaDropSTable(pVnode->pMeta, version, &req) < 0) {
+    rcode = terrno;
+    goto _exit;
+  }
 
   // return rsp
 _exit:

From 8353da6694408895df7f277aa69920968dc560ce Mon Sep 17 00:00:00 2001
From: Cary Xu <klxu@taosdata.com>
Date: Thu, 16 Jun 2022 20:17:36 +0800
Subject: [PATCH 39/40] other: revert the code

---
 source/dnode/vnode/src/tsdb/tsdbReadImpl.c | 11 -----------
 1 file changed, 11 deletions(-)

diff --git a/source/dnode/vnode/src/tsdb/tsdbReadImpl.c b/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
index e4c5ef8fc3..1c2514d46f 100644
--- a/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
+++ b/source/dnode/vnode/src/tsdb/tsdbReadImpl.c
@@ -233,17 +233,6 @@ int tsdbLoadBlockInfo(SReadH *pReadh, void *pTarget) {
     tsdbError("vgId:%d, SBlockInfo part in file %s is corrupted since wrong checksum, offset:%u len :%u",
               TSDB_READ_REPO_ID(pReadh), TSDB_FILE_FULL_NAME(pHeadf), pBlkIdx->offset, pBlkIdx->len);
     return -1;
-  } else {
-    SBlockInfo *pBlkInfo = pReadh->pBlkInfo;
-    SBlock *pBlk = NULL;
-    int         nBlks = (pBlkIdx->len - sizeof(SBlockInfo)) / sizeof(SBlock);
-    for (int n = 0; n < nBlks; ++n) {
-      pBlk = &pBlkInfo->blocks[n];
-      if (pBlk->numOfSubBlocks >= 1) {
-        tsdbDebug("prop:vgId:%d, file %s , offset:%u len :%u has %d rows", TSDB_READ_REPO_ID(pReadh),
-                  TSDB_FILE_FULL_NAME(pHeadf), pBlkIdx->offset, pBlkIdx->len, pBlk->numOfRows);
-      }
-    }
   }
 
   // ASSERT(pBlkIdx->tid == pReadh->pBlkInfo->tid && pBlkIdx->uid == pReadh->pBlkInfo->uid);

From cc638a2236ad0b802443e198c467dc41fff30d05 Mon Sep 17 00:00:00 2001
From: Haojun Liao <hjliao@taosdata.com>
Date: Thu, 16 Jun 2022 21:13:18 +0800
Subject: [PATCH 40/40] fix(query): set null for null value in groupby

---
 source/libs/executor/src/executorimpl.c  |  3 +--
 source/libs/executor/src/groupoperator.c |  1 -
 source/libs/function/src/builtinsimpl.c  | 19 +++++++++++++++----
 3 files changed, 16 insertions(+), 7 deletions(-)

diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c
index 66849e49b8..40b019eb5d 100644
--- a/source/libs/executor/src/executorimpl.c
+++ b/source/libs/executor/src/executorimpl.c
@@ -3507,8 +3507,7 @@ static int32_t handleLimitOffset(SOperatorInfo* pOperator, SSDataBlock* pBlock)
   // check for the limitation in each group
   if (pProjectInfo->limit.limit > 0 && pProjectInfo->curOutput + pRes->info.rows >= pProjectInfo->limit.limit) {
     pRes->info.rows = (int32_t)(pProjectInfo->limit.limit - pProjectInfo->curOutput);
-
-    if (pProjectInfo->slimit.limit != -1 && pProjectInfo->slimit.limit <= pProjectInfo->curGroupOutput) {
+    if (pProjectInfo->slimit.limit > 0 && pProjectInfo->slimit.limit <= pProjectInfo->curGroupOutput) {
       pOperator->status = OP_EXEC_DONE;
     }
 
diff --git a/source/libs/executor/src/groupoperator.c b/source/libs/executor/src/groupoperator.c
index 23226a0134..4fc25688c4 100644
--- a/source/libs/executor/src/groupoperator.c
+++ b/source/libs/executor/src/groupoperator.c
@@ -331,7 +331,6 @@ static SSDataBlock* hashGroupbyAggregate(SOperatorInfo* pOperator) {
   }
 
   pOperator->status = OP_RES_TO_RETURN;
-  closeAllResultRows(&pInfo->binfo.resultRowInfo);
 
 #if 0
   if(pOperator->fpSet.encodeResultRow){
diff --git a/source/libs/function/src/builtinsimpl.c b/source/libs/function/src/builtinsimpl.c
index 90700580bd..55900c7c7e 100644
--- a/source/libs/function/src/builtinsimpl.c
+++ b/source/libs/function/src/builtinsimpl.c
@@ -1384,8 +1384,9 @@ int32_t maxFunction(SqlFunctionCtx* pCtx) {
   return TSDB_CODE_SUCCESS;
 }
 
-static void setSelectivityValue(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, const STuplePos* pTuplePos,
-                                int32_t rowIndex);
+static void setNullSelectivityValue(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, int32_t rowIndex);
+
+static void setSelectivityValue(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, const STuplePos* pTuplePos, int32_t rIndex);
 
 int32_t minmaxFunctionFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
@@ -1407,11 +1408,23 @@ int32_t minmaxFunctionFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
 
   if (pEntryInfo->numOfRes > 0) {
     setSelectivityValue(pCtx, pBlock, &pRes->tuplePos, currentRow);
+  } else {
+    setNullSelectivityValue(pCtx, pBlock, currentRow);
   }
 
   return pEntryInfo->numOfRes;
 }
 
+void setNullSelectivityValue(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, int32_t rowIndex) {
+  for (int32_t j = 0; j < pCtx->subsidiaries.num; ++j) {
+    SqlFunctionCtx* pc = pCtx->subsidiaries.pCtx[j];
+    int32_t      dstSlotId = pc->pExpr->base.resSchema.slotId;
+
+    SColumnInfoData* pDstCol = taosArrayGet(pBlock->pDataBlock, dstSlotId);
+    colDataAppendNULL(pDstCol, rowIndex);
+  }
+}
+
 void setSelectivityValue(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, const STuplePos* pTuplePos, int32_t rowIndex) {
   int32_t pageId = pTuplePos->pageId;
   int32_t offset = pTuplePos->offset;
@@ -4627,8 +4640,6 @@ int32_t tailFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   for (int32_t i = 0; i < pEntryInfo->numOfRes; ++i) {
     STailItem* pItem = pInfo->pItems[i];
     colDataAppend(pCol, currentRow, pItem->data, false);
-
-    // setSelectivityValue(pCtx, pBlock, &pInfo->pItems[i].tuplePos, currentRow);
     currentRow += 1;
   }