Merge branch 'feat/tsdb_refact' of https://github.com/taosdata/TDengine into feat/tsdb_snapshot

Jenkinsfile2

@@ -127,6 +127,25 @@ def pre_test(){
     '''
     return 1
 }
+def pre_test_build_mac() {
+    sh '''
+        hostname
+        date
+    '''
+    sh '''
+        cd ${WK}
+        rm -rf debug
+        mkdir debug
+    '''
+    sh '''
+        cd ${WK}/debug
+        cmake ..
+        make -j8
+    '''
+    sh '''
+        date
+    '''
+}
 def pre_test_win(){
     bat '''
         hostname
@@ -334,6 +353,17 @@ pipeline {
                         }
                     }
                 }
+                stage('mac test') {
+                    agent{label " Mac_catalina "}
+                    steps {
+                        catchError(buildResult: 'FAILURE', stageResult: 'FAILURE') {
+                            timeout(time: 20, unit: 'MINUTES'){
+                                pre_test()
+                                pre_test_build_mac()
+                            }
+                        }
+                    }
+                }
                 stage('linux test') {
                     agent{label " worker03 || slave215 || slave217 || slave219 "}
                     options { skipDefaultCheckout() } 

include/common/tmsg.h

@@ -708,6 +708,7 @@ typedef struct {
   int32_t buffer;        // MB
   int32_t pageSize;
   int32_t pages;
+  int32_t lastRowMem;
   int32_t daysPerFile;
   int32_t daysToKeep0;
   int32_t daysToKeep1;
@@ -736,6 +737,7 @@ typedef struct {
   int32_t buffer;
   int32_t pageSize;
   int32_t pages;
+  int32_t lastRowMem;
   int32_t daysPerFile;
   int32_t daysToKeep0;
   int32_t daysToKeep1;
@@ -1023,8 +1025,10 @@ typedef struct {
   int64_t     clusterId;
   int64_t     rebootTime;
   int64_t     updateTime;
-  int32_t     numOfCores;
+  float       numOfCores;
   int32_t     numOfSupportVnodes;
+  int64_t     memTotal;
+  int64_t     memAvail;
   char        dnodeEp[TSDB_EP_LEN];
   SMnodeLoad  mload;
   SQnodeLoad  qload;
@@ -1079,6 +1083,7 @@ typedef struct {
   int32_t  buffer;
   int32_t  pageSize;
   int32_t  pages;
+  int32_t  lastRowMem;
   int32_t  daysPerFile;
   int32_t  daysToKeep0;
   int32_t  daysToKeep1;
@@ -1131,6 +1136,7 @@ typedef struct {
   int32_t  buffer;
   int32_t  pageSize;
   int32_t  pages;
+  int32_t  lastRowMem;
   int32_t  daysPerFile;
   int32_t  daysToKeep0;
   int32_t  daysToKeep1;

include/libs/executor/executor.h

@@ -37,9 +37,6 @@ typedef struct SReadHandle {
   void*   mnd;
   SMsgCb* pMsgCb;
 
-  /* XXXXXXXXXXXXXXXXXXXX */
-  int32_t deleteQuery;
-  /* XXXXXXXXXXXXXXXXXXXX */
 //  int8_t  initTsdbReader;
   bool    tqReader;
 } SReadHandle;

include/libs/nodes/nodes.h

@@ -23,7 +23,7 @@ extern "C" {
 #include "tdef.h"
 
 #define nodeType(nodeptr)              (((const SNode*)(nodeptr))->type)
-#define setNodeType(nodeptr, type) (((SNode*)(nodeptr))->type = (type))
+#define setNodeType(nodeptr, nodetype) (((SNode*)(nodeptr))->type = (nodetype))
 
 #define LIST_LENGTH(l) (NULL != (l) ? (l)->length : 0)
 
@@ -119,6 +119,7 @@ typedef enum ENodeType {
   QUERY_NODE_DROP_TABLE_STMT,
   QUERY_NODE_DROP_SUPER_TABLE_STMT,
   QUERY_NODE_ALTER_TABLE_STMT,
+  QUERY_NODE_ALTER_SUPER_TABLE_STMT,
   QUERY_NODE_CREATE_USER_STMT,
   QUERY_NODE_ALTER_USER_STMT,
   QUERY_NODE_DROP_USER_STMT,

include/libs/nodes/plannodes.h

@@ -320,6 +320,7 @@ typedef struct SInterpFuncPhysiNode {
   SNodeList*  pFuncs;
   STimeWindow timeRange;
   int64_t     interval;
+  int8_t      intervalUnit;
   EFillMode   fillMode;
   SNode*      pFillValues;  // SNodeListNode
   SNode*      pTimeSeries;  // SColumnNode

include/libs/scalar/scalar.h

@@ -25,6 +25,8 @@ extern "C" {
 
 typedef struct SFilterInfo SFilterInfo;
 
+int32_t scalarGetOperatorResultType(SDataType left, SDataType right, EOperatorType op, SDataType* pRes);
+
 /*
 pNode will be freed in API;
 *pRes need to freed in caller

include/libs/scheduler/scheduler.h

@@ -130,7 +130,7 @@ void schedulerStopQueryHb(void *pTrans);
  * Free the query job
  * @param pJob
  */
-void schedulerFreeJob(int64_t job, int32_t errCode);
+void schedulerFreeJob(int64_t* job, int32_t errCode);
 
 void schedulerDestroy(void);
 

include/libs/sync/sync.h

@@ -31,6 +31,12 @@ extern bool gRaftDetailLog;
 #define SYNC_INDEX_INVALID -1
 #define SYNC_TERM_INVALID 0xFFFFFFFFFFFFFFFF
 
+typedef enum {
+  SYNC_STRATEGY_NO_SNAPSHOT = 0,
+  SYNC_STRATEGY_STANDARD_SNAPSHOT = 1,
+  SYNC_STRATEGY_WAL_FIRST = 2,
+} ESyncStrategy;
+
 typedef uint64_t SyncNodeId;
 typedef int32_t  SyncGroupId;
 typedef int64_t  SyncIndex;
@@ -48,11 +54,6 @@ typedef enum {
   TAOS_SYNC_STATE_ERROR = 103,
 } ESyncState;
 
-typedef enum {
-  TAOS_SYNC_FSM_CB_SUCCESS = 0,
-  TAOS_SYNC_FSM_CB_OTHER_ERROR = 1,
-} ESyncFsmCbCode;
-
 typedef struct SNodeInfo {
   uint16_t nodePort;
   char     nodeFqdn[TSDB_FQDN_LEN];
@@ -96,6 +97,11 @@ typedef struct SReConfigCbMeta {
 
 } SReConfigCbMeta;
 
+typedef struct SSnapshotParam {
+  SyncIndex start;
+  SyncIndex end;
+} SSnapshotParam;
+
 typedef struct SSnapshot {
   void*     data;
   SyncIndex lastApplyIndex;
@@ -125,7 +131,7 @@ typedef struct SSyncFSM {
   int32_t (*FpSnapshotStopRead)(struct SSyncFSM* pFsm, void* pReader);
   int32_t (*FpSnapshotDoRead)(struct SSyncFSM* pFsm, void* pReader, void** ppBuf, int32_t* len);
 
-  int32_t (*FpSnapshotStartWrite)(struct SSyncFSM* pFsm, void** ppWriter);
+  int32_t (*FpSnapshotStartWrite)(struct SSyncFSM* pFsm, void* pWriterParam, void** ppWriter);
   int32_t (*FpSnapshotStopWrite)(struct SSyncFSM* pFsm, void* pWriter, bool isApply);
   int32_t (*FpSnapshotDoWrite)(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_t len);
 
@@ -179,7 +185,7 @@ typedef struct SSyncLogStore {
 
 typedef struct SSyncInfo {
   bool          isStandBy;
-  bool        snapshotEnable;
+  ESyncStrategy snapshotStrategy;
   SyncGroupId   vgId;
   int32_t       batchSize;
   SSyncCfg      syncCfg;
@@ -205,7 +211,7 @@ SyncGroupId syncGetVgId(int64_t rid);
 void        syncGetEpSet(int64_t rid, SEpSet* pEpSet);
 void        syncGetRetryEpSet(int64_t rid, SEpSet* pEpSet);
 int32_t     syncPropose(int64_t rid, SRpcMsg* pMsg, bool isWeak);
-// int32_t     syncProposeBatch(int64_t rid, SRpcMsg* pMsgArr, bool* pIsWeakArr, int32_t arrSize);
+int32_t     syncProposeBatch(int64_t rid, SRpcMsg* pMsgArr, bool* pIsWeakArr, int32_t arrSize);
 bool        syncEnvIsStart();
 const char* syncStr(ESyncState state);
 bool        syncIsRestoreFinish(int64_t rid);

include/libs/sync/syncTools.h

@@ -191,12 +191,12 @@ void syncTimeoutLog2(char* s, const SyncTimeout* pMsg);
 typedef struct SyncClientRequest {
   uint32_t bytes;
   int32_t  vgId;
-  uint32_t msgType;          // SyncClientRequest msgType
+  uint32_t msgType;          // TDMT_SYNC_CLIENT_REQUEST
-  uint32_t originalRpcType;  // user RpcMsg msgType
+  uint32_t originalRpcType;  // origin RpcMsg msgType
   uint64_t seqNum;
   bool     isWeak;
-  uint32_t dataLen;  // user RpcMsg.contLen
+  uint32_t dataLen;  // origin RpcMsg.contLen
-  char     data[];   // user RpcMsg.pCont
+  char     data[];   // origin RpcMsg.pCont
 } SyncClientRequest;
 
 SyncClientRequest* syncClientRequestBuild(uint32_t dataLen);
@@ -220,11 +220,6 @@ void syncClientRequestLog(const SyncClientRequest* pMsg);
 void syncClientRequestLog2(char* s, const SyncClientRequest* pMsg);
 
 // ---------------------------------------------
-typedef struct SOffsetAndContLen {
-  int32_t offset;
-  int32_t contLen;
-} SOffsetAndContLen;
-
 typedef struct SRaftMeta {
   uint64_t seqNum;
   bool     isWeak;
@@ -232,20 +227,33 @@ typedef struct SRaftMeta {
 
 // block1:
 // block2: SRaftMeta array
-// block3: rpc msg array (with pCont)
+// block3: rpc msg array (with pCont pointer)
 
 typedef struct SyncClientRequestBatch {
   uint32_t bytes;
   int32_t  vgId;
-  uint32_t msgType;  // SyncClientRequestBatch msgType
+  uint32_t msgType;  // TDMT_SYNC_CLIENT_REQUEST_BATCH
   uint32_t dataCount;
-  uint32_t dataLen;  // user RpcMsg.contLen
+  uint32_t dataLen;
-  char     data[];   // user RpcMsg.pCont
+  char     data[];  // block2, block3
 } SyncClientRequestBatch;
 
 SyncClientRequestBatch* syncClientRequestBatchBuild(SRpcMsg* rpcMsgArr, SRaftMeta* raftArr, int32_t arrSize,
                                                     int32_t vgId);
 void                    syncClientRequestBatch2RpcMsg(const SyncClientRequestBatch* pSyncMsg, SRpcMsg* pRpcMsg);
+void                    syncClientRequestBatchDestroy(SyncClientRequestBatch* pMsg);
+void                    syncClientRequestBatchDestroyDeep(SyncClientRequestBatch* pMsg);
+SRaftMeta*              syncClientRequestBatchMetaArr(const SyncClientRequestBatch* pSyncMsg);
+SRpcMsg*                syncClientRequestBatchRpcMsgArr(const SyncClientRequestBatch* pSyncMsg);
+SyncClientRequestBatch* syncClientRequestBatchFromRpcMsg(const SRpcMsg* pRpcMsg);
+cJSON*                  syncClientRequestBatch2Json(const SyncClientRequestBatch* pMsg);
+char*                   syncClientRequestBatch2Str(const SyncClientRequestBatch* pMsg);
+
+// for debug ----------------------
+void syncClientRequestBatchPrint(const SyncClientRequestBatch* pMsg);
+void syncClientRequestBatchPrint2(char* s, const SyncClientRequestBatch* pMsg);
+void syncClientRequestBatchLog(const SyncClientRequestBatch* pMsg);
+void syncClientRequestBatchLog2(char* s, const SyncClientRequestBatch* pMsg);
 
 // ---------------------------------------------
 typedef struct SyncClientRequestReply {
@@ -318,12 +326,15 @@ void syncRequestVoteReplyLog(const SyncRequestVoteReply* pMsg);
 void syncRequestVoteReplyLog2(char* s, const SyncRequestVoteReply* pMsg);
 
 // ---------------------------------------------
+// data: entry
+
 typedef struct SyncAppendEntries {
   uint32_t bytes;
   int32_t  vgId;
   uint32_t msgType;
   SRaftId  srcId;
   SRaftId  destId;
+
   // private data
   SyncTerm  term;
   SyncIndex prevLogIndex;
@@ -354,18 +365,14 @@ void syncAppendEntriesLog2(char* s, const SyncAppendEntries* pMsg);
 
 // ---------------------------------------------
 
-// define ahead
-/*
 typedef struct SOffsetAndContLen {
   int32_t offset;
   int32_t contLen;
 } SOffsetAndContLen;
-*/
 
-// block1: SOffsetAndContLen
+// data:
-// block2: SOffsetAndContLen Array
+// block1: SOffsetAndContLen Array
-// block3: SRpcMsg Array
+// block2: entry Array
-// block4: SRpcMsg pCont Array
 
 typedef struct SyncAppendEntriesBatch {
   uint32_t bytes;
@@ -382,10 +389,11 @@ typedef struct SyncAppendEntriesBatch {
   SyncTerm  privateTerm;
   int32_t   dataCount;
   uint32_t  dataLen;
-  char      data[];
+  char      data[];  // block1, block2
 } SyncAppendEntriesBatch;
 
-SyncAppendEntriesBatch* syncAppendEntriesBatchBuild(SRpcMsg* rpcMsgArr, int32_t arrSize, int32_t vgId);
+SyncAppendEntriesBatch* syncAppendEntriesBatchBuild(SSyncRaftEntry** entryPArr, int32_t arrSize, int32_t vgId);
+SOffsetAndContLen*      syncAppendEntriesBatchMetaTableArray(SyncAppendEntriesBatch* pMsg);
 void                    syncAppendEntriesBatchDestroy(SyncAppendEntriesBatch* pMsg);
 void                    syncAppendEntriesBatchSerialize(const SyncAppendEntriesBatch* pMsg, char* buf, uint32_t bufLen);
 void                    syncAppendEntriesBatchDeserialize(const char* buf, uint32_t len, SyncAppendEntriesBatch* pMsg);
@@ -396,8 +404,6 @@ void                    syncAppendEntriesBatchFromRpcMsg(const SRpcMsg* pRpcMsg,
 SyncAppendEntriesBatch* syncAppendEntriesBatchFromRpcMsg2(const SRpcMsg* pRpcMsg);
 cJSON*                  syncAppendEntriesBatch2Json(const SyncAppendEntriesBatch* pMsg);
 char*                   syncAppendEntriesBatch2Str(const SyncAppendEntriesBatch* pMsg);
-void syncAppendEntriesBatch2RpcMsgArray(SyncAppendEntriesBatch* pSyncMsg, SRpcMsg* rpcMsgArr, int32_t maxArrSize,
-                                        int32_t* pRetArrSize);
 
 // for debug ----------------------
 void syncAppendEntriesBatchPrint(const SyncAppendEntriesBatch* pMsg);
@@ -477,9 +483,10 @@ typedef struct SyncSnapshotSend {
   SRaftId  destId;
 
   SyncTerm  term;
-  SyncIndex lastIndex;  // lastIndex of snapshot
+  SyncIndex beginIndex;       // snapshot.beginIndex
-  SyncTerm  lastTerm;   // lastTerm of snapshot
+  SyncIndex lastIndex;        // snapshot.lastIndex
-  SyncIndex lastConfigIndex;
+  SyncTerm  lastTerm;         // snapshot.lastTerm
+  SyncIndex lastConfigIndex;  // snapshot.lastConfigIndex
   SSyncCfg  lastConfig;
   SyncTerm  privateTerm;
   int32_t   seq;
@@ -617,6 +624,9 @@ int32_t syncNodeOnRequestVoteReplySnapshotCb(SSyncNode* ths, SyncRequestVoteRepl
 int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg);
 int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
 
+int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatch* pMsg);
+int32_t syncNodeOnAppendEntriesReplySnapshot2Cb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
+
 int32_t syncNodeOnSnapshotSendCb(SSyncNode* ths, SyncSnapshotSend* pMsg);
 int32_t syncNodeOnSnapshotRspCb(SSyncNode* ths, SyncSnapshotRsp* pMsg);
 
@@ -634,6 +644,7 @@ typedef int32_t (*FpOnSnapshotRspCb)(SSyncNode* ths, SyncSnapshotRsp* pMsg);
 
 // option ----------------------------------
 bool          syncNodeSnapshotEnable(SSyncNode* pSyncNode);
+ESyncStrategy syncNodeStrategy(SSyncNode* pSyncNode);
 
 // ---------------------------------------------
 

include/libs/transport/trpc.h

@@ -110,9 +110,12 @@ typedef struct {
 } SRpcCtx;
 
 int32_t rpcInit();
+
 void  rpcCleanup();
 void *rpcOpen(const SRpcInit *pRpc);
+
 void  rpcClose(void *);
+void  rpcCloseImpl(void *);
 void *rpcMallocCont(int32_t contLen);
 void  rpcFreeCont(void *pCont);
 void *rpcReallocCont(void *ptr, int32_t contLen);

include/util/tRealloc.h

@@ -38,7 +38,7 @@ static FORCE_INLINE int32_t tRealloc(uint8_t **ppBuf, int64_t size) {
     bsize *= 2;
   }
 
-  pBuf = taosMemoryRealloc(*ppBuf ? (*ppBuf) - sizeof(int64_t) : *ppBuf, bsize + sizeof(int64_t));
+  pBuf = (uint8_t *)taosMemoryRealloc(*ppBuf ? (*ppBuf) - sizeof(int64_t) : *ppBuf, bsize + sizeof(int64_t));
   if (pBuf == NULL) {
     code = TSDB_CODE_OUT_OF_MEMORY;
     goto _exit;

include/util/taoserror.h

@@ -174,11 +174,12 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_MND_DNODE_NOT_EXIST           TAOS_DEF_ERROR_CODE(0, 0x0341)
 #define TSDB_CODE_MND_TOO_MANY_DNODES           TAOS_DEF_ERROR_CODE(0, 0x0342)
 #define TSDB_CODE_MND_NO_ENOUGH_DNODES          TAOS_DEF_ERROR_CODE(0, 0x0343)
-#define TSDB_CODE_MND_INVALID_CLUSTER_CFG       TAOS_DEF_ERROR_CODE(0, 0x0344)
+#define TSDB_CODE_MND_NO_ENOUGH_MEM_IN_DNODE    TAOS_DEF_ERROR_CODE(0, 0x0344)
-#define TSDB_CODE_MND_INVALID_CLUSTER_ID        TAOS_DEF_ERROR_CODE(0, 0x0345)
+#define TSDB_CODE_MND_INVALID_CLUSTER_CFG       TAOS_DEF_ERROR_CODE(0, 0x0345)
-#define TSDB_CODE_MND_INVALID_DNODE_CFG         TAOS_DEF_ERROR_CODE(0, 0x0346)
+#define TSDB_CODE_MND_INVALID_CLUSTER_ID        TAOS_DEF_ERROR_CODE(0, 0x0346)
-#define TSDB_CODE_MND_INVALID_DNODE_EP          TAOS_DEF_ERROR_CODE(0, 0x0347)
+#define TSDB_CODE_MND_INVALID_DNODE_CFG         TAOS_DEF_ERROR_CODE(0, 0x0347)
-#define TSDB_CODE_MND_INVALID_DNODE_ID          TAOS_DEF_ERROR_CODE(0, 0x0348)
+#define TSDB_CODE_MND_INVALID_DNODE_EP          TAOS_DEF_ERROR_CODE(0, 0x0348)
+#define TSDB_CODE_MND_INVALID_DNODE_ID          TAOS_DEF_ERROR_CODE(0, 0x0349)
 
 // mnode-node
 #define TSDB_CODE_MND_MNODE_ALREADY_EXIST       TAOS_DEF_ERROR_CODE(0, 0x0350)
@@ -429,6 +430,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_SYN_PROPOSE_NOT_READY         TAOS_DEF_ERROR_CODE(0, 0x0911)
 #define TSDB_CODE_SYN_STANDBY_NOT_READY         TAOS_DEF_ERROR_CODE(0, 0x0912)
 #define TSDB_CODE_SYN_BATCH_ERROR               TAOS_DEF_ERROR_CODE(0, 0x0913)
+#define TSDB_CODE_SYN_TIMEOUT                   TAOS_DEF_ERROR_CODE(0, 0x0914)
 #define TSDB_CODE_SYN_INTERNAL_ERROR            TAOS_DEF_ERROR_CODE(0, 0x09FF)
 
 // tq
@@ -579,6 +581,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_PAR_INVALID_TABLE_OPTION      TAOS_DEF_ERROR_CODE(0, 0x265C)
 #define TSDB_CODE_PAR_INVALID_INTERP_CLAUSE     TAOS_DEF_ERROR_CODE(0, 0x265D)
 #define TSDB_CODE_PAR_NO_VALID_FUNC_IN_WIN      TAOS_DEF_ERROR_CODE(0, 0x265E)
+#define TSDB_CODE_PAR_ONLY_SUPPORT_SINGLE_TABLE TAOS_DEF_ERROR_CODE(0, 0x265F)
 
 //planner
 #define TSDB_CODE_PLAN_INTERNAL_ERROR           TAOS_DEF_ERROR_CODE(0, 0x2700)

include/util/tdef.h

@@ -334,6 +334,9 @@ typedef enum ELogicConditionType {
 #define TSDB_MIN_DB_CACHE_LAST_ROW      0
 #define TSDB_MAX_DB_CACHE_LAST_ROW      3
 #define TSDB_DEFAULT_CACHE_LAST_ROW     0
+#define TSDB_MIN_DB_LAST_ROW_MEM        1  // MB
+#define TSDB_MAX_DB_LAST_ROW_MEM        65536
+#define TSDB_DEFAULT_LAST_ROW_MEM       1
 #define TSDB_DB_STREAM_MODE_OFF         0
 #define TSDB_DB_STREAM_MODE_ON          1
 #define TSDB_DEFAULT_DB_STREAM_MODE     0

source/client/inc/clientInt.h

@@ -284,6 +284,7 @@ static FORCE_INLINE SReqResultInfo* tscGetCurResInfo(TAOS_RES* res) {
 extern SAppInfo appInfo;
 extern int32_t  clientReqRefPool;
 extern int32_t  clientConnRefPool;
+extern void*    tscQhandle;
 
 __async_send_cb_fn_t getMsgRspHandle(int32_t msgType);
 
@@ -337,7 +338,8 @@ int  hbHandleRsp(SClientHbBatchRsp* hbRsp);
 SAppHbMgr* appHbMgrInit(SAppInstInfo* pAppInstInfo, char* key);
 void       appHbMgrCleanup(void);
 void       hbRemoveAppHbMrg(SAppHbMgr** pAppHbMgr);
-void       closeAllRequests(SHashObj *pRequests);
+void       destroyAllRequests(SHashObj* pRequests);
+void       stopAllRequests(SHashObj* pRequests);
 
 // conn level
 int  hbRegisterConn(SAppHbMgr* pAppHbMgr, int64_t tscRefId, int64_t clusterId, int8_t connType);
@@ -356,6 +358,9 @@ int32_t      removeMeta(STscObj* pTscObj, SArray* tbList);  // todo move to clie
 int32_t      handleAlterTbExecRes(void* res, struct SCatalog* pCatalog);  // todo move to xxx
 bool         qnodeRequired(SRequestObj* pRequest);
 
+void initTscQhandle();
+void cleanupTscQhandle();
+
 #ifdef __cplusplus
 }
 #endif

source/client/src/clientEnv.c

@@ -25,6 +25,7 @@
 #include "tmsg.h"
 #include "tref.h"
 #include "trpc.h"
+#include "tsched.h"
 #include "ttime.h"
 
 #define TSC_VAR_NOT_RELEASE 1
@@ -34,9 +35,20 @@ SAppInfo appInfo;
 int32_t  clientReqRefPool = -1;
 int32_t  clientConnRefPool = -1;
 
+void *tscQhandle = NULL;
+
 static TdThreadOnce tscinit = PTHREAD_ONCE_INIT;
 volatile int32_t    tscInitRes = 0;
 
+void initTscQhandle() {
+  // init handle
+  tscQhandle = taosInitScheduler(4096, 5, "tsc");
+}
+
+void cleanupTscQhandle() {
+  // destroy handle
+  taosCleanUpScheduler(tscQhandle);
+}
 static int32_t registerRequest(SRequestObj *pRequest) {
   STscObj *pTscObj = acquireTscObj(pRequest->pTscObj->id);
   if (NULL == pTscObj) {
@@ -121,12 +133,31 @@ void *openTransporter(const char *user, const char *auth, int32_t numOfThread) {
   return pDnodeConn;
 }
 
-void closeAllRequests(SHashObj *pRequests) {
+void destroyAllRequests(SHashObj *pRequests) {
   void *pIter = taosHashIterate(pRequests, NULL);
   while (pIter != NULL) {
     int64_t *rid = pIter;
 
-    removeRequest(*rid);
+    SRequestObj *pRequest = acquireRequest(*rid);
+    if (pRequest) {
+      destroyRequest(pRequest);
+      releaseRequest(*rid);
+    }
+
+    pIter = taosHashIterate(pRequests, pIter);
+  }
+}
+
+void stopAllRequests(SHashObj *pRequests) {
+  void *pIter = taosHashIterate(pRequests, NULL);
+  while (pIter != NULL) {
+    int64_t *rid = pIter;
+
+    SRequestObj *pRequest = acquireRequest(*rid);
+    if (pRequest) {
+      taos_stop_query(pRequest);
+      releaseRequest(*rid);
+    }
 
     pIter = taosHashIterate(pRequests, pIter);
   }
@@ -153,12 +184,18 @@ void destroyAppInst(SAppInstInfo *pAppInfo) {
 }
 
 void destroyTscObj(void *pObj) {
+  if (NULL == pObj) {
+    return;
+  }
+
   STscObj *pTscObj = pObj;
+  int64_t  tscId = pTscObj->id;
+  tscTrace("begin to destroy tscObj %" PRIx64 " p:%p", tscId, pTscObj);
 
   SClientHbKey connKey = {.tscRid = pTscObj->id, .connType = pTscObj->connType};
   hbDeregisterConn(pTscObj->pAppInfo->pAppHbMgr, connKey);
   int64_t connNum = atomic_sub_fetch_64(&pTscObj->pAppInfo->numOfConns, 1);
-  closeAllRequests(pTscObj->pRequests);
+  destroyAllRequests(pTscObj->pRequests);
   schedulerStopQueryHb(pTscObj->pAppInfo->pTransporter);
   tscDebug("connObj 0x%" PRIx64 " p:%p destroyed, remain inst totalConn:%" PRId64, pTscObj->id, pTscObj,
            pTscObj->pAppInfo->numOfConns);
@@ -167,7 +204,9 @@ void destroyTscObj(void *pObj) {
     destroyAppInst(pTscObj->pAppInfo);
   }
   taosThreadMutexDestroy(&pTscObj->mutex);
-  taosMemoryFreeClear(pTscObj);
+  taosMemoryFree(pTscObj);
+
+  tscTrace("end to destroy tscObj %" PRIx64 " p:%p", tscId, pTscObj);
 }
 
 void *createTscObj(const char *user, const char *auth, const char *db, int32_t connType, SAppInstInfo *pAppInfo) {
@@ -261,14 +300,18 @@ int32_t releaseRequest(int64_t rid) { return taosReleaseRef(clientReqRefPool, ri
 int32_t removeRequest(int64_t rid) { return taosRemoveRef(clientReqRefPool, rid); }
 
 void doDestroyRequest(void *p) {
-  assert(p != NULL);
+  if (NULL == p) {
+    return;
+  }
+
   SRequestObj *pRequest = (SRequestObj *)p;
 
+  int64_t reqId = pRequest->self;
+  tscTrace("begin to destroy request %" PRIx64 " p:%p", reqId, pRequest);
+
   taosHashRemove(pRequest->pTscObj->pRequests, &pRequest->self, sizeof(pRequest->self));
 
-  if (pRequest->body.queryJob != 0) {
+  schedulerFreeJob(&pRequest->body.queryJob, 0);
-    schedulerFreeJob(pRequest->body.queryJob, 0);
-  }
 
   taosMemoryFreeClear(pRequest->msgBuf);
   taosMemoryFreeClear(pRequest->sqlstr);
@@ -284,7 +327,9 @@ void doDestroyRequest(void *p) {
   if (pRequest->self) {
     deregisterRequest(pRequest);
   }
-  taosMemoryFreeClear(pRequest);
+  taosMemoryFree(pRequest);
+
+  tscTrace("end to destroy request %" PRIx64 " p:%p", reqId, pRequest);
 }
 
 void destroyRequest(SRequestObj *pRequest) {
@@ -292,6 +337,8 @@ void destroyRequest(SRequestObj *pRequest) {
     return;
   }
 
+  taos_stop_query(pRequest);
+
   removeRequest(pRequest->self);
 }
 
@@ -299,7 +346,7 @@ void taos_init_imp(void) {
   // In the APIs of other program language, taos_cleanup is not available yet.
   // So, to make sure taos_cleanup will be invoked to clean up the allocated resource to suppress the valgrind warning.
   atexit(taos_cleanup);
-
+  initTscQhandle();
   errno = TSDB_CODE_SUCCESS;
   taosSeedRand(taosGetTimestampSec());
 

source/client/src/clientImpl.c

@@ -25,6 +25,7 @@
 #include "tmsgtype.h"
 #include "tpagedbuf.h"
 #include "tref.h"
+#include "tsched.h"
 
 static int32_t       initEpSetFromCfg(const char* firstEp, const char* secondEp, SCorEpSet* pEpSet);
 static SMsgSendInfo* buildConnectMsg(SRequestObj* pRequest);
@@ -645,9 +646,7 @@ int32_t scheduleQuery(SRequestObj* pRequest, SQueryPlan* pDag, SArray* pNodeList
   pRequest->body.resInfo.execRes = res.res;
 
   if (code != TSDB_CODE_SUCCESS) {
-    if (pRequest->body.queryJob != 0) {
+    schedulerFreeJob(&pRequest->body.queryJob, 0);
-      schedulerFreeJob(pRequest->body.queryJob, 0);
-    }
 
     pRequest->code = code;
     terrno = code;
@@ -658,9 +657,7 @@ int32_t scheduleQuery(SRequestObj* pRequest, SQueryPlan* pDag, SArray* pNodeList
       TDMT_VND_CREATE_TABLE == pRequest->type) {
     pRequest->body.resInfo.numOfRows = res.numOfRows;
 
-    if (pRequest->body.queryJob != 0) {
+    schedulerFreeJob(&pRequest->body.queryJob, 0);
-      schedulerFreeJob(pRequest->body.queryJob, 0);
-    }
   }
 
   pRequest->code = res.code;
@@ -792,10 +789,7 @@ void schedulerExecCb(SQueryResult* pResult, void* param, int32_t code) {
       TDMT_VND_CREATE_TABLE == pRequest->type) {
     pRequest->body.resInfo.numOfRows = pResult->numOfRows;
 
-    if (pRequest->body.queryJob != 0) {
+    schedulerFreeJob(&pRequest->body.queryJob, 0);
-      schedulerFreeJob(pRequest->body.queryJob, 0);
-      pRequest->body.queryJob = 0;
-    }
   }
 
   taosMemoryFree(pResult);
@@ -1239,7 +1233,16 @@ void updateTargetEpSet(SMsgSendInfo* pSendInfo, STscObj* pTscObj, SRpcMsg* pMsg,
   }
 }
 
-void processMsgFromServer(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet) {
+typedef struct SchedArg {
+  SRpcMsg msg;
+  SEpSet* pEpset;
+} SchedArg;
+
+void doProcessMsgFromServer(SSchedMsg* schedMsg) {
+  SchedArg* arg = (SchedArg*)schedMsg->ahandle;
+  SRpcMsg*  pMsg = &arg->msg;
+  SEpSet*   pEpSet = arg->pEpset;
+
   SMsgSendInfo* pSendInfo = (SMsgSendInfo*)pMsg->info.ahandle;
   assert(pMsg->info.ahandle != NULL);
   STscObj* pTscObj = NULL;
@@ -1272,7 +1275,8 @@ void processMsgFromServer(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet) {
 
   updateTargetEpSet(pSendInfo, pTscObj, pMsg, pEpSet);
 
-  SDataBuf buf = {.msgType = pMsg->msgType, .len = pMsg->contLen, .pData = NULL, .handle = pMsg->info.handle, .pEpSet = pEpSet};
+  SDataBuf buf = {
+      .msgType = pMsg->msgType, .len = pMsg->contLen, .pData = NULL, .handle = pMsg->info.handle, .pEpSet = pEpSet};
 
   if (pMsg->contLen > 0) {
     buf.pData = taosMemoryCalloc(1, pMsg->contLen);
@@ -1287,6 +1291,25 @@ void processMsgFromServer(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet) {
   pSendInfo->fp(pSendInfo->param, &buf, pMsg->code);
   rpcFreeCont(pMsg->pCont);
   destroySendMsgInfo(pSendInfo);
+
+  taosMemoryFree(arg);
+}
+
+void processMsgFromServer(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet) {
+  SSchedMsg schedMsg = {0};
+
+  SEpSet* tEpSet = pEpSet != NULL ? taosMemoryCalloc(1, sizeof(SEpSet)) : NULL;
+  if (tEpSet != NULL) {
+    *tEpSet = *pEpSet;
+  }
+
+  SchedArg* arg = taosMemoryCalloc(1, sizeof(SchedArg));
+  arg->msg = *pMsg;
+  arg->pEpset = tEpSet;
+
+  schedMsg.fp = doProcessMsgFromServer;
+  schedMsg.ahandle = arg;
+  taosScheduleTask(tscQhandle, &schedMsg);
 }
 
 TAOS* taos_connect_auth(const char* ip, const char* user, const char* auth, const char* db, uint16_t port) {

source/client/src/clientMain.c

@@ -47,11 +47,9 @@ int taos_options(TSDB_OPTION option, const void *arg, ...) {
   atomic_store_32(&lock, 0);
   return ret;
 }
-
 // this function may be called by user or system, or by both simultaneously.
 void taos_cleanup(void) {
   tscInfo("start  to cleanup client environment");
-
   if (atomic_val_compare_exchange_32(&sentinel, TSC_VAR_NOT_RELEASE, TSC_VAR_RELEASED) != TSC_VAR_NOT_RELEASE) {
     return;
   }
@@ -74,8 +72,8 @@ void taos_cleanup(void) {
   catalogDestroy();
   schedulerDestroy();
 
+  cleanupTscQhandle();
   rpcCleanup();
-
   tscInfo("all local resources released");
   taosCleanupCfg();
   taosCloseLog();
@@ -196,10 +194,10 @@ void taos_kill_query(TAOS *taos) {
   if (NULL == taos) {
     return;
   }
-  int64_t rid = *(int64_t*)taos;
 
+  int64_t  rid = *(int64_t *)taos;
   STscObj *pTscObj = acquireTscObj(rid);
-  closeAllRequests(pTscObj->pRequests);
+  stopAllRequests(pTscObj->pRequests);
   releaseTscObj(rid);
 }
 
@@ -480,9 +478,7 @@ void taos_stop_query(TAOS_RES *res) {
     return;
   }
 
-  if (pRequest->body.queryJob) {
+  schedulerFreeJob(&pRequest->body.queryJob, TSDB_CODE_TSC_QUERY_KILLED);
-    schedulerFreeJob(pRequest->body.queryJob, TSDB_CODE_TSC_QUERY_KILLED);
-  }
 
   tscDebug("request %" PRIx64 " killed", pRequest->requestId);
 }
@@ -722,8 +718,7 @@ int32_t createParseContext(const SRequestObj *pRequest, SParseContext **pCxt) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
 
-  **pCxt = (SParseContext){
+  **pCxt = (SParseContext){.requestId = pRequest->requestId,
-      .requestId = pRequest->requestId,
                            .requestRid = pRequest->self,
                            .acctId = pTscObj->acctId,
                            .db = pRequest->pDb,
@@ -739,8 +734,7 @@ int32_t createParseContext(const SRequestObj *pRequest, SParseContext **pCxt) {
                            .isSuperUser = (0 == strcmp(pTscObj->user, TSDB_DEFAULT_USER)),
                            .async = true,
                            .svrVer = pTscObj->sVer,
-      .nodeOffline = (pTscObj->pAppInfo->onlineDnodes < pTscObj->pAppInfo->totalDnodes)
+                           .nodeOffline = (pTscObj->pAppInfo->onlineDnodes < pTscObj->pAppInfo->totalDnodes)};
-  };
   return TSDB_CODE_SUCCESS;
 }
 

source/common/src/tdatablock.c

@@ -923,6 +923,7 @@ int32_t blockDataSort(SSDataBlock* pDataBlock, SArray* pOrderInfo) {
     pInfo->pColData = taosArrayGet(pDataBlock->pDataBlock, pInfo->slotId);
   }
 
+  terrno = 0;
   taosqsort(index, rows, sizeof(int32_t), &helper, dataBlockCompar);
   if (terrno) return terrno;
 

source/common/src/tmsg.c

@@ -947,8 +947,10 @@ int32_t tSerializeSStatusReq(void *buf, int32_t bufLen, SStatusReq *pReq) {
   if (tEncodeI64(&encoder, pReq->clusterId) < 0) return -1;
   if (tEncodeI64(&encoder, pReq->rebootTime) < 0) return -1;
   if (tEncodeI64(&encoder, pReq->updateTime) < 0) return -1;
-  if (tEncodeI32(&encoder, pReq->numOfCores) < 0) return -1;
+  if (tEncodeFloat(&encoder, pReq->numOfCores) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->numOfSupportVnodes) < 0) return -1;
+  if (tEncodeI64(&encoder, pReq->memTotal) < 0) return -1;
+  if (tEncodeI64(&encoder, pReq->memAvail) < 0) return -1;
   if (tEncodeCStr(&encoder, pReq->dnodeEp) < 0) return -1;
 
   // cluster cfg
@@ -1008,8 +1010,10 @@ int32_t tDeserializeSStatusReq(void *buf, int32_t bufLen, SStatusReq *pReq) {
   if (tDecodeI64(&decoder, &pReq->clusterId) < 0) return -1;
   if (tDecodeI64(&decoder, &pReq->rebootTime) < 0) return -1;
   if (tDecodeI64(&decoder, &pReq->updateTime) < 0) return -1;
-  if (tDecodeI32(&decoder, &pReq->numOfCores) < 0) return -1;
+  if (tDecodeFloat(&decoder, &pReq->numOfCores) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->numOfSupportVnodes) < 0) return -1;
+  if (tDecodeI64(&decoder, &pReq->memTotal) < 0) return -1;
+  if (tDecodeI64(&decoder, &pReq->memAvail) < 0) return -1;
   if (tDecodeCStrTo(&decoder, pReq->dnodeEp) < 0) return -1;
 
   // cluster cfg
@@ -1974,6 +1978,7 @@ int32_t tSerializeSCreateDbReq(void *buf, int32_t bufLen, SCreateDbReq *pReq) {
   if (tEncodeI32(&encoder, pReq->buffer) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pageSize) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pages) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->lastRowMem) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysPerFile) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep0) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep1) < 0) return -1;
@@ -2015,6 +2020,7 @@ int32_t tDeserializeSCreateDbReq(void *buf, int32_t bufLen, SCreateDbReq *pReq)
   if (tDecodeI32(&decoder, &pReq->buffer) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pageSize) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pages) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->lastRowMem) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysPerFile) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep0) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep1) < 0) return -1;
@@ -2069,6 +2075,7 @@ int32_t tSerializeSAlterDbReq(void *buf, int32_t bufLen, SAlterDbReq *pReq) {
   if (tEncodeI32(&encoder, pReq->buffer) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pageSize) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pages) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->lastRowMem) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysPerFile) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep0) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep1) < 0) return -1;
@@ -2094,6 +2101,7 @@ int32_t tDeserializeSAlterDbReq(void *buf, int32_t bufLen, SAlterDbReq *pReq) {
   if (tDecodeI32(&decoder, &pReq->buffer) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pageSize) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pages) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->lastRowMem) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysPerFile) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep0) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep1) < 0) return -1;
@@ -2679,11 +2687,13 @@ int32_t tDeserializeSDbCfgRsp(void *buf, int32_t bufLen, SDbCfgRsp *pRsp) {
   if (tDecodeI8(&decoder, &pRsp->strict) < 0) return -1;
   if (tDecodeI8(&decoder, &pRsp->cacheLastRow) < 0) return -1;
   if (tDecodeI32(&decoder, &pRsp->numOfRetensions) < 0) return -1;
+  if (pRsp->numOfRetensions > 0) {
     pRsp->pRetensions = taosArrayInit(pRsp->numOfRetensions, sizeof(SRetention));
     if (pRsp->pRetensions == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
       return -1;
     }
+  }
 
   for (int32_t i = 0; i < pRsp->numOfRetensions; ++i) {
     SRetention rentension = {0};
@@ -3586,6 +3596,7 @@ int32_t tSerializeSCreateVnodeReq(void *buf, int32_t bufLen, SCreateVnodeReq *pR
   if (tEncodeI32(&encoder, pReq->buffer) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pageSize) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pages) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->lastRowMem) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysPerFile) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep0) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep1) < 0) return -1;
@@ -3643,6 +3654,7 @@ int32_t tDeserializeSCreateVnodeReq(void *buf, int32_t bufLen, SCreateVnodeReq *
   if (tDecodeI32(&decoder, &pReq->buffer) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pageSize) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pages) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->lastRowMem) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysPerFile) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep0) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep1) < 0) return -1;
@@ -3665,7 +3677,6 @@ int32_t tDeserializeSCreateVnodeReq(void *buf, int32_t bufLen, SCreateVnodeReq *
     SReplica *pReplica = &pReq->replicas[i];
     if (tDecodeSReplica(&decoder, pReplica) < 0) return -1;
   }
-
   if (tDecodeI32(&decoder, &pReq->numOfRetensions) < 0) return -1;
   pReq->pRetensions = taosArrayInit(pReq->numOfRetensions, sizeof(SRetention));
   if (pReq->pRetensions == NULL) {
@@ -3768,6 +3779,7 @@ int32_t tSerializeSAlterVnodeReq(void *buf, int32_t bufLen, SAlterVnodeReq *pReq
   if (tEncodeI32(&encoder, pReq->buffer) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pageSize) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->pages) < 0) return -1;
+  if (tEncodeI32(&encoder, pReq->lastRowMem) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysPerFile) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep0) < 0) return -1;
   if (tEncodeI32(&encoder, pReq->daysToKeep1) < 0) return -1;
@@ -3782,7 +3794,6 @@ int32_t tSerializeSAlterVnodeReq(void *buf, int32_t bufLen, SAlterVnodeReq *pReq
     SReplica *pReplica = &pReq->replicas[i];
     if (tEncodeSReplica(&encoder, pReplica) < 0) return -1;
   }
-
   tEndEncode(&encoder);
 
   int32_t tlen = encoder.pos;
@@ -3799,6 +3810,7 @@ int32_t tDeserializeSAlterVnodeReq(void *buf, int32_t bufLen, SAlterVnodeReq *pR
   if (tDecodeI32(&decoder, &pReq->buffer) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pageSize) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->pages) < 0) return -1;
+  if (tDecodeI32(&decoder, &pReq->lastRowMem) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysPerFile) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep0) < 0) return -1;
   if (tDecodeI32(&decoder, &pReq->daysToKeep1) < 0) return -1;

source/dnode/mgmt/mgmt_dnode/src/dmHandle.c

@@ -67,6 +67,8 @@ void dmSendStatusReq(SDnodeMgmt *pMgmt) {
   req.updateTime = pMgmt->pData->updateTime;
   req.numOfCores = tsNumOfCores;
   req.numOfSupportVnodes = tsNumOfSupportVnodes;
+  req.memTotal = tsTotalMemoryKB * 1024;
+  req.memAvail = req.memTotal - tsRpcQueueMemoryAllowed - 16 * 1024 * 1024;
   tstrncpy(req.dnodeEp, tsLocalEp, TSDB_EP_LEN);
 
   req.clusterCfg.statusInterval = tsStatusInterval;

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

@@ -353,7 +353,7 @@ static void *dmConsumParentQueue(void *param) {
       rpcRegisterBrokenLinkArg(pMsg);
     } else if (ftype == DND_FUNC_RELEASE) {
       dmRemoveProcRpcHandle(proc, pMsg->info.handle);
-      rpcReleaseHandle(pMsg->info.handle, (int8_t)pMsg->code);
+      rpcReleaseHandle(&pMsg->info, TAOS_CONN_SERVER);
     } else {
       dError("node:%s, invalid ftype:%d from pqueue", proc->name, ftype);
       rpcFreeCont(pMsg->pCont);

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

@@ -245,7 +245,7 @@ static inline void dmReleaseHandle(SRpcHandleInfo *pHandle, int8_t type) {
     SRpcMsg msg = {.code = type, .info = *pHandle};
     dmPutToProcPQueue(&pWrapper->proc, &msg, DND_FUNC_RELEASE);
   } else {
-    rpcReleaseHandle(pHandle->handle, type);
+    rpcReleaseHandle(pHandle, type);
   }
 }
 

source/dnode/mgmt/node_util/src/dmEps.c

@@ -280,7 +280,7 @@ static void dmPrintEps(SDnodeData *pData) {
   dDebug("print dnode list, num:%d", numOfEps);
   for (int32_t i = 0; i < numOfEps; i++) {
     SDnodeEp *pEp = taosArrayGet(pData->dnodeEps, i);
-    dDebug("dnode:%d, fqdn:%s port:%u is_mnode:%d", pEp->id, pEp->ep.fqdn, pEp->ep.port, pEp->isMnode);
+    dDebug("dnode:%d, fqdn:%s port:%u isMnode:%d", pEp->id, pEp->ep.fqdn, pEp->ep.port, pEp->isMnode);
   }
 }
 

source/dnode/mgmt/test/sut/src/sut.cpp

@@ -30,11 +30,13 @@ void Testbase::InitLog(const char* path) {
   tsdbDebugFlag = 0;
   tsLogEmbedded = 1;
   tsAsyncLog = 0;
-  tsRpcQueueMemoryAllowed = 1024 * 1024 * 64;
   
   taosRemoveDir(path);
   taosMkDir(path);
   tstrncpy(tsLogDir, path, PATH_MAX);
+
+  taosGetSystemInfo();
+  tsRpcQueueMemoryAllowed = tsTotalMemoryKB * 0.1;
 if (taosInitLog("taosdlog", 1) != 0) {
     printf("failed to init log file\n");
   }

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

@@ -148,7 +148,10 @@ typedef struct {
   int32_t    accessTimes;
   int32_t    numOfVnodes;
   int32_t    numOfSupportVnodes;
-  int32_t    numOfCores;
+  float      numOfCores;
+  int64_t    memTotal;
+  int64_t    memAvail;
+  int64_t    memUsed;
   EDndReason offlineReason;
   uint16_t   port;
   char       fqdn[TSDB_FQDN_LEN];
@@ -243,6 +246,7 @@ typedef struct {
   int32_t buffer;
   int32_t pageSize;
   int32_t pages;
+  int32_t lastRowMem;
   int32_t daysPerFile;
   int32_t daysToKeep0;
   int32_t daysToKeep1;
@@ -255,8 +259,8 @@ typedef struct {
   int8_t  compression;
   int8_t  replications;
   int8_t  strict;
-  int8_t  cacheLastRow;
   int8_t  hashMethod;  // default is 1
+  int8_t  cacheLastRow;
   int32_t numOfRetensions;
   SArray* pRetensions;
   int8_t  schemaless;

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

@@ -30,6 +30,8 @@ SSdbRaw *mndVgroupActionEncode(SVgObj *pVgroup);
 SEpSet   mndGetVgroupEpset(SMnode *pMnode, const SVgObj *pVgroup);
 int32_t  mndGetVnodesNum(SMnode *pMnode, int32_t dnodeId);
 void     mndSortVnodeGid(SVgObj *pVgroup);
+int64_t  mndGetVnodesMemory(SMnode *pMnode, int32_t dnodeId);
+int64_t  mndGetVgroupMemory(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup);
 
 SArray *mndBuildDnodesArray(SMnode *, int32_t exceptDnodeId);
 int32_t mndAllocSmaVgroup(SMnode *, SDbObj *pDb, SVgObj *pVgroup);

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

@@ -93,6 +93,7 @@ static SSdbRaw *mndDbActionEncode(SDbObj *pDb) {
   SDB_SET_INT32(pRaw, dataPos, pDb->cfg.buffer, _OVER)
   SDB_SET_INT32(pRaw, dataPos, pDb->cfg.pageSize, _OVER)
   SDB_SET_INT32(pRaw, dataPos, pDb->cfg.pages, _OVER)
+  SDB_SET_INT32(pRaw, dataPos, pDb->cfg.lastRowMem, _OVER)
   SDB_SET_INT32(pRaw, dataPos, pDb->cfg.daysPerFile, _OVER)
   SDB_SET_INT32(pRaw, dataPos, pDb->cfg.daysToKeep0, _OVER)
   SDB_SET_INT32(pRaw, dataPos, pDb->cfg.daysToKeep1, _OVER)
@@ -165,6 +166,7 @@ static SSdbRow *mndDbActionDecode(SSdbRaw *pRaw) {
   SDB_GET_INT32(pRaw, dataPos, &pDb->cfg.buffer, _OVER)
   SDB_GET_INT32(pRaw, dataPos, &pDb->cfg.pageSize, _OVER)
   SDB_GET_INT32(pRaw, dataPos, &pDb->cfg.pages, _OVER)
+  SDB_GET_INT32(pRaw, dataPos, &pDb->cfg.lastRowMem, _OVER)
   SDB_GET_INT32(pRaw, dataPos, &pDb->cfg.daysPerFile, _OVER)
   SDB_GET_INT32(pRaw, dataPos, &pDb->cfg.daysToKeep0, _OVER)
   SDB_GET_INT32(pRaw, dataPos, &pDb->cfg.daysToKeep1, _OVER)
@@ -230,8 +232,9 @@ static int32_t mndDbActionUpdate(SSdb *pSdb, SDbObj *pOld, SDbObj *pNew) {
   pOld->cfgVersion = pNew->cfgVersion;
   pOld->vgVersion = pNew->vgVersion;
   pOld->cfg.buffer = pNew->cfg.buffer;
-  pOld->cfg.pages = pNew->cfg.pages;
   pOld->cfg.pageSize = pNew->cfg.pageSize;
+  pOld->cfg.pages = pNew->cfg.pages;
+  pOld->cfg.lastRowMem = pNew->cfg.lastRowMem;
   pOld->cfg.daysPerFile = pNew->cfg.daysPerFile;
   pOld->cfg.daysToKeep0 = pNew->cfg.daysToKeep0;
   pOld->cfg.daysToKeep1 = pNew->cfg.daysToKeep1;
@@ -288,6 +291,7 @@ static int32_t mndCheckDbCfg(SMnode *pMnode, SDbCfg *pCfg) {
   if (pCfg->buffer < TSDB_MIN_BUFFER_PER_VNODE || pCfg->buffer > TSDB_MAX_BUFFER_PER_VNODE) return -1;
   if (pCfg->pageSize < TSDB_MIN_PAGESIZE_PER_VNODE || pCfg->pageSize > TSDB_MAX_PAGESIZE_PER_VNODE) return -1;
   if (pCfg->pages < TSDB_MIN_PAGES_PER_VNODE || pCfg->pages > TSDB_MAX_PAGES_PER_VNODE) return -1;
+  if (pCfg->lastRowMem < TSDB_MIN_DB_LAST_ROW_MEM || pCfg->lastRowMem > TSDB_MAX_DB_LAST_ROW_MEM) return -1;
   if (pCfg->daysPerFile < TSDB_MIN_DAYS_PER_FILE || pCfg->daysPerFile > TSDB_MAX_DAYS_PER_FILE) return -1;
   if (pCfg->daysToKeep0 < TSDB_MIN_KEEP || pCfg->daysToKeep0 > TSDB_MAX_KEEP) return -1;
   if (pCfg->daysToKeep1 < TSDB_MIN_KEEP || pCfg->daysToKeep1 > TSDB_MAX_KEEP) return -1;
@@ -336,6 +340,7 @@ static void mndSetDefaultDbCfg(SDbCfg *pCfg) {
   if (pCfg->replications < 0) pCfg->replications = TSDB_DEFAULT_DB_REPLICA;
   if (pCfg->strict < 0) pCfg->strict = TSDB_DEFAULT_DB_STRICT;
   if (pCfg->cacheLastRow < 0) pCfg->cacheLastRow = TSDB_DEFAULT_CACHE_LAST_ROW;
+  if (pCfg->lastRowMem <= 0) pCfg->lastRowMem = TSDB_DEFAULT_LAST_ROW_MEM;
   if (pCfg->numOfRetensions < 0) pCfg->numOfRetensions = 0;
   if (pCfg->schemaless < 0) pCfg->schemaless = TSDB_DB_SCHEMALESS_OFF;
 }
@@ -434,6 +439,7 @@ static int32_t mndCreateDb(SMnode *pMnode, SRpcMsg *pReq, SCreateDbReq *pCreate,
       .buffer = pCreate->buffer,
       .pageSize = pCreate->pageSize,
       .pages = pCreate->pages,
+      .lastRowMem = pCreate->lastRowMem,
       .daysPerFile = pCreate->daysPerFile,
       .daysToKeep0 = pCreate->daysToKeep0,
       .daysToKeep1 = pCreate->daysToKeep1,
@@ -475,7 +481,7 @@ static int32_t mndCreateDb(SMnode *pMnode, SRpcMsg *pReq, SCreateDbReq *pCreate,
   int32_t code = -1;
   STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_DB, pReq);
   if (pTrans == NULL) goto _OVER;
-
+  // mndTransSetSerial(pTrans);
   mDebug("trans:%d, used to create db:%s", pTrans->id, pCreate->db);
 
   mndTransSetDbName(pTrans, dbObj.name, NULL);
@@ -622,6 +628,11 @@ static int32_t mndSetDbCfgFromAlterDbReq(SDbObj *pDb, SAlterDbReq *pAlter) {
     terrno = 0;
   }
 
+  if (pAlter->lastRowMem > 0 && pAlter->lastRowMem != pDb->cfg.lastRowMem) {
+    pDb->cfg.lastRowMem = pAlter->lastRowMem;
+    terrno = 0;
+  }
+
   if (pAlter->replications > 0 && pAlter->replications != pDb->cfg.replications) {
 #if 1
     terrno = TSDB_CODE_OPS_NOT_SUPPORT;

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

@@ -15,8 +15,8 @@
 
 #define _DEFAULT_SOURCE
 #include "mndDnode.h"
-#include "mndPrivilege.h"
 #include "mndMnode.h"
+#include "mndPrivilege.h"
 #include "mndQnode.h"
 #include "mndShow.h"
 #include "mndSnode.h"
@@ -432,7 +432,8 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
     }
 
     if (!online) {
-      mInfo("dnode:%d, from offline to online", pDnode->id);
+      mInfo("dnode:%d, from offline to online, memory avail:%" PRId64 " total:%" PRId64 " cores:%.2f", pDnode->id,
+            statusReq.memAvail, statusReq.memTotal, statusReq.numOfCores);
     } else {
       mDebug("dnode:%d, send dnode epset, online:%d dnodeVer:%" PRId64 ":%" PRId64 " reboot:%d", pDnode->id, online,
              statusReq.dnodeVer, dnodeVer, reboot);
@@ -441,6 +442,8 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
     pDnode->rebootTime = statusReq.rebootTime;
     pDnode->numOfCores = statusReq.numOfCores;
     pDnode->numOfSupportVnodes = statusReq.numOfSupportVnodes;
+    pDnode->memAvail = statusReq.memAvail;
+    pDnode->memTotal = statusReq.memTotal;
 
     SStatusRsp statusRsp = {0};
     statusRsp.dnodeVer = dnodeVer;

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

@@ -427,7 +427,7 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
   } while (0);
 
   // ToDo: ugly! use function pointer
-  if (syncNodeSnapshotEnable(pSyncNode)) {
+  if (syncNodeStrategy(pSyncNode) == SYNC_STRATEGY_STANDARD_SNAPSHOT) {
     if (pMsg->msgType == TDMT_SYNC_TIMEOUT) {
       SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pMsg);
       code = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);

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

@@ -134,7 +134,7 @@ int32_t mndSnapshotDoRead(struct SSyncFSM *pFsm, void *pReader, void **ppBuf, in
   return sdbDoRead(pMnode->pSdb, pReader, ppBuf, len);
 }
 
-int32_t mndSnapshotStartWrite(struct SSyncFSM *pFsm, void **ppWriter) {
+int32_t mndSnapshotStartWrite(struct SSyncFSM *pFsm, void *pParam, void **ppWriter) {
   mInfo("start to apply snapshot to sdb");
   SMnode *pMnode = pFsm->data;
   return sdbStartWrite(pMnode->pSdb, (SSdbIter **)ppWriter);
@@ -178,7 +178,7 @@ int32_t mndInitSync(SMnode *pMnode) {
   syncInfo.pWal = pMnode->pWal;
   syncInfo.pFsm = mndSyncMakeFsm(pMnode);
   syncInfo.isStandBy = pMgmt->standby;
-  syncInfo.snapshotEnable = true;
+  syncInfo.snapshotStrategy = SYNC_STRATEGY_STANDARD_SNAPSHOT;
 
   mInfo("start to open mnode sync, standby:%d", pMgmt->standby);
   if (pMgmt->standby || pMgmt->replica.id > 0) {

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

@@ -207,6 +207,7 @@ void *mndBuildCreateVnodeReq(SMnode *pMnode, SDnodeObj *pDnode, SDbObj *pDb, SVg
   createReq.buffer = pDb->cfg.buffer;
   createReq.pageSize = pDb->cfg.pageSize;
   createReq.pages = pDb->cfg.pages;
+  createReq.lastRowMem = pDb->cfg.lastRowMem;
   createReq.daysPerFile = pDb->cfg.daysPerFile;
   createReq.daysToKeep0 = pDb->cfg.daysToKeep0;
   createReq.daysToKeep1 = pDb->cfg.daysToKeep1;
@@ -274,8 +275,9 @@ void *mndBuildAlterVnodeReq(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup, int32_
   SAlterVnodeReq alterReq = {0};
   alterReq.vgVersion = pVgroup->version;
   alterReq.buffer = pDb->cfg.buffer;
-  alterReq.pages = pDb->cfg.pages;
   alterReq.pageSize = pDb->cfg.pageSize;
+  alterReq.pages = pDb->cfg.pages;
+  alterReq.lastRowMem = pDb->cfg.lastRowMem;
   alterReq.daysPerFile = pDb->cfg.daysPerFile;
   alterReq.daysToKeep0 = pDb->cfg.daysToKeep0;
   alterReq.daysToKeep1 = pDb->cfg.daysToKeep1;
@@ -392,9 +394,10 @@ static bool mndBuildDnodesArrayFp(SMnode *pMnode, void *pObj, void *p1, void *p2
   bool    online = mndIsDnodeOnline(pDnode, curMs);
   bool    isMnode = mndIsMnode(pMnode, pDnode->id);
   pDnode->numOfVnodes = mndGetVnodesNum(pMnode, pDnode->id);
+  pDnode->memUsed = mndGetVnodesMemory(pMnode, pDnode->id);
 
-  mDebug("dnode:%d, vnodes:%d support_vnodes:%d is_mnode:%d online:%d", pDnode->id, pDnode->numOfVnodes,
+  mDebug("dnode:%d, vnodes:%d supportVnodes:%d isMnode:%d online:%d memory avail:%" PRId64 " used:%" PRId64, pDnode->id,
-         pDnode->numOfSupportVnodes, isMnode, online);
+         pDnode->numOfVnodes, pDnode->numOfSupportVnodes, isMnode, online, pDnode->memAvail, pDnode->memUsed);
 
   if (isMnode) {
     pDnode->numOfVnodes++;
@@ -426,16 +429,8 @@ static int32_t mndCompareDnodeId(int32_t *dnode1Id, int32_t *dnode2Id) { return
 static int32_t mndCompareDnodeVnodes(SDnodeObj *pDnode1, SDnodeObj *pDnode2) {
   float d1Score = (float)pDnode1->numOfVnodes / pDnode1->numOfSupportVnodes;
   float d2Score = (float)pDnode2->numOfVnodes / pDnode2->numOfSupportVnodes;
-#if 0  
-  if (d1Score == d2Score) {
-    return pDnode2->id - pDnode1->id;
-  } else {
   return d1Score >= d2Score ? 1 : 0;
 }
-#else
-  return d1Score >= d2Score ? 1 : 0;
-#endif
-}
 
 void mndSortVnodeGid(SVgObj *pVgroup) {
   for (int32_t i = 0; i < pVgroup->replica; ++i) {
@@ -447,7 +442,7 @@ void mndSortVnodeGid(SVgObj *pVgroup) {
   }
 }
 
-static int32_t mndGetAvailableDnode(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
+static int32_t mndGetAvailableDnode(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray) {
   SSdb   *pSdb = pMnode->pSdb;
   int32_t allocedVnodes = 0;
   void   *pIter = NULL;
@@ -470,6 +465,16 @@ static int32_t mndGetAvailableDnode(SMnode *pMnode, SVgObj *pVgroup, SArray *pAr
       return -1;
     }
 
+    int64_t vgMem = mndGetVgroupMemory(pMnode, pDb, pVgroup);
+    if (pDnode->memAvail - vgMem - pDnode->memUsed <= 0) {
+      mError("db:%s, vgId:%d, no enough memory:%" PRId64 " in dnode:%d, avail:%" PRId64 " used:%" PRId64,
+             pVgroup->dbName, pVgroup->vgId, vgMem, pDnode->id, pDnode->memAvail, pDnode->memUsed);
+      terrno = TSDB_CODE_MND_NO_ENOUGH_MEM_IN_DNODE;
+      return -1;
+    } else {
+      pDnode->memUsed += vgMem;
+    }
+
     pVgid->dnodeId = pDnode->id;
     if (pVgroup->replica == 1) {
       pVgid->role = TAOS_SYNC_STATE_LEADER;
@@ -477,7 +482,8 @@ static int32_t mndGetAvailableDnode(SMnode *pMnode, SVgObj *pVgroup, SArray *pAr
       pVgid->role = TAOS_SYNC_STATE_FOLLOWER;
     }
 
-    mInfo("db:%s, vgId:%d, vn:%d dnode:%d is alloced", pVgroup->dbName, pVgroup->vgId, v, pVgid->dnodeId);
+    mInfo("db:%s, vgId:%d, vn:%d is alloced, memory:%" PRId64 ", dnode:%d avail:%" PRId64 " used:%" PRId64,
+          pVgroup->dbName, pVgroup->vgId, v, vgMem, pVgid->dnodeId, pDnode->memAvail, pDnode->memUsed);
     pDnode->numOfVnodes++;
   }
 
@@ -498,7 +504,7 @@ int32_t mndAllocSmaVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup) {
   pVgroup->dbUid = pDb->uid;
   pVgroup->replica = 1;
 
-  if (mndGetAvailableDnode(pMnode, pVgroup, pArray) != 0) return -1;
+  if (mndGetAvailableDnode(pMnode, pDb, pVgroup, pArray) != 0) return -1;
 
   mInfo("db:%s, sma vgId:%d is alloced", pDb->name, pVgroup->vgId);
   return 0;
@@ -546,8 +552,7 @@ int32_t mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups) {
     pVgroup->dbUid = pDb->uid;
     pVgroup->replica = pDb->cfg.replications;
 
-    if (mndGetAvailableDnode(pMnode, pVgroup, pArray) != 0) {
+    if (mndGetAvailableDnode(pMnode, pDb, pVgroup, pArray) != 0) {
-      terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
       goto _OVER;
     }
 
@@ -728,6 +733,43 @@ int32_t mndGetVnodesNum(SMnode *pMnode, int32_t dnodeId) {
   return numOfVnodes;
 }
 
+int64_t mndGetVgroupMemory(SMnode *pMnode, SDbObj *pDbInput, SVgObj *pVgroup) {
+  SDbObj *pDb = pDbInput;
+  if (pDbInput == NULL) {
+    pDb = mndAcquireDb(pMnode, pVgroup->dbName);
+  }
+
+  int64_t vgroupMemroy = (int64_t)pDb->cfg.buffer * 1024 * 1024 + (int64_t)pDb->cfg.pages * pDb->cfg.pageSize * 1024;
+  if (pDb->cfg.cacheLastRow > 0) {
+    vgroupMemroy += (int64_t)pDb->cfg.lastRowMem * 1024 * 1024;
+  }
+
+  if (pDbInput == NULL) {
+    mndReleaseDb(pMnode, pDb);
+  }
+  return vgroupMemroy;
+}
+
+static bool mndGetVnodeMemroyFp(SMnode *pMnode, void *pObj, void *p1, void *p2, void *p3) {
+  SVgObj  *pVgroup = pObj;
+  int32_t  dnodeId = *(int32_t *)p1;
+  int64_t *pVnodeMemory = (int64_t *)p2;
+
+  for (int32_t v = 0; v < pVgroup->replica; ++v) {
+    if (pVgroup->vnodeGid[v].dnodeId == dnodeId) {
+      *pVnodeMemory += mndGetVgroupMemory(pMnode, NULL, pVgroup);
+    }
+  }
+
+  return true;
+}
+
+int64_t mndGetVnodesMemory(SMnode *pMnode, int32_t dnodeId) {
+  int64_t vnodeMemory = 0;
+  sdbTraverse(pMnode->pSdb, SDB_VGROUP, mndGetVnodeMemroyFp, &dnodeId, &vnodeMemory, NULL);
+  return vnodeMemory;
+}
+
 static int32_t mndRetrieveVnodes(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows) {
   SMnode *pMnode = pReq->info.node;
   SSdb   *pSdb = pMnode->pSdb;
@@ -807,9 +849,20 @@ int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
       return -1;
     }
 
+    int64_t vgMem = mndGetVgroupMemory(pMnode, NULL, pVgroup);
+    if (pDnode->memAvail - vgMem - pDnode->memUsed <= 0) {
+      mError("db:%s, vgId:%d, no enough memory:%" PRId64 " in dnode:%d avail:%" PRId64 " used:%" PRId64,
+             pVgroup->dbName, pVgroup->vgId, vgMem, pDnode->id, pDnode->memAvail, pDnode->memUsed);
+      terrno = TSDB_CODE_MND_NO_ENOUGH_MEM_IN_DNODE;
+      return -1;
+    } else {
+      pDnode->memUsed += vgMem;
+    }
+
     pVgid->dnodeId = pDnode->id;
     pVgid->role = TAOS_SYNC_STATE_ERROR;
-    mInfo("db:%s, vgId:%d, vn:%d dnode:%d, is added", pVgroup->dbName, pVgroup->vgId, pVgroup->replica, pVgid->dnodeId);
+    mInfo("db:%s, vgId:%d, vn:%d is added, memory:%" PRId64 ", dnode:%d avail:%" PRId64 " used:%" PRId64,
+          pVgroup->dbName, pVgroup->vgId, pVgroup->replica, vgMem, pVgid->dnodeId, pDnode->memAvail, pDnode->memUsed);
 
     pVgroup->replica++;
     pDnode->numOfVnodes++;
@@ -835,7 +888,10 @@ int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray
     for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
       SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
       if (pVgid->dnodeId == pDnode->id) {
-        mInfo("db:%s, vgId:%d, vn:%d dnode:%d, is removed", pVgroup->dbName, pVgroup->vgId, vn, pVgid->dnodeId);
+        int64_t vgMem = mndGetVgroupMemory(pMnode, NULL, pVgroup);
+        pDnode->memUsed -= vgMem;
+        mInfo("db:%s, vgId:%d, vn:%d is removed, memory:%" PRId64 ", dnode:%d avail:%" PRId64 " used:%" PRId64,
+              pVgroup->dbName, pVgroup->vgId, vn, vgMem, pVgid->dnodeId, pDnode->memAvail, pDnode->memUsed);
         pDnode->numOfVnodes--;
         pVgroup->replica--;
         *pDelVgid = *pVgid;
@@ -1161,6 +1217,17 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
       terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
       goto _OVER;
     }
+
+    int64_t vgMem = mndGetVgroupMemory(pMnode, NULL, pVgroup);
+    if (pNew1->memAvail - vgMem - pNew1->memUsed <= 0) {
+      mError("db:%s, vgId:%d, no enough memory:%" PRId64 " in dnode:%d avail:%" PRId64 " used:%" PRId64,
+             pVgroup->dbName, pVgroup->vgId, vgMem, pNew1->id, pNew1->memAvail, pNew1->memUsed);
+      terrno = TSDB_CODE_MND_NO_ENOUGH_MEM_IN_DNODE;
+      return -1;
+    } else {
+      pNew1->memUsed += vgMem;
+    }
+
     if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew1->id) != 0) goto _OVER;
     if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld1->id) != 0) goto _OVER;
   }
@@ -1173,6 +1240,15 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
       terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
       goto _OVER;
     }
+    int64_t vgMem = mndGetVgroupMemory(pMnode, NULL, pVgroup);
+    if (pNew2->memAvail - vgMem - pNew2->memUsed <= 0) {
+      mError("db:%s, vgId:%d, no enough memory:%" PRId64 " in dnode:%d avail:%" PRId64 " used:%" PRId64,
+             pVgroup->dbName, pVgroup->vgId, vgMem, pNew2->id, pNew2->memAvail, pNew2->memUsed);
+      terrno = TSDB_CODE_MND_NO_ENOUGH_MEM_IN_DNODE;
+      return -1;
+    } else {
+      pNew2->memUsed += vgMem;
+    }
     if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew2->id) != 0) goto _OVER;
     if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld2->id) != 0) goto _OVER;
   }
@@ -1185,6 +1261,15 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
       terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
       goto _OVER;
     }
+    int64_t vgMem = mndGetVgroupMemory(pMnode, NULL, pVgroup);
+    if (pNew3->memAvail - vgMem - pNew3->memUsed <= 0) {
+      mError("db:%s, vgId:%d, no enough memory:%" PRId64 " in dnode:%d avail:%" PRId64 " used:%" PRId64,
+             pVgroup->dbName, pVgroup->vgId, vgMem, pNew3->id, pNew3->memAvail, pNew3->memUsed);
+      terrno = TSDB_CODE_MND_NO_ENOUGH_MEM_IN_DNODE;
+      return -1;
+    } else {
+      pNew3->memUsed += vgMem;
+    }
     if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew3->id) != 0) goto _OVER;
     if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld3->id) != 0) goto _OVER;
   }

source/dnode/vnode/inc/vnode.h

@@ -128,7 +128,7 @@ bool    tsdbNextDataBlock(STsdbReader *pReader);
 void    tsdbRetrieveDataBlockInfo(STsdbReader *pReader, SDataBlockInfo *pDataBlockInfo);
 int32_t tsdbRetrieveDataBlockStatisInfo(STsdbReader *pReader, SColumnDataAgg ***pBlockStatis, bool *allHave);
 SArray *tsdbRetrieveDataBlock(STsdbReader *pTsdbReadHandle, SArray *pColumnIdList);
-void    tsdbResetReadHandle(STsdbReader *pReader, SQueryTableDataCond *pCond, int32_t tWinIdx);
+int32_t tsdbReaderReset(STsdbReader *pReader, SQueryTableDataCond *pCond, int32_t tWinIdx);
 int32_t tsdbGetFileBlocksDistInfo(STsdbReader *pReader, STableBlockDistInfo *pTableBlockInfo);
 int64_t tsdbGetNumOfRowsInMemTable(STsdbReader *pHandle);
 

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

@@ -83,8 +83,8 @@ int32_t smaBegin(SSma *pSma) {
 /**
  * @brief pre-commit for rollup sma.
  *  1) set trigger stat of rsma timer TASK_TRIGGER_STAT_PAUSED.
- *  2) perform persist task for qTaskInfo
+ *  2) wait all triggered fetch tasks finished
- *  3) wait all triggered fetch tasks finished
+ *  3) perform persist task for qTaskInfo
  *
  * @param pSma
  * @return int32_t
@@ -102,10 +102,7 @@ static int32_t tdProcessRSmaPreCommitImpl(SSma *pSma) {
   // step 1: set persistence task paused
   atomic_store_8(RSMA_TRIGGER_STAT(pRSmaStat), TASK_TRIGGER_STAT_PAUSED);
 
-  // step 2: perform persist task for qTaskInfo
+  // step 2: wait all triggered fetch tasks finished
-  tdRSmaPersistExecImpl(pRSmaStat);
-
-  // step 3: wait all triggered fetch tasks finished
   int32_t nLoops = 0;
   while (1) {
     if (T_REF_VAL_GET(pStat) == 0) {
@@ -121,6 +118,9 @@ static int32_t tdProcessRSmaPreCommitImpl(SSma *pSma) {
     }
   }
 
+  // step 3: perform persist task for qTaskInfo
+  tdRSmaPersistExecImpl(pRSmaStat);
+
   smaDebug("vgId:%d, rsma pre commit succeess", SMA_VID(pSma));
 
   return TSDB_CODE_SUCCESS;

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

@@ -915,9 +915,9 @@ static int32_t tdRSmaQTaskInfoItemRestore(SSma *pSma, const SRSmaQTaskInfoItem *
     return TSDB_CODE_SUCCESS;
   }
 
-  if (pItem->type == 1) {
+  if (pItem->type == TSDB_RETENTION_L1) {
     qTaskInfo = pRSmaInfo->items[0].taskInfo;
-  } else if (pItem->type == 2) {
+  } else if (pItem->type == TSDB_RETENTION_L2) {
     qTaskInfo = pRSmaInfo->items[1].taskInfo;
   } else {
     ASSERT(0);
@@ -1233,7 +1233,6 @@ static void tdRSmaPersistTask(SRSmaStat *pRSmaStat) {
     } else {
       smaWarn("vgId:%d, persist task in abnormal stat %" PRIi8, SMA_VID(pRSmaStat->pSma),
               atomic_load_8(RSMA_TRIGGER_STAT(pRSmaStat)));
-      ASSERT(0);
     }
     atomic_store_8(RSMA_RUNNING_STAT(pRSmaStat), 0);
     taosReleaseRef(smaMgmt.smaRef, pRSmaStat->refId);

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

@@ -16,17 +16,22 @@
 #include "tsdb.h"
 #define ASCENDING_TRAVERSE(o) (o == TSDB_ORDER_ASC)
 
+typedef struct {
+  STbDataIter *iter;
+  int32_t      index;
+  bool         hasVal;
+} SIterInfo;
+
 typedef struct STableBlockScanInfo {
   uint64_t     uid;
   TSKEY        lastKey;
   SBlockIdx    blockIdx;
   SArray*      pBlockList;        // block data index list
-  bool         iterInit;          // whether to initialize the in-memory skip list iterator or not
+  SIterInfo    iter;              // mem buffer skip list iterator
-  STbDataIter* iter;              // mem buffer skip list iterator
+  SIterInfo    iiter;             // imem buffer skip list iterator
-  STbDataIter* iiter;             // imem buffer skip list iterator
   SArray*      delSkyline;        // delete info for this table
-  bool         memHasVal;
+  int32_t      fileDelIndex;
-  bool         imemHasVal;
+  bool         iterInit;          // whether to initialize the in-memory skip list iterator or not
 } STableBlockScanInfo;
 
 typedef struct SBlockOrderWrapper {
@@ -112,14 +117,14 @@ struct STsdbReader {
   char*              idStr;  // query info handle, for debug purpose
   int32_t            type;   // query type: 1. retrieve all data blocks, 2. retrieve direct prev|next rows
   SBlockLoadSuppInfo suppInfo;
-  SArray*            prev;  // previous row which is before than time window
+
-  SArray*            next;  // next row which is after the query time window
   SIOCostSummary     cost;
   STSchema*          pSchema;
-
   SDataFReader*      pFileReader;
   SVersionRange      verRange;
 #if 0
+  SArray*            prev;  // previous row which is before than time window
+  SArray*            next;  // next row which is after the query time window
   SFileBlockInfo*   pDataBlockInfo;
   SDataCols*         pDataCols;         // in order to hold current file data block
   int32_t            allocSize;         // allocated data block size
@@ -136,19 +141,19 @@ struct STsdbReader {
 static SFileDataBlockInfo* getCurrentBlockInfo(SDataBlockIter* pBlockIter);
 static int      buildDataBlockFromBufImpl(STableBlockScanInfo* pBlockScanInfo, int64_t endKey, int32_t capacity,
                                           STsdbReader* pReader);
-static TSDBROW* getValidRow(STbDataIter* pIter, bool* hasVal, STsdbReader* pReader);
+static TSDBROW* getValidRow(SIterInfo* pIter, const SArray* pDelList, STsdbReader* pReader);
 static int32_t  doMergeRowsInFileBlocks(SBlockData* pBlockData, STableBlockScanInfo* pScanInfo, STsdbReader* pReader,
                                         SRowMerger* pMerger);
-static int32_t  doMergeRowsInBuf(STbDataIter* pIter, bool* hasVal, int64_t ts, SRowMerger* pMerger,
+static int32_t  doMergeRowsInBuf(SIterInfo *pIter, int64_t ts, SArray* pDelList, SRowMerger* pMerger, STsdbReader* pReader);
-                                 STsdbReader* pReader);
 static int32_t  doAppendOneRow(SSDataBlock* pBlock, STsdbReader* pReader, STSRow* pTSRow);
 static void     setComposedBlockFlag(STsdbReader* pReader, bool composed);
 static void     updateSchema(TSDBROW* pRow, uint64_t uid, STsdbReader* pReader);
+static bool     hasBeenDropped(const SArray* pDelList, int32_t* index, TSDBKEY* pKey);
 
-static void doMergeMultiRows(TSDBROW* pRow, uint64_t uid, STbDataIter* dIter, bool* hasVal, STSRow** pTSRow,
+static void doMergeMultiRows(TSDBROW* pRow, uint64_t uid, SIterInfo *pIter, SArray* pDelList, STSRow** pTSRow, STsdbReader* pReader);
-                             STsdbReader* pReader);
 static void doMergeMemIMemRows(TSDBROW* pRow, TSDBROW* piRow, STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader,
                                STSRow** pTSRow);
+static int32_t initDelSkylineIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader, STbData* pMemTbData, STbData* piMemTbData);
 
 static int32_t setColumnIdSlotList(STsdbReader* pReader, SSDataBlock* pBlock) {
   SBlockLoadSuppInfo* pSupInfo = &pReader->suppInfo;
@@ -372,7 +377,7 @@ static int32_t tsdbReaderCreate(SVnode* pVnode, SQueryTableDataCond* pCond, STsd
   pReader->suid        = pCond->suid;
   pReader->order       = pCond->order;
   pReader->capacity    = 4096;
-  pReader->idStr       = strdup(idstr);
+  pReader->idStr       = (idstr != NULL)? strdup(idstr):NULL;
   pReader->verRange    = (SVersionRange) {.minVer = pCond->startVersion, .maxVer = 10000};
   pReader->type        = pCond->type;
   pReader->window      = updateQueryTimeWindow(pVnode->pTsdb, pCond->twindows);
@@ -1708,7 +1713,7 @@ static bool fileBlockShouldLoad(STsdbReader* pReader, SFileDataBlockInfo* pFBloc
 }
 
 static int32_t buildDataBlockFromBuf(STsdbReader* pReader, STableBlockScanInfo* pBlockScanInfo, int64_t endKey) {
-  if (!(pBlockScanInfo->imemHasVal || pBlockScanInfo->memHasVal)) {
+  if (!(pBlockScanInfo->iiter.hasVal || pBlockScanInfo->iter.hasVal)) {
     return TSDB_CODE_SUCCESS;
   }
 
@@ -1728,13 +1733,14 @@ static int32_t buildDataBlockFromBuf(STsdbReader* pReader, STableBlockScanInfo*
 }
 
 static int32_t doMergeBufAndFileRows(STsdbReader* pReader, STableBlockScanInfo* pBlockScanInfo, TSDBROW* pRow,
-                                     STSRow* pTSRow, STbDataIter* pIter, bool* hasVal, int64_t key) {
+                                     STSRow* pTSRow, SIterInfo* pIter, int64_t key) {
   SRowMerger          merge = {0};
   SBlockData*         pBlockData = &pReader->status.fileBlockData;
   SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo;
 
   TSDBKEY k = TSDBROW_KEY(pRow);
   TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex);
+  SArray* pDelList = pBlockScanInfo->delSkyline;
 
   // ascending order traverse
   if (ASCENDING_TRAVERSE(pReader->order)) {
@@ -1744,19 +1750,19 @@ static int32_t doMergeBufAndFileRows(STsdbReader* pReader, STableBlockScanInfo*
       doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader, &merge);
       tRowMergerGetRow(&merge, &pTSRow);
     } else if (k.ts < key) {  // k.ts < key
-      doMergeMultiRows(pRow, pBlockScanInfo->uid, pIter, hasVal, &pTSRow, pReader);
+      doMergeMultiRows(pRow, pBlockScanInfo->uid, pIter, pDelList, &pTSRow, pReader);
     } else {  // k.ts == key, ascending order: file block ----> imem rows -----> mem rows
       tRowMergerInit(&merge, &fRow, pReader->pSchema);
       doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader, &merge);
 
       tRowMerge(&merge, pRow);
-      doMergeRowsInBuf(pIter, hasVal, k.ts, &merge, pReader);
+      doMergeRowsInBuf(pIter, k.ts, pBlockScanInfo->delSkyline, &merge, pReader);
 
       tRowMergerGetRow(&merge, &pTSRow);
     }
   } else {  // descending order scan
     if (key < k.ts) {
-      doMergeMultiRows(pRow, pBlockScanInfo->uid, pIter, hasVal, &pTSRow, pReader);
+      doMergeMultiRows(pRow, pBlockScanInfo->uid, pIter, pDelList, &pTSRow, pReader);
     } else if (k.ts < key) {
       tRowMergerInit(&merge, &fRow, pReader->pSchema);
 
@@ -1766,7 +1772,7 @@ static int32_t doMergeBufAndFileRows(STsdbReader* pReader, STableBlockScanInfo*
       updateSchema(pRow, pBlockScanInfo->uid, pReader);
 
       tRowMergerInit(&merge, pRow, pReader->pSchema);
-      doMergeRowsInBuf(pIter, hasVal, k.ts, &merge, pReader);
+      doMergeRowsInBuf(pIter, k.ts, pBlockScanInfo->delSkyline, &merge, pReader);
 
       tRowMerge(&merge, &fRow);
       doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader, &merge);
@@ -1786,9 +1792,10 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
 
   SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo;
   SBlockData*         pBlockData = &pReader->status.fileBlockData;
+  SArray* pDelList = pBlockScanInfo->delSkyline;
 
-  TSDBROW* pRow = getValidRow(pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, pReader);
+  TSDBROW* pRow = getValidRow(&pBlockScanInfo->iter, pDelList, pReader);
-  TSDBROW* piRow = getValidRow(pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, pReader);
+  TSDBROW* piRow = getValidRow(&pBlockScanInfo->iiter, pDelList, pReader);
   ASSERT(pRow != NULL && piRow != NULL);
 
   int64_t key = pBlockData->aTSKEY[pDumpInfo->rowIndex];
@@ -1808,12 +1815,12 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
 
       if (ik.ts == key) {
         tRowMerge(&merge, piRow);
-        doMergeRowsInBuf(pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, key, &merge, pReader);
+        doMergeRowsInBuf(&pBlockScanInfo->iiter, key, pBlockScanInfo->delSkyline, &merge, pReader);
       }
 
       if (k.ts == key) {
         tRowMerge(&merge, pRow);
-        doMergeRowsInBuf(pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, key, &merge, pReader);
+        doMergeRowsInBuf(&pBlockScanInfo->iter, key, pBlockScanInfo->delSkyline, &merge, pReader);
       }
 
       tRowMergerGetRow(&merge, &pTSRow);
@@ -1825,7 +1832,7 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
       // [3] ik.ts < key <= k.ts
       // [4] ik.ts < k.ts <= key
       if (ik.ts < k.ts) {
-        doMergeMultiRows(piRow, uid, pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, &pTSRow, pReader);
+        doMergeMultiRows(piRow, uid, &pBlockScanInfo->iiter, pDelList, &pTSRow, pReader);
         doAppendOneRow(pReader->pResBlock, pReader, pTSRow);
         return TSDB_CODE_SUCCESS;
       }
@@ -1833,7 +1840,7 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
       // [5] k.ts < key   <= ik.ts
       // [6] k.ts < ik.ts <= key
       if (k.ts < ik.ts) {
-        doMergeMultiRows(pRow, uid, pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, &pTSRow, pReader);
+        doMergeMultiRows(pRow, uid, &pBlockScanInfo->iter, pDelList, &pTSRow, pReader);
         doAppendOneRow(pReader->pResBlock, pReader, pTSRow);
         return TSDB_CODE_SUCCESS;
       }
@@ -1853,11 +1860,11 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
       updateSchema(pRow, uid, pReader);
 
       tRowMergerInit(&merge, pRow, pReader->pSchema);
-      doMergeRowsInBuf(pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, key, &merge, pReader);
+      doMergeRowsInBuf(&pBlockScanInfo->iter, key, pBlockScanInfo->delSkyline, &merge, pReader);
 
       if (ik.ts == k.ts) {
         tRowMerge(&merge, piRow);
-        doMergeRowsInBuf(pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, key, &merge, pReader);
+        doMergeRowsInBuf(&pBlockScanInfo->iiter, key, pBlockScanInfo->delSkyline, &merge, pReader);
       }
 
       if (k.ts == key) {
@@ -1875,7 +1882,7 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
       // [3] ik.ts > k.ts >= Key
       // [4] ik.ts > key >= k.ts
       if (ik.ts > key) {
-        doMergeMultiRows(piRow, uid, pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, &pTSRow, pReader);
+        doMergeMultiRows(piRow, uid, &pBlockScanInfo->iiter, pDelList, &pTSRow, pReader);
         doAppendOneRow(pReader->pResBlock, pReader, pTSRow);
         return TSDB_CODE_SUCCESS;
       }
@@ -1894,7 +1901,7 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
 
       //[7] key = ik.ts > k.ts
       if (key == ik.ts) {
-        doMergeMultiRows(piRow, uid, pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, &pTSRow, pReader);
+        doMergeMultiRows(piRow, uid, &pBlockScanInfo->iiter, pDelList, &pTSRow, pReader);
 
         TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex);
         tRowMerge(&merge, &fRow);
@@ -1909,7 +1916,8 @@ static int32_t doMergeThreeLevelRows(STsdbReader* pReader, STableBlockScanInfo*
   ASSERT(0);
 }
 
-static bool isValidFileBlockRow(SBlockData* pBlockData, SFileBlockDumpInfo* pDumpInfo, STsdbReader* pReader) {
+static bool isValidFileBlockRow(SBlockData* pBlockData, SFileBlockDumpInfo* pDumpInfo, STableBlockScanInfo* pBlockScanInfo,
+    STsdbReader* pReader) {
   // check for version and time range
   int64_t ver = pBlockData->aVersion[pDumpInfo->rowIndex];
   if (ver > pReader->verRange.maxVer || ver < pReader->verRange.minVer) {
@@ -1921,6 +1929,11 @@ static bool isValidFileBlockRow(SBlockData* pBlockData, SFileBlockDumpInfo* pDum
     return false;
   }
 
+  TSDBKEY k = {.ts = ts, .version = ver};
+  if (hasBeenDropped(pBlockScanInfo->delSkyline, &pBlockScanInfo->fileDelIndex, &k)) {
+    return false;
+  }
+
   return true;
 }
 
@@ -1934,22 +1947,20 @@ static int32_t buildComposedDataBlockImpl(STsdbReader* pReader, STableBlockScanI
   STSRow*    pTSRow = NULL;
 
   int64_t  key = pBlockData->aTSKEY[pDumpInfo->rowIndex];
-  TSDBROW* pRow = getValidRow(pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, pReader);
+  TSDBROW* pRow = getValidRow(&pBlockScanInfo->iter, pBlockScanInfo->delSkyline, pReader);
-  TSDBROW* piRow = getValidRow(pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, pReader);
+  TSDBROW* piRow = getValidRow(&pBlockScanInfo->iiter, pBlockScanInfo->delSkyline, pReader);
 
-  if (pBlockScanInfo->memHasVal && pBlockScanInfo->imemHasVal) {
+  if (pBlockScanInfo->iter.hasVal && pBlockScanInfo->iiter.hasVal) {
     return doMergeThreeLevelRows(pReader, pBlockScanInfo);
   } else {
     // imem + file
-    if (pBlockScanInfo->imemHasVal) {
+    if (pBlockScanInfo->iiter.hasVal) {
-      return doMergeBufAndFileRows(pReader, pBlockScanInfo, piRow, pTSRow, pBlockScanInfo->iiter,
+      return doMergeBufAndFileRows(pReader, pBlockScanInfo, piRow, pTSRow, &pBlockScanInfo->iiter, key);
-                                   &pBlockScanInfo->imemHasVal, key);
     }
 
     // mem + file
-    if (pBlockScanInfo->memHasVal) {
+    if (pBlockScanInfo->iter.hasVal) {
-      return doMergeBufAndFileRows(pReader, pBlockScanInfo, pRow, pTSRow, pBlockScanInfo->iter,
+      return doMergeBufAndFileRows(pReader, pBlockScanInfo, pRow, pTSRow, &pBlockScanInfo->iter,key);
-                                   &pBlockScanInfo->memHasVal, key);
     }
 
     // imem & mem are all empty, only file exist
@@ -1973,7 +1984,7 @@ static int32_t buildComposedDataBlock(STsdbReader* pReader, STableBlockScanInfo*
   while (1) {
     // todo check the validate of row in file block
     {
-      if (!isValidFileBlockRow(pBlockData, pDumpInfo, pReader)) {
+      if (!isValidFileBlockRow(pBlockData, pDumpInfo, pBlockScanInfo, pReader)) {
         pDumpInfo->rowIndex += step;
 
         SFileDataBlockInfo* pFBlock = getCurrentBlockInfo(&pReader->status.blockIter);
@@ -2018,7 +2029,7 @@ static int32_t buildComposedDataBlock(STsdbReader* pReader, STableBlockScanInfo*
 
 void setComposedBlockFlag(STsdbReader* pReader, bool composed) { pReader->status.composedDataBlock = composed; }
 
-static int32_t initMemIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader) {
+static int32_t initMemDataIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader) {
   if (pBlockScanInfo->iterInit) {
     return TSDB_CODE_SUCCESS;
   }
@@ -2038,9 +2049,9 @@ static int32_t initMemIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader*
   if (pReader->pTsdb->mem != NULL) {
     tsdbGetTbDataFromMemTable(pReader->pTsdb->mem, pReader->suid, pBlockScanInfo->uid, &d);
     if (d != NULL) {
-      code = tsdbTbDataIterCreate(d, &startKey, backward, &pBlockScanInfo->iter);
+      code = tsdbTbDataIterCreate(d, &startKey, backward, &pBlockScanInfo->iter.iter);
       if (code == TSDB_CODE_SUCCESS) {
-        pBlockScanInfo->memHasVal = (tsdbTbDataIterGet(pBlockScanInfo->iter) != NULL);
+        pBlockScanInfo->iter.hasVal = (tsdbTbDataIterGet(pBlockScanInfo->iter.iter) != NULL);
 
         tsdbDebug("%p uid:%" PRId64 ", check data in mem from skey:%" PRId64 ", order:%d, ts range in buf:%" PRId64
                   "-%" PRId64 " %s",
@@ -2059,9 +2070,9 @@ static int32_t initMemIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader*
   if (pReader->pTsdb->imem != NULL) {
     tsdbGetTbDataFromMemTable(pReader->pTsdb->imem, pReader->suid, pBlockScanInfo->uid, &di);
     if (di != NULL) {
-      code = tsdbTbDataIterCreate(di, &startKey, backward, &pBlockScanInfo->iiter);
+      code = tsdbTbDataIterCreate(di, &startKey, backward, &pBlockScanInfo->iiter.iter);
       if (code == TSDB_CODE_SUCCESS) {
-        pBlockScanInfo->imemHasVal = (tsdbTbDataIterGet(pBlockScanInfo->iiter) != NULL);
+        pBlockScanInfo->iiter.hasVal = (tsdbTbDataIterGet(pBlockScanInfo->iiter.iter) != NULL);
 
         tsdbDebug("%p uid:%" PRId64 ", check data in imem from skey:%" PRId64 ", order:%d, ts range in buf:%" PRId64
                   "-%" PRId64 " %s",
@@ -2076,18 +2087,22 @@ static int32_t initMemIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader*
     tsdbDebug("%p uid:%" PRId64 ", no data in imem, %s", pReader, pBlockScanInfo->uid, pReader->idStr);
   }
 
+  initDelSkylineIterator(pBlockScanInfo, pReader, d, di);
+
   pBlockScanInfo->iterInit = true;
   return TSDB_CODE_SUCCESS;
 }
 
-static int32_t initDelSkylineIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader) {
+int32_t initDelSkylineIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader, STbData* pMemTbData, STbData* piMemTbData) {
   if (pBlockScanInfo->delSkyline != NULL) {
     return TSDB_CODE_SUCCESS;
   }
-#if 0
+
   int32_t code = 0;
   STsdb*  pTsdb = pReader->pTsdb;
 
+  SArray* pDelData = taosArrayInit(4, sizeof(SDelData));
+
   SDelFile *pDelFile = tsdbFSStateGetDelFile(pTsdb->fs->cState);
   if (pDelFile) {
     SDelFReader* pDelFReader = NULL;
@@ -2101,27 +2116,51 @@ static int32_t initDelSkylineIterator(STableBlockScanInfo* pBlockScanInfo, STsdb
       goto _err;
     }
 
-    SDelFile* pDelFileR = pReader->pTsdb->fs->nState->pDelFile;
-    if (pDelFileR) {
-      code = tsdbDelFReaderOpen(&pDelFReader, pDelFileR, pTsdb, NULL);
-      if (code) {
-        goto _err;
-      }
-
     code = tsdbReadDelIdx(pDelFReader, aDelIdx, NULL);
     if (code) {
       goto _err;
     }
+
+    SDelIdx  idx = {.suid = pReader->suid, .uid = pBlockScanInfo->uid};
+    SDelIdx* pIdx = taosArraySearch(aDelIdx, &idx, tCmprDelIdx, TD_EQ);
+
+    code = tsdbReadDelData(pDelFReader, pIdx, pDelData, NULL);
+    if (code != TSDB_CODE_SUCCESS) {
+      goto _err;
+    }
   }
 
-    code = tsdbBuildDeleteSkyline(pBlockScanInfo->delSkyline, 0, (int32_t)(nDelData - 1), aSkyline);
+  SDelData* p = NULL;
+  if (pMemTbData != NULL) {
+    p = pMemTbData->pHead;
+    while (p) {
+      taosArrayPush(pDelData, p);
+      p = p->pNext;
+    }
   }
 
-  _err:
+  if (piMemTbData != NULL) {
+    p = piMemTbData->pHead;
+    while (p) {
+      taosArrayPush(pDelData, p);
+      p = p->pNext;
+    }
+  }
+
+  if (taosArrayGetSize(pDelData) > 0) {
+    pBlockScanInfo->delSkyline = taosArrayInit(4, sizeof(TSDBKEY));
+    code = tsdbBuildDeleteSkyline(pDelData, 0, (int32_t)(taosArrayGetSize(pDelData) - 1), pBlockScanInfo->delSkyline);
+  }
+
+  taosArrayDestroy(pDelData);
+  pBlockScanInfo->iter.index = ASCENDING_TRAVERSE(pReader->order)? 0:taosArrayGetSize(pBlockScanInfo->delSkyline) - 1;
+  pBlockScanInfo->iiter.index = pBlockScanInfo->iter.index;
+  pBlockScanInfo->fileDelIndex = pBlockScanInfo->iter.index;
   return code;
 
-#endif
+_err:
-  return 0;
+  taosArrayDestroy(pDelData);
+  return code;
 }
 
 static TSDBKEY getCurrentKeyInBuf(SDataBlockIter* pBlockIter, STsdbReader* pReader) {
@@ -2130,13 +2169,13 @@ static TSDBKEY getCurrentKeyInBuf(SDataBlockIter* pBlockIter, STsdbReader* pRead
   SFileDataBlockInfo*  pFBlock = getCurrentBlockInfo(pBlockIter);
   STableBlockScanInfo* pScanInfo = taosHashGet(pReader->status.pTableMap, &pFBlock->uid, sizeof(pFBlock->uid));
 
-  initMemIterator(pScanInfo, pReader);
+  initMemDataIterator(pScanInfo, pReader);
-  TSDBROW* pRow = getValidRow(pScanInfo->iter, &pScanInfo->memHasVal, pReader);
+  TSDBROW* pRow = getValidRow(&pScanInfo->iter, pScanInfo->delSkyline, pReader);
   if (pRow != NULL) {
     key = TSDBROW_KEY(pRow);
   }
 
-  pRow = getValidRow(pScanInfo->iiter, &pScanInfo->imemHasVal, pReader);
+  pRow = getValidRow(&pScanInfo->iiter, pScanInfo->delSkyline, pReader);
   if (pRow != NULL) {
     TSDBKEY k = TSDBROW_KEY(pRow);
     if (key.ts > k.ts) {
@@ -2230,7 +2269,7 @@ static int32_t buildBlockFromBufferSequentially(STsdbReader* pReader) {
     }
 
     STableBlockScanInfo* pBlockScanInfo = pStatus->pTableIter;
-    initMemIterator(pBlockScanInfo, pReader);
+    initMemDataIterator(pBlockScanInfo, pReader);
 
     int64_t endKey = (ASCENDING_TRAVERSE(pReader->order)) ? INT64_MAX : INT64_MIN;
     int32_t code = buildDataBlockFromBuf(pReader, pBlockScanInfo, endKey);
@@ -2369,51 +2408,89 @@ static int32_t buildBlockFromFiles(STsdbReader* pReader) {
 //   taosArrayPush(pTsdbReadHandle->pTableCheckInfo, &info);
 // }
 
-TSDBROW* getValidRow(STbDataIter* pIter, bool* hasVal, STsdbReader* pReader) {
+bool hasBeenDropped(const SArray* pDelList, int32_t* index, TSDBKEY* pKey) {
-  if (!(*hasVal)) {
+  ASSERT(pKey != NULL);
+  if (pDelList == NULL) {
+    return false;
+  }
+
+  if (*index >= taosArrayGetSize(pDelList) - 1) {
+    TSDBKEY* last = taosArrayGetLast(pDelList);
+    if (pKey->ts > last->ts) {
+      return false;
+    } else if (pKey->ts == last->ts) {
+      size_t size = taosArrayGetSize(pDelList);
+      TSDBKEY* prev = taosArrayGet(pDelList, size - 2);
+      if (prev->version >= pKey->version) {
+        return true;
+      } else {
+        return false;
+      }
+    } else {
+      ASSERT(0);
+    }
+  } else {
+    TSDBKEY* pCurrent = taosArrayGet(pDelList, *index);
+    TSDBKEY* pNext = taosArrayGet(pDelList, (*index) + 1);
+
+    if (pCurrent->ts <= pKey->ts && pNext->ts >= pKey->ts && pCurrent->version >= pKey->version) {
+      return true;
+    } else {
+      while (pNext->ts < pKey->ts && (*index) < taosArrayGetSize(pDelList) - 1) {
+        (*index) += 1;
+      }
+
+      return false;
+    }
+  }
+}
+
+TSDBROW* getValidRow(SIterInfo* pIter, const SArray* pDelList, STsdbReader* pReader) {
+  if (!pIter->hasVal) {
     return NULL;
   }
 
-  TSDBROW* pRow = tsdbTbDataIterGet(pIter);
+  TSDBROW* pRow = tsdbTbDataIterGet(pIter->iter);
   TSDBKEY  key = TSDBROW_KEY(pRow);
   if (outOfTimeWindow(key.ts, &pReader->window)) {
-    *hasVal = false;
+    pIter->hasVal = false;
     return NULL;
   }
 
-  if (key.version <= pReader->verRange.maxVer && key.version >= pReader->verRange.minVer) {
+  // it is a valid data version
+  if ((key.version <= pReader->verRange.maxVer && key.version >= pReader->verRange.minVer) && (!hasBeenDropped(pDelList, &pIter->index, &key))) {
     return pRow;
   }
 
   while (1) {
-    *hasVal = tsdbTbDataIterNext(pIter);
+    pIter->hasVal = tsdbTbDataIterNext(pIter->iter);
-    if (!(*hasVal)) {
+    if (!pIter->hasVal) {
       return NULL;
     }
 
-    pRow = tsdbTbDataIterGet(pIter);
+    pRow = tsdbTbDataIterGet(pIter->iter);
 
     key = TSDBROW_KEY(pRow);
     if (outOfTimeWindow(key.ts, &pReader->window)) {
-      *hasVal = false;
+      pIter->hasVal = false;
       return NULL;
     }
 
-    if (key.version <= pReader->verRange.maxVer && key.version >= pReader->verRange.minVer) {
+    if (key.version <= pReader->verRange.maxVer && key.version >= pReader->verRange.minVer && (!hasBeenDropped(pDelList, &pIter->index, &key))) {
       return pRow;
     }
   }
 }
 
-int32_t doMergeRowsInBuf(STbDataIter* pIter, bool* hasVal, int64_t ts, SRowMerger* pMerger, STsdbReader* pReader) {
+int32_t doMergeRowsInBuf(SIterInfo *pIter, int64_t ts, SArray* pDelList, SRowMerger* pMerger, STsdbReader* pReader) {
   while (1) {
-    *hasVal = tsdbTbDataIterNext(pIter);
+    pIter->hasVal = tsdbTbDataIterNext(pIter->iter);
-    if (!(*hasVal)) {
+    if (!pIter->hasVal) {
       break;
     }
 
     // data exists but not valid
-    TSDBROW* pRow = getValidRow(pIter, hasVal, pReader);
+    TSDBROW* pRow = getValidRow(pIter, pDelList, pReader);
     if (pRow == NULL) {
       break;
     }
@@ -2539,15 +2616,14 @@ void updateSchema(TSDBROW* pRow, uint64_t uid, STsdbReader* pReader) {
   }
 }
 
-void doMergeMultiRows(TSDBROW* pRow, uint64_t uid, STbDataIter* dIter, bool* hasVal, STSRow** pTSRow,
+void doMergeMultiRows(TSDBROW* pRow, uint64_t uid, SIterInfo *pIter, SArray* pDelList, STSRow** pTSRow, STsdbReader* pReader) {
-                      STsdbReader* pReader) {
   SRowMerger merge = {0};
 
   TSDBKEY k = TSDBROW_KEY(pRow);
   updateSchema(pRow, uid, pReader);
 
   tRowMergerInit(&merge, pRow, pReader->pSchema);
-  doMergeRowsInBuf(dIter, hasVal, k.ts, &merge, pReader);
+  doMergeRowsInBuf(pIter, k.ts, pDelList, &merge, pReader);
   tRowMergerGetRow(&merge, pTSRow);
 }
 
@@ -2562,18 +2638,18 @@ void doMergeMemIMemRows(TSDBROW* pRow, TSDBROW* piRow, STableBlockScanInfo* pBlo
     updateSchema(piRow, pBlockScanInfo->uid, pReader);
 
     tRowMergerInit(&merge, piRow, pReader->pSchema);
-    doMergeRowsInBuf(pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, ik.ts, &merge, pReader);
+    doMergeRowsInBuf(&pBlockScanInfo->iiter, ik.ts, pBlockScanInfo->delSkyline, &merge, pReader);
 
     tRowMerge(&merge, pRow);
-    doMergeRowsInBuf(pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, k.ts, &merge, pReader);
+    doMergeRowsInBuf(&pBlockScanInfo->iter, k.ts, pBlockScanInfo->delSkyline, &merge, pReader);
   } else {
     updateSchema(pRow, pBlockScanInfo->uid, pReader);
 
     tRowMergerInit(&merge, pRow, pReader->pSchema);
-    doMergeRowsInBuf(pBlockScanInfo->iiter, &pBlockScanInfo->memHasVal, ik.ts, &merge, pReader);
+    doMergeRowsInBuf(&pBlockScanInfo->iter, k.ts, pBlockScanInfo->delSkyline, &merge, pReader);
 
     tRowMerge(&merge, piRow);
-    doMergeRowsInBuf(pBlockScanInfo->iter, &pBlockScanInfo->imemHasVal, k.ts, &merge, pReader);
+    doMergeRowsInBuf(&pBlockScanInfo->iiter, k.ts, pBlockScanInfo->delSkyline, &merge, pReader);
   }
 
   tRowMergerGetRow(&merge, pTSRow);
@@ -2581,33 +2657,34 @@ void doMergeMemIMemRows(TSDBROW* pRow, TSDBROW* piRow, STableBlockScanInfo* pBlo
 
 int32_t tsdbGetNextRowInMem(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader, STSRow** pTSRow,
                             int64_t endKey) {
-  TSDBROW* pRow = getValidRow(pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, pReader);
+  TSDBROW* pRow = getValidRow(&pBlockScanInfo->iter, pBlockScanInfo->delSkyline, pReader);
-  TSDBROW* piRow = getValidRow(pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, pReader);
+  TSDBROW* piRow = getValidRow(&pBlockScanInfo->iiter, pBlockScanInfo->delSkyline, pReader);
+  SArray* pDelList = pBlockScanInfo->delSkyline;
 
   // todo refactor
   bool asc = ASCENDING_TRAVERSE(pReader->order);
-  if (pBlockScanInfo->memHasVal) {
+  if (pBlockScanInfo->iter.hasVal) {
     TSDBKEY k = TSDBROW_KEY(pRow);
     if ((k.ts >= endKey && asc) || (k.ts <= endKey && !asc)) {
       pRow = NULL;
     }
   }
 
-  if (pBlockScanInfo->imemHasVal) {
+  if (pBlockScanInfo->iiter.hasVal) {
     TSDBKEY k = TSDBROW_KEY(piRow);
     if ((k.ts >= endKey && asc) || (k.ts <= endKey && !asc)) {
       piRow = NULL;
     }
   }
 
-  if (pBlockScanInfo->memHasVal && pBlockScanInfo->imemHasVal && pRow != NULL && piRow != NULL) {
+  if (pBlockScanInfo->iter.hasVal && pBlockScanInfo->iiter.hasVal && pRow != NULL && piRow != NULL) {
     TSDBKEY k = TSDBROW_KEY(pRow);
     TSDBKEY ik = TSDBROW_KEY(piRow);
 
     if (ik.ts < k.ts) {  // ik.ts < k.ts
-      doMergeMultiRows(piRow, pBlockScanInfo->uid, pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, pTSRow, pReader);
+      doMergeMultiRows(piRow, pBlockScanInfo->uid, &pBlockScanInfo->iiter, pDelList, pTSRow, pReader);
     } else if (k.ts < ik.ts) {
-      doMergeMultiRows(pRow, pBlockScanInfo->uid, pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, pTSRow, pReader);
+      doMergeMultiRows(pRow, pBlockScanInfo->uid, &pBlockScanInfo->iter, pDelList, pTSRow, pReader);
     } else {  // ik.ts == k.ts
       doMergeMemIMemRows(pRow, piRow, pBlockScanInfo, pReader, pTSRow);
     }
@@ -2615,13 +2692,13 @@ int32_t tsdbGetNextRowInMem(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pR
     return TSDB_CODE_SUCCESS;
   }
 
-  if (pBlockScanInfo->memHasVal && pRow != NULL) {
+  if (pBlockScanInfo->iter.hasVal && pRow != NULL) {
-    doMergeMultiRows(pRow, pBlockScanInfo->uid, pBlockScanInfo->iter, &pBlockScanInfo->memHasVal, pTSRow, pReader);
+    doMergeMultiRows(pRow, pBlockScanInfo->uid, &pBlockScanInfo->iter, pDelList, pTSRow, pReader);
     return TSDB_CODE_SUCCESS;
   }
 
-  if (pBlockScanInfo->imemHasVal && piRow != NULL) {
+  if (pBlockScanInfo->iiter.hasVal && piRow != NULL) {
-    doMergeMultiRows(piRow, pBlockScanInfo->uid, pBlockScanInfo->iiter, &pBlockScanInfo->imemHasVal, pTSRow, pReader);
+    doMergeMultiRows(piRow, pBlockScanInfo->uid, &pBlockScanInfo->iiter, pDelList, pTSRow, pReader);
     return TSDB_CODE_SUCCESS;
   }
 
@@ -2686,7 +2763,7 @@ int32_t buildDataBlockFromBufImpl(STableBlockScanInfo* pBlockScanInfo, int64_t e
     doAppendOneRow(pBlock, pReader, pTSRow);
 
     // no data in buffer, return immediately
-    if (!(pBlockScanInfo->memHasVal || pBlockScanInfo->imemHasVal)) {
+    if (!(pBlockScanInfo->iter.hasVal || pBlockScanInfo->iiter.hasVal)) {
       break;
     }
 
@@ -2699,12 +2776,13 @@ int32_t buildDataBlockFromBufImpl(STableBlockScanInfo* pBlockScanInfo, int64_t e
   return TSDB_CODE_SUCCESS;
 }
 
+// todo refactor, use arraylist instead
 int32_t tsdbSetTableId(STsdbReader* pReader, int64_t uid) {
-  // if (pReader->pTableCheckInfo) taosArrayDestroy(pReader->pTableCheckInfo);
+  ASSERT(pReader != NULL);
-  // pReader->pTableCheckInfo = createCheckInfoFromUid(pReader, uid);
+  taosHashClear(pReader->status.pTableMap);
-  // if (pReader->pTableCheckInfo == NULL) {
+
-  //   return TSDB_CODE_TDB_OUT_OF_MEMORY;
+  STableBlockScanInfo info = {.lastKey = 0, .uid = uid};
-  // }
+  taosHashPut(pReader->status.pTableMap, &info.uid, sizeof(uint64_t), &info, sizeof(info));
   return TDB_CODE_SUCCESS;
 }
 
@@ -3165,40 +3243,46 @@ SArray* tsdbRetrieveDataBlock(STsdbReader* pReader, SArray* pIdList) {
   return pReader->pResBlock->pDataBlock;
 }
 
-void tsdbResetReadHandle(STsdbReader* pReader, SQueryTableDataCond* pCond, int32_t tWinIdx) {
+int32_t tsdbReaderReset(STsdbReader* pReader, SQueryTableDataCond* pCond, int32_t tWinIdx) {
-  // if (isEmptyQueryTimeWindow(pReader)) {
+  if (isEmptyQueryTimeWindow(&pReader->window)) {
-  //   if (pCond->order != pReader->order) {
+    return TSDB_CODE_SUCCESS;
-  //     pReader->order = pCond->order;
+  }
-  //     TSWAP(pReader->window.skey, pReader->window.ekey);
-  //   }
 
-  //   return;
+  setQueryTimewindow(pReader, pCond, tWinIdx);
-  // }
 
-  // pReader->order = pCond->order;
+  pReader->order = pCond->order;
-  // setQueryTimewindow(pReader, pCond, tWinIdx);
+  pReader->type = BLOCK_LOAD_OFFSET_ORDER;
-  // pReader->type = TSDB_QUERY_TYPE_ALL;
+  pReader->status.loadFromFile = true;
-  // pReader->cur.fid = -1;
+  pReader->status.pTableIter = NULL;
-  // pReader->cur.win = TSWINDOW_INITIALIZER;
-  // pReader->checkFiles = true;
-  // pReader->activeIndex = 0;  // current active table index
-  // pReader->locateStart = false;
-  // pReader->loadExternalRow = pCond->loadExternalRows;
 
-  // if (ASCENDING_TRAVERSE(pCond->order)) {
+  pReader->window = updateQueryTimeWindow(pReader->pTsdb, &pCond->twindows[tWinIdx]);
-  //   assert(pReader->window.skey <= pReader->window.ekey);
-  // } else {
-  //   assert(pReader->window.skey >= pReader->window.ekey);
-  // }
 
-  // // allocate buffer in order to load data blocks from file
+  // allocate buffer in order to load data blocks from file
-  // memset(pReader->suppInfo.pstatis, 0, sizeof(SColumnDataAgg));
+  memset(pReader->suppInfo.pstatis, 0, sizeof(SColumnDataAgg));
-  // memset(pReader->suppInfo.plist, 0, POINTER_BYTES);
+  memset(pReader->suppInfo.plist, 0, POINTER_BYTES);
 
-  // tsdbInitDataBlockLoadInfo(&pReader->dataBlockLoadInfo);
+  // todo set the correct numOfTables
-  // tsdbInitCompBlockLoadInfo(&pReader->compBlockLoadInfo);
+  int32_t         numOfTables = 1;
+  SDataBlockIter* pBlockIter = &pReader->status.blockIter;
 
-  // resetCheckInfo(pReader);
+  STsdbFSState* pFState = pReader->pTsdb->fs->cState;
+  initFilesetIterator(&pReader->status.fileIter, pFState, pReader->order, pReader->idStr);
+  resetDataBlockIterator(&pReader->status.blockIter, pReader->order);
+
+  int32_t code = 0;
+  // no data in files, let's try buffer in memory
+  if (pReader->status.fileIter.numOfFiles == 0) {
+    pReader->status.loadFromFile = false;
+  } else {
+    code = initForFirstBlockInFile(pReader, pBlockIter);
+    if (code != TSDB_CODE_SUCCESS) {
+      return code;
+    }
+  }
+
+  ASSERT(0);
+  tsdbDebug("%p reset tsdbreader in query %s", pReader, numOfTables, pReader->idStr);
+  return code;
 }
 
 int32_t tsdbGetFileBlocksDistInfo(STsdbReader* pReader, STableBlockDistInfo* pTableBlockInfo) {

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

@@ -345,8 +345,14 @@ int32_t tsdbReadDelData(SDelFReader *pReader, SDelIdx *pDelIdx, SArray *aDelData
   ASSERT(pHdr->suid == pDelIdx->suid);
   ASSERT(pHdr->uid == pDelIdx->uid);
   n += sizeof(*pHdr);
+  taosArrayClear(aDelData);
   while (n < size - sizeof(TSCKSUM)) {
     n += tGetDelData(*ppBuf + n, pDelData);
+
+    if (taosArrayPush(aDelData, pDelData) == NULL) {
+      code = TSDB_CODE_OUT_OF_MEMORY;
+      goto _err;
+    }
   }
 
   ASSERT(n == size - sizeof(TSCKSUM));

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

@@ -229,15 +229,15 @@ int32_t tCmprBlockIdx(void const *lhs, void const *rhs) {
   SBlockIdx *lBlockIdx = *(SBlockIdx **)lhs;
   SBlockIdx *rBlockIdx = *(SBlockIdx **)rhs;
 
-  if (lBlockIdx->suid < lBlockIdx->suid) {
+  if (lBlockIdx->suid < rBlockIdx->suid) {
     return -1;
-  } else if (lBlockIdx->suid > lBlockIdx->suid) {
+  } else if (lBlockIdx->suid > rBlockIdx->suid) {
     return 1;
   }
 
-  if (lBlockIdx->uid < lBlockIdx->uid) {
+  if (lBlockIdx->uid < rBlockIdx->uid) {
     return -1;
-  } else if (lBlockIdx->uid > lBlockIdx->uid) {
+  } else if (lBlockIdx->uid > rBlockIdx->uid) {
     return 1;
   }
 
@@ -385,15 +385,15 @@ int32_t tCmprDelIdx(void const *lhs, void const *rhs) {
   SDelIdx *lDelIdx = *(SDelIdx **)lhs;
   SDelIdx *rDelIdx = *(SDelIdx **)rhs;
 
-  if (lDelIdx->suid < lDelIdx->suid) {
+  if (lDelIdx->suid < rDelIdx->suid) {
     return -1;
-  } else if (lDelIdx->suid > lDelIdx->suid) {
+  } else if (lDelIdx->suid > rDelIdx->suid) {
     return 1;
   }
 
-  if (lDelIdx->uid < lDelIdx->uid) {
+  if (lDelIdx->uid < rDelIdx->uid) {
     return -1;
-  } else if (lDelIdx->uid > lDelIdx->uid) {
+  } else if (lDelIdx->uid > rDelIdx->uid) {
     return 1;
   }
 

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

@@ -256,59 +256,54 @@ int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
 
     SRpcMsg *pRpcMsg = pMsg;
 
+    // ToDo: ugly! use function pointer
+    // use different strategy
+    if (syncNodeStrategy(pSyncNode) == SYNC_STRATEGY_NO_SNAPSHOT) {
       if (pRpcMsg->msgType == TDMT_SYNC_TIMEOUT) {
         SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
         syncTimeoutDestroy(pSyncMsg);
 
       } else if (pRpcMsg->msgType == TDMT_SYNC_PING) {
         SyncPing *pSyncMsg = syncPingFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnPingCb(pSyncNode, pSyncMsg);
         syncPingDestroy(pSyncMsg);
 
       } else if (pRpcMsg->msgType == TDMT_SYNC_PING_REPLY) {
         SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
         syncPingReplyDestroy(pSyncMsg);
 
       } else if (pRpcMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
         SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg, NULL);
         syncClientRequestDestroy(pSyncMsg);
 
       } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
         SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
         syncRequestVoteDestroy(pSyncMsg);
 
       } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
         SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
         syncRequestVoteReplyDestroy(pSyncMsg);
 
       } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
         SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
         syncAppendEntriesDestroy(pSyncMsg);
 
       } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
         SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pRpcMsg);
-      assert(pSyncMsg != NULL);
+        ASSERT(pSyncMsg != NULL);
-
         ret = syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
         syncAppendEntriesReplyDestroy(pSyncMsg);
 
@@ -322,6 +317,74 @@ int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
         ret = -1;
       }
 
+    } else {
+      // use wal first strategy
+
+      if (pRpcMsg->msgType == TDMT_SYNC_TIMEOUT) {
+        SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
+        syncTimeoutDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_PING) {
+        SyncPing *pSyncMsg = syncPingFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnPingCb(pSyncNode, pSyncMsg);
+        syncPingDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_PING_REPLY) {
+        SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
+        syncPingReplyDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
+        SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg, NULL);
+        syncClientRequestDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_CLIENT_REQUEST_BATCH) {
+        SyncClientRequestBatch *pSyncMsg = syncClientRequestBatchFromRpcMsg(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnClientRequestBatchCb(pSyncNode, pSyncMsg);
+        syncClientRequestBatchDestroyDeep(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
+        SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
+        syncRequestVoteDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
+        SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
+        syncRequestVoteReplyDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_BATCH) {
+        SyncAppendEntriesBatch *pSyncMsg = syncAppendEntriesBatchFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnAppendEntriesSnapshot2Cb(pSyncNode, pSyncMsg);
+        syncAppendEntriesBatchDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
+        SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pRpcMsg);
+        ASSERT(pSyncMsg != NULL);
+        ret = syncNodeOnAppendEntriesReplySnapshot2Cb(pSyncNode, pSyncMsg);
+        syncAppendEntriesReplyDestroy(pSyncMsg);
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_SET_VNODE_STANDBY) {
+        ret = vnodeSetStandBy(pVnode);
+        if (ret != 0 && terrno != 0) ret = terrno;
+        SRpcMsg rsp = {.code = ret, .info = pMsg->info};
+        tmsgSendRsp(&rsp);
+      } else {
+        vError("==vnodeProcessSyncReq== error msg type:%d", pRpcMsg->msgType);
+        ret = -1;
+      }
+    }
+
     syncNodeRelease(pSyncNode);
   } else {
     vError("==vnodeProcessSyncReq== error syncEnv stop");
@@ -415,7 +478,7 @@ static int32_t vnodeSnapshotStopRead(struct SSyncFSM *pFsm, void *pReader) { ret
 
 static int32_t vnodeSnapshotDoRead(struct SSyncFSM *pFsm, void *pReader, void **ppBuf, int32_t *len) { return 0; }
 
-static int32_t vnodeSnapshotStartWrite(struct SSyncFSM *pFsm, void **ppWriter) { return 0; }
+static int32_t vnodeSnapshotStartWrite(struct SSyncFSM *pFsm, void *pParam, void **ppWriter) { return 0; }
 
 static int32_t vnodeSnapshotStopWrite(struct SSyncFSM *pFsm, void *pWriter, bool isApply) { return 0; }
 
@@ -442,7 +505,8 @@ static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
 
 int32_t vnodeSyncOpen(SVnode *pVnode, char *path) {
   SSyncInfo syncInfo = {
-      .snapshotEnable = false,
+      .snapshotStrategy = SYNC_STRATEGY_NO_SNAPSHOT,
+      .batchSize = 10,
       .vgId = pVnode->config.vgId,
       .isStandBy = pVnode->config.standby,
       .syncCfg = pVnode->config.syncCfg,

source/libs/command/inc/commandInt.h

@@ -29,13 +29,17 @@ extern "C" {
 #define EXPLAIN_TAG_SCAN_FORMAT "Tag Scan on %s"
 #define EXPLAIN_TBL_SCAN_FORMAT "Table Scan on %s"
 #define EXPLAIN_SYSTBL_SCAN_FORMAT "System Table Scan on %s"
+#define EXPLAIN_DISTBLK_SCAN_FORMAT "Block Dist Scan on %s"
+#define EXPLAIN_LASTROW_SCAN_FORMAT "Last Row Scan on %s"
 #define EXPLAIN_PROJECTION_FORMAT "Projection"
 #define EXPLAIN_JOIN_FORMAT "%s"
 #define EXPLAIN_AGG_FORMAT "Aggragate"
 #define EXPLAIN_INDEF_ROWS_FORMAT "Indefinite Rows Function"
 #define EXPLAIN_EXCHANGE_FORMAT "Data Exchange %d:1"
 #define EXPLAIN_SORT_FORMAT "Sort"
+#define EXPLAIN_GROUP_SORT_FORMAT "Group Sort"
 #define EXPLAIN_INTERVAL_FORMAT "Interval on Column %s"
+#define EXPLAIN_MERGE_INTERVAL_FORMAT "Merge Interval on Column %s"
 #define EXPLAIN_FILL_FORMAT "Fill"
 #define EXPLAIN_SESSION_FORMAT "Session"
 #define EXPLAIN_STATE_WINDOW_FORMAT "StateWindow on Column %s"
@@ -62,10 +66,12 @@ extern "C" {
 #define EXPLAIN_COST_FORMAT "cost=%.2f..%.2f"
 #define EXPLAIN_ROWS_FORMAT "rows=%" PRIu64
 #define EXPLAIN_COLUMNS_FORMAT "columns=%d"
+#define EXPLAIN_PSEUDO_COLUMNS_FORMAT "pseudo_columns=%d"
 #define EXPLAIN_WIDTH_FORMAT "width=%d"
 #define EXPLAIN_TABLE_SCAN_FORMAT "order=[asc|%d desc|%d]"
 #define EXPLAIN_GROUPS_FORMAT "groups=%d"
 #define EXPLAIN_WIDTH_FORMAT "width=%d"
+#define EXPLAIN_INTERVAL_VALUE_FORMAT "interval=%" PRId64 "%c"
 #define EXPLAIN_FUNCTIONS_FORMAT "functions=%d"
 #define EXPLAIN_EXECINFO_FORMAT "cost=%.3f..%.3f rows=%" PRIu64
 #define EXPLAIN_MODE_FORMAT "mode=%s"

source/libs/command/src/explain.c

@@ -199,6 +199,31 @@ int32_t qExplainGenerateResChildren(SPhysiNode *pNode, SExplainGroup *group, SNo
       pPhysiChildren = mergePhysiNode->scan.node.pChildren;
       break;
     }
+    case QUERY_NODE_PHYSICAL_PLAN_BLOCK_DIST_SCAN: {
+      SBlockDistScanPhysiNode *distPhysiNode = (SBlockDistScanPhysiNode *)pNode;
+      pPhysiChildren = distPhysiNode->node.pChildren;
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_LAST_ROW_SCAN: {
+      SLastRowScanPhysiNode *lastRowPhysiNode = (SLastRowScanPhysiNode *)pNode;
+      pPhysiChildren = lastRowPhysiNode->node.pChildren;
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT: {
+      SGroupSortPhysiNode *groupSortPhysiNode = (SGroupSortPhysiNode *)pNode;
+      pPhysiChildren = groupSortPhysiNode->node.pChildren;
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL: {
+      SMergeIntervalPhysiNode *mergeIntPhysiNode = (SMergeIntervalPhysiNode *)pNode;
+      pPhysiChildren = mergeIntPhysiNode->window.node.pChildren;
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_INTERP_FUNC: {
+      SInterpFuncPhysiNode *interpPhysiNode = (SInterpFuncPhysiNode *)pNode;
+      pPhysiChildren = interpPhysiNode->node.pChildren;
+      break;
+    }
     default:
       qError("not supported physical node type %d", pNode->type);
       QRY_ERR_RET(TSDB_CODE_QRY_APP_ERROR);
@@ -378,6 +403,10 @@ int32_t qExplainResNodeToRowsImpl(SExplainResNode *pResNode, SExplainCtx *ctx, i
       }
       EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT, pTagScanNode->pScanCols->length);
       EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      if (pTagScanNode->pScanPseudoCols) {
+        EXPLAIN_ROW_APPEND(EXPLAIN_PSEUDO_COLUMNS_FORMAT, pTagScanNode->pScanPseudoCols->length);
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
       EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pTagScanNode->node.pOutputDataBlockDesc->totalRowSize);
       EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
       EXPLAIN_ROW_END();
@@ -415,6 +444,10 @@ int32_t qExplainResNodeToRowsImpl(SExplainResNode *pResNode, SExplainCtx *ctx, i
 
       EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT, pTblScanNode->scan.pScanCols->length);
       EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      if (pTblScanNode->scan.pScanPseudoCols) {
+        EXPLAIN_ROW_APPEND(EXPLAIN_PSEUDO_COLUMNS_FORMAT, pTblScanNode->scan.pScanPseudoCols->length);
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
       EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pTblScanNode->scan.node.pOutputDataBlockDesc->totalRowSize);
       EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
       EXPLAIN_ROW_APPEND(EXPLAIN_TABLE_SCAN_FORMAT, pTblScanNode->scanSeq[0], pTblScanNode->scanSeq[1]);
@@ -516,6 +549,10 @@ int32_t qExplainResNodeToRowsImpl(SExplainResNode *pResNode, SExplainCtx *ctx, i
       }
       EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT, pSTblScanNode->scan.pScanCols->length);
       EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      if (pSTblScanNode->scan.pScanPseudoCols) {
+        EXPLAIN_ROW_APPEND(EXPLAIN_PSEUDO_COLUMNS_FORMAT, pSTblScanNode->scan.pScanPseudoCols->length);
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT); 
+      }
       EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pSTblScanNode->scan.node.pOutputDataBlockDesc->totalRowSize);
       EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
       EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
@@ -1131,6 +1168,258 @@ int32_t qExplainResNodeToRowsImpl(SExplainResNode *pResNode, SExplainCtx *ctx, i
       }
       break;
     }
+    case QUERY_NODE_PHYSICAL_PLAN_BLOCK_DIST_SCAN: {
+      SBlockDistScanPhysiNode *pDistScanNode = (SBlockDistScanPhysiNode *)pNode;
+      EXPLAIN_ROW_NEW(level, EXPLAIN_DISTBLK_SCAN_FORMAT, pDistScanNode->tableName.tname);
+      EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
+      if (pResNode->pExecInfo) {
+        QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
+      EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT, pDistScanNode->pScanCols->length);
+      EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      if (pDistScanNode->pScanPseudoCols) {
+        EXPLAIN_ROW_APPEND(EXPLAIN_PSEUDO_COLUMNS_FORMAT, pDistScanNode->pScanPseudoCols->length);
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);   
+      }
+      EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pDistScanNode->node.pOutputDataBlockDesc->totalRowSize);
+      EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
+      EXPLAIN_ROW_END();
+      QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
+
+      if (verbose) {
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
+                           nodesGetOutputNumFromSlotList(pDistScanNode->node.pOutputDataBlockDesc->pSlots));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pDistScanNode->node.pOutputDataBlockDesc->outputRowSize);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+
+        if (pDistScanNode->node.pConditions) {
+          EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
+          QRY_ERR_RET(nodesNodeToSQL(pDistScanNode->node.pConditions, tbuf + VARSTR_HEADER_SIZE,
+                                     TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
+          EXPLAIN_ROW_END();
+          QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        }
+      }
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_LAST_ROW_SCAN: {
+      SLastRowScanPhysiNode *pLastRowNode = (SLastRowScanPhysiNode *)pNode;
+      EXPLAIN_ROW_NEW(level, EXPLAIN_LASTROW_SCAN_FORMAT, pLastRowNode->tableName.tname);
+      EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
+      if (pResNode->pExecInfo) {
+        QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
+      EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT, pLastRowNode->pScanCols->length);
+      EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      if (pLastRowNode->pScanPseudoCols) {
+        EXPLAIN_ROW_APPEND(EXPLAIN_PSEUDO_COLUMNS_FORMAT, pLastRowNode->pScanPseudoCols->length);
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
+      EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pLastRowNode->node.pOutputDataBlockDesc->totalRowSize);
+      EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
+      EXPLAIN_ROW_END();
+      QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
+
+      if (verbose) {
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
+                           nodesGetOutputNumFromSlotList(pLastRowNode->node.pOutputDataBlockDesc->pSlots));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pLastRowNode->node.pOutputDataBlockDesc->outputRowSize);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+
+        if (pLastRowNode->node.pConditions) {
+          EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
+          QRY_ERR_RET(nodesNodeToSQL(pLastRowNode->node.pConditions, tbuf + VARSTR_HEADER_SIZE,
+                                     TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
+          EXPLAIN_ROW_END();
+          QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        }
+      }
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT: {
+      SGroupSortPhysiNode *pSortNode = (SGroupSortPhysiNode *)pNode;
+      EXPLAIN_ROW_NEW(level, EXPLAIN_GROUP_SORT_FORMAT);
+      EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
+      if (pResNode->pExecInfo) {
+        QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
+
+      SDataBlockDescNode *pDescNode = pSortNode->node.pOutputDataBlockDesc;
+      EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT, nodesGetOutputNumFromSlotList(pDescNode->pSlots));
+      EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pDescNode->totalRowSize);
+      EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
+      EXPLAIN_ROW_END();
+      QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
+
+      if (EXPLAIN_MODE_ANALYZE == ctx->mode) {
+        // sort key
+        EXPLAIN_ROW_NEW(level + 1, "Sort Key: ");
+        if (pResNode->pExecInfo) {
+          for (int32_t i = 0; i < LIST_LENGTH(pSortNode->pSortKeys); ++i) {
+            SOrderByExprNode *ptn = (SOrderByExprNode *)nodesListGetNode(pSortNode->pSortKeys, i);
+            EXPLAIN_ROW_APPEND("%s ", nodesGetNameFromColumnNode(ptn->pExpr));
+          }
+        }
+
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
+
+        // sort method
+        EXPLAIN_ROW_NEW(level + 1, "Sort Method: ");
+
+        int32_t           nodeNum = taosArrayGetSize(pResNode->pExecInfo);
+        SExplainExecInfo *execInfo = taosArrayGet(pResNode->pExecInfo, 0);
+        SSortExecInfo    *pExecInfo = (SSortExecInfo *)execInfo->verboseInfo;
+        EXPLAIN_ROW_APPEND("%s", pExecInfo->sortMethod == SORT_QSORT_T ? "quicksort" : "merge sort");
+        if (pExecInfo->sortBuffer > 1024 * 1024) {
+          EXPLAIN_ROW_APPEND("  Buffers:%.2f Mb", pExecInfo->sortBuffer / (1024 * 1024.0));
+        } else if (pExecInfo->sortBuffer > 1024) {
+          EXPLAIN_ROW_APPEND("  Buffers:%.2f Kb", pExecInfo->sortBuffer / (1024.0));
+        } else {
+          EXPLAIN_ROW_APPEND("  Buffers:%d b", pExecInfo->sortBuffer);
+        }
+
+        EXPLAIN_ROW_APPEND("  loops:%d", pExecInfo->loops);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
+      }
+
+      if (verbose) {
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
+                           nodesGetOutputNumFromSlotList(pSortNode->node.pOutputDataBlockDesc->pSlots));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pSortNode->node.pOutputDataBlockDesc->outputRowSize);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+
+        if (pSortNode->node.pConditions) {
+          EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
+          QRY_ERR_RET(nodesNodeToSQL(pSortNode->node.pConditions, tbuf + VARSTR_HEADER_SIZE,
+                                     TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
+          EXPLAIN_ROW_END();
+          QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        }
+      }
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL: {
+      SMergeIntervalPhysiNode *pIntNode = (SMergeIntervalPhysiNode *)pNode;
+      EXPLAIN_ROW_NEW(level, EXPLAIN_MERGE_INTERVAL_FORMAT, nodesGetNameFromColumnNode(pIntNode->window.pTspk));
+      EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
+      if (pResNode->pExecInfo) {
+        QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
+      EXPLAIN_ROW_APPEND(EXPLAIN_FUNCTIONS_FORMAT, pIntNode->window.pFuncs->length);
+      EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
+      EXPLAIN_ROW_END();
+      QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
+
+      if (verbose) {
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
+                           nodesGetOutputNumFromSlotList(pIntNode->window.node.pOutputDataBlockDesc->pSlots));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pIntNode->window.node.pOutputDataBlockDesc->outputRowSize);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        uint8_t precision = getIntervalPrecision(pIntNode);
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_TIME_WINDOWS_FORMAT,
+                        INVERAL_TIME_FROM_PRECISION_TO_UNIT(pIntNode->interval, pIntNode->intervalUnit, precision),
+                        pIntNode->intervalUnit, pIntNode->offset, getPrecisionUnit(precision),
+                        INVERAL_TIME_FROM_PRECISION_TO_UNIT(pIntNode->sliding, pIntNode->slidingUnit, precision),
+                        pIntNode->slidingUnit);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+
+        if (pIntNode->window.node.pConditions) {
+          EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
+          QRY_ERR_RET(nodesNodeToSQL(pIntNode->window.node.pConditions, tbuf + VARSTR_HEADER_SIZE,
+                                     TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
+          EXPLAIN_ROW_END();
+          QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        }
+      }
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_INTERP_FUNC: {
+      SInterpFuncPhysiNode *pInterpNode = (SInterpFuncPhysiNode *)pNode;
+      EXPLAIN_ROW_NEW(level, EXPLAIN_AGG_FORMAT);
+      EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
+      if (pResNode->pExecInfo) {
+        QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
+      if (pInterpNode->pFuncs) {
+        EXPLAIN_ROW_APPEND(EXPLAIN_FUNCTIONS_FORMAT, pInterpNode->pFuncs->length);
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+      }
+      
+      EXPLAIN_ROW_APPEND(EXPLAIN_MODE_FORMAT, nodesGetFillModeString(pInterpNode->fillMode));
+      EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+
+      EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
+      EXPLAIN_ROW_END();
+      QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
+
+      if (verbose) {
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
+                           nodesGetOutputNumFromSlotList(pInterpNode->node.pOutputDataBlockDesc->pSlots));
+        EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+        EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pInterpNode->node.pOutputDataBlockDesc->outputRowSize);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        if (pInterpNode->pFillValues) {
+          SNodeListNode *pValues = (SNodeListNode *)pInterpNode->pFillValues;
+          EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILL_VALUE_FORMAT);
+          SNode  *tNode = NULL;
+          int32_t i = 0;
+          FOREACH(tNode, pValues->pNodeList) {
+            if (i) {
+              EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
+            }
+            SValueNode *tValue = (SValueNode *)tNode;
+            char       *value = nodesGetStrValueFromNode(tValue);
+            EXPLAIN_ROW_APPEND(EXPLAIN_STRING_TYPE_FORMAT, value);
+            taosMemoryFree(value);
+            ++i;
+          }
+
+          EXPLAIN_ROW_END();
+          QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        }
+
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_INTERVAL_VALUE_FORMAT, pInterpNode->interval, pInterpNode->intervalUnit);
+        EXPLAIN_ROW_END();
+
+        EXPLAIN_ROW_NEW(level + 1, EXPLAIN_TIMERANGE_FORMAT, pInterpNode->timeRange.skey, pInterpNode->timeRange.ekey);
+        EXPLAIN_ROW_END();
+        QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+
+        if (pInterpNode->node.pConditions) {
+          EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
+          QRY_ERR_RET(nodesNodeToSQL(pInterpNode->node.pConditions, tbuf + VARSTR_HEADER_SIZE,
+                                     TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
+          EXPLAIN_ROW_END();
+          QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
+        }
+      }
+      break;
+    }    
     default:
       qError("not supported physical node type %d", pNode->type);
       return TSDB_CODE_QRY_APP_ERROR;

source/libs/executor/inc/executorimpl.h

@@ -151,10 +151,6 @@ typedef struct SExecTaskInfo {
   jmp_buf          env;                  // jump to this position when error happens.
   EOPTR_EXEC_MODEL execModel;            // operator execution model [batch model|stream model]
   struct SOperatorInfo* pRoot;
-
-  /* XXXXXXXXXXXXXXXXXXXX */
-  SReadHandle* pHandle;
-  /* XXXXXXXXXXXXXXXXXXXX */
 } SExecTaskInfo;
 
 enum {

source/libs/executor/src/executorMain.c

@@ -139,29 +139,8 @@ int32_t qExecTask(qTaskInfo_t tinfo, SSDataBlock** pRes, uint64_t* useconds) {
   qDebug("%s execTask is launched", GET_TASKID(pTaskInfo));
 
   int64_t st = taosGetTimestampUs();  
-  /* XXXXXXXXXXXXXXXXXXXX */
-  if (pTaskInfo->pHandle->deleteQuery) {
-    static int32_t first = 1;
-    if (first) {
-      *pRes = createDataBlock();
-      int64_t rows = 33;
-      SColumnInfoData infoData = createColumnInfoData(TSDB_DATA_TYPE_BIGINT, 8, 1);
-      blockDataAppendColInfo(*pRes, &infoData);
-      blockDataEnsureCapacity(*pRes, 1);
-      (*pRes)->info.rows = 1;
 
-      SColumnInfoData* pCol1 = taosArrayGet((*pRes)->pDataBlock, 0);
-      colDataAppend(pCol1, 0, (char*)&rows, false);
-      first = 0;
-    } else {
-      *pRes = NULL;
-    }
-  } else {
-  /* XXXXXXXXXXXXXXXXXXXX */
   *pRes = pTaskInfo->pRoot->fpSet.getNextFn(pTaskInfo->pRoot);
-  /* XXXXXXXXXXXXXXXXXXXX */
-  }
-  /* XXXXXXXXXXXXXXXXXXXX */
   uint64_t el = (taosGetTimestampUs() - st);
 
   pTaskInfo->cost.elapsedTime += el;

source/libs/executor/src/executorimpl.c

@@ -13,8 +13,6 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include <executorimpl.h>
-#include <vnode.h>
 #include "filter.h"
 #include "function.h"
 #include "functionMgt.h"
@@ -2845,11 +2843,18 @@ int32_t getTableScanInfo(SOperatorInfo* pOperator, int32_t* order, int32_t* scan
 
 int32_t doPrepareScan(SOperatorInfo* pOperator, uint64_t uid, int64_t ts) {
   int32_t type = pOperator->operatorType;
+
+  pOperator->status = OP_OPENED;
+
   if (type == QUERY_NODE_PHYSICAL_PLAN_STREAM_SCAN) {
     SStreamBlockScanInfo* pScanInfo = pOperator->info;
     pScanInfo->blockType = STREAM_INPUT__DATA_SCAN;
 
+    pScanInfo->pSnapshotReadOp->status = OP_OPENED;
+
     STableScanInfo* pInfo = pScanInfo->pSnapshotReadOp->info;
+    ASSERT(pInfo->scanMode == TABLE_SCAN__TABLE_ORDER);
+
     if (uid == 0) {
       pInfo->noTable = 1;
       return TSDB_CODE_SUCCESS;
@@ -2863,14 +2868,6 @@ int32_t doPrepareScan(SOperatorInfo* pOperator, uint64_t uid, int64_t ts) {
     pInfo->noTable = 0;
 
     if (pInfo->lastStatus.uid != uid || pInfo->lastStatus.ts != ts) {
-      tsdbSetTableId(pInfo->dataReader, uid);
-      int64_t oldSkey = pInfo->cond.twindows[0].skey;
-      pInfo->cond.twindows[0].skey = ts + 1;
-      tsdbResetReadHandle(pInfo->dataReader, &pInfo->cond, 0);
-      pInfo->cond.twindows[0].skey = oldSkey;
-      pInfo->scanTimes = 0;
-      pInfo->curTWinIdx = 0;
-
       SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
 
       int32_t tableSz = taosArrayGetSize(pTaskInfo->tableqinfoList.pTableList);
@@ -2882,8 +2879,17 @@ int32_t doPrepareScan(SOperatorInfo* pOperator, uint64_t uid, int64_t ts) {
           pInfo->currentTable = i;
         }
       }
-      // TODO after processing drop,
+      // TODO after processing drop, found can be false
       ASSERT(found);
+
+      tsdbSetTableId(pInfo->dataReader, uid);
+      int64_t oldSkey = pInfo->cond.twindows[0].skey;
+      pInfo->cond.twindows[0].skey = ts + 1;
+      tsdbReaderReset(pInfo->dataReader, &pInfo->cond, 0);
+      pInfo->cond.twindows[0].skey = oldSkey;
+      pInfo->scanTimes = 0;
+      pInfo->curTWinIdx = 0;
+
       qDebug("tsdb reader offset seek to uid %ld ts %ld, table cur set to %d , all table num %d", uid, ts,
              pInfo->currentTable, tableSz);
     }
@@ -4764,9 +4770,6 @@ int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SRead
                                            &(*pTaskInfo)->tableqinfoList, pPlan->user);
 
 
-  /* XXXXXXXXXXXXXXXXXXXX */
-  (*pTaskInfo)->pHandle = pHandle;
-  /* XXXXXXXXXXXXXXXXXXXX */
   if (NULL == (*pTaskInfo)->pRoot) {
     code = (*pTaskInfo)->code;
     goto _complete;

source/libs/executor/src/scanoperator.c

@@ -440,7 +440,7 @@ static SSDataBlock* doTableScanGroup(SOperatorInfo* pOperator) {
       }
       pTableScanInfo->curTWinIdx += 1;
       if (pTableScanInfo->curTWinIdx < pTableScanInfo->cond.numOfTWindows) {
-        tsdbResetReadHandle(pTableScanInfo->dataReader, &pTableScanInfo->cond, pTableScanInfo->curTWinIdx);
+        tsdbReaderReset(pTableScanInfo->dataReader, &pTableScanInfo->cond, pTableScanInfo->curTWinIdx);
       }
     }
 
@@ -455,7 +455,7 @@ static SSDataBlock* doTableScanGroup(SOperatorInfo* pOperator) {
         qDebug("%s qrange:%" PRId64 "-%" PRId64, GET_TASKID(pTaskInfo), pWin->skey, pWin->ekey);
       }
       // do prepare for the next round table scan operation
-      tsdbResetReadHandle(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
+      tsdbReaderReset(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
       pTableScanInfo->curTWinIdx = 0;
     }
   }
@@ -464,7 +464,7 @@ static SSDataBlock* doTableScanGroup(SOperatorInfo* pOperator) {
   if (pTableScanInfo->scanTimes < total) {
     if (pTableScanInfo->cond.order == TSDB_ORDER_ASC) {
       prepareForDescendingScan(pTableScanInfo, pTableScanInfo->pCtx, 0);
-      tsdbResetReadHandle(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
+      tsdbReaderReset(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
       pTableScanInfo->curTWinIdx = 0;
     }
 
@@ -482,7 +482,7 @@ static SSDataBlock* doTableScanGroup(SOperatorInfo* pOperator) {
         }
         pTableScanInfo->curTWinIdx += 1;
         if (pTableScanInfo->curTWinIdx < pTableScanInfo->cond.numOfTWindows) {
-          tsdbResetReadHandle(pTableScanInfo->dataReader, &pTableScanInfo->cond, pTableScanInfo->curTWinIdx);
+          tsdbReaderReset(pTableScanInfo->dataReader, &pTableScanInfo->cond, pTableScanInfo->curTWinIdx);
         }
       }
 
@@ -498,7 +498,7 @@ static SSDataBlock* doTableScanGroup(SOperatorInfo* pOperator) {
           STimeWindow* pWin = &pTableScanInfo->cond.twindows[i];
           qDebug("%s qrange:%" PRId64 "-%" PRId64, GET_TASKID(pTaskInfo), pWin->skey, pWin->ekey);
         }
-        tsdbResetReadHandle(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
+        tsdbReaderReset(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
         pTableScanInfo->curTWinIdx = 0;
       }
     }
@@ -526,7 +526,7 @@ static SSDataBlock* doTableScan(SOperatorInfo* pOperator) {
       }
       STableKeyInfo* pTableInfo = taosArrayGet(pTaskInfo->tableqinfoList.pTableList, pInfo->currentTable);
       tsdbSetTableId(pInfo->dataReader, pTableInfo->uid);
-      tsdbResetReadHandle(pInfo->dataReader, &pInfo->cond, 0);
+      tsdbReaderReset(pInfo->dataReader, &pInfo->cond, 0);
       pInfo->scanTimes = 0;
       pInfo->curTWinIdx = 0;
     }
@@ -560,7 +560,7 @@ static SSDataBlock* doTableScan(SOperatorInfo* pOperator) {
   SArray* tableList = taosArrayGetP(pTaskInfo->tableqinfoList.pGroupList, pInfo->currentGroupId);
   //  tsdbSetTableList(pInfo->dataReader, tableList);
 
-  tsdbResetReadHandle(pInfo->dataReader, &pInfo->cond, 0);
+  tsdbReaderReset(pInfo->dataReader, &pInfo->cond, 0);
   pInfo->curTWinIdx = 0;
   pInfo->scanTimes = 0;
 
@@ -859,7 +859,7 @@ static bool prepareDataScan(SStreamBlockScanInfo* pInfo, SSDataBlock* pSDB, int3
   STableScanInfo* pTableScanInfo = pInfo->pSnapshotReadOp->info;
   pTableScanInfo->cond.twindows[0] = win;
   pTableScanInfo->curTWinIdx = 0;
-  //  tsdbResetReadHandle(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
+  //  tsdbReaderReset(pTableScanInfo->dataReader, &pTableScanInfo->cond, 0);
   // if (!pTableScanInfo->dataReader) {
   //   return false;
   // }

source/libs/function/src/builtins.c

@@ -18,6 +18,7 @@
 #include "querynodes.h"
 #include "scalar.h"
 #include "taoserror.h"
+#include "cJSON.h"
 
 static int32_t buildFuncErrMsg(char* pErrBuf, int32_t len, int32_t errCode, const char* pFormat, ...) {
   va_list vArgList;
@@ -796,6 +797,165 @@ static int32_t translateLeastSQR(SFunctionNode* pFunc, char* pErrBuf, int32_t le
   return TSDB_CODE_SUCCESS;
 }
 
+typedef enum { UNKNOWN_BIN = 0, USER_INPUT_BIN, LINEAR_BIN, LOG_BIN } EHistoBinType;
+
+static int8_t validateHistogramBinType(char* binTypeStr) {
+  int8_t binType;
+  if (strcasecmp(binTypeStr, "user_input") == 0) {
+    binType = USER_INPUT_BIN;
+  } else if (strcasecmp(binTypeStr, "linear_bin") == 0) {
+    binType = LINEAR_BIN;
+  } else if (strcasecmp(binTypeStr, "log_bin") == 0) {
+    binType = LOG_BIN;
+  } else {
+    binType = UNKNOWN_BIN;
+  }
+
+  return binType;
+}
+
+static bool validateHistogramBinDesc(char* binDescStr, int8_t binType, char* errMsg, int32_t msgLen) {
+  const char *msg1 = "HISTOGRAM function requires four parameters";
+  const char *msg3 = "HISTOGRAM function invalid format for binDesc parameter";
+  const char *msg4 = "HISTOGRAM function binDesc parameter \"count\" should be in range [1, 1000]";
+  const char *msg5 = "HISTOGRAM function bin/parameter should be in range [-DBL_MAX, DBL_MAX]";
+  const char *msg6 = "HISTOGRAM function binDesc parameter \"width\" cannot be 0";
+  const char *msg7 = "HISTOGRAM function binDesc parameter \"start\" cannot be 0 with \"log_bin\" type";
+  const char *msg8 = "HISTOGRAM function binDesc parameter \"factor\" cannot be negative or equal to 0/1";
+
+  cJSON*  binDesc = cJSON_Parse(binDescStr);
+  int32_t numOfBins;
+  double* intervals;
+  if (cJSON_IsObject(binDesc)) { /* linaer/log bins */
+    int32_t numOfParams = cJSON_GetArraySize(binDesc);
+    int32_t startIndex;
+    if (numOfParams != 4) {
+      snprintf(errMsg, msgLen, "%s", msg1);
+      return false;
+    }
+
+    cJSON* start = cJSON_GetObjectItem(binDesc, "start");
+    cJSON* factor = cJSON_GetObjectItem(binDesc, "factor");
+    cJSON* width = cJSON_GetObjectItem(binDesc, "width");
+    cJSON* count = cJSON_GetObjectItem(binDesc, "count");
+    cJSON* infinity = cJSON_GetObjectItem(binDesc, "infinity");
+
+    if (!cJSON_IsNumber(start) || !cJSON_IsNumber(count) || !cJSON_IsBool(infinity)) {
+      snprintf(errMsg, msgLen, "%s", msg3);
+      return false;
+    }
+
+    if (count->valueint <= 0 || count->valueint > 1000) {  // limit count to 1000
+      snprintf(errMsg, msgLen, "%s", msg4);
+      return false;
+    }
+
+    if (isinf(start->valuedouble) || (width != NULL && isinf(width->valuedouble)) ||
+        (factor != NULL && isinf(factor->valuedouble)) || (count != NULL && isinf(count->valuedouble))) {
+      snprintf(errMsg, msgLen, "%s", msg5);
+      return false;
+    }
+
+    int32_t counter = (int32_t)count->valueint;
+    if (infinity->valueint == false) {
+      startIndex = 0;
+      numOfBins = counter + 1;
+    } else {
+      startIndex = 1;
+      numOfBins = counter + 3;
+    }
+
+    intervals = taosMemoryCalloc(numOfBins, sizeof(double));
+    if (cJSON_IsNumber(width) && factor == NULL && binType == LINEAR_BIN) {
+      // linear bin process
+      if (width->valuedouble == 0) {
+        snprintf(errMsg, msgLen, "%s", msg6);
+        taosMemoryFree(intervals);
+        return false;
+      }
+      for (int i = 0; i < counter + 1; ++i) {
+        intervals[startIndex] = start->valuedouble + i * width->valuedouble;
+        if (isinf(intervals[startIndex])) {
+          snprintf(errMsg, msgLen, "%s", msg5);
+          taosMemoryFree(intervals);
+          return false;
+        }
+        startIndex++;
+      }
+    } else if (cJSON_IsNumber(factor) && width == NULL && binType == LOG_BIN) {
+      // log bin process
+      if (start->valuedouble == 0) {
+        snprintf(errMsg, msgLen, "%s", msg7);
+        taosMemoryFree(intervals);
+        return false;
+      }
+      if (factor->valuedouble < 0 || factor->valuedouble == 0 || factor->valuedouble == 1) {
+        snprintf(errMsg, msgLen, "%s", msg8);
+        taosMemoryFree(intervals);
+        return false;
+      }
+      for (int i = 0; i < counter + 1; ++i) {
+        intervals[startIndex] = start->valuedouble * pow(factor->valuedouble, i * 1.0);
+        if (isinf(intervals[startIndex])) {
+          snprintf(errMsg, msgLen, "%s", msg5);
+          taosMemoryFree(intervals);
+          return false;
+        }
+        startIndex++;
+      }
+    } else {
+      snprintf(errMsg, msgLen, "%s", msg3);
+      taosMemoryFree(intervals);
+      return false;
+    }
+
+    if (infinity->valueint == true) {
+      intervals[0] = -INFINITY;
+      intervals[numOfBins - 1] = INFINITY;
+      // in case of desc bin orders, -inf/inf should be swapped
+      ASSERT(numOfBins >= 4);
+      if (intervals[1] > intervals[numOfBins - 2]) {
+        TSWAP(intervals[0], intervals[numOfBins - 1]);
+      }
+    }
+  } else if (cJSON_IsArray(binDesc)) { /* user input bins */
+    if (binType != USER_INPUT_BIN) {
+      snprintf(errMsg, msgLen, "%s", msg3);
+      return false;
+    }
+    numOfBins = cJSON_GetArraySize(binDesc);
+    intervals = taosMemoryCalloc(numOfBins, sizeof(double));
+    cJSON* bin = binDesc->child;
+    if (bin == NULL) {
+      snprintf(errMsg, msgLen, "%s", msg3);
+      taosMemoryFree(intervals);
+      return false;
+    }
+    int i = 0;
+    while (bin) {
+      intervals[i] = bin->valuedouble;
+      if (!cJSON_IsNumber(bin)) {
+        snprintf(errMsg, msgLen, "%s", msg3);
+        taosMemoryFree(intervals);
+        return false;
+      }
+      if (i != 0 && intervals[i] <= intervals[i - 1]) {
+        snprintf(errMsg, msgLen, "%s", msg3);
+        taosMemoryFree(intervals);
+        return false;
+      }
+      bin = bin->next;
+      i++;
+    }
+  } else {
+    snprintf(errMsg, msgLen, "%s", msg3);
+    return false;
+  }
+
+  taosMemoryFree(intervals);
+  return true;
+}
+
 static int32_t translateHistogram(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
   int32_t numOfParams = LIST_LENGTH(pFunc->pParameterList);
   if (4 != numOfParams) {
@@ -814,6 +974,8 @@ static int32_t translateHistogram(SFunctionNode* pFunc, char* pErrBuf, int32_t l
     return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
+  int8_t binType;
+  char*  binDesc;
   for (int32_t i = 1; i < numOfParams; ++i) {
     SNode* pParamNode = nodesListGetNode(pFunc->pParameterList, i);
     if (QUERY_NODE_VALUE != nodeType(pParamNode)) {
@@ -824,6 +986,23 @@ static int32_t translateHistogram(SFunctionNode* pFunc, char* pErrBuf, int32_t l
 
     pValue->notReserved = true;
 
+    if (i == 1) {
+      binType = validateHistogramBinType(varDataVal(pValue->datum.p));
+      if (binType == UNKNOWN_BIN) {
+        return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
+                               "HISTOGRAM function binType parameter should be "
+                               "\"user_input\", \"log_bin\" or \"linear_bin\"");
+      }
+    }
+
+    if (i == 2) {
+      char errMsg[128] = {0};
+      binDesc = varDataVal(pValue->datum.p);
+      if (!validateHistogramBinDesc(binDesc, binType, errMsg, (int32_t)sizeof(errMsg))) {
+        return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR, errMsg);
+      }
+    }
+
     if (i == 3 && pValue->datum.i != 1 && pValue->datum.i != 0) {
       return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
                              "HISTOGRAM function normalized parameter should be 0/1");
@@ -853,6 +1032,8 @@ static int32_t translateHistogramImpl(SFunctionNode* pFunc, char* pErrBuf, int32
       return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
     }
 
+    int8_t binType;
+    char*  binDesc;
     for (int32_t i = 1; i < numOfParams; ++i) {
       SNode* pParamNode = nodesListGetNode(pFunc->pParameterList, i);
       if (QUERY_NODE_VALUE != nodeType(pParamNode)) {
@@ -863,6 +1044,23 @@ static int32_t translateHistogramImpl(SFunctionNode* pFunc, char* pErrBuf, int32
 
       pValue->notReserved = true;
 
+      if (i == 1) {
+        binType = validateHistogramBinType(varDataVal(pValue->datum.p));
+        if (binType == UNKNOWN_BIN) {
+          return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
+                                 "HISTOGRAM function binType parameter should be "
+                                 "\"user_input\", \"log_bin\" or \"linear_bin\"");
+        }
+      }
+
+      if (i == 2) {
+        char errMsg[128] = {0};
+        binDesc = varDataVal(pValue->datum.p);
+        if (!validateHistogramBinDesc(binDesc, binType, errMsg, (int32_t)sizeof(errMsg))) {
+          return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR, errMsg);
+        }
+      }
+
       if (i == 3 && pValue->datum.i != 1 && pValue->datum.i != 0) {
         return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
                                "HISTOGRAM function normalized parameter should be 0/1");
@@ -2105,7 +2303,7 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
   {
     .name = "histogram",
     .type = FUNCTION_TYPE_HISTOGRAM,
-    .classification = FUNC_MGT_AGG_FUNC | FUNC_MGT_FORBID_FILL_FUNC,
+    .classification = FUNC_MGT_AGG_FUNC | FUNC_MGT_INDEFINITE_ROWS_FUNC | FUNC_MGT_FORBID_FILL_FUNC,
     .translateFunc = translateHistogram,
     .getEnvFunc   = getHistogramFuncEnv,
     .initFunc     = histogramFunctionSetup,

source/libs/function/src/functionMgt.c

@@ -260,7 +260,7 @@ bool fmIsSameInOutType(int32_t funcId) {
 }
 
 static int32_t getFuncInfo(SFunctionNode* pFunc) {
-  char msg[64] = {0};
+  char msg[128] = {0};
   return fmGetFuncInfo(pFunc, msg, sizeof(msg));
 }
 

source/libs/nodes/src/nodesUtilFuncs.c

@@ -19,8 +19,8 @@
 #include "querynodes.h"
 #include "taos.h"
 #include "taoserror.h"
-#include "thash.h"
 #include "tdatablock.h"
+#include "thash.h"
 
 static SNode* makeNode(ENodeType type, size_t size) {
   SNode* p = taosMemoryCalloc(1, size);
@@ -1501,13 +1501,18 @@ typedef struct SCollectFuncsCxt {
   int32_t         errCode;
   FFuncClassifier classifier;
   SNodeList*      pFuncs;
+  SHashObj*       pAliasName;
 } SCollectFuncsCxt;
 
 static EDealRes collectFuncs(SNode* pNode, void* pContext) {
   SCollectFuncsCxt* pCxt = (SCollectFuncsCxt*)pContext;
   if (QUERY_NODE_FUNCTION == nodeType(pNode) && pCxt->classifier(((SFunctionNode*)pNode)->funcId) &&
       !(((SExprNode*)pNode)->orderAlias)) {
+    SExprNode* pExpr = (SExprNode*)pNode;
+    if (NULL == taosHashGet(pCxt->pAliasName, pExpr->aliasName, strlen(pExpr->aliasName))) {
       pCxt->errCode = nodesListStrictAppend(pCxt->pFuncs, nodesCloneNode(pNode));
+      taosHashPut(pCxt->pAliasName, pExpr->aliasName, strlen(pExpr->aliasName), &pExpr, POINTER_BYTES);
+    }
     return (TSDB_CODE_SUCCESS == pCxt->errCode ? DEAL_RES_IGNORE_CHILD : DEAL_RES_ERROR);
   }
   return DEAL_RES_CONTINUE;
@@ -1519,23 +1524,27 @@ int32_t nodesCollectFuncs(SSelectStmt* pSelect, ESqlClause clause, FFuncClassifi
   }
 
   SCollectFuncsCxt cxt = {
-      .errCode = TSDB_CODE_SUCCESS, .classifier = classifier, .pFuncs = (NULL == *pFuncs ? nodesMakeList() : *pFuncs)};
+      .errCode = TSDB_CODE_SUCCESS,
+      .classifier = classifier,
+      .pFuncs = (NULL == *pFuncs ? nodesMakeList() : *pFuncs),
+      .pAliasName = taosHashInit(4, taosGetDefaultHashFunction(TSDB_DATA_TYPE_VARCHAR), false, false)};
   if (NULL == cxt.pFuncs) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
   *pFuncs = NULL;
   nodesWalkSelectStmt(pSelect, clause, collectFuncs, &cxt);
-  if (TSDB_CODE_SUCCESS != cxt.errCode) {
+  if (TSDB_CODE_SUCCESS == cxt.errCode) {
-    nodesDestroyList(cxt.pFuncs);
-    return cxt.errCode;
-  }
     if (LIST_LENGTH(cxt.pFuncs) > 0) {
       *pFuncs = cxt.pFuncs;
     } else {
       nodesDestroyList(cxt.pFuncs);
     }
+  } else {
+    nodesDestroyList(cxt.pFuncs);
+  }
+  taosHashCleanup(cxt.pAliasName);
 
-  return TSDB_CODE_SUCCESS;
+  return cxt.errCode;
 }
 
 typedef struct SCollectSpecialNodesCxt {

source/libs/parser/inc/parAst.h

@@ -155,6 +155,7 @@ SNode* createAlterTableDropCol(SAstCreateContext* pCxt, SNode* pRealTable, int8_
 SNode* createAlterTableRenameCol(SAstCreateContext* pCxt, SNode* pRealTable, int8_t alterType, SToken* pOldColName,
                                  SToken* pNewColName);
 SNode* createAlterTableSetTag(SAstCreateContext* pCxt, SNode* pRealTable, SToken* pTagName, SNode* pVal);
+SNode* setAlterSuperTableType(SNode* pStmt);
 SNode* createUseDatabaseStmt(SAstCreateContext* pCxt, SToken* pDbName);
 SNode* createShowStmt(SAstCreateContext* pCxt, ENodeType type);
 SNode* createShowStmtWithCond(SAstCreateContext* pCxt, ENodeType type, SNode* pDbName, SNode* pTbName,

source/libs/parser/inc/parUtil.h

@@ -53,6 +53,7 @@ typedef struct SParseMetaCache {
 } SParseMetaCache;
 
 int32_t generateSyntaxErrMsg(SMsgBuf* pBuf, int32_t errCode, ...);
+int32_t generateSyntaxErrMsgExt(SMsgBuf* pBuf, int32_t errCode, const char* pFormat, ...);
 int32_t buildInvalidOperationMsg(SMsgBuf* pMsgBuf, const char* msg);
 int32_t buildSyntaxErrMsg(SMsgBuf* pBuf, const char* additionalInfo, const char* sourceStr);
 

source/libs/parser/inc/sql.y

@@ -233,7 +233,7 @@ cmd ::= DROP TABLE multi_drop_clause(A).
 cmd ::= DROP STABLE exists_opt(A) full_table_name(B).                             { pCxt->pRootNode = createDropSuperTableStmt(pCxt, A, B); }
 
 cmd ::= ALTER TABLE alter_table_clause(A).                                        { pCxt->pRootNode = A; }
-cmd ::= ALTER STABLE alter_table_clause(A).                                       { pCxt->pRootNode = A; }
+cmd ::= ALTER STABLE alter_table_clause(A).                                       { pCxt->pRootNode = setAlterSuperTableType(A); }
 
 alter_table_clause(A) ::= full_table_name(B) alter_table_options(C).              { A = createAlterTableModifyOptions(pCxt, B, C); }
 alter_table_clause(A) ::=
@@ -260,7 +260,7 @@ multi_create_clause(A) ::= multi_create_clause(B) create_subtable_clause(C).
 
 create_subtable_clause(A) ::=
   not_exists_opt(B) full_table_name(C) USING full_table_name(D)
-  specific_tags_opt(E) TAGS NK_LP literal_list(F) NK_RP table_options(G).         { A = createCreateSubTableClause(pCxt, B, C, D, E, F, G); }
+  specific_tags_opt(E) TAGS NK_LP expression_list(F) NK_RP table_options(G).      { A = createCreateSubTableClause(pCxt, B, C, D, E, F, G); }
 
 %type multi_drop_clause                                                           { SNodeList* }
 %destructor multi_drop_clause                                                     { nodesDestroyList($$); }

source/libs/parser/src/parAstCreater.c

@@ -1138,6 +1138,11 @@ SNode* createAlterTableSetTag(SAstCreateContext* pCxt, SNode* pRealTable, SToken
   return createAlterTableStmtFinalize(pRealTable, pStmt);
 }
 
+SNode* setAlterSuperTableType(SNode* pStmt) {
+  setNodeType(pStmt, QUERY_NODE_ALTER_SUPER_TABLE_STMT);
+  return pStmt;
+}
+
 SNode* createUseDatabaseStmt(SAstCreateContext* pCxt, SToken* pDbName) {
   CHECK_PARSER_STATUS(pCxt);
   if (!checkDbName(pCxt, pDbName, false)) {

source/libs/parser/src/parAstParser.c

@@ -251,6 +251,10 @@ static int32_t collectMetaKeyFromAlterTable(SCollectMetaKeyCxt* pCxt, SAlterTabl
   return code;
 }
 
+static int32_t collectMetaKeyFromAlterStable(SCollectMetaKeyCxt* pCxt, SAlterTableStmt* pStmt) {
+  return reserveTableMetaInCache(pCxt->pParseCxt->acctId, pStmt->dbName, pStmt->tableName, pCxt->pMetaCache);
+}
+
 static int32_t collectMetaKeyFromUseDatabase(SCollectMetaKeyCxt* pCxt, SUseDatabaseStmt* pStmt) {
   return reserveDbVgVersionInCache(pCxt->pParseCxt->acctId, pStmt->dbName, pCxt->pMetaCache);
 }
@@ -489,6 +493,8 @@ static int32_t collectMetaKeyFromQuery(SCollectMetaKeyCxt* pCxt, SNode* pStmt) {
       return collectMetaKeyFromDropTable(pCxt, (SDropTableStmt*)pStmt);
     case QUERY_NODE_ALTER_TABLE_STMT:
       return collectMetaKeyFromAlterTable(pCxt, (SAlterTableStmt*)pStmt);
+    case QUERY_NODE_ALTER_SUPER_TABLE_STMT:
+      return collectMetaKeyFromAlterStable(pCxt, (SAlterTableStmt*)pStmt);
     case QUERY_NODE_USE_DATABASE_STMT:
       return collectMetaKeyFromUseDatabase(pCxt, (SUseDatabaseStmt*)pStmt);
     case QUERY_NODE_CREATE_INDEX_STMT:

source/libs/parser/src/parInsert.c

@@ -48,6 +48,12 @@
     pSql += sToken.n;                         \
   } while (TK_NK_SPACE == sToken.type)
 
+typedef struct SInsertParseBaseContext {
+  SParseContext* pComCxt;
+  char*          pSql;
+  SMsgBuf        msg;
+} SInsertParseBaseContext;
+
 typedef struct SInsertParseContext {
   SParseContext*     pComCxt;             // input
   char*              pSql;                // input
@@ -1105,6 +1111,32 @@ static int32_t storeTableMeta(SInsertParseContext* pCxt, SHashObj* pHash, SName*
   return taosHashPut(pHash, pName, len, &pBackup, POINTER_BYTES);
 }
 
+static int32_t skipParentheses(SInsertParseSyntaxCxt* pCxt) {
+  SToken  sToken;
+  int32_t expectRightParenthesis = 1;
+  while (1) {
+    NEXT_TOKEN(pCxt->pSql, sToken);
+    if (TK_NK_LP == sToken.type) {
+      ++expectRightParenthesis;
+    } else if (TK_NK_RP == sToken.type && 0 == --expectRightParenthesis) {
+      break;
+    }
+    if (0 == sToken.n) {
+      return buildSyntaxErrMsg(&pCxt->msg, ") expected", NULL);
+    }
+  }
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t skipBoundColumns(SInsertParseSyntaxCxt* pCxt) { return skipParentheses(pCxt); }
+
+static int32_t ignoreBoundColumns(SInsertParseContext* pCxt) {
+  SInsertParseSyntaxCxt cxt = {.pComCxt = pCxt->pComCxt, .pSql = pCxt->pSql, .msg = pCxt->msg, .pMetaCache = NULL};
+  int32_t               code = skipBoundColumns(&cxt);
+  pCxt->pSql = cxt.pSql;
+  return code;
+}
+
 static int32_t skipUsingClause(SInsertParseSyntaxCxt* pCxt);
 
 // pSql -> stb_name [(tag1_name, ...)] TAGS (tag1_value, ...)
@@ -1453,12 +1485,29 @@ static int32_t parseInsertBody(SInsertParseContext* pCxt) {
     tNameGetFullDbName(&name, dbFName);
     CHECK_CODE(taosHashPut(pCxt->pDbFNameHashObj, dbFName, strlen(dbFName), dbFName, sizeof(dbFName)));
 
+    bool existedUsing = false;
     // USING clause
+    if (TK_USING == sToken.type) {
+      existedUsing = true;
+      CHECK_CODE(parseUsingClause(pCxt, &name, tbFName));
+      NEXT_TOKEN(pCxt->pSql, sToken);
+      autoCreateTbl = true;
+    }
+
+    char* pBoundColsStart = NULL;
+    if (TK_NK_LP == sToken.type) {
+      // pSql -> field1_name, ...)
+      pBoundColsStart = pCxt->pSql;
+      CHECK_CODE(ignoreBoundColumns(pCxt));
+      // CHECK_CODE(parseBoundColumns(pCxt, &dataBuf->boundColumnInfo, getTableColumnSchema(pCxt->pTableMeta)));
+      NEXT_TOKEN(pCxt->pSql, sToken);
+    }
+
     if (TK_USING == sToken.type) {
       CHECK_CODE(parseUsingClause(pCxt, &name, tbFName));
       NEXT_TOKEN(pCxt->pSql, sToken);
       autoCreateTbl = true;
-    } else {
+    } else if (!existedUsing) {
       CHECK_CODE(getTableMeta(pCxt, &name, dbFName));
     }
 
@@ -1467,10 +1516,11 @@ static int32_t parseInsertBody(SInsertParseContext* pCxt) {
                                     sizeof(SSubmitBlk), getTableInfo(pCxt->pTableMeta).rowSize, pCxt->pTableMeta,
                                     &dataBuf, NULL, &pCxt->createTblReq));
 
-    if (TK_NK_LP == sToken.type) {
+    if (NULL != pBoundColsStart) {
-      // pSql -> field1_name, ...)
+      char* pCurrPos = pCxt->pSql;
+      pCxt->pSql = pBoundColsStart;
       CHECK_CODE(parseBoundColumns(pCxt, &dataBuf->boundColumnInfo, getTableColumnSchema(pCxt->pTableMeta)));
-      NEXT_TOKEN(pCxt->pSql, sToken);
+      pCxt->pSql = pCurrPos;
     }
 
     if (TK_VALUES == sToken.type) {
@@ -1610,25 +1660,6 @@ int32_t parseInsertSql(SParseContext* pContext, SQuery** pQuery, SParseMetaCache
   return code;
 }
 
-static int32_t skipParentheses(SInsertParseSyntaxCxt* pCxt) {
-  SToken  sToken;
-  int32_t expectRightParenthesis = 1;
-  while (1) {
-    NEXT_TOKEN(pCxt->pSql, sToken);
-    if (TK_NK_LP == sToken.type) {
-      ++expectRightParenthesis;
-    } else if (TK_NK_RP == sToken.type && 0 == --expectRightParenthesis) {
-      break;
-    }
-    if (0 == sToken.n) {
-      return buildSyntaxErrMsg(&pCxt->msg, ") expected", NULL);
-    }
-  }
-  return TSDB_CODE_SUCCESS;
-}
-
-static int32_t skipBoundColumns(SInsertParseSyntaxCxt* pCxt) { return skipParentheses(pCxt); }
-
 // pSql -> (field1_value, ...) [(field1_value2, ...) ...]
 static int32_t skipValuesClause(SInsertParseSyntaxCxt* pCxt) {
   int32_t numOfRows = 0;
@@ -1717,8 +1748,25 @@ static int32_t parseInsertBodySyntax(SInsertParseSyntaxCxt* pCxt) {
     SToken tbnameToken = sToken;
     NEXT_TOKEN(pCxt->pSql, sToken);
 
+    bool existedUsing = false;
     // USING clause
     if (TK_USING == sToken.type) {
+      existedUsing = true;
+      CHECK_CODE(collectAutoCreateTableMetaKey(pCxt, &tbnameToken));
+      NEXT_TOKEN(pCxt->pSql, sToken);
+      CHECK_CODE(collectTableMetaKey(pCxt, &sToken));
+      CHECK_CODE(skipUsingClause(pCxt));
+      NEXT_TOKEN(pCxt->pSql, sToken);
+    }
+
+    if (TK_NK_LP == sToken.type) {
+      // pSql -> field1_name, ...)
+      CHECK_CODE(skipBoundColumns(pCxt));
+      NEXT_TOKEN(pCxt->pSql, sToken);
+    }
+
+    if (TK_USING == sToken.type && !existedUsing) {
+      existedUsing = true;
       CHECK_CODE(collectAutoCreateTableMetaKey(pCxt, &tbnameToken));
       NEXT_TOKEN(pCxt->pSql, sToken);
       CHECK_CODE(collectTableMetaKey(pCxt, &sToken));
@@ -1728,12 +1776,6 @@ static int32_t parseInsertBodySyntax(SInsertParseSyntaxCxt* pCxt) {
       CHECK_CODE(collectTableMetaKey(pCxt, &tbnameToken));
     }
 
-    if (TK_NK_LP == sToken.type) {
-      // pSql -> field1_name, ...)
-      CHECK_CODE(skipBoundColumns(pCxt));
-      NEXT_TOKEN(pCxt->pSql, sToken);
-    }
-
     if (TK_VALUES == sToken.type) {
       // pSql -> (field1_value, ...) [(field1_value2, ...) ...]
       CHECK_CODE(skipValuesClause(pCxt));

source/libs/parser/src/parTranslater.c

@@ -121,8 +121,8 @@ static int32_t getTableMetaImpl(STranslateContext* pCxt, const SName* pName, STa
     }
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("catalogGetTableMeta error, code:%s, dbName:%s, tbName:%s", tstrerror(code), pName->dbname,
+    parserError("0x%" PRIx64 " catalogGetTableMeta error, code:%s, dbName:%s, tbName:%s", pCxt->pParseCxt->requestId,
-                pName->tname);
+                tstrerror(code), pName->dbname, pName->tname);
   }
   return code;
 }
@@ -150,8 +150,8 @@ static int32_t getTableCfg(STranslateContext* pCxt, const SName* pName, STableCf
     }
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("catalogRefreshGetTableCfg error, code:%s, dbName:%s, tbName:%s", tstrerror(code), pName->dbname,
+    parserError("0x%" PRIx64 " catalogRefreshGetTableCfg error, code:%s, dbName:%s, tbName:%s",
-                pName->tname);
+                pCxt->pParseCxt->requestId, tstrerror(code), pName->dbname, pName->tname);
   }
   return code;
 }
@@ -173,8 +173,8 @@ static int32_t refreshGetTableMeta(STranslateContext* pCxt, const char* pDbName,
     code = catalogRefreshGetTableMeta(pParCxt->pCatalog, &conn, &name, pMeta, false);
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("catalogRefreshGetTableMeta error, code:%s, dbName:%s, tbName:%s", tstrerror(code), pDbName,
+    parserError("0x%" PRIx64 " catalogRefreshGetTableMeta error, code:%s, dbName:%s, tbName:%s",
-                pTableName);
+                pCxt->pParseCxt->requestId, tstrerror(code), pDbName, pTableName);
   }
   return code;
 }
@@ -196,7 +196,8 @@ static int32_t getDBVgInfoImpl(STranslateContext* pCxt, const SName* pName, SArr
     }
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("catalogGetDBVgInfo error, code:%s, dbFName:%s", tstrerror(code), fullDbName);
+    parserError("0x%" PRIx64 " catalogGetDBVgInfo error, code:%s, dbFName:%s", pCxt->pParseCxt->requestId,
+                tstrerror(code), fullDbName);
   }
   return code;
 }
@@ -227,8 +228,8 @@ static int32_t getTableHashVgroupImpl(STranslateContext* pCxt, const SName* pNam
     }
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("catalogGetTableHashVgroup error, code:%s, dbName:%s, tbName:%s", tstrerror(code), pName->dbname,
+    parserError("0x%" PRIx64 " catalogGetTableHashVgroup error, code:%s, dbName:%s, tbName:%s",
-                pName->tname);
+                pCxt->pParseCxt->requestId, tstrerror(code), pName->dbname, pName->tname);
   }
   return code;
 }
@@ -251,7 +252,8 @@ static int32_t getDBVgVersion(STranslateContext* pCxt, const char* pDbFName, int
     }
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("catalogGetDBVgVersion error, code:%s, dbFName:%s", tstrerror(code), pDbFName);
+    parserError("0x%" PRIx64 " catalogGetDBVgVersion error, code:%s, dbFName:%s", pCxt->pParseCxt->requestId,
+                tstrerror(code), pDbFName);
   }
   return code;
 }
@@ -276,7 +278,8 @@ static int32_t getDBCfg(STranslateContext* pCxt, const char* pDbName, SDbCfgInfo
     }
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("catalogGetDBCfg error, code:%s, dbFName:%s", tstrerror(code), dbFname);
+    parserError("0x%" PRIx64 " catalogGetDBCfg error, code:%s, dbFName:%s", pCxt->pParseCxt->requestId, tstrerror(code),
+                dbFname);
   }
   return code;
 }
@@ -303,6 +306,10 @@ static int32_t getUdfInfo(STranslateContext* pCxt, SFunctionNode* pFunc) {
     pFunc->udfBufSize = funcInfo.bufSize;
     tFreeSFuncInfo(&funcInfo);
   }
+  if (TSDB_CODE_SUCCESS != code) {
+    parserError("0x%" PRIx64 " catalogGetUdfInfo error, code:%s, funcName:%s", pCxt->pParseCxt->requestId,
+                tstrerror(code), pFunc->functionName);
+  }
   return code;
 }
 
@@ -323,7 +330,8 @@ static int32_t getTableIndex(STranslateContext* pCxt, const SName* pName, SArray
     code = catalogGetTableIndex(pParCxt->pCatalog, &conn, pName, pIndexes);
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("getTableIndex error, code:%s, dbName:%s, tbName:%s", tstrerror(code), pName->dbname, pName->tname);
+    parserError("0x%" PRIx64 " getTableIndex error, code:%s, dbName:%s, tbName:%s", pCxt->pParseCxt->requestId,
+                tstrerror(code), pName->dbname, pName->tname);
   }
   return code;
 }
@@ -341,7 +349,7 @@ static int32_t getDnodeList(STranslateContext* pCxt, SArray** pDnodes) {
     code = catalogGetDnodeList(pParCxt->pCatalog, &conn, pDnodes);
   }
   if (TSDB_CODE_SUCCESS != code) {
-    parserError("getDnodeList error, code:%s", tstrerror(code));
+    parserError("0x%" PRIx64 " getDnodeList error, code:%s", pCxt->pParseCxt->requestId, tstrerror(code));
   }
   return code;
 }
@@ -707,7 +715,7 @@ static EDealRes translateColumnUseAlias(STranslateContext* pCxt, SColumnNode** p
 }
 
 static EDealRes translateColumn(STranslateContext* pCxt, SColumnNode** pCol) {
-  if (isSelectStmt(pCxt->pCurrStmt) && NULL == ((SSelectStmt*)pCxt->pCurrStmt)->pFromTable) {
+  if (NULL == pCxt->pCurrStmt || isSelectStmt(pCxt->pCurrStmt) && NULL == ((SSelectStmt*)pCxt->pCurrStmt)->pFromTable) {
     return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_INVALID_COLUMN, (*pCol)->colName);
   }
 
@@ -776,29 +784,18 @@ static int32_t parseBoolFromValueNode(STranslateContext* pCxt, SValueNode* pVal)
   }
 }
 
-static EDealRes translateValueImpl(STranslateContext* pCxt, SValueNode* pVal, SDataType targetDt) {
+static EDealRes translateDurationValue(STranslateContext* pCxt, SValueNode* pVal) {
-  uint8_t precision = getPrecisionFromCurrStmt(pCxt->pCurrStmt, targetDt.precision);
+  if (parseNatualDuration(pVal->literal, strlen(pVal->literal), &pVal->datum.i, &pVal->unit,
-  pVal->node.resType.precision = precision;
+                          pVal->node.resType.precision) != TSDB_CODE_SUCCESS) {
-  if (pVal->placeholderNo > 0 || pVal->isNull) {
-    return DEAL_RES_CONTINUE;
-  }
-  if (TSDB_DATA_TYPE_NULL == pVal->node.resType.type) {
-    // TODO
-    // pVal->node.resType = targetDt;
-    pVal->translate = true;
-    pVal->isNull = true;
-    return DEAL_RES_CONTINUE;
-  }
-  if (pVal->isDuration) {
-    if (parseNatualDuration(pVal->literal, strlen(pVal->literal), &pVal->datum.i, &pVal->unit, precision) !=
-        TSDB_CODE_SUCCESS) {
     return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
   }
   *(int64_t*)&pVal->typeData = pVal->datum.i;
-  } else {
+  return DEAL_RES_CONTINUE;
+}
+
+static EDealRes translateNormalValue(STranslateContext* pCxt, SValueNode* pVal, SDataType targetDt, bool strict) {
+  int32_t code = TSDB_CODE_SUCCESS;
   switch (targetDt.type) {
-      case TSDB_DATA_TYPE_NULL:
-        break;
     case TSDB_DATA_TYPE_BOOL:
       if (TSDB_CODE_SUCCESS != parseBoolFromValueNode(pCxt, pVal)) {
         return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
@@ -806,42 +803,66 @@ static EDealRes translateValueImpl(STranslateContext* pCxt, SValueNode* pVal, SD
       *(bool*)&pVal->typeData = pVal->datum.b;
       break;
     case TSDB_DATA_TYPE_TINYINT: {
-        pVal->datum.i = taosStr2Int64(pVal->literal, NULL, 10);
+      code = toInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.i);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_TINYINT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(int8_t*)&pVal->typeData = pVal->datum.i;
       break;
     }
     case TSDB_DATA_TYPE_SMALLINT: {
-        pVal->datum.i = taosStr2Int64(pVal->literal, NULL, 10);
+      code = toInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.i);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_SMALLINT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(int16_t*)&pVal->typeData = pVal->datum.i;
       break;
     }
     case TSDB_DATA_TYPE_INT: {
-        pVal->datum.i = taosStr2Int64(pVal->literal, NULL, 10);
+      code = toInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.i);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_INT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(int32_t*)&pVal->typeData = pVal->datum.i;
       break;
     }
     case TSDB_DATA_TYPE_BIGINT: {
-        pVal->datum.i = taosStr2Int64(pVal->literal, NULL, 10);
+      code = toInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.i);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_BIGINT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(int64_t*)&pVal->typeData = pVal->datum.i;
       break;
     }
     case TSDB_DATA_TYPE_UTINYINT: {
-        pVal->datum.u = taosStr2UInt64(pVal->literal, NULL, 10);
+      code = toUInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.u);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_UTINYINT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(uint8_t*)&pVal->typeData = pVal->datum.u;
       break;
     }
     case TSDB_DATA_TYPE_USMALLINT: {
-        pVal->datum.u = taosStr2UInt64(pVal->literal, NULL, 10);
+      code = toUInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.u);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_USMALLINT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(uint16_t*)&pVal->typeData = pVal->datum.u;
       break;
     }
     case TSDB_DATA_TYPE_UINT: {
-        pVal->datum.u = taosStr2UInt64(pVal->literal, NULL, 10);
+      code = toUInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.u);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_UINT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(uint32_t*)&pVal->typeData = pVal->datum.u;
       break;
     }
     case TSDB_DATA_TYPE_UBIGINT: {
-        pVal->datum.u = taosStr2UInt64(pVal->literal, NULL, 10);
+      code = toUInteger(pVal->literal, strlen(pVal->literal), 10, &pVal->datum.u);
+      if (strict && (TSDB_CODE_SUCCESS != code || !IS_VALID_UBIGINT(pVal->datum.i))) {
+        return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, pVal->literal);
+      }
       *(uint64_t*)&pVal->typeData = pVal->datum.u;
       break;
     }
@@ -893,10 +914,33 @@ static EDealRes translateValueImpl(STranslateContext* pCxt, SValueNode* pVal, SD
     default:
       break;
   }
+  return DEAL_RES_CONTINUE;
+}
+
+static EDealRes translateValueImpl(STranslateContext* pCxt, SValueNode* pVal, SDataType targetDt, bool strict) {
+  if (pVal->placeholderNo > 0 || pVal->isNull) {
+    return DEAL_RES_CONTINUE;
+  }
+
+  if (TSDB_DATA_TYPE_NULL == pVal->node.resType.type) {
+    // TODO
+    // pVal->node.resType = targetDt;
+    pVal->translate = true;
+    pVal->isNull = true;
+    return DEAL_RES_CONTINUE;
+  }
+
+  pVal->node.resType.precision = getPrecisionFromCurrStmt(pCxt->pCurrStmt, targetDt.precision);
+
+  EDealRes res = DEAL_RES_CONTINUE;
+  if (pVal->isDuration) {
+    res = translateDurationValue(pCxt, pVal);
+  } else {
+    res = translateNormalValue(pCxt, pVal, targetDt, strict);
   }
   pVal->node.resType = targetDt;
   pVal->translate = true;
-  return DEAL_RES_CONTINUE;
+  return res;
 }
 
 static int32_t calcTypeBytes(SDataType dt) {
@@ -912,7 +956,7 @@ static int32_t calcTypeBytes(SDataType dt) {
 static EDealRes translateValue(STranslateContext* pCxt, SValueNode* pVal) {
   SDataType dt = pVal->node.resType;
   dt.bytes = calcTypeBytes(dt);
-  return translateValueImpl(pCxt, pVal, dt);
+  return translateValueImpl(pCxt, pVal, dt, false);
 }
 
 static bool isMultiResFunc(SNode* pNode) {
@@ -1248,6 +1292,22 @@ static int32_t translateForbidGroupByFunc(STranslateContext* pCxt, SFunctionNode
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t translateRepeatScanFunc(STranslateContext* pCxt, SFunctionNode* pFunc) {
+  if (!fmIsRepeatScanFunc(pFunc->funcId)) {
+    return TSDB_CODE_SUCCESS;
+  }
+  if (isSelectStmt(pCxt->pCurrStmt) && NULL != ((SSelectStmt*)pCxt->pCurrStmt)->pFromTable) {
+    SNode* pTable = ((SSelectStmt*)pCxt->pCurrStmt)->pFromTable;
+    if (QUERY_NODE_REAL_TABLE == nodeType(pTable) &&
+        (TSDB_CHILD_TABLE == ((SRealTableNode*)pTable)->pMeta->tableType ||
+         TSDB_NORMAL_TABLE == ((SRealTableNode*)pTable)->pMeta->tableType)) {
+      return TSDB_CODE_SUCCESS;
+    }
+  }
+  return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_ONLY_SUPPORT_SINGLE_TABLE,
+                                 "%s is only supported in single table query", pFunc->functionName);
+}
+
 static void setFuncClassification(SNode* pCurrStmt, SFunctionNode* pFunc) {
   if (NULL != pCurrStmt && QUERY_NODE_SELECT_STMT == nodeType(pCurrStmt)) {
     SSelectStmt* pSelect = (SSelectStmt*)pCurrStmt;
@@ -1370,6 +1430,9 @@ static int32_t translateNoramlFunction(STranslateContext* pCxt, SFunctionNode* p
   if (TSDB_CODE_SUCCESS == code) {
     code = translateForbidGroupByFunc(pCxt, pFunc);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = translateRepeatScanFunc(pCxt, pFunc);
+  }
   if (TSDB_CODE_SUCCESS == code) {
     setFuncClassification(pCxt->pCurrStmt, pFunc);
   }
@@ -2760,6 +2823,7 @@ static int32_t translateDeleteWhere(STranslateContext* pCxt, SDeleteStmt* pDelet
 }
 
 static int32_t translateDelete(STranslateContext* pCxt, SDeleteStmt* pDelete) {
+  pCxt->pCurrStmt = (SNode*)pDelete;
   int32_t code = translateFrom(pCxt, pDelete->pFromTable);
   if (TSDB_CODE_SUCCESS == code) {
     code = translateDeleteWhere(pCxt, pDelete);
@@ -3110,6 +3174,7 @@ static void buildAlterDbReq(STranslateContext* pCxt, SAlterDatabaseStmt* pStmt,
   pReq->buffer = pStmt->pOptions->buffer;
   pReq->pageSize = -1;
   pReq->pages = pStmt->pOptions->pages;
+  pReq->lastRowMem = -1;
   pReq->daysPerFile = -1;
   pReq->daysToKeep0 = pStmt->pOptions->keep[0];
   pReq->daysToKeep1 = pStmt->pOptions->keep[1];
@@ -3187,13 +3252,30 @@ static bool validRollupFunc(const char* pFunc) {
   return false;
 }
 
-static int32_t checkTableRollupOption(STranslateContext* pCxt, SNodeList* pFuncs) {
+static int32_t checkTableRollupOption(STranslateContext* pCxt, SNodeList* pFuncs, bool createStable,
+                                      SDbCfgInfo* pDbCfg) {
   if (NULL == pFuncs) {
+    if (NULL != pDbCfg->pRetensions) {
+      return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_TABLE_OPTION,
+                                     "To create a super table in a database with the retensions parameter configured, "
+                                     "the 'ROLLUP' option must be present");
+    }
     return TSDB_CODE_SUCCESS;
   }
 
-  if (1 != LIST_LENGTH(pFuncs) || !validRollupFunc(((SFunctionNode*)nodesListGetNode(pFuncs, 0))->functionName)) {
+  if (!createStable || NULL == pDbCfg->pRetensions) {
-    return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ROLLUP_OPTION);
+    return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_TABLE_OPTION,
+                                   "Invalid option rollup: Only supported for create super table in databases "
+                                   "configured with the 'RETENTIONS' option");
+  }
+  if (1 != LIST_LENGTH(pFuncs)) {
+    return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ROLLUP_OPTION,
+                                   "Invalid option rollup: only one function is allowed");
+  }
+  const char* pFunc = ((SFunctionNode*)nodesListGetNode(pFuncs, 0))->functionName;
+  if (!validRollupFunc(pFunc)) {
+    return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ROLLUP_OPTION,
+                                   "Invalid option rollup: %s function is not supported", pFunc);
   }
   return TSDB_CODE_SUCCESS;
 }
@@ -3219,8 +3301,8 @@ static int32_t checkTableTagsSchema(STranslateContext* pCxt, SHashObj* pHash, SN
       code = generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_ONLY_ONE_JSON_TAG);
     }
     if (TSDB_CODE_SUCCESS == code) {
-      if ((TSDB_DATA_TYPE_VARCHAR == pTag->dataType.type && pTag->dataType.bytes > TSDB_MAX_BINARY_LEN) ||
+      if ((TSDB_DATA_TYPE_VARCHAR == pTag->dataType.type && calcTypeBytes(pTag->dataType) > TSDB_MAX_BINARY_LEN) ||
-          (TSDB_DATA_TYPE_NCHAR == pTag->dataType.type && pTag->dataType.bytes > TSDB_MAX_NCHAR_LEN)) {
+          (TSDB_DATA_TYPE_NCHAR == pTag->dataType.type && calcTypeBytes(pTag->dataType) > TSDB_MAX_NCHAR_LEN)) {
         code = generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_VAR_COLUMN_LEN);
       }
     }
@@ -3241,11 +3323,11 @@ static int32_t checkTableTagsSchema(STranslateContext* pCxt, SHashObj* pHash, SN
   return code;
 }
 
-static int32_t checkTableColsSchema(STranslateContext* pCxt, SHashObj* pHash, SNodeList* pCols) {
+static int32_t checkTableColsSchema(STranslateContext* pCxt, SHashObj* pHash, int32_t ntags, SNodeList* pCols) {
   int32_t ncols = LIST_LENGTH(pCols);
   if (ncols < TSDB_MIN_COLUMNS) {
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_COLUMNS_NUM);
-  } else if (ncols > TSDB_MAX_COLUMNS) {
+  } else if (ncols + ntags > TSDB_MAX_COLUMNS) {
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_TOO_MANY_COLUMNS);
   }
 
@@ -3301,7 +3383,7 @@ static int32_t checkTableSchema(STranslateContext* pCxt, SCreateTableStmt* pStmt
 
   int32_t code = checkTableTagsSchema(pCxt, pHash, pStmt->pTags);
   if (TSDB_CODE_SUCCESS == code) {
-    code = checkTableColsSchema(pCxt, pHash, pStmt->pCols);
+    code = checkTableColsSchema(pCxt, pHash, LIST_LENGTH(pStmt->pTags), pStmt->pCols);
   }
 
   taosHashCleanup(pHash);
@@ -3330,11 +3412,18 @@ static int32_t getTableMaxDelayOption(STranslateContext* pCxt, SValueNode* pVal,
                                         pMaxDelay);
 }
 
-static int32_t checkTableMaxDelayOption(STranslateContext* pCxt, STableOptions* pOptions) {
+static int32_t checkTableMaxDelayOption(STranslateContext* pCxt, STableOptions* pOptions, bool createStable,
+                                        SDbCfgInfo* pDbCfg) {
   if (NULL == pOptions->pMaxDelay) {
     return TSDB_CODE_SUCCESS;
   }
 
+  if (!createStable || NULL == pDbCfg->pRetensions) {
+    return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_TABLE_OPTION,
+                                   "Invalid option maxdelay: Only supported for create super table in databases "
+                                   "configured with the 'RETENTIONS' option");
+  }
+
   if (LIST_LENGTH(pOptions->pMaxDelay) > 2) {
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_TABLE_OPTION, "maxdelay");
   }
@@ -3353,11 +3442,18 @@ static int32_t getTableWatermarkOption(STranslateContext* pCxt, SValueNode* pVal
                                         pMaxDelay);
 }
 
-static int32_t checkTableWatermarkOption(STranslateContext* pCxt, STableOptions* pOptions) {
+static int32_t checkTableWatermarkOption(STranslateContext* pCxt, STableOptions* pOptions, bool createStable,
+                                         SDbCfgInfo* pDbCfg) {
   if (NULL == pOptions->pWatermark) {
     return TSDB_CODE_SUCCESS;
   }
 
+  if (!createStable || NULL == pDbCfg->pRetensions) {
+    return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_TABLE_OPTION,
+                                   "Invalid option watermark: Only supported for create super table in databases "
+                                   "configured with the 'RETENTIONS' option");
+  }
+
   if (LIST_LENGTH(pOptions->pWatermark) > 2) {
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_TABLE_OPTION, "watermark");
   }
@@ -3371,13 +3467,20 @@ static int32_t checkTableWatermarkOption(STranslateContext* pCxt, STableOptions*
   return code;
 }
 
-static int32_t checkCreateTable(STranslateContext* pCxt, SCreateTableStmt* pStmt) {
+static int32_t checkCreateTable(STranslateContext* pCxt, SCreateTableStmt* pStmt, bool createStable) {
-  int32_t code = checkTableMaxDelayOption(pCxt, pStmt->pOptions);
+  int32_t    code = TSDB_CODE_SUCCESS;
-  if (TSDB_CODE_SUCCESS == code) {
+  SDbCfgInfo dbCfg = {0};
-    code = checkTableWatermarkOption(pCxt, pStmt->pOptions);
+  if (createStable) {
+    code = getDBCfg(pCxt, pStmt->dbName, &dbCfg);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = checkTableRollupOption(pCxt, pStmt->pOptions->pRollupFuncs);
+    code = checkTableMaxDelayOption(pCxt, pStmt->pOptions, createStable, &dbCfg);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = checkTableWatermarkOption(pCxt, pStmt->pOptions, createStable, &dbCfg);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = checkTableRollupOption(pCxt, pStmt->pOptions->pRollupFuncs, createStable, &dbCfg);
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = checkTableSmaOption(pCxt, pStmt);
@@ -3656,7 +3759,7 @@ static int32_t buildCreateStbReq(STranslateContext* pCxt, SCreateTableStmt* pStm
 
 static int32_t translateCreateSuperTable(STranslateContext* pCxt, SCreateTableStmt* pStmt) {
   SMCreateStbReq createReq = {0};
-  int32_t        code = checkCreateTable(pCxt, pStmt);
+  int32_t        code = checkCreateTable(pCxt, pStmt, true);
   if (TSDB_CODE_SUCCESS == code) {
     code = buildCreateStbReq(pCxt, pStmt, &createReq);
   }
@@ -3743,12 +3846,24 @@ static int32_t buildAlterSuperTableReq(STranslateContext* pCxt, SAlterTableStmt*
   return TSDB_CODE_SUCCESS;
 }
 
-static SSchema* getColSchema(STableMeta* pTableMeta, const char* pTagName) {
+static SSchema* getColSchema(STableMeta* pTableMeta, const char* pColName) {
   int32_t numOfFields = getNumOfTags(pTableMeta) + getNumOfColumns(pTableMeta);
   for (int32_t i = 0; i < numOfFields; ++i) {
-    SSchema* pTagSchema = pTableMeta->schema + i;
+    SSchema* pSchema = pTableMeta->schema + i;
-    if (0 == strcmp(pTagName, pTagSchema->name)) {
+    if (0 == strcmp(pColName, pSchema->name)) {
-      return pTagSchema;
+      return pSchema;
+    }
+  }
+  return NULL;
+}
+
+static SSchema* getTagSchema(STableMeta* pTableMeta, const char* pTagName) {
+  int32_t  numOfTags = getNumOfTags(pTableMeta);
+  SSchema* pTagsSchema = getTableTagSchema(pTableMeta);
+  for (int32_t i = 0; i < numOfTags; ++i) {
+    SSchema* pSchema = pTagsSchema + i;
+    if (0 == strcmp(pTagName, pSchema->name)) {
+      return pSchema;
     }
   }
   return NULL;
@@ -3756,23 +3871,49 @@ static SSchema* getColSchema(STableMeta* pTableMeta, const char* pTagName) {
 
 static int32_t checkAlterSuperTable(STranslateContext* pCxt, SAlterTableStmt* pStmt) {
   if (TSDB_ALTER_TABLE_UPDATE_TAG_VAL == pStmt->alterType || TSDB_ALTER_TABLE_UPDATE_COLUMN_NAME == pStmt->alterType) {
+    return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ALTER_TABLE,
+                                   "Set tag value only available for child table");
+  }
+
+  if (pStmt->alterType == TSDB_ALTER_TABLE_UPDATE_OPTIONS && -1 != pStmt->pOptions->ttl) {
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ALTER_TABLE);
   }
-  if (TSDB_ALTER_TABLE_UPDATE_COLUMN_BYTES == pStmt->alterType ||
+
-      TSDB_ALTER_TABLE_UPDATE_TAG_BYTES == pStmt->alterType) {
+  if (pStmt->dataType.type == TSDB_DATA_TYPE_JSON && pStmt->alterType == TSDB_ALTER_TABLE_ADD_TAG) {
+    return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_ONLY_ONE_JSON_TAG);
+  }
+
+  if (pStmt->dataType.type == TSDB_DATA_TYPE_JSON && pStmt->alterType == TSDB_ALTER_TABLE_ADD_COLUMN) {
+    return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_COL_JSON);
+  }
+
   STableMeta* pTableMeta = NULL;
   int32_t     code = getTableMeta(pCxt, pStmt->dbName, pStmt->tableName, &pTableMeta);
-    if (TSDB_CODE_SUCCESS == code) {
+  if (TSDB_CODE_SUCCESS != code) {
+    return code;
+  }
+
+  SSchema* pTagsSchema = getTableTagSchema(pTableMeta);
+  if (getNumOfTags(pTableMeta) == 1 && pTagsSchema->type == TSDB_DATA_TYPE_JSON &&
+      (pStmt->alterType == TSDB_ALTER_TABLE_ADD_TAG || pStmt->alterType == TSDB_ALTER_TABLE_DROP_TAG ||
+       pStmt->alterType == TSDB_ALTER_TABLE_UPDATE_TAG_BYTES)) {
+    return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_ONLY_ONE_JSON_TAG);
+  }
+
+  if (TSDB_ALTER_TABLE_UPDATE_COLUMN_BYTES == pStmt->alterType ||
+      TSDB_ALTER_TABLE_UPDATE_TAG_BYTES == pStmt->alterType) {
+    if (TSDB_SUPER_TABLE != pTableMeta->tableType) {
+      return generateSyntaxErrMsgExt(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ALTER_TABLE, "Table is not super table");
+    }
+
     SSchema* pSchema = getColSchema(pTableMeta, pStmt->colName);
     if (NULL == pSchema) {
-        code = generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_COLUMN, pStmt->colName);
+      return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_COLUMN, pStmt->colName);
     } else if (!IS_VAR_DATA_TYPE(pSchema->type) || pSchema->type != pStmt->dataType.type ||
                pSchema->bytes >= calcTypeBytes(pStmt->dataType)) {
-        code = generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_MODIFY_COL);
+      return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_MODIFY_COL);
     }
   }
-    return code;
-  }
   return TSDB_CODE_SUCCESS;
 }
 
@@ -4198,6 +4339,39 @@ static void getSourceDatabase(SNode* pStmt, int32_t acctId, char* pDbFName) {
   tNameGetFullDbName(&name, pDbFName);
 }
 
+static int32_t addWstartTsToCreateStreamQuery(SNode* pStmt) {
+  SSelectStmt* pSelect = (SSelectStmt*)pStmt;
+  SNode*       pProj = nodesListGetNode(pSelect->pProjectionList, 0);
+  if (NULL == pSelect->pWindow ||
+      (QUERY_NODE_FUNCTION == nodeType(pProj) && 0 == strcmp("_wstartts", ((SFunctionNode*)pProj)->functionName))) {
+    return TSDB_CODE_SUCCESS;
+  }
+  SFunctionNode* pFunc = (SFunctionNode*)nodesMakeNode(QUERY_NODE_FUNCTION);
+  if (NULL == pFunc) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+  strcpy(pFunc->functionName, "_wstartts");
+  strcpy(pFunc->node.aliasName, pFunc->functionName);
+  int32_t code = nodesListPushFront(pSelect->pProjectionList, (SNode*)pFunc);
+  if (TSDB_CODE_SUCCESS != code) {
+    nodesDestroyNode((SNode*)pFunc);
+  }
+  return code;
+}
+
+static int32_t buildCreateStreamQuery(STranslateContext* pCxt, SNode* pStmt, SCMCreateStreamReq* pReq) {
+  pCxt->createStream = true;
+  int32_t code = addWstartTsToCreateStreamQuery(pStmt);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = translateQuery(pCxt, pStmt);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    getSourceDatabase(pStmt, pCxt->pParseCxt->acctId, pReq->sourceDB);
+    code = nodesNodeToString(pStmt, false, &pReq->ast, NULL);
+  }
+  return code;
+}
+
 static int32_t buildCreateStreamReq(STranslateContext* pCxt, SCreateStreamStmt* pStmt, SCMCreateStreamReq* pReq) {
   pReq->igExists = pStmt->ignoreExists;
 
@@ -4211,13 +4385,7 @@ static int32_t buildCreateStreamReq(STranslateContext* pCxt, SCreateStreamStmt*
     tNameExtractFullName(&name, pReq->targetStbFullName);
   }
 
-  pCxt->createStream = true;
+  int32_t code = buildCreateStreamQuery(pCxt, pStmt->pQuery, pReq);
-  int32_t code = translateQuery(pCxt, pStmt->pQuery);
-  if (TSDB_CODE_SUCCESS == code) {
-    getSourceDatabase(pStmt->pQuery, pCxt->pParseCxt->acctId, pReq->sourceDB);
-    code = nodesNodeToString(pStmt->pQuery, false, &pReq->ast, NULL);
-  }
-
   if (TSDB_CODE_SUCCESS == code) {
     pReq->sql = strdup(pCxt->pParseCxt->pSql);
     if (NULL == pReq->sql) {
@@ -4455,6 +4623,7 @@ static int32_t translateQuery(STranslateContext* pCxt, SNode* pNode) {
       code = translateDropSuperTable(pCxt, (SDropSuperTableStmt*)pNode);
       break;
     case QUERY_NODE_ALTER_TABLE_STMT:
+    case QUERY_NODE_ALTER_SUPER_TABLE_STMT:
       code = translateAlterSuperTable(pCxt, (SAlterTableStmt*)pNode);
       break;
     case QUERY_NODE_CREATE_USER_STMT:
@@ -5165,7 +5334,7 @@ static int32_t buildCreateTableDataBlock(int32_t acctId, const SCreateTableStmt*
 static int32_t rewriteCreateTable(STranslateContext* pCxt, SQuery* pQuery) {
   SCreateTableStmt* pStmt = (SCreateTableStmt*)pQuery->pRoot;
 
-  int32_t     code = checkCreateTable(pCxt, pStmt);
+  int32_t     code = checkCreateTable(pCxt, pStmt, false);
   SVgroupInfo info = {0};
   if (TSDB_CODE_SUCCESS == code) {
     code = getTableHashVgroup(pCxt, pStmt->dbName, pStmt->tableName, &info);
@@ -5222,30 +5391,62 @@ static void addCreateTbReqIntoVgroup(int32_t acctId, SHashObj* pVgroupHashmap, S
   }
 }
 
-static int32_t createValueFromFunction(STranslateContext* pCxt, SFunctionNode* pFunc, SValueNode** pVal) {
-  int32_t code = getFuncInfo(pCxt, pFunc);
-  if (TSDB_CODE_SUCCESS == code) {
-    code = scalarCalculateConstants((SNode*)pFunc, (SNode**)pVal);
-  }
-  return code;
-}
-
 static SDataType schemaToDataType(uint8_t precision, SSchema* pSchema) {
   SDataType dt = {.type = pSchema->type, .bytes = pSchema->bytes, .precision = precision, .scale = 0};
   return dt;
 }
 
-static int32_t translateTagVal(STranslateContext* pCxt, uint8_t precision, SSchema* pSchema, SNode* pNode,
+static int32_t createCastFuncForTag(STranslateContext* pCxt, SNode* pNode, SDataType dt, SNode** pCast) {
-                               SValueNode** pVal) {
+  SNode* pExpr = nodesCloneNode(pNode);
-  if (QUERY_NODE_FUNCTION == nodeType(pNode)) {
+  if (NULL == pExpr) {
-    return createValueFromFunction(pCxt, (SFunctionNode*)pNode, pVal);
+    return TSDB_CODE_OUT_OF_MEMORY;
-  } else if (QUERY_NODE_VALUE == nodeType(pNode)) {
+  }
-    return (DEAL_RES_ERROR == translateValueImpl(pCxt, (SValueNode*)pNode, schemaToDataType(precision, pSchema))
+  int32_t code = translateExpr(pCxt, &pExpr);
-                ? pCxt->errCode
+  if (TSDB_CODE_SUCCESS == code) {
-                : TSDB_CODE_SUCCESS);
+    code = createCastFunc(pCxt, pExpr, dt, pCast);
+  }
+  if (TSDB_CODE_SUCCESS != code) {
+    nodesDestroyNode(pExpr);
+  }
+  return code;
+}
+
+static int32_t createTagValFromExpr(STranslateContext* pCxt, SDataType targetDt, SNode* pNode, SValueNode** pVal) {
+  SNode*  pCast = NULL;
+  int32_t code = createCastFuncForTag(pCxt, pNode, targetDt, &pCast);
+  SNode*  pNew = NULL;
+  if (TSDB_CODE_SUCCESS == code) {
+    code = scalarCalculateConstants(pCast, &pNew);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    pCast = pNew;
+    if (QUERY_NODE_VALUE != nodeType(pCast)) {
+      code = generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_WRONG_VALUE_TYPE, ((SExprNode*)pNode)->aliasName);
+    }
+  }
+
+  if (TSDB_CODE_SUCCESS == code) {
+    *pVal = (SValueNode*)pCast;
   } else {
-    // return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_WRONG_VALUE_TYPE, ((SExprNode*)pNode)->aliasName);
+    nodesDestroyNode(pCast);
-    return TSDB_CODE_FAILED;
+  }
+  return code;
+}
+
+static int32_t createTagValFromVal(STranslateContext* pCxt, SDataType targetDt, SNode* pNode, SValueNode** pVal) {
+  *pVal = (SValueNode*)nodesCloneNode(pNode);
+  if (NULL == *pVal) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+  return DEAL_RES_ERROR == translateValueImpl(pCxt, *pVal, targetDt, true) ? pCxt->errCode : TSDB_CODE_SUCCESS;
+}
+
+static int32_t createTagVal(STranslateContext* pCxt, uint8_t precision, SSchema* pSchema, SNode* pNode,
+                            SValueNode** pVal) {
+  if (QUERY_NODE_VALUE == nodeType(pNode)) {
+    return createTagValFromVal(pCxt, schemaToDataType(precision, pSchema), pNode, pVal);
+  } else {
+    return createTagValFromExpr(pCxt, schemaToDataType(precision, pSchema), pNode, pVal);
   }
 }
 
@@ -5285,35 +5486,23 @@ static int32_t buildKVRowForBindTags(STranslateContext* pCxt, SCreateSubTableCla
   if (NULL == pTagArray) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
+
   int32_t code = TSDB_CODE_SUCCESS;
-  SSchema* pTagSchema = getTableTagSchema(pSuperTableMeta);
+
-  SNode *  pTag = NULL, *pNode = NULL;
   bool       isJson = false;
+  SNodeList* pVals = NULL;
+  SNode *    pTag = NULL, *pNode = NULL;
   FORBOTH(pTag, pStmt->pSpecificTags, pNode, pStmt->pValsOfTags) {
     SColumnNode* pCol = (SColumnNode*)pTag;
-    SSchema*     pSchema = NULL;
+    SSchema*     pSchema = getTagSchema(pSuperTableMeta, pCol->colName);
-    for (int32_t i = 0; i < numOfTags; ++i) {
-      if (0 == strcmp(pCol->colName, pTagSchema[i].name)) {
-        pSchema = pTagSchema + i;
-        break;
-      }
-    }
     if (NULL == pSchema) {
       code = generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_TAG_NAME, pCol->colName);
-      goto end;
     }
     SValueNode* pVal = NULL;
-    code = translateTagVal(pCxt, pSuperTableMeta->tableInfo.precision, pSchema, pNode, &pVal);
+    if (TSDB_CODE_SUCCESS == code) {
-    if (TSDB_CODE_SUCCESS != code) {
+      code = createTagVal(pCxt, pSuperTableMeta->tableInfo.precision, pSchema, pNode, &pVal);
-      goto end;
     }
-
+    if (TSDB_CODE_SUCCESS == code) {
-    if (NULL == pVal) {
-      pVal = (SValueNode*)pNode;
-    } else {
-      REPLACE_LIST2_NODE(pVal);
-    }
-
       if (pSchema->type == TSDB_DATA_TYPE_JSON) {
         isJson = true;
         code = buildJsonTagVal(pCxt, pSchema, pVal, pTagArray, ppTag);
@@ -5321,20 +5510,21 @@ static int32_t buildKVRowForBindTags(STranslateContext* pCxt, SCreateSubTableCla
         code = buildNormalTagVal(pCxt, pSchema, pVal, pTagArray);
       }
     }
+    if (TSDB_CODE_SUCCESS == code) {
+      code = nodesListMakeAppend(&pVals, (SNode*)pVal);
+    }
+    if (TSDB_CODE_SUCCESS != code) {
+      break;
+    }
+  }
 
-  if (!isJson) code = tTagNew(pTagArray, 1, false, ppTag);
+  if (TSDB_CODE_SUCCESS == code && !isJson) {
+    code = tTagNew(pTagArray, 1, false, ppTag);
+  }
 
-end:
+  nodesDestroyList(pVals);
-  if (isJson) {
-    for (int i = 0; i < taosArrayGetSize(pTagArray); ++i) {
-      STagVal* p = (STagVal*)taosArrayGet(pTagArray, i);
-      if (IS_VAR_DATA_TYPE(p->type)) {
-        taosMemoryFree(p->pData);
-      }
-    }
-  }
   taosArrayDestroy(pTagArray);
-  return TSDB_CODE_SUCCESS;
+  return code;
 }
 
 static int32_t buildKVRowForAllTags(STranslateContext* pCxt, SCreateSubTableClause* pStmt, STableMeta* pSuperTableMeta,
@@ -5343,39 +5533,26 @@ static int32_t buildKVRowForAllTags(STranslateContext* pCxt, SCreateSubTableClau
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_TAGS_NOT_MATCHED);
   }
 
-  SSchema* pTagSchemas = getTableTagSchema(pSuperTableMeta);
-  SNode*   pNode;
-  int32_t  code = TSDB_CODE_SUCCESS;
-  int32_t  index = 0;
   SArray* pTagArray = taosArrayInit(LIST_LENGTH(pStmt->pValsOfTags), sizeof(STagVal));
-  if (!pTagArray) {
+  if (NULL == pTagArray) {
-    return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_TSC_OUT_OF_MEMORY);
+    return TSDB_CODE_OUT_OF_MEMORY;
   }
 
+  int32_t code = TSDB_CODE_SUCCESS;
+
   bool       isJson = false;
+  int32_t    index = 0;
+  SSchema*   pTagSchemas = getTableTagSchema(pSuperTableMeta);
+  SNodeList* pVals = NULL;
+  SNode*     pNode;
   FOREACH(pNode, pStmt->pValsOfTags) {
     SValueNode* pVal = NULL;
     SSchema*    pTagSchema = pTagSchemas + index;
-    code = translateTagVal(pCxt, pSuperTableMeta->tableInfo.precision, pTagSchema, pNode, &pVal);
+    code = createTagVal(pCxt, pSuperTableMeta->tableInfo.precision, pTagSchema, pNode, &pVal);
-    if (TSDB_CODE_SUCCESS != code) {
+    if (TSDB_CODE_SUCCESS == code) {
-      goto end;
-    }
-    if (NULL == pVal) {
-      pVal = (SValueNode*)pNode;
-    } else {
-      REPLACE_NODE(pVal);
-    }
       if (pTagSchema->type == TSDB_DATA_TYPE_JSON) {
-      if (pVal->literal && strlen(pVal->literal) > (TSDB_MAX_JSON_TAG_LEN - VARSTR_HEADER_SIZE) / TSDB_NCHAR_SIZE) {
-        code = buildSyntaxErrMsg(&pCxt->msgBuf, "json string too long than 4095", pVal->literal);
-        goto end;
-      }
-
         isJson = true;
-      code = parseJsontoTagData(pVal->literal, pTagArray, ppTag, &pCxt->msgBuf);
+        code = buildJsonTagVal(pCxt, pTagSchema, pVal, pTagArray, ppTag);
-      if (code != TSDB_CODE_SUCCESS) {
-        goto end;
-      }
       } else if (pVal->node.resType.type != TSDB_DATA_TYPE_NULL && !pVal->isNull) {
         char*   tmpVal = nodesGetValueFromNode(pVal);
         STagVal val = {.cid = pTagSchema->colId, .type = pTagSchema->type};
@@ -5387,20 +5564,21 @@ static int32_t buildKVRowForAllTags(STranslateContext* pCxt, SCreateSubTableClau
         }
         taosArrayPush(pTagArray, &val);
       }
+    }
+    if (TSDB_CODE_SUCCESS == code) {
+      code = nodesListMakeAppend(&pVals, (SNode*)pVal);
+    }
+    if (TSDB_CODE_SUCCESS != code) {
+      break;
+    }
     ++index;
   }
-  if (!isJson) code = tTagNew(pTagArray, 1, false, ppTag);
 
-end:
+  if (TSDB_CODE_SUCCESS == code && !isJson) {
-  if (isJson) {
+    code = tTagNew(pTagArray, 1, false, ppTag);
-    for (int i = 0; i < taosArrayGetSize(pTagArray); ++i) {
-      STagVal* p = (STagVal*)taosArrayGet(pTagArray, i);
-      if (IS_VAR_DATA_TYPE(p->type)) {
-        taosMemoryFree(p->pData);
-      }
-    }
   }
 
+  nodesDestroyList(pVals);
   taosArrayDestroy(pTagArray);
   return code;
 }
@@ -5644,7 +5822,7 @@ static int32_t buildUpdateTagValReq(STranslateContext* pCxt, SAlterTableStmt* pS
   }
 
   SDataType targetDt = schemaToDataType(pTableMeta->tableInfo.precision, pSchema);
-  if (DEAL_RES_ERROR == translateValueImpl(pCxt, pStmt->pVal, targetDt)) {
+  if (DEAL_RES_ERROR == translateValueImpl(pCxt, pStmt->pVal, targetDt, true)) {
     return pCxt->errCode;
   }
 
@@ -5898,15 +6076,6 @@ static int32_t rewriteAlterTableImpl(STranslateContext* pCxt, SAlterTableStmt* p
   }
 
   if (TSDB_SUPER_TABLE == pTableMeta->tableType) {
-    SSchema* pTagsSchema = getTableTagSchema(pTableMeta);
-    if (getNumOfTags(pTableMeta) == 1 && pTagsSchema->type == TSDB_DATA_TYPE_JSON &&
-        (pStmt->alterType == TSDB_ALTER_TABLE_ADD_TAG || pStmt->alterType == TSDB_ALTER_TABLE_DROP_TAG ||
-         pStmt->alterType == TSDB_ALTER_TABLE_UPDATE_TAG_BYTES)) {
-      return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_ONLY_ONE_JSON_TAG);
-    }
-    if (pStmt->alterType == TSDB_ALTER_TABLE_UPDATE_OPTIONS && -1 != pStmt->pOptions->ttl) {
-      return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ALTER_TABLE);
-    }
     return TSDB_CODE_SUCCESS;
   } else if (TSDB_CHILD_TABLE != pTableMeta->tableType && TSDB_NORMAL_TABLE != pTableMeta->tableType) {
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_ALTER_TABLE);
@@ -5929,10 +6098,6 @@ static int32_t rewriteAlterTableImpl(STranslateContext* pCxt, SAlterTableStmt* p
 static int32_t rewriteAlterTable(STranslateContext* pCxt, SQuery* pQuery) {
   SAlterTableStmt* pStmt = (SAlterTableStmt*)pQuery->pRoot;
 
-  if (pStmt->dataType.type == TSDB_DATA_TYPE_JSON && pStmt->alterType == TSDB_ALTER_TABLE_ADD_TAG) {
-    return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_ONLY_ONE_JSON_TAG);
-  }
-
   if (pStmt->dataType.type == TSDB_DATA_TYPE_JSON && pStmt->alterType == TSDB_ALTER_TABLE_ADD_COLUMN) {
     return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INVALID_COL_JSON);
   }

source/libs/parser/src/parUtil.c

@@ -215,6 +215,14 @@ int32_t generateSyntaxErrMsg(SMsgBuf* pBuf, int32_t errCode, ...) {
   return errCode;
 }
 
+int32_t generateSyntaxErrMsgExt(SMsgBuf* pBuf, int32_t errCode, const char* pFormat, ...) {
+  va_list vArgList;
+  va_start(vArgList, pFormat);
+  vsnprintf(pBuf->buf, pBuf->len, pFormat, vArgList);
+  va_end(vArgList);
+  return errCode;
+}
+
 int32_t buildInvalidOperationMsg(SMsgBuf* pBuf, const char* msg) {
   strncpy(pBuf->buf, msg, pBuf->len);
   return TSDB_CODE_TSC_INVALID_OPERATION;

source/libs/parser/test/mockCatalog.cpp

@@ -159,7 +159,7 @@ void generatePerformanceSchema(MockCatalogService* mcs) {
  *          c4         |       column       |       DOUBLE       |    8     |
  *          c5         |       column       |       DOUBLE       |    8     |
  */
-void generateTestT1(MockCatalogService* mcs) {
+void generateTestTables(MockCatalogService* mcs) {
   ITableBuilder& builder = mcs->createTableBuilder("test", "t1", TSDB_NORMAL_TABLE, 6)
                                .setPrecision(TSDB_TIME_PRECISION_MILLI)
                                .setVgid(1)
@@ -183,8 +183,18 @@ void generateTestT1(MockCatalogService* mcs) {
  *         tag2        |        tag         |      VARCHAR       |    20    |
  *         tag3        |        tag         |     TIMESTAMP      |    8     |
  * Child Table: st1s1, st1s2
+ *
+ * Super Table: st2
+ *        Field        |        Type        |      DataType      |  Bytes   |
+ * ==========================================================================
+ *          ts         |       column       |     TIMESTAMP      |    8     |
+ *          c1         |       column       |        INT         |    4     |
+ *          c2         |       column       |      VARCHAR       |    20    |
+ *         jtag        |        tag         |        json        |    --    |
+ * Child Table: st2s1, st2s2
  */
-void generateTestST1(MockCatalogService* mcs) {
+void generateTestStables(MockCatalogService* mcs) {
+  {
     ITableBuilder& builder = mcs->createTableBuilder("test", "st1", TSDB_SUPER_TABLE, 3, 3)
                                  .setPrecision(TSDB_TIME_PRECISION_MILLI)
                                  .addColumn("ts", TSDB_DATA_TYPE_TIMESTAMP)
@@ -198,18 +208,7 @@ void generateTestST1(MockCatalogService* mcs) {
     mcs->createSubTable("test", "st1", "st1s2", 2);
     mcs->createSubTable("test", "st1", "st1s3", 1);
   }
-
+  {
-/*
- * Super Table: st2
- *        Field        |        Type        |      DataType      |  Bytes   |
- * ==========================================================================
- *          ts         |       column       |     TIMESTAMP      |    8     |
- *          c1         |       column       |        INT         |    4     |
- *          c2         |       column       |      VARCHAR       |    20    |
- *         jtag        |        tag         |        json        |    --    |
- * Child Table: st2s1, st2s2
- */
-void generateTestST2(MockCatalogService* mcs) {
     ITableBuilder& builder = mcs->createTableBuilder("test", "st2", TSDB_SUPER_TABLE, 3, 1)
                                  .setPrecision(TSDB_TIME_PRECISION_MILLI)
                                  .addColumn("ts", TSDB_DATA_TYPE_TIMESTAMP)
@@ -220,6 +219,7 @@ void generateTestST2(MockCatalogService* mcs) {
     mcs->createSubTable("test", "st2", "st2s1", 1);
     mcs->createSubTable("test", "st2", "st2s2", 2);
   }
+}
 
 void generateFunctions(MockCatalogService* mcs) {
   mcs->createFunction("udf1", TSDB_FUNC_TYPE_SCALAR, TSDB_DATA_TYPE_INT, tDataTypes[TSDB_DATA_TYPE_INT].bytes, 0);
@@ -233,6 +233,11 @@ void generateDnodes(MockCatalogService* mcs) {
   mcs->createDnode(3, "host3", 7030);
 }
 
+void generateDatabases(MockCatalogService* mcs) {
+  mcs->createDatabase("test");
+  mcs->createDatabase("rollup_db", true);
+}
+
 }  // namespace
 
 int32_t __catalogGetHandle(const char* clusterId, struct SCatalog** catalogHandle) { return 0; }
@@ -262,7 +267,7 @@ int32_t __catalogGetDBVgInfo(SCatalog* pCtg, SRequestConnInfo* pConn, const char
 }
 
 int32_t __catalogGetDBCfg(SCatalog* pCtg, SRequestConnInfo* pConn, const char* dbFName, SDbCfgInfo* pDbCfg) {
-  return 0;
+  return g_mockCatalogService->catalogGetDBCfg(dbFName, pDbCfg);
 }
 
 int32_t __catalogChkAuth(SCatalog* pCtg, SRequestConnInfo* pConn, const char* user, const char* dbFName, AUTH_TYPE type,
@@ -359,11 +364,11 @@ void initMetaDataEnv() {
 }
 
 void generateMetaData() {
+  generateDatabases(g_mockCatalogService.get());
   generateInformationSchema(g_mockCatalogService.get());
   generatePerformanceSchema(g_mockCatalogService.get());
-  generateTestT1(g_mockCatalogService.get());
+  generateTestTables(g_mockCatalogService.get());
-  generateTestST1(g_mockCatalogService.get());
+  generateTestStables(g_mockCatalogService.get());
-  generateTestST2(g_mockCatalogService.get());
   generateFunctions(g_mockCatalogService.get());
   generateDnodes(g_mockCatalogService.get());
   g_mockCatalogService->showTables();

source/libs/parser/test/mockCatalogService.cpp

@@ -140,6 +140,17 @@ class MockCatalogServiceImpl {
     return TSDB_CODE_SUCCESS;
   }
 
+  int32_t catalogGetDBCfg(const char* pDbFName, SDbCfgInfo* pDbCfg) const {
+    std::string                dbFName(pDbFName);
+    DbCfgCache::const_iterator it = dbCfg_.find(dbFName.substr(std::string(pDbFName).find_last_of('.') + 1));
+    if (dbCfg_.end() == it) {
+      return TSDB_CODE_FAILED;
+    }
+
+    memcpy(pDbCfg, &(it->second), sizeof(SDbCfgInfo));
+    return TSDB_CODE_SUCCESS;
+  }
+
   int32_t catalogGetUdfInfo(const std::string& funcName, SFuncInfo* pInfo) const {
     auto it = udf_.find(funcName);
     if (udf_.end() == it) {
@@ -323,12 +334,21 @@ class MockCatalogServiceImpl {
     dnode_.insert(std::make_pair(dnodeId, epSet));
   }
 
+  void createDatabase(const std::string& db, bool rollup) {
+    SDbCfgInfo cfg = {0};
+    if (rollup) {
+      cfg.pRetensions = taosArrayInit(TARRAY_MIN_SIZE, sizeof(SRetention));
+    }
+    dbCfg_.insert(std::make_pair(db, cfg));
+  }
+
  private:
   typedef std::map<std::string, std::shared_ptr<MockTableMeta>> TableMetaCache;
   typedef std::map<std::string, TableMetaCache>                 DbMetaCache;
   typedef std::map<std::string, std::shared_ptr<SFuncInfo>>     UdfMetaCache;
   typedef std::map<std::string, std::vector<STableIndexInfo>>   IndexMetaCache;
   typedef std::map<int32_t, SEpSet>                             DnodeCache;
+  typedef std::map<std::string, SDbCfgInfo>                     DbCfgCache;
 
   uint64_t getNextId() { return id_++; }
 
@@ -486,6 +506,7 @@ class MockCatalogServiceImpl {
       for (int32_t i = 0; i < ndbs; ++i) {
         SMetaRes res = {0};
         res.pRes = taosMemoryCalloc(1, sizeof(SDbCfgInfo));
+        res.code = catalogGetDBCfg((const char*)taosArrayGet(pDbCfgReq, i), (SDbCfgInfo*)res.pRes);
         taosArrayPush(*pDbCfgData, &res);
       }
     }
@@ -576,6 +597,7 @@ class MockCatalogServiceImpl {
   UdfMetaCache                  udf_;
   IndexMetaCache                index_;
   DnodeCache                    dnode_;
+  DbCfgCache                    dbCfg_;
 };
 
 MockCatalogService::MockCatalogService() : impl_(new MockCatalogServiceImpl()) {}
@@ -605,6 +627,8 @@ void MockCatalogService::createDnode(int32_t dnodeId, const std::string& host, i
   impl_->createDnode(dnodeId, host, port);
 }
 
+void MockCatalogService::createDatabase(const std::string& db, bool rollup) { impl_->createDatabase(db, rollup); }
+
 int32_t MockCatalogService::catalogGetTableMeta(const SName* pTableName, STableMeta** pTableMeta) const {
   return impl_->catalogGetTableMeta(pTableName, pTableMeta);
 }
@@ -621,6 +645,10 @@ int32_t MockCatalogService::catalogGetDBVgInfo(const char* pDbFName, SArray** pV
   return impl_->catalogGetDBVgInfo(pDbFName, pVgList);
 }
 
+int32_t MockCatalogService::catalogGetDBCfg(const char* pDbFName, SDbCfgInfo* pDbCfg) const {
+  return impl_->catalogGetDBCfg(pDbFName, pDbCfg);
+}
+
 int32_t MockCatalogService::catalogGetUdfInfo(const std::string& funcName, SFuncInfo* pInfo) const {
   return impl_->catalogGetUdfInfo(funcName, pInfo);
 }

source/libs/parser/test/mockCatalogService.h

@@ -63,11 +63,13 @@ class MockCatalogService {
   void createFunction(const std::string& func, int8_t funcType, int8_t outputType, int32_t outputLen, int32_t bufSize);
   void createSmaIndex(const SMCreateSmaReq* pReq);
   void createDnode(int32_t dnodeId, const std::string& host, int16_t port);
+  void createDatabase(const std::string& db, bool rollup = false);
 
   int32_t catalogGetTableMeta(const SName* pTableName, STableMeta** pTableMeta) const;
   int32_t catalogGetTableHashVgroup(const SName* pTableName, SVgroupInfo* vgInfo) const;
   int32_t catalogGetTableDistVgInfo(const SName* pTableName, SArray** pVgList) const;
   int32_t catalogGetDBVgInfo(const char* pDbFName, SArray** pVgList) const;
+  int32_t catalogGetDBCfg(const char* pDbFName, SDbCfgInfo* pDbCfg) const;
   int32_t catalogGetUdfInfo(const std::string& funcName, SFuncInfo* pInfo) const;
   int32_t catalogGetTableIndex(const SName* pTableName, SArray** pIndexes) const;
   int32_t catalogGetDnodeList(SArray** pDnodes) const;

source/libs/parser/test/parInitialATest.cpp

@@ -38,9 +38,9 @@ TEST_F(ParserInitialATest, alterDnode) {
 TEST_F(ParserInitialATest, alterDatabase) {
   useDb("root", "test");
 
-  run("ALTER DATABASE wxy_db CACHELAST 1 FSYNC 200 WAL 1");
+  run("ALTER DATABASE test CACHELAST 1 FSYNC 200 WAL 1");
 
-  run("ALTER DATABASE wxy_db KEEP 2400");
+  run("ALTER DATABASE test KEEP 2400");
 }
 
 TEST_F(ParserInitialATest, alterLocal) {
@@ -77,8 +77,6 @@ TEST_F(ParserInitialATest, alterLocal) {
   clearAlterLocal();
 }
 
-// todo ALTER stable
-
 /*
  * ALTER TABLE [db_name.]tb_name alter_table_clause
  *
@@ -157,7 +155,7 @@ TEST_F(ParserInitialATest, alterSTable) {
   };
 
   setCheckDdlFunc([&](const SQuery* pQuery, ParserStage stage) {
-    ASSERT_EQ(nodeType(pQuery->pRoot), QUERY_NODE_ALTER_TABLE_STMT);
+    ASSERT_EQ(nodeType(pQuery->pRoot), QUERY_NODE_ALTER_SUPER_TABLE_STMT);
     SMAlterStbReq req = {0};
     ASSERT_EQ(tDeserializeSMAlterStbReq(pQuery->pCmdMsg->pMsg, pQuery->pCmdMsg->msgLen, &req), TSDB_CODE_SUCCESS);
     ASSERT_EQ(std::string(req.name), std::string(expect.name));
@@ -181,44 +179,44 @@ TEST_F(ParserInitialATest, alterSTable) {
   });
 
   //  setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_UPDATE_OPTIONS, 0, nullptr, 0, 0, nullptr, nullptr, 10);
-  //  run("ALTER TABLE st1 TTL 10");
+  //  run("ALTER STABLE st1 TTL 10");
   //  clearAlterStbReq();
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_UPDATE_OPTIONS, 0, nullptr, 0, 0, nullptr, "test");
-  run("ALTER TABLE st1 COMMENT 'test'");
+  run("ALTER STABLE st1 COMMENT 'test'");
   clearAlterStbReq();
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_ADD_COLUMN, 1, "cc1", TSDB_DATA_TYPE_BIGINT);
-  run("ALTER TABLE st1 ADD COLUMN cc1 BIGINT");
+  run("ALTER STABLE st1 ADD COLUMN cc1 BIGINT");
   clearAlterStbReq();
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_DROP_COLUMN, 1, "c1");
-  run("ALTER TABLE st1 DROP COLUMN c1");
+  run("ALTER STABLE st1 DROP COLUMN c1");
   clearAlterStbReq();
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_UPDATE_COLUMN_BYTES, 1, "c2", TSDB_DATA_TYPE_VARCHAR,
                      30 + VARSTR_HEADER_SIZE);
-  run("ALTER TABLE st1 MODIFY COLUMN c2 VARCHAR(30)");
+  run("ALTER STABLE st1 MODIFY COLUMN c2 VARCHAR(30)");
   clearAlterStbReq();
 
   // setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_UPDATE_COLUMN_NAME, 2, "c1", 0, 0, "cc1");
-  // run("ALTER TABLE st1 RENAME COLUMN c1 cc1");
+  // run("ALTER STABLE st1 RENAME COLUMN c1 cc1");
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_ADD_TAG, 1, "tag11", TSDB_DATA_TYPE_BIGINT);
-  run("ALTER TABLE st1 ADD TAG tag11 BIGINT");
+  run("ALTER STABLE st1 ADD TAG tag11 BIGINT");
   clearAlterStbReq();
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_DROP_TAG, 1, "tag1");
-  run("ALTER TABLE st1 DROP TAG tag1");
+  run("ALTER STABLE st1 DROP TAG tag1");
   clearAlterStbReq();
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_UPDATE_TAG_BYTES, 1, "tag2", TSDB_DATA_TYPE_VARCHAR,
                      30 + VARSTR_HEADER_SIZE);
-  run("ALTER TABLE st1 MODIFY TAG tag2 VARCHAR(30)");
+  run("ALTER STABLE st1 MODIFY TAG tag2 VARCHAR(30)");
   clearAlterStbReq();
 
   setAlterStbReqFunc("st1", TSDB_ALTER_TABLE_UPDATE_TAG_NAME, 2, "tag1", 0, 0, "tag11");
-  run("ALTER TABLE st1 RENAME TAG tag1 tag11");
+  run("ALTER STABLE st1 RENAME TAG tag1 tag11");
   clearAlterStbReq();
 
   // todo
@@ -228,11 +226,11 @@ TEST_F(ParserInitialATest, alterSTable) {
 TEST_F(ParserInitialATest, alterSTableSemanticCheck) {
   useDb("root", "test");
 
-  run("ALTER TABLE st1 RENAME COLUMN c1 cc1", TSDB_CODE_PAR_INVALID_ALTER_TABLE);
+  run("ALTER STABLE st1 RENAME COLUMN c1 cc1", TSDB_CODE_PAR_INVALID_ALTER_TABLE);
 
-  run("ALTER TABLE st1 MODIFY COLUMN c2 NCHAR(10)", TSDB_CODE_PAR_INVALID_MODIFY_COL);
+  run("ALTER STABLE st1 MODIFY COLUMN c2 NCHAR(10)", TSDB_CODE_PAR_INVALID_MODIFY_COL);
 
-  run("ALTER TABLE st1 MODIFY TAG tag2 NCHAR(10)", TSDB_CODE_PAR_INVALID_MODIFY_COL);
+  run("ALTER STABLE st1 MODIFY TAG tag2 NCHAR(10)", TSDB_CODE_PAR_INVALID_MODIFY_COL);
 }
 
 TEST_F(ParserInitialATest, alterTable) {

source/libs/parser/test/parInitialCTest.cpp

@@ -359,11 +359,11 @@ TEST_F(ParserInitialCTest, createStable) {
     memset(&expect, 0, sizeof(SMCreateStbReq));
   };
 
-  auto setCreateStbReqFunc = [&](const char* pTbname, int8_t igExists = 0, int64_t delay1 = -1, int64_t delay2 = -1,
+  auto setCreateStbReqFunc = [&](const char* pDbName, const char* pTbName, int8_t igExists = 0, int64_t delay1 = -1,
-                                 int64_t watermark1 = TSDB_DEFAULT_ROLLUP_WATERMARK,
+                                 int64_t delay2 = -1, int64_t watermark1 = TSDB_DEFAULT_ROLLUP_WATERMARK,
                                  int64_t watermark2 = TSDB_DEFAULT_ROLLUP_WATERMARK,
                                  int32_t ttl = TSDB_DEFAULT_TABLE_TTL, const char* pComment = nullptr) {
-    int32_t len = snprintf(expect.name, sizeof(expect.name), "0.test.%s", pTbname);
+    int32_t len = snprintf(expect.name, sizeof(expect.name), "0.%s.%s", pDbName, pTbName);
     expect.name[len] = '\0';
     expect.igExists = igExists;
     expect.delay1 = delay1;
@@ -454,14 +454,14 @@ TEST_F(ParserInitialCTest, createStable) {
     tFreeSMCreateStbReq(&req);
   });
 
-  setCreateStbReqFunc("t1");
+  setCreateStbReqFunc("test", "t1");
   addFieldToCreateStbReqFunc(true, "ts", TSDB_DATA_TYPE_TIMESTAMP);
   addFieldToCreateStbReqFunc(true, "c1", TSDB_DATA_TYPE_INT);
   addFieldToCreateStbReqFunc(false, "id", TSDB_DATA_TYPE_INT);
   run("CREATE STABLE t1(ts TIMESTAMP, c1 INT) TAGS(id INT)");
   clearCreateStbReq();
 
-  setCreateStbReqFunc("t1", 1, 100 * MILLISECOND_PER_SECOND, 10 * MILLISECOND_PER_MINUTE, 10,
+  setCreateStbReqFunc("rollup_db", "t1", 1, 100 * MILLISECOND_PER_SECOND, 10 * MILLISECOND_PER_MINUTE, 10,
                       1 * MILLISECOND_PER_MINUTE, 100, "test create table");
   addFieldToCreateStbReqFunc(true, "ts", TSDB_DATA_TYPE_TIMESTAMP, 0, 0);
   addFieldToCreateStbReqFunc(true, "c1", TSDB_DATA_TYPE_INT);
@@ -493,7 +493,7 @@ TEST_F(ParserInitialCTest, createStable) {
   addFieldToCreateStbReqFunc(false, "a13", TSDB_DATA_TYPE_BOOL);
   addFieldToCreateStbReqFunc(false, "a14", TSDB_DATA_TYPE_NCHAR, 30 * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE);
   addFieldToCreateStbReqFunc(false, "a15", TSDB_DATA_TYPE_VARCHAR, 50 + VARSTR_HEADER_SIZE);
-  run("CREATE STABLE IF NOT EXISTS test.t1("
+  run("CREATE STABLE IF NOT EXISTS rollup_db.t1("
       "ts TIMESTAMP, c1 INT, c2 INT UNSIGNED, c3 BIGINT, c4 BIGINT UNSIGNED, c5 FLOAT, c6 DOUBLE, c7 BINARY(20), "
       "c8 SMALLINT, c9 SMALLINT UNSIGNED COMMENT 'test column comment', c10 TINYINT, c11 TINYINT UNSIGNED, c12 BOOL, "
       "c13 NCHAR(30), c14 VARCHAR(50)) "
@@ -507,12 +507,13 @@ TEST_F(ParserInitialCTest, createStable) {
 TEST_F(ParserInitialCTest, createStableSemanticCheck) {
   useDb("root", "test");
 
-  run("CREATE STABLE stb2 (ts TIMESTAMP, c1 INT) TAGS (tag1 INT) ROLLUP(CEIL)", TSDB_CODE_PAR_INVALID_ROLLUP_OPTION);
+  run("CREATE STABLE rollup_db.stb2 (ts TIMESTAMP, c1 INT) TAGS (tag1 INT) ROLLUP(CEIL)",
+      TSDB_CODE_PAR_INVALID_ROLLUP_OPTION);
 
-  run("CREATE STABLE stb2 (ts TIMESTAMP, c1 INT) TAGS (tag1 INT) ROLLUP(MAX) MAX_DELAY 0s WATERMARK 1m",
+  run("CREATE STABLE rollup_db.stb2 (ts TIMESTAMP, c1 INT) TAGS (tag1 INT) ROLLUP(MAX) MAX_DELAY 0s WATERMARK 1m",
       TSDB_CODE_PAR_INVALID_RANGE_OPTION);
 
-  run("CREATE STABLE stb2 (ts TIMESTAMP, c1 INT) TAGS (tag1 INT) ROLLUP(MAX) MAX_DELAY 10s WATERMARK 18m",
+  run("CREATE STABLE rollup_db.stb2 (ts TIMESTAMP, c1 INT) TAGS (tag1 INT) ROLLUP(MAX) MAX_DELAY 10s WATERMARK 18m",
       TSDB_CODE_PAR_INVALID_RANGE_OPTION);
 }
 
@@ -561,30 +562,33 @@ TEST_F(ParserInitialCTest, createStream) {
     tFreeSCMCreateStreamReq(&req);
   });
 
-  setCreateStreamReqFunc("s1", "test", "create stream s1 as select * from t1");
+  setCreateStreamReqFunc("s1", "test", "create stream s1 as select count(*) from t1 interval(10s)");
-  run("CREATE STREAM s1 AS SELECT * FROM t1");
+  run("CREATE STREAM s1 AS SELECT COUNT(*) FROM t1 INTERVAL(10S)");
   clearCreateStreamReq();
 
-  setCreateStreamReqFunc("s1", "test", "create stream if not exists s1 as select * from t1", nullptr, 1);
+  setCreateStreamReqFunc("s1", "test", "create stream if not exists s1 as select count(*) from t1 interval(10s)",
-  run("CREATE STREAM IF NOT EXISTS s1 AS SELECT * FROM t1");
+                         nullptr, 1);
+  run("CREATE STREAM IF NOT EXISTS s1 AS SELECT COUNT(*) FROM t1 INTERVAL(10S)");
   clearCreateStreamReq();
 
-  setCreateStreamReqFunc("s1", "test", "create stream s1 into st1 as select * from t1", "st1");
+  setCreateStreamReqFunc("s1", "test", "create stream s1 into st1 as select count(*) from t1 interval(10s)", "st1");
-  run("CREATE STREAM s1 INTO st1 AS SELECT * FROM t1");
+  run("CREATE STREAM s1 INTO st1 AS SELECT COUNT(*) FROM t1 INTERVAL(10S)");
   clearCreateStreamReq();
 
-  setCreateStreamReqFunc(
+  setCreateStreamReqFunc("s1", "test",
-      "s1", "test",
+                         "create stream if not exists s1 trigger max_delay 20s watermark 10s ignore expired into st1 "
-      "create stream if not exists s1 trigger max_delay 20s watermark 10s ignore expired into st1 as select * from t1",
+                         "as select count(*) from t1 interval(10s)",
-      "st1", 1, STREAM_TRIGGER_MAX_DELAY, 20 * MILLISECOND_PER_SECOND, 10 * MILLISECOND_PER_SECOND, 1);
+                         "st1", 1, STREAM_TRIGGER_MAX_DELAY, 20 * MILLISECOND_PER_SECOND, 10 * MILLISECOND_PER_SECOND,
-  run("CREATE STREAM IF NOT EXISTS s1 TRIGGER MAX_DELAY 20s WATERMARK 10s IGNORE EXPIRED INTO st1 AS SELECT * FROM t1");
+                         1);
+  run("CREATE STREAM IF NOT EXISTS s1 TRIGGER MAX_DELAY 20s WATERMARK 10s IGNORE EXPIRED INTO st1 AS SELECT COUNT(*) "
+      "FROM t1 INTERVAL(10S)");
   clearCreateStreamReq();
 }
 
 TEST_F(ParserInitialCTest, createStreamSemanticCheck) {
   useDb("root", "test");
 
-  run("CREATE STREAM s1 AS SELECT PERCENTILE(c1, 30) FROM t1", TSDB_CODE_PAR_STREAM_NOT_ALLOWED_FUNC);
+  run("CREATE STREAM s1 AS SELECT PERCENTILE(c1, 30) FROM t1 INTERVAL(10S)", TSDB_CODE_PAR_STREAM_NOT_ALLOWED_FUNC);
 }
 
 TEST_F(ParserInitialCTest, createTable) {
@@ -598,7 +602,7 @@ TEST_F(ParserInitialCTest, createTable) {
       "c13 NCHAR(30), c15 VARCHAR(50)) "
       "TTL 100 COMMENT 'test create table' SMA(c1, c2, c3)");
 
-  run("CREATE TABLE IF NOT EXISTS test.t1("
+  run("CREATE TABLE IF NOT EXISTS rollup_db.t1("
       "ts TIMESTAMP, c1 INT, c2 INT UNSIGNED, c3 BIGINT, c4 BIGINT UNSIGNED, c5 FLOAT, c6 DOUBLE, c7 BINARY(20), "
       "c8 SMALLINT, c9 SMALLINT UNSIGNED COMMENT 'test column comment', c10 TINYINT, c11 TINYINT UNSIGNED, c12 BOOL, "
       "c13 NCHAR(30), c14 VARCHAR(50)) "
@@ -617,6 +621,21 @@ TEST_F(ParserInitialCTest, createTable) {
   // run("CREATE TABLE IF NOT EXISTS t1 USING st1 TAGS(1, 'wxy', NOW + 1S)");
 }
 
+TEST_F(ParserInitialCTest, createTableSemanticCheck) {
+  useDb("root", "test");
+
+  string sql = "CREATE TABLE st1(ts TIMESTAMP, ";
+  for (int32_t i = 1; i < 4096; ++i) {
+    if (i > 1) {
+      sql.append(", ");
+    }
+    sql.append("c" + to_string(i) + " INT");
+  }
+  sql.append(") TAGS (t1 int)");
+
+  run(sql, TSDB_CODE_PAR_TOO_MANY_COLUMNS);
+}
+
 TEST_F(ParserInitialCTest, createTopic) {
   useDb("root", "test");
 

source/libs/planner/src/planOptimizer.c

@@ -1251,7 +1251,7 @@ static SNode* partTagsCreateWrapperFunc(const char* pFuncName, SNode* pNode) {
   }
 
   strcpy(pFunc->functionName, pFuncName);
-  if (QUERY_NODE_COLUMN == nodeType(pNode)) {
+  if (QUERY_NODE_COLUMN == nodeType(pNode) && COLUMN_TYPE_TBNAME != ((SColumnNode*)pNode)->colType) {
     SColumnNode* pCol = (SColumnNode*)pNode;
     partTagsSetAlias(pFunc->node.aliasName, sizeof(pFunc->node.aliasName), pCol->tableAlias, pCol->colName);
   } else {
@@ -1868,6 +1868,8 @@ static EDealRes mergeProjectionsExpr(SNode** pNode, void* pContext) {
               pCxt->errCode = terrno;
               return DEAL_RES_ERROR;
             }
+            snprintf(((SExprNode*)pExpr)->aliasName, sizeof(((SExprNode*)pExpr)->aliasName), "%s",
+                     ((SExprNode*)*pNode)->aliasName);
             nodesDestroyNode(*pNode);
             *pNode = pExpr;
           }

source/libs/planner/src/planPhysiCreater.c

@@ -552,6 +552,9 @@ static int32_t createSystemTableScanPhysiNode(SPhysiPlanContext* pCxt, SSubplan*
   if (0 == strcmp(pScanLogicNode->tableName.tname, TSDB_INS_TABLE_USER_TABLES) ||
       0 == strcmp(pScanLogicNode->tableName.tname, TSDB_INS_TABLE_USER_TABLE_DISTRIBUTED)) {
     vgroupInfoToNodeAddr(pScanLogicNode->pVgroupList->vgroups, &pSubplan->execNode);
+  } else {
+    pSubplan->execNode.nodeId = MNODE_HANDLE;
+    pSubplan->execNode.epSet = pCxt->pPlanCxt->mgmtEpSet;
   }
   SQueryNodeLoad node = {.addr = {.nodeId = MNODE_HANDLE, .epSet = pCxt->pPlanCxt->mgmtEpSet}, .load = 0};
   taosArrayPush(pCxt->pExecNodeList, &node);

source/libs/planner/src/planSpliter.c

@@ -986,6 +986,10 @@ static bool unionIsChildSubplan(SLogicNode* pLogicNode, int32_t groupId) {
     return ((SExchangeLogicNode*)pLogicNode)->srcGroupId == groupId;
   }
 
+  if (QUERY_NODE_LOGIC_PLAN_MERGE == nodeType(pLogicNode)) {
+    return ((SMergeLogicNode*)pLogicNode)->srcGroupId == groupId;
+  }
+
   SNode* pChild;
   FOREACH(pChild, pLogicNode->pChildren) {
     bool isChild = unionIsChildSubplan((SLogicNode*)pChild, groupId);
@@ -1014,14 +1018,14 @@ static int32_t unionMountSubplan(SLogicSubplan* pParent, SNodeList* pChildren) {
   return TSDB_CODE_SUCCESS;
 }
 
-static SLogicSubplan* unionCreateSubplan(SSplitContext* pCxt, SLogicNode* pNode) {
+static SLogicSubplan* unionCreateSubplan(SSplitContext* pCxt, SLogicNode* pNode, ESubplanType subplanType) {
   SLogicSubplan* pSubplan = (SLogicSubplan*)nodesMakeNode(QUERY_NODE_LOGIC_SUBPLAN);
   if (NULL == pSubplan) {
     return NULL;
   }
   pSubplan->id.queryId = pCxt->queryId;
   pSubplan->id.groupId = pCxt->groupId;
-  pSubplan->subplanType = SUBPLAN_TYPE_SCAN;
+  pSubplan->subplanType = subplanType;
   pSubplan->pNode = pNode;
   pNode->pParent = NULL;
   return pSubplan;
@@ -1035,7 +1039,7 @@ static int32_t unionSplitSubplan(SSplitContext* pCxt, SLogicSubplan* pUnionSubpl
 
   SNode* pChild = NULL;
   FOREACH(pChild, pSplitNode->pChildren) {
-    SLogicSubplan* pNewSubplan = unionCreateSubplan(pCxt, (SLogicNode*)pChild);
+    SLogicSubplan* pNewSubplan = unionCreateSubplan(pCxt, (SLogicNode*)pChild, pUnionSubplan->subplanType);
     code = nodesListMakeStrictAppend(&pUnionSubplan->pChildren, (SNode*)pNewSubplan);
     if (TSDB_CODE_SUCCESS == code) {
       REPLACE_NODE(NULL);

source/libs/planner/test/planOptimizeTest.cpp

@@ -68,6 +68,8 @@ TEST_F(PlanOptimizeTest, PartitionTags) {
   run("SELECT SUM(c1), tag1 FROM st1 GROUP BY tag1");
 
   run("SELECT SUM(c1), tag1 + 10 FROM st1 GROUP BY tag1 + 10");
+
+  run("SELECT SUM(c1), tbname FROM st1 GROUP BY tbname");
 }
 
 TEST_F(PlanOptimizeTest, eliminateProjection) {

source/libs/planner/test/planSetOpTest.cpp

@@ -97,7 +97,15 @@ TEST_F(PlanSetOpTest, unionSubquery) {
   run("SELECT * FROM (SELECT c1, c2 FROM t1 UNION SELECT c1, c2 FROM t1)");
 }
 
-TEST_F(PlanSetOpTest, bug001) {
+TEST_F(PlanSetOpTest, unionWithSubquery) {
+  useDb("root", "test");
+
+  run("SELECT c1 FROM (SELECT c1 FROM st1) UNION SELECT c2 FROM (SELECT c1 AS c2 FROM st2)");
+
+  run("SELECT c1 FROM (SELECT c1 FROM st1 ORDER BY c2) UNION SELECT c1 FROM (SELECT c1 FROM st2)");
+}
+
+TEST_F(PlanSetOpTest, unionDataTypeConversion) {
   useDb("root", "test");
 
   run("SELECT c2 FROM t1 WHERE c1 IS NOT NULL GROUP BY c2 "

source/libs/qworker/src/qworker.c

@@ -931,11 +931,6 @@ int32_t qwProcessDelete(QW_FPARAMS_DEF, SQWMsg *qwMsg, SDeleteRes *pRes) {
 
   ctx.plan = plan;
   
-  /* XXXXXXXXXXXXXXXXXXXX */
-  SReadHandle *phandle = (SReadHandle*)qwMsg->node;
-  phandle->deleteQuery = 1;
-  /* XXXXXXXXXXXXXXXXXXXX */
-  
   code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle, NULL, OPTR_EXEC_MODEL_BATCH);
   if (code) {
     QW_TASK_ELOG("qCreateExecTask failed, code:%x - %s", code, tstrerror(code));

source/libs/scalar/src/scalar.c

@@ -1035,3 +1035,72 @@ _return:
   sclFreeRes(ctx.pRes);
   return code;
 }
+
+int32_t scalarGetOperatorResultType(SDataType left, SDataType right, EOperatorType op, SDataType* pRes) {
+  switch (op) {
+    case OP_TYPE_ADD:
+      if (left.type == TSDB_DATA_TYPE_TIMESTAMP && right.type == TSDB_DATA_TYPE_TIMESTAMP) {
+        qError("invalid op %d, left type:%d, right type:%d", op, left.type, right.type);
+        return TSDB_CODE_TSC_INVALID_OPERATION;
+      }
+      if ((left.type == TSDB_DATA_TYPE_TIMESTAMP && (IS_INTEGER_TYPE(right.type) || right.type == TSDB_DATA_TYPE_BOOL)) ||
+          (right.type == TSDB_DATA_TYPE_TIMESTAMP && (IS_INTEGER_TYPE(left.type) || left.type == TSDB_DATA_TYPE_BOOL))) {
+        pRes->type = TSDB_DATA_TYPE_TIMESTAMP;
+        return TSDB_CODE_SUCCESS;
+      }
+      pRes->type = TSDB_DATA_TYPE_DOUBLE;
+      return TSDB_CODE_SUCCESS;
+    case OP_TYPE_SUB:
+      if ((left.type == TSDB_DATA_TYPE_TIMESTAMP && right.type == TSDB_DATA_TYPE_BIGINT) ||
+          (right.type == TSDB_DATA_TYPE_TIMESTAMP && left.type == TSDB_DATA_TYPE_BIGINT)) {
+        pRes->type = TSDB_DATA_TYPE_TIMESTAMP;
+        return TSDB_CODE_SUCCESS;
+      }
+      pRes->type = TSDB_DATA_TYPE_DOUBLE;
+      return TSDB_CODE_SUCCESS;
+    case OP_TYPE_MULTI:
+      if (left.type == TSDB_DATA_TYPE_TIMESTAMP && right.type == TSDB_DATA_TYPE_TIMESTAMP) {
+        qError("invalid op %d, left type:%d, right type:%d", op, left.type, right.type);
+        return TSDB_CODE_TSC_INVALID_OPERATION;
+      }
+    case OP_TYPE_DIV:
+      if (left.type == TSDB_DATA_TYPE_TIMESTAMP && right.type == TSDB_DATA_TYPE_TIMESTAMP) {
+        qError("invalid op %d, left type:%d, right type:%d", op, left.type, right.type);
+        return TSDB_CODE_TSC_INVALID_OPERATION;
+      }
+    case OP_TYPE_REM:
+    case OP_TYPE_MINUS:
+      pRes->type = TSDB_DATA_TYPE_DOUBLE;
+      return TSDB_CODE_SUCCESS;
+    case OP_TYPE_GREATER_THAN:
+    case OP_TYPE_GREATER_EQUAL:
+    case OP_TYPE_LOWER_THAN:
+    case OP_TYPE_LOWER_EQUAL:
+    case OP_TYPE_EQUAL:
+    case OP_TYPE_NOT_EQUAL:
+    case OP_TYPE_IN:
+    case OP_TYPE_NOT_IN:
+    case OP_TYPE_LIKE:
+    case OP_TYPE_NOT_LIKE:
+    case OP_TYPE_MATCH:
+    case OP_TYPE_NMATCH:
+    case OP_TYPE_IS_NULL:
+    case OP_TYPE_IS_NOT_NULL:
+    case OP_TYPE_IS_TRUE:
+    case OP_TYPE_JSON_CONTAINS:
+      pRes->type = TSDB_DATA_TYPE_BOOL;
+      return TSDB_CODE_SUCCESS;
+    case OP_TYPE_BIT_AND:
+    case OP_TYPE_BIT_OR:
+      pRes->type = TSDB_DATA_TYPE_BIGINT;
+      return TSDB_CODE_SUCCESS;
+    case OP_TYPE_JSON_GET_VALUE:
+      pRes->type = TSDB_DATA_TYPE_JSON;
+      return TSDB_CODE_SUCCESS;
+    default:
+      ASSERT(0);
+      return TSDB_CODE_APP_ERROR;
+  }
+}
+
+

source/libs/scalar/src/sclfunc.c

@@ -360,9 +360,6 @@ static void trtrim(char *input, char *output, int32_t type, int32_t charLen) {
 
 static int32_t doLengthFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput, _len_fn lenFn) {
   int32_t type = GET_PARAM_TYPE(pInput);
-  if (inputNum != 1 || !IS_VAR_DATA_TYPE(type)) {
-    return TSDB_CODE_FAILED;
-  }
 
   SColumnInfoData *pInputData  = pInput->columnData;
   SColumnInfoData *pOutputData = pOutput->columnData;
@@ -586,9 +583,6 @@ DONE:
 
 static int32_t doCaseConvFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput, _conv_fn convFn) {
   int32_t type = GET_PARAM_TYPE(pInput);
-  if (inputNum != 1 || !IS_VAR_DATA_TYPE(type)) {
-    return TSDB_CODE_FAILED;
-  }
 
   SColumnInfoData *pInputData  = pInput->columnData;
   SColumnInfoData *pOutputData = pOutput->columnData;
@@ -628,9 +622,6 @@ static int32_t doCaseConvFunction(SScalarParam *pInput, int32_t inputNum, SScala
 
 static int32_t doTrimFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput, _trim_fn trimFn) {
   int32_t type = GET_PARAM_TYPE(pInput);
-  if (inputNum != 1 || !IS_VAR_DATA_TYPE(type)) {
-    return TSDB_CODE_FAILED;
-  }
 
   SColumnInfoData *pInputData  = pInput->columnData;
   SColumnInfoData *pOutputData = pOutput->columnData;
@@ -664,16 +655,10 @@ static int32_t doTrimFunction(SScalarParam *pInput, int32_t inputNum, SScalarPar
 int32_t substrFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
   int32_t subPos = 0;
   GET_TYPED_DATA(subPos, int32_t, GET_PARAM_TYPE(&pInput[1]), pInput[1].columnData->pData);
-  if (subPos == 0) { //subPos needs to be positive or negative values;
-    return TSDB_CODE_FAILED;
-  }
 
   int32_t subLen = INT16_MAX;
   if (inputNum == 3) {
     GET_TYPED_DATA(subLen, int32_t, GET_PARAM_TYPE(&pInput[2]), pInput[2].columnData->pData);
-    if (subLen < 0 || subLen > INT16_MAX) { //subLen cannot be negative
-      return TSDB_CODE_FAILED;
-    }
     subLen = (GET_PARAM_TYPE(pInput) == TSDB_DATA_TYPE_VARCHAR) ? subLen : subLen * TSDB_NCHAR_SIZE;
   }
 
@@ -1149,13 +1134,6 @@ int32_t toUnixtimestampFunction(SScalarParam *pInput, int32_t inputNum, SScalarP
 
 int32_t toJsonFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
   int32_t type = GET_PARAM_TYPE(pInput);
-  if (type != TSDB_DATA_TYPE_BINARY && type != TSDB_DATA_TYPE_NCHAR) {
-    return TSDB_CODE_FAILED;
-  }
-
-  if (inputNum != 1) {
-    return TSDB_CODE_FAILED;
-  }
 
   char tmp[TSDB_MAX_JSON_TAG_LEN] = {0};
   for (int32_t i = 0; i < pInput[0].numOfRows; ++i) {
@@ -1196,6 +1174,8 @@ int32_t timeTruncateFunction(SScalarParam *pInput, int32_t inputNum, SScalarPara
   int64_t factor = (timePrec == TSDB_TIME_PRECISION_MILLI) ? 1000 :
                    (timePrec == TSDB_TIME_PRECISION_MICRO ? 1000000 : 1000000000);
 
+  timeUnit = timeUnit * 1000 / factor;
+
   for (int32_t i = 0; i < pInput[0].numOfRows; ++i) {
     if (colDataIsNull_s(pInput[0].columnData, i)) {
       colDataAppendNULL(pOutput->columnData, i);
@@ -1228,7 +1208,6 @@ int32_t timeTruncateFunction(SScalarParam *pInput, int32_t inputNum, SScalarPara
     char buf[20] = {0};
     NUM_TO_STRING(TSDB_DATA_TYPE_BIGINT, &timeVal, sizeof(buf), buf);
     int32_t tsDigits = (int32_t)strlen(buf);
-    timeUnit = timeUnit * 1000 / factor;
 
     switch (timeUnit) {
       case 0: { /* 1u */
@@ -1384,6 +1363,11 @@ int32_t timeDiffFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *p
     GET_TYPED_DATA(timePrec, int64_t, GET_PARAM_TYPE(&pInput[2]), pInput[2].columnData->pData);
   }
 
+  int64_t factor = (timePrec == TSDB_TIME_PRECISION_MILLI) ? 1000 :
+                   (timePrec == TSDB_TIME_PRECISION_MICRO ? 1000000 : 1000000000);
+
+  timeUnit = timeUnit * 1000 / factor;
+
   int32_t numOfRows = 0;
   for (int32_t i = 0; i < inputNum; ++i) {
     if (pInput[i].numOfRows > numOfRows) {
@@ -1463,9 +1447,6 @@ int32_t timeDiffFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *p
         }
       }
     } else {
-      int64_t factor = (timePrec == TSDB_TIME_PRECISION_MILLI) ? 1000 :
-                       (timePrec == TSDB_TIME_PRECISION_MICRO ? 1000000 : 1000000000);
-      timeUnit = timeUnit * 1000 / factor;
       switch(timeUnit) {
         case 0: { /* 1u */
           result = result / 1000;

source/libs/scheduler/src/schJob.c

@@ -184,9 +184,7 @@ FORCE_INLINE bool schJobNeedToStop(SSchJob *pJob, int8_t *pStatus) {
   }
 
   if ((*pJob->chkKillFp)(pJob->chkKillParam)) {
-    schUpdateJobStatus(pJob, JOB_TASK_STATUS_DROPPING);
     schUpdateJobErrCode(pJob, TSDB_CODE_TSC_QUERY_KILLED);
-
     return true;
   }  
 
@@ -669,6 +667,11 @@ int32_t schSetTaskCandidateAddrs(SSchJob *pJob, SSchTask *pTask) {
     return TSDB_CODE_SUCCESS;
   }
 
+  if (SCH_IS_DATA_SRC_QRY_TASK(pTask)) {
+    SCH_TASK_ELOG("no execNode specifed for data src task, numOfEps:%d", pTask->plan->execNode.epSet.numOfEps);
+    SCH_ERR_RET(TSDB_CODE_QRY_APP_ERROR);
+  }
+
   SCH_ERR_RET(schSetAddrsFromNodeList(pJob, pTask));
 
   /*
@@ -811,14 +814,6 @@ int32_t schMoveTaskToExecList(SSchJob *pJob, SSchTask *pTask, bool *moved) {
 */
 
 int32_t schTaskCheckSetRetry(SSchJob *pJob, SSchTask *pTask, int32_t errCode, bool *needRetry) {
-  int8_t status = 0;
-
-  if (schJobNeedToStop(pJob, &status)) {
-    *needRetry = false;
-    SCH_TASK_DLOG("task no more retry cause of job status, job status:%s", jobTaskStatusStr(status));
-    return TSDB_CODE_SUCCESS;
-  }
-
   if (TSDB_CODE_SCH_TIMEOUT_ERROR == errCode) {
     pTask->maxExecTimes++;
     if (pTask->timeoutUsec < SCH_MAX_TASK_TIMEOUT_USEC) {
@@ -1277,7 +1272,7 @@ int32_t schProcessOnTaskStatusRsp(SQueryNodeEpId* pEpId, SArray* pStatusList) {
   for (int32_t i = 0; i < taskNum; ++i) {
     STaskStatus *taskStatus = taosArrayGet(pStatusList, i);
 
-    qDebug("QID:%" PRIx64 ",TID:0x%" PRIx64 ",EID:%d task status in server: %s", 
+    qDebug("QID:0x%" PRIx64 ",TID:0x%" PRIx64 ",EID:%d task status in server: %s", 
       taskStatus->queryId, taskStatus->taskId, taskStatus->execId, jobTaskStatusStr(taskStatus->status));
 
     SSchJob *pJob = schAcquireJob(taskStatus->refId);
@@ -1495,6 +1490,8 @@ void schFreeJobImpl(void *job) {
   uint64_t queryId = pJob->queryId;
   int64_t  refId = pJob->refId;
 
+  qDebug("QID:0x%" PRIx64 " begin to free sch job, refId:0x%" PRIx64 ", pointer:%p", queryId, refId, pJob);
+
   if (pJob->status == JOB_TASK_STATUS_EXECUTING) {
     schCancelJob(pJob);
   }
@@ -1535,12 +1532,12 @@ void schFreeJobImpl(void *job) {
   taosMemoryFreeClear(pJob->resData);
   taosMemoryFree(pJob);
 
-  qDebug("QID:0x%" PRIx64 " sch job freed, refId:0x%" PRIx64 ", pointer:%p", queryId, refId, pJob);
-
   int32_t jobNum = atomic_sub_fetch_32(&schMgmt.jobNum, 1);
   if (jobNum == 0) {
     schCloseJobRef();
   }
+
+  qDebug("QID:0x%" PRIx64 " sch job freed, refId:0x%" PRIx64 ", pointer:%p", queryId, refId, pJob);
 }
 
 int32_t schLaunchStaticExplainJob(SSchedulerReq *pReq, SSchJob *pJob, bool sync) {
@@ -1687,11 +1684,6 @@ _return:
 
 int32_t schDoTaskRedirect(SSchJob *pJob, SSchTask *pTask, SDataBuf* pData, int32_t rspCode) {
   int32_t code = 0;
-  int8_t  status = 0;
-  if (schJobNeedToStop(pJob, &status)) {
-    SCH_TASK_ELOG("redirect will no continue cause of job status %s", jobTaskStatusStr(status));
-    SCH_RET(atomic_load_32(&pJob->errCode));
-  }
   
   if ((pTask->execId + 1) >= pTask->maxExecTimes) {
     SCH_TASK_DLOG("task no more retry since reach max try times, execId:%d", pTask->execId);

source/libs/scheduler/src/schRemote.c

@@ -406,12 +406,16 @@ int32_t schHandleCallback(void *param, SDataBuf *pMsg, int32_t rspCode) {
     goto _return;
   }
   
-  code = schHandleResponseMsg(pJob, pTask, msgType, pMsg->pData, pMsg->len, rspCode);
+  schHandleResponseMsg(pJob, pTask, msgType, pMsg->pData, pMsg->len, rspCode);
   pMsg->pData = NULL;
 
 _return:
 
   if (pTask) {
+    if (code) {
+      schProcessOnTaskFailure(pJob, pTask, code);
+    }
+
     SCH_UNLOCK_TASK(pTask);
   }
 

source/libs/scheduler/src/scheduler.c

@@ -225,26 +225,33 @@ void schedulerStopQueryHb(void *pTrans) {
   schCleanClusterHb(pTrans);
 }
 
-void schedulerFreeJob(int64_t job, int32_t errCode) {
+void schedulerFreeJob(int64_t* job, int32_t errCode) {
-  SSchJob *pJob = schAcquireJob(job);
+  if (0 == *job) {
+    return;
+  }
+  
+  SSchJob *pJob = schAcquireJob(*job);
   if (NULL == pJob) {
-    qError("acquire job from jobRef list failed, may be dropped, jobId:0x%" PRIx64, job);
+    qError("acquire sch job failed, may be dropped, jobId:0x%" PRIx64, *job);
+    *job = 0;
     return;
   }
 
   int32_t code = schProcessOnJobDropped(pJob, errCode);
   if (TSDB_CODE_SCH_JOB_IS_DROPPING == code) {
-    SCH_JOB_DLOG("sch job is already dropping, refId:0x%" PRIx64, job);
+    SCH_JOB_DLOG("sch job is already dropping, refId:0x%" PRIx64, *job);
+    *job = 0;
     return;
   }
 
-  SCH_JOB_DLOG("start to remove job from jobRef list, refId:0x%" PRIx64, job);
+  SCH_JOB_DLOG("start to remove job from jobRef list, refId:0x%" PRIx64, *job);
 
-  if (taosRemoveRef(schMgmt.jobRef, job)) {
+  if (taosRemoveRef(schMgmt.jobRef, *job)) {
-    SCH_JOB_ELOG("remove job from job list failed, refId:0x%" PRIx64, job);
+    SCH_JOB_ELOG("remove job from job list failed, refId:0x%" PRIx64, *job);
   }
 
-  schReleaseJob(job);
+  schReleaseJob(*job);
+  *job = 0;
 }
 
 void schedulerDestroy(void) {

source/libs/scheduler/test/schedulerTests.cpp

@@ -457,7 +457,7 @@ void schtFreeQueryJob(int32_t freeThread) {
   int64_t job = queryJobRefId;
   
   if (job && atomic_val_compare_exchange_64(&queryJobRefId, job, 0)) {
-    schedulerFreeJob(job, 0);
+    schedulerFreeJob(&job, 0);
     if (freeThread) {
       if (++freeNum % schtTestPrintNum == 0) {
         printf("FreeNum:%d\n", freeNum);
@@ -724,7 +724,7 @@ TEST(queryTest, normalCase) {
 
   schReleaseJob(job);
 
-  schedulerFreeJob(job, 0);
+  schedulerFreeJob(&job, 0);
 
   schtFreeQueryDag(&dag);
 
@@ -828,7 +828,7 @@ TEST(queryTest, readyFirstCase) {
 
   schReleaseJob(job);
 
-  schedulerFreeJob(job, 0);
+  schedulerFreeJob(&job, 0);
 
   schtFreeQueryDag(&dag);
 
@@ -940,7 +940,7 @@ TEST(queryTest, flowCtrlCase) {
 
   schReleaseJob(job);
 
-  schedulerFreeJob(job, 0);
+  schedulerFreeJob(&job, 0);
 
   schtFreeQueryDag(&dag);
 
@@ -994,7 +994,7 @@ TEST(insertTest, normalCase) {
   ASSERT_EQ(code, 0);
   ASSERT_EQ(res.numOfRows, 20);
 
-  schedulerFreeJob(insertJobRefId, 0);
+  schedulerFreeJob(&insertJobRefId, 0);
 
   schedulerDestroy();  
 }

source/libs/sync/inc/syncAppendEntriesReply.h

@@ -44,11 +44,6 @@ int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* p
 int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
 int32_t syncNodeOnAppendEntriesReplySnapshot2Cb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
 
-typedef struct SReaderParam {
-  SyncIndex start;
-  SyncIndex end;
-} SReaderParam;
-
 #ifdef __cplusplus
 }
 #endif

source/libs/sync/inc/syncInt.h

@@ -175,6 +175,7 @@ int32_t    syncNodeProposeBatch(SSyncNode* pSyncNode, SRpcMsg* pMsgArr, bool* pI
 
 // option
 bool          syncNodeSnapshotEnable(SSyncNode* pSyncNode);
+ESyncStrategy syncNodeStrategy(SSyncNode* pSyncNode);
 SyncIndex     syncNodeGetSnapshotConfigIndex(SSyncNode* pSyncNode, SyncIndex snapshotLastApplyIndex);
 
 // ping --------------

source/libs/sync/inc/syncRaftEntry.h

@@ -29,19 +29,20 @@ extern "C" {
 
 typedef struct SSyncRaftEntry {
   uint32_t  bytes;
-  uint32_t  msgType;          // SyncClientRequest msgType
+  uint32_t  msgType;          // TDMT_SYNC_CLIENT_REQUEST
-  uint32_t  originalRpcType;  // user RpcMsg msgType
+  uint32_t  originalRpcType;  // origin RpcMsg msgType
   uint64_t  seqNum;
   bool      isWeak;
   SyncTerm  term;
   SyncIndex index;
-  uint32_t  dataLen;  // user RpcMsg.contLen
+  uint32_t  dataLen;  // origin RpcMsg.contLen
-  char      data[];   // user RpcMsg.pCont
+  char      data[];   // origin RpcMsg.pCont
 } SSyncRaftEntry;
 
 SSyncRaftEntry* syncEntryBuild(uint32_t dataLen);
 SSyncRaftEntry* syncEntryBuild2(SyncClientRequest* pMsg, SyncTerm term, SyncIndex index);  // step 4
 SSyncRaftEntry* syncEntryBuild3(SyncClientRequest* pMsg, SyncTerm term, SyncIndex index);
+SSyncRaftEntry* syncEntryBuild4(SRpcMsg* pOriginalMsg, SyncTerm term, SyncIndex index);
 SSyncRaftEntry* syncEntryBuildNoop(SyncTerm term, SyncIndex index, int32_t vgId);
 void            syncEntryDestory(SSyncRaftEntry* pEntry);
 char*           syncEntrySerialize(const SSyncRaftEntry* pEntry, uint32_t* len);  // step 5

source/libs/sync/inc/syncReplication.h

@@ -54,6 +54,7 @@ extern "C" {
 int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode);
 int32_t syncNodeAppendEntriesPeersSnapshot(SSyncNode* pSyncNode);
 int32_t syncNodeAppendEntriesPeersSnapshot2(SSyncNode* pSyncNode);
+
 int32_t syncNodeReplicate(SSyncNode* pSyncNode);
 int32_t syncNodeAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncAppendEntries* pMsg);
 int32_t syncNodeAppendEntriesBatch(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncAppendEntriesBatch* pMsg);

source/libs/sync/inc/syncSnapshot.h

@@ -43,6 +43,7 @@ typedef struct SSyncSnapshotSender {
   void          *pReader;
   void          *pCurrentBlock;
   int32_t        blockLen;
+  SSnapshotParam snapshotParam;
   SSnapshot      snapshot;
   SSyncCfg       lastConfig;
   int64_t        sendingMS;
@@ -56,7 +57,8 @@ typedef struct SSyncSnapshotSender {
 SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaIndex);
 void                 snapshotSenderDestroy(SSyncSnapshotSender *pSender);
 bool                 snapshotSenderIsStart(SSyncSnapshotSender *pSender);
-int32_t              snapshotSenderStart(SSyncSnapshotSender *pSender, SSnapshot snapshot, void *pReader);
+int32_t              snapshotSenderStart(SSyncSnapshotSender *pSender, SSnapshotParam snapshotParam, SSnapshot snapshot,
+                                         void *pReader);
 int32_t              snapshotSenderStop(SSyncSnapshotSender *pSender, bool finish);
 int32_t              snapshotSend(SSyncSnapshotSender *pSender);
 int32_t              snapshotReSend(SSyncSnapshotSender *pSender);
@@ -72,6 +74,7 @@ typedef struct SSyncSnapshotReceiver {
   void          *pWriter;
   SyncTerm       term;
   SyncTerm       privateTerm;
+  SSnapshotParam snapshotParam;
   SSnapshot      snapshot;
   SRaftId        fromId;
   SSyncNode     *pSyncNode;

source/libs/sync/src/syncAppendEntries.c

@@ -628,8 +628,6 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
 
 #endif
 
-static int32_t syncNodeMakeLogSame2(SSyncNode* ths, SyncAppendEntriesBatch* pMsg) { return 0; }
-
 static int32_t syncNodeMakeLogSame(SSyncNode* ths, SyncAppendEntries* pMsg) {
   int32_t code;
 
@@ -675,6 +673,51 @@ static int32_t syncNodeMakeLogSame(SSyncNode* ths, SyncAppendEntries* pMsg) {
   return code;
 }
 
+static int32_t syncNodeDoMakeLogSame(SSyncNode* ths, SyncIndex FromIndex) {
+  int32_t code;
+
+  SyncIndex delBegin = FromIndex;
+  SyncIndex delEnd = ths->pLogStore->syncLogLastIndex(ths->pLogStore);
+
+  // invert roll back!
+  for (SyncIndex index = delEnd; index >= delBegin; --index) {
+    if (ths->pFsm->FpRollBackCb != NULL) {
+      SSyncRaftEntry* pRollBackEntry;
+      code = ths->pLogStore->syncLogGetEntry(ths->pLogStore, index, &pRollBackEntry);
+      ASSERT(code == 0);
+      ASSERT(pRollBackEntry != NULL);
+
+      if (syncUtilUserRollback(pRollBackEntry->msgType)) {
+        SRpcMsg rpcMsg;
+        syncEntry2OriginalRpc(pRollBackEntry, &rpcMsg);
+
+        SFsmCbMeta cbMeta = {0};
+        cbMeta.index = pRollBackEntry->index;
+        cbMeta.lastConfigIndex = syncNodeGetSnapshotConfigIndex(ths, cbMeta.index);
+        cbMeta.isWeak = pRollBackEntry->isWeak;
+        cbMeta.code = 0;
+        cbMeta.state = ths->state;
+        cbMeta.seqNum = pRollBackEntry->seqNum;
+        ths->pFsm->FpRollBackCb(ths->pFsm, &rpcMsg, cbMeta);
+        rpcFreeCont(rpcMsg.pCont);
+      }
+
+      syncEntryDestory(pRollBackEntry);
+    }
+  }
+
+  // delete confict entries
+  code = ths->pLogStore->syncLogTruncate(ths->pLogStore, delBegin);
+  ASSERT(code == 0);
+
+  char eventLog[128];
+  snprintf(eventLog, sizeof(eventLog), "log truncate, from %ld to %ld", delBegin, delEnd);
+  syncNodeEventLog(ths, eventLog);
+  logStoreSimpleLog2("after syncNodeMakeLogSame", ths->pLogStore);
+
+  return code;
+}
+
 static int32_t syncNodePreCommit(SSyncNode* ths, SSyncRaftEntry* pEntry) {
   SRpcMsg rpcMsg;
   syncEntry2OriginalRpc(pEntry, &rpcMsg);
@@ -694,6 +737,31 @@ static int32_t syncNodePreCommit(SSyncNode* ths, SSyncRaftEntry* pEntry) {
   return 0;
 }
 
+static bool syncNodeOnAppendEntriesBatchLogOK(SSyncNode* pSyncNode, SyncAppendEntriesBatch* pMsg) {
+  if (pMsg->prevLogIndex == SYNC_INDEX_INVALID) {
+    return true;
+  }
+
+  SyncIndex myLastIndex = syncNodeGetLastIndex(pSyncNode);
+  if (pMsg->prevLogIndex > myLastIndex) {
+    sDebug("vgId:%d sync log not ok, preindex:%ld", pSyncNode->vgId, pMsg->prevLogIndex);
+    return false;
+  }
+
+  SyncTerm myPreLogTerm = syncNodeGetPreTerm(pSyncNode, pMsg->prevLogIndex + 1);
+  if (myPreLogTerm == SYNC_TERM_INVALID) {
+    sDebug("vgId:%d sync log not ok2, preindex:%ld", pSyncNode->vgId, pMsg->prevLogIndex);
+    return false;
+  }
+
+  if (pMsg->prevLogIndex <= myLastIndex && pMsg->prevLogTerm == myPreLogTerm) {
+    return true;
+  }
+
+  sDebug("vgId:%d sync log not ok3, preindex:%ld", pSyncNode->vgId, pMsg->prevLogIndex);
+  return false;
+}
+
 // really pre log match
 // prevLogIndex == -1
 static bool syncNodeOnAppendEntriesLogOK(SSyncNode* pSyncNode, SyncAppendEntries* pMsg) {
@@ -767,7 +835,6 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
   // operation:
   // if hasAppendEntries && pMsg->prevLogIndex == ths->commitIndex, append entry
   // match my-commit-index or my-commit-index + 1
-  // no operation on log
   do {
     bool condition = (pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) &&
                      (pMsg->prevLogIndex <= ths->commitIndex);
@@ -782,12 +849,9 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
 
       SyncIndex          matchIndex = ths->commitIndex;
       bool               hasAppendEntries = pMsg->dataLen > 0;
-      if (hasAppendEntries && pMsg->prevLogIndex == ths->commitIndex) {
+      SOffsetAndContLen* metaTableArr = syncAppendEntriesBatchMetaTableArray(pMsg);
-        SRpcMsg rpcMsgArr[SYNC_MAX_BATCH_SIZE];
-        memset(rpcMsgArr, 0, sizeof(rpcMsgArr));
-        int32_t retArrSize = 0;
-        syncAppendEntriesBatch2RpcMsgArray(pMsg, rpcMsgArr, SYNC_MAX_BATCH_SIZE, &retArrSize);
 
+      if (hasAppendEntries && pMsg->prevLogIndex == ths->commitIndex) {
         // make log same
         do {
           SyncIndex logLastIndex = ths->pLogStore->syncLogLastIndex(ths->pLogStore);
@@ -795,15 +859,15 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
 
           if (hasExtraEntries) {
             // make log same, rollback deleted entries
-            code = syncNodeMakeLogSame2(ths, pMsg);
+            code = syncNodeDoMakeLogSame(ths, pMsg->prevLogIndex + 1);
             ASSERT(code == 0);
           }
 
         } while (0);
 
         // append entry batch
-        for (int32_t i = 0; i < retArrSize; ++i) {
+        for (int32_t i = 0; i < pMsg->dataCount; ++i) {
-          SSyncRaftEntry* pAppendEntry = syncEntryBuild(1234);
+          SSyncRaftEntry* pAppendEntry = (SSyncRaftEntry*)(pMsg->data + metaTableArr[i].offset);
           code = ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pAppendEntry);
           if (code != 0) {
             return -1;
@@ -821,7 +885,7 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
         walFsync(pWal, true);
 
         // update match index
-        matchIndex = pMsg->prevLogIndex + retArrSize;
+        matchIndex = pMsg->prevLogIndex + pMsg->dataCount;
       }
 
       // prepare response msg
@@ -839,13 +903,12 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
       syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
       syncAppendEntriesReplyDestroy(pReply);
 
-      return ret;
+      return 0;
     }
   } while (0);
 
   // calculate logOK here, before will coredump, due to fake match
-  // bool logOK = syncNodeOnAppendEntriesLogOK(ths, pMsg);
+  bool logOK = syncNodeOnAppendEntriesBatchLogOK(ths, pMsg);
-  bool logOK = true;
 
   // not match
   //
@@ -866,8 +929,9 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
 
     if (condition) {
       char logBuf[128];
-      snprintf(logBuf, sizeof(logBuf), "recv sync-append-entries, not match, pre-index:%ld, pre-term:%lu, datalen:%d",
+      snprintf(logBuf, sizeof(logBuf),
-               pMsg->prevLogIndex, pMsg->prevLogTerm, pMsg->dataLen);
+               "recv sync-append-entries-batch, not match, pre-index:%ld, pre-term:%lu, datalen:%d", pMsg->prevLogIndex,
+               pMsg->prevLogTerm, pMsg->dataLen);
       syncNodeEventLog(ths, logBuf);
 
       // prepare response msg
@@ -885,7 +949,7 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
       syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
       syncAppendEntriesReplyDestroy(pReply);
 
-      return ret;
+      return 0;
     }
   } while (0);
 
@@ -906,27 +970,25 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
 
       // has entries in SyncAppendEntries msg
       bool               hasAppendEntries = pMsg->dataLen > 0;
+      SOffsetAndContLen* metaTableArr = syncAppendEntriesBatchMetaTableArray(pMsg);
 
+      do {
         char logBuf[128];
         snprintf(logBuf, sizeof(logBuf), "recv sync-append-entries, match, pre-index:%ld, pre-term:%lu, datalen:%d",
                  pMsg->prevLogIndex, pMsg->prevLogTerm, pMsg->dataLen);
         syncNodeEventLog(ths, logBuf);
+      } while (0);
 
       if (hasExtraEntries) {
         // make log same, rollback deleted entries
-        // code = syncNodeMakeLogSame(ths, pMsg);
+        code = syncNodeDoMakeLogSame(ths, pMsg->prevLogIndex + 1);
         ASSERT(code == 0);
       }
 
-      int32_t retArrSize = 0;
       if (hasAppendEntries) {
-        SRpcMsg rpcMsgArr[SYNC_MAX_BATCH_SIZE];
-        memset(rpcMsgArr, 0, sizeof(rpcMsgArr));
-        syncAppendEntriesBatch2RpcMsgArray(pMsg, rpcMsgArr, SYNC_MAX_BATCH_SIZE, &retArrSize);
-
         // append entry batch
-        for (int32_t i = 0; i < retArrSize; ++i) {
+        for (int32_t i = 0; i < pMsg->dataCount; ++i) {
-          SSyncRaftEntry* pAppendEntry = syncEntryBuild(1234);
+          SSyncRaftEntry* pAppendEntry = (SSyncRaftEntry*)(pMsg->data + metaTableArr[i].offset);
           code = ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pAppendEntry);
           if (code != 0) {
             return -1;
@@ -951,7 +1013,7 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
       pReply->term = ths->pRaftStore->currentTerm;
       pReply->privateTerm = ths->pNewNodeReceiver->privateTerm;
       pReply->success = true;
-      pReply->matchIndex = hasAppendEntries ? pMsg->prevLogIndex + retArrSize : pMsg->prevLogIndex;
+      pReply->matchIndex = hasAppendEntries ? pMsg->prevLogIndex + pMsg->dataCount : pMsg->prevLogIndex;
 
       // send response
       SRpcMsg rpcMsg;
@@ -991,11 +1053,11 @@ int32_t syncNodeOnAppendEntriesSnapshot2Cb(SSyncNode* ths, SyncAppendEntriesBatc
           ASSERT(code == 0);
         }
       }
-      return ret;
+      return 0;
     }
   } while (0);
 
-  return ret;
+  return 0;
 }
 
 int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg) {

source/libs/sync/src/syncAppendEntriesReply.c

@@ -119,11 +119,11 @@ static void syncNodeStartSnapshot(SSyncNode* ths, SyncIndex beginIndex, SyncInde
       .data = NULL, .lastApplyIndex = endIndex, .lastApplyTerm = lastApplyTerm, .lastConfigIndex = SYNC_INDEX_INVALID};
 
   void*          pReader = NULL;
-  SReaderParam readerParam = {.start = beginIndex, .end = endIndex};
+  SSnapshotParam readerParam = {.start = beginIndex, .end = endIndex};
   ths->pFsm->FpSnapshotStartRead(ths->pFsm, &readerParam, &pReader);
   if (!snapshotSenderIsStart(pSender) && pMsg->privateTerm < pSender->privateTerm) {
     ASSERT(pReader != NULL);
-    snapshotSenderStart(pSender, snapshot, pReader);
+    snapshotSenderStart(pSender, readerParam, snapshot, pReader);
 
   } else {
     if (pReader != NULL) {
@@ -165,23 +165,22 @@ int32_t syncNodeOnAppendEntriesReplySnapshot2Cb(SSyncNode* ths, SyncAppendEntrie
 
     if (ths->pLogStore->syncLogExist(ths->pLogStore, newNextIndex) &&
         ths->pLogStore->syncLogExist(ths->pLogStore, newNextIndex - 1)) {
-      // nextIndex'  = [nextIndex  EXCEPT ![i][j] = m.mmatchIndex + 1]
+      // update next-index, match-index
       syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), newNextIndex);
-
-      // matchIndex' = [matchIndex EXCEPT ![i][j] = m.mmatchIndex]
       syncIndexMgrSetIndex(ths->pMatchIndex, &(pMsg->srcId), newMatchIndex);
 
       // maybe commit
       if (ths->state == TAOS_SYNC_STATE_LEADER) {
         syncMaybeAdvanceCommitIndex(ths);
       }
+
     } else {
       // start snapshot <match+1, old snapshot.end>
-      SSnapshot snapshot;
+      SSnapshot oldSnapshot;
-      ths->pFsm->FpGetSnapshotInfo(ths->pFsm, &snapshot);
+      ths->pFsm->FpGetSnapshotInfo(ths->pFsm, &oldSnapshot);
-      syncNodeStartSnapshot(ths, newMatchIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm, pMsg);
+      syncNodeStartSnapshot(ths, newMatchIndex + 1, oldSnapshot.lastApplyIndex, oldSnapshot.lastApplyTerm, pMsg);
 
-      syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), snapshot.lastApplyIndex + 1);
+      syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), oldSnapshot.lastApplyIndex + 1);
       syncIndexMgrSetIndex(ths->pMatchIndex, &(pMsg->srcId), newMatchIndex);
     }
 
@@ -301,7 +300,8 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
           !snapshotSenderIsStart(pSender) && pMsg->privateTerm < pSender->privateTerm) {
         // has snapshot
         ASSERT(pReader != NULL);
-        snapshotSenderStart(pSender, snapshot, pReader);
+        SSnapshotParam readerParam = {.start = 0, .end = snapshot.lastApplyIndex};
+        snapshotSenderStart(pSender, readerParam, snapshot, pReader);
 
       } else {
         // no snapshot

source/libs/sync/src/syncMain.c

@@ -815,7 +815,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
     // create a new raft config file
     SRaftCfgMeta meta;
     meta.isStandBy = pSyncInfo->isStandBy;
-    meta.snapshotEnable = pSyncInfo->snapshotEnable;
+    meta.snapshotEnable = pSyncInfo->snapshotStrategy;
     meta.lastConfigIndex = SYNC_INDEX_INVALID;
     ret = raftCfgCreateFile((SSyncCfg*)&(pSyncInfo->syncCfg), meta, pSyncNode->configPath);
     ASSERT(ret == 0);
@@ -1100,7 +1100,9 @@ void syncNodeClose(SSyncNode* pSyncNode) {
 }
 
 // option
-bool syncNodeSnapshotEnable(SSyncNode* pSyncNode) { return pSyncNode->pRaftCfg->snapshotEnable; }
+// bool syncNodeSnapshotEnable(SSyncNode* pSyncNode) { return pSyncNode->pRaftCfg->snapshotEnable; }
+
+ESyncStrategy syncNodeStrategy(SSyncNode* pSyncNode) { return pSyncNode->pRaftCfg->snapshotEnable; }
 
 // ping --------------
 int32_t syncNodePing(SSyncNode* pSyncNode, const SRaftId* destRaftId, SyncPing* pMsg) {

source/libs/sync/src/syncMessage.c

@@ -15,6 +15,7 @@
 
 #include "syncMessage.h"
 #include "syncRaftCfg.h"
+#include "syncRaftEntry.h"
 #include "syncUtil.h"
 #include "tcoding.h"
 
@@ -996,7 +997,135 @@ SyncClientRequestBatch* syncClientRequestBatchBuild(SRpcMsg* rpcMsgArr, SRaftMet
   return pMsg;
 }
 
-void syncClientRequestBatch2RpcMsg(const SyncClientRequestBatch* pSyncMsg, SRpcMsg* pRpcMsg) {}
+void syncClientRequestBatch2RpcMsg(const SyncClientRequestBatch* pSyncMsg, SRpcMsg* pRpcMsg) {
+  memset(pRpcMsg, 0, sizeof(*pRpcMsg));
+  pRpcMsg->msgType = pSyncMsg->msgType;
+  pRpcMsg->contLen = pSyncMsg->bytes;
+  pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
+  memcpy(pRpcMsg->pCont, pSyncMsg, pRpcMsg->contLen);
+}
+
+void syncClientRequestBatchDestroy(SyncClientRequestBatch* pMsg) {
+  if (pMsg != NULL) {
+    taosMemoryFree(pMsg);
+  }
+}
+
+void syncClientRequestBatchDestroyDeep(SyncClientRequestBatch* pMsg) {
+  if (pMsg != NULL) {
+    int32_t  arrSize = pMsg->dataCount;
+    int32_t  raftMetaArrayLen = sizeof(SRaftMeta) * arrSize;
+    SRpcMsg* msgArr = (SRpcMsg*)((char*)(pMsg->data) + raftMetaArrayLen);
+    for (int i = 0; i < arrSize; ++i) {
+      if (msgArr[i].pCont != NULL) {
+        rpcFreeCont(msgArr[i].pCont);
+      }
+    }
+
+    taosMemoryFree(pMsg);
+  }
+}
+
+SRaftMeta* syncClientRequestBatchMetaArr(const SyncClientRequestBatch* pSyncMsg) {
+  SRaftMeta* raftMetaArr = (SRaftMeta*)(pSyncMsg->data);
+  return raftMetaArr;
+}
+
+SRpcMsg* syncClientRequestBatchRpcMsgArr(const SyncClientRequestBatch* pSyncMsg) {
+  int32_t  arrSize = pSyncMsg->dataCount;
+  int32_t  raftMetaArrayLen = sizeof(SRaftMeta) * arrSize;
+  SRpcMsg* msgArr = (SRpcMsg*)((char*)(pSyncMsg->data) + raftMetaArrayLen);
+  return msgArr;
+}
+
+SyncClientRequestBatch* syncClientRequestBatchFromRpcMsg(const SRpcMsg* pRpcMsg) {
+  SyncClientRequestBatch* pSyncMsg = taosMemoryMalloc(pRpcMsg->contLen);
+  ASSERT(pSyncMsg != NULL);
+  memcpy(pSyncMsg, pRpcMsg->pCont, pRpcMsg->contLen);
+  ASSERT(pRpcMsg->contLen == pSyncMsg->bytes);
+
+  return pSyncMsg;
+}
+
+cJSON* syncClientRequestBatch2Json(const SyncClientRequestBatch* pMsg) {
+  char   u64buf[128] = {0};
+  cJSON* pRoot = cJSON_CreateObject();
+
+  if (pMsg != NULL) {
+    cJSON_AddNumberToObject(pRoot, "bytes", pMsg->bytes);
+    cJSON_AddNumberToObject(pRoot, "vgId", pMsg->vgId);
+    cJSON_AddNumberToObject(pRoot, "msgType", pMsg->msgType);
+    cJSON_AddNumberToObject(pRoot, "dataLen", pMsg->dataLen);
+    cJSON_AddNumberToObject(pRoot, "dataCount", pMsg->dataCount);
+
+    SRaftMeta* metaArr = syncClientRequestBatchMetaArr(pMsg);
+    SRpcMsg*   msgArr = syncClientRequestBatchRpcMsgArr(pMsg);
+
+    cJSON* pMetaArr = cJSON_CreateArray();
+    cJSON_AddItemToObject(pRoot, "metaArr", pMetaArr);
+    for (int i = 0; i < pMsg->dataCount; ++i) {
+      cJSON* pMeta = cJSON_CreateObject();
+      cJSON_AddNumberToObject(pMeta, "seqNum", metaArr[i].seqNum);
+      cJSON_AddNumberToObject(pMeta, "isWeak", metaArr[i].isWeak);
+      cJSON_AddItemToArray(pMetaArr, pMeta);
+    }
+
+    cJSON* pMsgArr = cJSON_CreateArray();
+    cJSON_AddItemToObject(pRoot, "msgArr", pMsgArr);
+    for (int i = 0; i < pMsg->dataCount; ++i) {
+      cJSON* pRpcMsgJson = syncRpcMsg2Json(&msgArr[i]);
+      cJSON_AddItemToArray(pMsgArr, pRpcMsgJson);
+    }
+
+    char* s;
+    s = syncUtilprintBin((char*)(pMsg->data), pMsg->dataLen);
+    cJSON_AddStringToObject(pRoot, "data", s);
+    taosMemoryFree(s);
+    s = syncUtilprintBin2((char*)(pMsg->data), pMsg->dataLen);
+    cJSON_AddStringToObject(pRoot, "data2", s);
+    taosMemoryFree(s);
+  }
+
+  cJSON* pJson = cJSON_CreateObject();
+  cJSON_AddItemToObject(pJson, "SyncClientRequestBatch", pRoot);
+  return pJson;
+}
+
+char* syncClientRequestBatch2Str(const SyncClientRequestBatch* pMsg) {
+  cJSON* pJson = syncClientRequestBatch2Json(pMsg);
+  char*  serialized = cJSON_Print(pJson);
+  cJSON_Delete(pJson);
+  return serialized;
+}
+
+// for debug ----------------------
+void syncClientRequestBatchPrint(const SyncClientRequestBatch* pMsg) {
+  char* serialized = syncClientRequestBatch2Str(pMsg);
+  printf("syncClientRequestBatchPrint | len:%lu | %s \n", strlen(serialized), serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void syncClientRequestBatchPrint2(char* s, const SyncClientRequestBatch* pMsg) {
+  char* serialized = syncClientRequestBatch2Str(pMsg);
+  printf("syncClientRequestBatchPrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void syncClientRequestBatchLog(const SyncClientRequestBatch* pMsg) {
+  char* serialized = syncClientRequestBatch2Str(pMsg);
+  sTrace("syncClientRequestBatchLog | len:%lu | %s", strlen(serialized), serialized);
+  taosMemoryFree(serialized);
+}
+
+void syncClientRequestBatchLog2(char* s, const SyncClientRequestBatch* pMsg) {
+  if (gRaftDetailLog) {
+    char* serialized = syncClientRequestBatch2Str(pMsg);
+    sTraceLong("syncClientRequestBatchLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
+}
 
 // ---- message process SyncRequestVote----
 SyncRequestVote* syncRequestVoteBuild(int32_t vgId) {
@@ -1472,21 +1601,20 @@ void syncAppendEntriesLog2(char* s, const SyncAppendEntries* pMsg) {
 
 // block1: SOffsetAndContLen
 // block2: SOffsetAndContLen Array
-// block3: SRpcMsg Array
+// block3: entry Array
-// block4: SRpcMsg pCont Array
 
-SyncAppendEntriesBatch* syncAppendEntriesBatchBuild(SRpcMsg* rpcMsgArr, int32_t arrSize, int32_t vgId) {
+SyncAppendEntriesBatch* syncAppendEntriesBatchBuild(SSyncRaftEntry** entryPArr, int32_t arrSize, int32_t vgId) {
-  ASSERT(rpcMsgArr != NULL);
+  ASSERT(entryPArr != NULL);
   ASSERT(arrSize > 0);
 
   int32_t dataLen = 0;
   int32_t metaArrayLen = sizeof(SOffsetAndContLen) * arrSize;  // <offset, contLen>
-  int32_t rpcArrayLen = sizeof(SRpcMsg) * arrSize;             // SRpcMsg
+  int32_t entryArrayLen = 0;
-  int32_t contArrayLen = 0;
   for (int i = 0; i < arrSize; ++i) {  // SRpcMsg pCont
-    contArrayLen += rpcMsgArr[i].contLen;
+    SSyncRaftEntry* pEntry = entryPArr[i];
+    entryArrayLen += pEntry->bytes;
   }
-  dataLen += (metaArrayLen + rpcArrayLen + contArrayLen);
+  dataLen += (metaArrayLen + entryArrayLen);
 
   uint32_t                bytes = sizeof(SyncAppendEntriesBatch) + dataLen;
   SyncAppendEntriesBatch* pMsg = taosMemoryMalloc(bytes);
@@ -1498,30 +1626,30 @@ SyncAppendEntriesBatch* syncAppendEntriesBatchBuild(SRpcMsg* rpcMsgArr, int32_t
   pMsg->dataLen = dataLen;
 
   SOffsetAndContLen* metaArr = (SOffsetAndContLen*)(pMsg->data);
-  SRpcMsg*           msgArr = (SRpcMsg*)((char*)(pMsg->data) + metaArrayLen);
   char*              pData = pMsg->data;
 
   for (int i = 0; i < arrSize; ++i) {
-    // init <offset, contLen>
+    // init meta <offset, contLen>
     if (i == 0) {
-      metaArr[i].offset = metaArrayLen + rpcArrayLen;
+      metaArr[i].offset = metaArrayLen;
-      metaArr[i].contLen = rpcMsgArr[i].contLen;
+      metaArr[i].contLen = entryPArr[i]->bytes;
     } else {
       metaArr[i].offset = metaArr[i - 1].offset + metaArr[i - 1].contLen;
-      metaArr[i].contLen = rpcMsgArr[i].contLen;
+      metaArr[i].contLen = entryPArr[i]->bytes;
     }
 
-    // init msgArr
+    // init entry array
-    msgArr[i] = rpcMsgArr[i];
+    ASSERT(metaArr[i].contLen == entryPArr[i]->bytes);
-
+    memcpy(pData + metaArr[i].offset, entryPArr[i], metaArr[i].contLen);
-    // init data
-    ASSERT(rpcMsgArr[i].contLen == metaArr[i].contLen);
-    memcpy(pData + metaArr[i].offset, rpcMsgArr[i].pCont, rpcMsgArr[i].contLen);
   }
 
   return pMsg;
 }
 
+SOffsetAndContLen* syncAppendEntriesBatchMetaTableArray(SyncAppendEntriesBatch* pMsg) {
+  return (SOffsetAndContLen*)(pMsg->data);
+}
+
 void syncAppendEntriesBatchDestroy(SyncAppendEntriesBatch* pMsg) {
   if (pMsg != NULL) {
     taosMemoryFree(pMsg);
@@ -1634,16 +1762,12 @@ cJSON* syncAppendEntriesBatch2Json(const SyncAppendEntriesBatch* pMsg) {
     cJSON_AddNumberToObject(pRoot, "dataLen", pMsg->dataLen);
 
     int32_t metaArrayLen = sizeof(SOffsetAndContLen) * pMsg->dataCount;  // <offset, contLen>
-    int32_t rpcArrayLen = sizeof(SRpcMsg) * pMsg->dataCount;             // SRpcMsg
+    int32_t entryArrayLen = pMsg->dataLen - metaArrayLen;
-    int32_t contArrayLen = pMsg->dataLen - metaArrayLen - rpcArrayLen;
 
     cJSON_AddNumberToObject(pRoot, "metaArrayLen", metaArrayLen);
-    cJSON_AddNumberToObject(pRoot, "rpcArrayLen", rpcArrayLen);
+    cJSON_AddNumberToObject(pRoot, "entryArrayLen", entryArrayLen);
-    cJSON_AddNumberToObject(pRoot, "contArrayLen", contArrayLen);
 
     SOffsetAndContLen* metaArr = (SOffsetAndContLen*)(pMsg->data);
-    SRpcMsg*           msgArr = (SRpcMsg*)(pMsg->data + metaArrayLen);
-    void*              pData = (void*)(pMsg->data + metaArrayLen + rpcArrayLen);
 
     cJSON* pMetaArr = cJSON_CreateArray();
     cJSON_AddItemToObject(pRoot, "metaArr", pMetaArr);
@@ -1654,14 +1778,12 @@ cJSON* syncAppendEntriesBatch2Json(const SyncAppendEntriesBatch* pMsg) {
       cJSON_AddItemToArray(pMetaArr, pMeta);
     }
 
-    cJSON* pMsgArr = cJSON_CreateArray();
+    cJSON* pEntryArr = cJSON_CreateArray();
-    cJSON_AddItemToObject(pRoot, "msgArr", pMsgArr);
+    cJSON_AddItemToObject(pRoot, "entryArr", pEntryArr);
     for (int i = 0; i < pMsg->dataCount; ++i) {
-      cJSON* pRpcMsgJson = cJSON_CreateObject();
+      SSyncRaftEntry* pEntry = (SSyncRaftEntry*)(pMsg->data + metaArr[i].offset);
-      cJSON_AddNumberToObject(pRpcMsgJson, "code", msgArr[i].code);
+      cJSON*          pEntryJson = syncEntry2Json(pEntry);
-      cJSON_AddNumberToObject(pRpcMsgJson, "contLen", msgArr[i].contLen);
+      cJSON_AddItemToArray(pEntryArr, pEntryJson);
-      cJSON_AddNumberToObject(pRpcMsgJson, "msgType", msgArr[i].msgType);
-      cJSON_AddItemToArray(pMsgArr, pRpcMsgJson);
     }
 
     char* s;
@@ -1685,33 +1807,6 @@ char* syncAppendEntriesBatch2Str(const SyncAppendEntriesBatch* pMsg) {
   return serialized;
 }
 
-void syncAppendEntriesBatch2RpcMsgArray(SyncAppendEntriesBatch* pSyncMsg, SRpcMsg* rpcMsgArr, int32_t maxArrSize,
-                                        int32_t* pRetArrSize) {
-  if (pRetArrSize != NULL) {
-    *pRetArrSize = pSyncMsg->dataCount;
-  }
-
-  int32_t arrSize = pSyncMsg->dataCount;
-  if (arrSize > maxArrSize) {
-    arrSize = maxArrSize;
-  }
-
-  int32_t metaArrayLen = sizeof(SOffsetAndContLen) * pSyncMsg->dataCount;  // <offset, contLen>
-  int32_t rpcArrayLen = sizeof(SRpcMsg) * pSyncMsg->dataCount;             // SRpcMsg
-  int32_t contArrayLen = pSyncMsg->dataLen - metaArrayLen - rpcArrayLen;
-
-  SOffsetAndContLen* metaArr = (SOffsetAndContLen*)(pSyncMsg->data);
-  SRpcMsg*           msgArr = (SRpcMsg*)(pSyncMsg->data + metaArrayLen);
-  void*              pData = pSyncMsg->data + metaArrayLen + rpcArrayLen;
-
-  for (int i = 0; i < arrSize; ++i) {
-    rpcMsgArr[i] = msgArr[i];
-    rpcMsgArr[i].pCont = rpcMallocCont(msgArr[i].contLen);
-    void* pRpcCont = pSyncMsg->data + metaArr[i].offset;
-    memcpy(rpcMsgArr[i].pCont, pRpcCont, rpcMsgArr[i].contLen);
-  }
-}
-
 // for debug ----------------------
 void syncAppendEntriesBatchPrint(const SyncAppendEntriesBatch* pMsg) {
   char* serialized = syncAppendEntriesBatch2Str(pMsg);
@@ -2159,6 +2254,9 @@ cJSON* syncSnapshotSend2Json(const SyncSnapshotSend* pMsg) {
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
     cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
 
+    snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->beginIndex);
+    cJSON_AddStringToObject(pRoot, "beginIndex", u64buf);
+
     snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->lastIndex);
     cJSON_AddStringToObject(pRoot, "lastIndex", u64buf);
 

source/libs/sync/src/syncRaftEntry.c

@@ -50,6 +50,22 @@ SSyncRaftEntry* syncEntryBuild3(SyncClientRequest* pMsg, SyncTerm term, SyncInde
   return pEntry;
 }
 
+SSyncRaftEntry* syncEntryBuild4(SRpcMsg* pOriginalMsg, SyncTerm term, SyncIndex index) {
+  SSyncRaftEntry* pEntry = syncEntryBuild(pOriginalMsg->contLen);
+  ASSERT(pEntry != NULL);
+
+  pEntry->msgType = TDMT_SYNC_CLIENT_REQUEST;
+  pEntry->originalRpcType = pOriginalMsg->msgType;
+  pEntry->seqNum = 0;
+  pEntry->isWeak = 0;
+  pEntry->term = term;
+  pEntry->index = index;
+  pEntry->dataLen = pOriginalMsg->contLen;
+  memcpy(pEntry->data, pOriginalMsg->pCont, pOriginalMsg->contLen);
+
+  return pEntry;
+}
+
 SSyncRaftEntry* syncEntryBuildNoop(SyncTerm term, SyncIndex index, int32_t vgId) {
   // init rpcMsg
   SMsgHead head;

source/libs/sync/src/syncRaftLog.c

@@ -32,6 +32,7 @@ static SyncTerm  raftLogLastTerm(struct SSyncLogStore* pLogStore);
 static int32_t   raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry);
 static int32_t   raftLogGetEntry(struct SSyncLogStore* pLogStore, SyncIndex index, SSyncRaftEntry** ppEntry);
 static int32_t   raftLogTruncate(struct SSyncLogStore* pLogStore, SyncIndex fromIndex);
+static bool      raftLogExist(struct SSyncLogStore* pLogStore, SyncIndex index);
 
 // private function
 static int32_t raftLogGetLastEntry(SSyncLogStore* pLogStore, SSyncRaftEntry** ppLastEntry);
@@ -83,6 +84,7 @@ SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
   pLogStore->syncLogGetEntry = raftLogGetEntry;
   pLogStore->syncLogTruncate = raftLogTruncate;
   pLogStore->syncLogWriteIndex = raftLogWriteIndex;
+  pLogStore->syncLogExist = raftLogExist;
 
   return pLogStore;
 }
@@ -168,6 +170,13 @@ static SyncIndex raftLogWriteIndex(struct SSyncLogStore* pLogStore) {
   return lastVer + 1;
 }
 
+static bool raftLogExist(struct SSyncLogStore* pLogStore, SyncIndex index) {
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  bool               b = walLogExist(pWal, index);
+  return b;
+}
+
 // if success, return last term
 // if not log, return 0
 // if error, return SYNC_TERM_INVALID

source/libs/sync/src/syncReplication.c

@@ -145,26 +145,34 @@ int32_t syncNodeAppendEntriesPeersSnapshot2(SSyncNode* pSyncNode) {
       return -1;
     }
 
-    SRpcMsg rpcMsgArr[SYNC_MAX_BATCH_SIZE];
+    SSyncRaftEntry* entryPArr[SYNC_MAX_BATCH_SIZE];
-    memset(rpcMsgArr, 0, sizeof(rpcMsgArr));
+    memset(entryPArr, 0, sizeof(entryPArr));
 
     int32_t   getCount = 0;
+    SyncIndex getEntryIndex = nextIndex;
     for (int32_t i = 0; i < pSyncNode->batchSize; ++i) {
       SSyncRaftEntry* pEntry;
-      int32_t         code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, nextIndex, &pEntry);
+      int32_t         code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, getEntryIndex, &pEntry);
       if (code == 0) {
         ASSERT(pEntry != NULL);
-        // get rpc msg [i] from entry
+        entryPArr[i] = pEntry;
-        syncEntryDestory(pEntry);
         getCount++;
       } else {
         break;
       }
     }
 
-    SyncAppendEntriesBatch* pMsg = syncAppendEntriesBatchBuild(rpcMsgArr, getCount, pSyncNode->vgId);
+    SyncAppendEntriesBatch* pMsg = syncAppendEntriesBatchBuild(entryPArr, getCount, pSyncNode->vgId);
     ASSERT(pMsg != NULL);
 
+    for (int32_t i = 0; i < pSyncNode->batchSize; ++i) {
+      SSyncRaftEntry* pEntry = entryPArr[i];
+      if (pEntry != NULL) {
+        syncEntryDestory(pEntry);
+        entryPArr[i] = NULL;
+      }
+    }
+
     // prepare msg
     pMsg->srcId = pSyncNode->myRaftId;
     pMsg->destId = *pDestId;

source/libs/sync/src/syncSnapshot.c

@@ -80,13 +80,15 @@ void snapshotSenderDestroy(SSyncSnapshotSender *pSender) {
 bool snapshotSenderIsStart(SSyncSnapshotSender *pSender) { return pSender->start; }
 
 // begin send snapshot by snapshot, pReader
-int32_t snapshotSenderStart(SSyncSnapshotSender *pSender, SSnapshot snapshot, void *pReader) {
+int32_t snapshotSenderStart(SSyncSnapshotSender *pSender, SSnapshotParam snapshotParam, SSnapshot snapshot,
+                            void *pReader) {
   ASSERT(!snapshotSenderIsStart(pSender));
 
-  // init snapshot and reader
+  // init snapshot, parm, reader
   ASSERT(pSender->pReader == NULL);
   pSender->pReader = pReader;
   pSender->snapshot = snapshot;
+  pSender->snapshotParam = snapshotParam;
 
   // init current block
   if (pSender->pCurrentBlock != NULL) {
@@ -162,6 +164,7 @@ int32_t snapshotSenderStart(SSyncSnapshotSender *pSender, SSnapshot snapshot, vo
   pMsg->srcId = pSender->pSyncNode->myRaftId;
   pMsg->destId = (pSender->pSyncNode->replicasId)[pSender->replicaIndex];
   pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
+  pMsg->beginIndex = pSender->snapshotParam.start;
   pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
   pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
   pMsg->lastConfigIndex = pSender->snapshot.lastConfigIndex;
@@ -439,10 +442,13 @@ static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm p
   pReceiver->snapshot.lastApplyIndex = pBeginMsg->lastIndex;
   pReceiver->snapshot.lastApplyTerm = pBeginMsg->lastTerm;
   pReceiver->snapshot.lastConfigIndex = pBeginMsg->lastConfigIndex;
+  pReceiver->snapshotParam.start = pBeginMsg->beginIndex;
+  pReceiver->snapshotParam.end = pBeginMsg->lastIndex;
 
   // write data
   ASSERT(pReceiver->pWriter == NULL);
-  int32_t ret = pReceiver->pSyncNode->pFsm->FpSnapshotStartWrite(pReceiver->pSyncNode->pFsm, &(pReceiver->pWriter));
+  int32_t ret = pReceiver->pSyncNode->pFsm->FpSnapshotStartWrite(pReceiver->pSyncNode->pFsm,
+                                                                 &(pReceiver->snapshotParam), &(pReceiver->pWriter));
   ASSERT(ret == 0);
 
   // event log

source/libs/sync/test/CMakeLists.txt

@@ -24,6 +24,7 @@ add_executable(syncAppendEntriesTest "")
 add_executable(syncAppendEntriesBatchTest "")
 add_executable(syncAppendEntriesReplyTest "")
 add_executable(syncClientRequestTest "")
+add_executable(syncClientRequestBatchTest "")
 add_executable(syncTimeoutTest "")
 add_executable(syncPingTest "")
 add_executable(syncPingReplyTest "")
@@ -159,6 +160,10 @@ target_sources(syncClientRequestTest
     PRIVATE
     "syncClientRequestTest.cpp"
 )
+target_sources(syncClientRequestBatchTest
+    PRIVATE
+    "syncClientRequestBatchTest.cpp"
+)
 target_sources(syncTimeoutTest
     PRIVATE
     "syncTimeoutTest.cpp"
@@ -407,6 +412,11 @@ target_include_directories(syncClientRequestTest
     "${TD_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
 )
+target_include_directories(syncClientRequestBatchTest
+    PUBLIC
+    "${TD_SOURCE_DIR}/include/libs/sync"
+    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+)
 target_include_directories(syncTimeoutTest
     PUBLIC
     "${TD_SOURCE_DIR}/include/libs/sync"
@@ -658,6 +668,10 @@ target_link_libraries(syncClientRequestTest
     sync
     gtest_main
 )
+target_link_libraries(syncClientRequestBatchTest
+    sync
+    gtest_main
+)
 target_link_libraries(syncTimeoutTest
     sync
     gtest_main

source/libs/sync/test/syncAppendEntriesBatchTest.cpp

@@ -3,6 +3,7 @@
 #include "syncIO.h"
 #include "syncInt.h"
 #include "syncMessage.h"
+#include "syncRaftEntry.h"
 #include "syncUtil.h"
 #include "trpc.h"
 
@@ -15,30 +16,29 @@ void logTest() {
   sFatal("--- sync log test: fatal");
 }
 
-SRpcMsg *createRpcMsg(int32_t i, int32_t dataLen) {
+SSyncRaftEntry *createEntry(int i) {
-  SRpcMsg *pRpcMsg = (SRpcMsg *)taosMemoryMalloc(sizeof(SRpcMsg));
+  SSyncRaftEntry *pEntry = syncEntryBuild(20);
-  memset(pRpcMsg, 0, sizeof(SRpcMsg));
+  assert(pEntry != NULL);
-
+  pEntry->msgType = 1;
-  pRpcMsg->msgType = TDMT_SYNC_PING;
+  pEntry->originalRpcType = 2;
-  pRpcMsg->contLen = dataLen;
+  pEntry->seqNum = 3;
-  pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
+  pEntry->isWeak = true;
-  pRpcMsg->code = 10 * i;
+  pEntry->term = 100;
-  snprintf((char *)pRpcMsg->pCont, pRpcMsg->contLen, "value_%d", i);
+  pEntry->index = 200;
-
+  snprintf(pEntry->data, pEntry->dataLen, "value_%d", i);
-  return pRpcMsg;
+  return pEntry;
 }
 
 SyncAppendEntriesBatch *createMsg() {
-  SRpcMsg rpcMsgArr[5];
+  SSyncRaftEntry *entryPArr[5];
-  memset(rpcMsgArr, 0, sizeof(rpcMsgArr));
+  memset(entryPArr, 0, sizeof(entryPArr));
 
   for (int32_t i = 0; i < 5; ++i) {
-    SRpcMsg *pRpcMsg = createRpcMsg(i, 20);
+    SSyncRaftEntry *pEntry = createEntry(i);
-    rpcMsgArr[i] = *pRpcMsg;
+    entryPArr[i] = pEntry;
-    taosMemoryFree(pRpcMsg);
   }
 
-  SyncAppendEntriesBatch *pMsg = syncAppendEntriesBatchBuild(rpcMsgArr, 5, 1234);
+  SyncAppendEntriesBatch *pMsg = syncAppendEntriesBatchBuild(entryPArr, 5, 1234);
   pMsg->srcId.addr = syncUtilAddr2U64("127.0.0.1", 1234);
   pMsg->srcId.vgId = 100;
   pMsg->destId.addr = syncUtilAddr2U64("127.0.0.1", 5678);
@@ -52,17 +52,17 @@ SyncAppendEntriesBatch *createMsg() {
 
 void test1() {
   SyncAppendEntriesBatch *pMsg = createMsg();
-  syncAppendEntriesBatchLog2((char *)"test1:", pMsg);
+  syncAppendEntriesBatchLog2((char *)"==test1==", pMsg);
 
-  SRpcMsg rpcMsgArr[5];
+/*
-  int32_t retArrSize;
+  SOffsetAndContLen *metaArr = syncAppendEntriesBatchMetaTableArray(pMsg);
-  syncAppendEntriesBatch2RpcMsgArray(pMsg, rpcMsgArr, 5, &retArrSize);
+  int32_t            retArrSize = pMsg->dataCount;
   for (int i = 0; i < retArrSize; ++i) {
-    char logBuf[128];
+    SSyncRaftEntry *pEntry = (SSyncRaftEntry*)(pMsg->data + metaArr[i].offset);
-    snprintf(logBuf, sizeof(logBuf), "==test1 decode rpc msg %d: msgType:%d, code:%d, contLen:%d, pCont:%s \n", i,
+    ASSERT(pEntry->bytes == metaArr[i].contLen);
-             rpcMsgArr[i].msgType, rpcMsgArr[i].code, rpcMsgArr[i].contLen, (char *)rpcMsgArr[i].pCont);
+    syncEntryPrint(pEntry);
-    sTrace("%s", logBuf);
   }
+*/
 
   syncAppendEntriesBatchDestroy(pMsg);
 }

source/libs/sync/test/syncClientRequestBatchTest.cpp

@@ -0,0 +1,125 @@
+#include <gtest/gtest.h>
+#include <stdio.h>
+#include "syncIO.h"
+#include "syncInt.h"
+#include "syncMessage.h"
+#include "syncUtil.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");
+}
+
+SRpcMsg *createRpcMsg(int32_t i, int32_t dataLen) {
+  SyncPing *pSyncMsg = syncPingBuild(20);
+  snprintf(pSyncMsg->data, pSyncMsg->dataLen, "value_%d", i);
+
+  SRpcMsg *pRpcMsg = (SRpcMsg *)taosMemoryMalloc(sizeof(SRpcMsg));
+  memset(pRpcMsg, 0, sizeof(SRpcMsg));
+  pRpcMsg->code = 10 * i;
+  syncPing2RpcMsg(pSyncMsg, pRpcMsg);
+
+  syncPingDestroy(pSyncMsg);
+  return pRpcMsg;
+}
+
+SyncClientRequestBatch *createMsg() {
+  SRpcMsg rpcMsgArr[5];
+  memset(rpcMsgArr, 0, sizeof(rpcMsgArr));
+  for (int32_t i = 0; i < 5; ++i) {
+    SRpcMsg *pRpcMsg = createRpcMsg(i, 20);
+    rpcMsgArr[i] = *pRpcMsg;
+    taosMemoryFree(pRpcMsg);
+  }
+
+  SRaftMeta raftArr[5];
+  memset(raftArr, 0, sizeof(raftArr));
+  for (int32_t i = 0; i < 5; ++i) {
+    raftArr[i].seqNum = i * 10;
+    raftArr[i].isWeak = i % 2;
+  }
+
+  SyncClientRequestBatch *pMsg = syncClientRequestBatchBuild(rpcMsgArr, raftArr, 5, 1234);
+  return pMsg;
+}
+
+void test1() {
+  SyncClientRequestBatch *pMsg = createMsg();
+  syncClientRequestBatchLog2((char *)"==test1==", pMsg);
+  syncClientRequestBatchDestroyDeep(pMsg);
+}
+
+/*
+void test2() {
+  SyncClientRequest *pMsg = createMsg();
+  uint32_t           len = pMsg->bytes;
+  char *             serialized = (char *)taosMemoryMalloc(len);
+  syncClientRequestSerialize(pMsg, serialized, len);
+  SyncClientRequest *pMsg2 = syncClientRequestBuild(pMsg->dataLen);
+  syncClientRequestDeserialize(serialized, len, pMsg2);
+  syncClientRequestLog2((char *)"test2: syncClientRequestSerialize -> syncClientRequestDeserialize ", pMsg2);
+
+  taosMemoryFree(serialized);
+  syncClientRequestDestroy(pMsg);
+  syncClientRequestDestroy(pMsg2);
+}
+
+void test3() {
+  SyncClientRequest *pMsg = createMsg();
+  uint32_t           len;
+  char *             serialized = syncClientRequestSerialize2(pMsg, &len);
+  SyncClientRequest *pMsg2 = syncClientRequestDeserialize2(serialized, len);
+  syncClientRequestLog2((char *)"test3: syncClientRequestSerialize3 -> syncClientRequestDeserialize2 ", pMsg2);
+
+  taosMemoryFree(serialized);
+  syncClientRequestDestroy(pMsg);
+  syncClientRequestDestroy(pMsg2);
+}
+
+void test4() {
+  SyncClientRequest *pMsg = createMsg();
+  SRpcMsg            rpcMsg;
+  syncClientRequest2RpcMsg(pMsg, &rpcMsg);
+  SyncClientRequest *pMsg2 = (SyncClientRequest *)taosMemoryMalloc(rpcMsg.contLen);
+  syncClientRequestFromRpcMsg(&rpcMsg, pMsg2);
+  syncClientRequestLog2((char *)"test4: syncClientRequest2RpcMsg -> syncClientRequestFromRpcMsg ", pMsg2);
+
+  rpcFreeCont(rpcMsg.pCont);
+  syncClientRequestDestroy(pMsg);
+  syncClientRequestDestroy(pMsg2);
+}
+
+void test5() {
+  SyncClientRequest *pMsg = createMsg();
+  SRpcMsg            rpcMsg;
+  syncClientRequest2RpcMsg(pMsg, &rpcMsg);
+  SyncClientRequest *pMsg2 = syncClientRequestFromRpcMsg2(&rpcMsg);
+  syncClientRequestLog2((char *)"test5: syncClientRequest2RpcMsg -> syncClientRequestFromRpcMsg2 ", pMsg2);
+
+  rpcFreeCont(rpcMsg.pCont);
+  syncClientRequestDestroy(pMsg);
+  syncClientRequestDestroy(pMsg2);
+}
+*/
+
+int main() {
+  gRaftDetailLog = true;
+  tsAsyncLog = 0;
+  sDebugFlag = DEBUG_DEBUG + DEBUG_TRACE + DEBUG_SCREEN + DEBUG_FILE;
+  logTest();
+
+  test1();
+
+  /*
+test2();
+test3();
+test4();
+test5();
+*/
+
+  return 0;
+}

source/libs/sync/test/syncConfigChangeSnapshotTest.cpp

@@ -114,7 +114,7 @@ int32_t SnapshotDoRead(struct SSyncFSM* pFsm, void* pReader, void** ppBuf, int32
   return 0;
 }
 
-int32_t SnapshotStartWrite(struct SSyncFSM* pFsm, void** ppWriter) {
+int32_t SnapshotStartWrite(struct SSyncFSM* pFsm, void *pParam, void** ppWriter) {
   *ppWriter = (void*)0xCDEF;
   char logBuf[256] = {0};
   snprintf(logBuf, sizeof(logBuf), "==callback== ==SnapshotStartWrite== pFsm:%p, *ppWriter:%p", pFsm, *ppWriter);
@@ -198,7 +198,7 @@ int64_t createSyncNode(int32_t replicaNum, int32_t myIndex, int32_t vgId, SWal*
   snprintf(syncInfo.path, sizeof(syncInfo.path), "%s_sync_replica%d_index%d", path, replicaNum, myIndex);
   syncInfo.pWal = pWal;
   syncInfo.isStandBy = isStandBy;
-  syncInfo.snapshotEnable = true;
+  syncInfo.snapshotStrategy = SYNC_STRATEGY_STANDARD_SNAPSHOT;
 
   SSyncCfg* pCfg = &syncInfo.syncCfg;
 

source/libs/sync/test/syncSnapshotReceiverTest.cpp

@@ -29,7 +29,7 @@ int32_t SnapshotStartRead(struct SSyncFSM* pFsm, void** ppReader) { return 0; }
 int32_t SnapshotStopRead(struct SSyncFSM* pFsm, void* pReader) { return 0; }
 int32_t SnapshotDoRead(struct SSyncFSM* pFsm, void* pReader, void** ppBuf, int32_t* len) { return 0; }
 
-int32_t SnapshotStartWrite(struct SSyncFSM* pFsm, void** ppWriter) { return 0; }
+int32_t SnapshotStartWrite(struct SSyncFSM* pFsm, void *pParam, void** ppWriter) { return 0; }
 int32_t SnapshotStopWrite(struct SSyncFSM* pFsm, void* pWriter, bool isApply) { return 0; }
 int32_t SnapshotDoWrite(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_t len) { return 0; }
 

source/libs/sync/test/syncTestTool.cpp

@@ -111,7 +111,7 @@ int32_t SnapshotDoRead(struct SSyncFSM* pFsm, void* pReader, void** ppBuf, int32
   return 0;
 }
 
-int32_t SnapshotStartWrite(struct SSyncFSM* pFsm, void** ppWriter) {
+int32_t SnapshotStartWrite(struct SSyncFSM* pFsm, void *pParam, void** ppWriter) {
   *ppWriter = (void*)0xCDEF;
   char logBuf[256] = {0};
 
@@ -203,7 +203,7 @@ SWal* createWal(char* path, int32_t vgId) {
 }
 
 int64_t createSyncNode(int32_t replicaNum, int32_t myIndex, int32_t vgId, SWal* pWal, char* path, bool isStandBy,
-                       bool enableSnapshot) {
+                       ESyncStrategy enableSnapshot) {
   SSyncInfo syncInfo;
   syncInfo.vgId = vgId;
   syncInfo.msgcb = &gSyncIO->msgcb;
@@ -213,7 +213,7 @@ int64_t createSyncNode(int32_t replicaNum, int32_t myIndex, int32_t vgId, SWal*
   snprintf(syncInfo.path, sizeof(syncInfo.path), "%s_sync_replica%d_index%d", path, replicaNum, myIndex);
   syncInfo.pWal = pWal;
   syncInfo.isStandBy = isStandBy;
-  syncInfo.snapshotEnable = enableSnapshot;
+  syncInfo.snapshotStrategy = enableSnapshot;
 
   SSyncCfg* pCfg = &syncInfo.syncCfg;
 
@@ -316,7 +316,7 @@ int main(int argc, char** argv) {
 
   int32_t replicaNum = atoi(argv[1]);
   int32_t myIndex = atoi(argv[2]);
-  bool    enableSnapshot = atoi(argv[3]);
+  ESyncStrategy    enableSnapshot = (ESyncStrategy)atoi(argv[3]);
   int32_t lastApplyIndex = atoi(argv[4]);
   int32_t lastApplyTerm = atoi(argv[5]);
   int32_t writeRecordNum = atoi(argv[6]);

source/libs/tfs/src/tfs.c

@@ -291,7 +291,7 @@ int32_t tfsRmdir(STfs *pTfs, const char *rname) {
     for (int32_t id = 0; id < pTier->ndisk; id++) {
       STfsDisk *pDisk = pTier->disks[id];
       snprintf(aname, TMPNAME_LEN, "%s%s%s", pDisk->path, TD_DIRSEP, rname);
-      uInfo("====> tfs remove dir : path:%s aname:%s rname:[%s]", pDisk->path, aname, rname);
+      uInfo("tfs remove dir : path:%s aname:%s rname:[%s]", pDisk->path, aname, rname);
       taosRemoveDir(aname);
     }
   }

source/libs/transport/inc/transComm.h

@@ -253,7 +253,7 @@ int         transAsyncSend(SAsyncPool* pool, queue* mq);
   do {                                                                                                             \
     if (id > 0) {                                                                                                  \
       tTrace("handle step1");                                                                                      \
-      SExHandle* exh2 = transAcquireExHandle(id);                                                                  \
+      SExHandle* exh2 = transAcquireExHandle(transGetRefMgt(), id);                                                \
       if (exh2 == NULL || id != exh2->refId) {                                                                     \
         tTrace("handle %p except, may already freed, ignore msg, ref1: %" PRIu64 ", ref2 : %" PRIu64 "", exh1,     \
                exh2 ? exh2->refId : 0, id);                                                                        \
@@ -261,7 +261,7 @@ int         transAsyncSend(SAsyncPool* pool, queue* mq);
       }                                                                                                            \
     } else if (id == 0) {                                                                                          \
       tTrace("handle step2");                                                                                      \
-      SExHandle* exh2 = transAcquireExHandle(id);                                                                  \
+      SExHandle* exh2 = transAcquireExHandle(transGetRefMgt(), id);                                                \
       if (exh2 == NULL || id == exh2->refId) {                                                                     \
         tTrace("handle %p except, may already freed, ignore msg, ref1: %" PRIu64 ", ref2 : %" PRIu64 "", exh1, id, \
                exh2 ? exh2->refId : 0);                                                                            \
@@ -391,14 +391,17 @@ void transThreadOnce();
 void transInit();
 void transCleanup();
 
-int32_t    transOpenExHandleMgt(int size);
+int32_t transOpenRefMgt(int size, void (*func)(void*));
-void       transCloseExHandleMgt();
+void    transCloseRefMgt(int32_t refMgt);
-int64_t    transAddExHandle(void* p);
+int64_t transAddExHandle(int32_t refMgt, void* p);
-int32_t    transRemoveExHandle(int64_t refId);
+int32_t transRemoveExHandle(int32_t refMgt, int64_t refId);
-SExHandle* transAcquireExHandle(int64_t refId);
+void*   transAcquireExHandle(int32_t refMgt, int64_t refId);
-int32_t    transReleaseExHandle(int64_t refId);
+int32_t transReleaseExHandle(int32_t refMgt, int64_t refId);
 void    transDestoryExHandle(void* handle);
 
+int32_t transGetRefMgt();
+int32_t transGetInstMgt();
+
 #ifdef __cplusplus
 }
 #endif

source/libs/transport/inc/transportInt.h

@@ -57,7 +57,7 @@ typedef struct {
 
   void*         parent;
   void*         tcphandle;  // returned handle from TCP initialization
-  int32_t       refMgt;
+  int64_t       refId;
   TdThreadMutex mutex;
 } SRpcInfo;
 

source/libs/transport/src/trans.c

@@ -76,16 +76,23 @@ void* rpcOpen(const SRpcInit* pInit) {
   if (pInit->user) {
     memcpy(pRpc->user, pInit->user, strlen(pInit->user));
   }
-  return pRpc;
+
+  int64_t refId = transAddExHandle(transGetInstMgt(), pRpc);
+  transAcquireExHandle(transGetInstMgt(), refId);
+  pRpc->refId = refId;
+  return (void*)refId;
 }
 void rpcClose(void* arg) {
   tInfo("start to close rpc");
+  transRemoveExHandle(transGetInstMgt(), (int64_t)arg);
+  transReleaseExHandle(transGetInstMgt(), (int64_t)arg);
+  tInfo("finish to close rpc");
+  return;
+}
+void rpcCloseImpl(void* arg) {
   SRpcInfo* pRpc = (SRpcInfo*)arg;
   (*taosCloseHandle[pRpc->connType])(pRpc->tcphandle);
   taosMemoryFree(pRpc);
-  tInfo("finish to close rpc");
-
-  return;
 }
 
 void* rpcMallocCont(int32_t contLen) {
@@ -144,7 +151,6 @@ void    rpcSendResponse(const SRpcMsg* pMsg) { transSendResponse(pMsg); }
 
 int32_t rpcGetConnInfo(void* thandle, SRpcConnInfo* pInfo) { return 0; }
 
-
 void rpcRefHandle(void* handle, int8_t type) {
   assert(type == TAOS_CONN_SERVER || type == TAOS_CONN_CLIENT);
   (*taosRefHandle[type])(handle);

source/libs/transport/src/transCli.c

@@ -47,6 +47,7 @@ typedef struct SCliMsg {
   queue          q;
   STransMsgType  type;
 
+  int64_t  refId;
   uint64_t st;
   int      sent;  //(0: no send, 1: alread sent)
 } SCliMsg;
@@ -262,13 +263,17 @@ static void cliReleaseUnfinishedMsg(SCliConn* conn) {
 #define REQUEST_PERSIS_HANDLE(msg)   ((msg)->info.persistHandle == 1)
 #define REQUEST_RELEASE_HANDLE(cmsg) ((cmsg)->type == Release)
 
+#define EPSET_IS_VALID(epSet)       ((epSet) != NULL && (epSet)->numOfEps != 0)
 #define EPSET_GET_SIZE(epSet)       (epSet)->numOfEps
 #define EPSET_GET_INUSE_IP(epSet)   ((epSet)->eps[(epSet)->inUse].fqdn)
 #define EPSET_GET_INUSE_PORT(epSet) ((epSet)->eps[(epSet)->inUse].port)
 #define EPSET_FORWARD_INUSE(epSet)                                 \
   do {                                                             \
+    if ((epSet)->numOfEps != 0) {                                  \
       (epSet)->inUse = (++((epSet)->inUse)) % ((epSet)->numOfEps); \
+    }                                                              \
   } while (0)
+
 #define EPSET_DEBUG_STR(epSet, tbuf)                                                                                   \
   do {                                                                                                                 \
     int len = snprintf(tbuf, sizeof(tbuf), "epset:{");                                                                 \
@@ -501,18 +506,17 @@ static SCliConn* getConnFromPool(void* pool, char* ip, uint32_t port) {
 }
 static void allocConnRef(SCliConn* conn, bool update) {
   if (update) {
-    transRemoveExHandle(conn->refId);
+    transRemoveExHandle(transGetRefMgt(), conn->refId);
     conn->refId = -1;
   }
   SExHandle* exh = taosMemoryCalloc(1, sizeof(SExHandle));
   exh->handle = conn;
   exh->pThrd = conn->hostThrd;
-  exh->refId = transAddExHandle(exh);
+  exh->refId = transAddExHandle(transGetRefMgt(), exh);
   conn->refId = exh->refId;
 }
 static void addConnToPool(void* pool, SCliConn* conn) {
   if (conn->status == ConnInPool) {
-    // assert(0);
     return;
   }
   SCliThrd* thrd = conn->hostThrd;
@@ -601,16 +605,14 @@ static void cliDestroyConn(SCliConn* conn, bool clear) {
   tTrace("%s conn %p remove from conn pool", CONN_GET_INST_LABEL(conn), conn);
   QUEUE_REMOVE(&conn->conn);
   QUEUE_INIT(&conn->conn);
-  transRemoveExHandle(conn->refId);
+  transRemoveExHandle(transGetRefMgt(), conn->refId);
   conn->refId = -1;
 
   if (clear) {
     if (!uv_is_closing((uv_handle_t*)conn->stream)) {
+      uv_read_stop(conn->stream);
       uv_close((uv_handle_t*)conn->stream, cliDestroy);
     }
-    //} else {
-    //  cliDestroy((uv_handle_t*)conn->stream);
-    //}
   }
 }
 static void cliDestroy(uv_handle_t* handle) {
@@ -619,7 +621,7 @@ static void cliDestroy(uv_handle_t* handle) {
   }
 
   SCliConn* conn = handle->data;
-  transRemoveExHandle(conn->refId);
+  transRemoveExHandle(transGetRefMgt(), conn->refId);
   taosMemoryFree(conn->ip);
   conn->stream->data = NULL;
   taosMemoryFree(conn->stream);
@@ -635,7 +637,6 @@ static bool cliHandleNoResp(SCliConn* conn) {
     SCliMsg* pMsg = transQueueGet(&conn->cliMsgs, 0);
     if (REQUEST_NO_RESP(&pMsg->msg)) {
       transQueuePop(&conn->cliMsgs);
-      // taosArrayRemove(msgs, 0);
       destroyCmsg(pMsg);
       res = true;
     }
@@ -668,7 +669,6 @@ static void cliSendCb(uv_write_t* req, int status) {
 void cliSend(SCliConn* pConn) {
   CONN_HANDLE_BROKEN(pConn);
 
-  // assert(taosArrayGetSize(pConn->cliMsgs) > 0);
   assert(!transQueueEmpty(&pConn->cliMsgs));
 
   SCliMsg* pCliMsg = NULL;
@@ -747,7 +747,7 @@ static void cliHandleQuit(SCliMsg* pMsg, SCliThrd* pThrd) {
 }
 static void cliHandleRelease(SCliMsg* pMsg, SCliThrd* pThrd) {
   int64_t    refId = (int64_t)(pMsg->msg.info.handle);
-  SExHandle* exh = transAcquireExHandle(refId);
+  SExHandle* exh = transAcquireExHandle(transGetRefMgt(), refId);
   if (exh == NULL) {
     tDebug("%" PRId64 " already release", refId);
   }
@@ -773,14 +773,14 @@ SCliConn* cliGetConn(SCliMsg* pMsg, SCliThrd* pThrd, bool* ignore) {
   SCliConn* conn = NULL;
   int64_t   refId = (int64_t)(pMsg->msg.info.handle);
   if (refId != 0) {
-    SExHandle* exh = transAcquireExHandle(refId);
+    SExHandle* exh = transAcquireExHandle(transGetRefMgt(), refId);
     if (exh == NULL) {
       *ignore = true;
       destroyCmsg(pMsg);
       return NULL;
     } else {
       conn = exh->handle;
-      transReleaseExHandle(refId);
+      transReleaseExHandle(transGetRefMgt(), refId);
     }
     return conn;
   };
@@ -810,6 +810,11 @@ void cliHandleReq(SCliMsg* pMsg, SCliThrd* pThrd) {
   STrans*        pTransInst = pThrd->pTransInst;
 
   cliMayCvtFqdnToIp(&pCtx->epSet, &pThrd->cvtAddr);
+  if (!EPSET_IS_VALID(&pCtx->epSet)) {
+    destroyCmsg(pMsg);
+    tError("invalid epset");
+    return;
+  }
 
   bool      ignore = false;
   SCliConn* conn = cliGetConn(pMsg, pThrd, &ignore);
@@ -977,6 +982,7 @@ void cliSendQuit(SCliThrd* thrd) {
 }
 void cliWalkCb(uv_handle_t* handle, void* arg) {
   if (!uv_is_closing(handle)) {
+    uv_read_stop((uv_stream_t*)handle);
     uv_close(handle, cliDestroy);
   }
 }
@@ -1066,7 +1072,7 @@ int cliAppCb(SCliConn* pConn, STransMsg* pResp, SCliMsg* pMsg) {
   if (retry) {
     pMsg->sent = 0;
     pCtx->retryCnt += 1;
-    if (code == TSDB_CODE_RPC_NETWORK_UNAVAIL) {
+    if (code == TSDB_CODE_RPC_NETWORK_UNAVAIL || code == TSDB_CODE_RPC_BROKEN_LINK) {
       cliCompareAndSwap(&pCtx->retryLimit, TRANS_RETRY_COUNT_LIMIT, EPSET_GET_SIZE(&pCtx->epSet) * 3);
       if (pCtx->retryCnt < pCtx->retryLimit) {
         transUnrefCliHandle(pConn);
@@ -1077,12 +1083,14 @@ int cliAppCb(SCliConn* pConn, STransMsg* pResp, SCliMsg* pMsg) {
     } else {
       cliCompareAndSwap(&pCtx->retryLimit, TRANS_RETRY_COUNT_LIMIT, TRANS_RETRY_COUNT_LIMIT);
       if (pCtx->retryCnt < pCtx->retryLimit) {
-        addConnToPool(pThrd->pool, pConn);
         if (pResp->contLen == 0) {
           EPSET_FORWARD_INUSE(&pCtx->epSet);
         } else {
-          tDeserializeSEpSet(pResp->pCont, pResp->contLen, &pCtx->epSet);
+          if (tDeserializeSEpSet(pResp->pCont, pResp->contLen, &pCtx->epSet) < 0) {
+            tError("%s conn %p failed to deserialize epset", CONN_GET_INST_LABEL(pConn));
           }
+        }
+        addConnToPool(pThrd->pool, pConn);
         transFreeMsg(pResp->pCont);
         cliSchedMsgToNextNode(pMsg, pThrd);
         return -1;
@@ -1152,12 +1160,12 @@ void transUnrefCliHandle(void* handle) {
 }
 SCliThrd* transGetWorkThrdFromHandle(int64_t handle) {
   SCliThrd*  pThrd = NULL;
-  SExHandle* exh = transAcquireExHandle(handle);
+  SExHandle* exh = transAcquireExHandle(transGetRefMgt(), handle);
   if (exh == NULL) {
     return NULL;
   }
   pThrd = exh->pThrd;
-  transReleaseExHandle(handle);
+  transReleaseExHandle(transGetRefMgt(), handle);
   return pThrd;
 }
 SCliThrd* transGetWorkThrd(STrans* trans, int64_t handle) {
@@ -1184,10 +1192,13 @@ void transReleaseCliHandle(void* handle) {
 }
 
 void transSendRequest(void* shandle, const SEpSet* pEpSet, STransMsg* pReq, STransCtx* ctx) {
-  STrans*   pTransInst = (STrans*)shandle;
+  STrans* pTransInst = (STrans*)transAcquireExHandle(transGetInstMgt(), (int64_t)shandle);
+  if (pTransInst == NULL) return;
+
   SCliThrd* pThrd = transGetWorkThrd(pTransInst, (int64_t)pReq->info.handle);
   if (pThrd == NULL) {
     transFreeMsg(pReq->pCont);
+    transReleaseExHandle(transGetInstMgt(), (int64_t)shandle);
     return;
   }
 
@@ -1208,19 +1219,24 @@ void transSendRequest(void* shandle, const SEpSet* pEpSet, STransMsg* pReq, STra
   cliMsg->msg = *pReq;
   cliMsg->st = taosGetTimestampUs();
   cliMsg->type = Normal;
+  cliMsg->refId = (int64_t)shandle;
 
   STraceId* trace = &pReq->info.traceId;
   tGTrace("%s send request at thread:%08" PRId64 ", dst: %s:%d, app:%p", transLabel(pTransInst), pThrd->pid,
           EPSET_GET_INUSE_IP(&pCtx->epSet), EPSET_GET_INUSE_PORT(&pCtx->epSet), pReq->info.ahandle);
   ASSERT(transAsyncSend(pThrd->asyncPool, &(cliMsg->q)) == 0);
+  transReleaseExHandle(transGetInstMgt(), (int64_t)shandle);
   return;
 }
 
 void transSendRecv(void* shandle, const SEpSet* pEpSet, STransMsg* pReq, STransMsg* pRsp) {
-  STrans*   pTransInst = (STrans*)shandle;
+  STrans* pTransInst = (STrans*)transAcquireExHandle(transGetInstMgt(), (int64_t)shandle);
+  if (pTransInst == NULL) return;
+
   SCliThrd* pThrd = transGetWorkThrd(pTransInst, (int64_t)pReq->info.handle);
   if (pThrd == NULL) {
     transFreeMsg(pReq->pCont);
+    transReleaseExHandle(transGetInstMgt(), (int64_t)shandle);
     return;
   }
   tsem_t* sem = taosMemoryCalloc(1, sizeof(tsem_t));
@@ -1241,6 +1257,7 @@ void transSendRecv(void* shandle, const SEpSet* pEpSet, STransMsg* pReq, STransM
   cliMsg->msg = *pReq;
   cliMsg->st = taosGetTimestampUs();
   cliMsg->type = Normal;
+  cliMsg->refId = (int64_t)shandle;
 
   STraceId* trace = &pReq->info.traceId;
   tGTrace("%s send request at thread:%08" PRId64 ", dst: %s:%d, app:%p", transLabel(pTransInst), pThrd->pid,
@@ -1250,13 +1267,16 @@ void transSendRecv(void* shandle, const SEpSet* pEpSet, STransMsg* pReq, STransM
   tsem_wait(sem);
   tsem_destroy(sem);
   taosMemoryFree(sem);
+
+  transReleaseExHandle(transGetInstMgt(), (int64_t)shandle);
   return;
 }
 /*
  *
  **/
 void transSetDefaultAddr(void* shandle, const char* ip, const char* fqdn) {
-  STrans* pTransInst = shandle;
+  STrans* pTransInst = (STrans*)transAcquireExHandle(transGetInstMgt(), (int64_t)shandle);
+  if (pTransInst == NULL) return;
 
   SCvtAddr cvtAddr = {0};
   if (ip != NULL && fqdn != NULL) {
@@ -1271,11 +1291,13 @@ void transSetDefaultAddr(void* shandle, const char* ip, const char* fqdn) {
     SCliMsg* cliMsg = taosMemoryCalloc(1, sizeof(SCliMsg));
     cliMsg->ctx = pCtx;
     cliMsg->type = Update;
+    cliMsg->refId = (int64_t)shandle;
 
     SCliThrd* thrd = ((SCliObj*)pTransInst->tcphandle)->pThreadObj[i];
     tDebug("%s update epset at thread:%08" PRId64 "", pTransInst->label, thrd->pid);
 
     transAsyncSend(thrd->asyncPool, &(cliMsg->q));
   }
+  transReleaseExHandle(transGetInstMgt(), (int64_t)shandle);
 }
 #endif

source/libs/transport/src/transComm.c

@@ -19,6 +19,7 @@
 static TdThreadOnce transModuleInit = PTHREAD_ONCE_INIT;
 
 static int32_t refMgt;
+static int32_t instMgt;
 
 int transAuthenticateMsg(void* pMsg, int msgLen, void* pAuth, void* pKey) {
   T_MD5_CTX context;
@@ -479,53 +480,54 @@ bool transEpSetIsEqual(SEpSet* a, SEpSet* b) {
   }
   return true;
 }
-static int32_t transGetRefMgt() {
-  //
-  return refMgt;
-}
 
 static void transInitEnv() {
-  refMgt = transOpenExHandleMgt(50000);
+  refMgt = transOpenRefMgt(50000, transDestoryExHandle);
+  instMgt = taosOpenRef(50, rpcCloseImpl);
   uv_os_setenv("UV_TCP_SINGLE_ACCEPT", "1");
 }
 static void transDestroyEnv() {
-  // close ref
+  transCloseRefMgt(refMgt);
-  transCloseExHandleMgt();
+  transCloseRefMgt(instMgt);
 }
 
 void transInit() {
   // init env
   taosThreadOnce(&transModuleInit, transInitEnv);
 }
+
+int32_t transGetRefMgt() { return refMgt; }
+int32_t transGetInstMgt() { return instMgt; }
+
 void transCleanup() {
   // clean env
   transDestroyEnv();
 }
-int32_t transOpenExHandleMgt(int size) {
+int32_t transOpenRefMgt(int size, void (*func)(void*)) {
   // added into once later
-  return taosOpenRef(size, transDestoryExHandle);
+  return taosOpenRef(size, func);
 }
-void transCloseExHandleMgt() {
+void transCloseRefMgt(int32_t mgt) {
   // close ref
-  taosCloseRef(transGetRefMgt());
+  taosCloseRef(mgt);
 }
-int64_t transAddExHandle(void* p) {
+int64_t transAddExHandle(int32_t refMgt, void* p) {
   // acquire extern handle
-  return taosAddRef(transGetRefMgt(), p);
+  return taosAddRef(refMgt, p);
 }
-int32_t transRemoveExHandle(int64_t refId) {
+int32_t transRemoveExHandle(int32_t refMgt, int64_t refId) {
   // acquire extern handle
-  return taosRemoveRef(transGetRefMgt(), refId);
+  return taosRemoveRef(refMgt, refId);
 }
 
-SExHandle* transAcquireExHandle(int64_t refId) {
+void* transAcquireExHandle(int32_t refMgt, int64_t refId) {
   // acquire extern handle
-  return (SExHandle*)taosAcquireRef(transGetRefMgt(), refId);
+  return (void*)taosAcquireRef(refMgt, refId);
 }
 
-int32_t transReleaseExHandle(int64_t refId) {
+int32_t transReleaseExHandle(int32_t refMgt, int64_t refId) {
   // release extern handle
-  return taosReleaseRef(transGetRefMgt(), refId);
+  return taosReleaseRef(refMgt, refId);
 }
 void transDestoryExHandle(void* handle) {
   if (handle == NULL) {

source/libs/transport/src/transSvr.c

@@ -261,7 +261,7 @@ static void uvHandleReq(SSvrConn* pConn) {
   // 2. once send out data, cli conn released to conn pool immediately
   // 3. not mixed with persist
   transMsg.info.ahandle = (void*)pHead->ahandle;
-  transMsg.info.handle = (void*)transAcquireExHandle(pConn->refId);
+  transMsg.info.handle = (void*)transAcquireExHandle(transGetRefMgt(), pConn->refId);
   transMsg.info.refId = pConn->refId;
   transMsg.info.traceId = pHead->traceId;
 
@@ -279,7 +279,7 @@ static void uvHandleReq(SSvrConn* pConn) {
   pConnInfo->clientPort = ntohs(pConn->addr.sin_port);
   tstrncpy(pConnInfo->user, pConn->user, sizeof(pConnInfo->user));
 
-  transReleaseExHandle(pConn->refId);
+  transReleaseExHandle(transGetRefMgt(), pConn->refId);
 
   STrans* pTransInst = pConn->pTransInst;
   (*pTransInst->cfp)(pTransInst->parent, &transMsg, NULL);
@@ -507,15 +507,15 @@ void uvWorkerAsyncCb(uv_async_t* handle) {
 
       SExHandle* exh1 = transMsg.info.handle;
       int64_t    refId = transMsg.info.refId;
-      SExHandle* exh2 = transAcquireExHandle(refId);
+      SExHandle* exh2 = transAcquireExHandle(transGetRefMgt(), refId);
       if (exh2 == NULL || exh1 != exh2) {
         tTrace("handle except msg %p, ignore it", exh1);
-        transReleaseExHandle(refId);
+        transReleaseExHandle(transGetRefMgt(), refId);
         destroySmsg(msg);
         continue;
       }
       msg->pConn = exh1->handle;
-      transReleaseExHandle(refId);
+      transReleaseExHandle(transGetRefMgt(), refId);
       (*transAsyncHandle[msg->type])(msg, pThrd);
     }
   }
@@ -757,8 +757,8 @@ static SSvrConn* createConn(void* hThrd) {
   SExHandle* exh = taosMemoryMalloc(sizeof(SExHandle));
   exh->handle = pConn;
   exh->pThrd = pThrd;
-  exh->refId = transAddExHandle(exh);
+  exh->refId = transAddExHandle(transGetRefMgt(), exh);
-  transAcquireExHandle(exh->refId);
+  transAcquireExHandle(transGetRefMgt(), exh->refId);
 
   pConn->refId = exh->refId;
   transRefSrvHandle(pConn);
@@ -789,14 +789,14 @@ static void destroyConnRegArg(SSvrConn* conn) {
   }
 }
 static int reallocConnRef(SSvrConn* conn) {
-  transReleaseExHandle(conn->refId);
+  transReleaseExHandle(transGetRefMgt(), conn->refId);
-  transRemoveExHandle(conn->refId);
+  transRemoveExHandle(transGetRefMgt(), conn->refId);
   // avoid app continue to send msg on invalid handle
   SExHandle* exh = taosMemoryMalloc(sizeof(SExHandle));
   exh->handle = conn;
   exh->pThrd = conn->hostThrd;
-  exh->refId = transAddExHandle(exh);
+  exh->refId = transAddExHandle(transGetRefMgt(), exh);
-  transAcquireExHandle(exh->refId);
+  transAcquireExHandle(transGetRefMgt(), exh->refId);
   conn->refId = exh->refId;
 
   return 0;
@@ -808,8 +808,8 @@ static void uvDestroyConn(uv_handle_t* handle) {
   }
   SWorkThrd* thrd = conn->hostThrd;
 
-  transReleaseExHandle(conn->refId);
+  transReleaseExHandle(transGetRefMgt(), conn->refId);
-  transRemoveExHandle(conn->refId);
+  transRemoveExHandle(transGetRefMgt(), conn->refId);
 
   tDebug("%s conn %p destroy", transLabel(thrd->pTransInst), conn);
   transQueueDestroy(&conn->srvMsgs);
@@ -1029,8 +1029,9 @@ void transUnrefSrvHandle(void* handle) {
 }
 
 void transReleaseSrvHandle(void* handle) {
-  SExHandle* exh = handle;
+  SRpcHandleInfo* info = handle;
-  int64_t    refId = exh->refId;
+  SExHandle*      exh = info->handle;
+  int64_t         refId = info->refId;
 
   ASYNC_CHECK_HANDLE(exh, refId);
 
@@ -1045,11 +1046,11 @@ void transReleaseSrvHandle(void* handle) {
 
   tTrace("%s conn %p start to release", transLabel(pThrd->pTransInst), exh->handle);
   transAsyncSend(pThrd->asyncPool, &m->q);
-  transReleaseExHandle(refId);
+  transReleaseExHandle(transGetRefMgt(), refId);
   return;
 _return1:
   tTrace("handle %p failed to send to release handle", exh);
-  transReleaseExHandle(refId);
+  transReleaseExHandle(transGetRefMgt(), refId);
   return;
 _return2:
   tTrace("handle %p failed to send to release handle", exh);
@@ -1074,12 +1075,12 @@ void transSendResponse(const STransMsg* msg) {
   STraceId* trace = (STraceId*)&msg->info.traceId;
   tGTrace("conn %p start to send resp (1/2)", exh->handle);
   transAsyncSend(pThrd->asyncPool, &m->q);
-  transReleaseExHandle(refId);
+  transReleaseExHandle(transGetRefMgt(), refId);
   return;
 _return1:
   tTrace("handle %p failed to send resp", exh);
   rpcFreeCont(msg->pCont);
-  transReleaseExHandle(refId);
+  transReleaseExHandle(transGetRefMgt(), refId);
   return;
 _return2:
   tTrace("handle %p failed to send resp", exh);
@@ -1103,13 +1104,13 @@ void transRegisterMsg(const STransMsg* msg) {
 
   tTrace("%s conn %p start to register brokenlink callback", transLabel(pThrd->pTransInst), exh->handle);
   transAsyncSend(pThrd->asyncPool, &m->q);
-  transReleaseExHandle(refId);
+  transReleaseExHandle(transGetRefMgt(), refId);
   return;
 
 _return1:
   tTrace("handle %p failed to register brokenlink", exh);
   rpcFreeCont(msg->pCont);
-  transReleaseExHandle(refId);
+  transReleaseExHandle(transGetRefMgt(), refId);
   return;
 _return2:
   tTrace("handle %p failed to register brokenlink", exh);

source/libs/transport/test/transUT.cpp

@@ -175,7 +175,7 @@ static void processReleaseHandleCb(void *parent, SRpcMsg *pMsg, SEpSet *pEpSet)
   rpcMsg.code = 0;
   rpcSendResponse(&rpcMsg);
 
-  rpcReleaseHandle(pMsg->info.handle, TAOS_CONN_SERVER);
+  rpcReleaseHandle(&pMsg->info, TAOS_CONN_SERVER);
 }
 static void processRegisterFailure(void *parent, SRpcMsg *pMsg, SEpSet *pEpSet) {
   {