Merge branch '3.0' into test/jcy

include/common/tmsgdef.h

@@ -194,6 +194,7 @@ enum {
   TD_DEF_MSG_TYPE(TDMT_VND_ALTER_CONFIG, "alter-config", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_ALTER_REPLICA, "alter-replica", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_ALTER_CONFIRM, "alter-confirm", NULL, NULL)
+  TD_DEF_MSG_TYPE(TDMT_VND_ALTER_HASHRANGE, "alter-hashrange", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_COMPACT, "compact", NULL, NULL)
 
   TD_NEW_MSG_SEG(TDMT_QND_MSG)
@@ -233,6 +234,8 @@ enum {
   TD_DEF_MSG_TYPE(TDMT_SYNC_COMMON_RESPONSE, "sync-common-response", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_SYNC_APPLY_MSG, "sync-apply-msg", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_SYNC_CONFIG_CHANGE, "sync-config-change", NULL, NULL)
+  TD_DEF_MSG_TYPE(TDMT_SYNC_SNAPSHOT_SEND, "sync-snapshot-send", NULL, NULL)
+  TD_DEF_MSG_TYPE(TDMT_SYNC_SNAPSHOT_RSP, "sync-snapshot-rsp", NULL, NULL)
   
 #if defined(TD_MSG_NUMBER_)
   TDMT_MAX

include/libs/function/function.h

@@ -146,6 +146,7 @@ typedef struct SqlFunctionCtx {
   struct SDiskbasedBuf  *pBuf;
   struct SSDataBlock    *pSrcBlock;
   int32_t                curBufPage;
+  bool                   increase;
 
   char                   udfName[TSDB_FUNC_NAME_LEN];
 } SqlFunctionCtx;

include/libs/stream/tstream.h

@@ -308,9 +308,11 @@ static FORCE_INLINE int32_t streamTaskOutput(SStreamTask* pTask, SStreamDataBloc
   if (pTask->sinkType == TASK_SINK__TABLE) {
     ASSERT(pTask->dispatchType == TASK_DISPATCH__NONE);
     pTask->tbSink.tbSinkFunc(pTask, pTask->tbSink.vnode, 0, pBlock->blocks);
+    taosFreeQitem(pBlock);
   } else if (pTask->sinkType == TASK_SINK__SMA) {
     ASSERT(pTask->dispatchType == TASK_DISPATCH__NONE);
     pTask->smaSink.smaSink(pTask->smaSink.vnode, pTask->smaSink.smaId, pBlock->blocks);
+    taosFreeQitem(pBlock);
   } else {
     ASSERT(pTask->dispatchType != TASK_DISPATCH__NONE);
     taosWriteQitem(pTask->outputQueue->queue, pBlock);

include/libs/sync/sync.h

@@ -88,11 +88,16 @@ typedef struct SReConfigCbMeta {
 } SReConfigCbMeta;
 
 typedef struct SSnapshot {
-  void *data;
+  void*     data;
   SyncIndex lastApplyIndex;
   SyncTerm  lastApplyTerm;
+  SyncIndex lastConfigIndex;
 } SSnapshot;
 
+typedef struct SSnapshotMeta {
+  SyncIndex lastConfigIndex;
+} SSnapshotMeta;
+
 typedef struct SSyncFSM {
   void* data;
 
@@ -141,10 +146,28 @@ typedef struct SSyncLogStore {
   // return commit index of log
   SyncIndex (*getCommitIndex)(struct SSyncLogStore* pLogStore);
 
+  // refactor, log[0 .. n] ==> log[m .. n]
+  int32_t (*syncLogSetBeginIndex)(struct SSyncLogStore* pLogStore, SyncIndex beginIndex);
+  int32_t (*syncLogResetBeginIndex)(struct SSyncLogStore* pLogStore);
+  SyncIndex (*syncLogBeginIndex)(struct SSyncLogStore* pLogStore);
+  SyncIndex (*syncLogEndIndex)(struct SSyncLogStore* pLogStore);
+  bool (*syncLogIsEmpty)(struct SSyncLogStore* pLogStore);
+  int32_t (*syncLogEntryCount)(struct SSyncLogStore* pLogStore);
+  bool (*syncLogInRange)(struct SSyncLogStore* pLogStore, SyncIndex index);
+
+  SyncIndex (*syncLogWriteIndex)(struct SSyncLogStore* pLogStore);
+  SyncIndex (*syncLogLastIndex)(struct SSyncLogStore* pLogStore);
+  SyncTerm (*syncLogLastTerm)(struct SSyncLogStore* pLogStore);
+
+  int32_t (*syncLogAppendEntry)(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry);
+  int32_t (*syncLogGetEntry)(struct SSyncLogStore* pLogStore, SyncIndex index, SSyncRaftEntry** ppEntry);
+  int32_t (*syncLogTruncate)(struct SSyncLogStore* pLogStore, SyncIndex fromIndex);
+
 } SSyncLogStore;
 
 typedef struct SSyncInfo {
   bool        isStandBy;
+  bool        snapshotEnable;
   SyncGroupId vgId;
   SSyncCfg    syncCfg;
   char        path[TSDB_FILENAME_LEN];
@@ -172,6 +195,8 @@ bool        syncEnvIsStart();
 const char* syncStr(ESyncState state);
 bool        syncIsRestoreFinish(int64_t rid);
 
+int32_t syncGetSnapshotMeta(int64_t rid, struct SSnapshotMeta* sMeta);
+
 // to be moved to static
 void syncStartNormal(int64_t rid);
 void syncStartStandBy(int64_t rid);

include/libs/sync/syncTools.h

@@ -301,6 +301,7 @@ typedef struct SyncAppendEntries {
   SyncIndex prevLogIndex;
   SyncTerm  prevLogTerm;
   SyncIndex commitIndex;
+  SyncTerm  privateTerm;
   uint32_t  dataLen;
   char      data[];
 } SyncAppendEntries;
@@ -332,6 +333,7 @@ typedef struct SyncAppendEntriesReply {
   SRaftId  destId;
   // private data
   SyncTerm  term;
+  SyncTerm  privateTerm;
   bool      success;
   SyncIndex matchIndex;
 } SyncAppendEntriesReply;
@@ -385,6 +387,75 @@ void syncApplyMsgPrint2(char* s, const SyncApplyMsg* pMsg);
 void syncApplyMsgLog(const SyncApplyMsg* pMsg);
 void syncApplyMsgLog2(char* s, const SyncApplyMsg* pMsg);
 
+// ---------------------------------------------
+typedef struct SyncSnapshotSend {
+  uint32_t bytes;
+  int32_t  vgId;
+  uint32_t msgType;
+  SRaftId  srcId;
+  SRaftId  destId;
+
+  SyncTerm  term;
+  SyncIndex lastIndex;  // lastIndex of snapshot
+  SyncTerm  lastTerm;   // lastTerm of snapshot
+  SyncTerm  privateTerm;
+  int32_t   seq;
+  uint32_t  dataLen;
+  char      data[];
+} SyncSnapshotSend;
+
+SyncSnapshotSend* syncSnapshotSendBuild(uint32_t dataLen, int32_t vgId);
+void              syncSnapshotSendDestroy(SyncSnapshotSend* pMsg);
+void              syncSnapshotSendSerialize(const SyncSnapshotSend* pMsg, char* buf, uint32_t bufLen);
+void              syncSnapshotSendDeserialize(const char* buf, uint32_t len, SyncSnapshotSend* pMsg);
+char*             syncSnapshotSendSerialize2(const SyncSnapshotSend* pMsg, uint32_t* len);
+SyncSnapshotSend* syncSnapshotSendDeserialize2(const char* buf, uint32_t len);
+void              syncSnapshotSend2RpcMsg(const SyncSnapshotSend* pMsg, SRpcMsg* pRpcMsg);
+void              syncSnapshotSendFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotSend* pMsg);
+SyncSnapshotSend* syncSnapshotSendFromRpcMsg2(const SRpcMsg* pRpcMsg);
+cJSON*            syncSnapshotSend2Json(const SyncSnapshotSend* pMsg);
+char*             syncSnapshotSend2Str(const SyncSnapshotSend* pMsg);
+
+// for debug ----------------------
+void syncSnapshotSendPrint(const SyncSnapshotSend* pMsg);
+void syncSnapshotSendPrint2(char* s, const SyncSnapshotSend* pMsg);
+void syncSnapshotSendLog(const SyncSnapshotSend* pMsg);
+void syncSnapshotSendLog2(char* s, const SyncSnapshotSend* pMsg);
+
+// ---------------------------------------------
+typedef struct SyncSnapshotRsp {
+  uint32_t bytes;
+  int32_t  vgId;
+  uint32_t msgType;
+  SRaftId  srcId;
+  SRaftId  destId;
+
+  SyncTerm  term;
+  SyncIndex lastIndex;
+  SyncTerm  lastTerm;
+  SyncTerm  privateTerm;
+  int32_t   ack;
+  int32_t   code;
+} SyncSnapshotRsp;
+
+SyncSnapshotRsp* syncSnapshotRspBuild(int32_t vgId);
+void             syncSnapshotRspDestroy(SyncSnapshotRsp* pMsg);
+void             syncSnapshotRspSerialize(const SyncSnapshotRsp* pMsg, char* buf, uint32_t bufLen);
+void             syncSnapshotRspDeserialize(const char* buf, uint32_t len, SyncSnapshotRsp* pMsg);
+char*            syncSnapshotRspSerialize2(const SyncSnapshotRsp* pMsg, uint32_t* len);
+SyncSnapshotRsp* syncSnapshotRspDeserialize2(const char* buf, uint32_t len);
+void             syncSnapshotRsp2RpcMsg(const SyncSnapshotRsp* pMsg, SRpcMsg* pRpcMsg);
+void             syncSnapshotRspFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotRsp* pMsg);
+SyncSnapshotRsp* syncSnapshotRspFromRpcMsg2(const SRpcMsg* pRpcMsg);
+cJSON*           syncSnapshotRsp2Json(const SyncSnapshotRsp* pMsg);
+char*            syncSnapshotRsp2Str(const SyncSnapshotRsp* pMsg);
+
+// for debug ----------------------
+void syncSnapshotRspPrint(const SyncSnapshotRsp* pMsg);
+void syncSnapshotRspPrint2(char* s, const SyncSnapshotRsp* pMsg);
+void syncSnapshotRspLog(const SyncSnapshotRsp* pMsg);
+void syncSnapshotRspLog2(char* s, const SyncSnapshotRsp* pMsg);
+
 // on message ----------------------
 int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg);
 int32_t syncNodeOnPingReplyCb(SSyncNode* ths, SyncPingReply* pMsg);
@@ -395,6 +466,29 @@ int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg)
 int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg);
 int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
 
+int32_t syncNodeOnRequestVoteSnapshotCb(SSyncNode* ths, SyncRequestVote* pMsg);
+int32_t syncNodeOnRequestVoteReplySnapshotCb(SSyncNode* ths, SyncRequestVoteReply* pMsg);
+int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg);
+int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
+
+int32_t syncNodeOnSnapshotSendCb(SSyncNode* ths, SyncSnapshotSend* pMsg);
+int32_t syncNodeOnSnapshotRspCb(SSyncNode* ths, SyncSnapshotRsp* pMsg);
+
+// -----------------------------------------
+typedef int32_t (*FpOnPingCb)(SSyncNode* ths, SyncPing* pMsg);
+typedef int32_t (*FpOnPingReplyCb)(SSyncNode* ths, SyncPingReply* pMsg);
+typedef int32_t (*FpOnClientRequestCb)(SSyncNode* ths, SyncClientRequest* pMsg);
+typedef int32_t (*FpOnRequestVoteCb)(SSyncNode* ths, SyncRequestVote* pMsg);
+typedef int32_t (*FpOnRequestVoteReplyCb)(SSyncNode* ths, SyncRequestVoteReply* pMsg);
+typedef int32_t (*FpOnAppendEntriesCb)(SSyncNode* ths, SyncAppendEntries* pMsg);
+typedef int32_t (*FpOnAppendEntriesReplyCb)(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
+typedef int32_t (*FpOnTimeoutCb)(SSyncNode* pSyncNode, SyncTimeout* pMsg);
+typedef int32_t (*FpOnSnapshotSendCb)(SSyncNode* ths, SyncSnapshotSend* pMsg);
+typedef int32_t (*FpOnSnapshotRspCb)(SSyncNode* ths, SyncSnapshotRsp* pMsg);
+
+// option ----------------------------------
+bool syncNodeSnapshotEnable(SSyncNode* pSyncNode);
+
 // ---------------------------------------------
 
 #ifdef __cplusplus

include/libs/wal/wal.h

@@ -141,6 +141,8 @@ typedef struct SWal {
   // ctl
   int64_t       refId;
   TdThreadMutex mutex;
+  // ref
+  SHashObj *pRefHash;  // ref -> SWalRef
   // path
   char path[WAL_PATH_LEN];
   // reusable write head
@@ -184,7 +186,7 @@ int32_t walRollback(SWal *, int64_t ver);
 // notify that previous logs can be pruned safely
 int32_t walBeginSnapshot(SWal *, int64_t ver);
 int32_t walEndSnapshot(SWal *);
-void    walRestoreFromSnapshot(SWal *, int64_t ver);
+int32_t walRestoreFromSnapshot(SWal *, int64_t ver);
 // int32_t  walDataCorrupted(SWal*);
 
 // read
@@ -199,6 +201,16 @@ int32_t walFetchHead(SWalReadHandle *pRead, int64_t ver, SWalHead *pHead);
 int32_t walFetchBody(SWalReadHandle *pRead, SWalHead **ppHead);
 int32_t walSkipFetchBody(SWalReadHandle *pRead, const SWalHead *pHead);
 
+typedef struct {
+  int64_t refId;
+  int64_t ver;
+} SWalRef;
+
+SWalRef *walOpenRef(SWal *);
+void     walCloseRef(SWalRef *);
+int32_t  walRefVer(SWalRef *, int64_t ver);
+int32_t  walUnrefVer(SWal *);
+
 // deprecated
 #if 0
 int32_t walRead(SWal *, SWalHead **, int64_t ver);

source/common/src/tdatablock.c

@@ -1752,7 +1752,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
       taosArrayClear(tagArray);
       taosArrayPush(tagArray, &tagVal);
       tTagNew(tagArray, 1, false, &pTag);
-      if (!pTag) {
+      if (pTag == NULL) {
         tdDestroySVCreateTbReq(&createTbReq);
         taosArrayDestroy(tagArray);
         return NULL;
@@ -1763,9 +1763,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
       tEncodeSize(tEncodeSVCreateTbReq, &createTbReq, schemaLen, code);
 
       tdDestroySVCreateTbReq(&createTbReq);
-
       if (code < 0) {
-        tdDestroySVCreateTbReq(&createTbReq);
         taosArrayDestroy(tagArray);
         return NULL;
       }
@@ -1775,6 +1773,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
   }
 
   // assign data
+  // TODO
   ret = taosMemoryCalloc(1, cap + 46);
   ret = POINTER_SHIFT(ret, 46);
   ret->header.vgId = vgId;
@@ -1804,8 +1803,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
     int32_t schemaLen = 0;
     if (createTb) {
       SVCreateTbReq createTbReq = {0};
-      char*         cname = taosMemoryCalloc(1, TSDB_TABLE_FNAME_LEN);
-      snprintf(cname, TSDB_TABLE_FNAME_LEN, "%s:%ld", stbFullName, pDataBlock->info.groupId);
+      char*         cname = buildCtbNameByGroupId(stbFullName, pDataBlock->info.groupId);
       createTbReq.name = cname;
       createTbReq.flags = 0;
       createTbReq.type = TSDB_CHILD_TABLE;
@@ -1819,7 +1817,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
       taosArrayPush(tagArray, &tagVal);
       STag* pTag = NULL;
       tTagNew(tagArray, 1, false, &pTag);
-      if (!pTag) {
+      if (pTag == NULL) {
         tdDestroySVCreateTbReq(&createTbReq);
         taosArrayDestroy(tagArray);
         taosMemoryFreeClear(ret);
@@ -1945,7 +1943,6 @@ void blockCompressEncode(const SSDataBlock* pBlock, char* data, int32_t* dataLen
 
 const char* blockCompressDecode(SSDataBlock* pBlock, int32_t numOfCols, int32_t numOfRows, const char* pData) {
   blockDataEnsureCapacity(pBlock, numOfRows);
-  pBlock->info.rows = numOfRows;
 
   const char* pStart = pData;
 
@@ -2019,6 +2016,7 @@ const char* blockCompressDecode(SSDataBlock* pBlock, int32_t numOfCols, int32_t
     pStart += colLen[i];
   }
 
+  pBlock->info.rows = numOfRows;
   ASSERT(pStart - pData == dataLen);
   return pStart;
 }

source/dnode/mgmt/mgmt_mnode/src/mmHandle.c

@@ -215,8 +215,12 @@ SArray *mmGetMsgHandles() {
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIG_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_REPLICA_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIRM_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_HASHRANGE_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_COMPACT_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
 
+  if (dmSetMgmtHandle(pArray, TDMT_MON_MM_INFO, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_MON_MM_LOAD, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
+
   if (dmSetMgmtHandle(pArray, TDMT_SYNC_TIMEOUT, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_SYNC_PING, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_SYNC_PING_REPLY, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
@@ -226,9 +230,8 @@ SArray *mmGetMsgHandles() {
   if (dmSetMgmtHandle(pArray, TDMT_SYNC_REQUEST_VOTE_REPLY, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_SYNC_APPEND_ENTRIES, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_SYNC_APPEND_ENTRIES_REPLY, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
-
-  if (dmSetMgmtHandle(pArray, TDMT_MON_MM_INFO, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
-  if (dmSetMgmtHandle(pArray, TDMT_MON_MM_LOAD, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_SYNC_SNAPSHOT_SEND, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_SYNC_SNAPSHOT_RSP, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
 
   code = 0;
 

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

@@ -71,7 +71,7 @@ static void mmProcessSyncQueue(SQueueInfo *pInfo, SRpcMsg *pMsg) {
 }
 
 static int32_t mmPutNodeMsgToWorker(SSingleWorker *pWorker, SRpcMsg *pMsg) {
-  dTrace("msg:%p, put into worker %s, type:%s", pMsg, pWorker->name, TMSG_INFO(pMsg->msgType));
+  dTrace("msg:%p, put into %s queue, type:%s", pMsg, pWorker->name, TMSG_INFO(pMsg->msgType));
   taosWriteQitem(pWorker->queue, pMsg);
   return 0;
 }

source/dnode/mgmt/mgmt_vnode/src/vmHandle.c

@@ -359,6 +359,7 @@ SArray *vmGetMsgHandles() {
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_REPLICA, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIG, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIRM, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_HASHRANGE, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_COMPACT, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_DND_CREATE_VNODE, vmPutMsgToMgmtQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_DND_DROP_VNODE, vmPutMsgToMgmtQueue, 0) == NULL) goto _OVER;

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

@@ -19,6 +19,7 @@
 #include "mndDef.h"
 
 #include "sdb.h"
+#include "sync.h"
 #include "syncTools.h"
 #include "tcache.h"
 #include "tdatablock.h"
@@ -75,7 +76,6 @@ typedef struct {
 } STelemMgmt;
 
 typedef struct {
-  SWal   *pWal;
   sem_t   syncSem;
   int64_t sync;
   bool    standby;
@@ -108,6 +108,7 @@ typedef struct SMnode {
   SQHandle      *pQuery;
   SHashObj      *infosMeta;
   SHashObj      *perfsMeta;
+  SWal          *pWal;
   SShowMgmt      showMgmt;
   SProfileMgmt   profileMgmt;
   STelemMgmt     telemMgmt;

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

@@ -41,6 +41,7 @@ int32_t mndAddAlterVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pV
 int32_t mndAddDropVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid, bool isRedo);
 int32_t mndSetMoveVgroupInfoToTrans(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, int32_t vn, SArray *pArray);
 int32_t mndSetMoveVgroupsInfoToTrans(SMnode *, STrans *pTrans, int32_t dropDnodeId);
+int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray);
 
 void *mndBuildCreateVnodeReq(SMnode *, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *cntlen, bool standby);
 void *mndBuildDropVnodeReq(SMnode *, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);

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

@@ -636,57 +636,6 @@ static int32_t mndSetAlterDbCommitLogs(SMnode *pMnode, STrans *pTrans, SDbObj *p
   return 0;
 }
 
-static int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray) {
-  if (pVgroup->replica <= 0 || pVgroup->replica == pDb->cfg.replications) {
-    if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, TDMT_VND_ALTER_CONFIG) != 0) {
-      return -1;
-    }
-  } else {
-    SVgObj newVgroup = {0};
-    memcpy(&newVgroup, pVgroup, sizeof(SVgObj));
-    mndTransSetSerial(pTrans);
-
-    if (newVgroup.replica < pDb->cfg.replications) {
-      mInfo("db:%s, vgId:%d, vn:0 dnode:%d, will add 2 vnodes", pVgroup->dbName, pVgroup->vgId,
-            pVgroup->vnodeGid[0].dnodeId);
-
-      if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
-      if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[1], true) != 0) return -1;
-      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
-      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
-
-      if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
-      if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[2], true) != 0) return -1;
-      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
-      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
-    } else {
-      mInfo("db:%s, vgId:%d, will remove 2 vnodes", pVgroup->dbName, pVgroup->vgId);
-
-      SVnodeGid del1 = {0};
-      if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del1) != 0) return -1;
-      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
-      if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del1, true) != 0) return -1;
-      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
-
-      SVnodeGid del2 = {0};
-      if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del2) != 0) return -1;
-      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
-      if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del2, true) != 0) return -1;
-      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
-    }
-
-    SSdbRaw *pVgRaw = mndVgroupActionEncode(&newVgroup);
-    if (pVgRaw == NULL) return -1;
-    if (mndTransAppendCommitlog(pTrans, pVgRaw) != 0) {
-      sdbFreeRaw(pVgRaw);
-      return -1;
-    }
-    sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
-  }
-
-  return 0;
-}
-
 static int32_t mndSetAlterDbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj *pOld, SDbObj *pNew) {
   SSdb   *pSdb = pMnode->pSdb;
   void   *pIter = NULL;

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

@@ -566,9 +566,11 @@ static int32_t mndDropDnode(SMnode *pMnode, SRpcMsg *pReq, SDnodeObj *pDnode, SM
   pRaw = NULL;
 
   if (pMObj != NULL) {
+    mDebug("trans:%d, mnode on dnode:%d will be dropped", pTrans->id, pDnode->id);
     if (mndSetDropMnodeInfoToTrans(pMnode, pTrans, pMObj) != 0) goto _OVER;
   }
   if (numOfVnodes > 0) {
+    mDebug("trans:%d, %d vnodes on dnode:%d will be dropped", pTrans->id, numOfVnodes, pDnode->id);
     if (mndSetMoveVgroupsInfoToTrans(pMnode, pTrans, pDnode->id) != 0) goto _OVER;
   }
   if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;

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

@@ -139,10 +139,40 @@ static int32_t mndCreateDir(SMnode *pMnode, const char *path) {
   return 0;
 }
 
+static int32_t mndInitWal(SMnode *pMnode) {
+  char path[PATH_MAX + 20] = {0};
+  snprintf(path, sizeof(path), "%s%swal", pMnode->path, TD_DIRSEP);
+  SWalCfg cfg = {
+      .vgId = 1,
+      .fsyncPeriod = 0,
+      .rollPeriod = -1,
+      .segSize = -1,
+      .retentionPeriod = -1,
+      .retentionSize = -1,
+      .level = TAOS_WAL_FSYNC,
+  };
+
+  pMnode->pWal = walOpen(path, &cfg);
+  if (pMnode->pWal == NULL) {
+    mError("failed to open wal since %s", terrstr());
+    return -1;
+  }
+
+  return 0;
+}
+
+static void mndCloseWal(SMnode *pMnode) {
+  if (pMnode->pWal != NULL) {
+    walClose(pMnode->pWal);
+    pMnode->pWal = NULL;
+  }
+}
+
 static int32_t mndInitSdb(SMnode *pMnode) {
   SSdbOpt opt = {0};
   opt.path = pMnode->path;
   opt.pMnode = pMnode;
+  opt.pWal = pMnode->pWal;
 
   pMnode->pSdb = sdbInit(&opt);
   if (pMnode->pSdb == NULL) {
@@ -156,7 +186,6 @@ static int32_t mndOpenSdb(SMnode *pMnode) {
   if (!pMnode->deploy) {
     return sdbReadFile(pMnode->pSdb);
   } else {
-    // return sdbDeploy(pMnode->pSdb);;
     return 0;
   }
 }
@@ -182,6 +211,7 @@ static int32_t mndAllocStep(SMnode *pMnode, char *name, MndInitFp initFp, MndCle
 }
 
 static int32_t mndInitSteps(SMnode *pMnode) {
+  if (mndAllocStep(pMnode, "mnode-wal", mndInitWal, mndCloseWal) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-sdb", mndInitSdb, mndCleanupSdb) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-trans", mndInitTrans, mndCleanupTrans) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-cluster", mndInitCluster, mndCleanupCluster) != 0) return -1;
@@ -201,7 +231,7 @@ static int32_t mndInitSteps(SMnode *pMnode) {
   if (mndAllocStep(pMnode, "mnode-offset", mndInitOffset, mndCleanupOffset) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-vgroup", mndInitVgroup, mndCleanupVgroup) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-stb", mndInitStb, mndCleanupStb) != 0) return -1;
-  if (mndAllocStep(pMnode, "mnode-stb", mndInitSma, mndCleanupSma) != 0) return -1;
+  if (mndAllocStep(pMnode, "mnode-sma", mndInitSma, mndCleanupSma) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-infos", mndInitInfos, mndCleanupInfos) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-perfs", mndInitPerfs, mndCleanupPerfs) != 0) return -1;
   if (mndAllocStep(pMnode, "mnode-db", mndInitDb, mndCleanupDb) != 0) return -1;
@@ -376,41 +406,93 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
   syncRpcMsgLog2(logBuf, pMsg);
   taosMemoryFree(syncNodeStr);
 
-  if (pMsg->msgType == TDMT_SYNC_TIMEOUT) {
-    SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pMsg);
-    code = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
-    syncTimeoutDestroy(pSyncMsg);
-  } else if (pMsg->msgType == TDMT_SYNC_PING) {
-    SyncPing *pSyncMsg = syncPingFromRpcMsg2(pMsg);
-    code = syncNodeOnPingCb(pSyncNode, pSyncMsg);
-    syncPingDestroy(pSyncMsg);
-  } else if (pMsg->msgType == TDMT_SYNC_PING_REPLY) {
-    SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pMsg);
-    code = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
-    syncPingReplyDestroy(pSyncMsg);
-  } else if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
-    SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pMsg);
-    code = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
-    syncClientRequestDestroy(pSyncMsg);
-  } else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
-    SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pMsg);
-    code = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
-    syncRequestVoteDestroy(pSyncMsg);
-  } else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
-    SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pMsg);
-    code = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
-    syncRequestVoteReplyDestroy(pSyncMsg);
-  } else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
-    SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pMsg);
-    code = syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
-    syncAppendEntriesDestroy(pSyncMsg);
-  } else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
-    SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pMsg);
-    code = syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
-    syncAppendEntriesReplyDestroy(pSyncMsg);
+  // ToDo: ugly! use function pointer
+  if (syncNodeSnapshotEnable(pSyncNode)) {
+    if (pMsg->msgType == TDMT_SYNC_TIMEOUT) {
+      SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pMsg);
+      code = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
+      syncTimeoutDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_PING) {
+      SyncPing *pSyncMsg = syncPingFromRpcMsg2(pMsg);
+      code = syncNodeOnPingCb(pSyncNode, pSyncMsg);
+      syncPingDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_PING_REPLY) {
+      SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pMsg);
+      code = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
+      syncPingReplyDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
+      SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pMsg);
+      code = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
+      syncClientRequestDestroy(pSyncMsg);
+
+    } else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
+      SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pMsg);
+      code = syncNodeOnRequestVoteSnapshotCb(pSyncNode, pSyncMsg);
+      syncRequestVoteDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
+      SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pMsg);
+      code = syncNodeOnRequestVoteReplySnapshotCb(pSyncNode, pSyncMsg);
+      syncRequestVoteReplyDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
+      SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pMsg);
+      code = syncNodeOnAppendEntriesSnapshotCb(pSyncNode, pSyncMsg);
+      syncAppendEntriesDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
+      SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pMsg);
+      code = syncNodeOnAppendEntriesReplySnapshotCb(pSyncNode, pSyncMsg);
+      syncAppendEntriesReplyDestroy(pSyncMsg);
+
+    } else if (pMsg->msgType == TDMT_SYNC_SNAPSHOT_SEND) {
+      SyncSnapshotSend *pSyncMsg = syncSnapshotSendFromRpcMsg2(pMsg);
+      code = syncNodeOnSnapshotSendCb(pSyncNode, pSyncMsg);
+      syncSnapshotSendDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_SNAPSHOT_RSP) {
+      SyncSnapshotRsp *pSyncMsg = syncSnapshotRspFromRpcMsg2(pMsg);
+      code = syncNodeOnSnapshotRspCb(pSyncNode, pSyncMsg);
+      syncSnapshotRspDestroy(pSyncMsg);
+
+    } else {
+      mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
+      code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    }
+
   } else {
-    mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
-    code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    if (pMsg->msgType == TDMT_SYNC_TIMEOUT) {
+      SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pMsg);
+      code = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
+      syncTimeoutDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_PING) {
+      SyncPing *pSyncMsg = syncPingFromRpcMsg2(pMsg);
+      code = syncNodeOnPingCb(pSyncNode, pSyncMsg);
+      syncPingDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_PING_REPLY) {
+      SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pMsg);
+      code = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
+      syncPingReplyDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
+      SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pMsg);
+      code = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
+      syncClientRequestDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
+      SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pMsg);
+      code = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
+      syncRequestVoteDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
+      SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pMsg);
+      code = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
+      syncRequestVoteReplyDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
+      SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pMsg);
+      code = syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
+      syncAppendEntriesDestroy(pSyncMsg);
+    } else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
+      SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pMsg);
+      code = syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
+      syncAppendEntriesReplyDestroy(pSyncMsg);
+    } else {
+      mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
+      code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
+    }
   }
 
   mndReleaseSyncRef(pMnode);

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

@@ -186,7 +186,7 @@ int32_t mndAddShuffledSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* p
     SVgObj* pVgroup;
     pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void**)&pVgroup);
     if (pIter == NULL) break;
-    if (pVgroup->dbUid != pStream->dbUid) {
+    if (strcmp(pVgroup->dbName, pStream->targetDb) != 0) {
       sdbRelease(pSdb, pVgroup);
       continue;
     }
@@ -286,7 +286,8 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
   pStream->tasks = taosArrayInit(totLevel, sizeof(void*));
 
   bool hasExtraSink = false;
-  if (totLevel == 2 || strcmp(pStream->sourceDb, pStream->targetDb) != 0) {
+  bool externalTargetDB = strcmp(pStream->sourceDb, pStream->targetDb) != 0;
+  if (totLevel == 2 || externalTargetDB) {
     SArray* taskOneLevel = taosArrayInit(0, sizeof(void*));
     taosArrayPush(pStream->tasks, &taskOneLevel);
     // add extra sink
@@ -405,7 +406,6 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
         if (pStream->fixedSinkVgId == 0) {
           pTask->dispatchType = TASK_DISPATCH__SHUFFLE;
 
-          /*pTask->dispatchMsgType = TDMT_VND_TASK_WRITE_EXEC;*/
           pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
           SDbObj* pDb = mndAcquireDb(pMnode, pStream->targetDb);
           ASSERT(pDb);
@@ -426,10 +426,8 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
             SVgroupInfo* pVgInfo = taosArrayGet(pVgs, i);
             for (int32_t j = 0; j < sinkLvSize; j++) {
               SStreamTask* pLastLevelTask = taosArrayGetP(sinkLv, j);
-              /*printf("vgid %d node id %d\n", pVgInfo->vgId, pTask->nodeId);*/
               if (pLastLevelTask->nodeId == pVgInfo->vgId) {
                 pVgInfo->taskId = pLastLevelTask->taskId;
-                /*printf("taskid %d set to %d\n", pVgInfo->taskId, pTask->taskId);*/
                 break;
               }
             }

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

@@ -17,15 +17,27 @@
 #include "mndSync.h"
 #include "mndTrans.h"
 
-int32_t mndSyncEqMsg(const SMsgCb *msgcb, SRpcMsg *pMsg) {
+static int32_t mndSyncEqMsg(const SMsgCb *msgcb, SRpcMsg *pMsg) {
   SMsgHead *pHead = pMsg->pCont;
   pHead->contLen = htonl(pHead->contLen);
   pHead->vgId = htonl(pHead->vgId);
 
-  return tmsgPutToQueue(msgcb, SYNC_QUEUE, pMsg);
+  int32_t code = tmsgPutToQueue(msgcb, SYNC_QUEUE, pMsg);
+  if (code != 0) {
+    rpcFreeCont(pMsg->pCont);
+    pMsg->pCont = NULL;
+  }
+  return code;
 }
 
-int32_t mndSyncSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) { return tmsgSendReq(pEpSet, pMsg); }
+static int32_t mndSyncSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) {
+  int32_t code = tmsgSendReq(pEpSet, pMsg);
+  if (code != 0) {
+    rpcFreeCont(pMsg->pCont);
+    pMsg->pCont = NULL;
+  }
+  return code;
+}
 
 void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta) {
   SMnode    *pMnode = pFsm->data;
@@ -34,7 +46,7 @@ void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbM
 
   int32_t transId = sdbGetIdFromRaw(pMnode->pSdb, pRaw);
   pMgmt->errCode = cbMeta.code;
-  mTrace("trans:%d, is proposed, savedTransId:%d code:0x%x, ver:%" PRId64 " term:%" PRId64 " role:%s raw:%p", transId,
+  mDebug("trans:%d, is proposed, saved:%d code:0x%x, index:%" PRId64 " term:%" PRId64 " role:%s raw:%p", transId,
          pMgmt->transId, cbMeta.code, cbMeta.index, cbMeta.term, syncStr(cbMeta.state), pRaw);
 
   if (pMgmt->errCode == 0) {
@@ -50,6 +62,10 @@ void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbM
     tsem_post(&pMgmt->syncSem);
   } else {
     if (cbMeta.index - sdbGetApplyIndex(pMnode->pSdb) > 100) {
+      SSnapshotMeta sMeta = {0};
+      if (syncGetSnapshotMeta(pMnode->syncMgmt.sync, &sMeta) == 0) {
+        sdbSetCurConfig(pMnode->pSdb, sMeta.lastConfigIndex);
+      }
       sdbWriteFile(pMnode->pSdb);
     }
   }
@@ -57,13 +73,20 @@ void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbM
 
 int32_t mndSyncGetSnapshot(struct SSyncFSM *pFsm, SSnapshot *pSnapshot) {
   SMnode *pMnode = pFsm->data;
-  pSnapshot->lastApplyIndex = sdbGetApplyIndex(pMnode->pSdb);
-  pSnapshot->lastApplyTerm = sdbGetApplyTerm(pMnode->pSdb);
+  pSnapshot->lastApplyIndex = sdbGetCommitIndex(pMnode->pSdb);
+  pSnapshot->lastApplyTerm = sdbGetCommitTerm(pMnode->pSdb);
+  pSnapshot->lastConfigIndex = sdbGetCurConfig(pMnode->pSdb);
   return 0;
 }
 
 void mndRestoreFinish(struct SSyncFSM *pFsm) {
   SMnode *pMnode = pFsm->data;
+
+  SSnapshotMeta sMeta = {0};
+  if (syncGetSnapshotMeta(pMnode->syncMgmt.sync, &sMeta) == 0) {
+    sdbSetCurConfig(pMnode->pSdb, sMeta.lastConfigIndex);
+  }
+
   if (!pMnode->deploy) {
     mInfo("mnode sync restore finished, and will handle outstanding transactions");
     mndTransPullup(pMnode);
@@ -78,8 +101,8 @@ void mndReConfig(struct SSyncFSM *pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta)
   SSyncMgmt *pMgmt = &pMnode->syncMgmt;
 
   pMgmt->errCode = cbMeta.code;
-  mInfo("trans:-1, sync reconfig is proposed, savedTransId:%d code:0x%x, curTerm:%" PRId64 " term:%" PRId64,
-        pMgmt->transId, cbMeta.code, cbMeta.index, cbMeta.term);
+  mInfo("trans:-1, sync reconfig is proposed, saved:%d code:0x%x, index:%" PRId64 " term:%" PRId64, pMgmt->transId,
+        cbMeta.code, cbMeta.index, cbMeta.term);
 
   if (pMgmt->transId == -1) {
     if (pMgmt->errCode != 0) {
@@ -144,29 +167,12 @@ SSyncFSM *mndSyncMakeFsm(SMnode *pMnode) {
 int32_t mndInitSync(SMnode *pMnode) {
   SSyncMgmt *pMgmt = &pMnode->syncMgmt;
 
-  char path[PATH_MAX + 20] = {0};
-  snprintf(path, sizeof(path), "%s%swal", pMnode->path, TD_DIRSEP);
-  SWalCfg cfg = {
-      .vgId = 1,
-      .fsyncPeriod = 0,
-      .rollPeriod = -1,
-      .segSize = -1,
-      .retentionPeriod = -1,
-      .retentionSize = -1,
-      .level = TAOS_WAL_FSYNC,
-  };
-
-  pMgmt->pWal = walOpen(path, &cfg);
-  if (pMgmt->pWal == NULL) {
-    mError("failed to open wal since %s", terrstr());
-    return -1;
-  }
-
   SSyncInfo syncInfo = {.vgId = 1, .FpSendMsg = mndSyncSendMsg, .FpEqMsg = mndSyncEqMsg};
   snprintf(syncInfo.path, sizeof(syncInfo.path), "%s%ssync", pMnode->path, TD_DIRSEP);
-  syncInfo.pWal = pMgmt->pWal;
+  syncInfo.pWal = pMnode->pWal;
   syncInfo.pFsm = mndSyncMakeFsm(pMnode);
   syncInfo.isStandBy = pMgmt->standby;
+  syncInfo.snapshotEnable = true;
 
   SSyncCfg *pCfg = &syncInfo.syncCfg;
   pCfg->replicaNum = pMnode->replica;
@@ -196,10 +202,6 @@ void mndCleanupSync(SMnode *pMnode) {
   mDebug("mnode sync is stopped, id:%" PRId64, pMgmt->sync);
 
   tsem_destroy(&pMgmt->syncSem);
-  if (pMgmt->pWal != NULL) {
-    walClose(pMgmt->pWal);
-  }
-
   memset(pMgmt, 0, sizeof(SSyncMgmt));
 }
 

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

@@ -922,7 +922,7 @@ static int32_t mndTransSendSingleMsg(SMnode *pMnode, STrans *pTrans, STransActio
   char    detail[1024] = {0};
   int32_t len = snprintf(detail, sizeof(detail), "msgType:%s numOfEps:%d inUse:%d", TMSG_INFO(pAction->msgType),
                          pAction->epSet.numOfEps, pAction->epSet.inUse);
-  for (int32_t i = 0; i < pTrans->lastErrorEpset.numOfEps; ++i) {
+  for (int32_t i = 0; i < pAction->epSet.numOfEps; ++i) {
     len += snprintf(detail + len, sizeof(detail) - len, " ep:%d-%s:%u", i, pAction->epSet.eps[i].fqdn,
                     pAction->epSet.eps[i].port);
   }
@@ -1085,6 +1085,8 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans)
     }
 
     if (code == 0) {
+      if (!pMnode->deploy && !mndIsMaster(pMnode)) break;
+
       pTrans->code = 0;
       pTrans->redoActionPos++;
       mDebug("trans:%d, %s:%d is executed and need sync to other mnodes", pTrans->id, mndTransStr(pAction->stage),
@@ -1386,6 +1388,10 @@ void mndTransPullup(SMnode *pMnode) {
     mndReleaseTrans(pMnode, pTrans);
   }
 
+  SSnapshotMeta sMeta = {0};
+  if (syncGetSnapshotMeta(pMnode->syncMgmt.sync, &sMeta) == 0) {
+    sdbSetCurConfig(pMnode->pSdb, sMeta.lastConfigIndex);
+  }
   sdbWriteFile(pMnode->pSdb);
   taosArrayDestroy(pArray);
 }

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

@@ -55,6 +55,7 @@ int32_t mndInitVgroup(SMnode *pMnode) {
   mndSetMsgHandle(pMnode, TDMT_VND_ALTER_REPLICA_RSP, mndTransProcessRsp);
   mndSetMsgHandle(pMnode, TDMT_VND_ALTER_CONFIG_RSP, mndTransProcessRsp);
   mndSetMsgHandle(pMnode, TDMT_VND_ALTER_CONFIRM_RSP, mndTransProcessRsp);
+  mndSetMsgHandle(pMnode, TDMT_VND_ALTER_HASHRANGE_RSP, mndTransProcessRsp);
   mndSetMsgHandle(pMnode, TDMT_DND_DROP_VNODE_RSP, mndTransProcessRsp);
   mndSetMsgHandle(pMnode, TDMT_VND_COMPACT_RSP, mndTransProcessRsp);
 
@@ -1166,7 +1167,155 @@ _OVER:
   return code;
 }
 
-static int32_t mndProcessSplitVgroupMsg(SRpcMsg *pReq) { return 0; }
+int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray) {
+  if (pVgroup->replica <= 0 || pVgroup->replica == pDb->cfg.replications) {
+    if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, TDMT_VND_ALTER_CONFIG) != 0) {
+      return -1;
+    }
+  } else {
+    SVgObj newVgroup = {0};
+    memcpy(&newVgroup, pVgroup, sizeof(SVgObj));
+    mndTransSetSerial(pTrans);
+
+    if (newVgroup.replica < pDb->cfg.replications) {
+      mInfo("db:%s, vgId:%d, vn:0 dnode:%d, will add 2 vnodes", pVgroup->dbName, pVgroup->vgId,
+            pVgroup->vnodeGid[0].dnodeId);
+
+      if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
+      if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[1], true) != 0) return -1;
+      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
+
+      if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
+      if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[2], true) != 0) return -1;
+      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
+    } else if (newVgroup.replica > pDb->cfg.replications) {
+      mInfo("db:%s, vgId:%d, will remove 2 vnodes", pVgroup->dbName, pVgroup->vgId);
+
+      SVnodeGid del1 = {0};
+      if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del1) != 0) return -1;
+      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+      if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del1, true) != 0) return -1;
+      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
+
+      SVnodeGid del2 = {0};
+      if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del2) != 0) return -1;
+      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+      if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del2, true) != 0) return -1;
+      if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
+    } else {
+    }
+
+    SSdbRaw *pVgRaw = mndVgroupActionEncode(&newVgroup);
+    if (pVgRaw == NULL) return -1;
+    if (mndTransAppendCommitlog(pTrans, pVgRaw) != 0) {
+      sdbFreeRaw(pVgRaw);
+      return -1;
+    }
+    sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
+  }
+
+  return 0;
+}
+
+static int32_t mndAddAdjustVnodeHashRangeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup) {
+  return 0;
+}
+
+static int32_t mndSplitVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj *pVgroup) {
+  int32_t  code = -1;
+  SSdbRaw *pRaw = NULL;
+  STrans  *pTrans = NULL;
+  SArray  *pArray = mndBuildDnodesArray(pMnode, 0);
+
+  pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
+  if (pTrans == NULL) goto _OVER;
+  mndTransSetSerial(pTrans);
+  mDebug("trans:%d, used to split vgroup, vgId:%d", pTrans->id, pVgroup->vgId);
+
+  SVgObj newVg1 = {0};
+  memcpy(&newVg1, pVgroup, sizeof(SVgObj));
+  mInfo("vgId:%d, vgroup info before split, replica:%d hashBegin:%u hashEnd:%u", newVg1.vgId, newVg1.replica,
+        newVg1.hashBegin, newVg1.hashEnd);
+  for (int32_t i = 0; i < newVg1.replica; ++i) {
+    mInfo("vgId:%d, vnode:%d dnode:%d", newVg1.vgId, i, newVg1.vnodeGid[i].dnodeId);
+  }
+
+  if (newVg1.replica == 1) {
+    if (mndAddVnodeToVgroup(pMnode, &newVg1, pArray) != 0) goto _OVER;
+    if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVg1, &newVg1.vnodeGid[1], true) != 0) goto _OVER;
+    if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg1, TDMT_VND_ALTER_REPLICA) != 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 (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg1, TDMT_VND_ALTER_REPLICA) != 0) goto _OVER;
+    if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVg1, &del1, true) != 0) goto _OVER;
+    if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVg1) != 0) goto _OVER;
+  } else {
+    goto _OVER;
+  }
+
+  SVgObj newVg2 = {0};
+  memcpy(&newVg1, &newVg2, sizeof(SVgObj));
+  newVg1.replica = 1;
+  newVg1.hashEnd = (newVg1.hashBegin + newVg1.hashEnd) / 2;
+  memset(&newVg1.vnodeGid[1], 0, sizeof(SVnodeGid));
+
+  newVg2.replica = 1;
+  newVg2.hashBegin = newVg1.hashEnd + 1;
+  memcpy(&newVg2.vnodeGid[0], &newVg2.vnodeGid[1], sizeof(SVnodeGid));
+  memset(&newVg1.vnodeGid[1], 0, sizeof(SVnodeGid));
+
+  if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg1, TDMT_VND_ALTER_HASHRANGE) != 0) goto _OVER;
+  if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg2, TDMT_VND_ALTER_HASHRANGE) != 0) goto _OVER;
+
+  // adjust vgroup
+  if (mndBuildAlterVgroupAction(pMnode, pTrans, pDb, &newVg1, pArray) != 0) goto _OVER;
+  if (mndBuildAlterVgroupAction(pMnode, pTrans, pDb, &newVg2, pArray) != 0) goto _OVER;
+
+_OVER:
+  mndTransDrop(pTrans);
+  sdbFreeRaw(pRaw);
+  return code;
+}
+
+static int32_t mndProcessSplitVgroupMsg(SRpcMsg *pReq) {
+  SMnode   *pMnode = pReq->info.node;
+  int32_t   code = -1;
+  int32_t   vgId = 2;
+  SUserObj *pUser = NULL;
+  SVgObj   *pVgroup = NULL;
+  SDbObj   *pDb = NULL;
+
+  mDebug("vgId:%d, start to split", vgId);
+
+  pVgroup = mndAcquireVgroup(pMnode, vgId);
+  if (pVgroup == NULL) goto _OVER;
+
+  pDb = mndAcquireDb(pMnode, pVgroup->dbName);
+  if (pDb == NULL) goto _OVER;
+
+  pUser = mndAcquireUser(pMnode, pReq->conn.user);
+  if (pUser == NULL) {
+    terrno = TSDB_CODE_MND_NO_USER_FROM_CONN;
+    goto _OVER;
+  }
+
+  if (mndCheckNodeAuth(pUser) != 0) {
+    goto _OVER;
+  }
+
+  code = mndSplitVgroup(pMnode, pReq, pDb, pVgroup);
+  if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;
+
+_OVER:
+  mndReleaseUser(pMnode, pUser);
+  mndReleaseVgroup(pMnode, pVgroup);
+  mndReleaseDb(pMnode, pDb);
+  return code;
+}
 
 static int32_t mndSetBalanceVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup,
                                               SDnodeObj *pSrc, SDnodeObj *pDst) {

source/dnode/mnode/sdb/inc/sdb.h

@@ -22,6 +22,7 @@
 #include "tlockfree.h"
 #include "tlog.h"
 #include "tmsg.h"
+#include "wal.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -165,12 +166,14 @@ typedef struct SSdbRow {
 
 typedef struct SSdb {
   SMnode        *pMnode;
+  SWal          *pWal;
   char          *currDir;
   char          *tmpDir;
   int64_t        lastCommitVer;
   int64_t        lastCommitTerm;
   int64_t        curVer;
   int64_t        curTerm;
+  int64_t        curConfig;
   int64_t        tableVer[SDB_MAX];
   int64_t        maxId[SDB_MAX];
   EKeyType       keyTypes[SDB_MAX];
@@ -205,6 +208,7 @@ typedef struct {
 typedef struct SSdbOpt {
   const char *path;
   SMnode     *pMnode;
+  SWal       *pWal;
 } SSdbOpt;
 
 /**
@@ -358,9 +362,13 @@ int64_t sdbGetTableVer(SSdb *pSdb, ESdbType type);
  * @return int32_t The current index of sdb
  */
 void    sdbSetApplyIndex(SSdb *pSdb, int64_t index);
-int64_t sdbGetApplyIndex(SSdb *pSdb);
 void    sdbSetApplyTerm(SSdb *pSdb, int64_t term);
+void    sdbSetCurConfig(SSdb *pSdb, int64_t config);
+int64_t sdbGetApplyIndex(SSdb *pSdb);
 int64_t sdbGetApplyTerm(SSdb *pSdb);
+int64_t sdbGetCommitIndex(SSdb *pSdb);
+int64_t sdbGetCommitTerm(SSdb *pSdb);
+int64_t sdbGetCurConfig(SSdb *pSdb);
 
 SSdbRaw *sdbAllocRaw(ESdbType type, int8_t sver, int32_t dataLen);
 void     sdbFreeRaw(SSdbRaw *pRaw);

source/dnode/mnode/sdb/src/sdb.c

@@ -52,10 +52,12 @@ SSdb *sdbInit(SSdbOpt *pOption) {
     pSdb->keyTypes[i] = SDB_KEY_INT32;
   }
 
+  pSdb->pWal = pOption->pWal;
   pSdb->curVer = -1;
   pSdb->curTerm = -1;
   pSdb->lastCommitVer = -1;
   pSdb->lastCommitTerm = -1;
+  pSdb->curConfig = -1;
   pSdb->pMnode = pOption->pMnode;
   taosThreadMutexInit(&pSdb->filelock, NULL);
   mDebug("sdb init successfully");
@@ -159,8 +161,16 @@ static int32_t sdbCreateDir(SSdb *pSdb) {
 
 void sdbSetApplyIndex(SSdb *pSdb, int64_t index) { pSdb->curVer = index; }
 
-int64_t sdbGetApplyIndex(SSdb *pSdb) { return pSdb->curVer; }
-
 void sdbSetApplyTerm(SSdb *pSdb, int64_t term) { pSdb->curTerm = term; }
 
+void sdbSetCurConfig(SSdb *pSdb, int64_t config) { pSdb->curConfig = config; }
+
+int64_t sdbGetApplyIndex(SSdb *pSdb) { return pSdb->curVer; }
+
 int64_t sdbGetApplyTerm(SSdb *pSdb) { return pSdb->curTerm; }
+
+int64_t sdbGetCommitIndex(SSdb *pSdb) { return pSdb->lastCommitVer; }
+
+int64_t sdbGetCommitTerm(SSdb *pSdb) { return pSdb->lastCommitTerm; }
+
+int64_t sdbGetCurConfig(SSdb *pSdb) { return pSdb->curConfig; }
\ No newline at end of file

source/dnode/mnode/sdb/src/sdbFile.c

@@ -110,6 +110,16 @@ static int32_t sdbReadFileHead(SSdb *pSdb, TdFilePtr pFile) {
     return -1;
   }
 
+  ret = taosReadFile(pFile, &pSdb->curConfig, sizeof(int64_t));
+  if (ret < 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    return -1;
+  }
+  if (ret != sizeof(int64_t)) {
+    terrno = TSDB_CODE_FILE_CORRUPTED;
+    return -1;
+  }
+
   for (int32_t i = 0; i < SDB_TABLE_SIZE; ++i) {
     int64_t maxId = 0;
     ret = taosReadFile(pFile, &maxId, sizeof(int64_t));
@@ -173,6 +183,11 @@ static int32_t sdbWriteFileHead(SSdb *pSdb, TdFilePtr pFile) {
     return -1;
   }
 
+  if (taosWriteFile(pFile, &pSdb->curConfig, sizeof(int64_t)) != sizeof(int64_t)) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    return -1;
+  }
+
   for (int32_t i = 0; i < SDB_TABLE_SIZE; ++i) {
     int64_t maxId = 0;
     if (i < SDB_MAX) {
@@ -288,8 +303,8 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
   pSdb->lastCommitVer = pSdb->curVer;
   pSdb->lastCommitTerm = pSdb->curTerm;
   memcpy(pSdb->tableVer, tableVer, sizeof(tableVer));
-  mDebug("read sdb file:%s successfully, ver:%" PRId64 " term:%" PRId64, file, pSdb->lastCommitVer,
-         pSdb->lastCommitTerm);
+  mDebug("read sdb file:%s successfully, index:%" PRId64 " term:%" PRId64 " config:%" PRId64, file, pSdb->lastCommitVer,
+         pSdb->lastCommitTerm, pSdb->curConfig);
 
 _OVER:
   taosCloseFile(&pFile);
@@ -426,12 +441,23 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
 }
 
 int32_t sdbWriteFile(SSdb *pSdb) {
+  int32_t code = 0;
   if (pSdb->curVer == pSdb->lastCommitVer) {
     return 0;
   }
 
   taosThreadMutexLock(&pSdb->filelock);
-  int32_t code = sdbWriteFileImp(pSdb);
+  if (pSdb->pWal != NULL) {
+    code = walBeginSnapshot(pSdb->pWal, pSdb->curVer);
+  }
+  if (code == 0) {
+    code = sdbWriteFileImp(pSdb);
+  }
+  if (code == 0) {
+    if (pSdb->pWal != NULL) {
+      code = walEndSnapshot(pSdb->pWal);
+    }
+  }
   if (code != 0) {
     mError("failed to write sdb file since %s", terrstr());
   }
@@ -496,6 +522,9 @@ int32_t sdbStartRead(SSdb *pSdb, SSdbIter **ppIter) {
   snprintf(datafile, sizeof(datafile), "%s%ssdb.data", pSdb->currDir, TD_DIRSEP);
 
   taosThreadMutexLock(&pSdb->filelock);
+  int64_t commitIndex = pSdb->lastCommitVer;
+  int64_t commitTerm = pSdb->lastCommitTerm;
+  int64_t curConfig = pSdb->curConfig;
   if (taosCopyFile(datafile, pIter->name) < 0) {
     taosThreadMutexUnlock(&pSdb->filelock);
     terrno = TAOS_SYSTEM_ERROR(errno);
@@ -514,7 +543,8 @@ int32_t sdbStartRead(SSdb *pSdb, SSdbIter **ppIter) {
   }
 
   *ppIter = pIter;
-  mInfo("sdbiter:%p, is created to read snapshot, file:%s", pIter, pIter->name);
+  mInfo("sdbiter:%p, is created to read snapshot, index:%" PRId64 " term:%" PRId64 " config:%" PRId64 " file:%s", pIter,
+        commitIndex, commitTerm, curConfig, pIter->name);
   return 0;
 }
 

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

@@ -24,6 +24,7 @@ static int32_t vnodeProcessDropTbReq(SVnode *pVnode, int64_t version, void *pReq
 static int32_t vnodeProcessSubmitReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
 static int32_t vnodeProcessCreateTSmaReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
 static int32_t vnodeProcessAlterConfirmReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
+static int32_t vnodeProcessAlterHasnRangeReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
 static int32_t vnodeProcessWriteMsg(SVnode *pVnode, int64_t version, SRpcMsg *pMsg, SRpcMsg *pRsp);
 
 int32_t vnodePreprocessReq(SVnode *pVnode, SRpcMsg *pMsg) {
@@ -163,6 +164,9 @@ int32_t vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
     case TDMT_VND_ALTER_CONFIRM:
       vnodeProcessAlterConfirmReq(pVnode, version, pReq, len, pRsp);
       break;
+    case TDMT_VND_ALTER_HASHRANGE:
+      vnodeProcessAlterHasnRangeReq(pVnode, version, pReq, len, pRsp);
+      break;
     case TDMT_VND_ALTER_CONFIG:
       break;
     default:
@@ -889,3 +893,13 @@ static int32_t vnodeProcessAlterConfirmReq(SVnode *pVnode, int64_t version, void
 
   return 0;
 }
+
+static int32_t vnodeProcessAlterHasnRangeReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp) {
+  vInfo("vgId:%d, alter hashrange msg will be processed", TD_VID(pVnode));
+
+  // todo 
+  // 1. stop work 
+  // 2. adjust hash range / compact / remove wals / rename vgroups
+  // 3. reload sync
+  return 0;
+}
\ No newline at end of file

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

@@ -252,6 +252,8 @@ static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
 
 int32_t vnodeSyncOpen(SVnode *pVnode, char *path) {
   SSyncInfo syncInfo = {
+      .isStandBy = false,
+      .snapshotEnable = false,
       .vgId = pVnode->config.vgId,
       .isStandBy = pVnode->config.standby,
       .syncCfg = pVnode->config.syncCfg,

source/libs/executor/inc/executorimpl.h

@@ -804,7 +804,7 @@ SOperatorInfo* createSortedMergeOperatorInfo(SOperatorInfo** downstream, int32_t
 
 SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
                                           SSDataBlock* pResBlock, SInterval* pInterval, int32_t primaryTsSlotId,
-                                          STimeWindowAggSupp *pTwAggSupp, SExecTaskInfo* pTaskInfo);
+                                          STimeWindowAggSupp *pTwAggSupp, SExecTaskInfo* pTaskInfo, bool isStream);
 SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
     SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo, int32_t numOfChild);
 SOperatorInfo* createStreamIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
@@ -892,8 +892,8 @@ STimeWindow getActiveTimeWindow(SDiskbasedBuf* pBuf, SResultRowInfo* pResultRowI
 int32_t getNumOfRowsInTimeWindow(SDataBlockInfo* pDataBlockInfo, TSKEY* pPrimaryColumn, int32_t startPos, TSKEY ekey,
     __block_search_fn_t searchFn, STableQueryInfo* item, int32_t order);
 int32_t binarySearchForKey(char* pValue, int num, TSKEY key, int order);
-int32_t initSessionAggSupporter(SStreamAggSupporter* pSup, const char* pKey);
-int32_t initStateAggSupporter(SStreamAggSupporter* pSup, const char* pKey);
+int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey,
+    SqlFunctionCtx* pCtx, int32_t numOfOutput, size_t size);
 SResultRow* getNewResultRow(SDiskbasedBuf* pResultBuf, int64_t tableGroupId, int32_t interBufSize);
 SResultWindowInfo* getSessionTimeWindow(SArray* pWinInfos, TSKEY ts, int64_t gap, int32_t* pIndex);
 int32_t updateSessionWindowInfo(SResultWindowInfo* pWinInfo, TSKEY* pTs, int32_t rows,

source/libs/executor/src/executorimpl.c

@@ -1130,6 +1130,7 @@ SqlFunctionCtx* createSqlFunctionCtx(SExprInfo* pExprInfo, int32_t numOfOutput,
     pCtx->start.key = INT64_MIN;
     pCtx->end.key = INT64_MIN;
     pCtx->numOfParams = pExpr->base.numOfParams;
+    pCtx->increase = false;
 
     pCtx->param = pFunct->pParam;
     //    for (int32_t j = 0; j < pCtx->numOfParams; ++j) {
@@ -2008,8 +2009,16 @@ int32_t doCopyToSDataBlock(SExecTaskInfo* pTaskInfo, SSDataBlock* pBlock, SExprI
         // the _wstartts needs to copy to 20 following rows, since the results of top-k expands to 20 different rows.
         SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, slotId);
         char*            in = GET_ROWCELL_INTERBUF(pCtx[j].resultInfo);
-        for (int32_t k = 0; k < pRow->numOfRows; ++k) {
-          colDataAppend(pColInfoData, pBlock->info.rows + k, in, pCtx[j].resultInfo->isNullRes);
+        if (pCtx[j].increase) {
+          int64_t ts = *(int64_t*) in;
+          for (int32_t k = 0; k < pRow->numOfRows; ++k) {
+            colDataAppend(pColInfoData, pBlock->info.rows + k, (const char *)&ts, pCtx[j].resultInfo->isNullRes);
+            ts++;
+          }
+        } else {
+          for (int32_t k = 0; k < pRow->numOfRows; ++k) {
+            colDataAppend(pColInfoData, pBlock->info.rows + k, in, pCtx[j].resultInfo->isNullRes);
+          }
         }
       }
     }
@@ -4676,7 +4685,8 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
     }
 
     int32_t tsSlotId = ((SColumnNode*)pIntervalPhyNode->window.pTspk)->slotId;
-    pOptr = createIntervalOperatorInfo(ops[0], pExprInfo, num, pResBlock, &interval, tsSlotId, &as, pTaskInfo);
+    bool isStream = (QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL == type);
+    pOptr = createIntervalOperatorInfo(ops[0], pExprInfo, num, pResBlock, &interval, tsSlotId, &as, pTaskInfo, isStream);
 
   } else if (QUERY_NODE_PHYSICAL_PLAN_STREAM_SEMI_INTERVAL == type) {
     int32_t children = 8;
@@ -5339,7 +5349,8 @@ int32_t getOperatorExplainExecInfo(SOperatorInfo* operatorInfo, SExplainExecInfo
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, size_t size) {
+int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput, size_t size) {
+  pSup->resultRowSize = getResultRowSize(pCtx, numOfOutput);
   pSup->keySize = sizeof(int64_t) + sizeof(TSKEY);
   pSup->pKeyBuf = taosMemoryCalloc(1, pSup->keySize);
   pSup->pResultRows = taosArrayInit(1024, size);
@@ -5358,15 +5369,11 @@ int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, size
   if (bufSize <= pageSize) {
     bufSize = pageSize * 4;
   }
-  return createDiskbasedBuf(&pSup->pResultBuf, pageSize, bufSize, pKey, TD_TMP_DIR_PATH);
-}
-
-int32_t initSessionAggSupporter(SStreamAggSupporter* pSup, const char* pKey) {
-  return initStreamAggSupporter(pSup, pKey, sizeof(SResultWindowInfo));
-}
-
-int32_t initStateAggSupporter(SStreamAggSupporter* pSup, const char* pKey) {
-  return initStreamAggSupporter(pSup, pKey, sizeof(SStateWindowInfo));
+  int32_t code = createDiskbasedBuf(&pSup->pResultBuf, pageSize, bufSize, pKey, TD_TMP_DIR_PATH);
+  for(int32_t i = 0; i < numOfOutput; ++i) {
+    pCtx[i].pBuf = pSup->pResultBuf;
+  }
+  return code;
 }
 
 int64_t getSmaWaterMark(int64_t interval, double filesFactor) {

source/libs/executor/src/timewindowoperator.c

@@ -1438,9 +1438,15 @@ static bool timeWindowinterpNeeded(SqlFunctionCtx* pCtx, int32_t numOfCols, SInt
   return needed;
 }
 
+void increaseTs(SqlFunctionCtx* pCtx) {
+  if (pCtx[0].pExpr->pExpr->_function.pFunctNode->funcType == FUNCTION_TYPE_WSTARTTS) {
+    pCtx[0].increase = true;
+  }
+}
+
 SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
                                           SSDataBlock* pResBlock, SInterval* pInterval, int32_t primaryTsSlotId,
-                                          STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo) {
+                                          STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo, bool isStream) {
   SIntervalAggOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SIntervalAggOperatorInfo));
   SOperatorInfo*            pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
   if (pInfo == NULL || pOperator == NULL) {
@@ -1460,6 +1466,11 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo*
 
   int32_t code =
       initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols, pResBlock, keyBufSize, pTaskInfo->id.str);
+  
+  if (isStream) {
+    ASSERT(numOfCols > 0);
+    increaseTs(pInfo->binfo.pCtx);
+  }
 
   initExecTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pInfo->win);
 
@@ -2128,6 +2139,8 @@ SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
   SSDataBlock* pResBlock = createResDataBlock(pPhyNode->pOutputDataBlockDesc);
   int32_t code = initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols,
       pResBlock, keyBufSize, pTaskInfo->id.str);
+  ASSERT(numOfCols > 0);
+  increaseTs(pInfo->binfo.pCtx);
   initExecTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pTaskInfo->window);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
@@ -2212,6 +2225,8 @@ int32_t initBiasicInfo(SOptrBasicInfo* pBasicInfo, SExprInfo* pExprInfo,
   for (int32_t i = 0; i < numOfCols; ++i) {
     pBasicInfo->pCtx[i].pBuf = NULL;
   }
+  ASSERT(numOfCols > 0);
+  increaseTs(pBasicInfo->pCtx);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -2228,6 +2243,10 @@ void initDownStream(SOperatorInfo* downstream, SStreamAggSupporter* pAggSup, int
   pScanInfo->pUpdateInfo = updateInfoInit(60000, TSDB_TIME_PRECISION_MILLI, waterMark);
 }
 
+int32_t initSessionAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput) {
+  return initStreamAggSupporter(pSup, pKey, pCtx, numOfOutput, sizeof(SResultWindowInfo));
+}
+
 SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
                                                   SSDataBlock* pResBlock, int64_t gap, int32_t tsSlotId,
                                                   STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo) {
@@ -2244,8 +2263,8 @@ SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream, SEx
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }
-  pInfo->streamAggSup.resultRowSize = getResultRowSize(pInfo->binfo.pCtx, numOfCols);
-  code = initSessionAggSupporter(&pInfo->streamAggSup, "StreamSessionAggOperatorInfo");
+  
+  code = initSessionAggSupporter(&pInfo->streamAggSup, "StreamSessionAggOperatorInfo", pInfo->binfo.pCtx, numOfCols);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }
@@ -3097,6 +3116,10 @@ static SSDataBlock* doStreamStateAgg(SOperatorInfo* pOperator) {
   return pBInfo->pRes->info.rows == 0 ? NULL : pBInfo->pRes;
 }
 
+int32_t initStateAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput) {
+  return initStreamAggSupporter(pSup, pKey, pCtx, numOfOutput, sizeof(SStateWindowInfo));
+}
+
 SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhysiNode* pPhyNode,
                                                 SExecTaskInfo* pTaskInfo) {
   SStreamStateWinodwPhysiNode* pStateNode = (SStreamStateWinodwPhysiNode*)pPhyNode;
@@ -3130,8 +3153,7 @@ SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhys
     goto _error;
   }
 
-  pInfo->streamAggSup.resultRowSize = getResultRowSize(pInfo->binfo.pCtx, numOfCols);
-  code = initStateAggSupporter(&pInfo->streamAggSup, "StreamStateAggOperatorInfo");
+  code = initStateAggSupporter(&pInfo->streamAggSup, "StreamStateAggOperatorInfo", pInfo->binfo.pCtx, numOfCols);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }

source/libs/function/inc/builtinsimpl.h

@@ -102,6 +102,8 @@ bool getTopBotFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv);
 int32_t topFunction(SqlFunctionCtx *pCtx);
 int32_t bottomFunction(SqlFunctionCtx *pCtx);
 int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
+int32_t topCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
+int32_t bottomCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
 
 bool getSpreadFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
 bool spreadFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);

source/libs/function/src/builtins.c

@@ -1045,6 +1045,22 @@ static int32_t translateCast(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
  *
  */
 
+static bool validateHourRange(int8_t hour) {
+  if (hour < 0 || hour > 12) {
+    return false;
+  }
+
+  return true;
+}
+
+static bool validateMinuteRange(int8_t hour, int8_t minute, char sign) {
+  if (minute == 0 || (minute == 30 && (hour == 3 || hour == 5) && sign == '-')) {
+    return true;
+  }
+
+  return false;
+}
+
 static bool validateTimezoneFormat(const SValueNode* pVal) {
   if (TSDB_DATA_TYPE_BINARY != pVal->node.resType.type) {
     return false;
@@ -1053,6 +1069,8 @@ static bool validateTimezoneFormat(const SValueNode* pVal) {
   char*   tz = varDataVal(pVal->datum.p);
   int32_t len = varDataLen(pVal->datum.p);
 
+  char buf[3] = {0};
+  int8_t hour = -1, minute = -1;
   if (len == 0) {
     return false;
   } else if (len == 1 && (tz[0] == 'z' || tz[0] == 'Z')) {
@@ -1065,6 +1083,20 @@ static bool validateTimezoneFormat(const SValueNode* pVal) {
           if (!isdigit(tz[i])) {
             return false;
           }
+
+          if (i == 2) {
+            memcpy(buf, &tz[i - 1], 2);
+            hour = taosStr2Int8(buf, NULL, 10);
+            if (!validateHourRange(hour)) {
+              return false;
+            }
+          } else if (i == 4) {
+            memcpy(buf, &tz[i - 1], 2);
+            minute = taosStr2Int8(buf, NULL, 10);
+            if (!validateMinuteRange(hour, minute, tz[0])) {
+              return false;
+            }
+          }
         }
         break;
       }
@@ -1076,9 +1108,24 @@ static bool validateTimezoneFormat(const SValueNode* pVal) {
             }
             continue;
           }
+
           if (!isdigit(tz[i])) {
             return false;
           }
+
+          if (i == 2) {
+            memcpy(buf, &tz[i - 1], 2);
+            hour = taosStr2Int8(buf, NULL, 10);
+            if (!validateHourRange(hour)) {
+              return false;
+            }
+          } else if (i == 5) {
+            memcpy(buf, &tz[i - 1], 2);
+            minute = taosStr2Int8(buf, NULL, 10);
+            if (!validateMinuteRange(hour, minute, tz[0])) {
+              return false;
+            }
+          }
         }
         break;
       }
@@ -1364,6 +1411,7 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
     .initFunc     = functionSetup,
     .processFunc  = topFunction,
     .finalizeFunc = topBotFinalize,
+    .combineFunc  = topCombine,
   },
   {
     .name = "bottom",
@@ -1373,7 +1421,8 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
     .getEnvFunc   = getTopBotFuncEnv,
     .initFunc     = functionSetup,
     .processFunc  = bottomFunction,
-    .finalizeFunc = topBotFinalize
+    .finalizeFunc = topBotFinalize,
+    .combineFunc  = bottomCombine,
   },
   {
     .name = "spread",

source/libs/function/src/builtinsimpl.c

@@ -1389,6 +1389,18 @@ void setSelectivityValue(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, const STuple
   }
 }
 
+void releaseSource(STuplePos* pPos) {
+  if (pPos->pageId == -1) {
+    return ;
+  }
+  // Todo(liuyao) relase row
+}
+
+void replaceTupleData(STuplePos* pDestPos, STuplePos* pSourcePos) {
+  releaseSource(pDestPos);
+  *pDestPos = *pSourcePos;
+}
+
 int32_t minMaxCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx, int32_t isMinFunc) {
   SResultRowEntryInfo* pDResInfo = GET_RES_INFO(pDestCtx);
   SMinmaxResInfo*      pDBuf = GET_ROWCELL_INTERBUF(pDResInfo);
@@ -1400,10 +1412,12 @@ int32_t minMaxCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx, int3
     if (pSBuf->assign && 
         ( (((*(double*)&pDBuf->v) < (*(double*)&pSBuf->v)) ^ isMinFunc) || !pDBuf->assign ) ) {
       *(double*) &pDBuf->v = *(double*) &pSBuf->v;
+      replaceTupleData(&pDBuf->tuplePos, &pSBuf->tuplePos);
     }
   } else {
     if ( pSBuf->assign && ( ((pDBuf->v < pSBuf->v) ^ isMinFunc) || !pDBuf->assign ) ) {
       pDBuf->v = pSBuf->v;
+      replaceTupleData(&pDBuf->tuplePos, &pSBuf->tuplePos);
     }
   }
   pDResInfo->numOfRes = TMAX(pDResInfo->numOfRes, pSResInfo->numOfRes);
@@ -2856,7 +2870,6 @@ void copyTupleData(SqlFunctionCtx* pCtx, int32_t rowIndex, const SSDataBlock* pS
 int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
   STopBotRes*          pRes = GET_ROWCELL_INTERBUF(pEntryInfo);
-  pEntryInfo->complete = true;
 
   int32_t type = pCtx->input.pData[0]->info.type;
   int32_t slotId = pCtx->pExpr->base.resSchema.slotId;
@@ -2881,6 +2894,67 @@ int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
   return pEntryInfo->numOfRes;
 }
 
+void addResult(SqlFunctionCtx* pCtx, STopBotResItem* pSourceItem, int16_t type,
+    bool isTopQuery) {
+  SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
+  STopBotRes* pRes = getTopBotOutputInfo(pCtx);
+  int32_t     maxSize = pCtx->param[1].param.i;
+  STopBotResItem* pItems = pRes->pItems;
+  assert(pItems != NULL);
+
+  // not full yet
+  if (pEntryInfo->numOfRes < maxSize) {
+    STopBotResItem* pItem = &pItems[pEntryInfo->numOfRes];
+    pItem->v = pSourceItem->v;
+    pItem->uid = pSourceItem->uid;
+    pItem->tuplePos.pageId = -1;
+    replaceTupleData(&pItem->tuplePos, &pSourceItem->tuplePos);
+    pEntryInfo->numOfRes++;
+    taosheapsort((void*)pItems, sizeof(STopBotResItem), pEntryInfo->numOfRes, (const void*)&type, topBotResComparFn,
+                 !isTopQuery);
+  } else {  // replace the minimum value in the result
+    if ((isTopQuery && (
+        (IS_SIGNED_NUMERIC_TYPE(type) && pSourceItem->v.i > pItems[0].v.i) ||
+        (IS_UNSIGNED_NUMERIC_TYPE(type) && pSourceItem->v.u > pItems[0].v.u) ||
+        (IS_FLOAT_TYPE(type) && pSourceItem->v.d > pItems[0].v.d)))
+        || (!isTopQuery && (
+        (IS_SIGNED_NUMERIC_TYPE(type) && pSourceItem->v.i < pItems[0].v.i) ||
+        (IS_UNSIGNED_NUMERIC_TYPE(type) && pSourceItem->v.u < pItems[0].v.u) ||
+        (IS_FLOAT_TYPE(type) && pSourceItem->v.d < pItems[0].v.d))
+        )) {
+      // replace the old data and the coresponding tuple data
+      STopBotResItem* pItem = &pItems[0];
+      pItem->v = pSourceItem->v;
+      pItem->uid = pSourceItem->uid;
+
+      // save the data of this tuple by over writing the old data
+      replaceTupleData(&pItem->tuplePos, &pSourceItem->tuplePos);
+      taosheapadjust((void*)pItems, sizeof(STopBotResItem), 0, pEntryInfo->numOfRes - 1, (const void*)&type,
+                     topBotResComparFn, NULL, !isTopQuery);
+    }
+  }
+}
+
+int32_t topCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
+  int32_t type = pDestCtx->input.pData[0]->info.type;
+  SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
+  STopBotRes*          pSBuf = getTopBotOutputInfo(pSourceCtx);
+  for (int32_t i = 0; i < pSResInfo->numOfRes; i++) {
+    addResult(pDestCtx, pSBuf->pItems + i, type, true);
+  }
+  return TSDB_CODE_SUCCESS;
+}
+
+int32_t bottomCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
+  int32_t type = pDestCtx->input.pData[0]->info.type;
+  SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
+  STopBotRes*          pSBuf = getTopBotOutputInfo(pSourceCtx);
+  for (int32_t i = 0; i < pSResInfo->numOfRes; i++) {
+    addResult(pDestCtx, pSBuf->pItems + i, type, false);
+  }
+  return TSDB_CODE_SUCCESS;
+}
+
 bool getSpreadFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) {
   pEnv->calcMemSize = sizeof(SSpreadInfo);
   return true;

source/libs/sync/inc/syncAppendEntries.h

@@ -93,6 +93,7 @@ extern "C" {
 //       /\ UNCHANGED <<candidateVars, leaderVars>>
 //
 int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg);
+int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg);
 
 #ifdef __cplusplus
 }

source/libs/sync/inc/syncAppendEntriesReply.h

@@ -41,6 +41,7 @@ extern "C" {
 //    /\ UNCHANGED <<serverVars, candidateVars, logVars, elections>>
 //
 int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
+int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
 
 #ifdef __cplusplus
 }

source/libs/sync/inc/syncElection.h

@@ -39,6 +39,8 @@ extern "C" {
 //    /\ UNCHANGED <<serverVars, candidateVars, leaderVars, logVars>>
 //
 int32_t syncNodeRequestVotePeers(SSyncNode* pSyncNode);
+int32_t syncNodeRequestVotePeersSnapshot(SSyncNode* pSyncNode);
+
 int32_t syncNodeElect(SSyncNode* pSyncNode);
 int32_t syncNodeRequestVote(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncRequestVote* pMsg);
 

source/libs/sync/inc/syncIO.h

@@ -57,6 +57,9 @@ typedef struct SSyncIO {
   int32_t (*FpOnSyncAppendEntriesReply)(SSyncNode *pSyncNode, SyncAppendEntriesReply *pMsg);
   int32_t (*FpOnSyncTimeout)(SSyncNode *pSyncNode, SyncTimeout *pMsg);
 
+  int32_t (*FpOnSyncSnapshotSend)(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg);
+  int32_t (*FpOnSyncSnapshotRsp)(SSyncNode *pSyncNode, SyncSnapshotRsp *pMsg);
+
   int8_t isStart;
 
 } SSyncIO;

source/libs/sync/inc/syncIndexMgr.h

@@ -30,6 +30,7 @@ extern "C" {
 typedef struct SSyncIndexMgr {
   SRaftId (*replicas)[TSDB_MAX_REPLICA];
   SyncIndex  index[TSDB_MAX_REPLICA];
+  SyncTerm   privateTerm[TSDB_MAX_REPLICA];  // for advanced function
   int32_t    replicaNum;
   SSyncNode *pSyncNode;
 } SSyncIndexMgr;
@@ -43,6 +44,9 @@ SyncIndex      syncIndexMgrGetIndex(SSyncIndexMgr *pSyncIndexMgr, const SRaftId
 cJSON *        syncIndexMgr2Json(SSyncIndexMgr *pSyncIndexMgr);
 char *         syncIndexMgr2Str(SSyncIndexMgr *pSyncIndexMgr);
 
+// void     syncIndexMgrSetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId, SyncTerm term);
+// SyncTerm syncIndexMgrGetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId);
+
 // for debug -------------------
 void syncIndexMgrPrint(SSyncIndexMgr *pObj);
 void syncIndexMgrPrint2(char *s, SSyncIndexMgr *pObj);

source/libs/sync/inc/syncInt.h

@@ -58,6 +58,8 @@ typedef struct SSyncRespMgr           SSyncRespMgr;
 typedef struct SSyncSnapshotSender    SSyncSnapshotSender;
 typedef struct SSyncSnapshotReceiver  SSyncSnapshotReceiver;
 
+extern bool gRaftDetailLog;
+
 typedef struct SSyncNode {
   // init by SSyncInfo
   SyncGroupId vgId;
@@ -137,24 +139,27 @@ typedef struct SSyncNode {
   uint64_t          heartbeatTimerCounter;
 
   // callback
-  int32_t (*FpOnPing)(SSyncNode* ths, SyncPing* pMsg);
-  int32_t (*FpOnPingReply)(SSyncNode* ths, SyncPingReply* pMsg);
-  int32_t (*FpOnClientRequest)(SSyncNode* ths, SyncClientRequest* pMsg);
-  int32_t (*FpOnRequestVote)(SSyncNode* ths, SyncRequestVote* pMsg);
-  int32_t (*FpOnRequestVoteReply)(SSyncNode* ths, SyncRequestVoteReply* pMsg);
-  int32_t (*FpOnAppendEntries)(SSyncNode* ths, SyncAppendEntries* pMsg);
-  int32_t (*FpOnAppendEntriesReply)(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
-  int32_t (*FpOnTimeout)(SSyncNode* pSyncNode, SyncTimeout* pMsg);
+  FpOnPingCb               FpOnPing;
+  FpOnPingReplyCb          FpOnPingReply;
+  FpOnClientRequestCb      FpOnClientRequest;
+  FpOnTimeoutCb            FpOnTimeout;
+  FpOnRequestVoteCb        FpOnRequestVote;
+  FpOnRequestVoteReplyCb   FpOnRequestVoteReply;
+  FpOnAppendEntriesCb      FpOnAppendEntries;
+  FpOnAppendEntriesReplyCb FpOnAppendEntriesReply;
+  FpOnSnapshotSendCb       FpOnSnapshotSend;
+  FpOnSnapshotRspCb        FpOnSnapshotRsp;
 
   // tools
   SSyncRespMgr* pSyncRespMgr;
 
   // restore state
-  // sem_t      restoreSem;
-  bool                   restoreFinish;
-  SSnapshot*             pSnapshot;
-  SSyncSnapshotSender*   pSender;
-  SSyncSnapshotReceiver* pReceiver;
+  bool restoreFinish;
+  // SSnapshot*             pSnapshot;
+  SSyncSnapshotSender*   senders[TSDB_MAX_REPLICA];
+  SSyncSnapshotReceiver* pNewNodeReceiver;
+
+  SSnapshotMeta sMeta;
 
 } SSyncNode;
 
@@ -164,6 +169,9 @@ void       syncNodeStart(SSyncNode* pSyncNode);
 void       syncNodeStartStandBy(SSyncNode* pSyncNode);
 void       syncNodeClose(SSyncNode* pSyncNode);
 
+// option
+bool syncNodeSnapshotEnable(SSyncNode* pSyncNode);
+
 // ping --------------
 int32_t syncNodePing(SSyncNode* pSyncNode, const SRaftId* destRaftId, SyncPing* pMsg);
 int32_t syncNodePingSelf(SSyncNode* pSyncNode);
@@ -205,6 +213,25 @@ void syncNodeCandidate2Follower(SSyncNode* pSyncNode);
 void syncNodeVoteForTerm(SSyncNode* pSyncNode, SyncTerm term, SRaftId* pRaftId);
 void syncNodeVoteForSelf(SSyncNode* pSyncNode);
 
+// snapshot --------------
+bool syncNodeHasSnapshot(SSyncNode* pSyncNode);
+bool syncNodeIsIndexInSnapshot(SSyncNode* pSyncNode, SyncIndex index);
+
+SyncIndex syncNodeGetLastIndex(SSyncNode* pSyncNode);
+SyncTerm  syncNodeGetLastTerm(SSyncNode* pSyncNode);
+int32_t   syncNodeGetLastIndexTerm(SSyncNode* pSyncNode, SyncIndex* pLastIndex, SyncTerm* pLastTerm);
+
+SyncIndex syncNodeSyncStartIndex(SSyncNode* pSyncNode);
+
+SyncIndex syncNodeGetPreIndex(SSyncNode* pSyncNode, SyncIndex index);
+SyncTerm  syncNodeGetPreTerm(SSyncNode* pSyncNode, SyncIndex index);
+int32_t   syncNodeGetPreIndexTerm(SSyncNode* pSyncNode, SyncIndex index, SyncIndex* pPreIndex, SyncTerm* pPreTerm);
+
+int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, uint64_t flag);
+
+bool                 syncNodeInRaftGroup(SSyncNode* ths, SRaftId* pRaftId);
+SSyncSnapshotSender* syncNodeGetSnapshotSender(SSyncNode* ths, SRaftId* pDestId);
+
 // for debug --------------
 void syncNodePrint(SSyncNode* pObj);
 void syncNodePrint2(char* s, SSyncNode* pObj);

source/libs/sync/inc/syncRaftCfg.h

@@ -34,6 +34,7 @@ typedef struct SRaftCfg {
   TdFilePtr pFile;
   char      path[TSDB_FILENAME_LEN * 2];
   int8_t    isStandBy;
+  int8_t    snapshotEnable;
 } SRaftCfg;
 
 SRaftCfg *raftCfgOpen(const char *path);
@@ -50,7 +51,12 @@ char *  raftCfg2Str(SRaftCfg *pRaftCfg);
 int32_t raftCfgFromJson(const cJSON *pRoot, SRaftCfg *pRaftCfg);
 int32_t raftCfgFromStr(const char *s, SRaftCfg *pRaftCfg);
 
-int32_t raftCfgCreateFile(SSyncCfg *pCfg, int8_t isStandBy, const char *path);
+typedef struct SRaftCfgMeta {
+  int8_t isStandBy;
+  int8_t snapshotEnable;
+} SRaftCfgMeta;
+
+int32_t raftCfgCreateFile(SSyncCfg *pCfg, SRaftCfgMeta meta, const char *path);
 
 // for debug ----------------------
 void syncCfgPrint(SSyncCfg *pCfg);

source/libs/sync/inc/syncRaftLog.h

@@ -30,6 +30,7 @@ extern "C" {
 typedef struct SSyncLogStoreData {
   SSyncNode* pSyncNode;
   SWal*      pWal;
+  SyncIndex  beginIndex;  // valid begin index, default 0, may be set beginIndex > 0
 } SSyncLogStoreData;
 
 SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode);
@@ -39,14 +40,7 @@ char*          logStore2Str(SSyncLogStore* pLogStore);
 cJSON*         logStoreSimple2Json(SSyncLogStore* pLogStore);
 char*          logStoreSimple2Str(SSyncLogStore* pLogStore);
 
-// SSyncRaftEntry* logStoreGetLastEntry(SSyncLogStore* pLogStore);
-// SyncIndex       logStoreLastIndex(SSyncLogStore* pLogStore);
-// SyncTerm        logStoreLastTerm(SSyncLogStore* pLogStore);
-// SSyncRaftEntry* logStoreGetEntry(SSyncLogStore* pLogStore, SyncIndex index);
-// int32_t         logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry);
-// int32_t         logStoreTruncate(SSyncLogStore* pLogStore, SyncIndex fromIndex);
-// int32_t         logStoreUpdateCommitIndex(SSyncLogStore* pLogStore, SyncIndex index);
-// SyncIndex       logStoreGetCommitIndex(SSyncLogStore* pLogStore);
+SyncIndex logStoreFirstIndex(SSyncLogStore* pLogStore);
 
 // for debug
 void logStorePrint(SSyncLogStore* pLogStore);

source/libs/sync/inc/syncRaftStore.h

@@ -49,8 +49,8 @@ void    raftStoreClearVote(SRaftStore *pRaftStore);
 void    raftStoreNextTerm(SRaftStore *pRaftStore);
 void    raftStoreSetTerm(SRaftStore *pRaftStore, SyncTerm term);
 int32_t raftStoreFromJson(SRaftStore *pRaftStore, cJSON *pJson);
-cJSON * raftStore2Json(SRaftStore *pRaftStore);
-char *  raftStore2Str(SRaftStore *pRaftStore);
+cJSON  *raftStore2Json(SRaftStore *pRaftStore);
+char   *raftStore2Str(SRaftStore *pRaftStore);
 
 // for debug -------------------
 void raftStorePrint(SRaftStore *pObj);

source/libs/sync/inc/syncReplication.h

@@ -52,6 +52,7 @@ extern "C" {
 //    /\ UNCHANGED <<serverVars, candidateVars, leaderVars, logVars>>
 //
 int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode);
+int32_t syncNodeAppendEntriesPeersSnapshot(SSyncNode* pSyncNode);
 int32_t syncNodeReplicate(SSyncNode* pSyncNode);
 int32_t syncNodeAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncAppendEntries* pMsg);
 

source/libs/sync/inc/syncRequestVote.h

@@ -50,6 +50,7 @@ extern "C" {
 //       /\ UNCHANGED <<state, currentTerm, candidateVars, leaderVars, logVars>>
 //
 int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg);
+int32_t syncNodeOnRequestVoteSnapshotCb(SSyncNode* ths, SyncRequestVote* pMsg);
 
 #ifdef __cplusplus
 }

source/libs/sync/inc/syncRequestVoteReply.h

@@ -45,6 +45,7 @@ extern "C" {
 //    /\ UNCHANGED <<serverVars, votedFor, leaderVars, logVars>>
 //
 int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg);
+int32_t syncNodeOnRequestVoteReplySnapshotCb(SSyncNode* ths, SyncRequestVoteReply* pMsg);
 
 #ifdef __cplusplus
 }

source/libs/sync/inc/syncSnapshot.h

@@ -25,40 +25,64 @@ extern "C" {
 #include <stdlib.h>
 #include "cJSON.h"
 #include "syncInt.h"
+#include "syncMessage.h"
 #include "taosdef.h"
 
+#define SYNC_SNAPSHOT_SEQ_INVALID -1
+#define SYNC_SNAPSHOT_SEQ_FORCE_CLOSE -2
+#define SYNC_SNAPSHOT_SEQ_BEGIN 0
+#define SYNC_SNAPSHOT_SEQ_END 0x7FFFFFFF
+
+#define SYNC_SNAPSHOT_RETRY_MS 5000
+
 typedef struct SSyncSnapshotSender {
-  int32_t    sending;
-  int32_t    received;
-  bool       finish;
-  void *     pCurrentBlock;
+  bool       start;
+  int32_t    seq;
+  int32_t    ack;
+  void      *pReader;
+  void      *pCurrentBlock;
   int32_t    blockLen;
+  SSnapshot  snapshot;
   int64_t    sendingMS;
-  SSnapshot *pSnapshot;
   SSyncNode *pSyncNode;
+  int32_t    replicaIndex;
+  SyncTerm   term;
+  SyncTerm   privateTerm;
+  bool       finish;
 } SSyncSnapshotSender;
 
-SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode);
+SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaIndex);
 void                 snapshotSenderDestroy(SSyncSnapshotSender *pSender);
+bool                 snapshotSenderIsStart(SSyncSnapshotSender *pSender);
+void                 snapshotSenderStart(SSyncSnapshotSender *pSender);
+void                 snapshotSenderStop(SSyncSnapshotSender *pSender);
 int32_t              snapshotSend(SSyncSnapshotSender *pSender);
-cJSON *              snapshotSender2Json(SSyncSnapshotSender *pSender);
-char *               snapshotSender2Str(SSyncSnapshotSender *pSender);
+int32_t              snapshotReSend(SSyncSnapshotSender *pSender);
+cJSON               *snapshotSender2Json(SSyncSnapshotSender *pSender);
+char                *snapshotSender2Str(SSyncSnapshotSender *pSender);
 
 typedef struct SSyncSnapshotReceiver {
-  bool       start;
-  int32_t    received;
-  int32_t    progressIndex;
-  void *     pCurrentBlock;
-  int32_t    len;
-  SSnapshot *pSnapshot;
+  bool start;
+
+  int32_t  ack;
+  void    *pWriter;
+  SyncTerm term;
+  SyncTerm privateTerm;
+
   SSyncNode *pSyncNode;
+  int32_t    replicaIndex;
 } SSyncSnapshotReceiver;
 
-SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode);
+SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, int32_t replicaIndex);
 void                   snapshotReceiverDestroy(SSyncSnapshotReceiver *pReceiver);
-int32_t                snapshotReceive(SSyncSnapshotReceiver *pReceiver);
-cJSON *                snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver);
-char *                 snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver);
+void                   snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm);
+bool                   snapshotReceiverIsStart(SSyncSnapshotReceiver *pReceiver);
+void                   snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply);
+cJSON                 *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver);
+char                  *snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver);
+
+int32_t syncNodeOnSnapshotSendCb(SSyncNode *ths, SyncSnapshotSend *pMsg);
+int32_t syncNodeOnSnapshotRspCb(SSyncNode *ths, SyncSnapshotRsp *pMsg);
 
 #ifdef __cplusplus
 }

source/libs/sync/inc/syncUtil.h

@@ -61,6 +61,7 @@ bool        syncUtilIsData(tmsg_t msgType);
 bool        syncUtilUserPreCommit(tmsg_t msgType);
 bool        syncUtilUserCommit(tmsg_t msgType);
 bool        syncUtilUserRollback(tmsg_t msgType);
+void        syncUtilJson2Line(char* jsonStr);
 
 #ifdef __cplusplus
 }

source/libs/sync/src/syncAppendEntries.c

@@ -18,8 +18,10 @@
 #include "syncRaftCfg.h"
 #include "syncRaftLog.h"
 #include "syncRaftStore.h"
+#include "syncSnapshot.h"
 #include "syncUtil.h"
 #include "syncVoteMgr.h"
+#include "wal.h"
 
 // TLA+ Spec
 // HandleAppendEntriesRequest(i, j, m) ==
@@ -335,8 +337,12 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
                 cbMeta.currentTerm = ths->pRaftStore->currentTerm;
                 cbMeta.flag = 0x11;
 
+                SSnapshot snapshot;
+                ASSERT(ths->pFsm->FpGetSnapshot != NULL);
+                ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
+
                 bool needExecute = true;
-                if (ths->pSnapshot != NULL && cbMeta.index <= ths->pSnapshot->lastApplyIndex) {
+                if (cbMeta.index <= snapshot.lastApplyIndex) {
                   needExecute = false;
                 }
 
@@ -427,3 +433,332 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
 
   return ret;
 }
+
+static int32_t syncNodeMakeLogSame(SSyncNode* ths, SyncAppendEntries* pMsg) {
+  int32_t code;
+
+  SyncIndex delBegin = pMsg->prevLogIndex + 1;
+  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;
+        cbMeta.index = pRollBackEntry->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);
+  sInfo("sync event log truncate, from %ld to %ld", delBegin, delEnd);
+  logStoreSimpleLog2("after syncNodeMakeLogSame", ths->pLogStore);
+
+  return code;
+}
+
+static int32_t syncNodePreCommit(SSyncNode* ths, SSyncRaftEntry* pEntry) {
+  SRpcMsg rpcMsg;
+  syncEntry2OriginalRpc(pEntry, &rpcMsg);
+  if (ths->pFsm != NULL) {
+    if (ths->pFsm->FpPreCommitCb != NULL && syncUtilUserPreCommit(pEntry->originalRpcType)) {
+      SFsmCbMeta cbMeta;
+      cbMeta.index = pEntry->index;
+      cbMeta.isWeak = pEntry->isWeak;
+      cbMeta.code = 2;
+      cbMeta.state = ths->state;
+      cbMeta.seqNum = pEntry->seqNum;
+      ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, cbMeta);
+    }
+  }
+  rpcFreeCont(rpcMsg.pCont);
+  return 0;
+}
+
+// really pre log match
+// prevLogIndex == -1
+static bool syncNodeOnAppendEntriesLogOK(SSyncNode* pSyncNode, SyncAppendEntries* pMsg) {
+  if (pMsg->prevLogIndex == SYNC_INDEX_INVALID) {
+    if (gRaftDetailLog) {
+      sTrace("syncNodeOnAppendEntriesLogOK true, pMsg->prevLogIndex:%ld", pMsg->prevLogIndex);
+    }
+    return true;
+  }
+
+  SyncIndex myLastIndex = syncNodeGetLastIndex(pSyncNode);
+  if (pMsg->prevLogIndex > myLastIndex) {
+    if (gRaftDetailLog) {
+      sTrace("syncNodeOnAppendEntriesLogOK false, pMsg->prevLogIndex:%ld, myLastIndex:%ld", pMsg->prevLogIndex,
+             myLastIndex);
+    }
+    return false;
+  }
+
+  SyncTerm myPreLogTerm = syncNodeGetPreTerm(pSyncNode, pMsg->prevLogIndex + 1);
+  if (pMsg->prevLogIndex <= myLastIndex && pMsg->prevLogTerm == myPreLogTerm) {
+    if (gRaftDetailLog) {
+      sTrace(
+          "syncNodeOnAppendEntriesLogOK true, pMsg->prevLogIndex:%ld, myLastIndex:%ld, pMsg->prevLogTerm:%lu, "
+          "myPreLogTerm:%lu",
+          pMsg->prevLogIndex, myLastIndex, pMsg->prevLogTerm, myPreLogTerm);
+    }
+    return true;
+  }
+
+  if (gRaftDetailLog) {
+    sTrace(
+        "syncNodeOnAppendEntriesLogOK false, pMsg->prevLogIndex:%ld, myLastIndex:%ld, pMsg->prevLogTerm:%lu, "
+        "myPreLogTerm:%lu",
+        pMsg->prevLogIndex, myLastIndex, pMsg->prevLogTerm, myPreLogTerm);
+  }
+
+  return false;
+}
+
+int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
+  int32_t ret = 0;
+  int32_t code = 0;
+
+  // print log
+  char logBuf[128] = {0};
+  snprintf(logBuf, sizeof(logBuf), "recv SyncAppendEntries, vgId:%d, term:%lu", ths->vgId,
+           ths->pRaftStore->currentTerm);
+  syncAppendEntriesLog2(logBuf, pMsg);
+
+  // if already drop replica, do not process
+  if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
+    sInfo("recv SyncAppendEntries maybe replica already dropped");
+    return ret;
+  }
+
+  // maybe update term
+  if (pMsg->term > ths->pRaftStore->currentTerm) {
+    syncNodeUpdateTerm(ths, pMsg->term);
+  }
+  ASSERT(pMsg->term <= ths->pRaftStore->currentTerm);
+
+  // reset elect timer
+  if (pMsg->term == ths->pRaftStore->currentTerm) {
+    ths->leaderCache = pMsg->srcId;
+    syncNodeResetElectTimer(ths);
+  }
+  ASSERT(pMsg->dataLen >= 0);
+
+  // candidate to follower
+  //
+  // operation:
+  // to follower
+  do {
+    bool condition = pMsg->term == ths->pRaftStore->currentTerm && ths->state == TAOS_SYNC_STATE_CANDIDATE;
+    if (condition) {
+      sTrace("recv SyncAppendEntries, candidate to follower");
+
+      syncNodeBecomeFollower(ths);
+      // do not reply?
+      return ret;
+    }
+  } while (0);
+
+  // fake match
+  //
+  // condition1:
+  // I have snapshot, no log, preIndex > myLastIndex
+  //
+  // condition2:
+  // I have snapshot, have log, log <= snapshot, preIndex > myLastIndex
+  //
+  // condition3:
+  // I have snapshot, preIndex < snapshot.lastApplyIndex
+  //
+  // condition4:
+  // I have snapshot, preIndex == snapshot.lastApplyIndex, no data
+  //
+  // operation:
+  // match snapshot.lastApplyIndex - 1;
+  // no operation on log
+  do {
+    SyncIndex myLastIndex = syncNodeGetLastIndex(ths);
+    SSnapshot snapshot;
+    ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
+
+    bool condition0 = (pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) &&
+                      syncNodeHasSnapshot(ths);
+    bool condition1 =
+        condition0 && (ths->pLogStore->syncLogEntryCount(ths->pLogStore) == 0) && (pMsg->prevLogIndex > myLastIndex);
+    bool condition2 = condition0 && (ths->pLogStore->syncLogLastIndex(ths->pLogStore) <= snapshot.lastApplyIndex) &&
+                      (pMsg->prevLogIndex > myLastIndex);
+    bool condition3 = condition0 && (pMsg->prevLogIndex < snapshot.lastApplyIndex);
+    bool condition4 = condition0 && (pMsg->prevLogIndex == snapshot.lastApplyIndex) && (pMsg->dataLen == 0);
+    bool condition = condition1 || condition2 || condition3 || condition4;
+
+    if (condition) {
+      sTrace(
+          "recv SyncAppendEntries, fake match, myLastIndex:%ld, syncLogBeginIndex:%ld, syncLogEndIndex:%ld, "
+          "condition1:%d, condition2:%d, condition3:%d, condition4:%d",
+          myLastIndex, ths->pLogStore->syncLogBeginIndex(ths->pLogStore),
+          ths->pLogStore->syncLogEndIndex(ths->pLogStore), condition1, condition2, condition3, condition4);
+
+      // prepare response msg
+      SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild(ths->vgId);
+      pReply->srcId = ths->myRaftId;
+      pReply->destId = pMsg->srcId;
+      pReply->term = ths->pRaftStore->currentTerm;
+      pReply->privateTerm = ths->pNewNodeReceiver->privateTerm;
+      pReply->success = true;
+      pReply->matchIndex = snapshot.lastApplyIndex;
+
+      // send response
+      SRpcMsg rpcMsg;
+      syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
+      syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
+      syncAppendEntriesReplyDestroy(pReply);
+
+      return ret;
+    }
+  } while (0);
+
+  // calculate logOK here, before will coredump, due to fake match
+  bool logOK = syncNodeOnAppendEntriesLogOK(ths, pMsg);
+
+  // not match
+  //
+  // condition1:
+  // term < myTerm
+  //
+  // condition2:
+  // !logOK
+  //
+  // operation:
+  // not match
+  // no operation on log
+  do {
+    bool condition1 = pMsg->term < ths->pRaftStore->currentTerm;
+    bool condition2 =
+        (pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) && !logOK;
+    bool condition = condition1 || condition2;
+
+    if (condition) {
+      sTrace(
+          "recv SyncAppendEntries, not match, syncLogBeginIndex:%ld, syncLogEndIndex:%ld, condition1:%d, "
+          "condition2:%d, logOK:%d",
+          ths->pLogStore->syncLogBeginIndex(ths->pLogStore), ths->pLogStore->syncLogEndIndex(ths->pLogStore),
+          condition1, condition2, logOK);
+
+      // prepare response msg
+      SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild(ths->vgId);
+      pReply->srcId = ths->myRaftId;
+      pReply->destId = pMsg->srcId;
+      pReply->term = ths->pRaftStore->currentTerm;
+      pReply->privateTerm = ths->pNewNodeReceiver->privateTerm;
+      pReply->success = false;
+      pReply->matchIndex = SYNC_INDEX_INVALID;
+
+      // send response
+      SRpcMsg rpcMsg;
+      syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
+      syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
+      syncAppendEntriesReplyDestroy(pReply);
+
+      return ret;
+    }
+  } while (0);
+
+  // really match
+  //
+  // condition:
+  // logOK
+  //
+  // operation:
+  // match
+  // make log same
+  do {
+    bool condition = (pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) && logOK;
+    if (condition) {
+      // has extra entries (> preIndex) in local log
+      SyncIndex myLastIndex = syncNodeGetLastIndex(ths);
+      bool      hasExtraEntries = myLastIndex > pMsg->prevLogIndex;
+
+      // has entries in SyncAppendEntries msg
+      bool hasAppendEntries = pMsg->dataLen > 0;
+
+      sTrace("recv SyncAppendEntries, match, myLastIndex:%ld, hasExtraEntries:%d, hasAppendEntries:%d", myLastIndex,
+             hasExtraEntries, hasAppendEntries);
+
+      if (hasExtraEntries) {
+        // make log same, rollback deleted entries
+        code = syncNodeMakeLogSame(ths, pMsg);
+        ASSERT(code == 0);
+      }
+
+      if (hasAppendEntries) {
+        // append entry
+        SSyncRaftEntry* pAppendEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
+        ASSERT(pAppendEntry != NULL);
+
+        code = ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pAppendEntry);
+        ASSERT(code == 0);
+
+        // pre commit
+        code = syncNodePreCommit(ths, pAppendEntry);
+        ASSERT(code == 0);
+
+        syncEntryDestory(pAppendEntry);
+      }
+
+      // prepare response msg
+      SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild(ths->vgId);
+      pReply->srcId = ths->myRaftId;
+      pReply->destId = pMsg->srcId;
+      pReply->term = ths->pRaftStore->currentTerm;
+      pReply->privateTerm = ths->pNewNodeReceiver->privateTerm;
+      pReply->success = true;
+      pReply->matchIndex = hasAppendEntries ? pMsg->prevLogIndex + 1 : pMsg->prevLogIndex;
+
+      // send response
+      SRpcMsg rpcMsg;
+      syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
+      syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
+      syncAppendEntriesReplyDestroy(pReply);
+
+      // maybe update commit index, leader notice me
+      if (pMsg->commitIndex > ths->commitIndex) {
+        // has commit entry in local
+        if (pMsg->commitIndex <= ths->pLogStore->syncLogLastIndex(ths->pLogStore)) {
+          SyncIndex beginIndex = ths->commitIndex + 1;
+          SyncIndex endIndex = pMsg->commitIndex;
+
+          // update commit index
+          ths->commitIndex = pMsg->commitIndex;
+
+          // call back Wal
+          code = ths->pLogStore->updateCommitIndex(ths->pLogStore, ths->commitIndex);
+          ASSERT(code == 0);
+
+          code = syncNodeCommit(ths, beginIndex, endIndex, ths->state);
+          ASSERT(code == 0);
+        }
+      }
+      return ret;
+    }
+  } while (0);
+
+  return ret;
+}

source/libs/sync/src/syncAppendEntriesReply.c

@@ -17,8 +17,10 @@
 #include "syncCommit.h"
 #include "syncIndexMgr.h"
 #include "syncInt.h"
+#include "syncRaftCfg.h"
 #include "syncRaftLog.h"
 #include "syncRaftStore.h"
+#include "syncSnapshot.h"
 #include "syncUtil.h"
 #include "syncVoteMgr.h"
 
@@ -94,3 +96,116 @@ int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* p
 
   return ret;
 }
+
+int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg) {
+  int32_t ret = 0;
+
+  // print log
+  char logBuf[128] = {0};
+  snprintf(logBuf, sizeof(logBuf), "recv SyncAppendEntriesReply, vgId:%d, term:%lu", ths->vgId,
+           ths->pRaftStore->currentTerm);
+  syncAppendEntriesReplyLog2(logBuf, pMsg);
+
+  // if already drop replica, do not process
+  if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
+    sInfo("recv SyncAppendEntriesReply,  maybe replica already dropped");
+    return ret;
+  }
+
+  // drop stale response
+  if (pMsg->term < ths->pRaftStore->currentTerm) {
+    sTrace("recv SyncAppendEntriesReply, drop stale response, receive_term:%lu current_term:%lu", pMsg->term,
+           ths->pRaftStore->currentTerm);
+    return ret;
+  }
+
+  syncIndexMgrLog2("recv SyncAppendEntriesReply, before pNextIndex:", ths->pNextIndex);
+  syncIndexMgrLog2("recv SyncAppendEntriesReply, before pMatchIndex:", ths->pMatchIndex);
+  {
+    SSnapshot snapshot;
+    ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
+    sTrace("recv SyncAppendEntriesReply, before snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
+           snapshot.lastApplyIndex, snapshot.lastApplyTerm);
+  }
+
+  // no need this code, because if I receive reply.term, then I must have sent for that term.
+  //  if (pMsg->term > ths->pRaftStore->currentTerm) {
+  //    syncNodeUpdateTerm(ths, pMsg->term);
+  //  }
+
+  if (pMsg->term > ths->pRaftStore->currentTerm) {
+    char logBuf[128] = {0};
+    snprintf(logBuf, sizeof(logBuf), "recv SyncAppendEntriesReply, error term, receive_term:%lu current_term:%lu",
+             pMsg->term, ths->pRaftStore->currentTerm);
+    syncNodeLog2(logBuf, ths);
+    sError("%s", logBuf);
+    return ret;
+  }
+
+  ASSERT(pMsg->term == ths->pRaftStore->currentTerm);
+
+  if (pMsg->success) {
+    // nextIndex'  = [nextIndex  EXCEPT ![i][j] = m.mmatchIndex + 1]
+    syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), pMsg->matchIndex + 1);
+    sTrace("update next match, index:%ld, success:%d", pMsg->matchIndex + 1, pMsg->success);
+
+    // matchIndex' = [matchIndex EXCEPT ![i][j] = m.mmatchIndex]
+    syncIndexMgrSetIndex(ths->pMatchIndex, &(pMsg->srcId), pMsg->matchIndex);
+
+    // maybe commit
+    if (ths->state == TAOS_SYNC_STATE_LEADER) {
+      syncMaybeAdvanceCommitIndex(ths);
+    }
+
+  } else {
+    SyncIndex nextIndex = syncIndexMgrGetIndex(ths->pNextIndex, &(pMsg->srcId));
+    sTrace("update next not match, begin, index:%ld, success:%d", nextIndex, pMsg->success);
+
+    // notice! int64, uint64
+    if (nextIndex > SYNC_INDEX_BEGIN) {
+      --nextIndex;
+
+      // get sender
+      SSyncSnapshotSender* pSender = syncNodeGetSnapshotSender(ths, &(pMsg->srcId));
+      ASSERT(pSender != NULL);
+      bool      hasSnapshot = syncNodeHasSnapshot(ths);
+      SSnapshot snapshot;
+      ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
+
+      // start sending snapshot first time
+      // start here, stop by receiver
+      if (hasSnapshot && nextIndex <= snapshot.lastApplyIndex + 1 && !snapshotSenderIsStart(pSender) &&
+          pMsg->privateTerm < pSender->privateTerm) {
+        snapshotSenderStart(pSender);
+
+        char* s = snapshotSender2Str(pSender);
+        sInfo("sync event snapshot send start sender first time, sender:%s", s);
+        taosMemoryFree(s);
+      }
+
+      SyncIndex sentryIndex = pSender->snapshot.lastApplyIndex + 1;
+
+      // update nextIndex to sentryIndex
+      if (nextIndex <= sentryIndex) {
+        nextIndex = sentryIndex;
+      }
+
+    } else {
+      nextIndex = SYNC_INDEX_BEGIN;
+    }
+
+    syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), nextIndex);
+    sTrace("update next not match, end, index:%ld, success:%d", nextIndex, pMsg->success);
+  }
+
+  syncIndexMgrLog2("recv SyncAppendEntriesReply, after pNextIndex:", ths->pNextIndex);
+  syncIndexMgrLog2("recv SyncAppendEntriesReply, after pMatchIndex:", ths->pMatchIndex);
+  {
+    SSnapshot snapshot;
+    ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
+    sTrace("recv SyncAppendEntriesReply, after snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
+           snapshot.lastApplyIndex, snapshot.lastApplyTerm);
+  }
+
+  return ret;
+}
\ No newline at end of file

source/libs/sync/src/syncCommit.c

@@ -94,109 +94,8 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
 
     // execute fsm
     if (pSyncNode->pFsm != NULL) {
-      for (SyncIndex i = beginIndex; i <= endIndex; ++i) {
-        if (i != SYNC_INDEX_INVALID) {
-          SSyncRaftEntry* pEntry = pSyncNode->pLogStore->getEntry(pSyncNode->pLogStore, i);
-          assert(pEntry != NULL);
-
-          SRpcMsg rpcMsg;
-          syncEntry2OriginalRpc(pEntry, &rpcMsg);
-
-          if (pSyncNode->pFsm->FpCommitCb != NULL && syncUtilUserCommit(pEntry->originalRpcType)) {
-            SFsmCbMeta cbMeta;
-            cbMeta.index = pEntry->index;
-            cbMeta.isWeak = pEntry->isWeak;
-            cbMeta.code = 0;
-            cbMeta.state = pSyncNode->state;
-            cbMeta.seqNum = pEntry->seqNum;
-            cbMeta.term = pEntry->term;
-            cbMeta.currentTerm = pSyncNode->pRaftStore->currentTerm;
-            cbMeta.flag = 0x1;
-
-            bool needExecute = true;
-            if (pSyncNode->pSnapshot != NULL && cbMeta.index <= pSyncNode->pSnapshot->lastApplyIndex) {
-              needExecute = false;
-            }
-
-            if (needExecute) {
-              pSyncNode->pFsm->FpCommitCb(pSyncNode->pFsm, &rpcMsg, cbMeta);
-            }
-          }
-
-          // config change
-          if (pEntry->originalRpcType == TDMT_SYNC_CONFIG_CHANGE) {
-            SSyncCfg oldSyncCfg = pSyncNode->pRaftCfg->cfg;
-
-            SSyncCfg newSyncCfg;
-            int32_t  ret = syncCfgFromStr(rpcMsg.pCont, &newSyncCfg);
-            ASSERT(ret == 0);
-
-            // update new config myIndex
-            bool hit = false;
-            for (int i = 0; i < newSyncCfg.replicaNum; ++i) {
-              if (strcmp(pSyncNode->myNodeInfo.nodeFqdn, (newSyncCfg.nodeInfo)[i].nodeFqdn) == 0 &&
-                  pSyncNode->myNodeInfo.nodePort == (newSyncCfg.nodeInfo)[i].nodePort) {
-                newSyncCfg.myIndex = i;
-                hit = true;
-                break;
-              }
-            }
-
-            if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
-              ASSERT(hit == true);
-            }
-
-            bool isDrop;
-            syncNodeUpdateConfig(pSyncNode, &newSyncCfg, &isDrop);
-
-            // change isStandBy to normal
-            if (!isDrop) {
-              if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
-                syncNodeBecomeLeader(pSyncNode);
-              } else {
-                syncNodeBecomeFollower(pSyncNode);
-              }
-            }
-
-            char* sOld = syncCfg2Str(&oldSyncCfg);
-            char* sNew = syncCfg2Str(&newSyncCfg);
-            sInfo("==config change== 0x1 old:%s new:%s isDrop:%d \n", sOld, sNew, isDrop);
-            taosMemoryFree(sOld);
-            taosMemoryFree(sNew);
-
-            if (pSyncNode->pFsm->FpReConfigCb != NULL) {
-              SReConfigCbMeta cbMeta = {0};
-              cbMeta.code = 0;
-              cbMeta.currentTerm = pSyncNode->pRaftStore->currentTerm;
-              cbMeta.index = pEntry->index;
-              cbMeta.term = pEntry->term;
-              cbMeta.oldCfg = oldSyncCfg;
-              cbMeta.flag = 0x1;
-              cbMeta.isDrop = isDrop;
-              pSyncNode->pFsm->FpReConfigCb(pSyncNode->pFsm, newSyncCfg, cbMeta);
-            }
-          }
-
-          // restore finish
-          if (pEntry->index == pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore)) {
-            if (pSyncNode->restoreFinish == false) {
-              if (pSyncNode->pFsm->FpRestoreFinishCb != NULL) {
-                pSyncNode->pFsm->FpRestoreFinishCb(pSyncNode->pFsm);
-              }
-              pSyncNode->restoreFinish = true;
-              sInfo("==syncMaybeAdvanceCommitIndex== restoreFinish set true %p vgId:%d", pSyncNode, pSyncNode->vgId);
-
-              /*
-              tsem_post(&pSyncNode->restoreSem);
-              sInfo("==syncMaybeAdvanceCommitIndex== RestoreFinish tsem_post %p", pSyncNode);
-              */
-            }
-          }
-
-          rpcFreeCont(rpcMsg.pCont);
-          syncEntryDestory(pEntry);
-        }
-      }
+      int32_t code = syncNodeCommit(pSyncNode, beginIndex, endIndex, pSyncNode->state);
+      ASSERT(code == 0);
     }
   }
 }

source/libs/sync/src/syncElection.c

@@ -15,6 +15,7 @@
 
 #include "syncElection.h"
 #include "syncMessage.h"
+#include "syncRaftCfg.h"
 #include "syncRaftStore.h"
 #include "syncVoteMgr.h"
 
@@ -49,6 +50,26 @@ int32_t syncNodeRequestVotePeers(SSyncNode* pSyncNode) {
   return ret;
 }
 
+int32_t syncNodeRequestVotePeersSnapshot(SSyncNode* pSyncNode) {
+  ASSERT(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
+
+  int32_t ret = 0;
+  for (int i = 0; i < pSyncNode->peersNum; ++i) {
+    SyncRequestVote* pMsg = syncRequestVoteBuild(pSyncNode->vgId);
+    pMsg->srcId = pSyncNode->myRaftId;
+    pMsg->destId = pSyncNode->peersId[i];
+    pMsg->term = pSyncNode->pRaftStore->currentTerm;
+
+    ret = syncNodeGetLastIndexTerm(pSyncNode, &(pMsg->lastLogIndex), &(pMsg->lastLogTerm));
+    ASSERT(ret == 0);
+
+    ret = syncNodeRequestVote(pSyncNode, &pSyncNode->peersId[i], pMsg);
+    ASSERT(ret == 0);
+    syncRequestVoteDestroy(pMsg);
+  }
+  return ret;
+}
+
 int32_t syncNodeElect(SSyncNode* pSyncNode) {
   int32_t ret = 0;
   if (pSyncNode->state == TAOS_SYNC_STATE_FOLLOWER) {
@@ -71,7 +92,12 @@ int32_t syncNodeElect(SSyncNode* pSyncNode) {
     return ret;
   }
 
-  ret = syncNodeRequestVotePeers(pSyncNode);
+  if (pSyncNode->pRaftCfg->snapshotEnable) {
+    ret = syncNodeRequestVotePeersSnapshot(pSyncNode);
+  } else {
+    ret = syncNodeRequestVotePeers(pSyncNode);
+  }
+
   assert(ret == 0);
   syncNodeResetElectTimer(pSyncNode);
 

source/libs/sync/src/syncIO.c

@@ -75,7 +75,8 @@ int32_t syncIOSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) {
     syncUtilMsgNtoH(pMsg->pCont);
 
     char logBuf[256] = {0};
-    snprintf(logBuf, sizeof(logBuf), "==syncIOSendMsg== %s:%d", pEpSet->eps[0].fqdn, pEpSet->eps[0].port);
+    snprintf(logBuf, sizeof(logBuf), "==syncIOSendMsg== %s:%d msgType:%d", pEpSet->eps[0].fqdn, pEpSet->eps[0].port,
+             pMsg->msgType);
     syncRpcMsgLog2(logBuf, pMsg);
 
     syncUtilMsgHtoN(pMsg->pCont);
@@ -89,8 +90,10 @@ int32_t syncIOSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) {
 
 int32_t syncIOEqMsg(const SMsgCb *msgcb, SRpcMsg *pMsg) {
   int32_t ret = 0;
-  char    logBuf[128] = {0};
-  syncRpcMsgLog2((char *)"==syncIOEqMsg==", pMsg);
+
+  char logBuf[256] = {0};
+  snprintf(logBuf, sizeof(logBuf), "==syncIOEqMsg== msgType:%d", pMsg->msgType);
+  syncRpcMsgLog2(logBuf, pMsg);
 
   SRpcMsg *pTemp;
   pTemp = taosAllocateQitem(sizeof(SRpcMsg), DEF_QITEM);
@@ -253,7 +256,9 @@ static void *syncIOConsumerFunc(void *param) {
 
     for (int i = 0; i < numOfMsgs; ++i) {
       taosGetQitem(qall, (void **)&pRpcMsg);
-      syncRpcMsgLog2((char *)"==syncIOConsumerFunc==", pRpcMsg);
+      char logBuf[128];
+      snprintf(logBuf, sizeof(logBuf), "==syncIOConsumMsg== msgType:%d", pRpcMsg->msgType);
+      syncRpcMsgLog2(logBuf, pRpcMsg);
 
       // use switch case instead of if else
       if (pRpcMsg->msgType == TDMT_SYNC_PING) {
@@ -319,6 +324,23 @@ static void *syncIOConsumerFunc(void *param) {
           io->FpOnSyncTimeout(io->pSyncNode, pSyncMsg);
           syncTimeoutDestroy(pSyncMsg);
         }
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_SEND) {
+        if (io->FpOnSyncSnapshotSend != NULL) {
+          SyncSnapshotSend *pSyncMsg = syncSnapshotSendFromRpcMsg2(pRpcMsg);
+          assert(pSyncMsg != NULL);
+          io->FpOnSyncSnapshotSend(io->pSyncNode, pSyncMsg);
+          syncSnapshotSendDestroy(pSyncMsg);
+        }
+
+      } else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_RSP) {
+        if (io->FpOnSyncSnapshotRsp != NULL) {
+          SyncSnapshotRsp *pSyncMsg = syncSnapshotRspFromRpcMsg2(pRpcMsg);
+          assert(pSyncMsg != NULL);
+          io->FpOnSyncSnapshotRsp(io->pSyncNode, pSyncMsg);
+          syncSnapshotRspDestroy(pSyncMsg);
+        }
+
       } else {
         sTrace("unknown msgType:%d, no operator", pRpcMsg->msgType);
       }

source/libs/sync/src/syncIndexMgr.c

@@ -46,6 +46,7 @@ void syncIndexMgrDestroy(SSyncIndexMgr *pSyncIndexMgr) {
 
 void syncIndexMgrClear(SSyncIndexMgr *pSyncIndexMgr) {
   memset(pSyncIndexMgr->index, 0, sizeof(pSyncIndexMgr->index));
+  memset(pSyncIndexMgr->privateTerm, 0, sizeof(pSyncIndexMgr->privateTerm));
   /*
   for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
     pSyncIndexMgr->index[i] = 0;
@@ -62,7 +63,7 @@ void syncIndexMgrSetIndex(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId,
   }
 
   // maybe config change
-  // assert(0);
+  assert(0);
 }
 
 SyncIndex syncIndexMgrGetIndex(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId) {
@@ -86,14 +87,27 @@ cJSON *syncIndexMgr2Json(SSyncIndexMgr *pSyncIndexMgr) {
     for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
       cJSON_AddItemToArray(pReplicas, syncUtilRaftId2Json(&(*(pSyncIndexMgr->replicas))[i]));
     }
-    int  respondNum = 0;
-    int *arr = (int *)taosMemoryMalloc(sizeof(int) * pSyncIndexMgr->replicaNum);
-    for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
-      arr[i] = pSyncIndexMgr->index[i];
+
+    {
+      int *arr = (int *)taosMemoryMalloc(sizeof(int) * pSyncIndexMgr->replicaNum);
+      for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
+        arr[i] = pSyncIndexMgr->index[i];
+      }
+      cJSON *pIndex = cJSON_CreateIntArray(arr, pSyncIndexMgr->replicaNum);
+      taosMemoryFree(arr);
+      cJSON_AddItemToObject(pRoot, "index", pIndex);
     }
-    cJSON *pIndex = cJSON_CreateIntArray(arr, pSyncIndexMgr->replicaNum);
-    taosMemoryFree(arr);
-    cJSON_AddItemToObject(pRoot, "index", pIndex);
+
+    {
+      int *arr = (int *)taosMemoryMalloc(sizeof(int) * pSyncIndexMgr->replicaNum);
+      for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
+        arr[i] = pSyncIndexMgr->privateTerm[i];
+      }
+      cJSON *pIndex = cJSON_CreateIntArray(arr, pSyncIndexMgr->replicaNum);
+      taosMemoryFree(arr);
+      cJSON_AddItemToObject(pRoot, "privateTerm", pIndex);
+    }
+
     snprintf(u64buf, sizeof(u64buf), "%p", pSyncIndexMgr->pSyncNode);
     cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
   }
@@ -105,7 +119,7 @@ cJSON *syncIndexMgr2Json(SSyncIndexMgr *pSyncIndexMgr) {
 
 char *syncIndexMgr2Str(SSyncIndexMgr *pSyncIndexMgr) {
   cJSON *pJson = syncIndexMgr2Json(pSyncIndexMgr);
-  char * serialized = cJSON_Print(pJson);
+  char  *serialized = cJSON_Print(pJson);
   cJSON_Delete(pJson);
   return serialized;
 }
@@ -132,7 +146,31 @@ void syncIndexMgrLog(SSyncIndexMgr *pObj) {
 }
 
 void syncIndexMgrLog2(char *s, SSyncIndexMgr *pObj) {
-  char *serialized = syncIndexMgr2Str(pObj);
-  sTrace("syncIndexMgrLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char *serialized = syncIndexMgr2Str(pObj);
+    sTrace("syncIndexMgrLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
+}
+
+void syncIndexMgrSetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId, SyncTerm term) {
+  for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
+    if (syncUtilSameId(&((*(pSyncIndexMgr->replicas))[i]), pRaftId)) {
+      (pSyncIndexMgr->privateTerm)[i] = term;
+      return;
+    }
+  }
+
+  // maybe config change
+  assert(0);
+}
+
+SyncTerm syncIndexMgrGetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId) {
+  for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
+    if (syncUtilSameId(&((*(pSyncIndexMgr->replicas))[i]), pRaftId)) {
+      SyncTerm term = (pSyncIndexMgr->privateTerm)[i];
+      return term;
+    }
+  }
+  assert(0);
 }
\ No newline at end of file

source/libs/sync/src/syncMessage.c

@@ -65,6 +65,16 @@ cJSON* syncRpcMsg2Json(SRpcMsg* pRpcMsg) {
     pRoot = syncAppendEntriesReply2Json(pSyncMsg);
     syncAppendEntriesReplyDestroy(pSyncMsg);
 
+  } else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_SEND) {
+    SyncSnapshotSend* pSyncMsg = syncSnapshotSendDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
+    pRoot = syncSnapshotSend2Json(pSyncMsg);
+    syncSnapshotSendDestroy(pSyncMsg);
+
+  } else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_RSP) {
+    SyncSnapshotRsp* pSyncMsg = syncSnapshotRspDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
+    pRoot = syncSnapshotRsp2Json(pSyncMsg);
+    syncSnapshotRspDestroy(pSyncMsg);
+
   } else if (pRpcMsg->msgType == TDMT_SYNC_COMMON_RESPONSE) {
     pRoot = cJSON_CreateObject();
     char* s;
@@ -135,9 +145,11 @@ void syncRpcMsgLog(SRpcMsg* pMsg) {
 }
 
 void syncRpcMsgLog2(char* s, SRpcMsg* pMsg) {
-  char* serialized = syncRpcMsg2Str(pMsg);
-  sTrace("syncRpcMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncRpcMsg2Str(pMsg);
+    sTrace("syncRpcMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncTimeout----
@@ -264,9 +276,11 @@ void syncTimeoutLog(const SyncTimeout* pMsg) {
 }
 
 void syncTimeoutLog2(char* s, const SyncTimeout* pMsg) {
-  char* serialized = syncTimeout2Str(pMsg);
-  sTrace("syncTimeoutLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncTimeout2Str(pMsg);
+    sTrace("syncTimeoutLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncPing----
@@ -524,9 +538,11 @@ void syncPingLog(const SyncPing* pMsg) {
 }
 
 void syncPingLog2(char* s, const SyncPing* pMsg) {
-  char* serialized = syncPing2Str(pMsg);
-  sTrace("syncPingLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncPing2Str(pMsg);
+    sTrace("syncPingLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncPingReply----
@@ -784,9 +800,11 @@ void syncPingReplyLog(const SyncPingReply* pMsg) {
 }
 
 void syncPingReplyLog2(char* s, const SyncPingReply* pMsg) {
-  char* serialized = syncPingReply2Str(pMsg);
-  sTrace("syncPingReplyLog2 | len:%zu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncPingReply2Str(pMsg);
+    sTrace("syncPingReplyLog2 | len:%zu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncClientRequest----
@@ -925,9 +943,11 @@ void syncClientRequestLog(const SyncClientRequest* pMsg) {
 }
 
 void syncClientRequestLog2(char* s, const SyncClientRequest* pMsg) {
-  char* serialized = syncClientRequest2Str(pMsg);
-  sTrace("syncClientRequestLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncClientRequest2Str(pMsg);
+    sTrace("syncClientRequestLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncRequestVote----
@@ -1074,9 +1094,11 @@ void syncRequestVoteLog(const SyncRequestVote* pMsg) {
 }
 
 void syncRequestVoteLog2(char* s, const SyncRequestVote* pMsg) {
-  char* serialized = syncRequestVote2Str(pMsg);
-  sTrace("syncRequestVoteLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncRequestVote2Str(pMsg);
+    sTrace("syncRequestVoteLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncRequestVoteReply----
@@ -1220,9 +1242,11 @@ void syncRequestVoteReplyLog(const SyncRequestVoteReply* pMsg) {
 }
 
 void syncRequestVoteReplyLog2(char* s, const SyncRequestVoteReply* pMsg) {
-  char* serialized = syncRequestVoteReply2Str(pMsg);
-  sTrace("syncRequestVoteReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncRequestVoteReply2Str(pMsg);
+    sTrace("syncRequestVoteReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncAppendEntries----
@@ -1333,6 +1357,9 @@ cJSON* syncAppendEntries2Json(const SyncAppendEntries* pMsg) {
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
     cJSON_AddStringToObject(pRoot, "term", u64buf);
 
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
+    cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
+
     snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->prevLogIndex);
     cJSON_AddStringToObject(pRoot, "prevLogIndex", u64buf);
 
@@ -1386,9 +1413,11 @@ void syncAppendEntriesLog(const SyncAppendEntries* pMsg) {
 }
 
 void syncAppendEntriesLog2(char* s, const SyncAppendEntries* pMsg) {
-  char* serialized = syncAppendEntries2Str(pMsg);
-  sTrace("syncAppendEntriesLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncAppendEntries2Str(pMsg);
+    sTrace("syncAppendEntriesLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncAppendEntriesReply----
@@ -1494,6 +1523,9 @@ cJSON* syncAppendEntriesReply2Json(const SyncAppendEntriesReply* pMsg) {
     cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
     cJSON_AddItemToObject(pRoot, "destId", pDestId);
 
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
+    cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
+
     snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
     cJSON_AddStringToObject(pRoot, "term", u64buf);
     cJSON_AddNumberToObject(pRoot, "success", pMsg->success);
@@ -1535,9 +1567,11 @@ void syncAppendEntriesReplyLog(const SyncAppendEntriesReply* pMsg) {
 }
 
 void syncAppendEntriesReplyLog2(char* s, const SyncAppendEntriesReply* pMsg) {
-  char* serialized = syncAppendEntriesReply2Str(pMsg);
-  sTrace("syncAppendEntriesReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncAppendEntriesReply2Str(pMsg);
+    sTrace("syncAppendEntriesReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }
 
 // ---- message process SyncApplyMsg----
@@ -1686,7 +1720,339 @@ void syncApplyMsgLog(const SyncApplyMsg* pMsg) {
 }
 
 void syncApplyMsgLog2(char* s, const SyncApplyMsg* pMsg) {
-  char* serialized = syncApplyMsg2Str(pMsg);
-  sTrace("syncApplyMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+  if (gRaftDetailLog) {
+    char* serialized = syncApplyMsg2Str(pMsg);
+    sTrace("syncApplyMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
+}
+
+// ---------------------------------------------
+SyncSnapshotSend* syncSnapshotSendBuild(uint32_t dataLen, int32_t vgId) {
+  uint32_t          bytes = sizeof(SyncSnapshotSend) + dataLen;
+  SyncSnapshotSend* pMsg = taosMemoryMalloc(bytes);
+  memset(pMsg, 0, bytes);
+  pMsg->bytes = bytes;
+  pMsg->vgId = vgId;
+  pMsg->msgType = TDMT_SYNC_SNAPSHOT_SEND;
+  pMsg->dataLen = dataLen;
+  return pMsg;
+}
+
+void syncSnapshotSendDestroy(SyncSnapshotSend* pMsg) {
+  if (pMsg != NULL) {
+    taosMemoryFree(pMsg);
+  }
+}
+
+void syncSnapshotSendSerialize(const SyncSnapshotSend* pMsg, char* buf, uint32_t bufLen) {
+  assert(pMsg->bytes <= bufLen);
+  memcpy(buf, pMsg, pMsg->bytes);
+}
+
+void syncSnapshotSendDeserialize(const char* buf, uint32_t len, SyncSnapshotSend* pMsg) {
+  memcpy(pMsg, buf, len);
+  assert(len == pMsg->bytes);
+  assert(pMsg->bytes == sizeof(SyncSnapshotSend) + pMsg->dataLen);
+}
+
+char* syncSnapshotSendSerialize2(const SyncSnapshotSend* pMsg, uint32_t* len) {
+  char* buf = taosMemoryMalloc(pMsg->bytes);
+  assert(buf != NULL);
+  syncSnapshotSendSerialize(pMsg, buf, pMsg->bytes);
+  if (len != NULL) {
+    *len = pMsg->bytes;
+  }
+  return buf;
+}
+
+SyncSnapshotSend* syncSnapshotSendDeserialize2(const char* buf, uint32_t len) {
+  uint32_t          bytes = *((uint32_t*)buf);
+  SyncSnapshotSend* pMsg = taosMemoryMalloc(bytes);
+  assert(pMsg != NULL);
+  syncSnapshotSendDeserialize(buf, len, pMsg);
+  assert(len == pMsg->bytes);
+  return pMsg;
+}
+
+void syncSnapshotSend2RpcMsg(const SyncSnapshotSend* pMsg, SRpcMsg* pRpcMsg) {
+  memset(pRpcMsg, 0, sizeof(*pRpcMsg));
+  pRpcMsg->msgType = pMsg->msgType;
+  pRpcMsg->contLen = pMsg->bytes;
+  pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
+  syncSnapshotSendSerialize(pMsg, pRpcMsg->pCont, pRpcMsg->contLen);
+}
+
+void syncSnapshotSendFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotSend* pMsg) {
+  syncSnapshotSendDeserialize(pRpcMsg->pCont, pRpcMsg->contLen, pMsg);
+}
+
+SyncSnapshotSend* syncSnapshotSendFromRpcMsg2(const SRpcMsg* pRpcMsg) {
+  SyncSnapshotSend* pMsg = syncSnapshotSendDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
+  assert(pMsg != NULL);
+  return pMsg;
+}
+
+cJSON* syncSnapshotSend2Json(const SyncSnapshotSend* pMsg) {
+  char   u64buf[128];
+  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* pSrcId = cJSON_CreateObject();
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->srcId.addr);
+    cJSON_AddStringToObject(pSrcId, "addr", u64buf);
+    {
+      uint64_t u64 = pMsg->srcId.addr;
+      cJSON*   pTmp = pSrcId;
+      char     host[128];
+      uint16_t port;
+      syncUtilU642Addr(u64, host, sizeof(host), &port);
+      cJSON_AddStringToObject(pTmp, "addr_host", host);
+      cJSON_AddNumberToObject(pTmp, "addr_port", port);
+    }
+    cJSON_AddNumberToObject(pSrcId, "vgId", pMsg->srcId.vgId);
+    cJSON_AddItemToObject(pRoot, "srcId", pSrcId);
+
+    cJSON* pDestId = cJSON_CreateObject();
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->destId.addr);
+    cJSON_AddStringToObject(pDestId, "addr", u64buf);
+    {
+      uint64_t u64 = pMsg->destId.addr;
+      cJSON*   pTmp = pDestId;
+      char     host[128];
+      uint16_t port;
+      syncUtilU642Addr(u64, host, sizeof(host), &port);
+      cJSON_AddStringToObject(pTmp, "addr_host", host);
+      cJSON_AddNumberToObject(pTmp, "addr_port", port);
+    }
+    cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
+    cJSON_AddItemToObject(pRoot, "destId", pDestId);
+
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
+    cJSON_AddStringToObject(pRoot, "term", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
+    cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->lastIndex);
+    cJSON_AddStringToObject(pRoot, "lastIndex", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastTerm);
+    cJSON_AddStringToObject(pRoot, "lastTerm", u64buf);
+
+    cJSON_AddNumberToObject(pRoot, "seq", pMsg->seq);
+
+    cJSON_AddNumberToObject(pRoot, "dataLen", pMsg->dataLen);
+    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, "SyncSnapshotSend", pRoot);
+  return pJson;
+}
+
+char* syncSnapshotSend2Str(const SyncSnapshotSend* pMsg) {
+  cJSON* pJson = syncSnapshotSend2Json(pMsg);
+  char*  serialized = cJSON_Print(pJson);
+  cJSON_Delete(pJson);
+  return serialized;
+}
+
+// for debug ----------------------
+void syncSnapshotSendPrint(const SyncSnapshotSend* pMsg) {
+  char* serialized = syncSnapshotSend2Str(pMsg);
+  printf("syncSnapshotSendPrint | len:%lu | %s \n", strlen(serialized), serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void syncSnapshotSendPrint2(char* s, const SyncSnapshotSend* pMsg) {
+  char* serialized = syncSnapshotSend2Str(pMsg);
+  printf("syncSnapshotSendPrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void syncSnapshotSendLog(const SyncSnapshotSend* pMsg) {
+  char* serialized = syncSnapshotSend2Str(pMsg);
+  sTrace("syncSnapshotSendLog | len:%lu | %s", strlen(serialized), serialized);
   taosMemoryFree(serialized);
 }
+
+void syncSnapshotSendLog2(char* s, const SyncSnapshotSend* pMsg) {
+  if (gRaftDetailLog) {
+    char* serialized = syncSnapshotSend2Str(pMsg);
+    sTrace("syncSnapshotSendLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
+}
+
+// ---------------------------------------------
+SyncSnapshotRsp* syncSnapshotRspBuild(int32_t vgId) {
+  uint32_t         bytes = sizeof(SyncSnapshotRsp);
+  SyncSnapshotRsp* pMsg = taosMemoryMalloc(bytes);
+  memset(pMsg, 0, bytes);
+  pMsg->bytes = bytes;
+  pMsg->vgId = vgId;
+  pMsg->msgType = TDMT_SYNC_SNAPSHOT_RSP;
+  return pMsg;
+}
+
+void syncSnapshotRspDestroy(SyncSnapshotRsp* pMsg) {
+  if (pMsg != NULL) {
+    taosMemoryFree(pMsg);
+  }
+}
+
+void syncSnapshotRspSerialize(const SyncSnapshotRsp* pMsg, char* buf, uint32_t bufLen) {
+  assert(pMsg->bytes <= bufLen);
+  memcpy(buf, pMsg, pMsg->bytes);
+}
+
+void syncSnapshotRspDeserialize(const char* buf, uint32_t len, SyncSnapshotRsp* pMsg) {
+  memcpy(pMsg, buf, len);
+  assert(len == pMsg->bytes);
+}
+
+char* syncSnapshotRspSerialize2(const SyncSnapshotRsp* pMsg, uint32_t* len) {
+  char* buf = taosMemoryMalloc(pMsg->bytes);
+  assert(buf != NULL);
+  syncSnapshotRspSerialize(pMsg, buf, pMsg->bytes);
+  if (len != NULL) {
+    *len = pMsg->bytes;
+  }
+  return buf;
+}
+
+SyncSnapshotRsp* syncSnapshotRspDeserialize2(const char* buf, uint32_t len) {
+  uint32_t         bytes = *((uint32_t*)buf);
+  SyncSnapshotRsp* pMsg = taosMemoryMalloc(bytes);
+  assert(pMsg != NULL);
+  syncSnapshotRspDeserialize(buf, len, pMsg);
+  assert(len == pMsg->bytes);
+  return pMsg;
+}
+
+void syncSnapshotRsp2RpcMsg(const SyncSnapshotRsp* pMsg, SRpcMsg* pRpcMsg) {
+  memset(pRpcMsg, 0, sizeof(*pRpcMsg));
+  pRpcMsg->msgType = pMsg->msgType;
+  pRpcMsg->contLen = pMsg->bytes;
+  pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
+  syncSnapshotRspSerialize(pMsg, pRpcMsg->pCont, pRpcMsg->contLen);
+}
+
+void syncSnapshotRspFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotRsp* pMsg) {
+  syncSnapshotRspDeserialize(pRpcMsg->pCont, pRpcMsg->contLen, pMsg);
+}
+
+SyncSnapshotRsp* syncSnapshotRspFromRpcMsg2(const SRpcMsg* pRpcMsg) {
+  SyncSnapshotRsp* pMsg = syncSnapshotRspDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
+  assert(pMsg != NULL);
+  return pMsg;
+}
+
+cJSON* syncSnapshotRsp2Json(const SyncSnapshotRsp* pMsg) {
+  char   u64buf[128];
+  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* pSrcId = cJSON_CreateObject();
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->srcId.addr);
+    cJSON_AddStringToObject(pSrcId, "addr", u64buf);
+    {
+      uint64_t u64 = pMsg->srcId.addr;
+      cJSON*   pTmp = pSrcId;
+      char     host[128];
+      uint16_t port;
+      syncUtilU642Addr(u64, host, sizeof(host), &port);
+      cJSON_AddStringToObject(pTmp, "addr_host", host);
+      cJSON_AddNumberToObject(pTmp, "addr_port", port);
+    }
+    cJSON_AddNumberToObject(pSrcId, "vgId", pMsg->srcId.vgId);
+    cJSON_AddItemToObject(pRoot, "srcId", pSrcId);
+
+    cJSON* pDestId = cJSON_CreateObject();
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->destId.addr);
+    cJSON_AddStringToObject(pDestId, "addr", u64buf);
+    {
+      uint64_t u64 = pMsg->destId.addr;
+      cJSON*   pTmp = pDestId;
+      char     host[128];
+      uint16_t port;
+      syncUtilU642Addr(u64, host, sizeof(host), &port);
+      cJSON_AddStringToObject(pTmp, "addr_host", host);
+      cJSON_AddNumberToObject(pTmp, "addr_port", port);
+    }
+    cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
+    cJSON_AddItemToObject(pRoot, "destId", pDestId);
+
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
+    cJSON_AddStringToObject(pRoot, "term", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
+    cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->lastIndex);
+    cJSON_AddStringToObject(pRoot, "lastIndex", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastTerm);
+    cJSON_AddStringToObject(pRoot, "lastTerm", u64buf);
+
+    cJSON_AddNumberToObject(pRoot, "ack", pMsg->ack);
+    cJSON_AddNumberToObject(pRoot, "code", pMsg->code);
+  }
+
+  cJSON* pJson = cJSON_CreateObject();
+  cJSON_AddItemToObject(pJson, "SyncSnapshotRsp", pRoot);
+  return pJson;
+}
+
+char* syncSnapshotRsp2Str(const SyncSnapshotRsp* pMsg) {
+  cJSON* pJson = syncSnapshotRsp2Json(pMsg);
+  char*  serialized = cJSON_Print(pJson);
+  cJSON_Delete(pJson);
+  return serialized;
+}
+
+// for debug ----------------------
+void syncSnapshotRspPrint(const SyncSnapshotRsp* pMsg) {
+  char* serialized = syncSnapshotRsp2Str(pMsg);
+  printf("syncSnapshotRspPrint | len:%lu | %s \n", strlen(serialized), serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void syncSnapshotRspPrint2(char* s, const SyncSnapshotRsp* pMsg) {
+  char* serialized = syncSnapshotRsp2Str(pMsg);
+  printf("syncSnapshotRspPrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void syncSnapshotRspLog(const SyncSnapshotRsp* pMsg) {
+  char* serialized = syncSnapshotRsp2Str(pMsg);
+  sTrace("syncSnapshotRspLog | len:%lu | %s", strlen(serialized), serialized);
+  taosMemoryFree(serialized);
+}
+
+void syncSnapshotRspLog2(char* s, const SyncSnapshotRsp* pMsg) {
+  if (gRaftDetailLog) {
+    char* serialized = syncSnapshotRsp2Str(pMsg);
+    sTrace("syncSnapshotRspLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
+}
\ No newline at end of file

source/libs/sync/src/syncRaftCfg.c

@@ -148,6 +148,7 @@ cJSON *raftCfg2Json(SRaftCfg *pRaftCfg) {
   cJSON *pRoot = cJSON_CreateObject();
   cJSON_AddItemToObject(pRoot, "SSyncCfg", syncCfg2Json(&(pRaftCfg->cfg)));
   cJSON_AddNumberToObject(pRoot, "isStandBy", pRaftCfg->isStandBy);
+  cJSON_AddNumberToObject(pRoot, "snapshotEnable", pRaftCfg->snapshotEnable);
 
   cJSON *pJson = cJSON_CreateObject();
   cJSON_AddItemToObject(pJson, "RaftCfg", pRoot);
@@ -161,7 +162,7 @@ char *raftCfg2Str(SRaftCfg *pRaftCfg) {
   return serialized;
 }
 
-int32_t raftCfgCreateFile(SSyncCfg *pCfg, int8_t isStandBy, const char *path) {
+int32_t raftCfgCreateFile(SSyncCfg *pCfg, SRaftCfgMeta meta, const char *path) {
   assert(pCfg != NULL);
 
   TdFilePtr pFile = taosOpenFile(path, TD_FILE_CREATE | TD_FILE_WRITE);
@@ -169,7 +170,8 @@ int32_t raftCfgCreateFile(SSyncCfg *pCfg, int8_t isStandBy, const char *path) {
 
   SRaftCfg raftCfg;
   raftCfg.cfg = *pCfg;
-  raftCfg.isStandBy = isStandBy;
+  raftCfg.isStandBy = meta.isStandBy;
+  raftCfg.snapshotEnable = meta.snapshotEnable;
   char *s = raftCfg2Str(&raftCfg);
 
   char buf[CONFIG_FILE_LEN] = {0};
@@ -194,6 +196,9 @@ int32_t raftCfgFromJson(const cJSON *pRoot, SRaftCfg *pRaftCfg) {
   cJSON *pJsonIsStandBy = cJSON_GetObjectItem(pJson, "isStandBy");
   pRaftCfg->isStandBy = cJSON_GetNumberValue(pJsonIsStandBy);
 
+  cJSON *pJsonSnapshotEnable = cJSON_GetObjectItem(pJson, "snapshotEnable");
+  pRaftCfg->snapshotEnable = cJSON_GetNumberValue(pJsonSnapshotEnable);
+
   cJSON * pJsonSyncCfg = cJSON_GetObjectItem(pJson, "SSyncCfg");
   int32_t code = syncCfgFromJson(pJsonSyncCfg, &(pRaftCfg->cfg));
   ASSERT(code == 0);

source/libs/sync/src/syncRaftLog.c

@@ -16,6 +16,23 @@
 #include "syncRaftLog.h"
 #include "wal.h"
 
+// refactor, log[0 .. n] ==> log[m .. n]
+static int32_t   raftLogSetBeginIndex(struct SSyncLogStore* pLogStore, SyncIndex beginIndex);
+static SyncIndex raftLogBeginIndex(struct SSyncLogStore* pLogStore);
+static SyncIndex raftLogEndIndex(struct SSyncLogStore* pLogStore);
+static SyncIndex raftLogWriteIndex(struct SSyncLogStore* pLogStore);
+static bool      raftLogIsEmpty(struct SSyncLogStore* pLogStore);
+static int32_t   raftLogEntryCount(struct SSyncLogStore* pLogStore);
+static bool      raftLogInRange(struct SSyncLogStore* pLogStore, SyncIndex index);
+static SyncIndex raftLogLastIndex(struct SSyncLogStore* pLogStore);
+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 int32_t raftLogGetLastEntry(SSyncLogStore* pLogStore, SSyncRaftEntry** ppLastEntry);
+
+//-------------------------------
 static SSyncRaftEntry* logStoreGetLastEntry(SSyncLogStore* pLogStore);
 static SyncIndex       logStoreLastIndex(SSyncLogStore* pLogStore);
 static SyncTerm        logStoreLastTerm(SSyncLogStore* pLogStore);
@@ -25,6 +42,202 @@ static int32_t         logStoreTruncate(SSyncLogStore* pLogStore, SyncIndex from
 static int32_t         logStoreUpdateCommitIndex(SSyncLogStore* pLogStore, SyncIndex index);
 static SyncIndex       logStoreGetCommitIndex(SSyncLogStore* pLogStore);
 
+// refactor, log[0 .. n] ==> log[m .. n]
+static int32_t raftLogSetBeginIndex(struct SSyncLogStore* pLogStore, SyncIndex beginIndex) {
+  sTrace("raftLogSetBeginIndex beginIndex:%ld", beginIndex);
+
+  // if beginIndex == 0, donot need call this funciton
+  ASSERT(beginIndex > 0);
+
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  pData->beginIndex = beginIndex;
+  walRestoreFromSnapshot(pWal, beginIndex - 1);
+  return 0;
+}
+
+int32_t raftLogResetBeginIndex(struct SSyncLogStore* pLogStore) { return 0; }
+
+static SyncIndex raftLogBeginIndex(struct SSyncLogStore* pLogStore) {
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  return pData->beginIndex;
+}
+
+static SyncIndex raftLogEndIndex(struct SSyncLogStore* pLogStore) { return raftLogLastIndex(pLogStore); }
+
+static bool raftLogIsEmpty(struct SSyncLogStore* pLogStore) {
+  SyncIndex beginIndex = raftLogBeginIndex(pLogStore);
+  SyncIndex endIndex = raftLogEndIndex(pLogStore);
+  return (endIndex < beginIndex);
+}
+
+static int32_t raftLogEntryCount(struct SSyncLogStore* pLogStore) {
+  SyncIndex beginIndex = raftLogBeginIndex(pLogStore);
+  SyncIndex endIndex = raftLogEndIndex(pLogStore);
+  int32_t   count = endIndex - beginIndex + 1;
+  return count > 0 ? count : 0;
+}
+
+static bool raftLogInRange(struct SSyncLogStore* pLogStore, SyncIndex index) {
+  SyncIndex beginIndex = raftLogBeginIndex(pLogStore);
+  SyncIndex endIndex = raftLogEndIndex(pLogStore);
+  if (index >= beginIndex && index <= endIndex) {
+    return true;
+  } else {
+    return false;
+  }
+}
+
+static SyncIndex raftLogLastIndex(struct SSyncLogStore* pLogStore) {
+  SyncIndex          lastIndex;
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  SyncIndex          lastVer = walGetLastVer(pWal);
+  SyncIndex          firstVer = walGetFirstVer(pWal);
+
+  if (lastVer < firstVer) {
+    // no record
+    lastIndex = -1;
+
+  } else {
+    if (firstVer >= 0) {
+      lastIndex = lastVer;
+    } else if (firstVer == -1) {
+      lastIndex = -1;
+    } else {
+      ASSERT(0);
+    }
+  }
+
+  return lastIndex;
+}
+
+static SyncIndex raftLogWriteIndex(struct SSyncLogStore* pLogStore) {
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  SyncIndex          lastVer = walGetLastVer(pWal);
+  return lastVer + 1;
+}
+
+static SyncTerm raftLogLastTerm(struct SSyncLogStore* pLogStore) {
+  SyncTerm lastTerm = 0;
+  if (raftLogEntryCount(pLogStore) == 0) {
+    lastTerm = 0;
+  } else {
+    SSyncRaftEntry* pLastEntry;
+    int32_t         code = raftLogGetLastEntry(pLogStore, &pLastEntry);
+    ASSERT(code == 0);
+    if (pLastEntry != NULL) {
+      lastTerm = pLastEntry->term;
+      taosMemoryFree(pLastEntry);
+    }
+  }
+  return lastTerm;
+}
+
+static int32_t raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+
+  SyncIndex writeIndex = raftLogWriteIndex(pLogStore);
+  ASSERT(pEntry->index == writeIndex);
+
+  int          code = 0;
+  SSyncLogMeta syncMeta;
+  syncMeta.isWeek = pEntry->isWeak;
+  syncMeta.seqNum = pEntry->seqNum;
+  syncMeta.term = pEntry->term;
+  code = walWriteWithSyncInfo(pWal, pEntry->index, pEntry->originalRpcType, syncMeta, pEntry->data, pEntry->dataLen);
+  if (code != 0) {
+    int32_t     err = terrno;
+    const char* errStr = tstrerror(err);
+    int32_t     linuxErr = errno;
+    const char* linuxErrMsg = strerror(errno);
+    sError("raftLogAppendEntry error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr,
+           linuxErrMsg);
+    ASSERT(0);
+  }
+
+  walFsync(pWal, true);
+
+  sTrace("sync event write index:%" PRId64, pEntry->index);
+
+  return code;
+}
+
+static int32_t raftLogGetEntry(struct SSyncLogStore* pLogStore, SyncIndex index, SSyncRaftEntry** ppEntry) {
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  int32_t            code;
+
+  *ppEntry = NULL;
+  if (raftLogInRange(pLogStore, index)) {
+    SWalReadHandle* pWalHandle = walOpenReadHandle(pWal);
+    ASSERT(pWalHandle != NULL);
+
+    code = walReadWithHandle(pWalHandle, index);
+    if (code != 0) {
+      int32_t     err = terrno;
+      const char* errStr = tstrerror(err);
+      int32_t     linuxErr = errno;
+      const char* linuxErrMsg = strerror(errno);
+      sError("raftLogGetEntry error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr,
+             linuxErrMsg);
+      ASSERT(0);
+      walCloseReadHandle(pWalHandle);
+      return code;
+    }
+
+    *ppEntry = syncEntryBuild(pWalHandle->pHead->head.bodyLen);
+    ASSERT(*ppEntry != NULL);
+    (*ppEntry)->msgType = TDMT_SYNC_CLIENT_REQUEST;
+    (*ppEntry)->originalRpcType = pWalHandle->pHead->head.msgType;
+    (*ppEntry)->seqNum = pWalHandle->pHead->head.syncMeta.seqNum;
+    (*ppEntry)->isWeak = pWalHandle->pHead->head.syncMeta.isWeek;
+    (*ppEntry)->term = pWalHandle->pHead->head.syncMeta.term;
+    (*ppEntry)->index = index;
+    ASSERT((*ppEntry)->dataLen == pWalHandle->pHead->head.bodyLen);
+    memcpy((*ppEntry)->data, pWalHandle->pHead->head.body, pWalHandle->pHead->head.bodyLen);
+
+    // need to hold, do not new every time!!
+    walCloseReadHandle(pWalHandle);
+
+  } else {
+    // index not in range
+    code = 0;
+  }
+
+  return code;
+}
+
+static int32_t raftLogTruncate(struct SSyncLogStore* pLogStore, SyncIndex fromIndex) {
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  int32_t            code = walRollback(pWal, fromIndex);
+  if (code != 0) {
+    int32_t     err = terrno;
+    const char* errStr = tstrerror(err);
+    int32_t     linuxErr = errno;
+    const char* linuxErrMsg = strerror(errno);
+    sError("raftLogTruncate error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr,
+           linuxErrMsg);
+    ASSERT(0);
+  }
+  return code;
+}
+
+static int32_t raftLogGetLastEntry(SSyncLogStore* pLogStore, SSyncRaftEntry** ppLastEntry) {
+  *ppLastEntry = NULL;
+  if (raftLogEntryCount(pLogStore) == 0) {
+    return 0;
+  }
+  SyncIndex lastIndex = raftLogLastIndex(pLogStore);
+  int32_t   code = raftLogGetEntry(pLogStore, lastIndex, ppLastEntry);
+  return code;
+}
+
+//-------------------------------
 SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
   SSyncLogStore* pLogStore = taosMemoryMalloc(sizeof(SSyncLogStore));
   assert(pLogStore != NULL);
@@ -36,6 +249,16 @@ SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
   pData->pSyncNode = pSyncNode;
   pData->pWal = pSyncNode->pWal;
 
+  SyncIndex firstVer = walGetFirstVer(pData->pWal);
+  SyncIndex lastVer = walGetLastVer(pData->pWal);
+  if (firstVer >= 0) {
+    pData->beginIndex = firstVer;
+  } else if (firstVer == -1) {
+    pData->beginIndex = lastVer + 1;
+  } else {
+    ASSERT(0);
+  }
+
   pLogStore->appendEntry = logStoreAppendEntry;
   pLogStore->getEntry = logStoreGetEntry;
   pLogStore->truncate = logStoreTruncate;
@@ -43,6 +266,20 @@ SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
   pLogStore->getLastTerm = logStoreLastTerm;
   pLogStore->updateCommitIndex = logStoreUpdateCommitIndex;
   pLogStore->getCommitIndex = logStoreGetCommitIndex;
+
+  pLogStore->syncLogSetBeginIndex = raftLogSetBeginIndex;
+  pLogStore->syncLogBeginIndex = raftLogBeginIndex;
+  pLogStore->syncLogEndIndex = raftLogEndIndex;
+  pLogStore->syncLogIsEmpty = raftLogIsEmpty;
+  pLogStore->syncLogEntryCount = raftLogEntryCount;
+  pLogStore->syncLogInRange = raftLogInRange;
+  pLogStore->syncLogLastIndex = raftLogLastIndex;
+  pLogStore->syncLogLastTerm = raftLogLastTerm;
+  pLogStore->syncLogAppendEntry = raftLogAppendEntry;
+  pLogStore->syncLogGetEntry = raftLogGetEntry;
+  pLogStore->syncLogTruncate = raftLogTruncate;
+  pLogStore->syncLogWriteIndex = raftLogWriteIndex;
+
   return pLogStore;
 }
 
@@ -53,6 +290,7 @@ void logStoreDestory(SSyncLogStore* pLogStore) {
   }
 }
 
+//-------------------------------
 int32_t logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
   SSyncLogStoreData* pData = pLogStore->data;
   SWal*              pWal = pData->pWal;
@@ -78,6 +316,8 @@ int32_t logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
   // assert(code == 0);
 
   walFsync(pWal, true);
+
+  sTrace("sync event old write wal: %ld", pEntry->index);
   return code;
 }
 
@@ -136,7 +376,7 @@ int32_t logStoreTruncate(SSyncLogStore* pLogStore, SyncIndex fromIndex) {
            linuxErrMsg);
     ASSERT(0);
   }
-  return 0;  // to avoid compiler error
+  return 0;
 }
 
 SyncIndex logStoreLastIndex(SSyncLogStore* pLogStore) {
@@ -169,7 +409,7 @@ int32_t logStoreUpdateCommitIndex(SSyncLogStore* pLogStore, SyncIndex index) {
     sError("walCommit error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr, linuxErrMsg);
     ASSERT(0);
   }
-  return 0;  // to avoid compiler error
+  return 0;
 }
 
 SyncIndex logStoreGetCommitIndex(SSyncLogStore* pLogStore) {
@@ -199,15 +439,32 @@ cJSON* logStore2Json(SSyncLogStore* pLogStore) {
     cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
     snprintf(u64buf, sizeof(u64buf), "%p", pData->pWal);
     cJSON_AddStringToObject(pRoot, "pWal", u64buf);
-    snprintf(u64buf, sizeof(u64buf), "%ld", logStoreLastIndex(pLogStore));
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", pData->beginIndex);
+    cJSON_AddStringToObject(pRoot, "beginIndex", u64buf);
+
+    SyncIndex endIndex = raftLogEndIndex(pLogStore);
+    snprintf(u64buf, sizeof(u64buf), "%ld", endIndex);
+    cJSON_AddStringToObject(pRoot, "endIndex", u64buf);
+
+    int32_t count = raftLogEntryCount(pLogStore);
+    cJSON_AddNumberToObject(pRoot, "entryCount", count);
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", raftLogWriteIndex(pLogStore));
+    cJSON_AddStringToObject(pRoot, "WriteIndex", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%d", raftLogIsEmpty(pLogStore));
+    cJSON_AddStringToObject(pRoot, "IsEmpty", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", raftLogLastIndex(pLogStore));
     cJSON_AddStringToObject(pRoot, "LastIndex", u64buf);
-    snprintf(u64buf, sizeof(u64buf), "%lu", logStoreLastTerm(pLogStore));
+    snprintf(u64buf, sizeof(u64buf), "%lu", raftLogLastTerm(pLogStore));
     cJSON_AddStringToObject(pRoot, "LastTerm", u64buf);
 
     cJSON* pEntries = cJSON_CreateArray();
     cJSON_AddItemToObject(pRoot, "pEntries", pEntries);
-    SyncIndex lastIndex = logStoreLastIndex(pLogStore);
-    for (SyncIndex i = 0; i <= lastIndex; ++i) {
+
+    for (SyncIndex i = pData->beginIndex; i <= endIndex; ++i) {
       SSyncRaftEntry* pEntry = logStoreGetEntry(pLogStore, i);
       cJSON_AddItemToArray(pEntries, syncEntry2Json(pEntry));
       syncEntryDestory(pEntry);
@@ -236,9 +493,26 @@ cJSON* logStoreSimple2Json(SSyncLogStore* pLogStore) {
     cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
     snprintf(u64buf, sizeof(u64buf), "%p", pData->pWal);
     cJSON_AddStringToObject(pRoot, "pWal", u64buf);
-    snprintf(u64buf, sizeof(u64buf), "%ld", logStoreLastIndex(pLogStore));
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", pData->beginIndex);
+    cJSON_AddStringToObject(pRoot, "beginIndex", u64buf);
+
+    SyncIndex endIndex = raftLogEndIndex(pLogStore);
+    snprintf(u64buf, sizeof(u64buf), "%ld", endIndex);
+    cJSON_AddStringToObject(pRoot, "endIndex", u64buf);
+
+    int32_t count = raftLogEntryCount(pLogStore);
+    cJSON_AddNumberToObject(pRoot, "entryCount", count);
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", raftLogWriteIndex(pLogStore));
+    cJSON_AddStringToObject(pRoot, "WriteIndex", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%d", raftLogIsEmpty(pLogStore));
+    cJSON_AddStringToObject(pRoot, "IsEmpty", u64buf);
+
+    snprintf(u64buf, sizeof(u64buf), "%ld", raftLogLastIndex(pLogStore));
     cJSON_AddStringToObject(pRoot, "LastIndex", u64buf);
-    snprintf(u64buf, sizeof(u64buf), "%lu", logStoreLastTerm(pLogStore));
+    snprintf(u64buf, sizeof(u64buf), "%lu", raftLogLastTerm(pLogStore));
     cJSON_AddStringToObject(pRoot, "LastTerm", u64buf);
   }
 
@@ -254,6 +528,12 @@ char* logStoreSimple2Str(SSyncLogStore* pLogStore) {
   return serialized;
 }
 
+SyncIndex logStoreFirstIndex(SSyncLogStore* pLogStore) {
+  SSyncLogStoreData* pData = pLogStore->data;
+  SWal*              pWal = pData->pWal;
+  return walGetFirstVer(pWal);
+}
+
 // for debug -----------------
 void logStorePrint(SSyncLogStore* pLogStore) {
   char* serialized = logStore2Str(pLogStore);
@@ -303,7 +583,9 @@ void logStoreSimpleLog(SSyncLogStore* pLogStore) {
 }
 
 void logStoreSimpleLog2(char* s, SSyncLogStore* pLogStore) {
-  char* serialized = logStoreSimple2Str(pLogStore);
-  sTrace("logStoreSimpleLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
-  taosMemoryFree(serialized);
+  if (gRaftDetailLog) {
+    char* serialized = logStoreSimple2Str(pLogStore);
+    sTrace("logStoreSimpleLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
 }

source/libs/sync/src/syncReplication.c

@@ -16,9 +16,11 @@
 #include "syncReplication.h"
 #include "syncIndexMgr.h"
 #include "syncMessage.h"
+#include "syncRaftCfg.h"
 #include "syncRaftEntry.h"
 #include "syncRaftLog.h"
 #include "syncRaftStore.h"
+#include "syncSnapshot.h"
 #include "syncUtil.h"
 
 // TLA+ Spec
@@ -59,6 +61,7 @@ int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode) {
 
     // set prevLogIndex
     SyncIndex nextIndex = syncIndexMgrGetIndex(pSyncNode->pNextIndex, pDestId);
+
     SyncIndex preLogIndex = nextIndex - 1;
 
     // set preLogTerm
@@ -113,9 +116,87 @@ int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode) {
   return ret;
 }
 
+int32_t syncNodeAppendEntriesPeersSnapshot(SSyncNode* pSyncNode) {
+  ASSERT(pSyncNode->state == TAOS_SYNC_STATE_LEADER);
+
+  syncIndexMgrLog2("begin append entries peers pNextIndex:", pSyncNode->pNextIndex);
+  syncIndexMgrLog2("begin append entries peers pMatchIndex:", pSyncNode->pMatchIndex);
+  logStoreSimpleLog2("begin append entries peers LogStore:", pSyncNode->pLogStore);
+  {
+    SSnapshot snapshot;
+    pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
+    sTrace("begin append entries peers, snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
+           snapshot.lastApplyIndex, snapshot.lastApplyTerm);
+  }
+
+  int32_t ret = 0;
+  for (int i = 0; i < pSyncNode->peersNum; ++i) {
+    SRaftId* pDestId = &(pSyncNode->peersId[i]);
+
+    // next index
+    SyncIndex nextIndex = syncIndexMgrGetIndex(pSyncNode->pNextIndex, pDestId);
+
+    // pre index, pre term
+    SyncIndex preLogIndex = syncNodeGetPreIndex(pSyncNode, nextIndex);
+    SyncTerm  preLogTerm = syncNodeGetPreTerm(pSyncNode, nextIndex);
+
+    // batch optimized
+    // SyncIndex lastIndex = syncUtilMinIndex(pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore), nextIndex);
+
+    // prepare entry
+    SyncAppendEntries* pMsg = NULL;
+
+    SSyncRaftEntry* pEntry;
+    int32_t         code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, nextIndex, &pEntry);
+    ASSERT(code == 0);
+
+    if (pEntry != NULL) {
+      pMsg = syncAppendEntriesBuild(pEntry->bytes, pSyncNode->vgId);
+      ASSERT(pMsg != NULL);
+
+      // add pEntry into msg
+      uint32_t len;
+      char*    serialized = syncEntrySerialize(pEntry, &len);
+      assert(len == pEntry->bytes);
+      memcpy(pMsg->data, serialized, len);
+
+      taosMemoryFree(serialized);
+      syncEntryDestory(pEntry);
+
+    } else {
+      // no entry in log
+      pMsg = syncAppendEntriesBuild(0, pSyncNode->vgId);
+      ASSERT(pMsg != NULL);
+    }
+
+    // prepare msg
+    ASSERT(pMsg != NULL);
+    pMsg->srcId = pSyncNode->myRaftId;
+    pMsg->destId = *pDestId;
+    pMsg->term = pSyncNode->pRaftStore->currentTerm;
+    pMsg->prevLogIndex = preLogIndex;
+    pMsg->prevLogTerm = preLogTerm;
+    pMsg->commitIndex = pSyncNode->commitIndex;
+    pMsg->privateTerm = 0;
+    // pMsg->privateTerm = syncIndexMgrGetTerm(pSyncNode->pNextIndex, pDestId);
+
+    // send msg
+    syncNodeAppendEntries(pSyncNode, pDestId, pMsg);
+    syncAppendEntriesDestroy(pMsg);
+  }
+
+  return ret;
+}
+
 int32_t syncNodeReplicate(SSyncNode* pSyncNode) {
   // start replicate
-  int32_t ret = syncNodeAppendEntriesPeers(pSyncNode);
+  int32_t ret = 0;
+
+  if (pSyncNode->pRaftCfg->snapshotEnable) {
+    ret = syncNodeAppendEntriesPeersSnapshot(pSyncNode);
+  } else {
+    ret = syncNodeAppendEntriesPeers(pSyncNode);
+  }
   return ret;
 }
 

source/libs/sync/src/syncRequestVote.c

@@ -15,6 +15,7 @@
 
 #include "syncRequestVote.h"
 #include "syncInt.h"
+#include "syncRaftCfg.h"
 #include "syncRaftStore.h"
 #include "syncUtil.h"
 #include "syncVoteMgr.h"
@@ -62,6 +63,9 @@ int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg) {
     // maybe has already voted for pMsg->srcId
     // vote again, no harm
     raftStoreVote(ths->pRaftStore, &(pMsg->srcId));
+
+    // forbid elect for this round
+    syncNodeResetElectTimer(ths);
   }
 
   SyncRequestVoteReply* pReply = syncRequestVoteReplyBuild(ths->vgId);
@@ -77,3 +81,64 @@ int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg) {
 
   return ret;
 }
+
+static bool syncNodeOnRequestVoteLogOK(SSyncNode* pSyncNode, SyncRequestVote* pMsg) {
+  SyncTerm  myLastTerm = syncNodeGetLastTerm(pSyncNode);
+  SyncIndex myLastIndex = syncNodeGetLastIndex(pSyncNode);
+
+  if (pMsg->lastLogTerm > myLastTerm) {
+    return true;
+  }
+  if (pMsg->lastLogTerm == myLastTerm && pMsg->lastLogIndex >= myLastIndex) {
+    return true;
+  }
+
+  return false;
+}
+
+int32_t syncNodeOnRequestVoteSnapshotCb(SSyncNode* ths, SyncRequestVote* pMsg) {
+  int32_t ret = 0;
+
+  // print log
+  char logBuf[128] = {0};
+  snprintf(logBuf, sizeof(logBuf), "recv SyncRequestVote, currentTerm:%lu", ths->pRaftStore->currentTerm);
+  syncRequestVoteLog2(logBuf, pMsg);
+
+  // if already drop replica, do not process
+  if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
+    sInfo("recv SyncRequestVote maybe replica already dropped");
+    return ret;
+  }
+
+  // maybe update term
+  if (pMsg->term > ths->pRaftStore->currentTerm) {
+    syncNodeUpdateTerm(ths, pMsg->term);
+  }
+  ASSERT(pMsg->term <= ths->pRaftStore->currentTerm);
+
+  bool logOK = syncNodeOnRequestVoteLogOK(ths, pMsg);
+  bool grant = (pMsg->term == ths->pRaftStore->currentTerm) && logOK &&
+               ((!raftStoreHasVoted(ths->pRaftStore)) || (syncUtilSameId(&(ths->pRaftStore->voteFor), &(pMsg->srcId))));
+  if (grant) {
+    // maybe has already voted for pMsg->srcId
+    // vote again, no harm
+    raftStoreVote(ths->pRaftStore, &(pMsg->srcId));
+
+    // forbid elect for this round
+    syncNodeResetElectTimer(ths);
+  }
+
+  // send msg
+  SyncRequestVoteReply* pReply = syncRequestVoteReplyBuild(ths->vgId);
+  pReply->srcId = ths->myRaftId;
+  pReply->destId = pMsg->srcId;
+  pReply->term = ths->pRaftStore->currentTerm;
+  pReply->voteGranted = grant;
+
+  SRpcMsg rpcMsg;
+  syncRequestVoteReply2RpcMsg(pReply, &rpcMsg);
+  syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
+  syncRequestVoteReplyDestroy(pReply);
+
+  return ret;
+}
\ No newline at end of file

source/libs/sync/src/syncRequestVoteReply.c

@@ -15,6 +15,7 @@
 
 #include "syncRequestVoteReply.h"
 #include "syncInt.h"
+#include "syncRaftCfg.h"
 #include "syncRaftStore.h"
 #include "syncUtil.h"
 #include "syncVoteMgr.h"
@@ -92,3 +93,68 @@ int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg)
 
   return ret;
 }
+
+int32_t syncNodeOnRequestVoteReplySnapshotCb(SSyncNode* ths, SyncRequestVoteReply* pMsg) {
+  int32_t ret = 0;
+
+  // print log
+  char logBuf[128] = {0};
+  snprintf(logBuf, sizeof(logBuf), "recv SyncRequestVoteReply, term:%lu", ths->pRaftStore->currentTerm);
+  syncRequestVoteReplyLog2(logBuf, pMsg);
+
+  // if already drop replica, do not process
+  if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
+    sInfo("recv SyncRequestVoteReply, maybe replica already dropped");
+    return ret;
+  }
+
+  // drop stale response
+  if (pMsg->term < ths->pRaftStore->currentTerm) {
+    sTrace("recv SyncRequestVoteReply, drop stale response, receive_term:%lu current_term:%lu", pMsg->term,
+           ths->pRaftStore->currentTerm);
+    return ret;
+  }
+
+  // assert(!(pMsg->term > ths->pRaftStore->currentTerm));
+  //  no need this code, because if I receive reply.term, then I must have sent for that term.
+  //   if (pMsg->term > ths->pRaftStore->currentTerm) {
+  //     syncNodeUpdateTerm(ths, pMsg->term);
+  //   }
+
+  if (pMsg->term > ths->pRaftStore->currentTerm) {
+    char logBuf[128] = {0};
+    snprintf(logBuf, sizeof(logBuf), "recv SyncRequestVoteReply, error term, receive_term:%lu current_term:%lu",
+             pMsg->term, ths->pRaftStore->currentTerm);
+    syncNodePrint2(logBuf, ths);
+    sError("%s", logBuf);
+    return ret;
+  }
+
+  ASSERT(pMsg->term == ths->pRaftStore->currentTerm);
+
+  // This tallies votes even when the current state is not Candidate,
+  // but they won't be looked at, so it doesn't matter.
+  if (ths->state == TAOS_SYNC_STATE_CANDIDATE) {
+    votesRespondAdd(ths->pVotesRespond, pMsg);
+    if (pMsg->voteGranted) {
+      // add vote
+      voteGrantedVote(ths->pVotesGranted, pMsg);
+
+      // maybe to leader
+      if (voteGrantedMajority(ths->pVotesGranted)) {
+        if (!ths->pVotesGranted->toLeader) {
+          syncNodeCandidate2Leader(ths);
+
+          // prevent to leader again!
+          ths->pVotesGranted->toLeader = true;
+        }
+      }
+    } else {
+      ;
+      // do nothing
+      // UNCHANGED <<votesGranted, voterLog>>
+    }
+  }
+
+  return ret;
+}
\ No newline at end of file

source/libs/sync/src/syncUtil.c

@@ -240,4 +240,26 @@ bool syncUtilUserRollback(tmsg_t msgType) {
     return true;
   }
   return false;
+}
+
+void syncUtilJson2Line(char* jsonStr) {
+  int p, q, len;
+  p = 0;
+  q = 1;
+  len = strlen(jsonStr);
+  while (1) {
+    if (jsonStr[q] == '\0') {
+      jsonStr[p + 1] = '\0';
+      break;
+    }
+
+    if (jsonStr[q] == '\n' || jsonStr[q] == ' ' || jsonStr[q] == '\t') {
+      q++;
+      continue;
+    } else {
+      jsonStr[p + 1] = jsonStr[q];
+      p++;
+      q++;
+    }
+  }
 }
\ No newline at end of file

source/libs/sync/test/syncAppendEntriesReplyTest.cpp

@@ -22,6 +22,8 @@ SyncAppendEntriesReply *createMsg() {
   pMsg->destId.vgId = 100;
   pMsg->success = true;
   pMsg->matchIndex = 77;
+  pMsg->term = 33;
+  pMsg->privateTerm = 44;
   return pMsg;
 }
 

source/libs/sync/test/syncAppendEntriesTest.cpp

@@ -23,6 +23,7 @@ SyncAppendEntries *createMsg() {
   pMsg->prevLogIndex = 11;
   pMsg->prevLogTerm = 22;
   pMsg->commitIndex = 33;
+  pMsg->privateTerm = 44;
   strcpy(pMsg->data, "hello world");
   return pMsg;
 }

source/libs/sync/test/syncConfigChangeTest.cpp

@@ -93,7 +93,6 @@ SSyncFSM* createFsm() {
   pFsm->FpGetSnapshot = GetSnapshotCb;
   pFsm->FpRestoreFinishCb = RestoreFinishCb;
 
-
   pFsm->FpReConfigCb = ReConfigCb;
 
   return pFsm;

source/libs/wal/inc/walInt.h

@@ -132,6 +132,7 @@ static inline void walResetVer(SWalVer* pVer) {
 
 int walLoadMeta(SWal* pWal);
 int walSaveMeta(SWal* pWal);
+int walRemoveMeta(SWal* pWal);
 int walRollFileInfo(SWal* pWal);
 
 int walCheckAndRepairMeta(SWal* pWal);