Merge pull request #26669 from taosdata/fix/syntax

refactor: check return value for each function.

include/common/tdatablock.h

@@ -253,7 +253,7 @@ int32_t assignOneDataBlock(SSDataBlock* dst, const SSDataBlock* src);
 int32_t copyDataBlock(SSDataBlock* pDst, const SSDataBlock* pSrc);
 
 SSDataBlock* createDataBlock();
-void*        blockDataDestroy(SSDataBlock* pBlock);
+void         blockDataDestroy(SSDataBlock* pBlock);
 void         blockDataFreeRes(SSDataBlock* pBlock);
 SSDataBlock* createOneDataBlock(const SSDataBlock* pDataBlock, bool copyData);
 SSDataBlock* createSpecialDataBlock(EStreamType type);

include/libs/stream/tstream.h

@@ -629,7 +629,7 @@ int32_t streamSetupScheduleTrigger(SStreamTask* pTask);
 int32_t streamProcessDispatchMsg(SStreamTask* pTask, SStreamDispatchReq* pReq, SRpcMsg* pMsg);
 int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, int32_t code);
 
-int32_t streamTaskGetUpstreamTaskEpInfo(SStreamTask* pTask, int32_t taskId, SStreamUpstreamEpInfo** pEpInfo);
+void streamTaskGetUpstreamTaskEpInfo(SStreamTask* pTask, int32_t taskId, SStreamUpstreamEpInfo** pEpInfo);
 #if 0
 SEpSet* streamTaskGetDownstreamEpInfo(SStreamTask* pTask, int32_t taskId);
 #endif
@@ -659,10 +659,10 @@ int8_t  streamTaskSetSchedStatusInactive(SStreamTask* pTask);
 int32_t streamTaskClearHTaskAttr(SStreamTask* pTask, int32_t clearRelHalt);
 
 int32_t streamExecTask(SStreamTask* pTask);
-int32_t streamResumeTask(SStreamTask* pTask);
+void    streamResumeTask(SStreamTask* pTask);
 int32_t streamTrySchedExec(SStreamTask* pTask);
 int32_t streamTaskSchedTask(SMsgCb* pMsgCb, int32_t vgId, int64_t streamId, int32_t taskId, int32_t execType);
-int32_t streamTaskResumeInFuture(SStreamTask* pTask);
+void    streamTaskResumeInFuture(SStreamTask* pTask);
 void    streamTaskClearSchedIdleInfo(SStreamTask* pTask);
 void    streamTaskSetIdleInfo(SStreamTask* pTask, int32_t idleTime);
 
@@ -675,8 +675,8 @@ int32_t streamTaskSendCheckRsp(const SStreamMeta* pMeta, int32_t vgId, SStreamTa
 int32_t streamTaskProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRsp);
 
 // check downstream status
-int32_t streamTaskStartMonitorCheckRsp(SStreamTask* pTask);
+void    streamTaskStartMonitorCheckRsp(SStreamTask* pTask);
-int32_t streamTaskStopMonitorCheckRsp(STaskCheckInfo* pInfo, const char* id);
+void    streamTaskStopMonitorCheckRsp(STaskCheckInfo* pInfo, const char* id);
 void    streamTaskCleanupCheckInfo(STaskCheckInfo* pInfo);
 
 // fill-history task

source/common/src/tdatablock.c

@@ -1523,9 +1523,9 @@ void blockDataFreeRes(SSDataBlock* pBlock) {
   memset(&pBlock->info, 0, sizeof(SDataBlockInfo));
 }
 
-void* blockDataDestroy(SSDataBlock* pBlock) {
+void blockDataDestroy(SSDataBlock* pBlock) {
   if (pBlock == NULL) {
-    return NULL;
+    return;
   }
 
   if (IS_VAR_DATA_TYPE(pBlock->info.pks[0].type)) {
@@ -1535,7 +1535,6 @@ void* blockDataDestroy(SSDataBlock* pBlock) {
 
   blockDataFreeRes(pBlock);
   taosMemoryFreeClear(pBlock);
-  return NULL;
 }
 
 // todo remove it
@@ -2468,19 +2467,18 @@ char* buildCtbNameByGroupId(const char* stbFullName, uint64_t groupId) {
 
 int32_t buildCtbNameByGroupIdImpl(const char* stbFullName, uint64_t groupId, char* cname) {
   if (stbFullName[0] == 0) {
-    terrno = TSDB_CODE_INVALID_PARA;
+    return TSDB_CODE_INVALID_PARA;
-    return TSDB_CODE_FAILED;
   }
 
   SArray* tags = taosArrayInit(0, sizeof(SSmlKv));
   if (tags == NULL) {
-    return TSDB_CODE_FAILED;
+    return TSDB_CODE_OUT_OF_MEMORY;
   }
 
   if (cname == NULL) {
     terrno = TSDB_CODE_INVALID_PARA;
     taosArrayDestroy(tags);
-    return TSDB_CODE_FAILED;
+    return terrno;
   }
 
   int8_t      type = TSDB_DATA_TYPE_UBIGINT;

source/dnode/vnode/src/tsdb/tsdbRead2.c

@@ -4648,7 +4648,8 @@ void tsdbReaderClose2(STsdbReader* pReader) {
   }
 
   if (pReader->resBlockInfo.freeBlock) {
-    pReader->resBlockInfo.pResBlock = blockDataDestroy(pReader->resBlockInfo.pResBlock);
+    blockDataDestroy(pReader->resBlockInfo.pResBlock);
+    pReader->resBlockInfo.pResBlock = NULL;
   }
 
   taosMemoryFree(pSupInfo->colId);

source/libs/executor/src/exchangeoperator.c

@@ -607,7 +607,7 @@ int32_t extractDataBlockFromFetchRsp(SSDataBlock* pRes, char* pData, SArray* pCo
       blockDataAppendColInfo(pBlock, &idata);
     }
 
-    blockDecode(pBlock, pStart);
+    (void) blockDecode(pBlock, pStart);
     blockDataEnsureCapacity(pRes, pBlock->info.rows);
 
     // data from mnode

source/libs/executor/src/executorInt.c

@@ -970,7 +970,7 @@ void cleanupExprSupp(SExprSupp* pSupp) {
   taosMemoryFree(pSupp->rowEntryInfoOffset);
 }
 
-void cleanupBasicInfo(SOptrBasicInfo* pInfo) { pInfo->pRes = blockDataDestroy(pInfo->pRes); }
+void cleanupBasicInfo(SOptrBasicInfo* pInfo) { blockDataDestroy(pInfo->pRes); pInfo->pRes = NULL;}
 
 bool groupbyTbname(SNodeList* pGroupList) {
   bool bytbname = false;

source/libs/executor/src/filloperator.c

@@ -310,8 +310,10 @@ static SSDataBlock* doFill(SOperatorInfo* pOperator) {
 void destroyFillOperatorInfo(void* param) {
   SFillOperatorInfo* pInfo = (SFillOperatorInfo*)param;
   pInfo->pFillInfo = taosDestroyFillInfo(pInfo->pFillInfo);
-  pInfo->pRes = blockDataDestroy(pInfo->pRes);
+  blockDataDestroy(pInfo->pRes);
-  pInfo->pFinalRes = blockDataDestroy(pInfo->pFinalRes);
+  pInfo->pRes = NULL;
+  blockDataDestroy(pInfo->pFinalRes);
+  pInfo->pFinalRes = NULL;
 
   cleanupExprSupp(&pInfo->noFillExprSupp);
 

source/libs/executor/src/hashjoinoperator.c

@@ -1065,7 +1065,8 @@ static void destroyHashJoinOperator(void* param) {
 
   hJoinFreeTableInfo(&pJoinOperator->tbs[0]);
   hJoinFreeTableInfo(&pJoinOperator->tbs[1]);
-  pJoinOperator->finBlk = blockDataDestroy(pJoinOperator->finBlk);
+  blockDataDestroy(pJoinOperator->finBlk);
+  pJoinOperator->finBlk = NULL;
   taosMemoryFreeClear(pJoinOperator->pResColMap);
   taosArrayDestroyEx(pJoinOperator->pRowBufs, hJoinFreeBufPage);
 

source/libs/executor/src/mergejoin.c

@@ -3305,8 +3305,10 @@ void mJoinDestroyWindowCtx(SMJoinOperatorInfo* pJoin) {
 
   mWinJoinResetWindowCache(pCtx, &pCtx->cache);
 
-  pCtx->finBlk = blockDataDestroy(pCtx->finBlk);
+  blockDataDestroy(pCtx->finBlk);
-  pCtx->cache.outBlk = blockDataDestroy(pCtx->cache.outBlk);
+  pCtx->finBlk = NULL;
+  blockDataDestroy(pCtx->cache.outBlk);
+  pCtx->cache.outBlk = NULL;
 
   taosArrayDestroy(pCtx->cache.grps);
 }
@@ -3378,9 +3380,11 @@ int32_t mJoinInitWindowCtx(SMJoinOperatorInfo* pJoin, SSortMergeJoinPhysiNode* p
 
 void mJoinDestroyMergeCtx(SMJoinOperatorInfo* pJoin) {
   SMJoinMergeCtx* pCtx = &pJoin->ctx.mergeCtx;
+  blockDataDestroy(pCtx->finBlk);
+  blockDataDestroy(pCtx->midBlk);
 
-  pCtx->finBlk = blockDataDestroy(pCtx->finBlk);
+  pCtx->finBlk = NULL;
-  pCtx->midBlk = blockDataDestroy(pCtx->midBlk);
+  pCtx->midBlk = NULL;
 }
 
 

source/libs/executor/src/mergeoperator.c

@@ -232,8 +232,11 @@ int32_t getSortMergeExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, u
 
 void destroySortMergeOperatorInfo(void* param) {
   SSortMergeInfo* pSortMergeInfo = param;
-  pSortMergeInfo->pInputBlock = blockDataDestroy(pSortMergeInfo->pInputBlock);
+  blockDataDestroy(pSortMergeInfo->pInputBlock);
-  pSortMergeInfo->pIntermediateBlock = blockDataDestroy(pSortMergeInfo->pIntermediateBlock);
+  pSortMergeInfo->pInputBlock = NULL;
+
+  blockDataDestroy(pSortMergeInfo->pIntermediateBlock);
+  pSortMergeInfo->pIntermediateBlock = NULL;
 
   taosArrayDestroy(pSortMergeInfo->matchInfo.pList);
 
@@ -429,7 +432,8 @@ SSDataBlock* doMultiwayMerge(SOperatorInfo* pOperator) {
 
 void destroyMultiwayMergeOperatorInfo(void* param) {
   SMultiwayMergeOperatorInfo* pInfo = (SMultiwayMergeOperatorInfo*)param;
-  pInfo->binfo.pRes = blockDataDestroy(pInfo->binfo.pRes);
+  blockDataDestroy(pInfo->binfo.pRes);
+  pInfo->binfo.pRes = NULL;
 
   if (NULL != gMultiwayMergeFps[pInfo->type].closeFn) {
     (*gMultiwayMergeFps[pInfo->type].closeFn)(&pInfo->sortMergeInfo);

source/libs/executor/src/scanoperator.c

@@ -3789,7 +3789,8 @@ static void destroyTagScanOperatorInfo(void* param) {
   taosArrayDestroy(pInfo->aFilterIdxs);
   taosArrayDestroyEx(pInfo->aUidTags, tagScanFreeUidTag);
 
-  pInfo->pRes = blockDataDestroy(pInfo->pRes);
+  blockDataDestroy(pInfo->pRes);
+  pInfo->pRes = NULL;
   taosArrayDestroy(pInfo->matchInfo.pList);
   pInfo->pTableListInfo = tableListDestroy(pInfo->pTableListInfo);
   taosMemoryFreeClear(param);
@@ -4742,15 +4743,19 @@ void destroyTableMergeScanOperatorInfo(void* param) {
   pTableScanInfo->pSortHandle = NULL;
   taosHashCleanup(pTableScanInfo->mSkipTables);
   pTableScanInfo->mSkipTables = NULL;
-  pTableScanInfo->pSortInputBlock = blockDataDestroy(pTableScanInfo->pSortInputBlock);
+  blockDataDestroy(pTableScanInfo->pSortInputBlock);
+  pTableScanInfo->pSortInputBlock = NULL;
   // end one reader variable
 
   cleanupQueryTableDataCond(&pTableScanInfo->base.cond);
   destroyTableScanBase(&pTableScanInfo->base, &pTableScanInfo->base.readerAPI);
 
-  pTableScanInfo->pResBlock = blockDataDestroy(pTableScanInfo->pResBlock);
+  blockDataDestroy(pTableScanInfo->pResBlock);
+  pTableScanInfo->pResBlock = NULL;
+
   // remove it from the task->result list
-  pTableScanInfo->pReaderBlock = blockDataDestroy(pTableScanInfo->pReaderBlock);
+  blockDataDestroy(pTableScanInfo->pReaderBlock);
+  pTableScanInfo->pReaderBlock = NULL;
   taosArrayDestroy(pTableScanInfo->pSortInfo);
 
   stopSubTablesTableMergeScan(pTableScanInfo);

source/libs/executor/src/sortoperator.c

@@ -368,7 +368,8 @@ SSDataBlock* doSort(SOperatorInfo* pOperator) {
 
 void destroySortOperatorInfo(void* param) {
   SSortOperatorInfo* pInfo = (SSortOperatorInfo*)param;
-  pInfo->binfo.pRes = blockDataDestroy(pInfo->binfo.pRes);
+  blockDataDestroy(pInfo->binfo.pRes);
+  pInfo->binfo.pRes = NULL;
 
   tsortDestroySortHandle(pInfo->pSortHandle);
   taosArrayDestroy(pInfo->pSortInfo);
@@ -611,7 +612,8 @@ int32_t getGroupSortExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, u
 
 void destroyGroupSortOperatorInfo(void* param) {
   SGroupSortOperatorInfo* pInfo = (SGroupSortOperatorInfo*)param;
-  pInfo->binfo.pRes = blockDataDestroy(pInfo->binfo.pRes);
+  blockDataDestroy(pInfo->binfo.pRes);
+  pInfo->binfo.pRes = NULL;
 
   taosArrayDestroy(pInfo->pSortInfo);
   taosArrayDestroy(pInfo->matchInfo.pList);

source/libs/executor/src/streamfilloperator.c

@@ -126,9 +126,12 @@ static void destroyStreamFillOperatorInfo(void* param) {
   SStreamFillOperatorInfo* pInfo = (SStreamFillOperatorInfo*)param;
   pInfo->pFillInfo = destroyStreamFillInfo(pInfo->pFillInfo);
   pInfo->pFillSup = destroyStreamFillSupporter(pInfo->pFillSup);
-  pInfo->pRes = blockDataDestroy(pInfo->pRes);
+  blockDataDestroy(pInfo->pRes);
-  pInfo->pSrcBlock = blockDataDestroy(pInfo->pSrcBlock);
+  pInfo->pRes = NULL;
-  pInfo->pDelRes = blockDataDestroy(pInfo->pDelRes);
+  blockDataDestroy(pInfo->pSrcBlock);
+  pInfo->pSrcBlock = NULL;
+  blockDataDestroy(pInfo->pDelRes);
+  pInfo->pDelRes = NULL;
   taosArrayDestroy(pInfo->matchInfo.pList);
   pInfo->matchInfo.pList = NULL;
   taosMemoryFree(pInfo);

source/libs/executor/src/timesliceoperator.c

@@ -1087,7 +1087,8 @@ _error:
 void destroyTimeSliceOperatorInfo(void* param) {
   STimeSliceOperatorInfo* pInfo = (STimeSliceOperatorInfo*)param;
 
-  pInfo->pRes = blockDataDestroy(pInfo->pRes);
+  blockDataDestroy(pInfo->pRes);
+  pInfo->pRes = NULL;
 
   for (int32_t i = 0; i < taosArrayGetSize(pInfo->pPrevRow); ++i) {
     SGroupKeys* pKey = taosArrayGet(pInfo->pPrevRow, i);

source/libs/executor/src/tsort.c

@@ -555,7 +555,8 @@ static int32_t adjustMergeTreeForNextTuple(SSortSource* pSource, SMultiwayMergeT
         (*numOfCompleted) += 1;
         pSource->src.rowIndex = -1;
         pSource->pageIndex = -1;
-        pSource->src.pBlock = blockDataDestroy(pSource->src.pBlock);
+        blockDataDestroy(pSource->src.pBlock);
+        pSource->src.pBlock = NULL;
       } else {
         if (pSource->pageIndex % 512 == 0) {
           qDebug("begin source %p page %d", pSource, pSource->pageIndex);

source/libs/stream/inc/streamInt.h

@@ -208,7 +208,7 @@ int32_t streamQueueOpen(int64_t cap, SStreamQueue** pQ);
 void    streamQueueClose(SStreamQueue* pQueue, int32_t taskId);
 void    streamQueueProcessSuccess(SStreamQueue* queue);
 void    streamQueueProcessFail(SStreamQueue* queue);
-void*   streamQueueNextItem(SStreamQueue* pQueue);
+void    streamQueueNextItem(SStreamQueue* pQueue, SStreamQueueItem** pItem);
 void    streamFreeQitem(SStreamQueueItem* data);
 int32_t streamQueueGetItemSize(const SStreamQueue* pQueue);
 
@@ -226,7 +226,7 @@ int32_t streamTaskDownloadCheckpointData(const char* id, char* path);
 int32_t streamTaskOnNormalTaskReady(SStreamTask* pTask);
 int32_t streamTaskOnScanHistoryTaskReady(SStreamTask* pTask);
 
-int32_t initCheckpointReadyInfo(STaskCheckpointReadyInfo* pReadyInfo, int32_t upstreamNodeId, int32_t upstreamTaskId,
+void    initCheckpointReadyInfo(STaskCheckpointReadyInfo* pReadyInfo, int32_t upstreamNodeId, int32_t upstreamTaskId,
                                 int32_t childId, SEpSet* pEpset, int64_t checkpointId);
 int32_t initCheckpointReadyMsg(SStreamTask* pTask, int32_t upstreamNodeId, int32_t upstreamTaskId, int32_t childId,
                                int64_t checkpointId, SRpcMsg* pMsg);
@@ -236,6 +236,10 @@ typedef int32_t (*__stream_async_exec_fn_t)(void* param);
 int32_t streamMetaAsyncExec(SStreamMeta* pMeta, __stream_async_exec_fn_t fn, void* param, int32_t* code);
 void    flushStateDataInExecutor(SStreamTask* pTask, SStreamQueueItem* pCheckpointBlock);
 
+void streamMutexLock(TdThreadMutex *pMutex);
+void streamMutexUnlock(TdThreadMutex *pMutex);
+void streamMutexDestroy(TdThreadMutex *pMutex);
+
 #ifdef __cplusplus
 }
 #endif

source/libs/stream/src/streamBackendRocksdb.c

@@ -887,8 +887,8 @@ _EXIT:
   rocksdb_options_destroy(opts);
   rocksdb_cache_destroy(cache);
   rocksdb_env_destroy(env);
-  taosThreadMutexDestroy(&pHandle->mutex);
+  streamMutexDestroy(&pHandle->mutex);
-  taosThreadMutexDestroy(&pHandle->cfMutex);
+  streamMutexDestroy(&pHandle->cfMutex);
   taosHashCleanup(pHandle->cfInst);
   tdListFree(pHandle->list);
   taosMemoryFree(pHandle);
@@ -923,9 +923,9 @@ void streamBackendCleanup(void* arg) {
   }
 
   tdListFree(pHandle->list);
-  taosThreadMutexDestroy(&pHandle->mutex);
+  streamMutexDestroy(&pHandle->mutex);
 
-  taosThreadMutexDestroy(&pHandle->cfMutex);
+  streamMutexDestroy(&pHandle->cfMutex);
   stDebug("destroy stream backend :%p", pHandle);
   taosMemoryFree(pHandle);
   return;
@@ -1393,7 +1393,7 @@ int32_t taskDbBuildSnap(void* arg, SArray* pSnap) {
   // vnode task->db
   SStreamMeta* pMeta = arg;
 
-  taosThreadMutexLock(&pMeta->backendMutex);
+  streamMutexLock(&pMeta->backendMutex);
   void*   pIter = taosHashIterate(pMeta->pTaskDbUnique, NULL);
   int32_t code = 0;
 
@@ -1434,14 +1434,14 @@ int32_t taskDbBuildSnap(void* arg, SArray* pSnap) {
 
     pIter = taosHashIterate(pMeta->pTaskDbUnique, pIter);
   }
-  taosThreadMutexUnlock(&pMeta->backendMutex);
+  streamMutexUnlock(&pMeta->backendMutex);
   return code;
 }
 int32_t taskDbDestroySnap(void* arg, SArray* pSnapInfo) {
   if (pSnapInfo == NULL) return 0;
   SStreamMeta* pMeta = arg;
   int32_t      code = 0;
-  taosThreadMutexLock(&pMeta->backendMutex);
+  streamMutexLock(&pMeta->backendMutex);
 
   char buf[128] = {0};
   for (int i = 0; i < taosArrayGetSize(pSnapInfo); i++) {
@@ -1457,7 +1457,7 @@ int32_t taskDbDestroySnap(void* arg, SArray* pSnapInfo) {
 
     taskDbUnRefChkp(*pTaskDb, pSnap->chkpId);
   }
-  taosThreadMutexUnlock(&pMeta->backendMutex);
+  streamMutexUnlock(&pMeta->backendMutex);
   return 0;
 }
 #ifdef BUILD_NO_CALL
@@ -1697,17 +1697,17 @@ int32_t streamBackendDoCheckpoint(void* arg, int64_t chkpId, int64_t processVer)
 SListNode* streamBackendAddCompare(void* backend, void* arg) {
   SBackendWrapper* pHandle = (SBackendWrapper*)backend;
   SListNode*       node = NULL;
-  taosThreadMutexLock(&pHandle->mutex);
+  streamMutexLock(&pHandle->mutex);
   node = tdListAdd(pHandle->list, arg);
-  taosThreadMutexUnlock(&pHandle->mutex);
+  streamMutexUnlock(&pHandle->mutex);
   return node;
 }
 void streamBackendDelCompare(void* backend, void* arg) {
   SBackendWrapper* pHandle = (SBackendWrapper*)backend;
   SListNode*       node = NULL;
-  taosThreadMutexLock(&pHandle->mutex);
+  streamMutexLock(&pHandle->mutex);
   node = tdListPopNode(pHandle->list, arg);
-  taosThreadMutexUnlock(&pHandle->mutex);
+  streamMutexUnlock(&pHandle->mutex);
   if (node) {
     streamStateDestroyCompar(node->data);
     taosMemoryFree(node);
@@ -2461,9 +2461,9 @@ int32_t taskDbBuildFullPath(char* path, char* key, char** dbFullPath, char** sta
 
 void taskDbUpdateChkpId(void* pTaskDb, int64_t chkpId) {
   STaskDbWrapper* p = pTaskDb;
-  taosThreadMutexLock(&p->mutex);
+  streamMutexLock(&p->mutex);
   p->chkpId = chkpId;
-  taosThreadMutexUnlock(&p->mutex);
+  streamMutexUnlock(&p->mutex);
 }
 
 STaskDbWrapper* taskDbOpenImpl(const char* key, char* statePath, char* dbPath) {
@@ -2622,7 +2622,7 @@ void taskDbDestroy(void* pDb, bool flush) {
   taosMemoryFree(wrapper->pCfOpts);
   taosMemoryFree(wrapper->pCfParams);
 
-  taosThreadMutexDestroy(&wrapper->mutex);
+  streamMutexDestroy(&wrapper->mutex);
 
   taskDbDestroyChkpOpt(wrapper);
 
@@ -2957,7 +2957,7 @@ int streamStateOpenBackend(void* backend, SStreamState* pState) {
   SBackendWrapper*   handle = backend;
   SBackendCfWrapper* pBackendCfWrapper = taosMemoryCalloc(1, sizeof(SBackendCfWrapper));
 
-  taosThreadMutexLock(&handle->cfMutex);
+  streamMutexLock(&handle->cfMutex);
   RocksdbCfInst** ppInst = taosHashGet(handle->cfInst, pState->pTdbState->idstr, strlen(pState->pTdbState->idstr) + 1);
   if (ppInst != NULL && *ppInst != NULL) {
     RocksdbCfInst* inst = *ppInst;
@@ -2970,7 +2970,7 @@ int streamStateOpenBackend(void* backend, SStreamState* pState) {
     pBackendCfWrapper->param = inst->param;
     pBackendCfWrapper->pBackend = handle;
     pBackendCfWrapper->pComparNode = inst->pCompareNode;
-    taosThreadMutexUnlock(&handle->cfMutex);
+    streamMutexUnlock(&handle->cfMutex);
     pBackendCfWrapper->backendId = pState->streamBackendRid;
     memcpy(pBackendCfWrapper->idstr, pState->pTdbState->idstr, sizeof(pState->pTdbState->idstr));
 
@@ -2987,7 +2987,7 @@ int streamStateOpenBackend(void* backend, SStreamState* pState) {
     inst->rOpt = NULL;
     return 0;
   }
-  taosThreadMutexUnlock(&handle->cfMutex);
+  streamMutexUnlock(&handle->cfMutex);
 
   char* err = NULL;
   int   cfLen = sizeof(ginitDict) / sizeof(ginitDict[0]);
@@ -3046,14 +3046,14 @@ void streamStateCloseBackend(SStreamState* pState, bool remove) {
 
   stInfo("start to close state on backend: %p", pHandle);
 
-  taosThreadMutexLock(&pHandle->cfMutex);
+  streamMutexLock(&pHandle->cfMutex);
   RocksdbCfInst** ppInst = taosHashGet(pHandle->cfInst, wrapper->idstr, strlen(pState->pTdbState->idstr) + 1);
   if (ppInst != NULL && *ppInst != NULL) {
     RocksdbCfInst* inst = *ppInst;
     taosMemoryFree(inst);
     taosHashRemove(pHandle->cfInst, pState->pTdbState->idstr, strlen(pState->pTdbState->idstr) + 1);
   }
-  taosThreadMutexUnlock(&pHandle->cfMutex);
+  streamMutexUnlock(&pHandle->cfMutex);
 
   char* status[] = {"close", "drop"};
   stInfo("start to %s state %p on backendWrapper %p %s", status[remove == false ? 0 : 1], pState, wrapper,
@@ -3085,7 +3085,7 @@ int streamStateGetCfIdx(SStreamState* pState, const char* funcName) {
       return -1;
     }
 
-    taosThreadMutexLock(&wrapper->mutex);
+    streamMutexLock(&wrapper->mutex);
 
     rocksdb_column_family_handle_t* cf = wrapper->pCf[idx];
     if (cf == NULL) {
@@ -3100,7 +3100,7 @@ int streamStateGetCfIdx(SStreamState* pState, const char* funcName) {
         wrapper->pCf[idx] = cf;
       }
     }
-    taosThreadMutexUnlock(&wrapper->mutex);
+    streamMutexUnlock(&wrapper->mutex);
   }
 
   return idx;

source/libs/stream/src/streamCheckStatus.c

@@ -23,10 +23,10 @@
 static void    processDownstreamReadyRsp(SStreamTask* pTask);
 static void    addIntoNodeUpdateList(SStreamTask* pTask, int32_t nodeId);
 static void    rspMonitorFn(void* param, void* tmrId);
-static int32_t streamTaskInitTaskCheckInfo(STaskCheckInfo* pInfo, STaskOutputInfo* pOutputInfo, int64_t startTs);
+static void    streamTaskInitTaskCheckInfo(STaskCheckInfo* pInfo, STaskOutputInfo* pOutputInfo, int64_t startTs);
 static int32_t streamTaskStartCheckDownstream(STaskCheckInfo* pInfo, const char* id);
-static int32_t streamTaskCompleteCheckRsp(STaskCheckInfo* pInfo, bool lock, const char* id);
+static void    streamTaskCompleteCheckRsp(STaskCheckInfo* pInfo, bool lock, const char* id);
-static int32_t streamTaskAddReqInfo(STaskCheckInfo* pInfo, int64_t reqId, int32_t taskId, int32_t vgId, const char* id);
+static void    streamTaskAddReqInfo(STaskCheckInfo* pInfo, int64_t reqId, int32_t taskId, int32_t vgId, const char* id);
 static void    doSendCheckMsg(SStreamTask* pTask, SDownstreamStatusInfo* p);
 static void    handleTimeoutDownstreamTasks(SStreamTask* pTask, SArray* pTimeoutList);
 static void    handleNotReadyDownstreamTask(SStreamTask* pTask, SArray* pNotReadyList);
@@ -41,9 +41,9 @@ static SDownstreamStatusInfo* findCheckRspStatus(STaskCheckInfo* pInfo, int32_t
 int32_t streamTaskCheckStatus(SStreamTask* pTask, int32_t upstreamTaskId, int32_t vgId, int64_t stage,
                               int64_t* oldStage) {
   SStreamUpstreamEpInfo* pInfo = NULL;
-  int32_t code = streamTaskGetUpstreamTaskEpInfo(pTask, upstreamTaskId, &pInfo);
+  streamTaskGetUpstreamTaskEpInfo(pTask, upstreamTaskId, &pInfo);
-  if (code != TSDB_CODE_SUCCESS) {
+  if (pInfo == NULL) {
-    return code;
+    return TSDB_CODE_STREAM_TASK_NOT_EXIST;
   }
 
   *oldStage = pInfo->stage;
@@ -65,21 +65,21 @@ int32_t streamTaskCheckStatus(SStreamTask* pTask, int32_t upstreamTaskId, int32_
             ", prev:%" PRId64,
             id, upstreamTaskId, vgId, stage, pInfo->stage);
     // record the checkpoint failure id and sent to mnode
-    taosThreadMutexLock(&pTask->lock);
+    streamMutexLock(&pTask->lock);
     ETaskStatus status = streamTaskGetStatus(pTask).state;
     if (status == TASK_STATUS__CK) {
       streamTaskSetFailedCheckpointId(pTask);
     }
-    taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
   }
 
   if (pInfo->stage != stage) {
-    taosThreadMutexLock(&pTask->lock);
+    streamMutexLock(&pTask->lock);
     ETaskStatus status = streamTaskGetStatus(pTask).state;
     if (status == TASK_STATUS__CK) {
       streamTaskSetFailedCheckpointId(pTask);
     }
-    taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     return TASK_UPSTREAM_NEW_STAGE;
   } else if (pTask->status.downstreamReady != 1) {
@@ -120,7 +120,7 @@ void streamTaskSendCheckMsg(SStreamTask* pTask) {
             idstr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, pRange->range.minVer,
             pRange->range.maxVer, pWindow->skey, pWindow->ekey, req.reqId);
 
-    streamSendCheckMsg(pTask, &req, pTask->outputInfo.fixedDispatcher.nodeId, &pTask->outputInfo.fixedDispatcher.epSet);
+    (void) streamSendCheckMsg(pTask, &req, pTask->outputInfo.fixedDispatcher.nodeId, &pTask->outputInfo.fixedDispatcher.epSet);
 
   } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) {
     streamTaskStartMonitorCheckRsp(pTask);
@@ -140,7 +140,7 @@ void streamTaskSendCheckMsg(SStreamTask* pTask) {
       stDebug("s-task:%s (vgId:%d) stage:%" PRId64
               " check downstream task:0x%x (vgId:%d) (shuffle), idx:%d, reqId:0x%" PRIx64,
               idstr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, i, req.reqId);
-      streamSendCheckMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet);
+      (void) streamSendCheckMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet);
     }
   } else {  // for sink task, set it ready directly.
     stDebug("s-task:%s (vgId:%d) set downstream ready, since no downstream", idstr, pTask->info.nodeId);
@@ -265,7 +265,7 @@ int32_t streamTaskSendCheckRsp(const SStreamMeta* pMeta, int32_t vgId, SStreamTa
 
   void* abuf = POINTER_SHIFT(buf, sizeof(SMsgHead));
   tEncoderInit(&encoder, (uint8_t*)abuf, len);
-  tEncodeStreamTaskCheckRsp(&encoder, pRsp);
+  (void) tEncodeStreamTaskCheckRsp(&encoder, pRsp);
   tEncoderClear(&encoder);
 
   SRpcMsg rspMsg = {.code = 0, .pCont = buf, .contLen = sizeof(SMsgHead) + len, .info = *pRpcInfo};
@@ -274,14 +274,16 @@ int32_t streamTaskSendCheckRsp(const SStreamMeta* pMeta, int32_t vgId, SStreamTa
   return 0;
 }
 
-int32_t streamTaskStartMonitorCheckRsp(SStreamTask* pTask) {
+void streamTaskStartMonitorCheckRsp(SStreamTask* pTask) {
+  int32_t         vgId = pTask->pMeta->vgId;
   STaskCheckInfo* pInfo = &pTask->taskCheckInfo;
-  taosThreadMutexLock(&pInfo->checkInfoLock);
+
+  streamMutexLock(&pInfo->checkInfoLock);
 
   int32_t code = streamTaskStartCheckDownstream(pInfo, pTask->id.idStr);
   if (code != TSDB_CODE_SUCCESS) {
-    taosThreadMutexUnlock(&pInfo->checkInfoLock);
+    streamMutexUnlock(&pInfo->checkInfoLock);
-    return TSDB_CODE_FAILED;
+    return;
   }
 
   /*SStreamTask* p = */ streamMetaAcquireOneTask(pTask);  // add task ref here
@@ -293,20 +295,18 @@ int32_t streamTaskStartMonitorCheckRsp(SStreamTask* pTask) {
   if (pInfo->checkRspTmr == NULL) {
     pInfo->checkRspTmr = taosTmrStart(rspMonitorFn, CHECK_RSP_CHECK_INTERVAL, pTask, streamTimer);
   } else {
-    taosTmrReset(rspMonitorFn, CHECK_RSP_CHECK_INTERVAL, pTask, streamTimer, &pInfo->checkRspTmr);
+    streamTmrReset(rspMonitorFn, CHECK_RSP_CHECK_INTERVAL, pTask, streamTimer, &pInfo->checkRspTmr, vgId, "check-status-monitor");
   }
 
-  taosThreadMutexUnlock(&pInfo->checkInfoLock);
+  streamMutexUnlock(&pInfo->checkInfoLock);
-  return 0;
 }
 
-int32_t streamTaskStopMonitorCheckRsp(STaskCheckInfo* pInfo, const char* id) {
+void streamTaskStopMonitorCheckRsp(STaskCheckInfo* pInfo, const char* id) {
-  taosThreadMutexLock(&pInfo->checkInfoLock);
+  streamMutexLock(&pInfo->checkInfoLock);
   pInfo->stopCheckProcess = 1;
-  taosThreadMutexUnlock(&pInfo->checkInfoLock);
+  streamMutexUnlock(&pInfo->checkInfoLock);
 
   stDebug("s-task:%s set stop check-rsp monitor flag", id);
-  return TSDB_CODE_SUCCESS;
 }
 
 void streamTaskCleanupCheckInfo(STaskCheckInfo* pInfo) {
@@ -316,21 +316,21 @@ void streamTaskCleanupCheckInfo(STaskCheckInfo* pInfo) {
   pInfo->pList = NULL;
 
   if (pInfo->checkRspTmr != NULL) {
-    /*bool ret = */ taosTmrStop(pInfo->checkRspTmr);
+    (void) taosTmrStop(pInfo->checkRspTmr);
     pInfo->checkRspTmr = NULL;
   }
 
-  taosThreadMutexDestroy(&pInfo->checkInfoLock);
+  streamMutexDestroy(&pInfo->checkInfoLock);
 }
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 void processDownstreamReadyRsp(SStreamTask* pTask) {
   EStreamTaskEvent event = (pTask->info.fillHistory == 0) ? TASK_EVENT_INIT : TASK_EVENT_INIT_SCANHIST;
-  streamTaskOnHandleEventSuccess(pTask->status.pSM, event, NULL, NULL);
+  (void) streamTaskOnHandleEventSuccess(pTask->status.pSM, event, NULL, NULL);
 
   int64_t checkTs = pTask->execInfo.checkTs;
   int64_t readyTs = pTask->execInfo.readyTs;
-  streamMetaAddTaskLaunchResult(pTask->pMeta, pTask->id.streamId, pTask->id.taskId, checkTs, readyTs, true);
+  (void) streamMetaAddTaskLaunchResult(pTask->pMeta, pTask->id.streamId, pTask->id.taskId, checkTs, readyTs, true);
 
   if (pTask->status.taskStatus == TASK_STATUS__HALT) {
     ASSERT(HAS_RELATED_FILLHISTORY_TASK(pTask) && (pTask->info.fillHistory == 0));
@@ -338,21 +338,25 @@ void processDownstreamReadyRsp(SStreamTask* pTask) {
     // halt it self for count window stream task until the related fill history task completed.
     stDebug("s-task:%s level:%d initial status is %s from mnode, set it to be halt", pTask->id.idStr,
             pTask->info.taskLevel, streamTaskGetStatusStr(pTask->status.taskStatus));
-    streamTaskHandleEvent(pTask->status.pSM, TASK_EVENT_HALT);
+    int32_t code = streamTaskHandleEvent(pTask->status.pSM, TASK_EVENT_HALT);
+    if (code != 0) {
+      // todo: handle error
+    }
   }
 
   // start the related fill-history task, when current task is ready
   // not invoke in success callback due to the deadlock.
+  // todo: let's retry
   if (HAS_RELATED_FILLHISTORY_TASK(pTask)) {
     stDebug("s-task:%s try to launch related fill-history task", pTask->id.idStr);
-    streamLaunchFillHistoryTask(pTask);
+    (void) streamLaunchFillHistoryTask(pTask);
   }
 }
 
 void addIntoNodeUpdateList(SStreamTask* pTask, int32_t nodeId) {
   int32_t vgId = pTask->pMeta->vgId;
 
-  taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   int32_t num = taosArrayGetSize(pTask->outputInfo.pNodeEpsetUpdateList);
   bool    existed = false;
   for (int i = 0; i < num; ++i) {
@@ -365,16 +369,18 @@ void addIntoNodeUpdateList(SStreamTask* pTask, int32_t nodeId) {
 
   if (!existed) {
     SDownstreamTaskEpset t = {.nodeId = nodeId};
-    taosArrayPush(pTask->outputInfo.pNodeEpsetUpdateList, &t);
+    void* p = taosArrayPush(pTask->outputInfo.pNodeEpsetUpdateList, &t);
-
+    if (p == NULL) {
+      // todo let's retry
+    }
     stInfo("s-task:%s vgId:%d downstream nodeId:%d needs to be updated, total needs updated:%d", pTask->id.idStr, vgId,
            t.nodeId, (num + 1));
   }
 
-  taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 }
 
-int32_t streamTaskInitTaskCheckInfo(STaskCheckInfo* pInfo, STaskOutputInfo* pOutputInfo, int64_t startTs) {
+void streamTaskInitTaskCheckInfo(STaskCheckInfo* pInfo, STaskOutputInfo* pOutputInfo, int64_t startTs) {
   taosArrayClear(pInfo->pList);
 
   if (pOutputInfo->type == TASK_OUTPUT__FIXED_DISPATCH) {
@@ -387,7 +393,6 @@ int32_t streamTaskInitTaskCheckInfo(STaskCheckInfo* pInfo, STaskOutputInfo* pOut
   pInfo->startTs = startTs;
   pInfo->timeoutStartTs = startTs;
   pInfo->stopCheckProcess = 0;
-  return TSDB_CODE_SUCCESS;
 }
 
 SDownstreamStatusInfo* findCheckRspStatus(STaskCheckInfo* pInfo, int32_t taskId) {
@@ -403,7 +408,7 @@ SDownstreamStatusInfo* findCheckRspStatus(STaskCheckInfo* pInfo, int32_t taskId)
 
 int32_t streamTaskUpdateCheckInfo(STaskCheckInfo* pInfo, int32_t taskId, int32_t status, int64_t rspTs, int64_t reqId,
                                   int32_t* pNotReady, const char* id) {
-  taosThreadMutexLock(&pInfo->checkInfoLock);
+  streamMutexLock(&pInfo->checkInfoLock);
 
   SDownstreamStatusInfo* p = findCheckRspStatus(pInfo, taskId);
   if (p != NULL) {
@@ -411,7 +416,7 @@ int32_t streamTaskUpdateCheckInfo(STaskCheckInfo* pInfo, int32_t taskId, int32_t
       stError("s-task:%s reqId:0x%" PRIx64 " expected:0x%" PRIx64
               " expired check-rsp recv from downstream task:0x%x, discarded",
               id, reqId, p->reqId, taskId);
-      taosThreadMutexUnlock(&pInfo->checkInfoLock);
+      streamMutexUnlock(&pInfo->checkInfoLock);
       return TSDB_CODE_FAILED;
     }
 
@@ -425,11 +430,11 @@ int32_t streamTaskUpdateCheckInfo(STaskCheckInfo* pInfo, int32_t taskId, int32_t
     p->status = status;
     p->rspTs = rspTs;
 
-    taosThreadMutexUnlock(&pInfo->checkInfoLock);
+    streamMutexUnlock(&pInfo->checkInfoLock);
     return TSDB_CODE_SUCCESS;
   }
 
-  taosThreadMutexUnlock(&pInfo->checkInfoLock);
+  streamMutexUnlock(&pInfo->checkInfoLock);
   stError("s-task:%s unexpected check rsp msg, invalid downstream task:0x%x, reqId:%" PRIx64 " discarded", id, taskId,
           reqId);
   return TSDB_CODE_FAILED;
@@ -450,9 +455,9 @@ int32_t streamTaskStartCheckDownstream(STaskCheckInfo* pInfo, const char* id) {
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t streamTaskCompleteCheckRsp(STaskCheckInfo* pInfo, bool lock, const char* id) {
+void streamTaskCompleteCheckRsp(STaskCheckInfo* pInfo, bool lock, const char* id) {
   if (lock) {
-    taosThreadMutexLock(&pInfo->checkInfoLock);
+    streamMutexLock(&pInfo->checkInfoLock);
   }
 
   if (pInfo->inCheckProcess) {
@@ -474,27 +479,28 @@ int32_t streamTaskCompleteCheckRsp(STaskCheckInfo* pInfo, bool lock, const char*
   }
 
   if (lock) {
-    taosThreadMutexUnlock(&pInfo->checkInfoLock);
+    streamMutexUnlock(&pInfo->checkInfoLock);
+  }
 }
 
-  return 0;
+// todo: retry until success
-}
+void streamTaskAddReqInfo(STaskCheckInfo* pInfo, int64_t reqId, int32_t taskId, int32_t vgId, const char* id) {
-
-int32_t streamTaskAddReqInfo(STaskCheckInfo* pInfo, int64_t reqId, int32_t taskId, int32_t vgId, const char* id) {
   SDownstreamStatusInfo info = {.taskId = taskId, .status = -1, .vgId = vgId, .reqId = reqId, .rspTs = 0};
-  taosThreadMutexLock(&pInfo->checkInfoLock);
+  streamMutexLock(&pInfo->checkInfoLock);
 
   SDownstreamStatusInfo* p = findCheckRspStatus(pInfo, taskId);
   if (p != NULL) {
     stDebug("s-task:%s check info to task:0x%x already sent", id, taskId);
-    taosThreadMutexUnlock(&pInfo->checkInfoLock);
+    streamMutexUnlock(&pInfo->checkInfoLock);
-    return TSDB_CODE_SUCCESS;
+    return;
   }
 
-  taosArrayPush(pInfo->pList, &info);
+  void* px = taosArrayPush(pInfo->pList, &info);
+  if (px == NULL) {
+    // todo: retry
+  }
 
-  taosThreadMutexUnlock(&pInfo->checkInfoLock);
+  streamMutexUnlock(&pInfo->checkInfoLock);
-  return TSDB_CODE_SUCCESS;
 }
 
 void doSendCheckMsg(SStreamTask* pTask, SDownstreamStatusInfo* p) {
@@ -519,7 +525,7 @@ void doSendCheckMsg(SStreamTask* pTask, SDownstreamStatusInfo* p) {
     stDebug("s-task:%s (vgId:%d) stage:%" PRId64 " re-send check downstream task:0x%x(vgId:%d) reqId:0x%" PRIx64, id,
             pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, req.reqId);
 
-    streamSendCheckMsg(pTask, &req, pOutputInfo->fixedDispatcher.nodeId, &pOutputInfo->fixedDispatcher.epSet);
+    (void) streamSendCheckMsg(pTask, &req, pOutputInfo->fixedDispatcher.nodeId, &pOutputInfo->fixedDispatcher.epSet);
   } else if (pOutputInfo->type == TASK_OUTPUT__SHUFFLE_DISPATCH) {
     SArray* vgInfo = pOutputInfo->shuffleDispatcher.dbInfo.pVgroupInfos;
     int32_t numOfVgs = taosArrayGetSize(vgInfo);
@@ -533,7 +539,7 @@ void doSendCheckMsg(SStreamTask* pTask, SDownstreamStatusInfo* p) {
         stDebug("s-task:%s (vgId:%d) stage:%" PRId64
                 " re-send check downstream task:0x%x(vgId:%d) (shuffle), idx:%d reqId:0x%" PRIx64,
                 id, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, i, p->reqId);
-        streamSendCheckMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet);
+        (void) streamSendCheckMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet);
         break;
       }
     }
@@ -556,12 +562,12 @@ void getCheckRspStatus(STaskCheckInfo* pInfo, int64_t el, int32_t* numOfReady, i
       if (p->rspTs == 0) {  // not response yet
         ASSERT(p->status == -1);
         if (el >= CHECK_NOT_RSP_DURATION) {  // not receive info for 10 sec.
-          taosArrayPush(pTimeoutList, &p->taskId);
+          (void) taosArrayPush(pTimeoutList, &p->taskId);
         } else {                // el < CHECK_NOT_RSP_DURATION
           (*numOfNotRsp) += 1;  // do nothing and continue waiting for their rsp
         }
       } else {
-        taosArrayPush(pNotReadyList, &p->taskId);
+        (void) taosArrayPush(pNotReadyList, &p->taskId);
       }
     }
   }
@@ -676,7 +682,7 @@ void rspMonitorFn(void* param, void* tmrId) {
     // not record the failed of the current task if try to close current vnode
     // otherwise, the put of message operation may incur invalid read of message queue.
     if (!pMeta->closeFlag) {
-      addDownstreamFailedStatusResultAsync(pTask->pMsgCb, vgId, pTask->id.streamId, pTask->id.taskId);
+      (void) addDownstreamFailedStatusResultAsync(pTask->pMsgCb, vgId, pTask->id.streamId, pTask->id.taskId);
     }
 
     streamMetaReleaseTask(pMeta, pTask);
@@ -692,14 +698,14 @@ void rspMonitorFn(void* param, void* tmrId) {
     return;
   }
 
-  taosThreadMutexLock(&pInfo->checkInfoLock);
+  streamMutexLock(&pInfo->checkInfoLock);
   if (pInfo->notReadyTasks == 0) {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s status:%s vgId:%d all downstream ready, quit from monitor rsp tmr, ref:%d", id, pStat.name,
             vgId, ref);
 
     streamTaskCompleteCheckRsp(pInfo, false, id);
-    taosThreadMutexUnlock(&pInfo->checkInfoLock);
+    streamMutexUnlock(&pInfo->checkInfoLock);
     streamMetaReleaseTask(pMeta, pTask);
     return;
   }
@@ -723,7 +729,7 @@ void rspMonitorFn(void* param, void* tmrId) {
           id, pStat.name, vgId, total, numOfNotRsp, numOfNotReady, numOfFault, numOfTimeout, numOfReady, ref);
 
       streamTaskCompleteCheckRsp(pInfo, false, id);
-      taosThreadMutexUnlock(&pInfo->checkInfoLock);
+      streamMutexUnlock(&pInfo->checkInfoLock);
       streamMetaReleaseTask(pMeta, pTask);
 
       taosArrayDestroy(pNotReadyList);
@@ -743,9 +749,9 @@ void rspMonitorFn(void* param, void* tmrId) {
         id, pStat.name, vgId, total, numOfNotRsp, numOfNotReady, numOfFault, numOfTimeout, numOfReady, ref);
 
     streamTaskCompleteCheckRsp(pInfo, false, id);
-    taosThreadMutexUnlock(&pInfo->checkInfoLock);
+    streamMutexUnlock(&pInfo->checkInfoLock);
 
-    addDownstreamFailedStatusResultAsync(pTask->pMsgCb, vgId, pTask->id.streamId, pTask->id.taskId);
+    (void) addDownstreamFailedStatusResultAsync(pTask->pMsgCb, vgId, pTask->id.streamId, pTask->id.taskId);
     streamMetaReleaseTask(pMeta, pTask);
 
     taosArrayDestroy(pNotReadyList);
@@ -761,8 +767,8 @@ void rspMonitorFn(void* param, void* tmrId) {
     handleTimeoutDownstreamTasks(pTask, pTimeoutList);
   }
 
-  taosTmrReset(rspMonitorFn, CHECK_RSP_CHECK_INTERVAL, pTask, streamTimer, &pInfo->checkRspTmr);
+  streamTmrReset(rspMonitorFn, CHECK_RSP_CHECK_INTERVAL, pTask, streamTimer, &pInfo->checkRspTmr, vgId, "check-status-monitor");
-  taosThreadMutexUnlock(&pInfo->checkInfoLock);
+  streamMutexUnlock(&pInfo->checkInfoLock);
 
   stDebug(
       "s-task:%s vgId:%d continue checking rsp in 300ms, total:%d, notRsp:%d, notReady:%d, fault:%d, timeout:%d, "

source/libs/stream/src/streamCheckpoint.c

@@ -77,6 +77,7 @@ int32_t createChkptTriggerBlock(SStreamTask* pTask, int32_t checkpointType, int6
 }
 
 // this message must be put into inputq successfully, continue retrying until it succeeds
+// todo must be success
 int32_t appendCheckpointIntoInputQ(SStreamTask* pTask, int32_t checkpointType, int64_t checkpointId, int32_t transId,
                                    int32_t srcTaskId) {
   SStreamDataBlock* pCheckpoint = NULL;
@@ -119,7 +120,7 @@ int32_t streamTaskProcessCheckpointTriggerRsp(SStreamTask* pTask, SCheckpointTri
     return TSDB_CODE_SUCCESS;
   }
 
-  appendCheckpointIntoInputQ(pTask, STREAM_INPUT__CHECKPOINT_TRIGGER, pRsp->checkpointId, pRsp->transId,
+  (void)appendCheckpointIntoInputQ(pTask, STREAM_INPUT__CHECKPOINT_TRIGGER, pRsp->checkpointId, pRsp->transId,
                                    pRsp->upstreamTaskId);
   return TSDB_CODE_SUCCESS;
 }
@@ -164,7 +165,7 @@ int32_t continueDispatchCheckpointTriggerBlock(SStreamDataBlock* pBlock, SStream
   int32_t code = taosWriteQitem(pTask->outputq.queue->pQueue, pBlock);
   if (code == 0) {
     ASSERT(pTask->chkInfo.pActiveInfo->dispatchTrigger == false);
-    streamDispatchStreamBlock(pTask);
+    code = streamDispatchStreamBlock(pTask);
   } else {
     stError("s-task:%s failed to put checkpoint into outputQ, code:%s", pTask->id.idStr, tstrerror(code));
     streamFreeQitem((SStreamQueueItem*)pBlock);
@@ -184,12 +185,12 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
 
   SActiveCheckpointInfo* pActiveInfo = pTask->chkInfo.pActiveInfo;
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   if (pTask->chkInfo.checkpointId > checkpointId) {
     stError("s-task:%s vgId:%d current checkpointId:%" PRId64
             " recv expired checkpoint-trigger block, checkpointId:%" PRId64 " transId:%d, discard",
             id, vgId, pTask->chkInfo.checkpointId, checkpointId, transId);
-    code = taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     streamFreeQitem((SStreamQueueItem*)pBlock);
     return code;
@@ -199,13 +200,16 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
     {  // send checkpoint-ready msg to upstream
       SRpcMsg msg = {0};
       SStreamUpstreamEpInfo* pInfo = NULL;
-      code = streamTaskGetUpstreamTaskEpInfo(pTask, pBlock->srcTaskId, &pInfo);
+      streamTaskGetUpstreamTaskEpInfo(pTask, pBlock->srcTaskId, &pInfo);
-      if (code != TSDB_CODE_SUCCESS) {
+      if (pInfo == NULL) {
-        return code;
+        streamMutexUnlock(&pTask->lock);
+        return TSDB_CODE_STREAM_TASK_NOT_EXIST;
       }
 
-      initCheckpointReadyMsg(pTask, pInfo->nodeId, pBlock->srcTaskId, pInfo->childId, checkpointId, &msg);
+      code = initCheckpointReadyMsg(pTask, pInfo->nodeId, pBlock->srcTaskId, pInfo->childId, checkpointId, &msg);
-      tmsgSendReq(&pInfo->epSet, &msg);
+      if (code == TSDB_CODE_SUCCESS) {
+        (void)tmsgSendReq(&pInfo->epSet, &msg);
+      }
     }
 
     stWarn(
@@ -214,7 +218,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
         id, vgId, pBlock->srcTaskId);
 
     streamTaskOpenUpstreamInput(pTask, pBlock->srcTaskId);
-    code = taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     streamFreeQitem((SStreamQueueItem*)pBlock);
     return code;
@@ -225,7 +229,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
       stError("s-task:%s vgId:%d active checkpointId:%" PRId64 ", recv invalid checkpoint-trigger checkpointId:%" PRId64
               " discard",
               id, vgId, pActiveInfo->activeId, checkpointId);
-      code = taosThreadMutexUnlock(&pTask->lock);
+      streamMutexUnlock(&pTask->lock);
 
       streamFreeQitem((SStreamQueueItem*)pBlock);
       return code;
@@ -235,7 +239,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
             "s-task:%s vgId:%d all upstream checkpoint-trigger recv, discard this checkpoint-trigger, "
             "checkpointId:%" PRId64 " transId:%d",
             id, vgId, checkpointId, transId);
-        code = taosThreadMutexUnlock(&pTask->lock);
+        streamMutexUnlock(&pTask->lock);
         streamFreeQitem((SStreamQueueItem*)pBlock);
         return code;
       }
@@ -250,7 +254,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
                    ", prev recvTs:%" PRId64 " discard",
                    pTask->id.idStr, p->upstreamTaskId, p->upstreamNodeId, p->checkpointId, p->recvTs);
 
-            code = taosThreadMutexUnlock(&pTask->lock);
+            streamMutexUnlock(&pTask->lock);
             streamFreeQitem((SStreamQueueItem*)pBlock);
             return code;
           }
@@ -259,7 +263,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
     }
   }
 
-  (void) taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   stDebug("s-task:%s vgId:%d start to handle the checkpoint-trigger block, checkpointId:%" PRId64 " ver:%" PRId64
           ", transId:%d current active checkpointId:%" PRId64,
@@ -284,7 +288,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
     if (pActiveInfo->pChkptTriggerTmr == NULL) {
       pActiveInfo->pChkptTriggerTmr = taosTmrStart(checkpointTriggerMonitorFn, 100, pTask, streamTimer);
     } else {
-      taosTmrReset(checkpointTriggerMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pChkptTriggerTmr);
+      streamTmrReset(checkpointTriggerMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pChkptTriggerTmr, vgId, "trigger-recv-monitor");
     }
   }
 
@@ -300,9 +304,10 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
 
     if (type == TASK_OUTPUT__FIXED_DISPATCH || type == TASK_OUTPUT__SHUFFLE_DISPATCH) {
       stDebug("s-task:%s set childIdx:%d, and add checkpoint-trigger block into outputQ", id, pTask->info.selfChildId);
-      continueDispatchCheckpointTriggerBlock(pBlock, pTask);
+      (void)continueDispatchCheckpointTriggerBlock(pBlock, pTask);   // todo handle this failure
     } else {  // only one task exists, no need to dispatch downstream info
-      appendCheckpointIntoInputQ(pTask, STREAM_INPUT__CHECKPOINT, pActiveInfo->activeId, pActiveInfo->transId, -1);
+      (void)appendCheckpointIntoInputQ(pTask, STREAM_INPUT__CHECKPOINT, pActiveInfo->activeId, pActiveInfo->transId,
+                                       -1);
       streamFreeQitem((SStreamQueueItem*)pBlock);
     }
   } else if (taskLevel == TASK_LEVEL__SINK || taskLevel == TASK_LEVEL__AGG) {
@@ -312,8 +317,9 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
       pTask->execInfo.checkpoint += 1;
     }
 
+    // todo: handle this
     // update the child Id for downstream tasks
-    streamAddCheckpointReadyMsg(pTask, pBlock->srcTaskId, pTask->info.selfChildId, checkpointId);
+    (void) streamAddCheckpointReadyMsg(pTask, pBlock->srcTaskId, pTask->info.selfChildId, checkpointId);
 
     // there are still some upstream tasks not send checkpoint request, do nothing and wait for then
     if (pActiveInfo->allUpstreamTriggerRecv != 1) {
@@ -325,7 +331,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
     if (taskLevel == TASK_LEVEL__SINK) {
       stDebug("s-task:%s process checkpoint-trigger block, all %d upstreams sent, send ready msg to upstream", id, num);
       streamFreeQitem((SStreamQueueItem*)pBlock);
-      streamTaskBuildCheckpoint(pTask);
+      (void)streamTaskBuildCheckpoint(pTask);   // todo: not handle error yet
     } else {  // source & agg tasks need to forward the checkpoint msg downwards
       stDebug("s-task:%s process checkpoint-trigger block, all %d upstreams sent, forwards to downstream", id, num);
 
@@ -369,7 +375,7 @@ int32_t streamProcessCheckpointReadyMsg(SStreamTask* pTask, int64_t checkpointId
     return -1;
   }
 
-  (void) taosThreadMutexLock(&pInfo->lock);
+  streamMutexLock(&pInfo->lock);
 
   // only when all downstream tasks are send checkpoint rsp, we can start the checkpoint procedure for the agg task
   int32_t size = taosArrayGetSize(pInfo->pCheckpointReadyRecvList);
@@ -391,16 +397,16 @@ int32_t streamProcessCheckpointReadyMsg(SStreamTask* pTask, int64_t checkpointId
                                      .transId = pInfo->transId,
                                      .streamId = pTask->id.streamId,
                                      .downstreamNodeId = downstreamNodeId};
-    taosArrayPush(pInfo->pCheckpointReadyRecvList, &info);
+    (void)taosArrayPush(pInfo->pCheckpointReadyRecvList, &info);
   }
 
   int32_t notReady = total - taosArrayGetSize(pInfo->pCheckpointReadyRecvList);
   int32_t transId = pInfo->transId;
-  (void) taosThreadMutexUnlock(&pInfo->lock);
+  streamMutexUnlock(&pInfo->lock);
 
   if (notReady == 0) {
     stDebug("s-task:%s all downstream tasks have completed build checkpoint, do checkpoint for current task", id);
-    appendCheckpointIntoInputQ(pTask, STREAM_INPUT__CHECKPOINT, checkpointId, transId, -1);
+    (void)appendCheckpointIntoInputQ(pTask, STREAM_INPUT__CHECKPOINT, checkpointId, transId, -1);
   }
 
   return 0;
@@ -411,7 +417,7 @@ int32_t streamTaskProcessCheckpointReadyRsp(SStreamTask* pTask, int32_t upstream
   int64_t                now = taosGetTimestampMs();
   int32_t                numOfConfirmed = 0;
 
-  (void) taosThreadMutexLock(&pInfo->lock);
+  streamMutexLock(&pInfo->lock);
   for (int32_t i = 0; i < taosArrayGetSize(pInfo->pReadyMsgList); ++i) {
     STaskCheckpointReadyInfo* pReadyInfo = taosArrayGet(pInfo->pReadyMsgList, i);
     if (pReadyInfo->upstreamTaskId == upstreamTaskId && pReadyInfo->checkpointId == checkpointId) {
@@ -432,7 +438,7 @@ int32_t streamTaskProcessCheckpointReadyRsp(SStreamTask* pTask, int32_t upstream
   stDebug("s-task:%s send checkpoint-ready msg to %d upstream confirmed, checkpointId:%" PRId64, pTask->id.idStr,
           numOfConfirmed, checkpointId);
 
-  (void) taosThreadMutexUnlock(&pInfo->lock);
+  streamMutexUnlock(&pInfo->lock);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -440,12 +446,12 @@ void streamTaskClearCheckInfo(SStreamTask* pTask, bool clearChkpReadyMsg) {
   pTask->chkInfo.startTs = 0;             // clear the recorded start time
   streamTaskOpenAllUpstreamInput(pTask);  // open inputQ for all upstream tasks
 
-  (void) taosThreadMutexLock(&pTask->chkInfo.pActiveInfo->lock);
+  streamMutexLock(&pTask->chkInfo.pActiveInfo->lock);
   streamTaskClearActiveInfo(pTask->chkInfo.pActiveInfo);
   if (clearChkpReadyMsg) {
     streamClearChkptReadyMsg(pTask->chkInfo.pActiveInfo);
   }
-  (void) taosThreadMutexUnlock(&pTask->chkInfo.pActiveInfo->lock);
+  streamMutexUnlock(&pTask->chkInfo.pActiveInfo->lock);
 }
 
 int32_t streamTaskUpdateTaskCheckpointInfo(SStreamTask* pTask, bool restored, SVUpdateCheckpointInfoReq* pReq) {
@@ -455,7 +461,7 @@ int32_t streamTaskUpdateTaskCheckpointInfo(SStreamTask* pTask, bool restored, SV
   const char*      id = pTask->id.idStr;
   SCheckpointInfo* pInfo = &pTask->chkInfo;
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
 
   if (pReq->checkpointId <= pInfo->checkpointId) {
     stDebug("s-task:%s vgId:%d latest checkpointId:%" PRId64 " checkpointVer:%" PRId64
@@ -463,7 +469,7 @@ int32_t streamTaskUpdateTaskCheckpointInfo(SStreamTask* pTask, bool restored, SV
             " transId:%d ignored",
             id, vgId, pInfo->checkpointId, pInfo->checkpointVer, pReq->checkpointId, pReq->checkpointVer,
             pReq->transId);
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     {  // destroy the related fill-history tasks
       // drop task should not in the meta-lock, and drop the related fill-history task now
@@ -525,13 +531,14 @@ int32_t streamTaskUpdateTaskCheckpointInfo(SStreamTask* pTask, bool restored, SV
   pTask->status.taskStatus = TASK_STATUS__READY;
 
   code = streamMetaSaveTask(pMeta, pTask);
+  streamMutexUnlock(&pTask->lock);
+
   if (code != TSDB_CODE_SUCCESS) {
     stError("s-task:%s vgId:%d failed to save task info after do checkpoint, checkpointId:%" PRId64 ", since %s", id,
             vgId, pReq->checkpointId, terrstr());
     return code;
   }
 
-  (void) taosThreadMutexUnlock(&pTask->lock);
   streamMetaWUnLock(pMeta);
 
   // drop task should not in the meta-lock, and drop the related fill-history task now
@@ -730,9 +737,9 @@ int32_t streamTaskBuildCheckpoint(SStreamTask* pTask) {
       code = streamSendChkptReportMsg(pTask, &pTask->chkInfo, dropRelHTask);
     }
   } else {  // clear the checkpoint info if failed
-    (void) taosThreadMutexLock(&pTask->lock);
+    streamMutexLock(&pTask->lock);
     streamTaskSetFailedCheckpointId(pTask);  // set failed checkpoint id before clear the checkpoint info
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     code = streamTaskHandleEvent(pTask->status.pSM, TASK_EVENT_CHECKPOINT_DONE);
     stDebug("s-task:%s clear checkpoint flag since gen checkpoint failed, checkpointId:%" PRId64, id, ckId);
@@ -763,20 +770,20 @@ void checkpointTriggerMonitorFn(void* param, void* tmrId) {
   }
 
   if (++pActiveInfo->checkCounter < 100) {
-    taosTmrReset(checkpointTriggerMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pChkptTriggerTmr);
+    streamTmrReset(checkpointTriggerMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pChkptTriggerTmr, vgId, "trigger-recv-monitor");
     return;
   }
 
   pActiveInfo->checkCounter = 0;
   stDebug("s-task:%s vgId:%d checkpoint-trigger monitor in tmr, ts:%" PRId64, id, vgId, now);
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   SStreamTaskState pState = streamTaskGetStatus(pTask);
   if (pState.state != TASK_STATUS__CK) {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s vgId:%d not in checkpoint status, quit from monitor checkpoint-trigger, ref:%d", id, vgId, ref);
 
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     streamMetaReleaseTask(pTask->pMeta, pTask);
     return;
   }
@@ -787,14 +794,14 @@ void checkpointTriggerMonitorFn(void* param, void* tmrId) {
     stDebug("s-task:%s vgId:%d all checkpoint-trigger recv, quit from monitor checkpoint-trigger, ref:%d", id, vgId,
             ref);
 
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     streamMetaReleaseTask(pTask->pMeta, pTask);
     return;
   }
 
-  (void) taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
-  (void) taosThreadMutexLock(&pActiveInfo->lock);
+  streamMutexLock(&pActiveInfo->lock);
 
   // send msg to retrieve checkpoint trigger msg
   SArray* pList = pTask->upstreamInfo.pList;
@@ -820,19 +827,19 @@ void checkpointTriggerMonitorFn(void* param, void* tmrId) {
 
     if (!recved) {  // make sure the inputQ is opened for not recv upstream checkpoint-trigger message
       streamTaskOpenUpstreamInput(pTask, pInfo->taskId);
-      taosArrayPush(pNotSendList, pInfo);
+      (void)taosArrayPush(pNotSendList, pInfo);
     }
   }
 
   // do send retrieve checkpoint trigger msg to upstream
   int32_t size = taosArrayGetSize(pNotSendList);
-  doSendRetrieveTriggerMsg(pTask, pNotSendList);
+  (void)doSendRetrieveTriggerMsg(pTask, pNotSendList);
-  (void) taosThreadMutexUnlock(&pActiveInfo->lock);
+  streamMutexUnlock(&pActiveInfo->lock);
 
   // check every 100ms
   if (size > 0) {
     stDebug("s-task:%s start to monitor checkpoint-trigger in 10s", id);
-    taosTmrReset(checkpointTriggerMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pChkptTriggerTmr);
+    streamTmrReset(checkpointTriggerMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pChkptTriggerTmr, vgId, "trigger-recv-monitor");
   } else {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s all checkpoint-trigger recved, quit from monitor checkpoint-trigger tmr, ref:%d", id, ref);
@@ -863,7 +870,7 @@ int32_t doSendRetrieveTriggerMsg(SStreamTask* pTask, SArray* pNotSendList) {
 
     SRetrieveChkptTriggerReq* pReq = rpcMallocCont(sizeof(SRetrieveChkptTriggerReq));
     if (pReq == NULL) {
-      terrno = TSDB_CODE_OUT_OF_MEMORY;
+      code = TSDB_CODE_OUT_OF_MEMORY;
       stError("vgId:%d failed to create msg to retrieve trigger msg for task:%s exec, code:out of memory", vgId, pId);
       continue;
     }
@@ -880,11 +887,16 @@ int32_t doSendRetrieveTriggerMsg(SStreamTask* pTask, SArray* pNotSendList) {
     initRpcMsg(&rpcMsg, TDMT_STREAM_RETRIEVE_TRIGGER, pReq, sizeof(SRetrieveChkptTriggerReq));
 
     code = tmsgSendReq(&pUpstreamTask->epSet, &rpcMsg);
-    stDebug("s-task:%s vgId:%d send checkpoint-trigger retrieve msg to 0x%x(vgId:%d) checkpointId:%" PRId64, pId, vgId,
+    if (code == TSDB_CODE_SUCCESS) {
-            pUpstreamTask->taskId, pUpstreamTask->nodeId, checkpointId);
+      stDebug("s-task:%s vgId:%d send checkpoint-trigger retrieve msg to 0x%x(vgId:%d) checkpointId:%" PRId64, pId,
+              vgId, pUpstreamTask->taskId, pUpstreamTask->nodeId, checkpointId);
+    } else {
+      stError("s-task:%s vgId:%d failed to send checkpoint-trigger retrieve msg to 0x%x(vgId:%d) checkpointId:%" PRId64,
+              pId, vgId, pUpstreamTask->taskId, pUpstreamTask->nodeId, checkpointId);
+    }
   }
 
-  return TSDB_CODE_SUCCESS;
+  return code;
 }
 
 bool streamTaskAlreadySendTrigger(SStreamTask* pTask, int32_t downstreamNodeId) {
@@ -897,9 +909,9 @@ bool streamTaskAlreadySendTrigger(SStreamTask* pTask, int32_t downstreamNodeId)
     return false;
   }
 
-  (void) taosThreadMutexLock(&pInfo->lock);
+  streamMutexLock(&pInfo->lock);
   if (!pInfo->dispatchTrigger) {
-    (void) taosThreadMutexUnlock(&pInfo->lock);
+    streamMutexUnlock(&pTask->lock);
     return false;
   }
 
@@ -921,7 +933,7 @@ bool streamTaskAlreadySendTrigger(SStreamTask* pTask, int32_t downstreamNodeId)
              id, pSendInfo->sendTs, before, pInfo->activeId, pInfo->transId);
     }
 
-    (void) taosThreadMutexUnlock(&pInfo->lock);
+    streamMutexUnlock(&pTask->lock);
     return true;
   }
 
@@ -945,7 +957,7 @@ void streamTaskInitTriggerDispatchInfo(SStreamTask* pTask) {
   SActiveCheckpointInfo* pInfo = pTask->chkInfo.pActiveInfo;
 
   int64_t now = taosGetTimestampMs();
-  (void) taosThreadMutexLock(&pInfo->lock);
+  streamMutexLock(&pInfo->lock);
 
   // outputQ should be empty here
   ASSERT(streamQueueGetNumOfUnAccessedItems(pTask->outputq.queue) == 0);
@@ -971,21 +983,21 @@ void streamTaskInitTriggerDispatchInfo(SStreamTask* pTask) {
     }
   }
 
-  (void) taosThreadMutexUnlock(&pInfo->lock);
+  streamMutexUnlock(&pInfo->lock);
 }
 
 int32_t streamTaskGetNumOfConfirmed(SStreamTask* pTask) {
   SActiveCheckpointInfo* pInfo = pTask->chkInfo.pActiveInfo;
 
   int32_t num = 0;
-  (void) taosThreadMutexLock(&pInfo->lock);
+  streamMutexLock(&pInfo->lock);
   for (int32_t i = 0; i < taosArrayGetSize(pInfo->pDispatchTriggerList); ++i) {
     STaskTriggerSendInfo* p = taosArrayGet(pInfo->pDispatchTriggerList, i);
     if (p->recved) {
       num++;
     }
   }
-  (void) taosThreadMutexUnlock(&pInfo->lock);
+  streamMutexUnlock(&pInfo->lock);
   return num;
 }
 
@@ -993,7 +1005,7 @@ void streamTaskSetTriggerDispatchConfirmed(SStreamTask* pTask, int32_t vgId) {
   SActiveCheckpointInfo* pInfo = pTask->chkInfo.pActiveInfo;
 
   int32_t taskId = 0;
-  (void) taosThreadMutexLock(&pInfo->lock);
+  streamMutexLock(&pInfo->lock);
 
   for (int32_t i = 0; i < taosArrayGetSize(pInfo->pDispatchTriggerList); ++i) {
     STaskTriggerSendInfo* p = taosArrayGet(pInfo->pDispatchTriggerList, i);
@@ -1007,7 +1019,7 @@ void streamTaskSetTriggerDispatchConfirmed(SStreamTask* pTask, int32_t vgId) {
     }
   }
 
-  (void) taosThreadMutexUnlock(&pInfo->lock);
+  streamMutexUnlock(&pInfo->lock);
 
   int32_t numOfConfirmed = streamTaskGetNumOfConfirmed(pTask);
   int32_t total = streamTaskGetNumOfDownstream(pTask);
@@ -1184,16 +1196,16 @@ int32_t deleteCheckpointFile(const char* id, const char* name) {
 int32_t streamTaskSendRestoreChkptMsg(SStreamTask* pTask) {
   const char* id = pTask->id.idStr;
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   if (pTask->status.sendConsensusChkptId == true) {
     stDebug("s-task:%s already start to consensus-checkpointId, not start again before it completed", id);
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     return TSDB_CODE_SUCCESS;
   } else {
     pTask->status.sendConsensusChkptId = true;
   }
 
-  (void) taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   ASSERT(pTask->pBackend == NULL);
   pTask->status.requireConsensusChkptId = true;
@@ -1207,12 +1219,12 @@ int32_t streamTaskSendCheckpointsourceRsp(SStreamTask* pTask) {
     return code;
   }
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   SStreamTaskState p = streamTaskGetStatus(pTask);
   if (p.state == TASK_STATUS__CK) {
     code = streamTaskSendCheckpointSourceRsp(pTask);
   }
-  (void) taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   return code;
 }

source/libs/stream/src/streamData.c

@@ -49,7 +49,7 @@ int32_t createStreamBlockFromDispatchMsg(const SStreamDispatchReq* pReq, int32_t
       pInput = p;
     }
 
-    blockDecode(pDataBlock, pInput);
+    (void) blockDecode(pDataBlock, pInput);
 
     if (pRetrieve->compressed && compLen < fullLen) {
       taosMemoryFree(pInput);
@@ -116,11 +116,11 @@ int32_t streamRetrieveReqToData(const SStreamRetrieveReq* pReq, SStreamDataBlock
     return terrno;
   }
 
-  taosArrayPush(pArray, &(SSDataBlock){0});
+  (void) taosArrayPush(pArray, &(SSDataBlock){0});
   SRetrieveTableRsp* pRetrieve = pReq->pRetrieve;
   SSDataBlock*       pDataBlock = taosArrayGet(pArray, 0);
 
-  blockDecode(pDataBlock, pRetrieve->data + PAYLOAD_PREFIX_LEN);
+  (void) blockDecode(pDataBlock, pRetrieve->data + PAYLOAD_PREFIX_LEN);
 
   // TODO: refactor
   pDataBlock->info.window.skey = be64toh(pRetrieve->skey);
@@ -156,28 +156,31 @@ void streamDataSubmitDestroy(SStreamDataSubmit* pDataSubmit) {
   taosFreeQitem(pDataSubmit);
 }
 
-SStreamMergedSubmit* streamMergedSubmitNew() {
+int32_t streamMergedSubmitNew(SStreamMergedSubmit** pSubmit) {
-  SStreamMergedSubmit* pMerged;
+  *pSubmit = NULL;
 
-  int32_t code = taosAllocateQitem(sizeof(SStreamMergedSubmit), DEF_QITEM, 0, (void**)&pMerged);
+  int32_t code = taosAllocateQitem(sizeof(SStreamMergedSubmit), DEF_QITEM, 0, (void**)pSubmit);
   if (code) {
-    terrno = code;
+    return TSDB_CODE_OUT_OF_MEMORY;
-    return NULL;
   }
 
-  pMerged->submits = taosArrayInit(0, sizeof(SPackedData));
+  (*pSubmit)->submits = taosArrayInit(0, sizeof(SPackedData));
-  if (pMerged->submits == NULL) {
+  if ((*pSubmit)->submits == NULL) {
-    taosArrayDestroy(pMerged->submits);
+    taosFreeQitem(*pSubmit);
-    taosFreeQitem(pMerged);
+    *pSubmit = NULL;
-    return NULL;
+    return TSDB_CODE_OUT_OF_MEMORY;
   }
 
-  pMerged->type = STREAM_INPUT__MERGED_SUBMIT;
+  (*pSubmit)->type = STREAM_INPUT__MERGED_SUBMIT;
-  return pMerged;
+  return TSDB_CODE_SUCCESS;
 }
 
 int32_t streamMergeSubmit(SStreamMergedSubmit* pMerged, SStreamDataSubmit* pSubmit) {
-  taosArrayPush(pMerged->submits, &pSubmit->submit);
+  void* p = taosArrayPush(pMerged->submits, &pSubmit->submit);
+  if (p == NULL) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
   if (pSubmit->ver > pMerged->ver) {
     pMerged->ver = pSubmit->ver;
   }
@@ -187,11 +190,12 @@ int32_t streamMergeSubmit(SStreamMergedSubmit* pMerged, SStreamDataSubmit* pSubm
 // todo handle memory error
 int32_t streamQueueMergeQueueItem(SStreamQueueItem* dst, SStreamQueueItem* pElem, SStreamQueueItem** pRes) {
   *pRes = NULL;
+  int32_t code = 0;
 
   if (dst->type == STREAM_INPUT__DATA_BLOCK && pElem->type == STREAM_INPUT__DATA_BLOCK) {
     SStreamDataBlock* pBlock = (SStreamDataBlock*)dst;
     SStreamDataBlock* pBlockSrc = (SStreamDataBlock*)pElem;
-    taosArrayAddAll(pBlock->blocks, pBlockSrc->blocks);
+    (void) taosArrayAddAll(pBlock->blocks, pBlockSrc->blocks);
     taosArrayDestroy(pBlockSrc->blocks);
     streamQueueItemIncSize(dst, streamQueueItemGetSize(pElem));
 
@@ -201,33 +205,38 @@ int32_t streamQueueMergeQueueItem(SStreamQueueItem* dst, SStreamQueueItem* pElem
   } else if (dst->type == STREAM_INPUT__MERGED_SUBMIT && pElem->type == STREAM_INPUT__DATA_SUBMIT) {
     SStreamMergedSubmit* pMerged = (SStreamMergedSubmit*)dst;
     SStreamDataSubmit*   pBlockSrc = (SStreamDataSubmit*)pElem;
-    streamMergeSubmit(pMerged, pBlockSrc);
+
+    code = streamMergeSubmit(pMerged, pBlockSrc);
     streamQueueItemIncSize(dst, streamQueueItemGetSize(pElem));
 
     taosFreeQitem(pElem);
     *pRes = dst;
     *pRes = dst;
-    return TSDB_CODE_SUCCESS;
+    return code;
   } else if (dst->type == STREAM_INPUT__DATA_SUBMIT && pElem->type == STREAM_INPUT__DATA_SUBMIT) {
-    SStreamMergedSubmit* pMerged = streamMergedSubmitNew();
+    SStreamMergedSubmit* pMerged = NULL;
-    if (pMerged == NULL) {
+    code = streamMergedSubmitNew(&pMerged);
-      return TSDB_CODE_OUT_OF_MEMORY;
+    if (code != 0) {
+      return code;
     }
 
     streamQueueItemIncSize((SStreamQueueItem*)pMerged, streamQueueItemGetSize(pElem));
 
-    streamMergeSubmit(pMerged, (SStreamDataSubmit*)dst);
+    code = streamMergeSubmit(pMerged, (SStreamDataSubmit*)dst);
-    streamMergeSubmit(pMerged, (SStreamDataSubmit*)pElem);
+    if (code == 0) {
+      code = streamMergeSubmit(pMerged, (SStreamDataSubmit*)pElem);
+    }
 
     taosFreeQitem(dst);
     taosFreeQitem(pElem);
 
     *pRes = (SStreamQueueItem*)pMerged;
-    return TSDB_CODE_SUCCESS;
+    return code;
   } else {
+    code = TSDB_CODE_FAILED;
     stDebug("block type:%s not merged with existed blocks list, type:%d", streamQueueItemGetTypeStr(pElem->type),
             dst->type);
-    return TSDB_CODE_FAILED;
+    return code;
   }
 }
 

source/libs/stream/src/streamDispatch.c

@@ -121,7 +121,7 @@ int32_t streamTaskBroadcastRetrieveReq(SStreamTask* pTask, SStreamRetrieveReq* r
     void*    abuf = POINTER_SHIFT(buf, sizeof(SMsgHead));
     SEncoder encoder;
     tEncoderInit(&encoder, abuf, len);
-    tEncodeStreamRetrieveReq(&encoder, req);
+    (void) tEncodeStreamRetrieveReq(&encoder, req);
     tEncoderClear(&encoder);
 
     SRpcMsg rpcMsg = {0};
@@ -189,6 +189,7 @@ int32_t streamBroadcastToUpTasks(SStreamTask* pTask, const SSDataBlock* pBlock)
   return code;
 }
 
+// no need to do anything if failed
 int32_t streamSendCheckMsg(SStreamTask* pTask, const SStreamTaskCheckReq* pReq, int32_t nodeId, SEpSet* pEpSet) {
   void*   buf = NULL;
   int32_t code = -1;
@@ -202,7 +203,7 @@ int32_t streamSendCheckMsg(SStreamTask* pTask, const SStreamTaskCheckReq* pReq,
 
   buf = rpcMallocCont(sizeof(SMsgHead) + tlen);
   if (buf == NULL) {
-    return -1;
+    return TSDB_CODE_OUT_OF_MEMORY;
   }
 
   ((SMsgHead*)buf)->vgId = htonl(nodeId);
@@ -220,8 +221,7 @@ int32_t streamSendCheckMsg(SStreamTask* pTask, const SStreamTaskCheckReq* pReq,
   stDebug("s-task:%s (level:%d) send check msg to s-task:0x%" PRIx64 ":0x%x (vgId:%d)", pTask->id.idStr,
           pTask->info.taskLevel, pReq->streamId, pReq->downstreamTaskId, nodeId);
 
-  tmsgSendReq(pEpSet, &msg);
+  return tmsgSendReq(pEpSet, &msg);
-  return 0;
 }
 
 void destroyDispatchMsg(SStreamDispatchReq* pReq, int32_t numOfVgroups) {
@@ -239,7 +239,7 @@ void clearBufferedDispatchMsg(SStreamTask* pTask) {
     destroyDispatchMsg(pMsgInfo->pData, streamTaskGetNumOfDownstream(pTask));
   }
 
-  taosThreadMutexLock(&pMsgInfo->lock);
+  streamMutexLock(&pMsgInfo->lock);
 
   pMsgInfo->checkpointId = -1;
   pMsgInfo->transId = -1;
@@ -249,7 +249,7 @@ void clearBufferedDispatchMsg(SStreamTask* pTask) {
   clearDispatchInfo(pMsgInfo);
 
   taosArrayClear(pTask->msgInfo.pSendInfo);
-  taosThreadMutexUnlock(&pMsgInfo->lock);
+  streamMutexUnlock(&pMsgInfo->lock);
 }
 
 static SStreamDispatchReq* createDispatchDataReq(SStreamTask* pTask, const SStreamDataBlock* pData) {
@@ -424,9 +424,9 @@ static int32_t sendDispatchMsg(SStreamTask* pTask, SStreamDispatchReq* pDispatch
 }
 
 static void setNotInDispatchMonitor(SDispatchMsgInfo* pMsgInfo) {
-  taosThreadMutexLock(&pMsgInfo->lock);
+  streamMutexLock(&pMsgInfo->lock);
   pMsgInfo->inMonitor = 0;
-  taosThreadMutexUnlock(&pMsgInfo->lock);
+  streamMutexUnlock(&pMsgInfo->lock);
 }
 
 static void setResendInfo(SDispatchEntry* pEntry, int64_t now) {
@@ -440,13 +440,13 @@ static void addDispatchEntry(SDispatchMsgInfo* pMsgInfo, int32_t nodeId, int64_t
   SDispatchEntry entry = {.nodeId = nodeId, .rspTs = -1, .status = 0, .sendTs = now};
 
   if (lock) {
-    taosThreadMutexLock(&pMsgInfo->lock);
+    streamMutexLock(&pMsgInfo->lock);
   }
 
-  taosArrayPush(pMsgInfo->pSendInfo, &entry);
+  (void) taosArrayPush(pMsgInfo->pSendInfo, &entry);
 
   if (lock) {
-    taosThreadMutexUnlock(&pMsgInfo->lock);
+    streamMutexUnlock(&pMsgInfo->lock);
   }
 }
 
@@ -496,16 +496,16 @@ static void doMonitorDispatchData(void* param, void* tmrId) {
     return;
   }
 
-  taosThreadMutexLock(&pMsgInfo->lock);
+  streamMutexLock(&pMsgInfo->lock);
   if (pTask->outputq.status == TASK_OUTPUT_STATUS__NORMAL) {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s not in dispatch procedure, abort from timer, ref:%d", pTask->id.idStr, ref);
 
     pTask->msgInfo.inMonitor = 0;
-    taosThreadMutexUnlock(&pMsgInfo->lock);
+    streamMutexUnlock(&pMsgInfo->lock);
     return;
   }
-  taosThreadMutexUnlock(&pMsgInfo->lock);
+  streamMutexUnlock(&pMsgInfo->lock);
 
   int32_t numOfFailed = getFailedDispatchInfo(pMsgInfo, now);
   if (numOfFailed == 0) {
@@ -577,8 +577,9 @@ static void doMonitorDispatchData(void* param, void* tmrId) {
 }
 
 void streamStartMonitorDispatchData(SStreamTask* pTask, int64_t waitDuration) {
+  int32_t vgId = pTask->pMeta->vgId;
   if (pTask->msgInfo.pRetryTmr != NULL) {
-    taosTmrReset(doMonitorDispatchData, waitDuration, pTask, streamTimer, &pTask->msgInfo.pRetryTmr);
+    streamTmrReset(doMonitorDispatchData, waitDuration, pTask, streamTimer, &pTask->msgInfo.pRetryTmr, vgId, "dispatch-monitor-tmr");
   } else {
     pTask->msgInfo.pRetryTmr = taosTmrStart(doMonitorDispatchData, waitDuration, pTask, streamTimer);
   }
@@ -612,7 +613,7 @@ int32_t streamSearchAndAddBlock(SStreamTask* pTask, SStreamDispatchReq* pReqs, S
         }
       }
     } else {
-      buildCtbNameByGroupIdImpl(pTask->outputInfo.shuffleDispatcher.stbFullName, groupId, pDataBlock->info.parTbName);
+      (void) buildCtbNameByGroupIdImpl(pTask->outputInfo.shuffleDispatcher.stbFullName, groupId, pDataBlock->info.parTbName);
     }
 
     snprintf(ctbName, TSDB_TABLE_NAME_LEN, "%s.%s", pTask->outputInfo.shuffleDispatcher.dbInfo.db,
@@ -624,8 +625,10 @@ int32_t streamSearchAndAddBlock(SStreamTask* pTask, SStreamDispatchReq* pReqs, S
     SBlockName bln = {0};
     bln.hashValue = hashValue;
     memcpy(bln.parTbName, pDataBlock->info.parTbName, strlen(pDataBlock->info.parTbName));
+
+    // failed to put into name buffer, no need to do anything
     if (tSimpleHashGetSize(pTask->pNameMap) < MAX_BLOCK_NAME_NUM) {
-      tSimpleHashPut(pTask->pNameMap, &groupId, sizeof(int64_t), &bln, sizeof(SBlockName));
+      (void) tSimpleHashPut(pTask->pNameMap, &groupId, sizeof(int64_t), &bln, sizeof(SBlockName));
     }
   }
 
@@ -633,14 +636,14 @@ int32_t streamSearchAndAddBlock(SStreamTask* pTask, SStreamDispatchReq* pReqs, S
   int32_t numOfVgroups = taosArrayGetSize(vgInfo);
 
   // TODO: optimize search
-  taosThreadMutexLock(&pTask->msgInfo.lock);
+  streamMutexLock(&pTask->msgInfo.lock);
 
   for (int32_t j = 0; j < numOfVgroups; j++) {
     SVgroupInfo* pVgInfo = taosArrayGet(vgInfo, j);
 
     if (hashValue >= pVgInfo->hashBegin && hashValue <= pVgInfo->hashEnd) {
       if (streamAddBlockIntoDispatchMsg(pDataBlock, &pReqs[j]) < 0) {
-        taosThreadMutexUnlock(&pTask->msgInfo.lock);
+        streamMutexUnlock(&pTask->msgInfo.lock);
         return -1;
       }
 
@@ -655,7 +658,7 @@ int32_t streamSearchAndAddBlock(SStreamTask* pTask, SStreamDispatchReq* pReqs, S
     }
   }
 
-  taosThreadMutexUnlock(&pTask->msgInfo.lock);
+  streamMutexUnlock(&pTask->msgInfo.lock);
   ASSERT(found);
   return 0;
 }
@@ -690,7 +693,8 @@ int32_t streamDispatchStreamBlock(SStreamTask* pTask) {
   ASSERT(pTask->msgInfo.pData == NULL);
   stDebug("s-task:%s start to dispatch msg, set output status:%d", id, pTask->outputq.status);
 
-  SStreamDataBlock* pBlock = streamQueueNextItem(pTask->outputq.queue);
+  SStreamDataBlock* pBlock = NULL;
+  streamQueueNextItem(pTask->outputq.queue, (SStreamQueueItem**)&pBlock);
   if (pBlock == NULL) {
     atomic_store_8(&pTask->outputq.status, TASK_OUTPUT_STATUS__NORMAL);
     stDebug("s-task:%s not dispatch since no elems in outputQ, output status:%d", id, pTask->outputq.status);
@@ -703,9 +707,9 @@ int32_t streamDispatchStreamBlock(SStreamTask* pTask) {
 
   pTask->execInfo.dispatch += 1;
 
-  taosThreadMutexLock(&pTask->msgInfo.lock);
+  streamMutexLock(&pTask->msgInfo.lock);
   initDispatchInfo(&pTask->msgInfo, pTask->execInfo.dispatch);
-  taosThreadMutexUnlock(&pTask->msgInfo.lock);
+  streamMutexUnlock(&pTask->msgInfo.lock);
 
   int32_t code = doBuildDispatchMsg(pTask, pBlock);
   if (code == 0) {
@@ -719,7 +723,7 @@ int32_t streamDispatchStreamBlock(SStreamTask* pTask) {
 
   code = sendDispatchMsg(pTask, pTask->msgInfo.pData);
 
-  taosThreadMutexLock(&pTask->msgInfo.lock);
+  streamMutexLock(&pTask->msgInfo.lock);
   if (pTask->msgInfo.inMonitor == 0) {
     int32_t ref = atomic_add_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s start dispatch monitor tmr in %dms, ref:%d, dispatch code:%s", id, DISPATCH_RETRY_INTERVAL_MS,
@@ -730,7 +734,7 @@ int32_t streamDispatchStreamBlock(SStreamTask* pTask) {
     stDebug("s-task:%s already in dispatch monitor tmr", id);
   }
 
-  taosThreadMutexUnlock(&pTask->msgInfo.lock);
+  streamMutexUnlock(&pTask->msgInfo.lock);
 
   // this block can not be deleted until it has been sent to downstream task successfully.
   return TSDB_CODE_SUCCESS;
@@ -758,7 +762,7 @@ int32_t initCheckpointReadyMsg(SStreamTask* pTask, int32_t upstreamNodeId, int32
 
   buf = rpcMallocCont(sizeof(SMsgHead) + tlen);
   if (buf == NULL) {
-    return -1;
+    return TSDB_CODE_OUT_OF_MEMORY;
   }
 
   ((SMsgHead*)buf)->vgId = htonl(req.upstreamNodeId);
@@ -791,33 +795,33 @@ static void checkpointReadyMsgSendMonitorFn(void* param, void* tmrId) {
 
   SActiveCheckpointInfo* pActiveInfo = pTask->chkInfo.pActiveInfo;
   if (++pActiveInfo->sendReadyCheckCounter < 100) {
-    taosTmrReset(checkpointReadyMsgSendMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pSendReadyMsgTmr);
+    streamTmrReset(checkpointReadyMsgSendMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pSendReadyMsgTmr, vgId, "chkpt-ready-monitor");
     return;
   }
 
   pActiveInfo->sendReadyCheckCounter = 0;
   stDebug("s-task:%s in sending checkpoint-ready msg monitor timer", id);
 
-  taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   SStreamTaskState pState = streamTaskGetStatus(pTask);
   if (pState.state != TASK_STATUS__CK) {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s vgId:%d status:%s not in checkpoint, quit from monitor checkpoint-ready send, ref:%d", id, vgId,
             pState.name, ref);
-    taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     streamMetaReleaseTask(pTask->pMeta, pTask);
     return;
   }
-  taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
-  taosThreadMutexLock(&pActiveInfo->lock);
+  streamMutexLock(&pActiveInfo->lock);
 
   SArray* pList = pActiveInfo->pReadyMsgList;
   int32_t num = taosArrayGetSize(pList);
 
   // active checkpoint info is cleared for now
   if ((pActiveInfo->activeId == 0) && (pActiveInfo->transId == 0) && (num == 0) && (pTask->chkInfo.startTs == 0)) {
-    taosThreadMutexUnlock(&pActiveInfo->lock);
+    streamMutexUnlock(&pActiveInfo->lock);
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stWarn("s-task:%s vgId:%d active checkpoint may be cleared, quit from readyMsg send tmr, ref:%d", id, vgId, ref);
 
@@ -835,7 +839,7 @@ static void checkpointReadyMsgSendMonitorFn(void* param, void* tmrId) {
       continue;
     }
 
-    taosArrayPush(pNotRspList, &pInfo->upstreamTaskId);
+    (void) taosArrayPush(pNotRspList, &pInfo->upstreamTaskId);
     stDebug("s-task:%s vgId:%d level:%d checkpoint-ready rsp from upstream:0x%x not confirmed yet", id, vgId,
             pTask->info.taskLevel, pInfo->upstreamTaskId);
   }
@@ -852,17 +856,25 @@ static void checkpointReadyMsgSendMonitorFn(void* param, void* tmrId) {
         if (taskId == pReadyInfo->upstreamTaskId) {  // send msg again
 
           SRpcMsg msg = {0};
-          initCheckpointReadyMsg(pTask, pReadyInfo->upstreamNodeId, pReadyInfo->upstreamTaskId, pReadyInfo->childId,
+          int32_t code = initCheckpointReadyMsg(pTask, pReadyInfo->upstreamNodeId, pReadyInfo->upstreamTaskId, pReadyInfo->childId,
                                  checkpointId, &msg);
-          tmsgSendReq(&pReadyInfo->upstreamNodeEpset, &msg);
+          if (code == TSDB_CODE_SUCCESS) {
+            code = tmsgSendReq(&pReadyInfo->upstreamNodeEpset, &msg);
+            if (code == TSDB_CODE_SUCCESS) {
               stDebug("s-task:%s level:%d checkpoint-ready msg sent to upstream:0x%x again", id, pTask->info.taskLevel,
                       pReadyInfo->upstreamTaskId);
+            } else {
+              stError("s-task:%s failed to send checkpoint-ready msg, try nex time in 10s", id);
+            }
+          } else {
+            stError("s-task:%s failed to prepare the checkpoint-ready msg, try nex time in 10s", id);
+          }
         }
       }
     }
 
-    taosTmrReset(checkpointReadyMsgSendMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pSendReadyMsgTmr);
+    streamTmrReset(checkpointReadyMsgSendMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pSendReadyMsgTmr, vgId, "chkpt-ready-monitor");
-    taosThreadMutexUnlock(&pActiveInfo->lock);
+    streamMutexUnlock(&pActiveInfo->lock);
   } else {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug(
@@ -871,7 +883,7 @@ static void checkpointReadyMsgSendMonitorFn(void* param, void* tmrId) {
         id, vgId, ref);
 
     streamClearChkptReadyMsg(pActiveInfo);
-    taosThreadMutexUnlock(&pActiveInfo->lock);
+    streamMutexUnlock(&pActiveInfo->lock);
     streamMetaReleaseTask(pTask->pMeta, pTask);
   }
 
@@ -883,9 +895,10 @@ int32_t streamTaskSendCheckpointReadyMsg(SStreamTask* pTask) {
   SActiveCheckpointInfo* pActiveInfo = pTask->chkInfo.pActiveInfo;
 
   const char* id = pTask->id.idStr;
+  int32_t     vgId = pTask->pMeta->vgId;
   SArray*     pList = pActiveInfo->pReadyMsgList;
 
-  taosThreadMutexLock(&pActiveInfo->lock);
+  streamMutexLock(&pActiveInfo->lock);
 
   int32_t num = taosArrayGetSize(pList);
   ASSERT(taosArrayGetSize(pTask->upstreamInfo.pList) == num);
@@ -894,15 +907,22 @@ int32_t streamTaskSendCheckpointReadyMsg(SStreamTask* pTask) {
     STaskCheckpointReadyInfo* pInfo = taosArrayGet(pList, i);
 
     SRpcMsg msg = {0};
-    initCheckpointReadyMsg(pTask, pInfo->upstreamNodeId, pInfo->upstreamTaskId, pInfo->childId, pInfo->checkpointId,
+    int32_t code = initCheckpointReadyMsg(pTask, pInfo->upstreamNodeId, pInfo->upstreamTaskId, pInfo->childId, pInfo->checkpointId,
                            &msg);
-    tmsgSendReq(&pInfo->upstreamNodeEpset, &msg);
+    if (code == TSDB_CODE_SUCCESS) {
-
+      code = tmsgSendReq(&pInfo->upstreamNodeEpset, &msg);
+      if (code == TSDB_CODE_SUCCESS) {
         stDebug("s-task:%s level:%d checkpoint-ready msg sent to upstream:0x%x", id, pTask->info.taskLevel,
                 pInfo->upstreamTaskId);
+      } else {
+        stError("s-task:%s failed to send checkpoint-ready msg, try nex time in 10s", id);
+      }
+    } else {
+      stError("s-task:%s failed to prepare the checkpoint-ready msg, try nex time in 10s", id);
+    }
   }
 
-  taosThreadMutexUnlock(&pActiveInfo->lock);
+  streamMutexUnlock(&pActiveInfo->lock);
   stDebug("s-task:%s level:%d checkpoint-ready msg sent to all %d upstreams", id, pTask->info.taskLevel, num);
 
   // start to check if checkpoint ready msg has successfully received by upstream tasks.
@@ -914,7 +934,7 @@ int32_t streamTaskSendCheckpointReadyMsg(SStreamTask* pTask) {
     if (pActiveInfo->pSendReadyMsgTmr == NULL) {
       pActiveInfo->pSendReadyMsgTmr = taosTmrStart(checkpointReadyMsgSendMonitorFn, 100, pTask, streamTimer);
     } else {
-      taosTmrReset(checkpointReadyMsgSendMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pSendReadyMsgTmr);
+      streamTmrReset(checkpointReadyMsgSendMonitorFn, 100, pTask, streamTimer, &pActiveInfo->pSendReadyMsgTmr, vgId, "chkpt-ready-monitor");
     }
   }
 
@@ -925,7 +945,7 @@ int32_t streamTaskSendCheckpointReadyMsg(SStreamTask* pTask) {
 int32_t streamTaskSendCheckpointSourceRsp(SStreamTask* pTask) {
   SArray* pList = pTask->chkInfo.pActiveInfo->pReadyMsgList;
 
-  taosThreadMutexLock(&pTask->chkInfo.pActiveInfo->lock);
+  streamMutexLock(&pTask->chkInfo.pActiveInfo->lock);
   ASSERT(pTask->info.taskLevel == TASK_LEVEL__SOURCE);
 
   if (taosArrayGetSize(pList) == 1) {
@@ -940,7 +960,7 @@ int32_t streamTaskSendCheckpointSourceRsp(SStreamTask* pTask) {
             pTask->info.taskLevel);
   }
 
-  taosThreadMutexUnlock(&pTask->chkInfo.pActiveInfo->lock);
+  streamMutexUnlock(&pTask->chkInfo.pActiveInfo->lock);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -978,8 +998,8 @@ int32_t streamAddBlockIntoDispatchMsg(const SSDataBlock* pBlock, SStreamDispatch
 
   payloadLen += sizeof(SRetrieveTableRsp);
 
-  taosArrayPush(pReq->dataLen, &payloadLen);
+  (void) taosArrayPush(pReq->dataLen, &payloadLen);
-  taosArrayPush(pReq->data, &buf);
+  (void) taosArrayPush(pReq->data, &buf);
 
   pReq->totalLen += dataStrLen;
   return 0;
@@ -1056,7 +1076,7 @@ int32_t streamTaskBuildCheckpointSourceRsp(SStreamCheckpointSourceReq* pReq, SRp
   void* abuf = POINTER_SHIFT(pBuf, sizeof(SMsgHead));
 
   tEncoderInit(&encoder, (uint8_t*)abuf, len);
-  tEncodeStreamCheckpointSourceRsp(&encoder, &rsp);
+  (void) tEncodeStreamCheckpointSourceRsp(&encoder, &rsp);
   tEncoderClear(&encoder);
 
   initRpcMsg(pMsg, 0, pBuf, sizeof(SMsgHead) + len);
@@ -1071,10 +1091,11 @@ int32_t streamAddCheckpointSourceRspMsg(SStreamCheckpointSourceReq* pReq, SRpcHa
   STaskCheckpointReadyInfo info = {
       .recvTs = taosGetTimestampMs(), .transId = pReq->transId, .checkpointId = pReq->checkpointId};
 
-  streamTaskBuildCheckpointSourceRsp(pReq, pRpcInfo, &info.msg, TSDB_CODE_SUCCESS);
+  // todo retry until it success
+  (void) streamTaskBuildCheckpointSourceRsp(pReq, pRpcInfo, &info.msg, TSDB_CODE_SUCCESS);
 
   SActiveCheckpointInfo* pActiveInfo = pTask->chkInfo.pActiveInfo;
-  taosThreadMutexLock(&pActiveInfo->lock);
+  streamMutexLock(&pActiveInfo->lock);
 
   int32_t size = taosArrayGetSize(pActiveInfo->pReadyMsgList);
   if (size > 0) {
@@ -1091,15 +1112,15 @@ int32_t streamAddCheckpointSourceRspMsg(SStreamCheckpointSourceReq* pReq, SRpcHa
       ASSERT(0);  // failed to handle it
     }
   } else {
-    taosArrayPush(pActiveInfo->pReadyMsgList, &info);
+    (void) taosArrayPush(pActiveInfo->pReadyMsgList, &info);
     stDebug("s-task:%s add checkpoint source rsp msg, total:%d", pTask->id.idStr, size + 1);
   }
 
-  taosThreadMutexUnlock(&pActiveInfo->lock);
+  streamMutexUnlock(&pActiveInfo->lock);
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t initCheckpointReadyInfo(STaskCheckpointReadyInfo* pReadyInfo, int32_t upstreamNodeId, int32_t upstreamTaskId,
+void initCheckpointReadyInfo(STaskCheckpointReadyInfo* pReadyInfo, int32_t upstreamNodeId, int32_t upstreamTaskId,
                                 int32_t childId, SEpSet* pEpset, int64_t checkpointId) {
   ASSERT(upstreamTaskId != 0);
 
@@ -1109,8 +1130,6 @@ int32_t initCheckpointReadyInfo(STaskCheckpointReadyInfo* pReadyInfo, int32_t up
   pReadyInfo->recvTs = taosGetTimestampMs();
   pReadyInfo->checkpointId = checkpointId;
   pReadyInfo->childId = childId;
-
-  return TSDB_CODE_SUCCESS;
 }
 
 int32_t streamAddCheckpointReadyMsg(SStreamTask* pTask, int32_t upstreamTaskId, int32_t index, int64_t checkpointId) {
@@ -1119,9 +1138,9 @@ int32_t streamAddCheckpointReadyMsg(SStreamTask* pTask, int32_t upstreamTaskId,
   }
 
   SStreamUpstreamEpInfo* pInfo = NULL;
-  int32_t code = streamTaskGetUpstreamTaskEpInfo(pTask, upstreamTaskId, &pInfo);
+  streamTaskGetUpstreamTaskEpInfo(pTask, upstreamTaskId, &pInfo);
-  if (code != TSDB_CODE_SUCCESS) {
+  if (pInfo == NULL) {
-    return code;
+    return TSDB_CODE_STREAM_TASK_NOT_EXIST;
   }
 
   STaskCheckpointReadyInfo info = {0};
@@ -1132,8 +1151,8 @@ int32_t streamAddCheckpointReadyMsg(SStreamTask* pTask, int32_t upstreamTaskId,
 
   SActiveCheckpointInfo* pActiveInfo = pTask->chkInfo.pActiveInfo;
 
-  taosThreadMutexLock(&pActiveInfo->lock);
+  streamMutexLock(&pActiveInfo->lock);
-  taosArrayPush(pActiveInfo->pReadyMsgList, &info);
+  (void) taosArrayPush(pActiveInfo->pReadyMsgList, &info);
 
   int32_t numOfRecv = taosArrayGetSize(pActiveInfo->pReadyMsgList);
   int32_t total = streamTaskGetNumOfUpstream(pTask);
@@ -1145,7 +1164,7 @@ int32_t streamAddCheckpointReadyMsg(SStreamTask* pTask, int32_t upstreamTaskId,
     stDebug("s-task:%s %d/%d checkpoint-trigger recv", pTask->id.idStr, numOfRecv, total);
   }
 
-  taosThreadMutexUnlock(&pActiveInfo->lock);
+  streamMutexUnlock(&pActiveInfo->lock);
   return 0;
 }
 
@@ -1187,10 +1206,8 @@ static int32_t handleDispatchSuccessRsp(SStreamTask* pTask, int32_t downstreamId
   if (delayDispatch) {
     return 0;
   } else {
-    streamDispatchStreamBlock(pTask);
+    return streamDispatchStreamBlock(pTask);
   }
-
-  return 0;
 }
 
 static bool setDispatchRspInfo(SDispatchMsgInfo* pMsgInfo, int32_t vgId, int32_t code, int64_t now, int32_t* pNotRsp, const char* id) {
@@ -1200,7 +1217,7 @@ static bool setDispatchRspInfo(SDispatchMsgInfo* pMsgInfo, int32_t vgId, int32_t
   bool    allRsp = false;
   *pNotRsp = 0;
 
-  taosThreadMutexLock(&pMsgInfo->lock);
+  streamMutexLock(&pMsgInfo->lock);
   int32_t numOfDispatchBranch = taosArrayGetSize(pMsgInfo->pSendInfo);
 
   for(int32_t i = 0; i < numOfDispatchBranch; ++i) {
@@ -1228,7 +1245,7 @@ static bool setDispatchRspInfo(SDispatchMsgInfo* pMsgInfo, int32_t vgId, int32_t
   *pNotRsp = numOfDispatchBranch - numOfRsp;
   allRsp = (numOfRsp == numOfDispatchBranch);
 
-  taosThreadMutexUnlock(&pMsgInfo->lock);
+  streamMutexUnlock(&pMsgInfo->lock);
 
   ASSERT(updated);
   return allRsp;
@@ -1240,7 +1257,7 @@ bool isDispatchRspTimeout(SDispatchEntry* pEntry, int64_t now) {
 
 int32_t getFailedDispatchInfo(SDispatchMsgInfo* pMsgInfo, int64_t now) {
   int32_t numOfFailed = 0;
-  taosThreadMutexLock(&pMsgInfo->lock);
+  streamMutexLock(&pMsgInfo->lock);
 
   for (int32_t j = 0; j < taosArrayGetSize(pMsgInfo->pSendInfo); ++j) {
     SDispatchEntry* pEntry = taosArrayGet(pMsgInfo->pSendInfo, j);
@@ -1248,7 +1265,7 @@ int32_t getFailedDispatchInfo(SDispatchMsgInfo* pMsgInfo, int64_t now) {
       numOfFailed += 1;
     }
   }
-  taosThreadMutexUnlock(&pMsgInfo->lock);
+  streamMutexUnlock(&pMsgInfo->lock);
   return numOfFailed;
 }
 
@@ -1260,9 +1277,9 @@ int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, i
   bool              allRsp = false;
   int32_t           notRsp = 0;
 
-  taosThreadMutexLock(&pMsgInfo->lock);
+  streamMutexLock(&pMsgInfo->lock);
   int32_t           msgId = pMsgInfo->msgId;
-  taosThreadMutexUnlock(&pMsgInfo->lock);
+  streamMutexUnlock(&pMsgInfo->lock);
 
   // follower not handle the dispatch rsp
   if ((pTask->pMeta->role == NODE_ROLE_FOLLOWER) || (pTask->status.downstreamReady != 1)) {
@@ -1315,7 +1332,7 @@ int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, i
       {
         bool delayDispatch = (pMsgInfo->dispatchMsgType == STREAM_INPUT__CHECKPOINT_TRIGGER);
         if (delayDispatch) {
-          taosThreadMutexLock(&pTask->lock);
+          streamMutexLock(&pTask->lock);
           // we only set the dispatch msg info for current checkpoint trans
           if (streamTaskGetStatus(pTask).state == TASK_STATUS__CK &&
               pTask->chkInfo.pActiveInfo->activeId == pMsgInfo->checkpointId) {
@@ -1329,7 +1346,7 @@ int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, i
                    " transId:%d discard, since expired",
                    pTask->id.idStr, pMsgInfo->checkpointId, pMsgInfo->transId);
           }
-          taosThreadMutexUnlock(&pTask->lock);
+          streamMutexUnlock(&pTask->lock);
         }
       }
     }
@@ -1374,12 +1391,12 @@ int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, i
         // now ready for next data output
         atomic_store_8(&pTask->outputq.status, TASK_OUTPUT_STATUS__NORMAL);
       } else {
-        handleDispatchSuccessRsp(pTask, pRsp->downstreamTaskId, pRsp->downstreamNodeId);
+        code = handleDispatchSuccessRsp(pTask, pRsp->downstreamTaskId, pRsp->downstreamNodeId);
       }
     }
   }
 
-  return 0;
+  return code;
 }
 
 static int32_t buildDispatchRsp(const SStreamTask* pTask, const SStreamDispatchReq* pReq, int32_t status, void** pBuf) {
@@ -1438,9 +1455,9 @@ int32_t streamProcessDispatchMsg(SStreamTask* pTask, SStreamDispatchReq* pReq, S
           pReq->upstreamTaskId, pReq->upstreamNodeId, pReq->totalLen, pReq->msgId);
 
   SStreamUpstreamEpInfo* pInfo = NULL;
-  int32_t code = streamTaskGetUpstreamTaskEpInfo(pTask, pReq->upstreamTaskId, &pInfo);
+  streamTaskGetUpstreamTaskEpInfo(pTask, pReq->upstreamTaskId, &pInfo);
-  if (code != TSDB_CODE_SUCCESS) {
+  if (pInfo == NULL) {
-    return code;
+    return TSDB_CODE_STREAM_TASK_NOT_EXIST;
   }
 
   if (pMeta->role == NODE_ROLE_FOLLOWER) {
@@ -1474,7 +1491,7 @@ int32_t streamProcessDispatchMsg(SStreamTask* pTask, SStreamDispatchReq* pReq, S
 
   {
     // do send response with the input status
-    code = buildDispatchRsp(pTask, pReq, status, &pRsp->pCont);
+    int32_t code = buildDispatchRsp(pTask, pReq, status, &pRsp->pCont);
     if (code != TSDB_CODE_SUCCESS) {
       stError("s-task:%s failed to build dispatch rsp, msgId:%d, code:%s", id, pReq->msgId, tstrerror(code));
       return code;
@@ -1484,6 +1501,5 @@ int32_t streamProcessDispatchMsg(SStreamTask* pTask, SStreamDispatchReq* pReq, S
     tmsgSendRsp(pRsp);
   }
 
-  streamTrySchedExec(pTask);
+  return streamTrySchedExec(pTask);
-  return code;
 }

source/libs/stream/src/streamExec.c

@@ -24,7 +24,7 @@
 #define FILL_HISTORY_TASK_EXEC_INTERVAL   5000            // 5 sec
 
 static int32_t streamTransferStateDoPrepare(SStreamTask* pTask);
-static int32_t streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t* totalSize, int32_t* totalBlocks);
+static void streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t* totalSize, int32_t* totalBlocks);
 
 bool streamTaskShouldStop(const SStreamTask* pTask) {
   SStreamTaskState pState = streamTaskGetStatus(pTask);
@@ -52,8 +52,9 @@ static int32_t doOutputResultBlockImpl(SStreamTask* pTask, SStreamDataBlock* pBl
       return code;
     }
 
+    // not handle error, if dispatch failed, try next time.
     // checkpoint trigger will be checked
-    streamDispatchStreamBlock(pTask);
+    (void) streamDispatchStreamBlock(pTask);
   }
 
   return code;
@@ -90,7 +91,7 @@ static int32_t doDumpResult(SStreamTask* pTask, SStreamQueueItem* pItem, SArray*
   return code;
 }
 
-int32_t streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t* totalSize, int32_t* totalBlocks) {
+void streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t* totalSize, int32_t* totalBlocks) {
   int32_t code = TSDB_CODE_SUCCESS;
   void*   pExecutor = pTask->exec.pExecutor;
 
@@ -108,7 +109,7 @@ int32_t streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t*
 
     if (streamTaskShouldStop(pTask)) {
       taosArrayDestroyEx(pRes, (FDelete)blockDataFreeRes);
-      return 0;
+      return;
     }
 
     SSDataBlock* output = NULL;
@@ -128,11 +129,16 @@ int32_t streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t*
         const SStreamDataBlock* pRetrieveBlock = (const SStreamDataBlock*)pItem;
         ASSERT(taosArrayGetSize(pRetrieveBlock->blocks) == 1);
 
-        assignOneDataBlock(&block, taosArrayGet(pRetrieveBlock->blocks, 0));
+        (void) assignOneDataBlock(&block, taosArrayGet(pRetrieveBlock->blocks, 0));
         block.info.type = STREAM_PULL_OVER;
         block.info.childId = pTask->info.selfChildId;
-        taosArrayPush(pRes, &block);
+
+        void* p = taosArrayPush(pRes, &block);
+        if (p != NULL) {
           numOfBlocks += 1;
+        } else {
+          stError("s-task:%s failed to add retrieve block", pTask->id.idStr);
+        }
 
         stDebug("s-task:%s(child %d) retrieve process completed, reqId:0x%" PRIx64 " dump results", pTask->id.idStr,
                pTask->info.selfChildId, pRetrieveBlock->reqId);
@@ -151,13 +157,21 @@ int32_t streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t*
     }
 
     SSDataBlock block = {0};
-    assignOneDataBlock(&block, output);
+    code = assignOneDataBlock(&block, output);
+    if (code) {
+      stError("s-task:%s failed to build result block due to out of memory", pTask->id.idStr);
+      continue;
+    }
+
     block.info.childId = pTask->info.selfChildId;
 
     size += blockDataGetSize(output) + sizeof(SSDataBlock) + sizeof(SColumnInfoData) * blockDataGetNumOfCols(&block);
     numOfBlocks += 1;
 
-    taosArrayPush(pRes, &block);
+    void* p = taosArrayPush(pRes, &block);
+    if (p == NULL) {
+      stError("s-task:%s failed to add computing results, the final res may be incorrect", pTask->id.idStr);
+    }
 
     stDebug("s-task:%s (child %d) executed and get %d result blocks, size:%.2fMiB", pTask->id.idStr,
            pTask->info.selfChildId, numOfBlocks, SIZE_IN_MiB(size));
@@ -166,8 +180,9 @@ int32_t streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t*
     if (numOfBlocks >= STREAM_RESULT_DUMP_THRESHOLD || size >= STREAM_RESULT_DUMP_SIZE_THRESHOLD) {
       ASSERT(numOfBlocks == taosArrayGetSize(pRes));
       code = doDumpResult(pTask, pItem, pRes, size, totalSize, totalBlocks);
+      // todo: here we need continue retry to put it into output buffer
       if (code != TSDB_CODE_SUCCESS) {
-        return code;
+        return;
       }
 
       pRes = NULL;
@@ -182,16 +197,16 @@ int32_t streamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pItem, int64_t*
   } else {
     taosArrayDestroyEx(pRes, (FDelete)blockDataFreeRes);
   }
-
-  return code;
 }
 
-static int32_t handleSanhistoryResultBlocks(SStreamTask* pTask, SArray* pRes, int32_t size) {
+// todo contiuous try to create result blocks
+static int32_t handleScanhistoryResultBlocks(SStreamTask* pTask, SArray* pRes, int32_t size) {
   int32_t code = TSDB_CODE_SUCCESS;
   if (taosArrayGetSize(pRes) > 0) {
     SStreamDataBlock* pStreamBlocks = NULL;
     code = createStreamBlockFromResults(NULL, pTask, size, pRes, &pStreamBlocks);
     if (code) {
+      stError("s-task:%s failed to build history result blocks", pTask->id.idStr);
       return code;
     }
 
@@ -236,9 +251,13 @@ static void streamScanHistoryDataImpl(SStreamTask* pTask, SArray* pRes, int32_t*
     }
 
     SSDataBlock block = {0};
-    assignOneDataBlock(&block, output);
+    (void) assignOneDataBlock(&block, output);
     block.info.childId = pTask->info.selfChildId;
-    taosArrayPush(pRes, &block);
+
+    void* p = taosArrayPush(pRes, &block);
+    if (p == NULL) {
+      stError("s-task:%s failed to add computing results, the final res may be incorrect", pTask->id.idStr);
+    }
 
     (*pSize) += blockDataGetSize(output) + sizeof(SSDataBlock) + sizeof(SColumnInfoData) * blockDataGetNumOfCols(&block);
     numOfBlocks += 1;
@@ -264,7 +283,7 @@ SScanhistoryDataInfo streamScanHistoryData(SStreamTask* pTask, int64_t st) {
   const char* id = pTask->id.idStr;
 
   if (!pTask->hTaskInfo.operatorOpen) {
-    qSetStreamOpOpen(exec);
+    (void) qSetStreamOpOpen(exec);
     pTask->hTaskInfo.operatorOpen = true;
   }
 
@@ -301,8 +320,8 @@ SScanhistoryDataInfo streamScanHistoryData(SStreamTask* pTask, int64_t st) {
       return buildScanhistoryExecRet(TASK_SCANHISTORY_QUIT, 0);
     }
 
-    // dispatch the generated results
+    // dispatch the generated results, todo fix error
-    /*int32_t code = */handleSanhistoryResultBlocks(pTask, pRes, size);
+    (void) handleScanhistoryResultBlocks(pTask, pRes, size);
 
     if (finished) {
       return buildScanhistoryExecRet(TASK_SCANHISTORY_CONT, 0);
@@ -323,14 +342,15 @@ int32_t streamTransferStateDoPrepare(SStreamTask* pTask) {
 
   SStreamTask* pStreamTask = NULL;
   int32_t code = streamMetaAcquireTask(pMeta, pTask->streamTaskId.streamId, pTask->streamTaskId.taskId, &pStreamTask);
-  if (pStreamTask == NULL) {
+  if (pStreamTask == NULL || code != TSDB_CODE_SUCCESS) {
     stError(
         "s-task:%s failed to find related stream task:0x%x, it may have been destroyed or closed, destroy the related "
         "fill-history task",
         id, (int32_t) pTask->streamTaskId.taskId);
 
     // 1. free it and remove fill-history task from disk meta-store
-    streamBuildAndSendDropTaskMsg(pTask->pMsgCb, pMeta->vgId, &pTask->id, 0);
+    // todo: this function should never be failed.
+    (void) streamBuildAndSendDropTaskMsg(pTask->pMsgCb, pMeta->vgId, &pTask->id, 0);
 
     // 2. save to disk
     streamMetaWLock(pMeta);
@@ -388,14 +408,14 @@ int32_t streamTransferStateDoPrepare(SStreamTask* pTask) {
             pStreamTask->id.idStr, TASK_LEVEL__SOURCE, pTimeWindow->skey, pTimeWindow->ekey, INT64_MIN,
             pTimeWindow->ekey, p, pStreamTask->status.schedStatus);
 
-    streamTaskResetTimewindowFilter(pStreamTask);
+    (void) streamTaskResetTimewindowFilter(pStreamTask);
   } else {
     stDebug("s-task:%s no need to update/reset filter time window for non-source tasks", pStreamTask->id.idStr);
   }
 
   // NOTE: transfer the ownership of executor state before handle the checkpoint block during stream exec
   // 2. send msg to mnode to launch a checkpoint to keep the state for current stream
-  streamTaskSendCheckpointReq(pStreamTask);
+  (void) streamTaskSendCheckpointReq(pStreamTask);
 
   // 3. assign the status to the value that will be kept in disk
   pStreamTask->status.taskStatus = streamTaskGetStatus(pStreamTask).state;
@@ -409,7 +429,7 @@ int32_t streamTransferStateDoPrepare(SStreamTask* pTask) {
 
 static int32_t haltCallback(SStreamTask* pTask, void* param) {
   streamTaskOpenAllUpstreamInput(pTask);
-  streamTaskSendCheckpointReq(pTask);
+  (void) streamTaskSendCheckpointReq(pTask);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -445,18 +465,19 @@ int32_t streamTransferStatePrepare(SStreamTask* pTask) {
 }
 
 // set input
-static void doSetStreamInputBlock(SStreamTask* pTask, const void* pInput, int64_t* pVer, const char* id) {
+static int32_t doSetStreamInputBlock(SStreamTask* pTask, const void* pInput, int64_t* pVer, const char* id) {
   void* pExecutor = pTask->exec.pExecutor;
+  int32_t code = 0;
 
   const SStreamQueueItem* pItem = pInput;
   if (pItem->type == STREAM_INPUT__GET_RES) {
     const SStreamTrigger* pTrigger = (const SStreamTrigger*)pInput;
-    qSetMultiStreamInput(pExecutor, pTrigger->pBlock, 1, STREAM_INPUT__DATA_BLOCK);
+    code = qSetMultiStreamInput(pExecutor, pTrigger->pBlock, 1, STREAM_INPUT__DATA_BLOCK);
 
   } else if (pItem->type == STREAM_INPUT__DATA_SUBMIT) {
     ASSERT(pTask->info.taskLevel == TASK_LEVEL__SOURCE);
     const SStreamDataSubmit* pSubmit = (const SStreamDataSubmit*)pInput;
-    qSetMultiStreamInput(pExecutor, &pSubmit->submit, 1, STREAM_INPUT__DATA_SUBMIT);
+    code = qSetMultiStreamInput(pExecutor, &pSubmit->submit, 1, STREAM_INPUT__DATA_SUBMIT);
     stDebug("s-task:%s set submit blocks as source block completed, %p %p len:%d ver:%" PRId64, id, pSubmit,
            pSubmit->submit.msgStr, pSubmit->submit.msgLen, pSubmit->submit.ver);
     ASSERT((*pVer) <= pSubmit->submit.ver);
@@ -468,7 +489,7 @@ static void doSetStreamInputBlock(SStreamTask* pTask, const void* pInput, int64_
     SArray* pBlockList = pBlock->blocks;
     int32_t numOfBlocks = taosArrayGetSize(pBlockList);
     stDebug("s-task:%s set sdata blocks as input num:%d, ver:%" PRId64, id, numOfBlocks, pBlock->sourceVer);
-    qSetMultiStreamInput(pExecutor, pBlockList->pData, numOfBlocks, STREAM_INPUT__DATA_BLOCK);
+    code = qSetMultiStreamInput(pExecutor, pBlockList->pData, numOfBlocks, STREAM_INPUT__DATA_BLOCK);
 
   } else if (pItem->type == STREAM_INPUT__MERGED_SUBMIT) {
     const SStreamMergedSubmit* pMerged = (const SStreamMergedSubmit*)pInput;
@@ -477,24 +498,26 @@ static void doSetStreamInputBlock(SStreamTask* pTask, const void* pInput, int64_
     int32_t numOfBlocks = taosArrayGetSize(pBlockList);
     stDebug("s-task:%s %p set (merged) submit blocks as a batch, numOfBlocks:%d, ver:%" PRId64, id, pTask, numOfBlocks,
            pMerged->ver);
-    qSetMultiStreamInput(pExecutor, pBlockList->pData, numOfBlocks, STREAM_INPUT__MERGED_SUBMIT);
+    code = qSetMultiStreamInput(pExecutor, pBlockList->pData, numOfBlocks, STREAM_INPUT__MERGED_SUBMIT);
     ASSERT((*pVer) <= pMerged->ver);
     (*pVer) = pMerged->ver;
 
   } else if (pItem->type == STREAM_INPUT__REF_DATA_BLOCK) {
     const SStreamRefDataBlock* pRefBlock = (const SStreamRefDataBlock*)pInput;
-    qSetMultiStreamInput(pExecutor, pRefBlock->pBlock, 1, STREAM_INPUT__DATA_BLOCK);
+    code = qSetMultiStreamInput(pExecutor, pRefBlock->pBlock, 1, STREAM_INPUT__DATA_BLOCK);
 
   } else if (pItem->type == STREAM_INPUT__CHECKPOINT || pItem->type == STREAM_INPUT__CHECKPOINT_TRIGGER) {
     const SStreamDataBlock* pCheckpoint = (const SStreamDataBlock*)pInput;
-    qSetMultiStreamInput(pExecutor, pCheckpoint->blocks, 1, pItem->type);
+    code = qSetMultiStreamInput(pExecutor, pCheckpoint->blocks, 1, pItem->type);
 
   } else {
     ASSERT(0);
   }
+
+  return code;
 }
 
-int32_t streamProcessTransstateBlock(SStreamTask* pTask, SStreamDataBlock* pBlock) {
+void streamProcessTransstateBlock(SStreamTask* pTask, SStreamDataBlock* pBlock) {
   const char* id = pTask->id.idStr;
   int32_t     code = TSDB_CODE_SUCCESS;
   int32_t     level = pTask->info.taskLevel;
@@ -505,7 +528,7 @@ int32_t streamProcessTransstateBlock(SStreamTask* pTask, SStreamDataBlock* pBloc
     if (remain > 0) {
       streamFreeQitem((SStreamQueueItem*)pBlock);
       stDebug("s-task:%s receive upstream trans-state msg, not sent remain:%d", id, remain);
-      return 0;
+      return;
     }
   }
 
@@ -526,7 +549,7 @@ int32_t streamProcessTransstateBlock(SStreamTask* pTask, SStreamDataBlock* pBloc
       pBlock->srcVgId = pTask->pMeta->vgId;
       code = taosWriteQitem(pTask->outputq.queue->pQueue, pBlock);
       if (code == 0) {
-        streamDispatchStreamBlock(pTask);
+        (void) streamDispatchStreamBlock(pTask);
       } else {  // todo put into queue failed, retry
         streamFreeQitem((SStreamQueueItem*)pBlock);
       }
@@ -540,11 +563,9 @@ int32_t streamProcessTransstateBlock(SStreamTask* pTask, SStreamDataBlock* pBloc
 
     code = streamTransferStatePrepare(pTask);
     if (code != TSDB_CODE_SUCCESS) {
-      /*int8_t status = */ streamTaskSetSchedStatusInactive(pTask);
+      (void)streamTaskSetSchedStatusInactive(pTask);
     }
   }
-
-  return code;
 }
 
 //static void streamTaskSetIdleInfo(SStreamTask* pTask, int32_t idleTime) { pTask->status.schedIdleTime = idleTime; }
@@ -559,7 +580,11 @@ static void doStreamTaskExecImpl(SStreamTask* pTask, SStreamQueueItem* pBlock, i
 
   stDebug("s-task:%s start to process batch blocks, num:%d, type:%s", id, num, streamQueueItemGetTypeStr(pBlock->type));
 
-  doSetStreamInputBlock(pTask, pBlock, &ver, id);
+  int32_t code = doSetStreamInputBlock(pTask, pBlock, &ver, id);
+  if(code) {
+    stError("s-task:%s failed to set input block, not exec for these blocks", id);
+    return;
+  }
 
   int64_t totalSize = 0;
   int32_t totalBlocks = 0;
@@ -601,9 +626,9 @@ void flushStateDataInExecutor(SStreamTask* pTask, SStreamQueueItem* pCheckpointB
     STaskId*     pHTaskId = &pTask->hTaskInfo.id;
     SStreamTask* pHTask = NULL;
     int32_t code = streamMetaAcquireTask(pTask->pMeta, pHTaskId->streamId, pHTaskId->taskId, &pHTask);
-    if (pHTask != NULL) {
+    if (code == TSDB_CODE_SUCCESS) {   // ignore the error code.
-      streamTaskReleaseState(pHTask);
+      (void) streamTaskReleaseState(pHTask);
-      streamTaskReloadState(pTask);
+      (void) streamTaskReloadState(pTask);
       stDebug("s-task:%s transfer state from fill-history task:%s, status:%s completed", id, pHTask->id.idStr,
               streamTaskGetStatus(pHTask).name);
 
@@ -676,7 +701,7 @@ static int32_t doStreamExecTask(SStreamTask* pTask) {
     // dispatch checkpoint msg to all downstream tasks
     int32_t type = pInput->type;
     if (type == STREAM_INPUT__CHECKPOINT_TRIGGER) {
-      streamProcessCheckpointTriggerBlock(pTask, (SStreamDataBlock*)pInput);
+      (void) streamProcessCheckpointTriggerBlock(pTask, (SStreamDataBlock*)pInput);
       continue;
     }
 
@@ -694,7 +719,10 @@ static int32_t doStreamExecTask(SStreamTask* pTask) {
       if (type == STREAM_INPUT__DATA_BLOCK) {
         pTask->execInfo.sink.dataSize += blockSize;
         stDebug("s-task:%s sink task start to sink %d blocks, size:%.2fKiB", id, numOfBlocks, SIZE_IN_KiB(blockSize));
-        doOutputResultBlockImpl(pTask, (SStreamDataBlock*)pInput);
+        int32_t code = doOutputResultBlockImpl(pTask, (SStreamDataBlock*)pInput);
+        if (code != TSDB_CODE_SUCCESS) {
+          // todo handle error.
+        }
 
         double el = (taosGetTimestampMs() - st) / 1000.0;
         if (fabs(el - 0.0) <= DBL_EPSILON) {
@@ -712,11 +740,11 @@ static int32_t doStreamExecTask(SStreamTask* pTask) {
       streamFreeQitem(pInput);
     } else { // todo other thread may change the status
     // do nothing after sync executor state to storage backend, untill the vnode-level checkpoint is completed.
-      taosThreadMutexLock(&pTask->lock);
+      streamMutexLock(&pTask->lock);
       SStreamTaskState pState = streamTaskGetStatus(pTask);
       if (pState.state == TASK_STATUS__CK) {
         stDebug("s-task:%s checkpoint block received, set status:%s", id, pState.name);
-        streamTaskBuildCheckpoint(pTask);
+        (void) streamTaskBuildCheckpoint(pTask);   // ignore this error msg, and continue
       } else { // todo refactor
         int32_t code = 0;
         if (pTask->info.taskLevel == TASK_LEVEL__SOURCE) {
@@ -732,7 +760,7 @@ static int32_t doStreamExecTask(SStreamTask* pTask) {
         }
       }
 
-      taosThreadMutexUnlock(&pTask->lock);
+      streamMutexUnlock(&pTask->lock);
       streamFreeQitem(pInput);
       return 0;
     }
@@ -765,21 +793,21 @@ bool streamTaskReadyToRun(const SStreamTask* pTask, char** pStatus) {
   }
 }
 
-int32_t streamResumeTask(SStreamTask* pTask) {
+void streamResumeTask(SStreamTask* pTask) {
   ASSERT(pTask->status.schedStatus == TASK_SCHED_STATUS__ACTIVE);
   const char* id = pTask->id.idStr;
 
   while (1) {
-    /*int32_t code = */ doStreamExecTask(pTask);
+    (void) doStreamExecTask(pTask);
 
     // check if continue
-    taosThreadMutexLock(&pTask->lock);
+    streamMutexLock(&pTask->lock);
 
     int32_t numOfItems = streamQueueGetNumOfItems(pTask->inputq.queue);
     if ((numOfItems == 0) || streamTaskShouldStop(pTask) || streamTaskShouldPause(pTask)) {
       atomic_store_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE);
       streamTaskClearSchedIdleInfo(pTask);
-      taosThreadMutexUnlock(&pTask->lock);
+      streamMutexUnlock(&pTask->lock);
 
       setLastExecTs(pTask, taosGetTimestampMs());
 
@@ -787,19 +815,19 @@ int32_t streamResumeTask(SStreamTask* pTask) {
       stDebug("s-task:%s exec completed, status:%s, sched-status:%d, lastExecTs:%" PRId64, id, p,
               pTask->status.schedStatus, pTask->status.lastExecTs);
 
-      return 0;
+      return;
     } else {
       // check if this task needs to be idle for a while
       if (pTask->status.schedIdleTime > 0) {
         streamTaskResumeInFuture(pTask);
 
-        taosThreadMutexUnlock(&pTask->lock);
+        streamMutexUnlock(&pTask->lock);
         setLastExecTs(pTask, taosGetTimestampMs());
-        return 0;
+        return;
       }
     }
 
-    taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
   }
 }
 

source/libs/stream/src/streamHb.c

@@ -54,7 +54,7 @@ static bool existInHbMsg(SStreamHbMsg* pMsg, SDownstreamTaskEpset* pTaskEpset) {
 static void addUpdateNodeIntoHbMsg(SStreamTask* pTask, SStreamHbMsg* pMsg) {
   SStreamMeta* pMeta = pTask->pMeta;
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
 
   int32_t num = taosArrayGetSize(pTask->outputInfo.pNodeEpsetUpdateList);
   for (int j = 0; j < num; ++j) {
@@ -73,7 +73,7 @@ static void addUpdateNodeIntoHbMsg(SStreamTask* pTask, SStreamHbMsg* pMsg) {
   }
 
   taosArrayClear(pTask->outputInfo.pNodeEpsetUpdateList);
-  (void) taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 }
 
 static int32_t doSendHbMsgInfo(SStreamHbMsg* pMsg, SStreamMeta* pMeta, SEpSet* pEpset) {
@@ -170,9 +170,9 @@ int32_t streamMetaSendHbHelper(SStreamMeta* pMeta) {
       continue;
     }
 
-    (void) taosThreadMutexLock(&(*pTask)->lock);
+    streamMutexLock(&(*pTask)->lock);
     STaskStatusEntry entry = streamTaskGetStatusEntry(*pTask);
-    (void) taosThreadMutexUnlock(&(*pTask)->lock);
+    streamMutexUnlock(&(*pTask)->lock);
 
     entry.inputRate = entry.inputQUsed * 100.0 / (2 * STREAM_TASK_QUEUE_CAPACITY_IN_SIZE);
     if ((*pTask)->info.taskLevel == TASK_LEVEL__SINK) {
@@ -190,9 +190,9 @@ int32_t streamMetaSendHbHelper(SStreamMeta* pMeta) {
         stInfo("s-task:%s set kill checkpoint trans in hbMsg, transId:%d, clear the active checkpointInfo",
                (*pTask)->id.idStr, p->transId);
 
-        (void) taosThreadMutexLock(&(*pTask)->lock);
+        streamMutexLock(&(*pTask)->lock);
         streamTaskClearCheckInfo((*pTask), true);
-        (void) taosThreadMutexUnlock(&(*pTask)->lock);
+        streamMutexUnlock(&(*pTask)->lock);
       }
     }
 

source/libs/stream/src/streamMeta.c

@@ -57,10 +57,13 @@ static void streamMetaEnvInit() {
   streamMetaId = taosOpenRef(64, streamMetaCloseImpl);
 
   metaRefMgtInit();
-  streamTimerInit();
+  int32_t code = streamTimerInit();
+  if (code != 0) {
+    stError("failed to init stream meta env, start failed");
+  }
 }
 
-void streamMetaInit() { taosThreadOnce(&streamMetaModuleInit, streamMetaEnvInit); }
+void streamMetaInit() { (void) taosThreadOnce(&streamMetaModuleInit, streamMetaEnvInit); }
 
 void streamMetaCleanup() {
   taosCloseRef(streamBackendId);
@@ -89,11 +92,11 @@ void metaRefMgtCleanup() {
   }
   taosHashCleanup(gMetaRefMgt.pTable);
 
-  taosThreadMutexDestroy(&gMetaRefMgt.mutex);
+  streamMutexDestroy(&gMetaRefMgt.mutex);
 }
 
 int32_t metaRefMgtAdd(int64_t vgId, int64_t* rid) {
-  taosThreadMutexLock(&gMetaRefMgt.mutex);
+  streamMutexLock(&gMetaRefMgt.mutex);
 
   void* p = taosHashGet(gMetaRefMgt.pTable, &vgId, sizeof(vgId));
   if (p == NULL) {
@@ -105,7 +108,7 @@ int32_t metaRefMgtAdd(int64_t vgId, int64_t* rid) {
     taosArrayPush(list, &rid);
   }
 
-  taosThreadMutexUnlock(&gMetaRefMgt.mutex);
+  streamMutexUnlock(&gMetaRefMgt.mutex);
   return 0;
 }
 
@@ -238,7 +241,7 @@ int32_t streamMetaMayCvtDbFormat(SStreamMeta* pMeta) {
 int32_t streamTaskSetDb(SStreamMeta* pMeta, SStreamTask* pTask, const char* key) {
   int64_t chkpId = pTask->chkInfo.checkpointId;
 
-  taosThreadMutexLock(&pMeta->backendMutex);
+  streamMutexLock(&pMeta->backendMutex);
   void** ppBackend = taosHashGet(pMeta->pTaskDbUnique, key, strlen(key));
   if ((ppBackend != NULL) && (*ppBackend != NULL)) {
     taskDbAddRef(*ppBackend);
@@ -247,7 +250,7 @@ int32_t streamTaskSetDb(SStreamMeta* pMeta, SStreamTask* pTask, const char* key)
     pBackend->pMeta = pMeta;
     pTask->pBackend = pBackend;
 
-    taosThreadMutexUnlock(&pMeta->backendMutex);
+    streamMutexUnlock(&pMeta->backendMutex);
     stDebug("s-task:0x%x set backend %p", pTask->id.taskId, pBackend);
     return 0;
   }
@@ -260,11 +263,11 @@ int32_t streamTaskSetDb(SStreamMeta* pMeta, SStreamTask* pTask, const char* key)
       break;
     }
 
-    taosThreadMutexUnlock(&pMeta->backendMutex);
+    streamMutexUnlock(&pMeta->backendMutex);
     taosMsleep(1000);
 
     stDebug("backend held by other task, restart later, path:%s, key:%s", pMeta->path, key);
-    taosThreadMutexLock(&pMeta->backendMutex);
+    streamMutexLock(&pMeta->backendMutex);
   }
 
   int64_t tref = taosAddRef(taskDbWrapperId, pBackend);
@@ -276,7 +279,7 @@ int32_t streamTaskSetDb(SStreamMeta* pMeta, SStreamTask* pTask, const char* key)
   if (processVer != -1) pTask->chkInfo.processedVer = processVer;
 
   taosHashPut(pMeta->pTaskDbUnique, key, strlen(key), &pBackend, sizeof(void*));
-  taosThreadMutexUnlock(&pMeta->backendMutex);
+  streamMutexUnlock(&pMeta->backendMutex);
 
   stDebug("s-task:0x%x set backend %p", pTask->id.taskId, pBackend);
   return 0;
@@ -286,10 +289,10 @@ void streamMetaRemoveDB(void* arg, char* key) {
   if (arg == NULL || key == NULL) return;
 
   SStreamMeta* pMeta = arg;
-  taosThreadMutexLock(&pMeta->backendMutex);
+  streamMutexLock(&pMeta->backendMutex);
   taosHashRemove(pMeta->pTaskDbUnique, key, strlen(key));
 
-  taosThreadMutexUnlock(&pMeta->backendMutex);
+  streamMutexUnlock(&pMeta->backendMutex);
 }
 
 int32_t streamMetaOpen(const char* path, void* ahandle, FTaskBuild buildTaskFn, FTaskExpand expandTaskFn, int32_t vgId,
@@ -541,7 +544,7 @@ void streamMetaCloseImpl(void* arg) {
   pMeta->pHbInfo = NULL;
 
   taosMemoryFree(pMeta->path);
-  taosThreadMutexDestroy(&pMeta->backendMutex);
+  streamMutexDestroy(&pMeta->backendMutex);
 
   taosCleanUpScheduler(pMeta->qHandle);
   taosMemoryFree(pMeta->qHandle);
@@ -1132,7 +1135,7 @@ int32_t streamMetaSendMsgBeforeCloseTasks(SStreamMeta* pMeta, SArray** pList) {
       continue;
     }
 
-    taosThreadMutexLock(&pTask->lock);
+    streamMutexLock(&pTask->lock);
 
     SStreamTaskState pState = streamTaskGetStatus(pTask);
     if (pState.state == TASK_STATUS__CK) {
@@ -1141,7 +1144,7 @@ int32_t streamMetaSendMsgBeforeCloseTasks(SStreamMeta* pMeta, SArray** pList) {
       stDebug("s-task:%s status:%s not reset the checkpoint", pTask->id.idStr, pState.name);
     }
 
-    taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     streamMetaReleaseTask(pMeta, pTask);
   }
 
@@ -1271,7 +1274,7 @@ int32_t streamMetaStartAllTasks(SStreamMeta* pMeta) {
         streamLaunchFillHistoryTask(pTask);
       }
 
-      streamMetaAddTaskLaunchResult(pMeta, pTaskId->streamId, pTaskId->taskId, pInfo->checkTs, pInfo->readyTs, true);
+      (void) streamMetaAddTaskLaunchResult(pMeta, pTaskId->streamId, pTaskId->taskId, pInfo->checkTs, pInfo->readyTs, true);
       streamMetaReleaseTask(pMeta, pTask);
       continue;
     }
@@ -1388,16 +1391,16 @@ int32_t streamMetaStartOneTask(SStreamMeta* pMeta, int64_t streamId, int32_t tas
   ASSERT(pTask->status.downstreamReady == 0);
 
   // avoid initialization and destroy running concurrently.
-  taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   if (pTask->pBackend == NULL) {
     code = pMeta->expandTaskFn(pTask);
-    taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     if (code != TSDB_CODE_SUCCESS) {
       streamMetaAddFailedTaskSelf(pTask, pInfo->readyTs);
     }
   } else {
-    taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
   }
 
   if (code == TSDB_CODE_SUCCESS) {
@@ -1440,7 +1443,6 @@ int32_t streamMetaAddTaskLaunchResult(SStreamMeta* pMeta, int64_t streamId, int3
   STaskId         id = {.streamId = streamId, .taskId = taskId};
 
   streamMetaWLock(pMeta);
-
   SStreamTask** p = taosHashGet(pMeta->pTasksMap, &id, sizeof(id));
   if (p == NULL) {  // task does not exists in current vnode, not record the complete info
     stError("vgId:%d s-task:0x%x not exists discard the check downstream info", pMeta->vgId, taskId);
@@ -1449,9 +1451,9 @@ int32_t streamMetaAddTaskLaunchResult(SStreamMeta* pMeta, int64_t streamId, int3
   }
 
   // clear the send consensus-checkpointId flag
-  taosThreadMutexLock(&(*p)->lock);
+  streamMutexLock(&(*p)->lock);
   (*p)->status.sendConsensusChkptId = false;
-  taosThreadMutexUnlock(&(*p)->lock);
+  streamMutexUnlock(&(*p)->lock);
 
   if (pStartInfo->startAllTasks != 1) {
     int64_t el = endTs - startTs;
@@ -1537,9 +1539,9 @@ int32_t streamMetaAddFailedTask(SStreamMeta* pMeta, int64_t streamId, int32_t ta
     streamMetaRUnLock(pMeta);
 
     // add the failed task info, along with the related fill-history task info into tasks list.
-    streamMetaAddTaskLaunchResult(pMeta, streamId, taskId, startTs, now, false);
+    (void) streamMetaAddTaskLaunchResult(pMeta, streamId, taskId, startTs, now, false);
     if (hasFillhistoryTask) {
-      streamMetaAddTaskLaunchResult(pMeta, hId.streamId, hId.taskId, startTs, now, false);
+      (void) streamMetaAddTaskLaunchResult(pMeta, hId.streamId, hId.taskId, startTs, now, false);
     }
   } else {
     streamMetaRUnLock(pMeta);
@@ -1554,12 +1556,12 @@ int32_t streamMetaAddFailedTask(SStreamMeta* pMeta, int64_t streamId, int32_t ta
 
 void streamMetaAddFailedTaskSelf(SStreamTask* pTask, int64_t failedTs) {
   int32_t startTs = pTask->execInfo.checkTs;
-  streamMetaAddTaskLaunchResult(pTask->pMeta, pTask->id.streamId, pTask->id.taskId, startTs, failedTs, false);
+  (void) streamMetaAddTaskLaunchResult(pTask->pMeta, pTask->id.streamId, pTask->id.taskId, startTs, failedTs, false);
 
   // automatically set the related fill-history task to be failed.
   if (HAS_RELATED_FILLHISTORY_TASK(pTask)) {
     STaskId* pId = &pTask->hTaskInfo.id;
-    streamMetaAddTaskLaunchResult(pTask->pMeta, pId->streamId, pId->taskId, startTs, failedTs, false);
+    (void) streamMetaAddTaskLaunchResult(pTask->pMeta, pId->streamId, pId->taskId, startTs, failedTs, false);
   }
 }
 

source/libs/stream/src/streamQueue.c

@@ -27,20 +27,24 @@ typedef struct SQueueReader {
   int32_t       waitDuration;  // maximum wait time to format several block into a batch to process, unit: ms
 } SQueueReader;
 
+#define streamQueueCurItem(_q) ((_q)->qItem)
+
 static bool streamTaskExtractAvailableToken(STokenBucket* pBucket, const char* id);
 static void streamTaskPutbackToken(STokenBucket* pBucket);
 static void streamTaskConsumeQuota(STokenBucket* pBucket, int32_t bytes);
 
 static void streamQueueCleanup(SStreamQueue* pQueue) {
-  void* qItem = NULL;
+  SStreamQueueItem* qItem = NULL;
-  while ((qItem = streamQueueNextItem(pQueue)) != NULL) {
+  while (1) {
+    streamQueueNextItem(pQueue, &qItem);
+    if (qItem == NULL) {
+      break;
+    }
     streamFreeQitem(qItem);
   }
   pQueue->status = STREAM_QUEUE__SUCESS;
 }
 
-static void* streamQueueCurItem(SStreamQueue* queue) { return queue->qItem; }
-
 int32_t streamQueueOpen(int64_t cap, SStreamQueue** pQ) {
   *pQ = NULL;
   int32_t code = 0;
@@ -81,21 +85,22 @@ void streamQueueClose(SStreamQueue* pQueue, int32_t taskId) {
   taosMemoryFree(pQueue);
 }
 
-void* streamQueueNextItem(SStreamQueue* pQueue) {
+void streamQueueNextItem(SStreamQueue* pQueue, SStreamQueueItem** pItem) {
+  *pItem = NULL;
   int8_t flag = atomic_exchange_8(&pQueue->status, STREAM_QUEUE__PROCESSING);
 
   if (flag == STREAM_QUEUE__FAILED) {
     ASSERT(pQueue->qItem != NULL);
-    return streamQueueCurItem(pQueue);
+    *pItem = streamQueueCurItem(pQueue);
   } else {
     pQueue->qItem = NULL;
-    taosGetQitem(pQueue->qall, &pQueue->qItem);
+    (void) taosGetQitem(pQueue->qall, &pQueue->qItem);
     if (pQueue->qItem == NULL) {
-      taosReadAllQitems(pQueue->pQueue, pQueue->qall);
+      (void) taosReadAllQitems(pQueue->pQueue, pQueue->qall);
-      taosGetQitem(pQueue->qall, &pQueue->qItem);
+      (void) taosGetQitem(pQueue->qall, &pQueue->qItem);
     }
 
-    return streamQueueCurItem(pQueue);
+    *pItem = streamQueueCurItem(pQueue);
   }
 }
 
@@ -181,7 +186,8 @@ EExtractDataCode streamTaskGetDataFromInputQ(SStreamTask* pTask, SStreamQueueIte
       return EXEC_CONTINUE;
     }
 
-    SStreamQueueItem* qItem = streamQueueNextItem(pTask->inputq.queue);
+    SStreamQueueItem* qItem = NULL;
+    streamQueueNextItem(pTask->inputq.queue, (SStreamQueueItem**)&qItem);
     if (qItem == NULL) {
       // restore the token to bucket
       if (*numOfBlocks > 0) {
@@ -338,7 +344,8 @@ int32_t streamTaskPutDataIntoInputQ(SStreamTask* pTask, SStreamQueueItem* pItem)
 
   if (type != STREAM_INPUT__GET_RES && type != STREAM_INPUT__CHECKPOINT && type != STREAM_INPUT__CHECKPOINT_TRIGGER &&
       (pTask->info.delaySchedParam != 0)) {
-    atomic_val_compare_exchange_8(&pTask->schedInfo.status, TASK_TRIGGER_STATUS__INACTIVE, TASK_TRIGGER_STATUS__ACTIVE);
+    (void)atomic_val_compare_exchange_8(&pTask->schedInfo.status, TASK_TRIGGER_STATUS__INACTIVE,
+                                        TASK_TRIGGER_STATUS__ACTIVE);
     stDebug("s-task:%s new data arrived, active the sched-trigger, triggerStatus:%d", pTask->id.idStr,
             pTask->schedInfo.status);
   }
@@ -347,18 +354,19 @@ int32_t streamTaskPutDataIntoInputQ(SStreamTask* pTask, SStreamQueueItem* pItem)
 }
 
 int32_t streamTaskPutTranstateIntoInputQ(SStreamTask* pTask) {
-  int32_t           code;
+  int32_t           code = 0;
-  SStreamDataBlock* pTranstate;
+  SStreamDataBlock* pTranstate = NULL;
+  SSDataBlock*      pBlock = NULL;
 
   code = taosAllocateQitem(sizeof(SStreamDataBlock), DEF_QITEM, sizeof(SSDataBlock), (void**)&pTranstate);
   if (code) {
     return code;
   }
 
-  SSDataBlock* pBlock = taosMemoryCalloc(1, sizeof(SSDataBlock));
+  pBlock = taosMemoryCalloc(1, sizeof(SSDataBlock));
   if (pBlock == NULL) {
-    taosFreeQitem(pTranstate);
+    code = TSDB_CODE_OUT_OF_MEMORY;
-    return TSDB_CODE_OUT_OF_MEMORY;
+    goto _err;
   }
 
   pTranstate->type = STREAM_INPUT__TRANS_STATE;
@@ -368,15 +376,30 @@ int32_t streamTaskPutTranstateIntoInputQ(SStreamTask* pTask) {
   pBlock->info.childId = pTask->info.selfChildId;
 
   pTranstate->blocks = taosArrayInit(4, sizeof(SSDataBlock));  // pBlock;
-  taosArrayPush(pTranstate->blocks, pBlock);
+  if (pTranstate->blocks == NULL) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto _err;
+  }
+
+  void* p = taosArrayPush(pTranstate->blocks, pBlock);
+  if (p == NULL) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto _err;
+  }
 
   taosMemoryFree(pBlock);
   if (streamTaskPutDataIntoInputQ(pTask, (SStreamQueueItem*)pTranstate) < 0) {
-    return TSDB_CODE_OUT_OF_MEMORY;
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto _err;
   }
 
   pTask->status.appendTranstateBlock = true;
   return TSDB_CODE_SUCCESS;
+
+_err:
+  taosMemoryFree(pBlock);
+  taosFreeQitem(pTranstate);
+  return code;
 }
 
 // the result should be put into the outputQ in any cases, the result may be lost otherwise.

source/libs/stream/src/streamSched.c

@@ -36,7 +36,7 @@ int32_t streamSetupScheduleTrigger(SStreamTask* pTask) {
 
 int32_t streamTrySchedExec(SStreamTask* pTask) {
   if (streamTaskSetSchedStatusWait(pTask)) {
-    streamTaskSchedTask(pTask->pMsgCb, pTask->info.nodeId, pTask->id.streamId, pTask->id.taskId, 0);
+    return streamTaskSchedTask(pTask->pMsgCb, pTask->info.nodeId, pTask->id.streamId, pTask->id.taskId, 0);
   } else {
     stTrace("s-task:%s not launch task since sched status:%d", pTask->id.idStr, pTask->status.schedStatus);
   }
@@ -47,10 +47,9 @@ int32_t streamTrySchedExec(SStreamTask* pTask) {
 int32_t streamTaskSchedTask(SMsgCb* pMsgCb, int32_t vgId, int64_t streamId, int32_t taskId, int32_t execType) {
   SStreamTaskRunReq* pRunReq = rpcMallocCont(sizeof(SStreamTaskRunReq));
   if (pRunReq == NULL) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
     stError("vgId:%d failed to create msg to start stream task:0x%x exec, type:%d, code:%s", vgId, taskId, execType,
             terrstr());
-    return -1;
+    return TSDB_CODE_OUT_OF_MEMORY;
   }
 
   if (streamId != 0) {
@@ -65,29 +64,27 @@ int32_t streamTaskSchedTask(SMsgCb* pMsgCb, int32_t vgId, int64_t streamId, int3
   pRunReq->reqType = execType;
 
   SRpcMsg msg = {.msgType = TDMT_STREAM_TASK_RUN, .pCont = pRunReq, .contLen = sizeof(SStreamTaskRunReq)};
-  tmsgPutToQueue(pMsgCb, STREAM_QUEUE, &msg);
+  return tmsgPutToQueue(pMsgCb, STREAM_QUEUE, &msg);
-  return TSDB_CODE_SUCCESS;
 }
 
 void streamTaskClearSchedIdleInfo(SStreamTask* pTask) { pTask->status.schedIdleTime = 0; }
 
 void streamTaskSetIdleInfo(SStreamTask* pTask, int32_t idleTime) { pTask->status.schedIdleTime = idleTime; }
 
-int32_t streamTaskResumeInFuture(SStreamTask* pTask) {
+void streamTaskResumeInFuture(SStreamTask* pTask) {
   int32_t ref = atomic_add_fetch_32(&pTask->status.timerActive, 1);
   stDebug("s-task:%s task should idle, add into timer to retry in %dms, ref:%d", pTask->id.idStr,
           pTask->status.schedIdleTime, ref);
 
   // add one ref count for task
-  /*SStreamTask* pAddRefTask = */ streamMetaAcquireOneTask(pTask);
+  streamMetaAcquireOneTask(pTask);
 
   if (pTask->schedInfo.pIdleTimer == NULL) {
     pTask->schedInfo.pIdleTimer = taosTmrStart(streamTaskResumeHelper, pTask->status.schedIdleTime, pTask, streamTimer);
   } else {
-    taosTmrReset(streamTaskResumeHelper, pTask->status.schedIdleTime, pTask, streamTimer, &pTask->schedInfo.pIdleTimer);
+    streamTmrReset(streamTaskResumeHelper, pTask->status.schedIdleTime, pTask, streamTimer,
+                   &pTask->schedInfo.pIdleTimer, pTask->pMeta->vgId, "resume-task-tmr");
   }
-
-  return TSDB_CODE_SUCCESS;
 }
 
 //////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -97,7 +94,7 @@ void streamTaskResumeHelper(void* param, void* tmrId) {
   SStreamTaskState  p = streamTaskGetStatus(pTask);
 
   if (p.state == TASK_STATUS__DROPPING || p.state == TASK_STATUS__STOP) {
-    streamTaskSetSchedStatusInactive(pTask);
+    (void) streamTaskSetSchedStatusInactive(pTask);
 
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s status:%s not resume task, ref:%d", pId->idStr, p.name, ref);
@@ -106,8 +103,10 @@ void streamTaskResumeHelper(void* param, void* tmrId) {
     return;
   }
 
-  streamTaskSchedTask(pTask->pMsgCb, pTask->info.nodeId, pId->streamId, pId->taskId, STREAM_EXEC_T_RESUME_TASK);
+  int32_t code = streamTaskSchedTask(pTask->pMsgCb, pTask->info.nodeId, pId->streamId, pId->taskId, STREAM_EXEC_T_RESUME_TASK);
-
+  if (code) {
+    stError("s-task:%s sched task failed, code:%s", pTask->id.idStr, strerror(code));
+  } else {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("trigger to resume s-task:%s after being idled for %dms, ref:%d", pId->idStr, pTask->status.schedIdleTime,
             ref);
@@ -116,11 +115,13 @@ void streamTaskResumeHelper(void* param, void* tmrId) {
     streamTaskClearSchedIdleInfo(pTask);
     streamMetaReleaseTask(pTask->pMeta, pTask);
   }
+}
 
 void streamTaskSchedHelper(void* param, void* tmrId) {
   SStreamTask* pTask = (void*)param;
   const char*  id = pTask->id.idStr;
   int32_t      nextTrigger = (int32_t)pTask->info.delaySchedParam;
+  int32_t      vgId = pTask->pMeta->vgId;
 
   int8_t status = atomic_load_8(&pTask->schedInfo.status);
   stTrace("s-task:%s in scheduler, trigger status:%d, next:%dms", id, status, nextTrigger);
@@ -140,7 +141,7 @@ void streamTaskSchedHelper(void* param, void* tmrId) {
       if (code) {
         stError("s-task:%s failed to prepare retrieve data trigger, code:%s, try again in %dms", id, "out of memory",
                 nextTrigger);
-        taosTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer);
+        streamTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer, vgId, "sched-run-tmr");
         terrno = code;
         return;
       }
@@ -152,7 +153,7 @@ void streamTaskSchedHelper(void* param, void* tmrId) {
 
         stError("s-task:%s failed to prepare retrieve data trigger, code:%s, try again in %dms", id, "out of memory",
                 nextTrigger);
-        taosTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer);
+        streamTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer, vgId, "sched-run-tmr");
         return;
       }
 
@@ -161,13 +162,16 @@ void streamTaskSchedHelper(void* param, void* tmrId) {
 
       code = streamTaskPutDataIntoInputQ(pTask, (SStreamQueueItem*)pTrigger);
       if (code != TSDB_CODE_SUCCESS) {
-        taosTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer);
+        streamTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer, vgId, "sched-run-tmr");
         return;
       }
 
-      streamTrySchedExec(pTask);
+      code = streamTrySchedExec(pTask);
+      if (code != TSDB_CODE_SUCCESS) {
+        stError("s-task:%s failed to sched to run, wait for next time", pTask->id.idStr);
+      }
     }
   }
 
-  taosTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer);
+  streamTmrReset(streamTaskSchedHelper, nextTrigger, pTask, streamTimer, &pTask->schedInfo.pDelayTimer, vgId, "sched-run-tmr");
 }

source/libs/stream/src/streamStartHistory.c

@@ -212,7 +212,7 @@ int32_t streamLaunchFillHistoryTask(SStreamTask* pTask) {
     stDebug("s-task:%s not launch related fill-history task:0x%" PRIx64 "-0x%x, status:%s", idStr, hStreamId, hTaskId,
             pStatus.name);
 
-    streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
+    (void) streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
     return -1;  // todo set the correct error code
   }
 
@@ -228,11 +228,11 @@ int32_t streamLaunchFillHistoryTask(SStreamTask* pTask) {
     code = streamMetaAcquireTask(pMeta, hStreamId, hTaskId, &pHisTask);
     if (pHisTask == NULL) {
       stDebug("s-task:%s failed acquire and start fill-history task, it may have been dropped/stopped", idStr);
-      streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
+      (void) streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
     } else {
       if (pHisTask->status.downstreamReady == 1) {  // it's ready now, do nothing
         stDebug("s-task:%s fill-history task is ready, no need to check downstream", pHisTask->id.idStr);
-        streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, true);
+        (void) streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, true);
       } else {  // exist, but not ready, continue check downstream task status
         if (pHisTask->pBackend == NULL) {
           code = pMeta->expandTaskFn(pHisTask);
@@ -289,7 +289,7 @@ void notRetryLaunchFillHistoryTask(SStreamTask* pTask, SLaunchHTaskInfo* pInfo,
   SHistoryTaskInfo* pHTaskInfo = &pTask->hTaskInfo;
 
   int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
-  streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
+  (void) streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
 
   stError("s-task:%s max retry:%d reached, quit from retrying launch related fill-history task:0x%x, ref:%d",
           pTask->id.idStr, MAX_RETRY_LAUNCH_HISTORY_TASK, (int32_t)pHTaskInfo->id.taskId, ref);
@@ -307,7 +307,7 @@ void doRetryLaunchFillHistoryTask(SStreamTask* pTask, SLaunchHTaskInfo* pInfo, i
     stDebug("s-task:0x%" PRIx64 " stopped, not launch rel history task:0x%" PRIx64 ", ref:%d", pInfo->id.taskId,
             pInfo->hTaskId.taskId, ref);
 
-    streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
+    (void) streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
     taosMemoryFree(pInfo);
   } else {
     char*   p = streamTaskGetStatus(pTask).name;
@@ -350,7 +350,7 @@ void tryLaunchHistoryTask(void* param, void* tmrId) {
     streamMetaWUnLock(pMeta);
 
     // record the related fill-history task failed
-    streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
+    (void) streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
     taosMemoryFree(pInfo);
     return;
   }
@@ -407,7 +407,7 @@ void tryLaunchHistoryTask(void* param, void* tmrId) {
 
     streamMetaReleaseTask(pMeta, pTask);
   } else {
-    streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
+    (void) streamMetaAddTaskLaunchResult(pMeta, pInfo->hTaskId.streamId, pInfo->hTaskId.taskId, 0, now, false);
 
     int32_t ref = atomic_sub_fetch_32(&(*ppTask)->status.timerActive, 1);
     stError("s-task:0x%x rel fill-history task:0x%" PRIx64 " may have been destroyed, not launch, ref:%d",
@@ -448,7 +448,7 @@ int32_t launchNotBuiltFillHistoryTask(SStreamTask* pTask) {
   SLaunchHTaskInfo* pInfo = createHTaskLaunchInfo(pMeta, &id, hStreamId, hTaskId);
   if (pInfo == NULL) {
     stError("s-task:%s failed to launch related fill-history task, since Out Of Memory", idStr);
-    streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
+    (void) streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
     return terrno;
   }
 
@@ -465,7 +465,7 @@ int32_t launchNotBuiltFillHistoryTask(SStreamTask* pTask) {
       stError("s-task:%s failed to start timer, related fill-history task not launched, ref:%d", idStr, ref);
 
       taosMemoryFree(pInfo);
-      streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
+      (void) streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
       return terrno;
     }
 

source/libs/stream/src/streamTask.c

@@ -197,13 +197,13 @@ void tFreeStreamTask(SStreamTask* pTask) {
   STaskExecStatisInfo* pStatis = &pTask->execInfo;
 
   ETaskStatus status1 = TASK_STATUS__UNINIT;
-  taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   if (pTask->status.pSM != NULL) {
     SStreamTaskState pStatus = streamTaskGetStatus(pTask);
     p = pStatus.name;
     status1 = pStatus.state;
   }
-  taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   stDebug("start to free s-task:0x%x %p, state:%s", taskId, pTask, p);
 
@@ -284,11 +284,11 @@ void tFreeStreamTask(SStreamTask* pTask) {
   streamTaskDestroyUpstreamInfo(&pTask->upstreamInfo);
 
   taosMemoryFree(pTask->outputInfo.pTokenBucket);
-  taosThreadMutexDestroy(&pTask->lock);
+  streamMutexDestroy(&pTask->lock);
 
   taosArrayDestroy(pTask->msgInfo.pSendInfo);
   pTask->msgInfo.pSendInfo = NULL;
-  taosThreadMutexDestroy(&pTask->msgInfo.lock);
+  streamMutexDestroy(&pTask->msgInfo.lock);
 
   taosArrayDestroy(pTask->outputInfo.pNodeEpsetUpdateList);
   pTask->outputInfo.pNodeEpsetUpdateList = NULL;
@@ -644,11 +644,7 @@ void streamTaskOpenAllUpstreamInput(SStreamTask* pTask) {
 
 void streamTaskCloseUpstreamInput(SStreamTask* pTask, int32_t taskId) {
   SStreamUpstreamEpInfo* pInfo = NULL;
-
+  streamTaskGetUpstreamTaskEpInfo(pTask, taskId, &pInfo);
-  int32_t code = streamTaskGetUpstreamTaskEpInfo(pTask, taskId, &pInfo);
-  if (code != TSDB_CODE_SUCCESS) {
-    return;
-  }
 
   if ((pInfo != NULL) && pInfo->dataAllowed) {
     pInfo->dataAllowed = false;
@@ -659,11 +655,7 @@ void streamTaskCloseUpstreamInput(SStreamTask* pTask, int32_t taskId) {
 
 void streamTaskOpenUpstreamInput(SStreamTask* pTask, int32_t taskId) {
   SStreamUpstreamEpInfo* pInfo = NULL;
-
+  streamTaskGetUpstreamTaskEpInfo(pTask, taskId, &pInfo);
-  int32_t code = streamTaskGetUpstreamTaskEpInfo(pTask, taskId, &pInfo);
-  if (code != TSDB_CODE_SUCCESS) {
-    return;
-  }
 
   if (pInfo != NULL && (!pInfo->dataAllowed)) {
     int32_t t = atomic_sub_fetch_32(&pTask->upstreamInfo.numOfClosed, 1);
@@ -679,34 +671,34 @@ bool streamTaskIsAllUpstreamClosed(SStreamTask* pTask) {
 bool streamTaskSetSchedStatusWait(SStreamTask* pTask) {
   bool ret = false;
 
-  taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   if (pTask->status.schedStatus == TASK_SCHED_STATUS__INACTIVE) {
     pTask->status.schedStatus = TASK_SCHED_STATUS__WAITING;
     ret = true;
   }
-  taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   return ret;
 }
 
 int8_t streamTaskSetSchedStatusActive(SStreamTask* pTask) {
-  taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   int8_t status = pTask->status.schedStatus;
   if (status == TASK_SCHED_STATUS__WAITING) {
     pTask->status.schedStatus = TASK_SCHED_STATUS__ACTIVE;
   }
-  taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   return status;
 }
 
 int8_t streamTaskSetSchedStatusInactive(SStreamTask* pTask) {
-  taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   int8_t status = pTask->status.schedStatus;
   ASSERT(status == TASK_SCHED_STATUS__WAITING || status == TASK_SCHED_STATUS__ACTIVE ||
          status == TASK_SCHED_STATUS__INACTIVE);
   pTask->status.schedStatus = TASK_SCHED_STATUS__INACTIVE;
-  taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   return status;
 }
@@ -723,7 +715,7 @@ int32_t streamTaskClearHTaskAttr(SStreamTask* pTask, int32_t resetRelHalt) {
     stDebug("s-task:%s clear the related stream task:0x%x attr to fill-history task", pTask->id.idStr,
             (int32_t)sTaskId.taskId);
 
-    taosThreadMutexLock(&(*ppStreamTask)->lock);
+    streamMutexLock(&(*ppStreamTask)->lock);
     CLEAR_RELATED_FILLHISTORY_TASK((*ppStreamTask));
 
     if (resetRelHalt) {
@@ -734,7 +726,7 @@ int32_t streamTaskClearHTaskAttr(SStreamTask* pTask, int32_t resetRelHalt) {
     }
 
     streamMetaSaveTask(pMeta, *ppStreamTask);
-    taosThreadMutexUnlock(&(*ppStreamTask)->lock);
+    streamMutexUnlock(&(*ppStreamTask)->lock);
   }
 
   return TSDB_CODE_SUCCESS;
@@ -923,6 +915,7 @@ void streamTaskResume(SStreamTask* pTask) {
 
 bool streamTaskIsSinkTask(const SStreamTask* pTask) { return pTask->info.taskLevel == TASK_LEVEL__SINK; }
 
+// this task must success
 int32_t streamTaskSendCheckpointReq(SStreamTask* pTask) {
   int32_t     code;
   int32_t     tlen = 0;
@@ -960,24 +953,23 @@ int32_t streamTaskSendCheckpointReq(SStreamTask* pTask) {
   return 0;
 }
 
-int32_t streamTaskGetUpstreamTaskEpInfo(SStreamTask* pTask, int32_t taskId, SStreamUpstreamEpInfo** pEpInfo) {
+void streamTaskGetUpstreamTaskEpInfo(SStreamTask* pTask, int32_t taskId, SStreamUpstreamEpInfo** pEpInfo) {
   *pEpInfo = NULL;
 
   int32_t num = taosArrayGetSize(pTask->upstreamInfo.pList);
   for (int32_t i = 0; i < num; ++i) {
     SStreamUpstreamEpInfo* pInfo = taosArrayGetP(pTask->upstreamInfo.pList, i);
     if (pInfo == NULL) {
-      return TSDB_CODE_FAILED;
+      return;
     }
 
     if (pInfo->taskId == taskId) {
       *pEpInfo = pInfo;
-      return TSDB_CODE_SUCCESS;
+      return;
     }
   }
 
   stError("s-task:%s failed to find upstream task:0x%x", pTask->id.idStr, taskId);
-  return TSDB_CODE_FAILED;
 }
 
 SEpSet* streamTaskGetDownstreamEpInfo(SStreamTask* pTask, int32_t taskId) {
@@ -1100,7 +1092,7 @@ void streamTaskDestroyActiveChkptInfo(SActiveCheckpointInfo* pInfo) {
     return;
   }
 
-  taosThreadMutexDestroy(&pInfo->lock);
+  streamMutexDestroy(&pInfo->lock);
   taosArrayDestroy(pInfo->pDispatchTriggerList);
   pInfo->pDispatchTriggerList = NULL;
   taosArrayDestroy(pInfo->pReadyMsgList);

source/libs/stream/src/streamTaskSm.c

@@ -201,7 +201,7 @@ static int32_t doHandleWaitingEvent(SStreamTaskSM* pSM, const char* pEventName,
 
     pSM->pActiveTrans = pNextTrans;
     pSM->startTs = taosGetTimestampMs();
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     code = pNextTrans->pAction(pSM->pTask);
     if (pNextTrans->autoInvokeEndFn) {
@@ -210,7 +210,7 @@ static int32_t doHandleWaitingEvent(SStreamTaskSM* pSM, const char* pEventName,
       return code;
     }
   } else {
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     stDebug("s-task:%s state:%s event:%s in waiting list, req state:%s not fulfilled, put it back", pTask->id.idStr,
             pSM->current.name, GET_EVT_NAME(pEvtInfo->event),
             StreamTaskStatusList[pEvtInfo->status].name);
@@ -247,7 +247,7 @@ int32_t streamTaskRestoreStatus(SStreamTask* pTask) {
   SStreamTaskSM* pSM = pTask->status.pSM;
   int32_t        code = 0;
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
 
   if (pSM->current.state == TASK_STATUS__PAUSE && pSM->pActiveTrans == NULL) {
     SStreamTaskState state = pSM->current;
@@ -326,13 +326,13 @@ static int32_t doHandleEvent(SStreamTaskSM* pSM, EStreamTaskEvent event, STaskSt
       return code;
     }
 
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     while (1) {
       // wait for the task to be here
-      (void) taosThreadMutexLock(&pTask->lock);
+      streamMutexLock(&pTask->lock);
       ETaskStatus s = streamTaskGetStatus(pTask).state;
-      (void) taosThreadMutexUnlock(&pTask->lock);
+      streamMutexUnlock(&pTask->lock);
 
       if ((s == pTrans->next.state) && (pSM->prev.evt == pTrans->event)) {// this event has been handled already
         stDebug("s-task:%s attached event:%s handled", id, GET_EVT_NAME(pTrans->event));
@@ -349,7 +349,7 @@ static int32_t doHandleEvent(SStreamTaskSM* pSM, EStreamTaskEvent event, STaskSt
   } else {  // override current active trans
     pSM->pActiveTrans = pTrans;
     pSM->startTs = taosGetTimestampMs();
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     code = pTrans->pAction(pTask);
 
@@ -374,11 +374,11 @@ static int32_t doHandleEventAsync(SStreamTaskSM* pSM, EStreamTaskEvent event, ST
     info.callBackFn = callbackFn;
 
     code = attachWaitedEvent(pTask, &info);
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
   } else {  // override current active trans
     pSM->pActiveTrans = pTrans;
     pSM->startTs = taosGetTimestampMs();
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     code = pTrans->pAction(pTask);
     // todo handle error code;
@@ -400,11 +400,11 @@ int32_t streamTaskHandleEvent(SStreamTaskSM* pSM, EStreamTaskEvent event) {
   STaskStateTrans* pTrans = NULL;
 
   while (1) {
-    (void) taosThreadMutexLock(&pTask->lock);
+    streamMutexLock(&pTask->lock);
 
     if (pSM->pActiveTrans != NULL && pSM->pActiveTrans->autoInvokeEndFn) {
       EStreamTaskEvent evt = pSM->pActiveTrans->event;
-      (void) taosThreadMutexUnlock(&pTask->lock);
+      streamMutexUnlock(&pTask->lock);
 
       stDebug("s-task:%s status:%s handling event:%s by some other thread, wait for 100ms and check if completed",
               pTask->id.idStr, pSM->current.name, GET_EVT_NAME(evt));
@@ -414,7 +414,7 @@ int32_t streamTaskHandleEvent(SStreamTaskSM* pSM, EStreamTaskEvent event) {
       pTrans = streamTaskFindTransform(pSM->current.state, event);
       if (pTrans == NULL) {
         stDebug("s-task:%s failed to handle event:%s", pTask->id.idStr, GET_EVT_NAME(event));
-        (void) taosThreadMutexUnlock(&pTask->lock);
+        streamMutexUnlock(&pTask->lock);
         return TSDB_CODE_STREAM_INVALID_STATETRANS;
       }
 
@@ -439,11 +439,11 @@ int32_t streamTaskHandleEventAsync(SStreamTaskSM* pSM, EStreamTaskEvent event, _
   STaskStateTrans* pTrans = NULL;
 
   while (1) {
-    (void) taosThreadMutexLock(&pTask->lock);
+    streamMutexLock(&pTask->lock);
 
     if (pSM->pActiveTrans != NULL && pSM->pActiveTrans->autoInvokeEndFn) {
       EStreamTaskEvent evt = pSM->pActiveTrans->event;
-      (void) taosThreadMutexUnlock(&pTask->lock);
+      streamMutexUnlock(&pTask->lock);
 
       stDebug("s-task:%s status:%s handling event:%s by some other thread, wait for 100ms and check if completed",
               pTask->id.idStr, pSM->current.name, GET_EVT_NAME(evt));
@@ -453,7 +453,7 @@ int32_t streamTaskHandleEventAsync(SStreamTaskSM* pSM, EStreamTaskEvent event, _
       pTrans = streamTaskFindTransform(pSM->current.state, event);
       if (pTrans == NULL) {
         stDebug("s-task:%s failed to handle event:%s, status:%s", pTask->id.idStr, GET_EVT_NAME(event), pSM->current.name);
-        (void) taosThreadMutexUnlock(&pTask->lock);
+        streamMutexUnlock(&pTask->lock);
         return TSDB_CODE_STREAM_INVALID_STATETRANS;
       }
 
@@ -485,7 +485,7 @@ int32_t streamTaskOnHandleEventSuccess(SStreamTaskSM* pSM, EStreamTaskEvent even
   int32_t code = 0;
 
   // do update the task status
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
 
   STaskStateTrans* pTrans = pSM->pActiveTrans;
   if (pTrans == NULL) {
@@ -497,14 +497,14 @@ int32_t streamTaskOnHandleEventSuccess(SStreamTaskSM* pSM, EStreamTaskEvent even
     stDebug("s-task:%s event:%s handled failed, current status:%s, trigger event:%s", id, GET_EVT_NAME(event),
             pSM->current.name, GET_EVT_NAME(pSM->prev.evt));
 
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     return TSDB_CODE_STREAM_INVALID_STATETRANS;
   }
 
   if (pTrans->event != event) {
     stWarn("s-task:%s handle event:%s failed, current status:%s, active trans evt:%s", id, GET_EVT_NAME(event),
            pSM->current.name, GET_EVT_NAME(pTrans->event));
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
     return TSDB_CODE_STREAM_INVALID_STATETRANS;
   }
 
@@ -518,7 +518,7 @@ int32_t streamTaskOnHandleEventSuccess(SStreamTaskSM* pSM, EStreamTaskEvent even
   // on success callback, add into lock if necessary, or maybe we should add an option for this?
   code = pTrans->pSuccAction(pTask);
 
-  (void) taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   // todo: add parameter to control lock
   // after handling the callback function assigned by invoker, go on handling the waiting tasks
@@ -532,13 +532,13 @@ int32_t streamTaskOnHandleEventSuccess(SStreamTaskSM* pSM, EStreamTaskEvent even
     stDebug("s-task:%s handle user-specified callback fn for event:%s completed", id, GET_EVT_NAME(pTrans->event));
   }
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
 
   // tasks in waiting list
   if (taosArrayGetSize(pSM->pWaitingEventList) > 0) {
     code = doHandleWaitingEvent(pSM, GET_EVT_NAME(pTrans->event), pTask);
   } else {
-    (void) taosThreadMutexUnlock(&pTask->lock);
+    streamMutexUnlock(&pTask->lock);
 
     int64_t el = (taosGetTimestampMs() - pSM->startTs);
     stDebug("s-task:%s handle event:%s completed, elapsed time:%" PRId64 "ms state:%s -> %s", id,
@@ -563,14 +563,14 @@ const char* streamTaskGetStatusStr(ETaskStatus status) {
 void streamTaskResetStatus(SStreamTask* pTask) {
   SStreamTaskSM* pSM = pTask->status.pSM;
 
-  (void) taosThreadMutexLock(&pTask->lock);
+  streamMutexLock(&pTask->lock);
   stDebug("s-task:%s level:%d fill-history:%d vgId:%d set uninit, prev status:%s", pTask->id.idStr,
           pTask->info.taskLevel, pTask->info.fillHistory, pTask->pMeta->vgId, pSM->current.name);
 
   pSM->current = StreamTaskStatusList[TASK_STATUS__UNINIT];
   pSM->pActiveTrans = NULL;
   taosArrayClear(pSM->pWaitingEventList);
-  (void) taosThreadMutexUnlock(&pTask->lock);
+  streamMutexUnlock(&pTask->lock);
 
   // clear the downstream ready status
   pTask->status.downstreamReady = 0;

source/libs/stream/src/streamTimer.c

@@ -41,11 +41,11 @@ tmr_h streamTimerGetInstance() {
 
 void streamTmrReset(TAOS_TMR_CALLBACK fp, int32_t mseconds, void* param, void* handle, tmr_h* pTmrId, int32_t vgId,
                     const char* pMsg) {
-  while (1) {
+//  while (1) {
     bool ret = taosTmrReset(fp, mseconds, param, handle, pTmrId);
     if (ret) {
-      break;
+//      break;
-    }
-    stError("vgId:%d failed to reset %s, try again", vgId, pMsg);
     }
+//    stError("vgId:%d failed to reset tmr: %s, try again", vgId, pMsg);
+//  }
 }

source/libs/stream/src/streamUtil.c

@@ -0,0 +1,26 @@
+/*
+ * 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/>.
+ */
+
+#include "streamInt.h"
+
+void streamMutexLock(TdThreadMutex *pMutex) {
+  (void) taosThreadMutexLock(pMutex);
+}
+
+void streamMutexUnlock(TdThreadMutex *pMutex) {
+  (void) taosThreadMutexUnlock(pMutex);
+}
+
+void streamMutexDestroy(TdThreadMutex *pMutex) { (void) taosThreadMutexDestroy(pMutex); }