Merge branch 'feature/scheduler' of https://github.com/taosdata/TDengine into sc

.gitignore

@@ -88,6 +88,7 @@ tests/examples/JDBC/JDBCDemo/.classpath
 tests/examples/JDBC/JDBCDemo/.project
 tests/examples/JDBC/JDBCDemo/.settings/
 source/libs/parser/inc/sql.*
+tests/script/tmqResult.txt
 
 # Emacs
 # -*- mode: gitignore; -*-

include/client/taos.h

@@ -85,11 +85,7 @@ typedef struct taosField {
   int32_t bytes;
 } TAOS_FIELD;
 
-#ifdef _TD_GO_DLL_
-#define DLL_EXPORT __declspec(dllexport)
-#else
 #define DLL_EXPORT
-#endif
 
 typedef void (*__taos_async_fn_t)(void *param, TAOS_RES *, int code);
 

include/common/tcommon.h

@@ -63,10 +63,9 @@ typedef struct SDataBlockInfo {
   int64_t        groupId;     // no need to serialize
 } SDataBlockInfo;
 
-// info.numOfCols = taosArrayGetSize(pDataBlock)
 typedef struct SSDataBlock {
-  SColumnDataAgg *pBlockAgg;
-  SArray         *pDataBlock;    // SArray<SColumnInfoData>
+  SColumnDataAgg* pBlockAgg;
+  SArray*         pDataBlock;  // SArray<SColumnInfoData>
   SDataBlockInfo  info;
 } SSDataBlock;
 
@@ -92,23 +91,37 @@ void*   blockDataDestroy(SSDataBlock* pBlock);
 int32_t tEncodeDataBlock(void** buf, const SSDataBlock* pBlock);
 void*   tDecodeDataBlock(const void* buf, SSDataBlock* pBlock);
 
-static FORCE_INLINE void tDeleteSSDataBlock(SSDataBlock* pBlock) {
-  if (pBlock == NULL) {
-    return;
+static FORCE_INLINE void blockDestroyInner(SSDataBlock* pBlock) {
+  // WARNING: do not use info.numOfCols,
+  // sometimes info.numOfCols != array size
+  int32_t numOfOutput = taosArrayGetSize(pBlock->pDataBlock);
+  for (int32_t i = 0; i < numOfOutput; ++i) {
+    SColumnInfoData* pColInfoData = (SColumnInfoData*)taosArrayGet(pBlock->pDataBlock, i);
+    if (IS_VAR_DATA_TYPE(pColInfoData->info.type)) {
+      tfree(pColInfoData->varmeta.offset);
+    } else {
+      tfree(pColInfoData->nullbitmap);
+    }
+
+    tfree(pColInfoData->pData);
   }
-  blockDataDestroy(pBlock);
+
+  taosArrayDestroy(pBlock->pDataBlock);
+  tfree(pBlock->pBlockAgg);
 }
 
+static FORCE_INLINE void tDeleteSSDataBlock(SSDataBlock* pBlock) { blockDestroyInner(pBlock); }
+
 static FORCE_INLINE int32_t tEncodeSMqPollRsp(void** buf, const SMqPollRsp* pRsp) {
   int32_t tlen = 0;
   int32_t sz = 0;
-  tlen += taosEncodeFixedI64(buf, pRsp->consumerId);
+  // tlen += taosEncodeFixedI64(buf, pRsp->consumerId);
   tlen += taosEncodeFixedI64(buf, pRsp->reqOffset);
   tlen += taosEncodeFixedI64(buf, pRsp->rspOffset);
   tlen += taosEncodeFixedI32(buf, pRsp->skipLogNum);
   tlen += taosEncodeFixedI32(buf, pRsp->numOfTopics);
   if (pRsp->numOfTopics == 0) return tlen;
-  tlen += tEncodeSSchemaWrapper(buf, pRsp->schemas);
+  tlen += tEncodeSSchemaWrapper(buf, pRsp->schema);
   if (pRsp->pBlockData) {
     sz = taosArrayGetSize(pRsp->pBlockData);
   }
@@ -122,15 +135,15 @@ static FORCE_INLINE int32_t tEncodeSMqPollRsp(void** buf, const SMqPollRsp* pRsp
 
 static FORCE_INLINE void* tDecodeSMqPollRsp(void* buf, SMqPollRsp* pRsp) {
   int32_t sz;
-  buf = taosDecodeFixedI64(buf, &pRsp->consumerId);
+  // buf = taosDecodeFixedI64(buf, &pRsp->consumerId);
   buf = taosDecodeFixedI64(buf, &pRsp->reqOffset);
   buf = taosDecodeFixedI64(buf, &pRsp->rspOffset);
   buf = taosDecodeFixedI32(buf, &pRsp->skipLogNum);
   buf = taosDecodeFixedI32(buf, &pRsp->numOfTopics);
   if (pRsp->numOfTopics == 0) return buf;
-  pRsp->schemas = (SSchemaWrapper*)calloc(1, sizeof(SSchemaWrapper));
-  if (pRsp->schemas == NULL) return NULL;
-  buf = tDecodeSSchemaWrapper(buf, pRsp->schemas);
+  pRsp->schema = (SSchemaWrapper*)calloc(1, sizeof(SSchemaWrapper));
+  if (pRsp->schema == NULL) return NULL;
+  buf = tDecodeSSchemaWrapper(buf, pRsp->schema);
   buf = taosDecodeFixedI32(buf, &sz);
   pRsp->pBlockData = taosArrayInit(sz, sizeof(SSDataBlock));
   for (int32_t i = 0; i < sz; i++) {
@@ -142,13 +155,13 @@ static FORCE_INLINE void* tDecodeSMqPollRsp(void* buf, SMqPollRsp* pRsp) {
 }
 
 static FORCE_INLINE void tDeleteSMqConsumeRsp(SMqPollRsp* pRsp) {
-  if (pRsp->schemas) {
-    if (pRsp->schemas->nCols) {
-      tfree(pRsp->schemas->pSchema);
+  if (pRsp->schema) {
+    if (pRsp->schema->nCols) {
+      tfree(pRsp->schema->pSchema);
     }
-    free(pRsp->schemas);
+    free(pRsp->schema);
   }
-  taosArrayDestroyEx(pRsp->pBlockData, (void (*)(void*))tDeleteSSDataBlock);
+  taosArrayDestroyEx(pRsp->pBlockData, (void (*)(void*))blockDestroyInner);
   pRsp->pBlockData = NULL;
 }
 
@@ -160,10 +173,8 @@ typedef struct SColumn {
     int64_t  dataBlockId;
   };
 
-  union {
-    int16_t colId;
-    int16_t slotId;
-  };
+  int16_t colId;
+  int16_t slotId;
 
   char    name[TSDB_COL_NAME_LEN];
   int8_t  flag;  // column type: normal column, tag, or user-input column (integer/float/string)
@@ -190,7 +201,7 @@ typedef struct SGroupbyExpr {
 
 typedef struct SFunctParam {
   int32_t  type;
-  SColumn *pCol;
+  SColumn* pCol;
   SVariant param;
 } SFunctParam;
 
@@ -208,12 +219,12 @@ typedef struct SResSchame {
 typedef struct SExprBasicInfo {
   SResSchema   resSchema;
   int16_t      numOfParams;  // argument value of each function
-  SFunctParam *pParam;
+  SFunctParam* pParam;
 } SExprBasicInfo;
 
 typedef struct SExprInfo {
-  struct SExprBasicInfo  base;
-  struct tExprNode      *pExpr;
+  struct SExprBasicInfo base;
+  struct tExprNode*     pExpr;
 } SExprInfo;
 
 typedef struct SStateWindow {

include/common/tmsg.h

@@ -1299,7 +1299,7 @@ static FORCE_INLINE SMqRebSubscribe* tNewSMqRebSubscribe(const char* key) {
   if (pRebSub == NULL) {
     goto _err;
   }
-  pRebSub->key = key;
+  pRebSub->key = strdup(key);
   pRebSub->lostConsumers = taosArrayInit(0, sizeof(int64_t));
   if (pRebSub->lostConsumers == NULL) {
     goto _err;
@@ -2138,25 +2138,16 @@ typedef struct {
   int8_t  mqMsgType;
   int32_t code;
   int32_t epoch;
+  int64_t consumerId;
 } SMqRspHead;
 
-typedef struct {
-  int64_t         consumerId;
-  SSchemaWrapper* schemas;
-  int64_t         reqOffset;
-  int64_t         rspOffset;
-  int32_t         skipLogNum;
-  int32_t         numOfTopics;
-  SArray*         pBlockData;  // SArray<SSDataBlock>
-} SMqPollRsp;
-
-// one req for one vg+topic
 typedef struct {
   SMsgHead head;
 
   int64_t consumerId;
   int64_t blockingTime;
   int32_t epoch;
+  int8_t  withSchema;
   char    cgroup[TSDB_CGROUP_LEN];
 
   int64_t currentOffset;
@@ -2175,19 +2166,22 @@ typedef struct {
 } SMqSubTopicEp;
 
 typedef struct {
-  int64_t consumerId;
-  char    cgroup[TSDB_CGROUP_LEN];
-  SArray* topics;  // SArray<SMqSubTopicEp>
-} SMqCMGetSubEpRsp;
+  SMqRspHead head;
+  int64_t    reqOffset;
+  int64_t    rspOffset;
+  int32_t    skipLogNum;
+  // TODO: replace with topic name
+  int32_t numOfTopics;
+  // TODO: remove from msg
+  SSchemaWrapper* schema;
+  SArray*         pBlockData;  // SArray<SSDataBlock>
+} SMqPollRsp;
 
 typedef struct {
   SMqRspHead head;
-  union {
-    SMqPollRsp       consumeRsp;
-    SMqCMGetSubEpRsp getEpRsp;
-  };
-  void* extra;
-} SMqMsgWrapper;
+  char       cgroup[TSDB_CGROUP_LEN];
+  SArray*    topics;  // SArray<SMqSubTopicEp>
+} SMqCMGetSubEpRsp;
 
 typedef struct {
   int32_t curBlock;
@@ -2195,11 +2189,13 @@ typedef struct {
   void**  uData;
 } SMqRowIter;
 
-struct tmq_message_t_v1 {
-  SMqPollRsp rsp;
+struct tmq_message_t {
+  SMqPollRsp msg;
+  void*      vg;
   SMqRowIter iter;
 };
 
+#if 0
 struct tmq_message_t {
   SMqRspHead head;
   union {
@@ -2211,6 +2207,7 @@ struct tmq_message_t {
   int32_t curRow;
   void**  uData;
 };
+#endif
 
 static FORCE_INLINE void tDeleteSMqSubTopicEp(SMqSubTopicEp* pSubTopicEp) { taosArrayDestroy(pSubTopicEp->vgs); }
 
@@ -2263,8 +2260,7 @@ static FORCE_INLINE void* tDecodeSMqSubTopicEp(void* buf, SMqSubTopicEp* pTopicE
 
 static FORCE_INLINE int32_t tEncodeSMqCMGetSubEpRsp(void** buf, const SMqCMGetSubEpRsp* pRsp) {
   int32_t tlen = 0;
-  tlen += taosEncodeFixedI64(buf, pRsp->consumerId);
-  tlen += taosEncodeString(buf, pRsp->cgroup);
+  // tlen += taosEncodeString(buf, pRsp->cgroup);
   int32_t sz = taosArrayGetSize(pRsp->topics);
   tlen += taosEncodeFixedI32(buf, sz);
   for (int32_t i = 0; i < sz; i++) {
@@ -2275,8 +2271,7 @@ static FORCE_INLINE int32_t tEncodeSMqCMGetSubEpRsp(void** buf, const SMqCMGetSu
 }
 
 static FORCE_INLINE void* tDecodeSMqCMGetSubEpRsp(void* buf, SMqCMGetSubEpRsp* pRsp) {
-  buf = taosDecodeFixedI64(buf, &pRsp->consumerId);
-  buf = taosDecodeStringTo(buf, pRsp->cgroup);
+  // buf = taosDecodeStringTo(buf, pRsp->cgroup);
   int32_t sz;
   buf = taosDecodeFixedI32(buf, &sz);
   pRsp->topics = taosArrayInit(sz, sizeof(SMqSubTopicEp));
@@ -2297,8 +2292,8 @@ enum {
 };
 
 typedef struct {
-  void*  inputHandle;
-  void** executor;
+  void* inputHandle;
+  void* executor[4];
 } SStreamTaskParRunner;
 
 typedef struct {
@@ -2352,14 +2347,6 @@ typedef struct {
 
 #pragma pack(pop)
 
-struct SRpcMsg;
-struct SEpSet;
-struct SMgmtWrapper;
-typedef int32_t (*PutToQueueFp)(struct SMgmtWrapper* pWrapper, struct SRpcMsg* pReq);
-typedef int32_t (*SendReqFp)(struct SMgmtWrapper* pWrapper, struct SEpSet* epSet, struct SRpcMsg* pReq);
-typedef int32_t (*SendMnodeReqFp)(struct SMgmtWrapper* pWrapper, struct SRpcMsg* pReq);
-typedef void (*SendRspFp)(struct SMgmtWrapper* pWrapper, struct SRpcMsg* pRsp);
-
 #ifdef __cplusplus
 }
 #endif

include/common/tmsgcb.h

+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_COMMON_MSG_CB_H_
+#define _TD_COMMON_MSG_CB_H_
+
+#include "os.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct SRpcMsg      SRpcMsg;
+typedef struct SEpSet       SEpSet;
+typedef struct SMgmtWrapper SMgmtWrapper;
+
+typedef int32_t (*PutToQueueFp)(SMgmtWrapper* pWrapper, SRpcMsg* pReq);
+typedef int32_t (*SendReqFp)(SMgmtWrapper* pWrapper, SEpSet* epSet, SRpcMsg* pReq);
+typedef int32_t (*SendMnodeReqFp)(SMgmtWrapper* pWrapper, SRpcMsg* pReq);
+typedef void (*SendRspFp)(SMgmtWrapper* pWrapper, SRpcMsg* pRsp);
+
+typedef enum { QUERY_QUEUE, FETCH_QUEUE, WRITE_QUEUE, APPLY_QUEUE, SYNC_QUEUE, QUEUE_MAX } EQueueType;
+
+typedef struct {
+  SMgmtWrapper*  pWrapper;
+  PutToQueueFp   queueFps[QUEUE_MAX];
+  SendReqFp      sendReqFp;
+  SendMnodeReqFp sendMnodeReqFp;
+  SendRspFp      sendRspFp;
+} SMsgCb;
+
+int32_t tmsgPutToQueue(const SMsgCb* pMsgCb, EQueueType qtype, SRpcMsg* pReq);
+int32_t tmsgSendReq(const SMsgCb* pMsgCb, SEpSet* epSet, SRpcMsg* pReq);
+int32_t tmsgSendMnodeReq(const SMsgCb* pMsgCb, SRpcMsg* pReq);
+void    tmsgSendRsp(const SMsgCb* pMsgCb, SRpcMsg* pRsp);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_COMMON_MSG_CB_H_*/

include/dnode/bnode/bnode.h

@@ -16,24 +16,20 @@
 #ifndef _TD_BNODE_H_
 #define _TD_BNODE_H_
 
+#include "tmsgcb.h"
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
 /* ------------------------ TYPES EXPOSED ------------------------ */
-typedef struct SMgmtWrapper SMgmtWrapper;
-typedef struct SBnode       SBnode;
+typedef struct SBnode SBnode;
 
 typedef struct {
 } SBnodeLoad;
 
 typedef struct {
-  int32_t        dnodeId;
-  int64_t        clusterId;
-  SMgmtWrapper  *pWrapper;
-  SendReqFp      sendReqFp;
-  SendMnodeReqFp sendMnodeReqFp;
-  SendRspFp      sendRspFp;
+  SMsgCb msgCb;
 } SBnodeOpt;
 
 /* ------------------------ SBnode ------------------------ */

include/dnode/mnode/mnode.h

@@ -17,27 +17,22 @@
 #define _TD_MND_H_
 
 #include "monitor.h"
+#include "tmsgcb.h"
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
 /* ------------------------ TYPES EXPOSED ------------------------ */
-typedef struct SMgmtWrapper SMgmtWrapper;
-typedef struct SMnode       SMnode;
+typedef struct SMnode SMnode;
 
 typedef struct {
-  int32_t        dnodeId;
-  int64_t        clusterId;
-  int8_t         replica;
-  int8_t         selfIndex;
-  SReplica       replicas[TSDB_MAX_REPLICA];
-  SMgmtWrapper  *pWrapper;
-  PutToQueueFp   putToWriteQFp;
-  PutToQueueFp   putToReadQFp;
-  SendReqFp      sendReqFp;
-  SendMnodeReqFp sendMnodeReqFp;
-  SendRspFp      sendRspFp;
+  int32_t  dnodeId;
+  int64_t  clusterId;
+  int8_t   replica;
+  int8_t   selfIndex;
+  SReplica replicas[TSDB_MAX_REPLICA];
+  SMsgCb   msgCb;
 } SMnodeOpt;
 
 /* ------------------------ SMnode ------------------------ */

include/dnode/qnode/qnode.h

@@ -16,13 +16,14 @@
 #ifndef _TD_QNODE_H_
 #define _TD_QNODE_H_
 
+#include "tmsgcb.h"
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
 /* ------------------------ TYPES EXPOSED ------------------------ */
-typedef struct SMgmtWrapper SMgmtWrapper;
-typedef struct SQnode       SQnode;
+typedef struct SQnode SQnode;
 
 typedef struct {
   int64_t numOfStartTask;
@@ -36,12 +37,7 @@ typedef struct {
 } SQnodeLoad;
 
 typedef struct {
-  int32_t        dnodeId;
-  int64_t        clusterId;
-  SMgmtWrapper  *pWrapper;
-  SendReqFp      sendReqFp;
-  SendMnodeReqFp sendMnodeReqFp;
-  SendRspFp      sendRspFp;
+  SMsgCb msgCb;
 } SQnodeOpt;
 
 /* ------------------------ SQnode ------------------------ */
@@ -74,10 +70,9 @@ int32_t qndGetLoad(SQnode *pQnode, SQnodeLoad *pLoad);
  *
  * @param pQnode The qnode object.
  * @param pMsg The request message
- * @param pRsp The response message
- * @return int32_t 0 for success, -1 for failure
  */
-int32_t qndProcessMsg(SQnode *pQnode, SRpcMsg *pMsg, SRpcMsg **pRsp);
+int32_t qndProcessQueryMsg(SQnode *pQnode, SRpcMsg *pMsg);
+int32_t qndProcessFetchMsg(SQnode *pQnode, SRpcMsg *pMsg);
 
 #ifdef __cplusplus
 }

include/dnode/snode/snode.h

@@ -16,7 +16,7 @@
 #ifndef _TD_SNODE_H_
 #define _TD_SNODE_H_
 
-#include "tcommon.h"
+#include "tmsgcb.h"
 #include "tmsg.h"
 #include "trpc.h"
 
@@ -25,20 +25,14 @@ extern "C" {
 #endif
 
 /* ------------------------ TYPES EXPOSED ------------------------ */
-typedef struct SMgmtWrapper SMgmtWrapper;
-typedef struct SSnode       SSnode;
+typedef struct SSnode SSnode;
 
 typedef struct {
   int32_t reserved;
 } SSnodeLoad;
 
 typedef struct {
-  int32_t        dnodeId;
-  int64_t        clusterId;
-  SMgmtWrapper  *pWrapper;
-  SendReqFp      sendReqFp;
-  SendMnodeReqFp sendMnodeReqFp;
-  SendRspFp      sendRspFp;
+  SMsgCb msgCb;
 } SSnodeOpt;
 
 /* ------------------------ SSnode ------------------------ */

include/libs/qcom/query.h

@@ -25,7 +25,7 @@ extern "C" {
 #include "tlog.h"
 #include "tmsg.h"
 
-enum {
+typedef enum {
   JOB_TASK_STATUS_NULL = 0,
   JOB_TASK_STATUS_NOT_START = 1,
   JOB_TASK_STATUS_EXECUTING,
@@ -35,12 +35,12 @@ enum {
   JOB_TASK_STATUS_CANCELLING,
   JOB_TASK_STATUS_CANCELLED,
   JOB_TASK_STATUS_DROPPING,
-};
+} EJobTaskType;
 
-enum {
+typedef enum {
   TASK_TYPE_PERSISTENT = 1,
   TASK_TYPE_TEMP,
-};
+} ETaskType;
 
 typedef struct STableComInfo {
   uint8_t numOfTags;     // the number of tags in schema

include/libs/qworker/qworker.h

@@ -20,6 +20,7 @@
 extern "C" {
 #endif
 
+#include "tmsgcb.h"
 #include "trpc.h"
 
 
@@ -48,11 +49,7 @@ typedef struct {
   uint64_t numOfErrors;
 } SQWorkerStat;
 
-typedef int32_t (*putReqToQueryQFp)(void *, struct SRpcMsg *);
-typedef int32_t (*sendReqFp)(void *, struct SEpSet *, struct SRpcMsg *);
-
-int32_t qWorkerInit(int8_t nodeType, int32_t nodeId, SQWorkerCfg *cfg, void **qWorkerMgmt, void *nodeObj,
-                    putReqToQueryQFp fp1, sendReqFp fp2);
+int32_t qWorkerInit(int8_t nodeType, int32_t nodeId, SQWorkerCfg *cfg, void **qWorkerMgmt, const SMsgCb *pMsgCb);
 
 int32_t qWorkerProcessQueryMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg);
 

include/os/osDir.h

@@ -17,6 +17,7 @@
 #define _TD_OS_DIR_H_
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define opendir OPENDIR_FUNC_TAOS_FORBID
     #define readdir READDIR_FUNC_TAOS_FORBID

include/os/osFile.h

@@ -23,6 +23,7 @@ extern "C" {
 #include "osSocket.h"
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following sectio
 #ifndef ALLOW_FORBID_FUNC
     #define open OPEN_FUNC_TAOS_FORBID
     #define fopen FOPEN_FUNC_TAOS_FORBID

include/os/osLocale.h

@@ -23,6 +23,7 @@ extern "C" {
 #endif
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define setlocale SETLOCALE_FUNC_TAOS_FORBID
 #endif

include/os/osRand.h

@@ -21,6 +21,7 @@ extern "C" {
 #endif
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define rand RAND_FUNC_TAOS_FORBID
     #define srand SRAND_FUNC_TAOS_FORBID

include/os/osSleep.h

@@ -21,6 +21,7 @@ extern "C" {
 #endif
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define Sleep SLEEP_FUNC_TAOS_FORBID
     #define sleep SLEEP_FUNC_TAOS_FORBID

include/os/osSocket.h

@@ -17,6 +17,7 @@
 #define _TD_OS_SOCKET_H_
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define socket SOCKET_FUNC_TAOS_FORBID
     #define bind   BIND_FUNC_TAOS_FORBID
@@ -52,9 +53,6 @@ extern "C" {
 
 #if (defined(_TD_WINDOWS_64) || defined(_TD_WINDOWS_32)) 
   #define htobe64 htonll
-  #if defined(_TD_GO_DLL_)
-    uint64_t htonll(uint64_t val);
-  #endif
 #endif
 
 #if defined(_TD_DARWIN_64)

include/os/osString.h

@@ -24,6 +24,7 @@ typedef wchar_t TdWchar;
 typedef int32_t TdUcs4;
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define iconv_open ICONV_OPEN_FUNC_TAOS_FORBID
     #define iconv_close ICONV_CLOSE_FUNC_TAOS_FORBID

include/os/osSystem.h

@@ -21,6 +21,7 @@ extern "C" {
 #endif
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define popen POPEN_FUNC_TAOS_FORBID
     #define pclose PCLOSE_FUNC_TAOS_FORBID

include/os/osThread.h

@@ -37,6 +37,7 @@ typedef pthread_condattr_t TdThreadCondAttr;
 #define taosThreadCleanupPop pthread_cleanup_pop
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define pthread_t PTHREAD_T_TYPE_TAOS_FORBID
     #define pthread_spinlock_t PTHREAD_SPINLOCK_T_TYPE_TAOS_FORBID

include/os/osTime.h

@@ -21,6 +21,7 @@ extern "C" {
 #endif
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define strptime STRPTIME_FUNC_TAOS_FORBID
     #define gettimeofday GETTIMEOFDAY_FUNC_TAOS_FORBID
@@ -33,11 +34,7 @@ extern "C" {
 
   #define CLOCK_REALTIME 	0
 
-  #ifdef _TD_GO_DLL_
-    #define MILLISECOND_PER_SECOND (1000LL)
-  #else
-    #define MILLISECOND_PER_SECOND (1000i64)
-  #endif
+  #define MILLISECOND_PER_SECOND (1000i64)
 #else
   #define MILLISECOND_PER_SECOND ((int64_t)1000L)
 #endif

include/os/osTimer.h

@@ -21,6 +21,7 @@ extern "C" {
 #endif
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define timer_create TIMER_CREATE_FUNC_TAOS_FORBID
     #define timer_settime TIMER_SETTIME_FUNC_TAOS_FORBID

include/os/osTimezone.h

@@ -21,6 +21,7 @@ extern "C" {
 #endif
 
 // If the error is in a third-party library, place this header file under the third-party library header file.
+// When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
     #define tzset TZSET_FUNC_TAOS_FORBID
 #endif

include/util/tqueue.h

@@ -70,10 +70,7 @@ int32_t    taosGetQueueNumber(STaosQset *qset);
 
 int32_t taosReadQitemFromQset(STaosQset *qset, void **ppItem, void **ahandle, FItem *itemFp);
 int32_t taosReadAllQitemsFromQset(STaosQset *qset, STaosQall *qall, void **ahandle, FItems *itemsFp);
-
-int32_t taosReadQitemFromQsetByThread(STaosQset *qset, void **ppItem, void **ahandle, FItem *itemFp, int32_t threadId);
 void    taosResetQsetThread(STaosQset *qset, void *pItem);
-
 int32_t taosGetQueueItemsNumber(STaosQueue *queue);
 int32_t taosGetQsetItemsNumber(STaosQset *qset);
 

include/util/tworker.h

@@ -27,33 +27,33 @@ typedef struct SWWorkerPool SWWorkerPool;
 
 typedef struct SQWorker {
   int32_t       id;      // worker ID
-  TdThread     thread;  // thread
+  TdThread      thread;  // thread
   SQWorkerPool *pool;
-} SQWorker, SFWorker;
+} SQWorker;
 
 typedef struct SQWorkerPool {
-  int32_t         max;  // max number of workers
-  int32_t         min;  // min number of workers
-  int32_t         num;  // current number of workers
-  STaosQset      *qset;
-  const char     *name;
-  SQWorker       *workers;
+  int32_t       max;  // max number of workers
+  int32_t       min;  // min number of workers
+  int32_t       num;  // current number of workers
+  STaosQset    *qset;
+  const char   *name;
+  SQWorker     *workers;
   TdThreadMutex mutex;
-} SQWorkerPool, SFWorkerPool;
+} SQWorkerPool;
 
 typedef struct SWWorker {
   int32_t       id;      // worker id
-  TdThread     thread;  // thread
+  TdThread      thread;  // thread
   STaosQall    *qall;
   STaosQset    *qset;  // queue set
   SWWorkerPool *pool;
 } SWWorker;
 
 typedef struct SWWorkerPool {
-  int32_t         max;     // max number of workers
-  int32_t         nextId;  // from 0 to max-1, cyclic
-  const char     *name;
-  SWWorker       *workers;
+  int32_t       max;     // max number of workers
+  int32_t       nextId;  // from 0 to max-1, cyclic
+  const char   *name;
+  SWWorker     *workers;
   TdThreadMutex mutex;
 } SWWorkerPool;
 
@@ -62,16 +62,43 @@ void        tQWorkerCleanup(SQWorkerPool *pool);
 STaosQueue *tQWorkerAllocQueue(SQWorkerPool *pool, void *ahandle, FItem fp);
 void        tQWorkerFreeQueue(SQWorkerPool *pool, STaosQueue *queue);
 
-int32_t     tFWorkerInit(SFWorkerPool *pool);
-void        tFWorkerCleanup(SFWorkerPool *pool);
-STaosQueue *tFWorkerAllocQueue(SFWorkerPool *pool, void *ahandle, FItem fp);
-void        tFWorkerFreeQueue(SFWorkerPool *pool, STaosQueue *queue);
-
 int32_t     tWWorkerInit(SWWorkerPool *pool);
 void        tWWorkerCleanup(SWWorkerPool *pool);
 STaosQueue *tWWorkerAllocQueue(SWWorkerPool *pool, void *ahandle, FItems fp);
 void        tWWorkerFreeQueue(SWWorkerPool *pool, STaosQueue *queue);
 
+typedef struct {
+  const char *name;
+  int32_t     minNum;
+  int32_t     maxNum;
+  FItem       fp;
+  void       *param;
+} SQWorkerAllCfg;
+
+typedef struct {
+  const char  *name;
+  STaosQueue  *queue;
+  SQWorkerPool pool;
+} SQWorkerAll;
+
+typedef struct {
+  const char *name;
+  int32_t     maxNum;
+  FItems      fp;
+  void       *param;
+} SWWorkerAllCfg;
+
+typedef struct {
+  const char  *name;
+  STaosQueue  *queue;
+  SWWorkerPool pool;
+} SWWorkerAll;
+
+int32_t tQWorkerAllInit(SQWorkerAll *pWorker, const SQWorkerAllCfg *pCfg);
+void    tQWorkerAllCleanup(SQWorkerAll *pWorker);
+int32_t tWWorkerAllInit(SWWorkerAll *pWorker, const SWWorkerAllCfg *pCfg);
+void    tWWorkerAllCleanup(SWWorkerAll *pWorker);
+
 #ifdef __cplusplus
 }
 #endif

source/client/src/clientEnv.c

@@ -48,8 +48,8 @@ static void registerRequest(SRequestObj *pRequest) {
   if (pTscObj->pAppInfo) {
     SInstanceSummary *pSummary = &pTscObj->pAppInfo->summary;
 
-    int32_t total = atomic_add_fetch_32(&pSummary->totalRequests, 1);
-    int32_t currentInst = atomic_add_fetch_32(&pSummary->currentRequests, 1);
+    int32_t total = atomic_add_fetch_64(&pSummary->totalRequests, 1);
+    int32_t currentInst = atomic_add_fetch_64(&pSummary->currentRequests, 1);
     tscDebug("0x%" PRIx64 " new Request from connObj:0x%" PRIx64
              ", current:%d, app current:%d, total:%d, reqId:0x%" PRIx64,
              pRequest->self, pRequest->pTscObj->id, num, currentInst, total, pRequest->requestId);
@@ -62,7 +62,7 @@ static void deregisterRequest(SRequestObj *pRequest) {
   STscObj *         pTscObj = pRequest->pTscObj;
   SInstanceSummary *pActivity = &pTscObj->pAppInfo->summary;
 
-  int32_t currentInst = atomic_sub_fetch_32(&pActivity->currentRequests, 1);
+  int32_t currentInst = atomic_sub_fetch_64(&pActivity->currentRequests, 1);
   int32_t num = atomic_sub_fetch_32(&pTscObj->numOfReqs, 1);
 
   int64_t duration = taosGetTimestampMs() - pRequest->metric.start;

source/client/src/tmq.c

@@ -683,7 +683,7 @@ static char* formatTimestamp(char* buf, int64_t val, int precision) {
 
 int32_t tmqGetSkipLogNum(tmq_message_t* tmq_message) {
   if (tmq_message == NULL) return 0;
-  SMqPollRsp* pRsp = &tmq_message->consumeRsp;
+  SMqPollRsp* pRsp = &tmq_message->msg;
   return pRsp->skipLogNum;
 }
 
@@ -692,15 +692,15 @@ void tmqShowMsg(tmq_message_t* tmq_message) {
 
   static bool noPrintSchema;
   char        pBuf[128];
-  SMqPollRsp* pRsp = &tmq_message->consumeRsp;
-  int32_t     colNum = pRsp->schemas->nCols;
+  SMqPollRsp* pRsp = &tmq_message->msg;
+  int32_t     colNum = pRsp->schema->nCols;
   if (!noPrintSchema) {
     printf("|");
     for (int32_t i = 0; i < colNum; i++) {
       if (i == 0)
-        printf(" %25s |", pRsp->schemas->pSchema[i].name);
+        printf(" %25s |", pRsp->schema->pSchema[i].name);
       else
-        printf(" %15s |", pRsp->schemas->pSchema[i].name);
+        printf(" %15s |", pRsp->schema->pSchema[i].name);
     }
     printf("\n");
     printf("===============================================\n");
@@ -737,7 +737,7 @@ int32_t tmqPollCb(void* param, const SDataBuf* pMsg, int32_t code) {
   SMqClientVg*    pVg = pParam->pVg;
   tmq_t*          tmq = pParam->tmq;
   if (code != 0) {
-    printf("msg discard\n");
+    printf("msg discard %x\n", code);
     goto WRITE_QUEUE_FAIL;
   }
 
@@ -780,19 +780,19 @@ int32_t tmqPollCb(void* param, const SDataBuf* pMsg, int32_t code) {
     goto WRITE_QUEUE_FAIL;
   }
   memcpy(pRsp, pMsg->pData, sizeof(SMqRspHead));
-  tDecodeSMqPollRsp(POINTER_SHIFT(pMsg->pData, sizeof(SMqRspHead)), &pRsp->consumeRsp);
-  pRsp->curBlock = 0;
-  pRsp->curRow = 0;
+  tDecodeSMqPollRsp(POINTER_SHIFT(pMsg->pData, sizeof(SMqRspHead)), &pRsp->msg);
+  pRsp->iter.curBlock = 0;
+  pRsp->iter.curRow = 0;
   // TODO: alloc mem
   /*pRsp->*/
   /*printf("rsp commit off:%ld rsp off:%ld has data:%d\n", pRsp->committedOffset, pRsp->rspOffset, pRsp->numOfTopics);*/
-  if (pRsp->consumeRsp.numOfTopics == 0) {
+  if (pRsp->msg.numOfTopics == 0) {
     /*printf("no data\n");*/
     taosFreeQitem(pRsp);
     goto WRITE_QUEUE_FAIL;
   }
 
-  pRsp->extra = pParam->pVg;
+  pRsp->vg = pParam->pVg;
   taosWriteQitem(tmq->mqueue, pRsp);
   atomic_add_fetch_32(&tmq->readyRequest, 1);
   tsem_post(&tmq->rspSem);
@@ -862,14 +862,14 @@ int32_t tmqAskEpCb(void* param, const SDataBuf* pMsg, int32_t code) {
     }
     tDeleteSMqCMGetSubEpRsp(&rsp);
   } else {
-    tmq_message_t* pRsp = taosAllocateQitem(sizeof(tmq_message_t));
+    SMqCMGetSubEpRsp* pRsp = taosAllocateQitem(sizeof(SMqCMGetSubEpRsp));
     if (pRsp == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
       code = -1;
       goto END;
     }
     memcpy(pRsp, pMsg->pData, sizeof(SMqRspHead));
-    tDecodeSMqCMGetSubEpRsp(POINTER_SHIFT(pMsg->pData, sizeof(SMqRspHead)), &pRsp->getEpRsp);
+    tDecodeSMqCMGetSubEpRsp(POINTER_SHIFT(pMsg->pData, sizeof(SMqRspHead)), pRsp);
 
     taosWriteQitem(tmq->mqueue, pRsp);
     tsem_post(&tmq->rspSem);
@@ -985,6 +985,7 @@ SMqPollReq* tmqBuildConsumeReqImpl(tmq_t* tmq, int64_t blockingTime, SMqClientTo
   return pReq;
 }
 
+#if 0
 tmq_message_t* tmqSyncPollImpl(tmq_t* tmq, int64_t blockingTime) {
   tmq_message_t* msg = NULL;
   for (int i = 0; i < taosArrayGetSize(tmq->clientTopics); i++) {
@@ -1052,6 +1053,7 @@ tmq_message_t* tmqSyncPollImpl(tmq_t* tmq, int64_t blockingTime) {
   }
   return NULL;
 }
+#endif
 
 int32_t tmqPollImpl(tmq_t* tmq, int64_t blockingTime) {
   /*printf("call poll\n");*/
@@ -1113,11 +1115,12 @@ int32_t tmqPollImpl(tmq_t* tmq, int64_t blockingTime) {
 }
 
 // return
-int32_t tmqHandleRes(tmq_t* tmq, tmq_message_t* rspMsg, bool* pReset) {
-  if (rspMsg->head.mqMsgType == TMQ_MSG_TYPE__EP_RSP) {
+int32_t tmqHandleRes(tmq_t* tmq, SMqRspHead* rspHead, bool* pReset) {
+  if (rspHead->mqMsgType == TMQ_MSG_TYPE__EP_RSP) {
     /*printf("ep %d %d\n", rspMsg->head.epoch, tmq->epoch);*/
-    if (rspMsg->head.epoch > atomic_load_32(&tmq->epoch)) {
-      tmqUpdateEp(tmq, rspMsg->head.epoch, &rspMsg->getEpRsp);
+    if (rspHead->epoch > atomic_load_32(&tmq->epoch)) {
+      SMqCMGetSubEpRsp* rspMsg = (SMqCMGetSubEpRsp*)rspHead;
+      tmqUpdateEp(tmq, rspHead->epoch, rspMsg);
       tmqClearUnhandleMsg(tmq);
       *pReset = true;
     } else {
@@ -1131,21 +1134,22 @@ int32_t tmqHandleRes(tmq_t* tmq, tmq_message_t* rspMsg, bool* pReset) {
 
 tmq_message_t* tmqHandleAllRsp(tmq_t* tmq, int64_t blockingTime, bool pollIfReset) {
   while (1) {
-    tmq_message_t* rspMsg = NULL;
-    taosGetQitem(tmq->qall, (void**)&rspMsg);
-    if (rspMsg == NULL) {
+    SMqRspHead* rspHead = NULL;
+    taosGetQitem(tmq->qall, (void**)&rspHead);
+    if (rspHead == NULL) {
       taosReadAllQitems(tmq->mqueue, tmq->qall);
-      taosGetQitem(tmq->qall, (void**)&rspMsg);
-      if (rspMsg == NULL) return NULL;
+      taosGetQitem(tmq->qall, (void**)&rspHead);
+      if (rspHead == NULL) return NULL;
     }
 
-    if (rspMsg->head.mqMsgType == TMQ_MSG_TYPE__POLL_RSP) {
+    if (rspHead->mqMsgType == TMQ_MSG_TYPE__POLL_RSP) {
+      tmq_message_t* rspMsg = (tmq_message_t*)rspHead;
       atomic_sub_fetch_32(&tmq->readyRequest, 1);
       /*printf("handle poll rsp %d\n", rspMsg->head.mqMsgType);*/
-      if (rspMsg->head.epoch == atomic_load_32(&tmq->epoch)) {
+      if (rspMsg->msg.head.epoch == atomic_load_32(&tmq->epoch)) {
         /*printf("epoch match\n");*/
-        SMqClientVg* pVg = rspMsg->extra;
-        pVg->currentOffset = rspMsg->consumeRsp.rspOffset;
+        SMqClientVg* pVg = rspMsg->vg;
+        pVg->currentOffset = rspMsg->msg.rspOffset;
         atomic_store_32(&pVg->vgStatus, TMQ_VG_STATUS__IDLE);
         return rspMsg;
       } else {
@@ -1155,8 +1159,8 @@ tmq_message_t* tmqHandleAllRsp(tmq_t* tmq, int64_t blockingTime, bool pollIfRese
     } else {
       /*printf("handle ep rsp %d\n", rspMsg->head.mqMsgType);*/
       bool reset = false;
-      tmqHandleRes(tmq, rspMsg, &reset);
-      taosFreeQitem(rspMsg);
+      tmqHandleRes(tmq, rspHead, &reset);
+      taosFreeQitem(rspHead);
       if (pollIfReset && reset) {
         printf("reset and repoll\n");
         tmqPollImpl(tmq, blockingTime);
@@ -1165,6 +1169,7 @@ tmq_message_t* tmqHandleAllRsp(tmq_t* tmq, int64_t blockingTime, bool pollIfRese
   }
 }
 
+#if 0
 tmq_message_t* tmq_consumer_poll_v1(tmq_t* tmq, int64_t blocking_time) {
   tmq_message_t* rspMsg = NULL;
   int64_t        startTime = taosGetTimestampMs();
@@ -1187,6 +1192,7 @@ tmq_message_t* tmq_consumer_poll_v1(tmq_t* tmq, int64_t blocking_time) {
       return NULL;
   }
 }
+#endif
 
 tmq_message_t* tmq_consumer_poll(tmq_t* tmq, int64_t blocking_time) {
   tmq_message_t* rspMsg;
@@ -1352,7 +1358,7 @@ tmq_resp_err_t tmq_commit(tmq_t* tmq, const tmq_topic_vgroup_list_t* tmq_topic_v
 
 void tmq_message_destroy(tmq_message_t* tmq_message) {
   if (tmq_message == NULL) return;
-  SMqPollRsp* pRsp = &tmq_message->consumeRsp;
+  SMqPollRsp* pRsp = &tmq_message->msg;
   tDeleteSMqConsumeRsp(pRsp);
   /*free(tmq_message);*/
   taosFreeQitem(tmq_message);
@@ -1368,24 +1374,24 @@ const char* tmq_err2str(tmq_resp_err_t err) {
 }
 
 TAOS_ROW tmq_get_row(tmq_message_t* message) {
-  SMqPollRsp* rsp = &message->consumeRsp;
+  SMqPollRsp* rsp = &message->msg;
   while (1) {
-    if (message->curBlock < taosArrayGetSize(rsp->pBlockData)) {
-      SSDataBlock* pBlock = taosArrayGet(rsp->pBlockData, message->curBlock);
-      if (message->curRow < pBlock->info.rows) {
+    if (message->iter.curBlock < taosArrayGetSize(rsp->pBlockData)) {
+      SSDataBlock* pBlock = taosArrayGet(rsp->pBlockData, message->iter.curBlock);
+      if (message->iter.curRow < pBlock->info.rows) {
         for (int i = 0; i < pBlock->info.numOfCols; i++) {
           SColumnInfoData* pData = taosArrayGet(pBlock->pDataBlock, i);
-          if (colDataIsNull_s(pData, message->curRow))
-            message->uData[i] = NULL;
+          if (colDataIsNull_s(pData, message->iter.curRow))
+            message->iter.uData[i] = NULL;
           else {
-            message->uData[i] = colDataGetData(pData, message->curRow);
+            message->iter.uData[i] = colDataGetData(pData, message->iter.curRow);
           }
         }
-        message->curRow++;
-        return message->uData;
+        message->iter.curRow++;
+        return message->iter.uData;
       } else {
-        message->curBlock++;
-        message->curRow = 0;
+        message->iter.curBlock++;
+        message->iter.curRow = 0;
         continue;
       }
     }

source/common/src/tmsg.c

@@ -2795,6 +2795,8 @@ int32_t tEncodeSStreamTask(SCoder *pEncoder, const SStreamTask *pTask) {
   if (tEncodeI32(pEncoder, pTask->taskId) < 0) return -1;
   if (tEncodeI32(pEncoder, pTask->level) < 0) return -1;
   if (tEncodeI8(pEncoder, pTask->status) < 0) return -1;
+  if (tEncodeI8(pEncoder, pTask->pipeEnd) < 0) return -1;
+  if (tEncodeI8(pEncoder, pTask->parallel) < 0) return -1;
   if (tEncodeSEpSet(pEncoder, &pTask->NextOpEp) < 0) return -1;
   if (tEncodeCStr(pEncoder, pTask->qmsg) < 0) return -1;
   tEndEncode(pEncoder);
@@ -2807,6 +2809,8 @@ int32_t tDecodeSStreamTask(SCoder *pDecoder, SStreamTask *pTask) {
   if (tDecodeI32(pDecoder, &pTask->taskId) < 0) return -1;
   if (tDecodeI32(pDecoder, &pTask->level) < 0) return -1;
   if (tDecodeI8(pDecoder, &pTask->status) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pTask->pipeEnd) < 0) return -1;
+  if (tDecodeI8(pDecoder, &pTask->parallel) < 0) return -1;
   if (tDecodeSEpSet(pDecoder, &pTask->NextOpEp) < 0) return -1;
   if (tDecodeCStrAlloc(pDecoder, &pTask->qmsg) < 0) return -1;
   tEndDecode(pDecoder);

source/common/src/tmsgcb.c

+/*
+ * 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/>.
+ */
+
+#define _DEFAULT_SOURCE
+#include "tmsgcb.h"
+
+int32_t tmsgPutToQueue(const SMsgCb* pMsgCb, EQueueType qtype, SRpcMsg* pReq) {
+  return (*pMsgCb->queueFps[qtype])(pMsgCb->pWrapper, pReq);
+}
+
+int32_t tmsgSendReq(const SMsgCb* pMsgCb, SEpSet* epSet, SRpcMsg* pReq) {
+  return (*pMsgCb->sendReqFp)(pMsgCb->pWrapper, epSet, pReq);
+}
+
+int32_t tmsgSendMnodeReq(const SMsgCb* pMsgCb, SRpcMsg* pReq) {
+  return (*pMsgCb->sendMnodeReqFp)(pMsgCb->pWrapper, pReq);
+}
+
+void tmsgSendRsp(const SMsgCb* pMsgCb, SRpcMsg* pRsp) { return (*pMsgCb->sendRspFp)(pMsgCb->pWrapper, pRsp); }

source/dnode/bnode/inc/bndInt.h

@@ -30,7 +30,7 @@ extern "C" {
 #endif
 
 typedef struct SBnode {
-  SBnodeOpt opt;
+  SMsgCb msgCb;
 } SBnode;
 
 #ifdef __cplusplus

source/dnode/bnode/src/bnode.c

@@ -17,6 +17,7 @@
 
 SBnode *bndOpen(const char *path, const SBnodeOpt *pOption) {
   SBnode *pBnode = calloc(1, sizeof(SBnode));
+  pBnode->msgCb = pOption->msgCb;
   return pBnode;
 }
 

source/dnode/mgmt/bnode/src/bmInt.c

@@ -19,13 +19,12 @@
 static int32_t bmRequire(SMgmtWrapper *pWrapper, bool *required) { return dndReadFile(pWrapper, required); }
 
 static void bmInitOption(SBnodeMgmt *pMgmt, SBnodeOpt *pOption) {
-  SDnode *pDnode = pMgmt->pDnode;
-  pOption->pWrapper = pMgmt->pWrapper;
-  pOption->sendReqFp = dndSendReqToDnode;
-  pOption->sendMnodeReqFp = dndSendReqToMnode;
-  pOption->sendRspFp = dndSendRsp;
-  pOption->dnodeId = pDnode->dnodeId;
-  pOption->clusterId = pDnode->clusterId;
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = pMgmt->pWrapper;
+  msgCb.sendReqFp = dndSendReqToDnode;
+  msgCb.sendMnodeReqFp = dndSendReqToMnode;
+  msgCb.sendRspFp = dndSendRsp;
+  pOption->msgCb = msgCb;
 }
 
 static int32_t bmOpenImp(SBnodeMgmt *pMgmt) {

source/dnode/mgmt/container/inc/dnd.h

@@ -89,6 +89,9 @@ typedef struct {
 } SDnodeWorker;
 
 typedef struct SMsgHandle {
+  int32_t       vgId;
+  NodeMsgFp     vgIdMsgFp;
+  SMgmtWrapper *pVgIdWrapper;  // Handle the case where the same message type is distributed to qnode or vnode
   NodeMsgFp     msgFp;
   SMgmtWrapper *pWrapper;
 } SMsgHandle;
@@ -114,6 +117,7 @@ typedef struct SMgmtWrapper {
   void       *pMgmt;
   SDnode     *pDnode;
   NodeMsgFp   msgFps[TDMT_MAX];
+  int32_t     msgVgIds[TDMT_MAX];  // Handle the case where the same message type is distributed to qnode or vnode
   SMgmtFp     fp;
 } SMgmtWrapper;
 
@@ -149,7 +153,7 @@ typedef struct SDnode {
 EDndStatus    dndGetStatus(SDnode *pDnode);
 void          dndSetStatus(SDnode *pDnode, EDndStatus stat);
 SMgmtWrapper *dndAcquireWrapper(SDnode *pDnode, ENodeType nodeType);
-void          dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp);
+void          dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId);
 void          dndReportStartup(SDnode *pDnode, char *pName, char *pDesc);
 void          dndSendMonitorReport(SDnode *pDnode);
 

source/dnode/mgmt/container/src/dndInt.c

@@ -65,8 +65,9 @@ void dndCleanup() {
   dInfo("dnode env is cleaned up");
 }
 
-void dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp) {
+void dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId) {
   pWrapper->msgFps[TMSG_INDEX(msgType)] = nodeMsgFp;
+  pWrapper->msgVgIds[TMSG_INDEX(msgType)] = vgId;
 }
 
 EDndStatus dndGetStatus(SDnode *pDnode) { return pDnode->status; }

source/dnode/mgmt/container/src/dndTransport.c

@@ -20,13 +20,26 @@
 #define INTERNAL_CKEY   "_key"
 #define INTERNAL_SECRET "_pwd"
 
+static inline void dndProcessQVnodeRpcMsg(SMsgHandle *pHandle, SRpcMsg *pMsg, SEpSet *pEpSet) {
+  SMsgHead *pHead = pMsg->pCont;
+  int32_t   vgId = htonl(pHead->vgId);
+
+  SMgmtWrapper *pWrapper = pHandle->pWrapper;
+  if (vgId == pHandle->vgId && pHandle->pVgIdWrapper != NULL) {
+    pWrapper = pHandle->pVgIdWrapper;
+  }
+
+  dTrace("msg:%s will be processed by %s, handle:%p app:%p vgId:%d", TMSG_INFO(pMsg->msgType), pWrapper->name,
+         pMsg->handle, pMsg->ahandle, vgId);
+  dndProcessRpcMsg(pWrapper, pMsg, pEpSet);
+}
+
 static void dndProcessResponse(void *parent, SRpcMsg *pRsp, SEpSet *pEpSet) {
   SDnode     *pDnode = parent;
   STransMgmt *pMgmt = &pDnode->trans;
   tmsg_t      msgType = pRsp->msgType;
 
   if (dndGetStatus(pDnode) != DND_STAT_RUNNING) {
-    // if (pRsp == NULL || pRsp->pCont == NULL) return;
     dTrace("rsp:%s ignored since dnode not running, handle:%p app:%p", TMSG_INFO(msgType), pRsp->handle, pRsp->ahandle);
     rpcFreeCont(pRsp->pCont);
     return;
@@ -34,9 +47,13 @@ static void dndProcessResponse(void *parent, SRpcMsg *pRsp, SEpSet *pEpSet) {
 
   SMsgHandle *pHandle = &pMgmt->msgHandles[TMSG_INDEX(msgType)];
   if (pHandle->msgFp != NULL) {
-    dTrace("rsp:%s will be processed by %s, handle:%p app:%p code:0x%04x:%s", TMSG_INFO(msgType),
-           pHandle->pWrapper->name, pRsp->handle, pRsp->ahandle, pRsp->code & 0XFFFF, tstrerror(pRsp->code));
-    dndProcessRpcMsg(pHandle->pWrapper, pRsp, pEpSet);
+    if (pHandle->vgId == 0) {
+      dTrace("rsp:%s will be processed by %s, handle:%p app:%p code:0x%04x:%s", TMSG_INFO(msgType),
+             pHandle->pWrapper->name, pRsp->handle, pRsp->ahandle, pRsp->code & 0XFFFF, tstrerror(pRsp->code));
+      dndProcessRpcMsg(pHandle->pWrapper, pRsp, pEpSet);
+    } else {
+      dndProcessQVnodeRpcMsg(pHandle, pRsp, pEpSet);
+    }
   } else {
     dError("rsp:%s not processed since no handle, handle:%p app:%p", TMSG_INFO(msgType), pRsp->handle, pRsp->ahandle);
     rpcFreeCont(pRsp->pCont);
@@ -110,9 +127,13 @@ static void dndProcessRequest(void *param, SRpcMsg *pReq, SEpSet *pEpSet) {
 
   SMsgHandle *pHandle = &pMgmt->msgHandles[TMSG_INDEX(msgType)];
   if (pHandle->msgFp != NULL) {
-    dTrace("req:%s will be processed by %s, handle:%p app:%p", TMSG_INFO(msgType), pHandle->pWrapper->name,
-           pReq->handle, pReq->ahandle);
-    dndProcessRpcMsg(pHandle->pWrapper, pReq, pEpSet);
+    if (pHandle->vgId == 0) {
+      dTrace("req:%s will be processed by %s, handle:%p app:%p", TMSG_INFO(msgType), pHandle->pWrapper->name,
+             pReq->handle, pReq->ahandle);
+      dndProcessRpcMsg(pHandle->pWrapper, pReq, pEpSet);
+    } else {
+      dndProcessQVnodeRpcMsg(pHandle, pReq, pEpSet);
+    }
   } else {
     dError("req:%s not processed since no handle, handle:%p app:%p", TMSG_INFO(msgType), pReq->handle, pReq->ahandle);
     SRpcMsg rspMsg = {.handle = pReq->handle, .code = TSDB_CODE_MSG_NOT_PROCESSED, .ahandle = pReq->ahandle};
@@ -245,17 +266,24 @@ int32_t dndInitMsgHandle(SDnode *pDnode) {
 
     for (int32_t msgIndex = 0; msgIndex < TDMT_MAX; ++msgIndex) {
       NodeMsgFp msgFp = pWrapper->msgFps[msgIndex];
+      int32_t   vgId = pWrapper->msgVgIds[msgIndex];
       if (msgFp == NULL) continue;
 
       SMsgHandle *pHandle = &pMgmt->msgHandles[msgIndex];
-      if (pHandle->msgFp != NULL) {
-        dError("msg:%s has multiple process nodes, prev node:%s, curr node:%s", tMsgInfo[msgIndex],
-               pHandle->pWrapper->name, pWrapper->name);
+      if (pHandle->msgFp != NULL && pHandle->vgId == vgId) {
+        dError("msg:%s has multiple process nodes, prev node:%s:%d, curr node:%s:%d", tMsgInfo[msgIndex],
+               pHandle->pWrapper->name, pHandle->pWrapper->msgVgIds[msgIndex], pWrapper->name, vgId);
         return -1;
       } else {
-        dTrace("msg:%s will be processed by %s", tMsgInfo[msgIndex], pWrapper->name);
-        pHandle->msgFp = msgFp;
-        pHandle->pWrapper = pWrapper;
+        dTrace("msg:%s will be processed by %s, vgId:%d", tMsgInfo[msgIndex], pWrapper->name, vgId);
+        if (vgId == 0) {
+          pHandle->msgFp = msgFp;
+          pHandle->pWrapper = pWrapper;
+        } else {
+          pHandle->vgId = vgId;
+          pHandle->vgIdMsgFp = msgFp;
+          pHandle->pVgIdWrapper = pWrapper;
+        }
       }
     }
   }

source/dnode/mgmt/dnode/src/dmMsg.c

@@ -114,19 +114,19 @@ int32_t dmProcessConfigReq(SDnodeMgmt *pMgmt, SNodeMsg *pMsg) {
 
 void dmInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by DNODE
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_NETWORK_TEST, (NodeMsgFp)dmProcessMgmtMsg);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_NETWORK_TEST, (NodeMsgFp)dmProcessMgmtMsg, 0);
 
   // Requests handled by MNODE
-  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS_RSP, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT_RSP, (NodeMsgFp)dmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH_RSP, (NodeMsgFp)dmProcessMgmtMsg);
+  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS_RSP, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT_RSP, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH_RSP, (NodeMsgFp)dmProcessMgmtMsg, 0);
 }

source/dnode/mgmt/mnode/inc/mmInt.h

@@ -28,9 +28,9 @@ typedef struct SMnodeMgmt {
   SDnode       *pDnode;
   SMgmtWrapper *pWrapper;
   const char   *path;
-  SDnodeWorker  readWorker;
-  SDnodeWorker  writeWorker;
-  SDnodeWorker  syncWorker;
+  SQWorkerAll   readWorker;
+  SQWorkerAll   writeWorker;
+  SQWorkerAll   syncWorker;
   SReplica      replicas[TSDB_MAX_REPLICA];
   int8_t        replica;
   int8_t        selfIndex;

source/dnode/mgmt/mnode/src/mmInt.c

@@ -39,14 +39,17 @@ static int32_t mmRequire(SMgmtWrapper *pWrapper, bool *required) {
 
 static void mmInitOption(SMnodeMgmt *pMgmt, SMnodeOpt *pOption) {
   SDnode *pDnode = pMgmt->pDnode;
-  pOption->pWrapper = pMgmt->pWrapper;
-  pOption->putToWriteQFp = mmPutMsgToWriteQueue;
-  pOption->putToReadQFp = mmPutMsgToReadQueue;
-  pOption->sendReqFp = dndSendReqToDnode;
-  pOption->sendMnodeReqFp = dndSendReqToMnode;
-  pOption->sendRspFp = dndSendRsp;
   pOption->dnodeId = pDnode->dnodeId;
   pOption->clusterId = pDnode->clusterId;
+
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = pMgmt->pWrapper;
+  msgCb.queueFps[QUERY_QUEUE] = mmPutMsgToReadQueue;
+  msgCb.queueFps[WRITE_QUEUE] = mmPutMsgToWriteQueue;
+  msgCb.sendReqFp = dndSendReqToDnode;
+  msgCb.sendMnodeReqFp = dndSendReqToMnode;
+  msgCb.sendRspFp = dndSendRsp;
+  pOption->msgCb = msgCb;
 }
 
 static void mmBuildOptionForDeploy(SMnodeMgmt *pMgmt, SMnodeOpt *pOption) {

source/dnode/mgmt/mnode/src/mmMsg.c

@@ -75,79 +75,79 @@ int32_t mmProcessAlterReq(SMnodeMgmt *pMgmt, SNodeMsg *pMsg) {
 
 void mmInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by DNODE
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE_RSP, (NodeMsgFp)mmProcessWriteMsg);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
 
   // Requests handled by MNODE
-  dndSetMsgHandle(pWrapper, TDMT_MND_CONNECT, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_ACCT, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_ACCT, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_ACCT, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_USER, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_USER, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_USER, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GET_USER_AUTH, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CONFIG_DNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_MNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_MNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_QNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_QNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_SNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_SNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_BNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_BNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_USE_DB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_DB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SYNC_DB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_COMPACT_DB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_FUNC, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_RETRIEVE_FUNC, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_FUNC, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_STB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_STB, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_TABLE_META, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_VGROUP_LIST, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_QUERY, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_CONN, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_HEARTBEAT, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW_RETRIEVE, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SYSTABLE_RETRIEVE, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_TRANS, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH, (NodeMsgFp)mmProcessReadMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_TOPIC, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_TOPIC, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_TOPIC, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SUBSCRIBE, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_MQ_COMMIT_OFFSET, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GET_SUB_EP, (NodeMsgFp)mmProcessReadMsg);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CONNECT, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_ACCT, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_ACCT, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_ACCT, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_USER, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_USER, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_USER, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GET_USER_AUTH, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CONFIG_DNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_MNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_MNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_QNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_QNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_SNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_SNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_BNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_BNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_USE_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SYNC_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_COMPACT_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_FUNC, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_RETRIEVE_FUNC, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_FUNC, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_TABLE_META, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_VGROUP_LIST, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_QUERY, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_CONN, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_HEARTBEAT, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW_RETRIEVE, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SYSTABLE_RETRIEVE, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_TRANS, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH, (NodeMsgFp)mmProcessReadMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_TOPIC, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_TOPIC, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_TOPIC, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SUBSCRIBE, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_MQ_COMMIT_OFFSET, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GET_SUB_EP, (NodeMsgFp)mmProcessReadMsg, 0);
 
   // Requests handled by VNODE
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB_RSP, (NodeMsgFp)mmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY, (NodeMsgFp)mmProcessWriteMsg);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY, (NodeMsgFp)mmProcessWriteMsg, 0);
 }

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

@@ -42,34 +42,9 @@ static void mmProcessQueue(SMnodeMgmt *pMgmt, SNodeMsg *pMsg) {
   taosFreeQitem(pMsg);
 }
 
-int32_t mmStartWorker(SMnodeMgmt *pMgmt) {
-  if (dndInitWorker(pMgmt, &pMgmt->readWorker, DND_WORKER_SINGLE, "mnode-read", 0, 1, mmProcessQueue) != 0) {
-    dError("failed to start mnode read worker since %s", terrstr());
-    return -1;
-  }
-
-  if (dndInitWorker(pMgmt, &pMgmt->writeWorker, DND_WORKER_SINGLE, "mnode-write", 0, 1, mmProcessQueue) != 0) {
-    dError("failed to start mnode write worker since %s", terrstr());
-    return -1;
-  }
-
-  if (dndInitWorker(pMgmt, &pMgmt->syncWorker, DND_WORKER_SINGLE, "mnode-sync", 0, 1, mmProcessQueue) != 0) {
-    dError("failed to start mnode sync worker since %s", terrstr());
-    return -1;
-  }
-
-  return 0;
-}
-
-void mmStopWorker(SMnodeMgmt *pMgmt) {
-  dndCleanupWorker(&pMgmt->readWorker);
-  dndCleanupWorker(&pMgmt->writeWorker);
-  dndCleanupWorker(&pMgmt->syncWorker);
-}
-
-static int32_t mmPutMsgToWorker(SMnodeMgmt *pMgmt, SDnodeWorker *pWorker, SNodeMsg *pMsg) {
+static int32_t mmPutMsgToWorker(SMnodeMgmt *pMgmt, SQWorkerAll *pWorker, SNodeMsg *pMsg) {
   dTrace("msg:%p, put into worker %s", pMsg, pWorker->name);
-  return dndWriteMsgToWorker(pWorker, pMsg);
+  return taosWriteQitem(pWorker->queue, pMsg);
 }
 
 int32_t mmProcessWriteMsg(SMnodeMgmt *pMgmt, SNodeMsg *pMsg) {
@@ -84,16 +59,16 @@ int32_t mmProcessReadMsg(SMnodeMgmt *pMgmt, SNodeMsg *pMsg) {
   return mmPutMsgToWorker(pMgmt, &pMgmt->readWorker, pMsg);
 }
 
-static int32_t mmPutRpcMsgToWorker(SMnodeMgmt *pMgmt, SDnodeWorker *pWorker, SRpcMsg *pRpc) {
+static int32_t mmPutRpcMsgToWorker(SMnodeMgmt *pMgmt, SQWorkerAll *pWorker, SRpcMsg *pRpc) {
   SNodeMsg *pMsg = taosAllocateQitem(sizeof(SNodeMsg));
   if (pMsg == NULL) {
     return -1;
   }
 
-  dTrace("msg:%p, is created, type:%s", pMsg, TMSG_INFO(pRpc->msgType));
+  dTrace("msg:%p, is created and put into worker:%s, type:%s", pMsg, pWorker->name, TMSG_INFO(pRpc->msgType));
   pMsg->rpcMsg = *pRpc;
 
-  int32_t code = mmPutMsgToWorker(pMgmt, pWorker, pMsg);
+  int32_t code = taosWriteQitem(pWorker->queue, pMsg);
   if (code != 0) {
     dTrace("msg:%p, is freed", pMsg);
     taosFreeQitem(pMsg);
@@ -112,3 +87,30 @@ int32_t mmPutMsgToReadQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
   SMnodeMgmt *pMgmt = pWrapper->pMgmt;
   return mmPutRpcMsgToWorker(pMgmt, &pMgmt->readWorker, pRpc);
 }
+
+int32_t mmStartWorker(SMnodeMgmt *pMgmt) {
+  SQWorkerAllCfg cfg = {.minNum = 0, .maxNum = 1, .name = "mnode-read", .fp = (FItem)mmProcessQueue, .param = pMgmt};
+
+  if (tQWorkerAllInit(&pMgmt->readWorker, &cfg) != 0) {
+    dError("failed to start mnode-read worker since %s", terrstr());
+    return -1;
+  }
+
+  if (tQWorkerAllInit(&pMgmt->writeWorker, &cfg) != 0) {
+    dError("failed to start mnode-write worker since %s", terrstr());
+    return -1;
+  }
+
+  if (tQWorkerAllInit(&pMgmt->syncWorker, &cfg) != 0) {
+    dError("failed to start mnode sync-worker since %s", terrstr());
+    return -1;
+  }
+
+  return 0;
+}
+
+void mmStopWorker(SMnodeMgmt *pMgmt) {
+  tQWorkerAllCleanup(&pMgmt->readWorker);
+  tQWorkerAllCleanup(&pMgmt->writeWorker);
+  tQWorkerAllCleanup(&pMgmt->syncWorker);
+}

source/dnode/mgmt/qnode/inc/qmInt.h

@@ -42,6 +42,9 @@ int32_t qmProcessCreateReq(SMgmtWrapper *pWrapper, SNodeMsg *pMsg);
 int32_t qmProcessDropReq(SMgmtWrapper *pWrapper, SNodeMsg *pMsg);
 
 // qmWorker.c
+int32_t qmPutMsgToQueryQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
+int32_t qmPutMsgToFetchQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
+
 int32_t qmStartWorker(SQnodeMgmt *pMgmt);
 void    qmStopWorker(SQnodeMgmt *pMgmt);
 int32_t qmProcessQueryMsg(SQnodeMgmt *pMgmt, SNodeMsg *pMsg);

source/dnode/mgmt/qnode/src/qmInt.c

@@ -19,13 +19,14 @@
 static int32_t qmRequire(SMgmtWrapper *pWrapper, bool *required) { return dndReadFile(pWrapper, required); }
 
 static void qmInitOption(SQnodeMgmt *pMgmt, SQnodeOpt *pOption) {
-  SDnode *pDnode = pMgmt->pDnode;
-  pOption->pWrapper = pMgmt->pWrapper;
-  pOption->sendReqFp = dndSendReqToDnode;
-  pOption->sendMnodeReqFp = dndSendReqToMnode;
-  pOption->sendRspFp = dndSendRsp;
-  pOption->dnodeId = pDnode->dnodeId;
-  pOption->clusterId = pDnode->clusterId;
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = pMgmt->pWrapper;
+  msgCb.queueFps[QUERY_QUEUE] = qmPutMsgToQueryQueue;
+  msgCb.queueFps[FETCH_QUEUE] = qmPutMsgToFetchQueue;
+  msgCb.sendReqFp = dndSendReqToDnode;
+  msgCb.sendMnodeReqFp = dndSendReqToMnode;
+  msgCb.sendRspFp = dndSendRsp;
+  pOption->msgCb = msgCb;
 }
 
 static int32_t qmOpenImp(SQnodeMgmt *pMgmt) {

source/dnode/mgmt/qnode/src/qmMsg.c

@@ -54,4 +54,16 @@ int32_t qmProcessDropReq(SMgmtWrapper *pWrapper, SNodeMsg *pMsg) {
   }
 }
 
-void qmInitMsgHandles(SMgmtWrapper *pWrapper) {}
+void qmInitMsgHandles(SMgmtWrapper *pWrapper) {
+  // Requests handled by VNODE
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)qmProcessQueryMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)qmProcessQueryMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)qmProcessFetchMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)qmProcessFetchMsg, 1);
+
+  dndSetMsgHandle(pWrapper, TDMT_VND_RES_READY, (NodeMsgFp)qmProcessFetchMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASKS_STATUS, (NodeMsgFp)qmProcessFetchMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_TASK, (NodeMsgFp)qmProcessFetchMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TASK, (NodeMsgFp)qmProcessFetchMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)qmProcessFetchMsg, 1);
+}

source/dnode/mgmt/qnode/src/qmWorker.c

@@ -16,50 +16,87 @@
 #define _DEFAULT_SOURCE
 #include "qmInt.h"
 
-static void qmProcessQueue(SQnodeMgmt *pMgmt, SNodeMsg *pMsg) {
-  dTrace("msg:%p, will be processed in qnode queue", pMsg);
-  SRpcMsg *pRsp = NULL;
-  SRpcMsg *pRpc = &pMsg->rpcMsg;
-  int32_t  code = qndProcessMsg(pMgmt->pQnode, pRpc, &pRsp);
-
-  if (pRpc->msgType & 1u) {
-    if (pRsp != NULL) {
-      pRsp->ahandle = pRpc->ahandle;
-      dndSendRsp(pMgmt->pWrapper, pRsp);
-      free(pRsp);
-    } else {
-      if (code != 0) code = terrno;
-      SRpcMsg rpcRsp = {.handle = pRpc->handle, .ahandle = pRpc->ahandle, .code = code};
-      dndSendRsp(pMgmt->pWrapper, &rpcRsp);
-    }
+static void qmSendRsp(SMgmtWrapper *pWrapper, SNodeMsg *pMsg, int32_t code) {
+  SRpcMsg rsp = {.handle = pMsg->rpcMsg.handle, .ahandle = pMsg->rpcMsg.ahandle, .code = code};
+  dndSendRsp(pWrapper, &rsp);
+}
+
+static void qmProcessQueryQueue(SQnodeMgmt *pMgmt, SNodeMsg *pMsg) {
+  dTrace("msg:%p, will be processed in qnode-query queue", pMsg);
+  int32_t code = qndProcessQueryMsg(pMgmt->pQnode, &pMsg->rpcMsg);
+  if (code != 0) {
+    qmSendRsp(pMgmt->pWrapper, pMsg, code);
   }
 
   dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
-  rpcFreeCont(pRpc->pCont);
+  rpcFreeCont(pMsg->rpcMsg.pCont);
   taosFreeQitem(pMsg);
 }
 
-int32_t qmProcessQueryMsg(SQnodeMgmt *pMgmt, SNodeMsg *pMsg) {
-  SDnodeWorker *pWorker = &pMgmt->queryWorker;
+static void qmProcessFetchQueue(SQnodeMgmt *pMgmt, SNodeMsg *pMsg) {
+  dTrace("msg:%p, will be processed in qnode-fetch queue", pMsg);
+  int32_t code = qndProcessFetchMsg(pMgmt->pQnode, &pMsg->rpcMsg);
+  if (code != 0) {
+    qmSendRsp(pMgmt->pWrapper, pMsg, code);
+  }
+
+  dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
+  rpcFreeCont(pMsg->rpcMsg.pCont);
+  taosFreeQitem(pMsg);
+}
 
-  dTrace("msg:%p, put into worker:%s", pMsg, pWorker->name);
+static int32_t qmPutMsgToWorker(SDnodeWorker *pWorker, SNodeMsg *pMsg) {
+  dTrace("msg:%p, put into worker %s", pMsg, pWorker->name);
   return dndWriteMsgToWorker(pWorker, pMsg);
 }
 
-int32_t qmProcessFetchMsg(SQnodeMgmt *pMgmt, SNodeMsg *pMsg) {
-  SDnodeWorker *pWorker = &pMgmt->fetchWorker;
+int32_t qmProcessQueryMsg(SQnodeMgmt *pMgmt, SNodeMsg *pMsg) { return qmPutMsgToWorker(&pMgmt->queryWorker, pMsg); }
 
-  dTrace("msg:%p, put into worker:%s", pMsg, pWorker->name);
-  return dndWriteMsgToWorker(pWorker, pMsg);
+int32_t qmProcessFetchMsg(SQnodeMgmt *pMgmt, SNodeMsg *pMsg) { return qmPutMsgToWorker(&pMgmt->fetchWorker, pMsg); }
+
+static int32_t qmPutRpcMsgToWorker(SQnodeMgmt *pMgmt, SDnodeWorker *pWorker, SRpcMsg *pRpc) {
+  SNodeMsg *pMsg = taosAllocateQitem(sizeof(SNodeMsg));
+  if (pMsg == NULL) {
+    return -1;
+  }
+
+  dTrace("msg:%p, is created and put into worker:%s, type:%s", pMsg, pWorker->name, TMSG_INFO(pRpc->msgType));
+  pMsg->rpcMsg = *pRpc;
+
+  int32_t code = dndWriteMsgToWorker(pWorker, pMsg);
+  if (code != 0) {
+    dTrace("msg:%p, is freed", pMsg);
+    taosFreeQitem(pMsg);
+    rpcFreeCont(pRpc->pCont);
+  }
+
+  return code;
+}
+
+int32_t qmPutMsgToQueryQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
+  SQnodeMgmt *pMgmt = pWrapper->pMgmt;
+  return qmPutRpcMsgToWorker(pMgmt, &pMgmt->queryWorker, pRpc);
+}
+
+int32_t qmPutMsgToFetchQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
+  SQnodeMgmt *pMgmt = pWrapper->pMgmt;
+  return qmPutRpcMsgToWorker(pMgmt, &pMgmt->fetchWorker, pRpc);
 }
 
 int32_t qmStartWorker(SQnodeMgmt *pMgmt) {
-  if (dndInitWorker(pMgmt, &pMgmt->queryWorker, DND_WORKER_SINGLE, "qnode-query", 0, 1, qmProcessQueue) != 0) {
+  int32_t maxFetchThreads = 4;
+  int32_t minFetchThreads = TMIN(maxFetchThreads, tsNumOfCores);
+  int32_t minQueryThreads = TMAX((int32_t)(tsNumOfCores * tsRatioOfQueryCores), 1);
+  int32_t maxQueryThreads = minQueryThreads;
+
+  if (dndInitWorker(pMgmt, &pMgmt->queryWorker, DND_WORKER_SINGLE, "qnode-query", minQueryThreads, maxQueryThreads,
+                    qmProcessQueryQueue) != 0) {
     dError("failed to start qnode query worker since %s", terrstr());
     return -1;
   }
 
-  if (dndInitWorker(pMgmt, &pMgmt->fetchWorker, DND_WORKER_SINGLE, "qnode-fetch", 0, 1, qmProcessQueue) != 0) {
+  if (dndInitWorker(pMgmt, &pMgmt->fetchWorker, DND_WORKER_SINGLE, "qnode-fetch", minFetchThreads, maxFetchThreads,
+                    qmProcessFetchQueue) != 0) {
     dError("failed to start qnode fetch worker since %s", terrstr());
     return -1;
   }

source/dnode/mgmt/snode/src/smInt.c

@@ -19,13 +19,12 @@
 static int32_t smRequire(SMgmtWrapper *pWrapper, bool *required) { return dndReadFile(pWrapper, required); }
 
 static void smInitOption(SSnodeMgmt *pMgmt, SSnodeOpt *pOption) {
-  SDnode *pDnode = pMgmt->pDnode;
-  pOption->pWrapper = pMgmt->pWrapper;
-  pOption->sendReqFp = dndSendReqToDnode;
-  pOption->sendMnodeReqFp = dndSendReqToMnode;
-  pOption->sendRspFp = dndSendRsp;
-  pOption->dnodeId = pDnode->dnodeId;
-  pOption->clusterId = pDnode->clusterId;
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = pMgmt->pWrapper;
+  msgCb.sendReqFp = dndSendReqToDnode;
+  msgCb.sendMnodeReqFp = dndSendReqToMnode;
+  msgCb.sendRspFp = dndSendRsp;
+  pOption->msgCb = msgCb;
 }
 
 static int32_t smOpenImp(SSnodeMgmt *pMgmt) {

source/dnode/mgmt/snode/src/smMsg.c

@@ -56,6 +56,6 @@ int32_t smProcessDropReq(SMgmtWrapper *pWrapper, SNodeMsg *pMsg) {
 
 void smInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by SNODE
-  dndSetMsgHandle(pWrapper, TDMT_SND_TASK_DEPLOY, (NodeMsgFp)smProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_SND_TASK_EXEC, (NodeMsgFp)smProcessExecMsg);
+  dndSetMsgHandle(pWrapper, TDMT_SND_TASK_DEPLOY, (NodeMsgFp)smProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_SND_TASK_EXEC, (NodeMsgFp)smProcessExecMsg, 0);
 }

source/dnode/mgmt/vnode/inc/vmInt.h

@@ -24,15 +24,13 @@
 extern "C" {
 #endif
 
-typedef enum { VND_WRITE_QUEUE, VND_QUERY_QUEUE, VND_FETCH_QUEUE, VND_APPLY_QUEUE, VND_SYNC_QUEUE } EVndQueueType;
-
 typedef struct SVnodesMgmt {
   SHashObj     *hash;
   SRWLatch      latch;
   SVnodesStat   state;
   STfs         *pTfs;
   SQWorkerPool  queryPool;
-  SFWorkerPool  fetchPool;
+  SQWorkerPool  fetchPool;
   SWWorkerPool  syncPool;
   SWWorkerPool  writePool;
   const char   *path;

source/dnode/mgmt/vnode/src/vmInt.c

@@ -128,7 +128,15 @@ static void *vmOpenVnodeFunc(void *param) {
              pMgmt->state.openVnodes, pMgmt->state.totalVnodes);
     dndReportStartup(pDnode, "open-vnodes", stepDesc);
 
-    SVnodeCfg cfg = {.pWrapper = pMgmt->pWrapper, .pTfs = pMgmt->pTfs, .vgId = pCfg->vgId, .dbId = pCfg->dbUid};
+    SMsgCb msgCb = {0};
+    msgCb.pWrapper = pMgmt->pWrapper;
+    msgCb.queueFps[QUERY_QUEUE] = vmPutMsgToQueryQueue;
+    msgCb.queueFps[FETCH_QUEUE] = vmPutMsgToFetchQueue;
+    msgCb.queueFps[APPLY_QUEUE] = vmPutMsgToApplyQueue;
+    msgCb.sendReqFp = dndSendReqToDnode;
+    msgCb.sendMnodeReqFp = dndSendReqToMnode;
+    msgCb.sendRspFp = dndSendRsp;
+    SVnodeCfg cfg = {.msgCb = msgCb, .pTfs = pMgmt->pTfs, .vgId = pCfg->vgId, .dbId = pCfg->dbUid};
     SVnode   *pImpl = vnodeOpen(pCfg->path, &cfg);
     if (pImpl == NULL) {
       dError("vgId:%d, failed to open vnode by thread:%d", pCfg->vgId, pThread->threadIndex);
@@ -262,7 +270,6 @@ static int32_t vmInit(SMgmtWrapper *pWrapper) {
   SDnode      *pDnode = pWrapper->pDnode;
   SVnodesMgmt *pMgmt = calloc(1, sizeof(SVnodesMgmt));
   int32_t      code = -1;
-  SVnodeOpt    vnodeOpt = {0};
 
   dInfo("vnodes-mgmt start to init");
   if (pMgmt == NULL) goto _OVER;
@@ -294,13 +301,7 @@ static int32_t vmInit(SMgmtWrapper *pWrapper) {
     goto _OVER;
   }
 
-  vnodeOpt.nthreads = tsNumOfCommitThreads;
-  vnodeOpt.putToQueryQFp = vmPutMsgToQueryQueue;
-  vnodeOpt.putToFetchQFp = vmPutMsgToQueryQueue;
-  vnodeOpt.sendReqFp = dndSendReqToDnode;
-  vnodeOpt.sendMnodeReqFp = dndSendReqToMnode;
-  vnodeOpt.sendRspFp = dndSendRsp;
-  if (vnodeInit(&vnodeOpt) != 0) {
+  if (vnodeInit() != 0) {
     dError("failed to init vnode since %s", terrstr());
     goto _OVER;
   }

source/dnode/mgmt/vnode/src/vmMsg.c

@@ -82,7 +82,16 @@ int32_t vmProcessCreateVnodeReq(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
     return -1;
   }
 
-  vnodeCfg.pWrapper = pMgmt->pWrapper;
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = pMgmt->pWrapper;
+  msgCb.queueFps[QUERY_QUEUE] = vmPutMsgToQueryQueue;
+  msgCb.queueFps[FETCH_QUEUE] = vmPutMsgToFetchQueue;
+  msgCb.queueFps[APPLY_QUEUE] = vmPutMsgToApplyQueue;
+  msgCb.sendReqFp = dndSendReqToDnode;
+  msgCb.sendMnodeReqFp = dndSendReqToMnode;
+  msgCb.sendRspFp = dndSendRsp;
+
+  vnodeCfg.msgCb = msgCb;
   vnodeCfg.pTfs = pMgmt->pTfs;
   vnodeCfg.dbId = wrapperCfg.dbUid;
   SVnode *pImpl = vnodeOpen(wrapperCfg.path, &vnodeCfg);
@@ -234,42 +243,42 @@ int32_t vmProcessCompactVnodeReq(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
 
 void vmInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by VNODE
-  dndSetMsgHandle(pWrapper, TDMT_VND_SUBMIT, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)vmProcessQueryMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)vmProcessQueryMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_UPDATE_TAG_VAL, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TABLE_META, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TABLES_META, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONSUME, (NodeMsgFp)vmProcessQueryMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_QUERY, (NodeMsgFp)vmProcessQueryMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONNECT, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_DISCONNECT, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_RES_READY, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASKS_STATUS, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_TASK, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TASK, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_TABLE, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TABLE, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES_FETCH, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB, (NodeMsgFp)vmProcessWriteMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CONSUME, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_HEARTBEAT, (NodeMsgFp)vmProcessFetchMsg);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_EXEC, (NodeMsgFp)vmProcessFetchMsg);
-
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE, (NodeMsgFp)vmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE, (NodeMsgFp)vmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE, (NodeMsgFp)vmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE, (NodeMsgFp)vmProcessMgmtMsg);
-  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE, (NodeMsgFp)vmProcessMgmtMsg);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SUBMIT, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)vmProcessQueryMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)vmProcessQueryMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_UPDATE_TAG_VAL, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TABLE_META, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TABLES_META, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONSUME, (NodeMsgFp)vmProcessQueryMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_QUERY, (NodeMsgFp)vmProcessQueryMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONNECT, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_DISCONNECT, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_RES_READY, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASKS_STATUS, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_TASK, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TASK, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES_FETCH, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB, (NodeMsgFp)vmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CONSUME, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_HEARTBEAT, (NodeMsgFp)vmProcessFetchMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_EXEC, (NodeMsgFp)vmProcessFetchMsg, 0);
+
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
 }

source/dnode/mgmt/vnode/src/vmWorker.c

@@ -62,11 +62,10 @@ static void vmProcessQueryQueue(SVnodeObj *pVnode, SNodeMsg *pMsg) {
   int32_t code = vnodeProcessQueryMsg(pVnode->pImpl, &pMsg->rpcMsg);
   if (code != 0) {
     vmSendRsp(pVnode->pWrapper, pMsg, code);
+    dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
+    rpcFreeCont(pMsg->rpcMsg.pCont);
+    taosFreeQitem(pMsg);
   }
-
-  dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
-  rpcFreeCont(pMsg->rpcMsg.pCont);
-  taosFreeQitem(pMsg);
 }
 
 static void vmProcessFetchQueue(SVnodeObj *pVnode, SNodeMsg *pMsg) {
@@ -74,11 +73,10 @@ static void vmProcessFetchQueue(SVnodeObj *pVnode, SNodeMsg *pMsg) {
   int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg);
   if (code != 0) {
     vmSendRsp(pVnode->pWrapper, pMsg, code);
+    dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
+    rpcFreeCont(pMsg->rpcMsg.pCont);
+    taosFreeQitem(pMsg);
   }
-
-  dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
-  rpcFreeCont(pMsg->rpcMsg.pCont);
-  taosFreeQitem(pMsg);
 }
 
 static void vmProcessWriteQueue(SVnodeObj *pVnode, STaosQall *qall, int32_t numOfMsgs) {
@@ -152,7 +150,7 @@ static void vmProcessSyncQueue(SVnodeObj *pVnode, STaosQall *qall, int32_t numOf
   }
 }
 
-static int32_t vmPutNodeMsgToQueue(SVnodesMgmt *pMgmt, SNodeMsg *pMsg, EVndQueueType qtype) {
+static int32_t vmPutNodeMsgToQueue(SVnodesMgmt *pMgmt, SNodeMsg *pMsg, EQueueType qtype) {
   SRpcMsg *pRpc = &pMsg->rpcMsg;
   int32_t  code = -1;
 
@@ -162,25 +160,27 @@ static int32_t vmPutNodeMsgToQueue(SVnodesMgmt *pMgmt, SNodeMsg *pMsg, EVndQueue
 
   SVnodeObj *pVnode = vmAcquireVnode(pMgmt, pHead->vgId);
   if (pVnode == NULL) {
-    dError("vgId:%d, failed to write msg:%p to queue since %s", pHead->vgId, pMsg, terrstr());
+    dError("vgId:%d, failed to write msg:%p to vnode-queue since %s", pHead->vgId, pMsg, terrstr());
     return -1;
   }
 
   switch (qtype) {
-    case VND_QUERY_QUEUE:
+    case QUERY_QUEUE:
       dTrace("msg:%p, will be written into vnode-query queue", pMsg);
       code = taosWriteQitem(pVnode->pQueryQ, pMsg);
       break;
-    case VND_FETCH_QUEUE:
+    case FETCH_QUEUE:
       dTrace("msg:%p, will be written into vnode-fetch queue", pMsg);
       code = taosWriteQitem(pVnode->pFetchQ, pMsg);
       break;
-    case VND_WRITE_QUEUE:
+    case WRITE_QUEUE:
       dTrace("msg:%p, will be written into vnode-write queue", pMsg);
       code = taosWriteQitem(pVnode->pWriteQ, pMsg);
-    case VND_SYNC_QUEUE:
+      break;
+    case SYNC_QUEUE:
       dTrace("msg:%p, will be written into vnode-sync queue", pMsg);
       code = taosWriteQitem(pVnode->pSyncQ, pMsg);
+      break;
     default:
       terrno = TSDB_CODE_INVALID_PARA;
       break;
@@ -191,19 +191,19 @@ static int32_t vmPutNodeMsgToQueue(SVnodesMgmt *pMgmt, SNodeMsg *pMsg, EVndQueue
 }
 
 int32_t vmProcessSyncMsg(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
-  return vmPutNodeMsgToQueue(pMgmt, pMsg, VND_SYNC_QUEUE);
+  return vmPutNodeMsgToQueue(pMgmt, pMsg, SYNC_QUEUE);
 }
 
 int32_t vmProcessWriteMsg(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
-  return vmPutNodeMsgToQueue(pMgmt, pMsg, VND_WRITE_QUEUE);
+  return vmPutNodeMsgToQueue(pMgmt, pMsg, WRITE_QUEUE);
 }
 
 int32_t vmProcessQueryMsg(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
-  return vmPutNodeMsgToQueue(pMgmt, pMsg, VND_QUERY_QUEUE);
+  return vmPutNodeMsgToQueue(pMgmt, pMsg, QUERY_QUEUE);
 }
 
 int32_t vmProcessFetchMsg(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
-  return vmPutNodeMsgToQueue(pMgmt, pMsg, VND_FETCH_QUEUE);
+  return vmPutNodeMsgToQueue(pMgmt, pMsg, FETCH_QUEUE);
 }
 
 int32_t vmProcessMgmtMsg(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
@@ -212,7 +212,7 @@ int32_t vmProcessMgmtMsg(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
   return dndWriteMsgToWorker(pWorker, pMsg);
 }
 
-static int32_t vmPutRpcMsgToQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc, EVndQueueType qtype) {
+static int32_t vmPutRpcMsgToQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc, EQueueType qtype) {
   SVnodesMgmt *pMgmt = pWrapper->pMgmt;
   int32_t      code = -1;
   SMsgHead    *pHead = pRpc->pCont;
@@ -225,20 +225,18 @@ static int32_t vmPutRpcMsgToQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc, EVndQue
     dTrace("msg:%p, is created, type:%s", pMsg, TMSG_INFO(pRpc->msgType));
     pMsg->rpcMsg = *pRpc;
     switch (qtype) {
-      case VND_QUERY_QUEUE:
+      case QUERY_QUEUE:
         dTrace("msg:%p, will be put into vnode-query queue", pMsg);
         code = taosWriteQitem(pVnode->pQueryQ, pMsg);
         break;
-      case VND_FETCH_QUEUE:
+      case FETCH_QUEUE:
         dTrace("msg:%p, will be put into vnode-fetch queue", pMsg);
         code = taosWriteQitem(pVnode->pFetchQ, pMsg);
         break;
-      case VND_APPLY_QUEUE:
+      case APPLY_QUEUE:
         dTrace("msg:%p, will be put into vnode-apply queue", pMsg);
         code = taosWriteQitem(pVnode->pApplyQ, pMsg);
         break;
-      case VND_WRITE_QUEUE:
-      case VND_SYNC_QUEUE:
       default:
         terrno = TSDB_CODE_INVALID_PARA;
         break;
@@ -249,22 +247,22 @@ static int32_t vmPutRpcMsgToQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc, EVndQue
 }
 
 int32_t vmPutMsgToQueryQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
-  return vmPutRpcMsgToQueue(pWrapper, pRpc, VND_QUERY_QUEUE);
+  return vmPutRpcMsgToQueue(pWrapper, pRpc, QUERY_QUEUE);
 }
 
 int32_t vmPutMsgToFetchQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
-  return vmPutRpcMsgToQueue(pWrapper, pRpc, VND_FETCH_QUEUE);
+  return vmPutRpcMsgToQueue(pWrapper, pRpc, FETCH_QUEUE);
 }
 
 int32_t vmPutMsgToApplyQueue(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {
-  return vmPutRpcMsgToQueue(pWrapper, pRpc, VND_APPLY_QUEUE);
+  return vmPutRpcMsgToQueue(pWrapper, pRpc, APPLY_QUEUE);
 }
 
 int32_t vmAllocQueue(SVnodesMgmt *pMgmt, SVnodeObj *pVnode) {
   pVnode->pWriteQ = tWWorkerAllocQueue(&pMgmt->writePool, pVnode, (FItems)vmProcessWriteQueue);
   pVnode->pApplyQ = tWWorkerAllocQueue(&pMgmt->writePool, pVnode, (FItems)vmProcessApplyQueue);
   pVnode->pSyncQ = tWWorkerAllocQueue(&pMgmt->syncPool, pVnode, (FItems)vmProcessSyncQueue);
-  pVnode->pFetchQ = tFWorkerAllocQueue(&pMgmt->fetchPool, pVnode, (FItem)vmProcessFetchQueue);
+  pVnode->pFetchQ = tQWorkerAllocQueue(&pMgmt->fetchPool, pVnode, (FItem)vmProcessFetchQueue);
   pVnode->pQueryQ = tQWorkerAllocQueue(&pMgmt->queryPool, pVnode, (FItem)vmProcessQueryQueue);
 
   if (pVnode->pApplyQ == NULL || pVnode->pWriteQ == NULL || pVnode->pSyncQ == NULL || pVnode->pFetchQ == NULL ||
@@ -279,7 +277,7 @@ int32_t vmAllocQueue(SVnodesMgmt *pMgmt, SVnodeObj *pVnode) {
 
 void vmFreeQueue(SVnodesMgmt *pMgmt, SVnodeObj *pVnode) {
   tQWorkerFreeQueue(&pMgmt->queryPool, pVnode->pQueryQ);
-  tFWorkerFreeQueue(&pMgmt->fetchPool, pVnode->pFetchQ);
+  tQWorkerFreeQueue(&pMgmt->fetchPool, pVnode->pFetchQ);
   tWWorkerFreeQueue(&pMgmt->writePool, pVnode->pWriteQ);
   tWWorkerFreeQueue(&pMgmt->writePool, pVnode->pApplyQ);
   tWWorkerFreeQueue(&pMgmt->syncPool, pVnode->pSyncQ);
@@ -305,11 +303,11 @@ int32_t vmStartWorker(SVnodesMgmt *pMgmt) {
   pQPool->max = maxQueryThreads;
   if (tQWorkerInit(pQPool) != 0) return -1;
 
-  SFWorkerPool *pFPool = &pMgmt->fetchPool;
+  SQWorkerPool *pFPool = &pMgmt->fetchPool;
   pFPool->name = "vnode-fetch";
   pFPool->min = minFetchThreads;
   pFPool->max = maxFetchThreads;
-  if (tFWorkerInit(pFPool) != 0) return -1;
+  if (tQWorkerInit(pFPool) != 0) return -1;
 
   SWWorkerPool *pWPool = &pMgmt->writePool;
   pWPool->name = "vnode-write";
@@ -332,7 +330,7 @@ int32_t vmStartWorker(SVnodesMgmt *pMgmt) {
 
 void vmStopWorker(SVnodesMgmt *pMgmt) {
   dndCleanupWorker(&pMgmt->mgmtWorker);
-  tFWorkerCleanup(&pMgmt->fetchPool);
+  tQWorkerCleanup(&pMgmt->fetchPool);
   tQWorkerCleanup(&pMgmt->queryPool);
   tWWorkerCleanup(&pMgmt->writePool);
   tWWorkerCleanup(&pMgmt->syncPool);

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

@@ -119,19 +119,12 @@ typedef struct SMnode {
   SHashObj         *infosMeta;
   SGrantInfo        grant;
   MndMsgFp          msgFp[TDMT_MAX];
-  SendReqFp         sendReqFp;
-  SendMnodeReqFp    sendMnodeReqFp;
-  PutToQueueFp      putToWriteQFp;
-  PutToQueueFp      putToReadQFp;
+  SMsgCb            msgCb;
 } SMnode;
 
-int32_t mndSendReqToDnode(SMnode *pMnode, SEpSet *pEpSet, SRpcMsg *rpcMsg);
-int32_t mndSendReqToMnode(SMnode *pMnode, SRpcMsg *pMsg);
-void    mndSetMsgHandle(SMnode *pMnode, tmsg_t msgType, MndMsgFp fp);
-
+void     mndSetMsgHandle(SMnode *pMnode, tmsg_t msgType, MndMsgFp fp);
 uint64_t mndGenerateUid(char *name, int32_t len);
-
-void mndGetLoad(SMnode *pMnode, SMnodeLoad *pLoad);
+void     mndGetLoad(SMnode *pMnode, SMnodeLoad *pLoad);
 
 #ifdef __cplusplus
 }

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

@@ -623,7 +623,7 @@ static int32_t mndProcessConfigDnodeReq(SNodeMsg *pReq) {
       .msgType = TDMT_DND_CONFIG_DNODE, .pCont = pBuf, .contLen = bufLen, .ahandle = pReq->rpcMsg.ahandle};
 
   mInfo("dnode:%d, app:%p config:%s req send to dnode", cfgReq.dnodeId, rpcMsg.ahandle, cfgReq.config);
-  mndSendReqToDnode(pMnode, &epSet, &rpcMsg);
+  tmsgSendReq(&pMnode->msgCb, &epSet, &rpcMsg);
 
   return 0;
 }

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

@@ -166,7 +166,7 @@ static int32_t mndStreamActionDelete(SSdb *pSdb, SStreamObj *pStream) {
 
 static int32_t mndStreamActionUpdate(SSdb *pSdb, SStreamObj *pOldStream, SStreamObj *pNewStream) {
   mTrace("stream:%s, perform update action", pOldStream->name);
-  atomic_exchange_32(&pOldStream->updateTime, pNewStream->updateTime);
+  atomic_exchange_64(&pOldStream->updateTime, pNewStream->updateTime);
   atomic_exchange_32(&pOldStream->version, pNewStream->version);
 
   taosWLockLatch(&pOldStream->lock);

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

@@ -272,7 +272,6 @@ static int32_t mndProcessGetSubEpReq(SNodeMsg *pMsg) {
   /*sdbWrite(pMnode->pSdb, pConsumerRaw);*/
 
   strcpy(rsp.cgroup, pReq->cgroup);
-  rsp.consumerId = consumerId;
   if (epoch != pConsumer->epoch) {
     mInfo("send new assignment to consumer, consumer epoch %d, server epoch %d", epoch, pConsumer->epoch);
     SArray *pTopics = pConsumer->currentTopics;
@@ -322,6 +321,7 @@ static int32_t mndProcessGetSubEpReq(SNodeMsg *pMsg) {
   }
   ((SMqRspHead *)buf)->mqMsgType = TMQ_MSG_TYPE__EP_RSP;
   ((SMqRspHead *)buf)->epoch = pConsumer->epoch;
+  ((SMqRspHead *)buf)->consumerId = pConsumer->consumerId;
 
   void *abuf = POINTER_SHIFT(buf, sizeof(SMqRspHead));
   tEncodeSMqCMGetSubEpRsp(&abuf, &rsp);
@@ -337,21 +337,21 @@ static int32_t mndSplitSubscribeKey(const char *key, char *topic, char *cgroup)
   while (key[i] != TMQ_SEPARATOR) {
     i++;
   }
-  memcpy(topic, key, i - 1);
-  topic[i] = 0;
-  strcpy(cgroup, &key[i + 1]);
+  memcpy(cgroup, key, i);
+  cgroup[i] = 0;
+  strcpy(topic, &key[i + 1]);
   return 0;
 }
 
 static SMqRebSubscribe *mndGetOrCreateRebSub(SHashObj *pHash, const char *key) {
-  SMqRebSubscribe *pRebSub = taosHashGet(pHash, key, strlen(key));
+  SMqRebSubscribe *pRebSub = taosHashGet(pHash, key, strlen(key) + 1);
   if (pRebSub == NULL) {
     pRebSub = tNewSMqRebSubscribe(key);
     if (pRebSub == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
       return NULL;
     }
-    taosHashPut(pHash, key, strlen(key), pRebSub, sizeof(SMqRebSubscribe));
+    taosHashPut(pHash, key, strlen(key) + 1, pRebSub, sizeof(SMqRebSubscribe));
   }
   return pRebSub;
 }
@@ -420,7 +420,7 @@ static int32_t mndProcessMqTimerMsg(SNodeMsg *pMsg) {
         .pCont = pRebMsg,
         .contLen = sizeof(SMqDoRebalanceMsg),
     };
-    (*pMnode->putToWriteQFp)(pMnode->pWrapper, &rpcMsg);
+    tmsgPutToQueue(&pMnode->msgCb, WRITE_QUEUE, &rpcMsg);
   } else {
     taosHashCleanup(pRebMsg->rebSubHash);
     rpcFreeCont(pRebMsg);
@@ -441,6 +441,7 @@ static int32_t mndProcessDoRebalanceMsg(SNodeMsg *pMsg) {
     if (pIter == NULL) break;
     SMqRebSubscribe *pRebSub = (SMqRebSubscribe *)pIter;
     SMqSubscribeObj *pSub = mndAcquireSubscribeByKey(pMnode, pRebSub->key);
+    tfree(pRebSub->key);
 
     mInfo("mq rebalance subscription: %s", pSub->key);
 
@@ -503,7 +504,8 @@ static int32_t mndProcessDoRebalanceMsg(SNodeMsg *pMsg) {
             atomic_store_32(&pRebConsumer->status, MQ_CONSUMER_STATUS__IDLE);
           }
 
-          mInfo("mq consumer:%" PRId64 ", status change from %d to %d", pRebConsumer->consumerId, status, pRebConsumer->status);
+          mInfo("mq consumer:%" PRId64 ", status change from %d to %d", pRebConsumer->consumerId, status,
+                pRebConsumer->status);
 
           SSdbRaw *pConsumerRaw = mndConsumerActionEncode(pRebConsumer);
           sdbSetRawStatus(pConsumerRaw, SDB_STATUS_READY);
@@ -537,14 +539,14 @@ static int32_t mndProcessDoRebalanceMsg(SNodeMsg *pMsg) {
               mndSplitSubscribeKey(pSub->key, topic, cgroup);
               SMqTopicObj *pTopic = mndAcquireTopic(pMnode, topic);
 
-              mInfo("mq set conn: assign vgroup %d of topic %s to consumer %" PRId64 "", pConsumerEp->vgId, topic,
-                    pConsumerEp->consumerId);
+              mInfo("mq set conn: assign vgroup %d of topic %s to consumer %" PRId64 " cgroup: %s", pConsumerEp->vgId,
+                    topic, pConsumerEp->consumerId, cgroup);
 
               mndPersistMqSetConnReq(pMnode, pTrans, pTopic, cgroup, pConsumerEp);
               mndReleaseTopic(pMnode, pTopic);
             } else {
-              mInfo("mq rebalance: assign vgroup %d, from consumer %" PRId64 " to consumer %" PRId64 "", pConsumerEp->vgId,
-                    pConsumerEp->oldConsumerId, pConsumerEp->consumerId);
+              mInfo("mq rebalance: assign vgroup %d, from consumer %" PRId64 " to consumer %" PRId64 "",
+                    pConsumerEp->vgId, pConsumerEp->oldConsumerId, pConsumerEp->consumerId);
 
               mndPersistRebalanceMsg(pMnode, pTrans, pConsumerEp);
             }
@@ -1099,7 +1101,8 @@ static int32_t mndProcessSubscribeReq(SNodeMsg *pMsg) {
       SMqSubscribeObj *pSub = mndAcquireSubscribe(pMnode, cgroup, newTopicName);
       bool             createSub = false;
       if (pSub == NULL) {
-        mDebug("create new subscription by consumer %" PRId64 ", group: %s, topic %s", consumerId, cgroup, newTopicName);
+        mDebug("create new subscription by consumer %" PRId64 ", group: %s, topic %s", consumerId, cgroup,
+               newTopicName);
         pSub = mndCreateSubscription(pMnode, pTopic, cgroup);
         createSub = true;
 

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

@@ -176,7 +176,7 @@ static int32_t mndTopicActionDelete(SSdb *pSdb, SMqTopicObj *pTopic) {
 
 static int32_t mndTopicActionUpdate(SSdb *pSdb, SMqTopicObj *pOldTopic, SMqTopicObj *pNewTopic) {
   mTrace("topic:%s, perform update action", pOldTopic->name);
-  atomic_exchange_32(&pOldTopic->updateTime, pNewTopic->updateTime);
+  atomic_exchange_64(&pOldTopic->updateTime, pNewTopic->updateTime);
   atomic_exchange_32(&pOldTopic->version, pNewTopic->version);
 
   taosWLockLatch(&pOldTopic->lock);
@@ -241,12 +241,12 @@ static int32_t mndGetPlanString(SCMCreateTopicReq *pCreate, char **pStr) {
     return TSDB_CODE_SUCCESS;
   }
 
-  SNode* pAst = NULL;
+  SNode  *pAst = NULL;
   int32_t code = nodesStringToNode(pCreate->ast, &pAst);
 
-  SQueryPlan* pPlan = NULL;
+  SQueryPlan *pPlan = NULL;
   if (TSDB_CODE_SUCCESS == code) {
-    SPlanContext cxt = { .pAstRoot = pAst, .topicQuery = true };
+    SPlanContext cxt = {.pAstRoot = pAst, .topicQuery = true};
     code = qCreateQueryPlan(&cxt, &pPlan, NULL);
   }
 
@@ -274,7 +274,7 @@ static int32_t mndCreateTopic(SMnode *pMnode, SNodeMsg *pReq, SCMCreateTopicReq
   topicObj.logicalPlan = "";
   topicObj.sqlLen = strlen(pCreate->sql);
 
-  char* pPlanStr = NULL;
+  char *pPlanStr = NULL;
   if (TSDB_CODE_SUCCESS != mndGetPlanString(pCreate, &pPlanStr)) {
     mError("topic:%s, failed to get plan since %s", pCreate->name, terrstr());
     return -1;

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

@@ -888,7 +888,7 @@ static int32_t mndTransSendActionMsg(SMnode *pMnode, STrans *pTrans, SArray *pAr
     }
     memcpy(rpcMsg.pCont, pAction->pCont, pAction->contLen);
 
-    if (mndSendReqToDnode(pMnode, &pAction->epSet, &rpcMsg) == 0) {
+    if (tmsgSendReq(&pMnode->msgCb, &pAction->epSet, &rpcMsg) == 0) {
       mDebug("trans:%d, action:%d is sent", pTrans->id, action);
       pAction->msgSent = 1;
       pAction->msgReceived = 0;

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

@@ -43,24 +43,6 @@
 #define TRNAS_TIMER_MS 6000
 #define TELEM_TIMER_MS 86400000
 
-int32_t mndSendReqToDnode(SMnode *pMnode, SEpSet *pEpSet, SRpcMsg *pMsg) {
-  if (pMnode == NULL || pMnode->sendReqFp == NULL) {
-    terrno = TSDB_CODE_MND_NOT_READY;
-    return -1;
-  }
-
-  return (*pMnode->sendReqFp)(pMnode->pWrapper, pEpSet, pMsg);
-}
-
-int32_t mndSendReqToMnode(SMnode *pMnode, SRpcMsg *pMsg) {
-  if (pMnode == NULL || pMnode->sendReqFp == NULL) {
-    terrno = TSDB_CODE_MND_NOT_READY;
-    return -1;
-  }
-
-  return (*pMnode->sendMnodeReqFp)(pMnode->pWrapper, pMsg);
-}
-
 static void *mndBuildTimerMsg(int32_t *pContLen) {
   SMTimerReq timerReq = {0};
 
@@ -80,7 +62,7 @@ static void mndPullupTrans(void *param, void *tmrId) {
     int32_t contLen = 0;
     void   *pReq = mndBuildTimerMsg(&contLen);
     SRpcMsg rpcMsg = {.msgType = TDMT_MND_TRANS_TIMER, .pCont = pReq, .contLen = contLen};
-    pMnode->putToWriteQFp(pMnode->pWrapper, &rpcMsg);
+    tmsgPutToQueue(&pMnode->msgCb, WRITE_QUEUE, &rpcMsg);
   }
 
   taosTmrReset(mndPullupTrans, TRNAS_TIMER_MS, pMnode, pMnode->timer, &pMnode->transTimer);
@@ -96,7 +78,7 @@ static void mndCalMqRebalance(void *param, void *tmrId) {
         .pCont = pReq,
         .contLen = contLen,
     };
-    pMnode->putToReadQFp(pMnode->pWrapper, &rpcMsg);
+    tmsgPutToQueue(&pMnode->msgCb, QUERY_QUEUE, &rpcMsg);
   }
 
   taosTmrReset(mndCalMqRebalance, MQ_TIMER_MS, pMnode, pMnode->timer, &pMnode->mqTimer);
@@ -108,7 +90,7 @@ static void mndPullupTelem(void *param, void *tmrId) {
     int32_t contLen = 0;
     void   *pReq = mndBuildTimerMsg(&contLen);
     SRpcMsg rpcMsg = {.msgType = TDMT_MND_TELEM_TIMER, .pCont = pReq, .contLen = contLen};
-    pMnode->putToReadQFp(pMnode->pWrapper, &rpcMsg);
+    tmsgPutToQueue(&pMnode->msgCb, QUERY_QUEUE, &rpcMsg);
   }
 
   taosTmrReset(mndPullupTelem, TELEM_TIMER_MS, pMnode, pMnode->timer, &pMnode->telemTimer);
@@ -286,14 +268,9 @@ static int32_t mndSetOptions(SMnode *pMnode, const SMnodeOpt *pOption) {
   pMnode->replica = pOption->replica;
   pMnode->selfIndex = pOption->selfIndex;
   memcpy(&pMnode->replicas, pOption->replicas, sizeof(SReplica) * TSDB_MAX_REPLICA);
-  pMnode->pWrapper = pOption->pWrapper;
-  pMnode->putToWriteQFp = pOption->putToWriteQFp;
-  pMnode->putToReadQFp = pOption->putToReadQFp;
-  pMnode->sendReqFp = pOption->sendReqFp;
-  pMnode->sendMnodeReqFp = pOption->sendMnodeReqFp;
-
-  if (pMnode->sendReqFp == NULL || pMnode->sendMnodeReqFp == NULL  ||
-      pMnode->putToWriteQFp == NULL || pMnode->dnodeId < 0 || pMnode->clusterId < 0) {
+  pMnode->msgCb = pOption->msgCb;
+
+  if (pMnode->dnodeId < 0 || pMnode->clusterId < 0) {
     terrno = TSDB_CODE_MND_INVALID_OPTIONS;
     return -1;
   }

source/dnode/qnode/inc/qndInt.h

@@ -21,6 +21,7 @@
 #include "tlog.h"
 #include "tmsg.h"
 #include "trpc.h"
+#include "tmsgcb.h"
 
 #include "qnode.h"
 
@@ -32,8 +33,8 @@ typedef struct SQWorkerMgmt SQHandle;
 
 typedef struct SQnode {
   int32_t   qndId;
-  SQnodeOpt opt;
-  SQHandle*  pQuery;  
+  SMsgCb    msgCb;
+  SQHandle* pQuery;
 } SQnode;
 
 #ifdef __cplusplus

source/dnode/qnode/src/qnode.c

@@ -13,14 +13,10 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include "executor.h"
 #include "qndInt.h"
 #include "query.h"
 #include "qworker.h"
-#include "executor.h"
-
-int32_t qnodePutReqToVQueryQ(SQnode* pQnode, struct SRpcMsg* pReq) {}
-void    qnodeSendReqToDnode(SQnode* pQnode, struct SEpSet* epSet, struct SRpcMsg* pReq) {}
-
 
 SQnode *qndOpen(const SQnodeOpt *pOption) {
   SQnode *pQnode = calloc(1, sizeof(SQnode));
@@ -29,34 +25,29 @@ SQnode *qndOpen(const SQnodeOpt *pOption) {
     return NULL;
   }
 
-  if (qWorkerInit(NODE_TYPE_QNODE, pQnode->qndId, NULL, (void **)&pQnode->pQuery, pQnode,
-                     (putReqToQueryQFp)qnodePutReqToVQueryQ, (sendReqFp)qnodeSendReqToDnode)) {
+  if (qWorkerInit(NODE_TYPE_QNODE, pQnode->qndId, NULL, (void **)&pQnode->pQuery, &pOption->msgCb)) {
     tfree(pQnode);
     return NULL;
   }
-  
+
+  pQnode->msgCb = pOption->msgCb;
   return pQnode;
 }
 
-void qndClose(SQnode *pQnode) { 
+void qndClose(SQnode *pQnode) {
   qWorkerDestroy((void **)&pQnode->pQuery);
 
-  free(pQnode); 
+  free(pQnode);
 }
 
 int32_t qndGetLoad(SQnode *pQnode, SQnodeLoad *pLoad) { return 0; }
 
-int32_t qndProcessMsg(SQnode *pQnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
-  *pRsp = NULL;
-  return 0;
-}
-
-int qnodeProcessQueryMsg(SQnode *pQnode, SRpcMsg *pMsg) {
+int32_t qndProcessQueryMsg(SQnode *pQnode, SRpcMsg *pMsg) {
   qTrace("message in query queue is processing");
   SReadHandle handle = {0};
 
   switch (pMsg->msgType) {
-    case TDMT_VND_QUERY:{
+    case TDMT_VND_QUERY: {
       return qWorkerProcessQueryMsg(&handle, pQnode->pQuery, pMsg);
     }
     case TDMT_VND_QUERY_CONTINUE:
@@ -67,7 +58,7 @@ int qnodeProcessQueryMsg(SQnode *pQnode, SRpcMsg *pMsg) {
   }
 }
 
-int qnodeProcessFetchMsg(SQnode *pQnode, SRpcMsg *pMsg) {
+int32_t qndProcessFetchMsg(SQnode *pQnode, SRpcMsg *pMsg) {
   qTrace("message in fetch queue is processing");
   switch (pMsg->msgType) {
     case TDMT_VND_FETCH:
@@ -85,17 +76,13 @@ int qnodeProcessFetchMsg(SQnode *pQnode, SRpcMsg *pMsg) {
     case TDMT_VND_SHOW_TABLES:
       return qWorkerProcessShowMsg(pQnode, pQnode->pQuery, pMsg);
     case TDMT_VND_SHOW_TABLES_FETCH:
-      //return vnodeGetTableList(pQnode, pMsg);
+      // return vnodeGetTableList(pQnode, pMsg);
     case TDMT_VND_TABLE_META:
-      //return vnodeGetTableMeta(pQnode, pMsg);
+      // return vnodeGetTableMeta(pQnode, pMsg);
     case TDMT_VND_CONSUME:
-      //return tqProcessConsumeReq(pQnode->pTq, pMsg);
+      // return tqProcessConsumeReq(pQnode->pTq, pMsg);
     default:
       qError("unknown msg type:%d in fetch queue", pMsg->msgType);
       return TSDB_CODE_VND_APP_ERROR;
   }
 }
-
-
-
-

source/dnode/snode/inc/sndInt.h

@@ -44,7 +44,7 @@ typedef struct {
 
 typedef struct SSnode {
   SStreamMeta* pMeta;
-  SSnodeOpt    cfg;
+  SMsgCb       msgCb;
 } SSnode;
 
 SStreamMeta* sndMetaNew();

source/dnode/snode/src/snode.c

@@ -22,7 +22,7 @@ SSnode *sndOpen(const char *path, const SSnodeOpt *pOption) {
   if (pSnode == NULL) {
     return NULL;
   }
-  memcpy(&pSnode->cfg, pOption, sizeof(SSnodeOpt));
+  pSnode->msgCb = pOption->msgCb;
   pSnode->pMeta = sndMetaNew();
   if (pSnode->pMeta == NULL) {
     free(pSnode);
@@ -57,7 +57,9 @@ void sndMetaDelete(SStreamMeta *pMeta) {
 }
 
 int32_t sndMetaDeployTask(SStreamMeta *pMeta, SStreamTask *pTask) {
-  pTask->runner.executor = qCreateStreamExecTaskInfo(pTask->qmsg, NULL);
+  for (int i = 0; i < pTask->parallel; i++) {
+    pTask->runner.executor[i] = qCreateStreamExecTaskInfo(pTask->qmsg, NULL);
+  }
   return taosHashPut(pMeta->pHash, &pTask->taskId, sizeof(int32_t), pTask, sizeof(void *));
 }
 
@@ -95,6 +97,7 @@ void sndProcessUMsg(SSnode *pSnode, SRpcMsg *pMsg) {
     SStreamTask *pTask = malloc(sizeof(SStreamTask));
     if (pTask == NULL) {
       ASSERT(0);
+      return;
     }
     SCoder decoder;
     tCoderInit(&decoder, TD_LITTLE_ENDIAN, msg, pMsg->contLen - sizeof(SMsgHead), TD_DECODER);

source/dnode/vnode/inc/vnode.h

@@ -18,6 +18,7 @@
 
 #include "os.h"
 #include "trpc.h"
+#include "tmsgcb.h"
 
 #include "meta.h"
 #include "tarray.h"
@@ -40,7 +41,6 @@ typedef struct {
 typedef struct {
   int32_t  vgId;
   uint64_t dbId;
-  void    *pWrapper;
   STfs    *pTfs;
   uint64_t wsize;
   uint64_t ssize;
@@ -54,20 +54,12 @@ typedef struct {
   SMetaCfg metaCfg;
   STqCfg   tqCfg;
   SWalCfg  walCfg;
+  SMsgCb   msgCb;
   uint32_t hashBegin;
   uint32_t hashEnd;
   int8_t   hashMethod;
 } SVnodeCfg;
 
-typedef struct {
-  uint16_t       nthreads;  // number of commit threads. 0 for no threads and a schedule queue should be given (TODO)
-  PutToQueueFp   putToQueryQFp;
-  PutToQueueFp   putToFetchQFp;
-  SendReqFp      sendReqFp;
-  SendMnodeReqFp sendMnodeReqFp;
-  SendRspFp      sendRspFp;
-} SVnodeOpt;
-
 typedef struct {
   int64_t           ver;
   int64_t           tbUid;
@@ -87,10 +79,9 @@ typedef struct {
 /**
  * @brief Initialize the vnode module
  *
- * @param pOption Option of the vnode mnodule
  * @return int 0 for success and -1 for failure
  */
-int vnodeInit(const SVnodeOpt *pOption);
+int vnodeInit();
 
 /**
  * @brief Cleanup the vnode module

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

@@ -52,12 +52,6 @@ typedef struct SVnodeMgr {
   TdThreadMutex mutex;
   TdThreadCond  hasTask;
   TD_DLIST(SVnodeTask) queue;
-  // For vnode Mgmt
-  PutToQueueFp   putToQueryQFp;
-  PutToQueueFp   putToFetchQFp;
-  SendReqFp      sendReqFp;
-  SendMnodeReqFp sendMnodeReqFp;
-  SendRspFp      sendRspFp;
 } SVnodeMgr;
 
 extern SVnodeMgr vnodeMgr;
@@ -81,7 +75,7 @@ struct SVnode {
   SWal*      pWal;
   tsem_t     canCommit;
   SQHandle*  pQuery;
-  void*      pWrapper;
+  SMsgCb     msgCb;
   STfs*      pTfs;
 };
 

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

@@ -245,7 +245,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
   }
 
   SMqPollRsp rsp = {
-      .consumerId = consumerId,
+      /*.consumerId = consumerId,*/
       .numOfTopics = 0,
       .pBlockData = NULL,
   };
@@ -298,7 +298,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
         }
 
         taosArrayPush(pRes, pDataBlock);
-        rsp.schemas = pTopic->buffer.output[pos].pReadHandle->pSchemaWrapper;
+        rsp.schema = pTopic->buffer.output[pos].pReadHandle->pSchemaWrapper;
         rsp.rspOffset = fetchOffset;
 
         rsp.numOfTopics = 1;
@@ -312,10 +312,11 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
         }
         ((SMqRspHead*)buf)->mqMsgType = TMQ_MSG_TYPE__POLL_RSP;
         ((SMqRspHead*)buf)->epoch = pReq->epoch;
+        ((SMqRspHead*)buf)->consumerId = consumerId;
 
         void* abuf = POINTER_SHIFT(buf, sizeof(SMqRspHead));
         tEncodeSMqPollRsp(&abuf, &rsp);
-        taosArrayDestroyEx(rsp.pBlockData, (void (*)(void*))tDeleteSSDataBlock);
+        /*taosArrayDestroyEx(rsp.pBlockData, (void (*)(void*))tDeleteSSDataBlock);*/
         pMsg->pCont = buf;
         pMsg->contLen = tlen;
         pMsg->code = 0;

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

@@ -14,8 +14,8 @@
  */
 
 #define _DEFAULT_SOURCE
-#include "vnd.h"
 #include "sync.h"
+#include "vnd.h"
 // #include "vnodeInt.h"
 
 int32_t vnodeAlter(SVnode *pVnode, const SVnodeCfg *pCfg) { return 0; }
@@ -41,6 +41,6 @@ int32_t vnodeGetLoad(SVnode *pVnode, SVnodeLoad *pLoad) {
 }
 
 int vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
-  vInfo("sync message is processed");
+  /*vInfo("sync message is processed");*/
   return 0;
 }

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

@@ -27,7 +27,7 @@ SVnode *vnodeOpen(const char *path, const SVnodeCfg *pVnodeCfg) {
   SVnodeCfg cfg = defaultVnodeOptions;
   if (pVnodeCfg != NULL) {
     cfg.vgId = pVnodeCfg->vgId;
-    cfg.pWrapper = pVnodeCfg->pWrapper;
+    cfg.msgCb = pVnodeCfg->msgCb;
     cfg.pTfs = pVnodeCfg->pTfs;
     cfg.dbId = pVnodeCfg->dbId;
     cfg.hashBegin = pVnodeCfg->hashBegin;
@@ -79,7 +79,7 @@ static SVnode *vnodeNew(const char *path, const SVnodeCfg *pVnodeCfg) {
   }
 
   pVnode->vgId = pVnodeCfg->vgId;
-  pVnode->pWrapper = pVnodeCfg->pWrapper;
+  pVnode->msgCb = pVnodeCfg->msgCb;
   pVnode->pTfs = pVnodeCfg->pTfs;
   pVnode->path = strdup(path);
   vnodeOptionsCopy(&(pVnode->config), pVnodeCfg);

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

@@ -14,43 +14,33 @@
  */
 
 #include "vnd.h"
+#include "tglobal.h"
 
 SVnodeMgr vnodeMgr = {.vnodeInitFlag = TD_MOD_UNINITIALIZED};
 
 static void* loop(void* arg);
 
-int vnodeInit(const SVnodeOpt *pOption) {
+int vnodeInit() {
   if (TD_CHECK_AND_SET_MODE_INIT(&(vnodeMgr.vnodeInitFlag)) == TD_MOD_INITIALIZED) {
     return 0;
   }
 
   vnodeMgr.stop = false;
-  vnodeMgr.putToQueryQFp = pOption->putToQueryQFp;
-  vnodeMgr.putToFetchQFp = pOption->putToFetchQFp;
-  vnodeMgr.sendReqFp = pOption->sendReqFp;
-  vnodeMgr.sendMnodeReqFp = pOption->sendMnodeReqFp;
-  vnodeMgr.sendRspFp = pOption->sendRspFp;
 
   // Start commit handers
-  if (pOption->nthreads > 0) {
-    vnodeMgr.nthreads = pOption->nthreads;
-    vnodeMgr.threads = (TdThread*)calloc(pOption->nthreads, sizeof(TdThread));
-    if (vnodeMgr.threads == NULL) {
-      return -1;
-    }
+  vnodeMgr.nthreads = tsNumOfCommitThreads;
+  vnodeMgr.threads = calloc(vnodeMgr.nthreads, sizeof(TdThread));
+  if (vnodeMgr.threads == NULL) {
+    return -1;
+  }
 
-    taosThreadMutexInit(&(vnodeMgr.mutex), NULL);
-    taosThreadCondInit(&(vnodeMgr.hasTask), NULL);
-    TD_DLIST_INIT(&(vnodeMgr.queue));
+  taosThreadMutexInit(&(vnodeMgr.mutex), NULL);
+  taosThreadCondInit(&(vnodeMgr.hasTask), NULL);
+  TD_DLIST_INIT(&(vnodeMgr.queue));
 
-    for (uint16_t i = 0; i < pOption->nthreads; i++) {
-      taosThreadCreate(&(vnodeMgr.threads[i]), NULL, loop, NULL);
-      // pthread_setname_np(vnodeMgr.threads[i], "VND Commit Thread");
-    }
-  } else {
-    // TODO: if no commit thread is set, then another mechanism should be
-    // given. Otherwise, it is a false.
-    ASSERT(0);
+  for (uint16_t i = 0; i < vnodeMgr.nthreads; i++) {
+    taosThreadCreate(&(vnodeMgr.threads[i]), NULL, loop, NULL);
+    // pthread_setname_np(vnodeMgr.threads[i], "VND Commit Thread");
   }
 
   if (walInit() < 0) {
@@ -92,26 +82,6 @@ int vnodeScheduleTask(SVnodeTask* pTask) {
   return 0;
 }
 
-int32_t vnodePutToVQueryQ(SVnode* pVnode, struct SRpcMsg* pReq) {
-  return (*vnodeMgr.putToQueryQFp)(pVnode->pWrapper, pReq);
-}
-
-int32_t vnodePutToVFetchQ(SVnode* pVnode, struct SRpcMsg* pReq) {
-  return (*vnodeMgr.putToFetchQFp)(pVnode->pWrapper, pReq);
-}
-
-int32_t vnodeSendReq(SVnode* pVnode, struct SEpSet* epSet, struct SRpcMsg* pReq) {
-  return (*vnodeMgr.sendReqFp)(pVnode->pWrapper, epSet, pReq);
-}
-
-int32_t vnodeSendMnodeReq(SVnode* pVnode, struct SRpcMsg* pReq) {
-  return (*vnodeMgr.sendMnodeReqFp)(pVnode->pWrapper, pReq);
-}
-
-void vnodeSendRsp(SVnode* pVnode, struct SEpSet* epSet, struct SRpcMsg* pRsp) {
-  (*vnodeMgr.sendRspFp)(pVnode->pWrapper, pRsp);
-}
-
 /* ------------------------ STATIC METHODS ------------------------ */
 static void* loop(void* arg) {
   setThreadName("vnode-commit");

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

@@ -20,8 +20,7 @@ static int32_t vnodeGetTableList(SVnode *pVnode, SRpcMsg *pMsg);
 static int     vnodeGetTableMeta(SVnode *pVnode, SRpcMsg *pMsg);
 
 int vnodeQueryOpen(SVnode *pVnode) {
-  return qWorkerInit(NODE_TYPE_VNODE, pVnode->vgId, NULL, (void **)&pVnode->pQuery, pVnode,
-                     (putReqToQueryQFp)vnodePutToVQueryQ, (sendReqFp)vnodeSendReq);
+  return qWorkerInit(NODE_TYPE_VNODE, pVnode->vgId, NULL, (void **)&pVnode->pQuery, &pVnode->msgCb);
 }
 
 void vnodeQueryClose(SVnode *pVnode) { qWorkerDestroy((void **)&pVnode->pQuery); }

source/libs/catalog/inc/catalogInt.h

@@ -288,7 +288,7 @@ typedef struct SCtgAction {
 #define CTG_ERR_JRET(c) do { code = c; if (code != TSDB_CODE_SUCCESS) { terrno = code; goto _return; } } while (0)
 
 #define CTG_API_LEAVE(c) do { int32_t __code = c; CTG_UNLOCK(CTG_READ, &gCtgMgmt.lock); CTG_API_DEBUG("CTG API leave %s", __FUNCTION__); CTG_RET(__code); } while (0)
-#define CTG_API_ENTER() do { CTG_API_DEBUG("CTG API enter %s", __FUNCTION__); CTG_LOCK(CTG_READ, &gCtgMgmt.lock); if (atomic_load_8(&gCtgMgmt.exit)) { CTG_API_LEAVE(TSDB_CODE_CTG_OUT_OF_SERVICE); }  } while (0)
+#define CTG_API_ENTER() do { CTG_API_DEBUG("CTG API enter %s", __FUNCTION__); CTG_LOCK(CTG_READ, &gCtgMgmt.lock); if (atomic_load_8((int8_t*)&gCtgMgmt.exit)) { CTG_API_LEAVE(TSDB_CODE_CTG_OUT_OF_SERVICE); }  } while (0)
 
 
 

source/libs/catalog/src/catalog.c

@@ -316,7 +316,7 @@ void ctgWaitAction(SCtgMetaAction *action) {
   while (true) {
     tsem_wait(&gCtgMgmt.queue.rspSem);
     
-    if (atomic_load_8(&gCtgMgmt.exit)) {
+    if (atomic_load_8((int8_t*)&gCtgMgmt.exit)) {
       tsem_post(&gCtgMgmt.queue.rspSem);
       break;
     }
@@ -2141,7 +2141,7 @@ void* ctgUpdateThreadFunc(void* param) {
   while (true) {
     tsem_wait(&gCtgMgmt.queue.reqSem);
     
-    if (atomic_load_8(&gCtgMgmt.exit)) {
+    if (atomic_load_8((int8_t*)&gCtgMgmt.exit)) {
       tsem_post(&gCtgMgmt.queue.rspSem);
       break;
     }
@@ -2270,7 +2270,7 @@ int32_t catalogInit(SCatalogCfg *cfg) {
     CTG_ERR_RET(TSDB_CODE_CTG_INVALID_INPUT);
   }
 
-  atomic_store_8(&gCtgMgmt.exit, false);
+  atomic_store_8((int8_t*)&gCtgMgmt.exit, false);
 
   if (cfg) {
     memcpy(&gCtgMgmt.cfg, cfg, sizeof(*cfg));
@@ -2844,11 +2844,11 @@ int32_t catalogGetExpiredDBs(SCatalog* pCtg, SDbVgVersion **dbs, uint32_t *num)
 void catalogDestroy(void) {
   qInfo("start to destroy catalog");
   
-  if (NULL == gCtgMgmt.pCluster || atomic_load_8(&gCtgMgmt.exit)) {
+  if (NULL == gCtgMgmt.pCluster || atomic_load_8((int8_t*)&gCtgMgmt.exit)) {
     return;
   }
 
-  atomic_store_8(&gCtgMgmt.exit, true);
+  atomic_store_8((int8_t*)&gCtgMgmt.exit, true);
 
   tsem_post(&gCtgMgmt.queue.reqSem);
   tsem_post(&gCtgMgmt.queue.rspSem);

source/libs/executor/inc/executorimpl.h

@@ -543,9 +543,10 @@ typedef struct SFillOperatorInfo {
 
 typedef struct SGroupbyOperatorInfo {
   SOptrBasicInfo binfo;
-  int32_t        colIndex;
+  SArray*        pGroupCols;
   char*          prevData;  // previous group by value
   SGroupResInfo  groupResInfo;
+  SAggSupporter  aggSup;
 } SGroupbyOperatorInfo;
 
 typedef struct SSessionAggOperatorInfo {
@@ -643,6 +644,8 @@ SOperatorInfo* createLimitOperatorInfo(SOperatorInfo* downstream, int32_t numOfD
 SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResBlock, SInterval* pInterval,
                                           const STableGroupInfo* pTableGroupInfo, SExecTaskInfo* pTaskInfo);
 SOperatorInfo* createSessionAggOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResBlock, SExecTaskInfo* pTaskInfo);
+SOperatorInfo* createGroupOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResultBlock,
+                                       SArray* pGroupColList, SExecTaskInfo* pTaskInfo, const STableGroupInfo* pTableGroupInfo);
 
 SOperatorInfo* createTableSeqScanOperatorInfo(void* pTsdbReadHandle, STaskRuntimeEnv* pRuntimeEnv);
 SOperatorInfo* createAllTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,
@@ -650,7 +653,6 @@ SOperatorInfo* createAllTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, S
 
 SOperatorInfo* createFillOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream, SExprInfo* pExpr,
                                       int32_t numOfOutput, bool multigroupResult);
-SOperatorInfo* createGroupOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExpr, int32_t numOfOutput, SSDataBlock* pResBlock, SExecTaskInfo* pTaskInfo);
 
 SOperatorInfo* createMultiTableTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,
                                                         SExprInfo* pExpr, int32_t numOfOutput);

source/libs/executor/src/tsimplehash.c

@@ -40,7 +40,7 @@ typedef struct SHNode {
 typedef struct SSHashObj {
   SHNode         **hashList;
   size_t           capacity;     // number of slots
-  size_t           size;         // number of elements in hash table
+  int64_t          size;         // number of elements in hash table
   _hash_fn_t       hashFp;       // hash function
   _equal_fn_t      equalFp;      // equal function
   int32_t          keyLen;
@@ -91,7 +91,7 @@ int32_t tSimpleHashGetSize(const SSHashObj *pHashObj) {
   if (pHashObj == NULL) {
     return 0;
   }
-  return (int32_t)atomic_load_64(&pHashObj->size);
+  return (int32_t)atomic_load_64((int64_t*)&pHashObj->size);
 }
 
 static SHNode *doCreateHashNode(const void *key, size_t keyLen, const void *pData, size_t dsize, uint32_t hashVal) {

source/libs/function/src/builtinsimpl.c

@@ -456,7 +456,7 @@ void firstFunction(SqlFunctionCtx *pCtx) {
 
   int32_t numOfElems = 0;
 
-  struct SResultRowEntryInfo *pResInfo = GET_RES_INFO(pCtx);
+  SResultRowEntryInfo *pResInfo = GET_RES_INFO(pCtx);
   char* buf = GET_ROWCELL_INTERBUF(pResInfo);
 
   SInputColumnInfoData* pInput = &pCtx->input;
@@ -500,7 +500,7 @@ void lastFunction(SqlFunctionCtx *pCtx) {
 
   int32_t numOfElems = 0;
 
-  struct SResultRowEntryInfo *pResInfo = GET_RES_INFO(pCtx);
+  SResultRowEntryInfo *pResInfo = GET_RES_INFO(pCtx);
   char* buf = GET_ROWCELL_INTERBUF(pResInfo);
 
   SInputColumnInfoData* pInput = &pCtx->input;

source/libs/function/src/functionMgt.c

@@ -97,7 +97,7 @@ bool fmIsAggFunc(int32_t funcId) {
 
 void fmFuncMgtDestroy() {
   void* m = gFunMgtService.pFuncNameHashTable;
-  if (m != NULL && atomic_val_compare_exchange_ptr(&gFunMgtService.pFuncNameHashTable, m, 0) == m) {
+  if (m != NULL && atomic_val_compare_exchange_ptr((void**)&gFunMgtService.pFuncNameHashTable, m, 0) == m) {
     taosHashCleanup(m);
   }
 }

source/libs/parser/src/parAstCreater.c

@@ -920,7 +920,7 @@ SDataType createDataType(uint8_t type) {
 }
 
 SDataType createVarLenDataType(uint8_t type, const SToken* pLen) {
-  SDataType dt = { .type = type, .precision = 0, .scale = 0, .bytes = tDataTypes[type].bytes };
+  SDataType dt = { .type = type, .precision = 0, .scale = 0, .bytes = strtol(pLen->z, NULL, 10) };
   return dt;
 }
 

source/libs/parser/src/parInsert.c

@@ -28,6 +28,13 @@
     pSql += index; \
   } while (0)
 
+#define NEXT_TOKEN_WITH_PREV(pSql, sToken) \
+  do { \
+    int32_t index = 0; \
+    sToken = tStrGetToken(pSql, &index, true); \
+    pSql += index; \
+  } while (0)
+
 #define NEXT_TOKEN_KEEP_SQL(pSql, sToken, index) \
   do { \
     sToken = tStrGetToken(pSql, &index, false); \
@@ -357,7 +364,7 @@ static int parseTime(char **end, SToken *pToken, int16_t timePrec, int64_t *time
   sToken = tStrGetToken(pTokenEnd, &index, false);
   pTokenEnd += index;
 
-  if (sToken.type == TK_MINUS || sToken.type == TK_NK_PLUS) {
+  if (sToken.type == TK_NK_MINUS || sToken.type == TK_NK_PLUS) {
     index = 0;
     valueToken = tStrGetToken(pTokenEnd, &index, false);
     pTokenEnd += index;
@@ -753,7 +760,7 @@ static int32_t parseTagsClause(SInsertParseContext* pCxt, SSchema* pTagsSchema,
   SToken sToken;
   char tmpTokenBuf[TSDB_MAX_BYTES_PER_ROW] = {0};  // used for deleting Escape character: \\, \', \"
   for (int i = 0; i < pCxt->tags.numOfBound; ++i) {
-    NEXT_TOKEN(pCxt->pSql, sToken);
+    NEXT_TOKEN_WITH_PREV(pCxt->pSql, sToken);
     SSchema* pSchema = &pTagsSchema[pCxt->tags.boundedColumns[i]];
     param.schema = pSchema;
     CHECK_CODE(parseValueToken(&pCxt->pSql, &sToken, pSchema, precision, tmpTokenBuf, KvRowAppend, &param, &pCxt->msg));
@@ -819,7 +826,7 @@ static int parseOneRow(SInsertParseContext* pCxt, STableDataBlocks* pDataBlocks,
   SToken sToken = {0};
   // 1. set the parsed value from sql string
   for (int i = 0; i < spd->numOfBound; ++i) {
-    NEXT_TOKEN(pCxt->pSql, sToken);
+    NEXT_TOKEN_WITH_PREV(pCxt->pSql, sToken);
     SSchema *pSchema = &schema[spd->boundedColumns[i] - 1];
     param.schema = pSchema;
     getMemRowAppendInfo(schema, pBuilder->rowType, spd, i, &param.toffset, &param.colIdx);

source/libs/parser/src/parTokenizer.c

@@ -321,7 +321,7 @@ uint32_t tGetToken(const char* z, uint32_t* tokenId) {
         *tokenId = TK_NK_COMMENT;
         return i;
       }
-      *tokenId = TK_MINUS;
+      *tokenId = TK_NK_MINUS;
       return 1;
     }
     case '(': {
@@ -675,7 +675,7 @@ SToken tStrGetToken(const char* str, int32_t* i, bool isPrevOptr) {
 
   } else {
     // support parse the -/+number format
-    if ((isPrevOptr) && (t0.type == TK_MINUS || t0.type == TK_NK_PLUS)) {
+    if ((isPrevOptr) && (t0.type == TK_NK_MINUS || t0.type == TK_NK_PLUS)) {
       len = tGetToken(&str[*i + t0.n], &type);
       if (type == TK_NK_INTEGER || type == TK_NK_FLOAT) {
         t0.type = type;

source/libs/parser/src/parTranslater.c

@@ -987,12 +987,22 @@ static int32_t translateAlterDatabase(STranslateContext* pCxt, SAlterDatabaseStm
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t calcTypeBytes(SDataType dt) {
+  if (TSDB_DATA_TYPE_BINARY == dt.type) {
+    return dt.bytes + VARSTR_HEADER_SIZE;
+  } else if (TSDB_DATA_TYPE_NCHAR == dt.type) {
+    return dt.bytes * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE;
+  } else {
+    return dt.bytes;
+  }
+}
+
 static int32_t columnNodeToField(SNodeList* pList, SArray** pArray) {
   *pArray = taosArrayInit(LIST_LENGTH(pList), sizeof(SField));
   SNode* pNode;
   FOREACH(pNode, pList) {
     SColumnDefNode* pCol = (SColumnDefNode*)pNode;
-    SField field = { .type = pCol->dataType.type, .bytes = pCol->dataType.bytes };
+    SField field = { .type = pCol->dataType.type, .bytes = calcTypeBytes(pCol->dataType) };
     strcpy(field.name, pCol->colName);
     taosArrayPush(*pArray, &field);
   }

source/libs/parser/test/mockCatalogService.cpp

@@ -16,8 +16,8 @@
 #include <iomanip>
 #include <iostream>
 #include <map>
-#include "tdatablock.h"
 #include "mockCatalogService.h"
+#include "tdatablock.h"
 
 #include "tname.h"
 #include "ttypes.h"

source/libs/qworker/inc/qworkerInt.h

@@ -20,6 +20,7 @@
 extern "C" {
 #endif
 
+#include "qworker.h"
 #include "tlockfree.h"
 #include "ttimer.h"
 
@@ -131,17 +132,16 @@ typedef struct SQWSchStatus {
 
 // Qnode/Vnode level task management
 typedef struct SQWorkerMgmt {
-  SQWorkerCfg      cfg;
-  int8_t           nodeType;
-  int32_t          nodeId;
-  void            *timer;
-  tmr_h            hbTimer;
-  SRWLatch         schLock;
-  SHashObj        *schHash;       //key: schedulerId,    value: SQWSchStatus
-  SHashObj        *ctxHash;       //key: queryId+taskId, value: SQWTaskCtx
-  void            *nodeObj;
-  putReqToQueryQFp putToQueueFp;
-  sendReqFp sendReqFp;
+  SQWorkerCfg cfg;
+  int8_t      nodeType;
+  int32_t     nodeId;
+  void       *timer;
+  tmr_h       hbTimer;
+  SRWLatch    schLock;
+  // SRWLatch ctxLock;
+  SHashObj   *schHash;  // key: schedulerId,    value: SQWSchStatus
+  SHashObj   *ctxHash;  // key: queryId+taskId, value: SQWTaskCtx
+  SMsgCb      msgCb;
 } SQWorkerMgmt;
 
 #define QW_FPARAMS_DEF SQWorkerMgmt *mgmt, uint64_t sId, uint64_t qId, uint64_t tId, int64_t rId

source/libs/qworker/src/qworker.c

@@ -711,12 +711,12 @@ int32_t qwHandlePrePhaseEvents(QW_FPARAMS_DEF, int8_t phase, SQWPhaseInput *inpu
   QW_LOCK(QW_WRITE, &ctx->lock);
 
   if (QW_PHASE_PRE_FETCH == phase) {
-    atomic_store_8(&ctx->queryFetched, true);
+    atomic_store_8((int8_t*)&ctx->queryFetched, true);
   } else {
     atomic_store_8(&ctx->phase, phase);
   }
 
-  if (atomic_load_8(&ctx->queryEnd)) {
+  if (atomic_load_8((int8_t*)&ctx->queryEnd)) {
     QW_TASK_ELOG_E("query already end");
     QW_ERR_JRET(TSDB_CODE_QW_MSG_ERROR);
   }
@@ -978,8 +978,8 @@ int32_t qwProcessReady(QW_FPARAMS_DEF, SQWMsg *qwMsg) {
 
   QW_SET_EVENT_PROCESSED(ctx, QW_EVENT_READY);
 
-  if (atomic_load_8(&ctx->queryEnd) || atomic_load_8(&ctx->queryFetched)) {
-    QW_TASK_ELOG("got ready msg at wrong status, queryEnd:%d, queryFetched:%d", atomic_load_8(&ctx->queryEnd), atomic_load_8(&ctx->queryFetched));
+  if (atomic_load_8((int8_t*)&ctx->queryEnd) || atomic_load_8((int8_t*)&ctx->queryFetched)) {
+    QW_TASK_ELOG("got ready msg at wrong status, queryEnd:%d, queryFetched:%d", atomic_load_8((int8_t*)&ctx->queryEnd), atomic_load_8((int8_t*)&ctx->queryFetched));
     QW_ERR_JRET(TSDB_CODE_QW_MSG_ERROR);
   }
 
@@ -1029,8 +1029,8 @@ int32_t qwProcessCQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg) {
 
     QW_ERR_JRET(qwGetTaskCtx(QW_FPARAMS(), &ctx));
 
-    atomic_store_8(&ctx->queryInQueue, 0);
-    atomic_store_8(&ctx->queryContinue, 0);
+    atomic_store_8((int8_t*)&ctx->queryInQueue, 0);
+    atomic_store_8((int8_t*)&ctx->queryContinue, 0);
 
     QW_ERR_JRET(qwExecTask(QW_FPARAMS(), ctx, &queryEnd));
 
@@ -1041,20 +1041,20 @@ int32_t qwProcessCQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg) {
       if ((!sOutput.queryEnd) && (DS_BUF_LOW == sOutput.bufStatus || DS_BUF_EMPTY == sOutput.bufStatus)) {    
         QW_TASK_DLOG("task not end and buf is %s, need to continue query", qwBufStatusStr(sOutput.bufStatus));
         
-        atomic_store_8(&ctx->queryContinue, 1);
+        atomic_store_8((int8_t*)&ctx->queryContinue, 1);
       }
       
       if (rsp) {
         bool qComplete = (DS_BUF_EMPTY == sOutput.bufStatus && sOutput.queryEnd);
         qwBuildFetchRsp(rsp, &sOutput, dataLen, qComplete);
-        atomic_store_8(&ctx->queryEnd, qComplete);
+        atomic_store_8((int8_t*)&ctx->queryEnd, qComplete);
         
         QW_SET_EVENT_PROCESSED(ctx, QW_EVENT_FETCH);            
         
         qwBuildAndSendFetchRsp(&qwMsg->connInfo, rsp, dataLen, code);                
         QW_TASK_DLOG("fetch rsp send, handle:%p, code:%x - %s, dataLen:%d", qwMsg->connInfo.handle, code, tstrerror(code), dataLen);
       } else {
-        atomic_store_8(&ctx->queryContinue, 1);
+        atomic_store_8((int8_t*)&ctx->queryContinue, 1);
       }
     }
 
@@ -1073,7 +1073,7 @@ _return:
     }
 
     QW_LOCK(QW_WRITE, &ctx->lock);
-    if (queryEnd || code || 0 == atomic_load_8(&ctx->queryContinue)) {
+    if (queryEnd || code || 0 == atomic_load_8((int8_t*)&ctx->queryContinue)) {
       // Note: if necessary, fetch need to put cquery to queue again
       atomic_store_8(&ctx->phase, 0);
       QW_UNLOCK(QW_WRITE,&ctx->lock);
@@ -1111,7 +1111,7 @@ int32_t qwProcessFetch(QW_FPARAMS_DEF, SQWMsg *qwMsg) {
   } else {
     bool qComplete = (DS_BUF_EMPTY == sOutput.bufStatus && sOutput.queryEnd);
     qwBuildFetchRsp(rsp, &sOutput, dataLen, qComplete);
-    atomic_store_8(&ctx->queryEnd, qComplete);
+    atomic_store_8((int8_t*)&ctx->queryEnd, qComplete);
   }
 
   if ((!sOutput.queryEnd) && (DS_BUF_LOW == sOutput.bufStatus || DS_BUF_EMPTY == sOutput.bufStatus)) {    
@@ -1122,11 +1122,11 @@ int32_t qwProcessFetch(QW_FPARAMS_DEF, SQWMsg *qwMsg) {
 
     // RC WARNING
     if (QW_IS_QUERY_RUNNING(ctx)) {
-      atomic_store_8(&ctx->queryContinue, 1);
-    } else if (0 == atomic_load_8(&ctx->queryInQueue)) {
+      atomic_store_8((int8_t*)&ctx->queryContinue, 1);
+    } else if (0 == atomic_load_8((int8_t*)&ctx->queryInQueue)) {
       qwUpdateTaskStatus(QW_FPARAMS(), JOB_TASK_STATUS_EXECUTING);      
 
-      atomic_store_8(&ctx->queryInQueue, 1);
+      atomic_store_8((int8_t*)&ctx->queryInQueue, 1);
       
       QW_ERR_JRET(qwBuildAndSendCQueryMsg(QW_FPARAMS(), &qwMsg->connInfo));
     }
@@ -1309,9 +1309,8 @@ _return:
   taosTmrReset(qwProcessHbTimerEvent, QW_DEFAULT_HEARTBEAT_MSEC, param, mgmt->timer, &mgmt->hbTimer);  
 }
 
-int32_t qWorkerInit(int8_t nodeType, int32_t nodeId, SQWorkerCfg *cfg, void **qWorkerMgmt, void *nodeObj,
-                    putReqToQueryQFp fp1, sendReqFp fp2) {
-  if (NULL == qWorkerMgmt || NULL == nodeObj || NULL == fp1 || NULL == fp2) {
+int32_t qWorkerInit(int8_t nodeType, int32_t nodeId, SQWorkerCfg *cfg, void **qWorkerMgmt, const SMsgCb *pMsgCb) {
+  if (NULL == qWorkerMgmt || pMsgCb->pWrapper == NULL) {
     qError("invalid param to init qworker");
     QW_RET(TSDB_CODE_QRY_INVALID_INPUT);
   }
@@ -1367,9 +1366,7 @@ int32_t qWorkerInit(int8_t nodeType, int32_t nodeId, SQWorkerCfg *cfg, void **qW
 
   mgmt->nodeType = nodeType;
   mgmt->nodeId = nodeId;
-  mgmt->nodeObj = nodeObj;
-  mgmt->putToQueueFp = fp1;
-  mgmt->sendReqFp = fp2;
+  mgmt->msgCb = *pMsgCb;
 
   *qWorkerMgmt = mgmt;
 

source/libs/qworker/src/qworkerMsg.c

@@ -252,7 +252,7 @@ int32_t qwBuildAndSendCQueryMsg(QW_FPARAMS_DEF, SQWConnInfo *pConn) {
     .code    = 0,
   };
 
-  int32_t code = (*mgmt->putToQueueFp)(mgmt->nodeObj, &pNewMsg);
+  int32_t code = tmsgPutToQueue(&mgmt->msgCb, QUERY_QUEUE, &pNewMsg);
   if (TSDB_CODE_SUCCESS != code) {
     QW_SCH_TASK_ELOG("put query continue msg to queue failed, vgId:%d, code:%s", mgmt->nodeId, tstrerror(code));
     rpcFreeCont(req);

source/libs/qworker/test/qworkerTests.cpp

@@ -52,7 +52,7 @@ bool qwtEnableLog = true;
 int32_t qwtTestMaxExecTaskUsec = 2;
 int32_t qwtTestReqMaxDelayUsec = 2;
 
-uint64_t qwtTestQueryId = 0;
+int64_t qwtTestQueryId = 0;
 bool qwtTestEnableSleep = true;
 bool qwtTestStop = false;
 bool qwtTestDeadLoop = false;
@@ -878,7 +878,10 @@ TEST(seqTest, normalCase) {
   stubSetPutDataBlock();
   stubSetGetDataBlock();
   
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   code = qWorkerProcessQueryMsg(mockPointer, mgmt, &queryRpc);
@@ -917,7 +920,10 @@ TEST(seqTest, cancelFirst) {
   stubSetStringToPlan();
   stubSetRpcSendResponse();
   
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   qwtBuildStatusReqMsg(&qwtstatusMsg, &statusRpc);
@@ -963,7 +969,10 @@ TEST(seqTest, randCase) {
 
   taosSeedRand(taosGetTimestampSec());
   
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   int32_t t = 0;
@@ -1034,7 +1043,10 @@ TEST(seqTest, multithreadRand) {
 
   taosSeedRand(taosGetTimestampSec());
   
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   TdThreadAttr thattr;
@@ -1095,7 +1107,10 @@ TEST(rcTest, shortExecshortDelay) {
   qwtTestStop = false;
   qwtTestQuitThreadNum = 0;
 
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   qwtTestMaxExecTaskUsec = 0;
@@ -1176,7 +1191,10 @@ TEST(rcTest, longExecshortDelay) {
   qwtTestStop = false;
   qwtTestQuitThreadNum = 0;
 
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   qwtTestMaxExecTaskUsec = 1000000;
@@ -1259,7 +1277,10 @@ TEST(rcTest, shortExeclongDelay) {
   qwtTestStop = false;
   qwtTestQuitThreadNum = 0;
 
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   qwtTestMaxExecTaskUsec = 0;
@@ -1340,7 +1361,10 @@ TEST(rcTest, dropTest) {
 
   taosSeedRand(taosGetTimestampSec());
   
-  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, mockPointer, qwtPutReqToQueue, (sendReqFp)qwtSendReqToDnode);
+  SMsgCb msgCb = {0};
+  msgCb.pWrapper = (struct SMgmtWrapper *)mockPointer;
+  msgCb.queueFps[QUERY_QUEUE] = (PutToQueueFp)qwtPutReqToQueue;
+  code = qWorkerInit(NODE_TYPE_VNODE, 1, NULL, &mgmt, &msgCb);
   ASSERT_EQ(code, 0);
 
   tsem_init(&qwtTestQuerySem, 0, 0);

source/libs/scheduler/src/scheduler.c

@@ -1158,6 +1158,8 @@ int32_t schHandleHbCallback(void *param, const SDataBuf *pMsg, int32_t code) {
   SCH_RET(schUpdateHbConnection(&rsp.epId, &trans));
 
   int32_t taskNum = (int32_t)taosArrayGetSize(rsp.taskStatus);
+  qDebug("%d task status in hb rsp, nodeId:%d, fqdn:%s, port:%d", taskNum, rsp.epId.nodeId, rsp.epId.ep.fqdn, rsp.epId.ep.port);
+
   for (int32_t i = 0; i < taskNum; ++i) {
     STaskStatus *taskStatus = taosArrayGet(rsp.taskStatus, i);
 

source/libs/sync/inc/syncEnv.h

@@ -31,10 +31,10 @@ extern "C" {
 #define TIMER_MAX_MS 0x7FFFFFFF
 #define ENV_TICK_TIMER_MS 1000
 #define PING_TIMER_MS 1000
-#define ELECT_TIMER_MS_MIN 1500
-#define ELECT_TIMER_MS_MAX 3000
+#define ELECT_TIMER_MS_MIN 150
+#define ELECT_TIMER_MS_MAX (ELECT_TIMER_MS_MIN * 2)
 #define ELECT_TIMER_MS_RANGE (ELECT_TIMER_MS_MAX - ELECT_TIMER_MS_MIN)
-#define HEARTBEAT_TIMER_MS 300
+#define HEARTBEAT_TIMER_MS 30
 
 #define EMPTY_RAFT_ID ((SRaftId){.addr = 0, .vgId = 0})
 

source/libs/sync/inc/syncIO.h

@@ -35,7 +35,7 @@ extern "C" {
 typedef struct SSyncIO {
   STaosQueue *pMsgQ;
   STaosQset * pQset;
-  TdThread   consumerTid;
+  TdThread    consumerTid;
 
   void * serverRpc;
   void * clientRpc;

source/libs/sync/inc/syncInt.h

@@ -160,6 +160,11 @@ typedef struct SSyncNode {
   SSyncLogStore* pLogStore;
   SyncIndex      commitIndex;
 
+  // timer ms init
+  int32_t pingBaseLine;
+  int32_t electBaseLine;
+  int32_t hbBaseLine;
+
   // ping timer
   tmr_h             pPingTimer;
   int32_t           pingTimerMS;
@@ -217,10 +222,12 @@ int32_t syncNodeStartHeartbeatTimer(SSyncNode* pSyncNode);
 int32_t syncNodeStopHeartbeatTimer(SSyncNode* pSyncNode);
 
 // utils --------------
-int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRpcMsg* pMsg);
-int32_t syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, SRpcMsg* pMsg);
-cJSON*  syncNode2Json(const SSyncNode* pSyncNode);
-char*   syncNode2Str(const SSyncNode* pSyncNode);
+int32_t    syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRpcMsg* pMsg);
+int32_t    syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, SRpcMsg* pMsg);
+cJSON*     syncNode2Json(const SSyncNode* pSyncNode);
+char*      syncNode2Str(const SSyncNode* pSyncNode);
+SSyncNode* syncNodeAcquire(int64_t rid);
+void       syncNodeRelease(SSyncNode* pNode);
 
 // raft state change --------------
 void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term);

source/libs/sync/inc/syncUtil.h

@@ -44,7 +44,7 @@ void syncUtilbufCopyDeep(const SSyncBuffer* src, SSyncBuffer* dest);
 
 // ---- misc ----
 int32_t     syncUtilRand(int32_t max);
-int32_t     syncUtilElectRandomMS();
+int32_t     syncUtilElectRandomMS(int32_t min, int32_t max);
 int32_t     syncUtilQuorum(int32_t replicaNum);
 cJSON*      syncUtilNodeInfo2Json(const SNodeInfo* p);
 cJSON*      syncUtilRaftId2Json(const SRaftId* p);

source/libs/sync/src/syncMain.c

@@ -242,9 +242,14 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
   assert(pSyncNode->pLogStore != NULL);
   pSyncNode->commitIndex = SYNC_INDEX_INVALID;
 
+  // timer ms init
+  pSyncNode->pingBaseLine = PING_TIMER_MS;
+  pSyncNode->electBaseLine = ELECT_TIMER_MS_MIN;
+  pSyncNode->hbBaseLine = HEARTBEAT_TIMER_MS;
+
   // init ping timer
   pSyncNode->pPingTimer = NULL;
-  pSyncNode->pingTimerMS = PING_TIMER_MS;
+  pSyncNode->pingTimerMS = pSyncNode->pingBaseLine;
   atomic_store_64(&pSyncNode->pingTimerLogicClock, 0);
   atomic_store_64(&pSyncNode->pingTimerLogicClockUser, 0);
   pSyncNode->FpPingTimerCB = syncNodeEqPingTimer;
@@ -252,7 +257,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
 
   // init elect timer
   pSyncNode->pElectTimer = NULL;
-  pSyncNode->electTimerMS = syncUtilElectRandomMS();
+  pSyncNode->electTimerMS = syncUtilElectRandomMS(pSyncNode->electBaseLine, 2 * pSyncNode->electBaseLine);
   atomic_store_64(&pSyncNode->electTimerLogicClock, 0);
   atomic_store_64(&pSyncNode->electTimerLogicClockUser, 0);
   pSyncNode->FpElectTimerCB = syncNodeEqElectTimer;
@@ -260,7 +265,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo) {
 
   // init heartbeat timer
   pSyncNode->pHeartbeatTimer = NULL;
-  pSyncNode->heartbeatTimerMS = HEARTBEAT_TIMER_MS;
+  pSyncNode->heartbeatTimerMS = pSyncNode->hbBaseLine;
   atomic_store_64(&pSyncNode->heartbeatTimerLogicClock, 0);
   atomic_store_64(&pSyncNode->heartbeatTimerLogicClockUser, 0);
   pSyncNode->FpHeartbeatTimerCB = syncNodeEqHeartbeatTimer;
@@ -394,7 +399,7 @@ int32_t syncNodeRestartElectTimer(SSyncNode* pSyncNode, int32_t ms) {
 
 int32_t syncNodeResetElectTimer(SSyncNode* pSyncNode) {
   int32_t ret = 0;
-  int32_t electMS = syncUtilElectRandomMS();
+  int32_t electMS = syncUtilElectRandomMS(pSyncNode->electBaseLine, 2 * pSyncNode->electBaseLine);
   ret = syncNodeRestartElectTimer(pSyncNode, electMS);
   return ret;
 }
@@ -575,6 +580,17 @@ char* syncNode2Str(const SSyncNode* pSyncNode) {
   return serialized;
 }
 
+SSyncNode* syncNodeAcquire(int64_t rid) {
+  SSyncNode* pNode = taosAcquireRef(tsNodeRefId, rid);
+  if (pNode == NULL) {
+    sTrace("failed to acquire node from refId:%" PRId64, rid);
+  }
+
+  return pNode;
+}
+
+void syncNodeRelease(SSyncNode* pNode) { taosReleaseRef(tsNodeRefId, pNode->rid); }
+
 // raft state change --------------
 void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term) {
   if (term > pSyncNode->pRaftStore->currentTerm) {
@@ -752,7 +768,7 @@ static void syncNodeEqElectTimer(void* param, void* tmrId) {
     syncTimeoutDestroy(pSyncMsg);
 
     // reset timer ms
-    pSyncNode->electTimerMS = syncUtilElectRandomMS();
+    pSyncNode->electTimerMS = syncUtilElectRandomMS(pSyncNode->electBaseLine, 2 * pSyncNode->electBaseLine);
     taosTmrReset(syncNodeEqPingTimer, pSyncNode->pingTimerMS, pSyncNode, gSyncEnv->pTimerManager,
                  &pSyncNode->pPingTimer);
   } else {

source/libs/sync/src/syncUtil.c

@@ -96,7 +96,10 @@ void syncUtilbufCopyDeep(const SSyncBuffer* src, SSyncBuffer* dest) {
 
 int32_t syncUtilRand(int32_t max) { return taosRand() % max; }
 
-int32_t syncUtilElectRandomMS() { return ELECT_TIMER_MS_MIN + syncUtilRand(ELECT_TIMER_MS_RANGE); }
+int32_t syncUtilElectRandomMS(int32_t min, int32_t max) {
+  assert(min > 0 && max > 0 && max >= min);
+  return min + syncUtilRand(max - min);
+}
 
 int32_t syncUtilQuorum(int32_t replicaNum) { return replicaNum / 2 + 1; }
 

source/libs/sync/test/CMakeLists.txt

@@ -28,6 +28,7 @@ add_executable(syncRpcMsgTest "")
 add_executable(syncPingTimerTest2 "")
 add_executable(syncPingSelfTest "")
 add_executable(syncElectTest "")
+add_executable(syncElectTest2 "")
 add_executable(syncEncodeTest "")
 add_executable(syncWriteTest "")
 add_executable(syncReplicateTest "")
@@ -154,6 +155,10 @@ target_sources(syncElectTest
     PRIVATE
     "syncElectTest.cpp"
 )
+target_sources(syncElectTest2
+    PRIVATE
+    "syncElectTest2.cpp"
+)
 target_sources(syncEncodeTest
     PRIVATE
     "syncEncodeTest.cpp"
@@ -322,7 +327,7 @@ target_include_directories(syncElectTest
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
 )
-target_include_directories(syncElectTest
+target_include_directories(syncElectTest2
     PUBLIC
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
@@ -469,6 +474,10 @@ target_link_libraries(syncElectTest
     sync
     gtest_main
 )
+target_link_libraries(syncElectTest2
+    sync
+    gtest_main
+)
 target_link_libraries(syncEncodeTest
     sync
     gtest_main

source/libs/sync/test/syncElectTest.cpp

@@ -120,6 +120,7 @@ int main(int argc, char** argv) {
         "elect sleep, state: %d, %s, term:%lu electTimerLogicClock:%lu, electTimerLogicClockUser:%lu, electTimerMS:%d",
         gSyncNode->state, syncUtilState2String(gSyncNode->state), gSyncNode->pRaftStore->currentTerm,
         gSyncNode->electTimerLogicClock, gSyncNode->electTimerLogicClockUser, gSyncNode->electTimerMS);
+    taosMsleep(1000);
   }
 
   return 0;

source/libs/sync/test/syncElectTest2.cpp

+#include <gtest/gtest.h>
+#include <stdio.h>
+#include "syncEnv.h"
+#include "syncIO.h"
+#include "syncInt.h"
+#include "syncRaftLog.h"
+#include "syncRaftStore.h"
+#include "syncUtil.h"
+#include "tref.h"
+
+void logTest() {
+  sTrace("--- sync log test: trace");
+  sDebug("--- sync log test: debug");
+  sInfo("--- sync log test: info");
+  sWarn("--- sync log test: warn");
+  sError("--- sync log test: error");
+  sFatal("--- sync log test: fatal");
+}
+
+uint16_t ports[] = {7010, 7110, 7210, 7310, 7410};
+int32_t  replicaNum = 3;
+int32_t  myIndex = 0;
+
+SRaftId   ids[TSDB_MAX_REPLICA];
+SSyncInfo syncInfo;
+SSyncFSM* pFsm;
+SWal*     pWal;
+
+int64_t syncNodeInit() {
+  syncInfo.vgId = 1234;
+  syncInfo.rpcClient = gSyncIO->clientRpc;
+  syncInfo.FpSendMsg = syncIOSendMsg;
+  syncInfo.queue = gSyncIO->pMsgQ;
+  syncInfo.FpEqMsg = syncIOEqMsg;
+  syncInfo.pFsm = pFsm;
+  snprintf(syncInfo.path, sizeof(syncInfo.path), "./elect2_test_%d", myIndex);
+
+  int code = walInit();
+  assert(code == 0);
+  SWalCfg walCfg;
+  memset(&walCfg, 0, sizeof(SWalCfg));
+  walCfg.vgId = syncInfo.vgId;
+  walCfg.fsyncPeriod = 1000;
+  walCfg.retentionPeriod = 1000;
+  walCfg.rollPeriod = 1000;
+  walCfg.retentionSize = 1000;
+  walCfg.segSize = 1000;
+  walCfg.level = TAOS_WAL_FSYNC;
+
+  char tmpdir[128];
+  snprintf(tmpdir, sizeof(tmpdir), "./elect2_test_wal_%d", myIndex);
+  pWal = walOpen(tmpdir, &walCfg);
+  assert(pWal != NULL);
+
+  syncInfo.pWal = pWal;
+
+  SSyncCfg* pCfg = &syncInfo.syncCfg;
+  pCfg->myIndex = myIndex;
+  pCfg->replicaNum = replicaNum;
+
+  for (int i = 0; i < replicaNum; ++i) {
+    pCfg->nodeInfo[i].nodePort = ports[i];
+    snprintf(pCfg->nodeInfo[i].nodeFqdn, sizeof(pCfg->nodeInfo[i].nodeFqdn), "%s", "127.0.0.1");
+    // taosGetFqdn(pCfg->nodeInfo[0].nodeFqdn);
+  }
+
+  int64_t rid = syncStart(&syncInfo);
+  assert(rid > 0);
+
+  SSyncNode* pSyncNode = (SSyncNode*)syncNodeAcquire(rid);
+  assert(pSyncNode != NULL);
+
+  gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
+  gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
+  gSyncIO->FpOnSyncRequestVote = pSyncNode->FpOnRequestVote;
+  gSyncIO->FpOnSyncRequestVoteReply = pSyncNode->FpOnRequestVoteReply;
+  gSyncIO->FpOnSyncAppendEntries = pSyncNode->FpOnAppendEntries;
+  gSyncIO->FpOnSyncAppendEntriesReply = pSyncNode->FpOnAppendEntriesReply;
+  gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
+  gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
+  gSyncIO->FpOnSyncTimeout = pSyncNode->FpOnTimeout;
+  gSyncIO->pSyncNode = pSyncNode;
+
+  syncNodeRelease(pSyncNode);
+
+  return rid;
+}
+
+void initRaftId(SSyncNode* pSyncNode) {
+  for (int i = 0; i < replicaNum; ++i) {
+    ids[i] = pSyncNode->replicasId[i];
+    char* s = syncUtilRaftId2Str(&ids[i]);
+    printf("raftId[%d] : %s\n", i, s);
+    free(s);
+  }
+}
+
+int main(int argc, char** argv) {
+  // taosInitLog((char *)"syncTest.log", 100000, 10);
+  tsAsyncLog = 0;
+  sDebugFlag = 143 + 64;
+
+  myIndex = 0;
+  if (argc >= 2) {
+    myIndex = atoi(argv[1]);
+  }
+
+  int32_t ret = syncIOStart((char*)"127.0.0.1", ports[myIndex]);
+  assert(ret == 0);
+
+  ret = syncInit();
+  assert(ret == 0);
+
+  int64_t rid = syncNodeInit();
+  assert(rid > 0);
+
+  SSyncNode* pSyncNode = (SSyncNode*)syncNodeAcquire(rid);
+  assert(pSyncNode != NULL);
+
+  syncNodePrint2((char*)"", pSyncNode);
+  initRaftId(pSyncNode);
+
+  //---------------------------
+  while (1) {
+    sTrace(
+        "elect sleep, state: %d, %s, term:%lu electTimerLogicClock:%lu, electTimerLogicClockUser:%lu, electTimerMS:%d",
+        pSyncNode->state, syncUtilState2String(pSyncNode->state), pSyncNode->pRaftStore->currentTerm,
+        pSyncNode->electTimerLogicClock, pSyncNode->electTimerLogicClockUser, pSyncNode->electTimerMS);
+    taosMsleep(1000);
+  }
+
+  syncNodeRelease(pSyncNode);
+
+  return 0;
+}