Merge pull request #18561 from taosdata/fix/TD-20805

enh: show more info while drop dnode

include/libs/sync/sync.h

@@ -67,6 +67,7 @@ typedef struct SWal           SWal;
 typedef struct SSyncRaftEntry SSyncRaftEntry;
 
 typedef enum {
+  TAOS_SYNC_STATE_OFFLINE = 0,
   TAOS_SYNC_STATE_FOLLOWER = 100,
   TAOS_SYNC_STATE_CANDIDATE = 101,
   TAOS_SYNC_STATE_LEADER = 102,

source/common/src/systable.c

@@ -212,6 +212,8 @@ static const SSysDbTableSchema vgroupsSchema[] = {
     {.name = "v2_status", .bytes = 10 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true},
     {.name = "v3_dnode", .bytes = 4, .type = TSDB_DATA_TYPE_INT, .sysInfo = true},
     {.name = "v3_status", .bytes = 10 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true},
+    {.name = "v4_dnode", .bytes = 4, .type = TSDB_DATA_TYPE_INT, .sysInfo = true},
+    {.name = "v4_status", .bytes = 10 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true},
     {.name = "status", .bytes = 12 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true},
     {.name = "cacheload", .bytes = 4, .type = TSDB_DATA_TYPE_INT, .sysInfo = true},
     {.name = "nfiles", .bytes = 4, .type = TSDB_DATA_TYPE_INT, .sysInfo = true},

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

@@ -150,7 +150,7 @@ static void dmGetServerRunStatus(SDnodeMgmt *pMgmt, SServerStatusRsp *pStatus) {
   SServerStatusRsp statusRsp = {0};
   SMonMloadInfo    minfo = {0};
   (*pMgmt->getMnodeLoadsFp)(&minfo);
-  if (minfo.isMnode && minfo.load.syncState == TAOS_SYNC_STATE_ERROR) {
+  if (minfo.isMnode && (minfo.load.syncState == TAOS_SYNC_STATE_ERROR || minfo.load.syncState == TAOS_SYNC_STATE_OFFLINE)) {
     pStatus->statusCode = TSDB_SRV_STATUS_SERVICE_DEGRADED;
     snprintf(pStatus->details, sizeof(pStatus->details), "mnode sync state is %s", syncStr(minfo.load.syncState));
     return;
@@ -160,7 +160,7 @@ static void dmGetServerRunStatus(SDnodeMgmt *pMgmt, SServerStatusRsp *pStatus) {
   (*pMgmt->getVnodeLoadsFp)(&vinfo);
   for (int32_t i = 0; i < taosArrayGetSize(vinfo.pVloads); ++i) {
     SVnodeLoad *pLoad = taosArrayGet(vinfo.pVloads, i);
-    if (pLoad->syncState == TAOS_SYNC_STATE_ERROR) {
+    if (pLoad->syncState == TAOS_SYNC_STATE_ERROR || pLoad->syncState == TAOS_SYNC_STATE_OFFLINE) {
       pStatus->statusCode = TSDB_SRV_STATUS_SERVICE_DEGRADED;
       snprintf(pStatus->details, sizeof(pStatus->details), "vnode:%d sync state is %s", pLoad->vgId,
                syncStr(pLoad->syncState));

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

@@ -36,8 +36,6 @@ int64_t  mndGetVgroupMemory(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup);
 SArray *mndBuildDnodesArray(SMnode *, int32_t exceptDnodeId);
 int32_t mndAllocSmaVgroup(SMnode *, SDbObj *pDb, SVgObj *pVgroup);
 int32_t mndAllocVgroup(SMnode *, SDbObj *pDb, SVgObj **ppVgroups);
-int32_t mndAddVnodeToVgroup(SMnode *, SVgObj *pVgroup, SArray *pArray);
-int32_t mndRemoveVnodeFromVgroup(SMnode *, SVgObj *pVgroup, SArray *pArray, SVnodeGid *pDelVgid);
 int32_t mndAddCreateVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid);
 int32_t mndAddAlterVnodeConfirmAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup);
 int32_t mndAddAlterVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, tmsg_t msgType);

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

@@ -151,10 +151,10 @@ static void mndSetVgroupOffline(SMnode *pMnode, int32_t dnodeId, int64_t curMs)
     bool roleChanged = false;
     for (int32_t vg = 0; vg < pVgroup->replica; ++vg) {
       if (pVgroup->vnodeGid[vg].dnodeId == dnodeId) {
-        if (pVgroup->vnodeGid[vg].syncState != TAOS_SYNC_STATE_ERROR) {
+        if (pVgroup->vnodeGid[vg].syncState != TAOS_SYNC_STATE_OFFLINE) {
           mInfo("vgId:%d, state changed by offline check, old state:%s restored:%d new state:error restored:0",
                 pVgroup->vgId, syncStr(pVgroup->vnodeGid[vg].syncState), pVgroup->vnodeGid[vg].syncRestore);
-          pVgroup->vnodeGid[vg].syncState = TAOS_SYNC_STATE_ERROR;
+          pVgroup->vnodeGid[vg].syncState = TAOS_SYNC_STATE_OFFLINE;
           pVgroup->vnodeGid[vg].syncRestore = 0;
           roleChanged = true;
         }
@@ -491,6 +491,15 @@ void mndPreClose(SMnode *pMnode) {
   if (pMnode != NULL) {
     syncLeaderTransfer(pMnode->syncMgmt.sync);
     syncPreStop(pMnode->syncMgmt.sync);
+
+    while (syncSnapshotRecving(pMnode->syncMgmt.sync)) {
+      mInfo("vgId:1, snapshot is recving");
+      taosMsleep(300);
+    }
+    while (syncSnapshotSending(pMnode->syncMgmt.sync)) {
+      mInfo("vgId:1, snapshot is sending");
+      taosMsleep(300);
+    }
   }
 }
 
@@ -747,7 +756,7 @@ int32_t mndGetMonitorInfo(SMnode *pMnode, SMonClusterInfo *pClusterInfo, SMonVgr
         tstrncpy(desc.status, "ready", sizeof(desc.status));
         pClusterInfo->vgroups_alive++;
       }
-      if (pVgid->syncState != TAOS_SYNC_STATE_ERROR) {
+      if (pVgid->syncState != TAOS_SYNC_STATE_ERROR && pVgid->syncState != TAOS_SYNC_STATE_OFFLINE) {
         pClusterInfo->vnodes_alive++;
       }
       pClusterInfo->vnodes_total++;

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

@@ -185,7 +185,7 @@ static int32_t mndMnodeActionInsert(SSdb *pSdb, SMnodeObj *pObj) {
     return -1;
   }
 
-  pObj->syncState = TAOS_SYNC_STATE_ERROR;
+  pObj->syncState = TAOS_SYNC_STATE_OFFLINE;
   mndReloadSyncConfig(pSdb->pMnode);
   return 0;
 }

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

@@ -179,6 +179,16 @@ static int32_t mndVgroupActionUpdate(SSdb *pSdb, SVgObj *pOld, SVgObj *pNew) {
   pOld->hashEnd = pNew->hashEnd;
   pOld->replica = pNew->replica;
   pOld->isTsma = pNew->isTsma;
+  for (int32_t i = 0; i < pNew->replica; ++i) {
+    SVnodeGid *pNewGid = &pNew->vnodeGid[i];
+    for (int32_t j = 0; j < pOld->replica; ++j) {
+      SVnodeGid *pOldGid = &pOld->vnodeGid[j];
+      if (pNewGid->dnodeId == pOldGid->dnodeId) {
+        pNewGid->syncState = pOldGid->syncState;
+        pNewGid->syncRestore = pOldGid->syncRestore;
+      }
+    }
+  }
   memcpy(pOld->vnodeGid, pNew->vnodeGid, TSDB_MAX_REPLICA * sizeof(SVnodeGid));
   return 0;
 }
@@ -659,8 +669,8 @@ static int32_t mndRetrieveVgroups(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *p
     pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
     colDataAppend(pColInfo, numOfRows, (const char *)&pVgroup->numOfTables, false);
 
-    // default 3 replica
+    // default 3 replica, add 1 replica if move vnode
-    for (int32_t i = 0; i < 3; ++i) {
+    for (int32_t i = 0; i < 4; ++i) {
       pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
       if (i < pVgroup->replica) {
         colDataAppend(pColInfo, numOfRows, (const char *)&pVgroup->vnodeGid[i].dnodeId, false);
@@ -851,7 +861,7 @@ static void mndCancelGetNextVnode(SMnode *pMnode, void *pIter) {
   sdbCancelFetch(pSdb, pIter);
 }
 
-int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
+static int32_t mndAddVnodeToVgroup(SMnode *pMnode, STrans *pTrans, SVgObj *pVgroup, SArray *pArray) {
   taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
   for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
     SDnodeObj *pDnode = taosArrayGet(pArray, i);
@@ -887,12 +897,21 @@ int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
     }
 
     pVgid->dnodeId = pDnode->id;
-    pVgid->syncState = TAOS_SYNC_STATE_ERROR;
+    pVgid->syncState = TAOS_SYNC_STATE_OFFLINE;
     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++;
+
+    SSdbRaw *pVgRaw = mndVgroupActionEncode(pVgroup);
+    if (pVgRaw == NULL) return -1;
+    if (mndTransAppendRedolog(pTrans, pVgRaw) != 0) {
+      sdbFreeRaw(pVgRaw);
+      return -1;
+    }
+    (void)sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
+
     return 0;
   }
 
@@ -901,7 +920,8 @@ int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
   return -1;
 }
 
-int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray, SVnodeGid *pDelVgid) {
+static int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, STrans *pTrans, SVgObj *pVgroup, SArray *pArray,
+                                        SVnodeGid *pDelVgid) {
   taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
   for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
     SDnodeObj *pDnode = taosArrayGet(pArray, i);
@@ -941,6 +961,15 @@ _OVER:
     SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
     mInfo("db:%s, vgId:%d, vn:%d dnode:%d is reserved", pVgroup->dbName, pVgroup->vgId, vn, pVgid->dnodeId);
   }
+
+  SSdbRaw *pVgRaw = mndVgroupActionEncode(pVgroup);
+  if (pVgRaw == NULL) return -1;
+  if (mndTransAppendRedolog(pTrans, pVgRaw) != 0) {
+    sdbFreeRaw(pVgRaw);
+    return -1;
+  }
+  (void)sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
+
   return 0;
 }
 
@@ -1088,7 +1117,7 @@ int32_t mndSetMoveVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb,
 
   if (!force) {
     mInfo("vgId:%d, will add 1 vnode", pVgroup->vgId);
-    if (mndAddVnodeToVgroup(pMnode, &newVg, pArray) != 0) return -1;
+    if (mndAddVnodeToVgroup(pMnode, pTrans, &newVg, pArray) != 0) return -1;
     for (int32_t i = 0; i < newVg.replica - 1; ++i) {
       if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pDb, &newVg, newVg.vnodeGid[i].dnodeId) != 0) return -1;
     }
@@ -1100,6 +1129,16 @@ int32_t mndSetMoveVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb,
     SVnodeGid del = newVg.vnodeGid[vnIndex];
     newVg.vnodeGid[vnIndex] = newVg.vnodeGid[newVg.replica];
     memset(&newVg.vnodeGid[newVg.replica], 0, sizeof(SVnodeGid));
+    {
+      SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
+      if (pRaw == NULL) return -1;
+      if (mndTransAppendRedolog(pTrans, pRaw) != 0) {
+        sdbFreeRaw(pRaw);
+        return -1;
+      }
+      (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
+    }
+
     if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVg, &del, true) != 0) return -1;
     for (int32_t i = 0; i < newVg.replica; ++i) {
       if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pDb, &newVg, newVg.vnodeGid[i].dnodeId) != 0) return -1;
@@ -1107,11 +1146,20 @@ int32_t mndSetMoveVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb,
     if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVg) != 0) return -1;
   } else {
     mInfo("vgId:%d, will add 1 vnode and force remove 1 vnode", pVgroup->vgId);
-    if (mndAddVnodeToVgroup(pMnode, &newVg, pArray) != 0) return -1;
+    if (mndAddVnodeToVgroup(pMnode, pTrans, &newVg, pArray) != 0) return -1;
     newVg.replica--;
     SVnodeGid del = newVg.vnodeGid[vnIndex];
     newVg.vnodeGid[vnIndex] = newVg.vnodeGid[newVg.replica];
     memset(&newVg.vnodeGid[newVg.replica], 0, sizeof(SVnodeGid));
+    {
+      SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
+      if (pRaw == NULL) return -1;
+      if (mndTransAppendRedolog(pTrans, pRaw) != 0) {
+        sdbFreeRaw(pRaw);
+        return -1;
+      }
+      (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
+    }
 
     for (int32_t i = 0; i < newVg.replica; ++i) {
       if (i != vnIndex) {
@@ -1128,16 +1176,12 @@ int32_t mndSetMoveVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb,
 
   {
     SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
-    if (pRaw == NULL || mndTransAppendRedolog(pTrans, pRaw) != 0) return -1;
+    if (pRaw == NULL) return -1;
+    if (mndTransAppendCommitlog(pTrans, pRaw) != 0) {
+      sdbFreeRaw(pRaw);
+      return -1;
+    }
     (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-    pRaw = NULL;
-  }
-
-  {
-    SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
-    if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) return -1;
-    (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-    pRaw = NULL;
   }
 
   mInfo("vgId:%d, vgroup info after move, replica:%d", newVg.vgId, newVg.replica);
@@ -1193,7 +1237,15 @@ static int32_t mndAddIncVgroupReplicaToTrans(SMnode *pMnode, STrans *pTrans, SDb
   SVnodeGid *pGid = &pVgroup->vnodeGid[pVgroup->replica];
   pVgroup->replica++;
   pGid->dnodeId = newDnodeId;
-  pGid->syncState = TAOS_SYNC_STATE_ERROR;
+  pGid->syncState = TAOS_SYNC_STATE_OFFLINE;
+
+  SSdbRaw *pVgRaw = mndVgroupActionEncode(pVgroup);
+  if (pVgRaw == NULL) return -1;
+  if (mndTransAppendRedolog(pTrans, pVgRaw) != 0) {
+    sdbFreeRaw(pVgRaw);
+    return -1;
+  }
+  (void)sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
 
   for (int32_t i = 0; i < pVgroup->replica - 1; ++i) {
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pDb, pVgroup, pVgroup->vnodeGid[i].dnodeId) != 0) return -1;
@@ -1224,6 +1276,14 @@ static int32_t mndAddDecVgroupReplicaFromTrans(SMnode *pMnode, STrans *pTrans, S
   memcpy(pGid, &pVgroup->vnodeGid[pVgroup->replica], sizeof(SVnodeGid));
   memset(&pVgroup->vnodeGid[pVgroup->replica], 0, sizeof(SVnodeGid));
 
+  SSdbRaw *pVgRaw = mndVgroupActionEncode(pVgroup);
+  if (pVgRaw == NULL) return -1;
+  if (mndTransAppendRedolog(pTrans, pVgRaw) != 0) {
+    sdbFreeRaw(pVgRaw);
+    return -1;
+  }
+  (void)sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
+
   if (mndAddDropVnodeAction(pMnode, pTrans, pDb, pVgroup, &delGid, true) != 0) return -1;
   for (int32_t i = 0; i < pVgroup->replica; ++i) {
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pDb, pVgroup, pVgroup->vnodeGid[i].dnodeId) != 0) return -1;
@@ -1236,9 +1296,8 @@ static int32_t mndAddDecVgroupReplicaFromTrans(SMnode *pMnode, STrans *pTrans, S
 static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj *pVgroup, SDnodeObj *pNew1,
                                      SDnodeObj *pOld1, SDnodeObj *pNew2, SDnodeObj *pOld2, SDnodeObj *pNew3,
                                      SDnodeObj *pOld3) {
-  int32_t  code = -1;
+  int32_t code = -1;
-  SSdbRaw *pRaw = NULL;
+  STrans *pTrans = NULL;
-  STrans  *pTrans = NULL;
 
   pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq, "red-vgroup");
   if (pTrans == NULL) goto _OVER;
@@ -1319,17 +1378,13 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
   }
 
   {
-    pRaw = mndVgroupActionEncode(&newVg);
+    SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
-    if (pRaw == NULL || mndTransAppendRedolog(pTrans, pRaw) != 0) goto _OVER;
+    if (pRaw == NULL) return -1;
+    if (mndTransAppendCommitlog(pTrans, pRaw) != 0) {
+      sdbFreeRaw(pRaw);
+      return -1;
+    }
     (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-    pRaw = NULL;
-  }
-
-  {
-    pRaw = mndVgroupActionEncode(&newVg);
-    if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) goto _OVER;
-    (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-    pRaw = NULL;
   }
 
   mInfo("vgId:%d, vgroup info after redistribute, replica:%d", newVg.vgId, newVg.replica);
@@ -1342,7 +1397,6 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
 
 _OVER:
   mndTransDrop(pTrans);
-  sdbFreeRaw(pRaw);
   mndReleaseDb(pMnode, pDb);
   return code;
 }
@@ -1593,13 +1647,13 @@ int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pOldDb
     mInfo("db:%s, vgId:%d, will add 2 vnodes, vn:0 dnode:%d", pVgroup->dbName, pVgroup->vgId,
           pVgroup->vnodeGid[0].dnodeId);
 
-    if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
+    if (mndAddVnodeToVgroup(pMnode, pTrans, &newVgroup, pArray) != 0) return -1;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pNewDb, &newVgroup, newVgroup.vnodeGid[0].dnodeId) != 0)
       return -1;
     if (mndAddCreateVnodeAction(pMnode, pTrans, pNewDb, &newVgroup, &newVgroup.vnodeGid[1]) != 0) return -1;
     if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pNewDb, &newVgroup) != 0) return -1;
 
-    if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
+    if (mndAddVnodeToVgroup(pMnode, pTrans, &newVgroup, pArray) != 0) return -1;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pNewDb, &newVgroup, newVgroup.vnodeGid[0].dnodeId) != 0)
       return -1;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pNewDb, &newVgroup, newVgroup.vnodeGid[1].dnodeId) != 0)
@@ -1612,7 +1666,7 @@ int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pOldDb
 
     SVnodeGid del1 = {0};
     SVnodeGid del2 = {0};
-    if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del1) != 0) return -1;
+    if (mndRemoveVnodeFromVgroup(pMnode, pTrans, &newVgroup, pArray, &del1) != 0) return -1;
     if (mndAddDropVnodeAction(pMnode, pTrans, pNewDb, &newVgroup, &del1, true) != 0) return -1;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pNewDb, &newVgroup, newVgroup.vnodeGid[0].dnodeId) != 0)
       return -1;
@@ -1620,7 +1674,7 @@ int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pOldDb
       return -1;
     if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pNewDb, &newVgroup) != 0) return -1;
 
-    if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del2) != 0) return -1;
+    if (mndRemoveVnodeFromVgroup(pMnode, pTrans, &newVgroup, pArray, &del2) != 0) return -1;
     if (mndAddDropVnodeAction(pMnode, pTrans, pNewDb, &newVgroup, &del2, true) != 0) return -1;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pNewDb, &newVgroup, newVgroup.vnodeGid[0].dnodeId) != 0)
       return -1;
@@ -1629,16 +1683,6 @@ int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pOldDb
     return -1;
   }
 
-  {
-    SSdbRaw *pVgRaw = mndVgroupActionEncode(&newVgroup);
-    if (pVgRaw == NULL) return -1;
-    if (mndTransAppendRedolog(pTrans, pVgRaw) != 0) {
-      sdbFreeRaw(pVgRaw);
-      return -1;
-    }
-    (void)sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
-  }
-
   {
     SSdbRaw *pVgRaw = mndVgroupActionEncode(&newVgroup);
     if (pVgRaw == NULL) return -1;
@@ -1657,10 +1701,9 @@ static int32_t mndAddAdjustVnodeHashRangeAction(SMnode *pMnode, STrans *pTrans,
 }
 
 static int32_t mndSplitVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj *pVgroup) {
-  int32_t  code = -1;
+  int32_t code = -1;
-  SSdbRaw *pRaw = NULL;
+  STrans *pTrans = NULL;
-  STrans  *pTrans = NULL;
+  SArray *pArray = mndBuildDnodesArray(pMnode, 0);
-  SArray  *pArray = mndBuildDnodesArray(pMnode, 0);
 
   pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq, "split-vgroup");
   if (pTrans == NULL) goto _OVER;
@@ -1676,13 +1719,13 @@ static int32_t mndSplitVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj
   }
 
   if (newVg1.replica == 1) {
-    if (mndAddVnodeToVgroup(pMnode, &newVg1, pArray) != 0) goto _OVER;
+    if (mndAddVnodeToVgroup(pMnode, pTrans, &newVg1, pArray) != 0) goto _OVER;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pDb, &newVg1, newVg1.vnodeGid[0].dnodeId) != 0) goto _OVER;
     if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVg1, &newVg1.vnodeGid[1]) != 0) goto _OVER;
     if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVg1) != 0) goto _OVER;
   } else if (newVg1.replica == 3) {
     SVnodeGid del1 = {0};
-    if (mndRemoveVnodeFromVgroup(pMnode, &newVg1, pArray, &del1) != 0) goto _OVER;
+    if (mndRemoveVnodeFromVgroup(pMnode, pTrans, &newVg1, pArray, &del1) != 0) goto _OVER;
     if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVg1, &del1, true) != 0) goto _OVER;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pDb, &newVg1, newVg1.vnodeGid[0].dnodeId) != 0) goto _OVER;
     if (mndAddAlterVnodeReplicaAction(pMnode, pTrans, pDb, &newVg1, newVg1.vnodeGid[1].dnodeId) != 0) goto _OVER;
@@ -1727,17 +1770,23 @@ static int32_t mndSplitVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj
 #endif
 
   {
-    pRaw = mndVgroupActionEncode(&newVg1);
+    SSdbRaw *pRaw = mndVgroupActionEncode(&newVg1);
-    if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) goto _OVER;
+    if (pRaw == NULL) return -1;
+    if (mndTransAppendCommitlog(pTrans, pRaw) != 0) {
+      sdbFreeRaw(pRaw);
+      return -1;
+    }
     (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-    pRaw = NULL;
   }
 
   {
-    pRaw = mndVgroupActionEncode(&newVg2);
+    SSdbRaw *pRaw = mndVgroupActionEncode(&newVg2);
-    if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) goto _OVER;
+    if (pRaw == NULL) return -1;
+    if (mndTransAppendCommitlog(pTrans, pRaw) != 0) {
+      sdbFreeRaw(pRaw);
+      return -1;
+    }
     (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-    pRaw = NULL;
   }
 
   mInfo("vgId:%d, vgroup info after adjust hash, replica:%d hashBegin:%u hashEnd:%u vnode:0 dnode:%d", newVg1.vgId,
@@ -1757,7 +1806,6 @@ static int32_t mndSplitVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj
 _OVER:
   taosArrayDestroy(pArray);
   mndTransDrop(pTrans);
-  sdbFreeRaw(pRaw);
   return code;
 }
 
@@ -1802,16 +1850,8 @@ static int32_t mndSetBalanceVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SD
 
   {
     SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
-    if (pRaw == NULL || mndTransAppendRedolog(pTrans, pRaw) != 0) {
+    if (pRaw == NULL) return -1;
-      sdbFreeRaw(pRaw);
+    if (mndTransAppendCommitlog(pTrans, pRaw) != 0) {
-      return -1;
-    }
-    (void)sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-  }
-
-  {
-    SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
-    if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) {
       sdbFreeRaw(pRaw);
       return -1;
     }

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

@@ -515,6 +515,16 @@ void vnodeSyncPreClose(SVnode *pVnode) {
   vInfo("vgId:%d, pre close sync", pVnode->config.vgId);
   syncLeaderTransfer(pVnode->sync);
   syncPreStop(pVnode->sync);
+
+  while (syncSnapshotRecving(pVnode->sync)) {
+    vInfo("vgId:%d, snapshot is recving", pVnode->config.vgId);
+    taosMsleep(300);
+  }
+  while (syncSnapshotSending(pVnode->sync)) {
+    vInfo("vgId:%d, snapshot is sending", pVnode->config.vgId);
+    taosMsleep(300);
+  }
+
   taosThreadMutexLock(&pVnode->lock);
   if (pVnode->blocked) {
     vInfo("vgId:%d, post block after close sync", pVnode->config.vgId);

source/libs/sync/src/syncMain.c

@@ -2467,8 +2467,12 @@ const char* syncStr(ESyncState state) {
       return "candidate";
     case TAOS_SYNC_STATE_LEADER:
       return "leader";
-    default:
+    case TAOS_SYNC_STATE_ERROR:
       return "error";
+    case TAOS_SYNC_STATE_OFFLINE:
+      return "offline";
+    default:
+      return "unknown";
   }
 }