Merge branch '3.0' into feature/tq

include/common/taosmsg.h

@@ -706,41 +706,31 @@ typedef struct {
 } SStatusRsp;
 
 typedef struct {
-  uint32_t vgId;
-  int32_t  dbCfgVersion;
-  int32_t  maxTables;
-  int32_t  cacheBlockSize;
-  int32_t  totalBlocks;
-  int32_t  daysPerFile;
-  int32_t  daysToKeep;
-  int32_t  daysToKeep1;
-  int32_t  daysToKeep2;
-  int32_t  minRowsPerFileBlock;
-  int32_t  maxRowsPerFileBlock;
-  int32_t  commitTime;
-  int32_t  fsyncPeriod;
-  int8_t   precision;
-  int8_t   compression;
-  int8_t   walLevel;
-  int8_t   vgReplica;
-  int8_t   wals;
-  int8_t   quorum;
-  int8_t   update;
-  int8_t   cacheLastRow;
-  int32_t  vgCfgVersion;
-  int8_t   dbReplica;
-  int8_t   dbType;
-  int8_t   reserved[8];
-} SVnodeCfg;
-
-typedef struct {
-  int32_t  nodeId;
-  char     nodeEp[TSDB_EP_LEN];
+  uint16_t port;
+  char     fqdn[TSDB_FQDN_LEN];
 } SVnodeDesc;
 
 typedef struct {
   char       db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN];
-  SVnodeCfg  cfg;
+  uint32_t   vgId;
+  int32_t    cacheBlockSize;
+  int32_t    totalBlocks;
+  int32_t    daysPerFile;
+  int32_t    daysToKeep0;
+  int32_t    daysToKeep1;
+  int32_t    daysToKeep2;
+  int32_t    minRowsPerFileBlock;
+  int32_t    maxRowsPerFileBlock;
+  int32_t    fsyncPeriod;
+  int8_t     reserved[16];
+  int8_t     precision;
+  int8_t     compression;
+  int8_t     cacheLastRow;
+  int8_t     update;
+  int8_t     walLevel;
+  int8_t     replica;
+  int8_t     quorum;
+  int8_t     selfIndex;
   SVnodeDesc nodes[TSDB_MAX_REPLICA];
 } SCreateVnodeMsg, SAlterVnodeMsg;
 

include/common/tdataformat.h

@@ -308,7 +308,7 @@ static FORCE_INLINE void tdCopyColOfRowBySchema(SDataRow dst, STSchema *pDstSche
       SET_DOUBLE_PTR(pData, value);
       break;
     case TSDB_DATA_TYPE_TIMESTAMP:
-      if (pSrcSchema->columns[srcIdx].colId == PRIMARYKEY_TIMESTAMP_COL_INDEX) {
+      if (pSrcSchema->columns[srcIdx].colId == PRIMARYKEY_TIMESTAMP_COL_ID) {
         *(TSKEY *)pData = tdGetKey(*(TKEY *)value);
       } else {
         *(TSKEY *)pData = *(TSKEY *)value;

include/common/ttime.h

@@ -23,6 +23,8 @@ extern "C" {
 #include "taosdef.h"
 #include "taosmsg.h"
 
+#define TIME_IS_VAR_DURATION(_t) ((_t) == 'n' || (_t) == 'y' || (_t) == 'N' || (_t) == 'Y')
+
 /*
  * @return timestamp decided by global conf variable, tsTimePrecision
  * if precision == TSDB_TIME_PRECISION_MICRO, it returns timestamp in microsecond.
@@ -50,7 +52,6 @@ void    deltaToUtcInitOnce();
 
 int64_t convertTimePrecision(int64_t time, int32_t fromPrecision, int32_t toPrecision);
 
-
 #ifdef __cplusplus
 }
 #endif

include/libs/function/function.h

@@ -13,8 +13,8 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef TDENGINE_TFUNCTION_H
-#define TDENGINE_TFUNCTION_H
+#ifndef TDENGINE_FUNCTION_H
+#define TDENGINE_FUNCTION_H
 
 #ifdef __cplusplus
 extern "C" {
@@ -24,6 +24,8 @@ extern "C" {
 #include "tvariant.h"
 #include "tbuffer.h"
 
+#define MAX_INTERVAL_TIME_WINDOW 1000000  // maximum allowed time windows in final results
+
 #define FUNCTION_SCALAR       1
 #define FUNCTION_AGG          2
 
@@ -184,6 +186,25 @@ typedef struct SResultDataInfo {
   int32_t intermediateBytes;
 } SResultDataInfo;
 
+
+typedef struct SMultiFunctionsDesc {
+  bool stableQuery;
+  bool groupbyColumn;
+  bool simpleAgg;
+  bool arithmeticOnAgg;
+  bool projectionQuery;
+  bool hasFilter;
+  bool onlyTagQuery;
+  bool orderProjectQuery;
+  bool stateWindow;
+  bool globalMerge;
+  bool multigroupResult;
+  bool blockDistribution;
+  bool timewindow;
+  bool topbotQuery;
+  bool interpQuery;
+} SMultiFunctionsDesc;
+
 int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionId, int32_t param, SResultDataInfo* pInfo, int16_t extLength,
                           bool isSuperTable);
 
@@ -199,8 +220,10 @@ bool qIsValidUdf(SArray* pUdfInfo, const char* name, int32_t len, int32_t* funct
 
 const char* qGetFunctionName(int32_t functionId);
 
+void extractFunctionDesc(SArray* pFunctionIdList, SMultiFunctionsDesc* pDesc);
+
 #ifdef __cplusplus
 }
 #endif
 
-#endif  // TDENGINE_TFUNCTION_H
+#endif  // TDENGINE_FUNCTION_H

include/libs/parser/parser.h

@@ -24,6 +24,7 @@ extern "C" {
 #include "common.h"
 #include "tname.h"
 #include "tvariant.h"
+#include "function.h"
 
 typedef struct SColumn {
   uint64_t     tableUid;
@@ -130,20 +131,6 @@ typedef struct STableMetaInfo {
   SArray       *tagColList;                        // SArray<SColumn*>, involved tag columns
 } STableMetaInfo;
 
-typedef struct SQueryAttrInfo {
-  bool             stableQuery;
-  bool             groupbyColumn;
-  bool             simpleAgg;
-  bool             arithmeticOnAgg;
-  bool             projectionQuery;
-  bool             hasFilter;
-  bool             onlyTagQuery;
-  bool             orderProjectQuery;
-  bool             stateWindow;
-  bool             globalMerge;
-  bool             multigroupResult;
-} SQueryAttrInfo;
-
 typedef struct SQueryStmtInfo {
   int16_t          command;       // the command may be different for each subclause, so keep it seperately.
   uint32_t         type;          // query/insert type
@@ -177,7 +164,6 @@ typedef struct SQueryStmtInfo {
 
   int32_t          udColumnId;    // current user-defined constant output field column id, monotonically decreases from TSDB_UD_COLUMN_INDEX
   bool             distinct;   // distinct tag or not
-  bool             onlyHasTagCond;
   int32_t          bufLen;
   char*            buf;
   SArray          *pUdfInfo;
@@ -186,7 +172,7 @@ typedef struct SQueryStmtInfo {
   SArray            *pUpstream;   // SArray<struct SQueryStmtInfo>
   struct SQueryStmtInfo *pDownstream;
   int32_t            havingFieldNum;
-  SQueryAttrInfo     info;
+  SMultiFunctionsDesc     info;
 } SQueryStmtInfo;
 
 typedef struct SColumnIndex {

include/libs/sync/sync.h

@@ -22,11 +22,10 @@ extern "C" {
 
 #include <stdint.h>
 #include "taosdef.h"
-#include "wal.h"
 
-typedef int64_t SyncNodeId;
-typedef int32_t SyncGroupId;
-typedef int64_t   SyncIndex;
+typedef int32_t  SyncNodeId;
+typedef int32_t  SyncGroupId;
+typedef int64_t  SyncIndex;
 typedef uint64_t SSyncTerm;
 
 typedef enum {
@@ -41,41 +40,42 @@ typedef struct {
 } SSyncBuffer;
 
 typedef struct {
-  uint16_t  nodePort;  // node sync Port
-  char      nodeFqdn[TSDB_FQDN_LEN]; // node FQDN  
+  SyncNodeId nodeId;
+  uint16_t   nodePort;                 // node sync Port
+  char       nodeFqdn[TSDB_FQDN_LEN];  // node FQDN
 } SNodeInfo;
 
 typedef struct {
-  int        selfIndex;
-  int        nNode;
-  SNodeInfo* nodeInfo;
+  int32_t   selfIndex;
+  int32_t   replica;
+  SNodeInfo nodeInfo[TSDB_MAX_REPLICA];
 } SSyncCluster;
 
 typedef struct {
-  int32_t  selfIndex;
-  int nNode;
-  SNodeInfo* node;
-  ESyncRole*  role;
+  int32_t   selfIndex;
+  int32_t   replica;
+  SNodeInfo node[TSDB_MAX_REPLICA];
+  ESyncRole role[TSDB_MAX_REPLICA];
 } SNodesRole;
 
 typedef struct SSyncFSM {
   void* pData;
 
   // apply committed log, bufs will be free by raft module
-  int (*applyLog)(struct SSyncFSM* fsm, SyncIndex index, const SSyncBuffer* buf, void* pData);
+  int32_t (*applyLog)(struct SSyncFSM* fsm, SyncIndex index, const SSyncBuffer* buf, void* pData);
 
   // cluster commit callback
-  int (*onClusterChanged)(struct SSyncFSM* fsm, const SSyncCluster* cluster, void* pData);
+  int32_t (*onClusterChanged)(struct SSyncFSM* fsm, const SSyncCluster* cluster, void* pData);
 
   // fsm return snapshot in ppBuf, bufs will be free by raft module
   // TODO: getSnapshot SHOULD be async?
-  int (*getSnapshot)(struct SSyncFSM* fsm, SSyncBuffer** ppBuf, int* objId, bool* isLast);
+  int32_t (*getSnapshot)(struct SSyncFSM* fsm, SSyncBuffer** ppBuf, int32_t* objId, bool* isLast);
 
   // fsm apply snapshot with pBuf data
-  int (*applySnapshot)(struct SSyncFSM* fsm, SSyncBuffer* pBuf, int objId, bool isLast);
+  int32_t (*applySnapshot)(struct SSyncFSM* fsm, SSyncBuffer* pBuf, int32_t objId, bool isLast);
 
   // call when restore snapshot and log done
-  int (*onRestoreDone)(struct SSyncFSM* fsm);
+  int32_t (*onRestoreDone)(struct SSyncFSM* fsm);
 
   void (*onRollback)(struct SSyncFSM* fsm, SyncIndex index, const SSyncBuffer* buf);
 
@@ -83,52 +83,79 @@ typedef struct SSyncFSM {
 
 } SSyncFSM;
 
+typedef struct SSyncLogStore {
+  void* pData;
+
+  // write log with given index
+  int32_t (*logWrite)(struct SSyncLogStore* logStore, SyncIndex index, SSyncBuffer* pBuf);
+
+  // mark log with given index has been commtted
+  int32_t (*logCommit)(struct SSyncLogStore* logStore, SyncIndex index);
+
+  // prune log before given index
+  int32_t (*logPrune)(struct SSyncLogStore* logStore, SyncIndex index);
+
+  // rollback log after given index
+  int32_t (*logRollback)(struct SSyncLogStore* logStore, SyncIndex index);
+} SSyncLogStore;
+
 typedef struct SSyncServerState {
-  SNodeInfo voteFor;
-  SSyncTerm term;
+  SyncNodeId voteFor;
+  SSyncTerm  term;
 } SSyncServerState;
 
+typedef struct SSyncClusterConfig {
+  // Log index number of current cluster config.
+  SyncIndex index;
+
+  // Log index number of previous cluster config.
+  SyncIndex prevIndex;
+
+  // current cluster
+  const SSyncCluster* cluster;
+} SSyncClusterConfig;
+
 typedef struct SStateManager {
   void* pData;
 
-  void (*saveServerState)(struct SStateManager* stateMng, const SSyncServerState* state);
+  int32_t (*saveServerState)(struct SStateManager* stateMng, SSyncServerState* state);
 
-  const SSyncServerState* (*readServerState)(struct SStateManager* stateMng);
+  int32_t (*readServerState)(struct SStateManager* stateMng, SSyncServerState* state);
 
-  void (*saveCluster)(struct SStateManager* stateMng, const SSyncCluster* cluster);
+  // void (*saveCluster)(struct SStateManager* stateMng, const SSyncClusterConfig* cluster);
 
-  const SSyncCluster* (*readCluster)(struct SStateManager* stateMng);
+  // const SSyncClusterConfig* (*readCluster)(struct SStateManager* stateMng);
 } SStateManager;
 
 typedef struct {
-  SyncGroupId vgId;
-
-  twalh walHandle;
-
-  SyncIndex snapshotIndex;
-  SSyncCluster syncCfg;
-
-  SSyncFSM fsm;
-
+  SyncGroupId   vgId;
+  SyncIndex     snapshotIndex;
+  SSyncCluster  syncCfg;
+  SSyncFSM      fsm;
+  SSyncLogStore logStore;
   SStateManager stateManager;
 } SSyncInfo;
 
+struct SSyncNode;
+typedef struct SSyncNode SSyncNode;
+
 int32_t syncInit();
 void    syncCleanUp();
 
-SyncNodeId syncStart(const SSyncInfo*);
-void       syncStop(SyncNodeId);
+SSyncNode* syncStart(const SSyncInfo*);
+void       syncReconfig(const SSyncNode*, const SSyncCluster*);
+void       syncStop(const SSyncNode*);
 
-int32_t syncPropose(SyncNodeId nodeId, SSyncBuffer buffer, void* pData, bool isWeak);
+int32_t syncPropose(SSyncNode* syncNode, SSyncBuffer buffer, void* pData, bool isWeak);
 
-int32_t syncAddNode(SyncNodeId nodeId, const SNodeInfo *pNode);
+// int32_t syncAddNode(SSyncNode syncNode, const SNodeInfo *pNode);
 
-int32_t syncRemoveNode(SyncNodeId nodeId, const SNodeInfo *pNode);
+// int32_t syncRemoveNode(SSyncNode syncNode, const SNodeInfo *pNode);
 
-extern int32_t  syncDebugFlag;
+extern int32_t syncDebugFlag;
 
 #ifdef __cplusplus
 }
 #endif
 
-#endif  /*_TD_LIBS_SYNC_H*/
+#endif /*_TD_LIBS_SYNC_H*/

include/libs/wal/wal.h

@@ -44,41 +44,41 @@ typedef struct {
   EWalType walLevel;     // wal level
 } SWalCfg;
 
-typedef void *  twalh;  // WAL HANDLE
-typedef int32_t FWalWrite(void *ahandle, void *pHead, int32_t qtype, void *pMsg);
+struct SWal;
+typedef struct SWal SWal;  // WAL HANDLE
+typedef int32_t (*FWalWrite)(void *ahandle, void *pHead, int32_t qtype, void *pMsg);
 
-//module initialization
-int32_t  walInit();
-void     walCleanUp();
+// module initialization
+int32_t walInit();
+void    walCleanUp();
 
-//handle open and ctl
-twalh    walOpen(char *path, SWalCfg *pCfg);
-int32_t  walAlter(twalh, SWalCfg *pCfg);
-void     walStop(twalh);
-void     walClose(twalh);
+// handle open and ctl
+SWal   *walOpen(char *path, SWalCfg *pCfg);
+int32_t walAlter(SWal *, SWalCfg *pCfg);
+void    walClose(SWal *);
 
-//write
-//int64_t  walWriteWithMsgType(twalh, int8_t msgType, void* body, int32_t bodyLen);
-int64_t  walWrite(twalh, void* body, int32_t bodyLen);
-int64_t  walWriteBatch(twalh, void** bodies, int32_t* bodyLen, int32_t batchSize);
+// write
+// int64_t  walWriteWithMsgType(SWal*, int8_t msgType, void* body, int32_t bodyLen);
+int64_t walWrite(SWal *, int64_t index, void *body, int32_t bodyLen);
+int64_t walWriteBatch(SWal *, void **bodies, int32_t *bodyLen, int32_t batchSize);
 
-//apis for lifecycle management
-void     walFsync(twalh, bool force);
-int32_t  walCommit(twalh, int64_t ver);
-//truncate after
-int32_t  walRollback(twalh, int64_t ver);
-//notify that previous log can be pruned safely
-int32_t  walPrune(twalh, int64_t ver);
+// apis for lifecycle management
+void    walFsync(SWal *, bool force);
+int32_t walCommit(SWal *, int64_t ver);
+// truncate after
+int32_t walRollback(SWal *, int64_t ver);
+// notify that previous log can be pruned safely
+int32_t walPrune(SWal *, int64_t ver);
 
-//read
-int32_t  walRead(twalh, SWalHead **, int64_t ver);
-int32_t  walReadWithFp(twalh, FWalWrite writeFp, int64_t verStart, int readNum);
+// read
+int32_t walRead(SWal *, SWalHead **, int64_t ver);
+int32_t walReadWithFp(SWal *, FWalWrite writeFp, int64_t verStart, int32_t readNum);
 
-//lifecycle check
-int32_t  walFirstVer(twalh);
-int32_t  walPersistedVer(twalh);
-int32_t  walLastVer(twalh);
-//int32_t  walDataCorrupted(twalh);
+// lifecycle check
+int32_t walFirstVer(SWal *);
+int32_t walPersistedVer(SWal *);
+int32_t walLastVer(SWal *);
+// int32_t  walDataCorrupted(SWal*);
 
 #ifdef __cplusplus
 }

include/server/dnode/dnode.h

@@ -67,6 +67,11 @@ void dnodeSendRedirectMsg(struct SRpcMsg *rpcMsg, bool forShell);
  */
 void dnodeGetEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port);
 
+/**
+ * Report the startup progress.
+ */
+void dnodeReportStartup(char *name, char *desc);
+
 #ifdef __cplusplus
 }
 #endif

include/server/vnode/vnode.h

@@ -46,6 +46,11 @@ typedef struct {
    */
   void (*GetDnodeEp)(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port);
 
+  /**
+   * Report the startup progress.
+   */
+  void (*ReportStartup)(char *name, char *desc);
+
 } SVnodeFp;
 
 typedef struct {

include/util/taoserror.h

@@ -233,11 +233,11 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_VND_NO_SUCH_FILE_OR_DIR       TAOS_DEF_ERROR_CODE(0, 0x0507)  //"Missing data file")
 #define TSDB_CODE_VND_OUT_OF_MEMORY             TAOS_DEF_ERROR_CODE(0, 0x0508)  //"Out of memory")
 #define TSDB_CODE_VND_APP_ERROR                 TAOS_DEF_ERROR_CODE(0, 0x0509)  //"Unexpected generic error in vnode")
-#define TSDB_CODE_VND_INVALID_VRESION_FILE      TAOS_DEF_ERROR_CODE(0, 0x050A)  //"Invalid version file")
-#define TSDB_CODE_VND_IS_FULL                   TAOS_DEF_ERROR_CODE(0, 0x050B)  //"Database memory is full for commit failed")
-#define TSDB_CODE_VND_IS_FLOWCTRL               TAOS_DEF_ERROR_CODE(0, 0x050C)  //"Database memory is full for waiting commit")
+#define TSDB_CODE_VND_INVALID_CFG_FILE          TAOS_DEF_ERROR_CODE(0, 0x050A)  //"Invalid config file)
+#define TSDB_CODE_VND_INVALID_TERM_FILE         TAOS_DEF_ERROR_CODE(0, 0x050B)  //"Invalid term file")
+#define TSDB_CODE_VND_IS_FLOWCTRL               TAOS_DEF_ERROR_CODE(0, 0x050C)  //"Database memory is full")
 #define TSDB_CODE_VND_IS_DROPPING               TAOS_DEF_ERROR_CODE(0, 0x050D)  //"Database is dropping")
-#define TSDB_CODE_VND_IS_BALANCING              TAOS_DEF_ERROR_CODE(0, 0x050E)  //"Database is balancing")
+#define TSDB_CODE_VND_IS_UPDATING               TAOS_DEF_ERROR_CODE(0, 0x050E)  //"Database is updating")
 #define TSDB_CODE_VND_IS_CLOSING                TAOS_DEF_ERROR_CODE(0, 0x0510)  //"Database is closing")
 #define TSDB_CODE_VND_NOT_SYNCED                TAOS_DEF_ERROR_CODE(0, 0x0511)  //"Database suspended")
 #define TSDB_CODE_VND_NO_WRITE_AUTH             TAOS_DEF_ERROR_CODE(0, 0x0512)  //"Database write operation denied")

include/util/tdef.h

@@ -303,7 +303,7 @@ do { \
 #define TSDB_MAX_FIELD_LEN              16384
 #define TSDB_MAX_BINARY_LEN            (TSDB_MAX_FIELD_LEN-TSDB_KEYSIZE) // keep 16384
 #define TSDB_MAX_NCHAR_LEN             (TSDB_MAX_FIELD_LEN-TSDB_KEYSIZE) // keep 16384
-#define PRIMARYKEY_TIMESTAMP_COL_INDEX  0
+#define PRIMARYKEY_TIMESTAMP_COL_ID     0
 
 #define TSDB_MAX_RPC_THREADS            5
 
@@ -382,6 +382,9 @@ do { \
 #define TSDB_DATA_TYPE_UINT       13    // 4 bytes
 #define TSDB_DATA_TYPE_UBIGINT    14    // 8 bytes
 
+enum { TRANS_STAT_INIT = 0, TRANS_STAT_EXECUTING, TRANS_STAT_EXECUTED, TRANS_STAT_ROLLBACKING, TRANS_STAT_ROLLBACKED };
+enum { TRANS_OPER_INIT = 0, TRANS_OPER_EXECUTE, TRANS_OPER_ROLLBACK };
+
 #ifdef __cplusplus
 }
 #endif

source/common/src/ttime.c

@@ -48,7 +48,7 @@
  * An encoding of midnight at the end of the day as 24:00:00 - ie. midnight
  * tomorrow - (allowable under ISO 8601) is supported.
  */
-int64_t user_mktime64(const unsigned int year0, const unsigned int mon0,
+static int64_t user_mktime64(const unsigned int year0, const unsigned int mon0,
 		const unsigned int day, const unsigned int hour,
 		const unsigned int min, const unsigned int sec, int64_t time_zone)
 {
@@ -79,19 +79,18 @@ void deltaToUtcInitOnce() {
   (void)strptime("1970-01-01 00:00:00", (const char *)("%Y-%m-%d %H:%M:%S"), &tm);
   m_deltaUtc = (int64_t)mktime(&tm);
   //printf("====delta:%lld\n\n", seconds);
-  return;
 }
 
 static int64_t parseFraction(char* str, char** end, int32_t timePrec);
 static int32_t parseTimeWithTz(char* timestr, int64_t* time, int32_t timePrec, char delim);
 static int32_t parseLocaltime(char* timestr, int64_t* time, int32_t timePrec);
-static int32_t parseLocaltimeWithDst(char* timestr, int64_t* time, int32_t timePrec);
+static int32_t parseLocaltimeDst(char* timestr, int64_t* time, int32_t timePrec);
 static char* forwardToTimeStringEnd(char* str);
 static bool checkTzPresent(char *str, int32_t len);
 
 static int32_t (*parseLocaltimeFp[]) (char* timestr, int64_t* time, int32_t timePrec) = {
   parseLocaltime,
-  parseLocaltimeWithDst
+  parseLocaltimeDst
 };
 
 int32_t taosParseTime(char* timestr, int64_t* time, int32_t len, int32_t timePrec, int8_t day_light) {
@@ -116,8 +115,8 @@ bool checkTzPresent(char *str, int32_t len) {
     }
     c--;
   }
-  return false;
 
+  return false;
 }
 
 char* forwardToTimeStringEnd(char* str) {
@@ -344,7 +343,7 @@ int32_t parseLocaltime(char* timestr, int64_t* time, int32_t timePrec) {
   return 0;
 }
 
-int32_t parseLocaltimeWithDst(char* timestr, int64_t* time, int32_t timePrec) {
+int32_t parseLocaltimeDst(char* timestr, int64_t* time, int32_t timePrec) {
   *time = 0;
   struct tm tm = {0};
   tm.tm_isdst = -1;

source/libs/function/CMakeLists.txt

@@ -7,6 +7,6 @@ target_include_directories(
 )
 
 target_link_libraries(
-        function
-        PRIVATE os util common
+    function
+    PRIVATE os util common
 )
\ No newline at end of file

source/libs/function/inc/taggfunction.h

@@ -56,8 +56,6 @@ typedef struct SResultRowCellInfo {
 #define QUERY_DESC_FORWARD_STEP -1
 
 #define GET_FORWARD_DIRECTION_FACTOR(ord) (((ord) == TSDB_ORDER_ASC) ? QUERY_ASC_FORWARD_STEP : QUERY_DESC_FORWARD_STEP)
-
-#define MAX_INTERVAL_TIME_WINDOW 1000000  // maximum allowed time windows in final results
 #define TOP_BOTTOM_QUERY_LIMIT   100
 
 enum {

source/libs/function/src/taggfunction.c

@@ -513,7 +513,7 @@ static void count_func_merge(SQLFunctionCtx *pCtx) {
  * @return
  */
 int32_t countRequired(SQLFunctionCtx *pCtx, STimeWindow* w, int32_t colId) {
-  if (colId == PRIMARYKEY_TIMESTAMP_COL_INDEX) {
+  if (colId == PRIMARYKEY_TIMESTAMP_COL_ID) {
     return BLK_DATA_NO_NEEDED;
   } else {
     return BLK_DATA_STATIS_NEEDED;
@@ -2303,10 +2303,10 @@ static void top_bottom_func_finalizer(SQLFunctionCtx *pCtx) {
   tValuePair **tvp = pRes->res;
   
   // user specify the order of output by sort the result according to timestamp
-  if (pCtx->param[1].i64 == PRIMARYKEY_TIMESTAMP_COL_INDEX) {
+  if (pCtx->param[1].i64 == PRIMARYKEY_TIMESTAMP_COL_ID) {
     __compar_fn_t comparator = (pCtx->param[2].i64 == TSDB_ORDER_ASC) ? resAscComparFn : resDescComparFn;
     qsort(tvp, (size_t)pResInfo->numOfRes, POINTER_BYTES, comparator);
-  } else /*if (pCtx->param[1].i64 > PRIMARYKEY_TIMESTAMP_COL_INDEX)*/ {
+  } else /*if (pCtx->param[1].i64 > PRIMARYKEY_TIMESTAMP_COL_ID)*/ {
     __compar_fn_t comparator = (pCtx->param[2].i64 == TSDB_ORDER_ASC) ? resDataAscComparFn : resDataDescComparFn;
     qsort(tvp, (size_t)pResInfo->numOfRes, POINTER_BYTES, comparator);
   }

source/libs/function/src/texpr.c

@@ -128,18 +128,26 @@ static void doExprTreeDestroy(tExprNode **pExpr, void (*fp)(void *)) {
   if (*pExpr == NULL) {
     return;
   }
-  
-  if ((*pExpr)->nodeType == TEXPR_BINARYEXPR_NODE) {
+
+  int32_t type = (*pExpr)->nodeType;
+  if (type == TEXPR_BINARYEXPR_NODE) {
     doExprTreeDestroy(&(*pExpr)->_node.pLeft, fp);
     doExprTreeDestroy(&(*pExpr)->_node.pRight, fp);
   
     if (fp != NULL) {
       fp((*pExpr)->_node.info);
     }
-  } else if ((*pExpr)->nodeType == TEXPR_VALUE_NODE) {
+  } else if (type == TEXPR_UNARYEXPR_NODE) {
+    doExprTreeDestroy(&(*pExpr)->_node.pLeft, fp);
+    if (fp != NULL) {
+      fp((*pExpr)->_node.info);
+    }
+
+    assert((*pExpr)->_node.pRight == NULL);
+  } else if (type == TEXPR_VALUE_NODE) {
     taosVariantDestroy((*pExpr)->pVal);
     free((*pExpr)->pVal);
-  } else if ((*pExpr)->nodeType == TEXPR_COL_NODE) {
+  } else if (type == TEXPR_COL_NODE) {
     free((*pExpr)->pSchema);
   }
 

source/libs/function/src/tfunction.c

@@ -52,7 +52,7 @@ bool isTagsQuery(SArray* pFunctionIdList) {
     int16_t f = *(int16_t*) taosArrayGet(pFunctionIdList, i);
 
     // "select count(tbname)" query
-//    if (functId == FUNCTION_COUNT && pExpr->base.colInfo.colId == TSDB_TBNAME_COLUMN_INDEX) {
+//    if (functId == FUNCTION_COUNT && pExpr->base.colpDesc->colId == TSDB_TBNAME_COLUMN_INDEX) {
 //      continue;
 //    }
 
@@ -80,19 +80,6 @@ bool isTagsQuery(SArray* pFunctionIdList) {
 //  return false;
 //}
 
-bool isBlockInfoQuery(SArray* pFunctionIdList) {
-  int32_t num = (int32_t) taosArrayGetSize(pFunctionIdList);
-  for (int32_t i = 0; i < num; ++i) {
-    int32_t f = *(int16_t*) taosArrayGet(pFunctionIdList, i);
-
-    if (f == FUNCTION_BLKINFO) {
-      return true;
-    }
-  }
-
-  return false;
-}
-
 bool isProjectionQuery(SArray* pFunctionIdList) {
   int32_t num = (int32_t) taosArrayGetSize(pFunctionIdList);
   for (int32_t i = 0; i < num; ++i) {
@@ -101,8 +88,12 @@ bool isProjectionQuery(SArray* pFunctionIdList) {
       continue;
     }
 
-    if (f != FUNCTION_PRJ && f != FUNCTION_TAGPRJ && f != FUNCTION_TAG &&
-        f != FUNCTION_TS && f != FUNCTION_ARITHM && f != FUNCTION_DIFF &&
+    if (f != FUNCTION_PRJ &&
+        f != FUNCTION_TAGPRJ &&
+        f != FUNCTION_TAG &&
+        f != FUNCTION_TS &&
+        f != FUNCTION_ARITHM &&
+        f != FUNCTION_DIFF &&
         f != FUNCTION_DERIVATIVE) {
       return false;
     }
@@ -111,7 +102,7 @@ bool isProjectionQuery(SArray* pFunctionIdList) {
   return true;
 }
 
-bool isDiffDerivQuery(SArray* pFunctionIdList) {
+bool isDiffDerivativeQuery(SArray* pFunctionIdList) {
   int32_t num = (int32_t) taosArrayGetSize(pFunctionIdList);
   for (int32_t i = 0; i < num; ++i) {
     int32_t f = *(int16_t*) taosArrayGet(pFunctionIdList, i);
@@ -127,7 +118,7 @@ bool isDiffDerivQuery(SArray* pFunctionIdList) {
   return false;
 }
 
-bool isPointInterpQuery(SArray* pFunctionIdList) {
+bool isInterpQuery(SArray* pFunctionIdList) {
   int32_t num = (int32_t) taosArrayGetSize(pFunctionIdList);
   for (int32_t i = 0; i < num; ++i) {
     int32_t f = *(int16_t*) taosArrayGet(pFunctionIdList, i);
@@ -264,8 +255,6 @@ bool needReverseScan(SArray* pFunctionIdList) {
 }
 
 bool isSimpleAggregateRv(SArray* pFunctionIdList) {
-  assert(0);
-
 //  if (pQueryInfo->interval.interval > 0 || pQueryInfo->sessionWindow.gap > 0) {
 //    return false;
 //  }
@@ -380,33 +369,17 @@ bool isProjectionQueryOnSTable(SArray* pFunctionIdList, int32_t tableIndex) {
 }
 
 bool hasTagValOutput(SArray* pFunctionIdList) {
-//  size_t    numOfExprs = getNumOfExprs(pQueryInfo);
-//  SExprInfo* pExpr1 = getExprInfo(pQueryInfo, 0);
-//
-//  if (numOfExprs == 1 && pExpr1->base.functionId == FUNCTION_TS_COMP) {
+  size_t size = taosArrayGetSize(pFunctionIdList);
+
+  //  if (numOfExprs == 1 && pExpr1->base.functionId == FUNCTION_TS_COMP) {
 //    return true;
-//  }
-//
-//  for (int32_t i = 0; i < numOfExprs; ++i) {
-//    SExprInfo* pExpr = getExprInfo(pQueryInfo, i);
-//    if (pExpr == NULL) {
-//      continue;
-//    }
-//
-//    // ts_comp column required the tag value for join filter
-//    if (TSDB_COL_IS_TAG(pExpr->base.colInfo.flag)) {
-//      return true;
-//    }
 //  }
 
-  return false;
-}
+  for (int32_t i = 0; i < size; ++i) {
+    int32_t functionId = *(int16_t*) taosArrayGet(pFunctionIdList, i);
 
-bool timeWindowInterpoRequired(SArray* pFunctionIdList) {
-  int32_t num = (int32_t) taosArrayGetSize(pFunctionIdList);
-  for (int32_t i = 0; i < num; ++i) {
-    int32_t f = *(int16_t*) taosArrayGet(pFunctionIdList, i);
-    if (f == FUNCTION_TWA || f == FUNCTION_INTERP) {
+    // ts_comp column required the tag value for join filter
+    if (functionId == FUNCTION_TAG || functionId == FUNCTION_TAGPRJ) {
       return true;
     }
   }
@@ -414,8 +387,28 @@ bool timeWindowInterpoRequired(SArray* pFunctionIdList) {
   return false;
 }
 
-//SQueryAttrInfo setQueryType(SArray* pFunctionIdList) {
-//  assert(pFunctionIdList != NULL);
-//
+//bool timeWindowInterpoRequired(SArray* pFunctionIdList) {
+//  int32_t num = (int32_t) taosArrayGetSize(pFunctionIdList);
+//  for (int32_t i = 0; i < num; ++i) {
+//    int32_t f = *(int16_t*) taosArrayGet(pFunctionIdList, i);
+//    if (f == FUNCTION_TWA || f == FUNCTION_INTERP) {
+//      return true;
+//    }
+//  }
 //
-//}
+//  return false;
\ No newline at end of file
+//}
+
+void extractFunctionDesc(SArray* pFunctionIdList, SMultiFunctionsDesc* pDesc) {
+  assert(pFunctionIdList != NULL);
+
+
+  pDesc->blockDistribution = isBlockDistQuery(pFunctionIdList);
+  if (pDesc->blockDistribution) {
+    return;
+  }
+
+  pDesc->projectionQuery = isProjectionQuery(pFunctionIdList);
+  pDesc->onlyTagQuery    = isTagsQuery(pFunctionIdList);
+  pDesc->interpQuery     = isInterpQuery(pFunctionIdList);
+}

source/libs/parser/inc/astGenerator.h

@@ -298,7 +298,7 @@ void* destroyCreateTableSql(SCreateTableSql* pCreate);
 void setDropFuncInfo(SSqlInfo *pInfo, int32_t type, SToken* pToken);
 void setCreateFuncInfo(SSqlInfo *pInfo, int32_t type, SToken *pName, SToken *pPath, SField *output, SToken* bufSize, int32_t funcType);
 
-void SqlInfoDestroy(SSqlInfo *pInfo);
+void destroySqlInfo(SSqlInfo *pInfo);
 
 void setDCLSqlElems(SSqlInfo *pInfo, int32_t type, int32_t nParams, ...);
 void setDropDbTableInfo(SSqlInfo *pInfo, int32_t type, SToken* pToken, SToken* existsCheck,int16_t dbType,int16_t tableType);

source/libs/parser/inc/parserInt.h

@@ -73,7 +73,9 @@ int32_t evaluateSqlNode(SSqlNode* pNode, int32_t tsPrecision, SMsgBuf* pMsgBuf);
 
 int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf* pMsgBuf);
 
-void initQueryInfo(SQueryStmtInfo* pQueryInfo);
+SQueryStmtInfo* createQueryInfo();
+
+void destroyQueryInfo(SQueryStmtInfo* pQueryInfo);
 
 int32_t checkForInvalidExpr(SQueryStmtInfo* pQueryInfo, SMsgBuf* pMsgBuf);
 
@@ -87,6 +89,12 @@ int32_t checkForInvalidExpr(SQueryStmtInfo* pQueryInfo, SMsgBuf* pMsgBuf);
  */
 int32_t qParserExtractRequestedMetaInfo(const SSqlInfo* pSqlInfo, SMetaReq* pMetaInfo, char* msg, int32_t msgBufLen);
 
+/**
+ * Destroy the meta data request structure.
+ * @param pMetaInfo
+ */
+void qParserClearupMetaRequestInfo(SMetaReq* pMetaInfo);
+
 #ifdef __cplusplus
 }
 #endif

source/libs/parser/inc/queryInfoUtil.h

@@ -30,9 +30,7 @@ SSchema *getTableTagSchema(const STableMeta* pTableMeta);
 SSchema *getOneColumnSchema(const STableMeta* pTableMeta, int32_t colIndex);
 
 size_t     getNumOfExprs(SQueryStmtInfo* pQueryInfo);
-//SExprInfo* createExprInfo(STableMetaInfo* pTableMetaInfo, int16_t functionId, SColumnIndex* pColIndex, struct tExprNode* pParamExpr, SSchema* pResSchema, int16_t interSize);
 SExprInfo* createBinaryExprInfo(struct tExprNode* pNode, SSchema* pResSchema);
-void       destroyExprInfoList();
 
 void       addExprInfo(SQueryStmtInfo* pQueryInfo, int32_t index, SExprInfo* pExprInfo);
 void       updateExprInfo(SExprInfo* pExprInfo, int16_t functionId, int32_t colId, int16_t srcColumnIndex, int16_t resType, int16_t resSize);
@@ -42,9 +40,11 @@ int32_t    copyAllExprInfo(SArray* dst, const SArray* src, bool deepcopy);
 
 void       addExprInfoParam(SSqlExpr* pExpr, char* argument, int32_t type, int32_t bytes);
 
+int32_t    getExprFunctionId(SExprInfo *pExprInfo);
 void       cleanupFieldInfo(SFieldInfo* pFieldInfo);
 
 STableComInfo getTableInfo(const STableMeta* pTableMeta);
+SArray* extractFunctionIdList(SArray* pExprInfoList);
 
 #ifdef __cplusplus
 }

source/libs/parser/src/astGenerator.c

@@ -242,7 +242,6 @@ tSqlExpr *tSqlExprClone(tSqlExpr *pSrc) {
 }
 
 void      tSqlExprCompact(tSqlExpr **pExpr) {
-
   if (*pExpr == NULL || tSqlExprIsParentOfLeaf(*pExpr)) {
     return;
   }
@@ -770,8 +769,11 @@ void setCreateFuncInfo(SSqlInfo *pInfo, int32_t type, SToken *pName, SToken *pPa
   }
 }
 
-void SqlInfoDestroy(SSqlInfo *pInfo) {
-  if (pInfo == NULL) return;;
+void destroySqlInfo(SSqlInfo *pInfo) {
+  if (pInfo == NULL) {
+    return;
+  }
+
   taosArrayDestroy(pInfo->funcs);
   if (pInfo->type == TSDB_SQL_SELECT) {
     destroyAllSqlNode(pInfo->list);

source/libs/parser/src/astValidate.c

@@ -14,6 +14,7 @@
  */
 
 #include <astGenerator.h>
+#include <function.h>
 #include "astGenerator.h"
 #include "function.h"
 #include "parserInt.h"
@@ -28,6 +29,8 @@
 #define DEFAULT_PRIMARY_TIMESTAMP_COL_NAME "_c0"
 #define VALID_COLUMN_INDEX(index) (((index).tableIndex >= 0) && ((index).columnIndex >= TSDB_TBNAME_COLUMN_INDEX))
 
+#define TSWINDOW_IS_EQUAL(t1, t2) (((t1).skey == (t2).skey) && ((t1).ekey == (t2).ekey))
+
 // -1 is tbname column index, so here use the -2 as the initial value
 #define COLUMN_INDEX_INITIAL_VAL (-2)
 #define COLUMN_INDEX_INITIALIZER { COLUMN_INDEX_INITIAL_VAL, COLUMN_INDEX_INITIAL_VAL }
@@ -213,7 +216,9 @@ static STableMeta* extractTempTableMetaFromSubquery(SQueryStmtInfo* pUpstream) {
   return meta;
 }
 
-void initQueryInfo(SQueryStmtInfo* pQueryInfo) {
+SQueryStmtInfo *createQueryInfo() {
+  SQueryStmtInfo* pQueryInfo = calloc(1, sizeof(SQueryStmtInfo));
+
   pQueryInfo->fieldsInfo.internalField = taosArrayInit(4, sizeof(SInternalField));
   pQueryInfo->exprList       = taosArrayInit(4, POINTER_BYTES);
   pQueryInfo->colList        = taosArrayInit(4, POINTER_BYTES);
@@ -225,6 +230,58 @@ void initQueryInfo(SQueryStmtInfo* pQueryInfo) {
   pQueryInfo->slimit.offset  = 0;
   pQueryInfo->pUpstream      = taosArrayInit(4, POINTER_BYTES);
   pQueryInfo->window         = TSWINDOW_INITIALIZER;
+
+  return pQueryInfo;
+}
+
+static void destroyQueryInfoImpl(SQueryStmtInfo* pQueryInfo) {
+  cleanupTagCond(&pQueryInfo->tagCond);
+  cleanupColumnCond(&pQueryInfo->colCond);
+  cleanupFieldInfo(&pQueryInfo->fieldsInfo);
+
+  dropAllExprInfo(pQueryInfo->exprList);
+  pQueryInfo->exprList = NULL;
+
+  if (pQueryInfo->exprList1 != NULL) {
+    dropAllExprInfo(pQueryInfo->exprList1);
+    pQueryInfo->exprList1 = NULL;
+  }
+
+  columnListDestroy(pQueryInfo->colList);
+  pQueryInfo->colList = NULL;
+
+  if (pQueryInfo->groupbyExpr.columnInfo != NULL) {
+    taosArrayDestroy(pQueryInfo->groupbyExpr.columnInfo);
+    pQueryInfo->groupbyExpr.columnInfo = NULL;
+  }
+
+  pQueryInfo->fillType = 0;
+
+  tfree(pQueryInfo->fillVal);
+  tfree(pQueryInfo->buf);
+
+  taosArrayDestroy(pQueryInfo->pUpstream);
+  pQueryInfo->pUpstream = NULL;
+  pQueryInfo->bufLen = 0;
+}
+
+void destroyQueryInfo(SQueryStmtInfo* pQueryInfo) {
+  while (pQueryInfo != NULL) {
+    SQueryStmtInfo* p = pQueryInfo->sibling;
+
+    size_t numOfUpstream = taosArrayGetSize(pQueryInfo->pUpstream);
+    for (int32_t i = 0; i < numOfUpstream; ++i) {
+      SQueryStmtInfo* pUpQueryInfo = taosArrayGetP(pQueryInfo->pUpstream, i);
+      destroyQueryInfoImpl(pUpQueryInfo);
+      clearAllTableMetaInfo(pUpQueryInfo, false, 0);
+      tfree(pUpQueryInfo);
+    }
+
+    destroyQueryInfoImpl(pQueryInfo);
+    clearAllTableMetaInfo(pQueryInfo, false, 0);
+    tfree(pQueryInfo);
+    pQueryInfo = p;
+  }
 }
 
 static int32_t doValidateSubquery(SSqlNode* pSqlNode, int32_t index, SQueryStmtInfo* pQueryInfo, SMsgBuf* pMsgBuf) {
@@ -236,8 +293,7 @@ static int32_t doValidateSubquery(SSqlNode* pSqlNode, int32_t index, SQueryStmtI
     return buildInvalidOperationMsg(pMsgBuf, "not support union in subquery");
   }
 
-  SQueryStmtInfo* pSub = calloc(1, sizeof(SQueryStmtInfo));
-  initQueryInfo(pSub);
+  SQueryStmtInfo* pSub = createQueryInfo();
 
   SArray *pUdfInfo = NULL;
   if (pQueryInfo->pUdfInfo) {
@@ -391,9 +447,9 @@ int32_t doGetColumnIndexByName(SToken* pToken, SQueryStmtInfo* pQueryInfo, SColu
     pIndex->type = TSDB_COL_TAG;
   } else if (strlen(DEFAULT_PRIMARY_TIMESTAMP_COL_NAME) == pToken->n &&
              strncasecmp(pToken->z, DEFAULT_PRIMARY_TIMESTAMP_COL_NAME, pToken->n) == 0) {
-    pIndex->columnIndex = PRIMARYKEY_TIMESTAMP_COL_INDEX; // just make runtime happy, need fix java test case InsertSpecialCharacterJniTest
+    pIndex->columnIndex = PRIMARYKEY_TIMESTAMP_COL_ID; // just make runtime happy, need fix java test case InsertSpecialCharacterJniTest
   } else if (pToken->n == 0) {
-    pIndex->columnIndex = PRIMARYKEY_TIMESTAMP_COL_INDEX; // just make runtime happy, need fix java test case InsertSpecialCharacterJniTest
+    pIndex->columnIndex = PRIMARYKEY_TIMESTAMP_COL_ID; // just make runtime happy, need fix java test case InsertSpecialCharacterJniTest
   } else {
     // not specify the table name, try to locate the table index by column name
     if (pIndex->tableIndex == COLUMN_INDEX_INITIAL_VAL) {
@@ -458,39 +514,32 @@ int32_t validateGroupbyNode(SQueryStmtInfo* pQueryInfo, SArray* pList, SMsgBuf*
     return buildInvalidOperationMsg(pMsgBuf, msg4);
   }
 
-  SGroupbyExpr* pGroupExpr = &pQueryInfo->groupbyExpr;
+  SGroupbyExpr* pGroupExpr = &(pQueryInfo->groupbyExpr);
+  pGroupExpr->columnInfo = taosArrayInit(4, sizeof(SColIndex));
   if (pGroupExpr->columnInfo == NULL) {
-    pGroupExpr->columnInfo = taosArrayInit(4, sizeof(SColIndex));
-  }
-
-  if (pQueryInfo->colList == NULL) {
-    pQueryInfo->colList = taosArrayInit(4, POINTER_BYTES);
-  }
-
-  if (pGroupExpr->columnInfo == NULL || pQueryInfo->colList == NULL) {
     return TSDB_CODE_TSC_OUT_OF_MEMORY;
   }
 
-  int32_t numOfGroupCols = (int16_t) taosArrayGetSize(pList);
-  if (numOfGroupCols > TSDB_MAX_TAGS) {
+  size_t num = taosArrayGetSize(pList);
+  if (num > TSDB_MAX_TAGS) {
     return buildInvalidOperationMsg(pMsgBuf, msg1);
   }
 
-  SSchema *pSchema       = NULL;
-  int32_t tableIndex     = COLUMN_INDEX_INITIAL_VAL;
+  int32_t  numOfGroupbyCols = 0;
+  SSchema *pSchema          = NULL;
+  int32_t  tableIndex       = COLUMN_INDEX_INITIAL_VAL;
 
-  size_t num = taosArrayGetSize(pList);
   for (int32_t i = 0; i < num; ++i) {
     SListItem * pItem = taosArrayGet(pList, i);
     SVariant* pVar = &pItem->pVar;
 
-    SToken token = {pVar->nLen, pVar->nType, pVar->pz};
-
     SColumnIndex index = COLUMN_INDEX_INITIALIZER;
+    SToken token = {pVar->nLen, pVar->nType, pVar->pz};
     if (getColumnIndexByName(&token, pQueryInfo, &index, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return buildInvalidOperationMsg(pMsgBuf, msg2);
     }
 
+    // Group by multiple tables is not supported.
     if (tableIndex == COLUMN_INDEX_INITIAL_VAL) {
       tableIndex = index.tableIndex;
     } else if (tableIndex != index.tableIndex) {
@@ -506,9 +555,7 @@ int32_t validateGroupbyNode(SQueryStmtInfo* pQueryInfo, SArray* pList, SMsgBuf*
       pSchema = getOneColumnSchema(pTableMeta, index.columnIndex);
     }
 
-    int32_t numOfCols = getNumOfColumns(pTableMeta);
-    bool groupTag = (index.columnIndex == TSDB_TBNAME_COLUMN_INDEX || index.columnIndex >= numOfCols);
-
+    bool groupTag = TSDB_COL_IS_TAG(index.type);
     if (groupTag) {
       if (!UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
         return buildInvalidOperationMsg(pMsgBuf, msg6);
@@ -516,7 +563,7 @@ int32_t validateGroupbyNode(SQueryStmtInfo* pQueryInfo, SArray* pList, SMsgBuf*
 
       int32_t relIndex = index.columnIndex;
       if (index.columnIndex != TSDB_TBNAME_COLUMN_INDEX) {
-        relIndex -= numOfCols;
+        relIndex -= getNumOfColumns(pTableMeta);
       }
 
       SColIndex colIndex = { .colIndex = relIndex, .flag = TSDB_COL_TAG, .colId = pSchema->colId, };
@@ -527,7 +574,7 @@ int32_t validateGroupbyNode(SQueryStmtInfo* pQueryInfo, SArray* pList, SMsgBuf*
       columnListInsert(pTableMetaInfo->tagColList, index.columnIndex, pTableMeta->uid, pSchema);
     } else {
       // check if the column type is valid, here only support the bool/tinyint/smallint/bigint group by
-      if (pSchema->type == TSDB_DATA_TYPE_TIMESTAMP || pSchema->type == TSDB_DATA_TYPE_FLOAT || pSchema->type == TSDB_DATA_TYPE_DOUBLE) {
+      if (pSchema->type == TSDB_DATA_TYPE_FLOAT || pSchema->type == TSDB_DATA_TYPE_DOUBLE) {
         return buildInvalidOperationMsg(pMsgBuf, msg5);
       }
 
@@ -537,14 +584,15 @@ int32_t validateGroupbyNode(SQueryStmtInfo* pQueryInfo, SArray* pList, SMsgBuf*
       strncpy(colIndex.name, pSchema->name, tListLen(colIndex.name));
 
       taosArrayPush(pGroupExpr->columnInfo, &colIndex);
-      pQueryInfo->groupbyExpr.orderType = TSDB_ORDER_ASC;
-      numOfGroupCols++;
+
+      numOfGroupbyCols++;
+      pQueryInfo->info.groupbyColumn = true;
     }
   }
 
   // 1. only one normal column allowed in the group by clause
-  // 2. the normal column in the group by clause can only located in the end position
-  if (numOfGroupCols > 1) {
+  // 2. the normal column in the group by clause can only located at the end position
+  if (numOfGroupbyCols > 1) {
     return buildInvalidOperationMsg(pMsgBuf, msg7);
   }
 
@@ -555,7 +603,8 @@ int32_t validateGroupbyNode(SQueryStmtInfo* pQueryInfo, SArray* pList, SMsgBuf*
     }
   }
 
-  pQueryInfo->groupbyExpr.tableIndex = tableIndex;
+  pGroupExpr->orderType  = TSDB_ORDER_ASC;
+  pGroupExpr->tableIndex = tableIndex;
   return TSDB_CODE_SUCCESS;
 }
 
@@ -589,18 +638,249 @@ int32_t validateWhereNode(SQueryStmtInfo *pQueryInfo, tSqlExpr* pWhereExpr, SMsg
   return 0;
 }
 
+static int32_t parseIntervalOffset(SQueryStmtInfo* pQueryInfo, SToken* offsetToken, int32_t precision, SMsgBuf* pMsgBuf) {
+  const char* msg1 = "interval offset cannot be negative";
+  const char* msg2 = "interval offset should be shorter than interval";
+  const char* msg3 = "cannot use 'year' as offset when interval is 'month'";
+
+  SToken* t = offsetToken;
+  if (t->n == 0) {
+    pQueryInfo->interval.offsetUnit = pQueryInfo->interval.intervalUnit;
+    pQueryInfo->interval.offset = 0;
+    return TSDB_CODE_SUCCESS;
+  }
+
+  if (parseNatualDuration(t->z, t->n, &pQueryInfo->interval.offset, &pQueryInfo->interval.offsetUnit, precision) != TSDB_CODE_SUCCESS) {
+    return TSDB_CODE_TSC_INVALID_OPERATION;
+  }
+
+  if (pQueryInfo->interval.offset < 0) {
+    return buildInvalidOperationMsg(pMsgBuf, msg1);
+  }
+
+  if (!TIME_IS_VAR_DURATION(pQueryInfo->interval.offsetUnit)) {
+    if (!TIME_IS_VAR_DURATION(pQueryInfo->interval.intervalUnit)) {
+      if (pQueryInfo->interval.offset > pQueryInfo->interval.interval) {
+        return buildInvalidOperationMsg(pMsgBuf, msg2);
+      }
+    }
+  } else if (pQueryInfo->interval.offsetUnit == pQueryInfo->interval.intervalUnit) {
+    if (pQueryInfo->interval.offset >= pQueryInfo->interval.interval) {
+      return buildInvalidOperationMsg(pMsgBuf, msg2);
+    }
+  } else if (pQueryInfo->interval.intervalUnit == 'n' && pQueryInfo->interval.offsetUnit == 'y') {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  } else if (pQueryInfo->interval.intervalUnit == 'y' && pQueryInfo->interval.offsetUnit == 'n') {
+    if (pQueryInfo->interval.interval * 12 <= pQueryInfo->interval.offset) {
+      return buildInvalidOperationMsg(pMsgBuf, msg2);
+    }
+  } else {
+    // TODO: offset should be shorter than interval, but how to check
+    // conflicts like 30days offset and 1 month interval
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t parseSlidingClause(SQueryStmtInfo* pQueryInfo, SToken* pSliding, int32_t precision, SMsgBuf* pMsgBuf) {
+  const char* msg1 = "sliding value no larger than the interval value";
+  const char* msg2 = "sliding value can not less than 1% of interval value";
+  const char* msg3 = "does not support sliding when interval is natural month/year";
+  const char* msg4 = "sliding value too small";
+
+  const static int32_t INTERVAL_SLIDING_FACTOR = 100;
+
+  SInterval* pInterval = &pQueryInfo->interval;
+  if (pSliding->n == 0) {
+    pInterval->slidingUnit = pInterval->intervalUnit;
+    pInterval->sliding     = pInterval->interval;
+    return TSDB_CODE_SUCCESS;
+  }
+
+  if (TIME_IS_VAR_DURATION(pInterval->intervalUnit)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  }
+
+  parseAbsoluteDuration(pSliding->z, pSliding->n, &pInterval->sliding, &pInterval->slidingUnit, precision);
+
+  // less than the threshold
+  if (pInterval->sliding < convertTimePrecision(tsMinSlidingTime, TSDB_TIME_PRECISION_MILLI, precision)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg4);
+  }
+
+  if (pInterval->sliding > pInterval->interval) {
+    return buildInvalidOperationMsg(pMsgBuf, msg1);
+  }
+
+  if ((pInterval->interval != 0) && (pInterval->interval/pInterval->sliding > INTERVAL_SLIDING_FACTOR)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg2);
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+
 // validate the interval info
 int32_t validateIntervalNode(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgBuf* pMsgBuf) {
-  return 0;
+  const char* msg1 = "sliding cannot be used without interval";
+  const char* msg2 = "only point interpolation query requires keyword EVERY";
+  const char* msg3 = "interval value is too small";
+
+  STableMetaInfo* pTableMetaInfo = getMetaInfo(pQueryInfo, 0);
+  STableComInfo tinfo = getTableInfo(pTableMetaInfo->pTableMeta);
+
+  if (!TPARSER_HAS_TOKEN(pSqlNode->interval.interval)) {
+    if (TPARSER_HAS_TOKEN(pSqlNode->sliding)) {
+      return buildInvalidOperationMsg(pMsgBuf, msg1);
+    } else {
+      return TSDB_CODE_SUCCESS;
+    }
+  }
+
+  // orderby column not set yet, set it to be the primary timestamp column
+  if (pQueryInfo->order.orderColId == INT32_MIN) {
+    pQueryInfo->order.orderColId = PRIMARYKEY_TIMESTAMP_COL_ID;
+  }
+
+  // interval is not null
+  SToken *t = &pSqlNode->interval.interval;
+  if (parseNatualDuration(t->z, t->n, &pQueryInfo->interval.interval,
+                          &pQueryInfo->interval.intervalUnit, tinfo.precision) != TSDB_CODE_SUCCESS) {
+    return TSDB_CODE_TSC_INVALID_OPERATION;
+  }
+
+  if (pQueryInfo->interval.interval <= 0) {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  }
+
+  if (!TIME_IS_VAR_DURATION(pQueryInfo->interval.intervalUnit)) {
+    // interval cannot be less than 10 milliseconds
+    if (convertTimePrecision(pQueryInfo->interval.interval, tinfo.precision, TSDB_TIME_PRECISION_MICRO) < tsMinIntervalTime) {
+      char msg[50] = {0};
+      snprintf(msg, 50, "interval time window can not be less than %d %s", tsMinIntervalTime, TSDB_TIME_PRECISION_MICRO_STR);
+      return buildInvalidOperationMsg(pMsgBuf, msg);
+    }
+  }
+
+  if (parseIntervalOffset(pQueryInfo, &pSqlNode->interval.offset, tinfo.precision, pMsgBuf) != TSDB_CODE_SUCCESS) {
+    return TSDB_CODE_TSC_INVALID_OPERATION;
+  }
+
+  if (parseSlidingClause(pQueryInfo, &pSqlNode->sliding, tinfo.precision, pMsgBuf) != TSDB_CODE_SUCCESS) {
+    return TSDB_CODE_TSC_INVALID_OPERATION;
+  }
+
+  // It is a time window query
+  pQueryInfo->info.timewindow = true;
+  return TSDB_CODE_SUCCESS;
+  // disable it temporarily
+//  bool interpQuery = tscIsPointInterpQuery(pQueryInfo);
+//  if ((pSqlNode->interval.token == TK_EVERY && (!interpQuery)) || (pSqlNode->interval.token == TK_INTERVAL && interpQuery)) {
+//    return buildInvalidOperationMsg(pMsgBuf, msg4);
+//  }
+
+  // The following part is used to check for the invalid query expression.
+//  return checkInvalidExprForTimeWindow(pCmd, pQueryInfo);
+
 }
 
-int32_t validateSessionNode(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgBuf* pMsgBuf) {
-  return 0;
+int32_t validateSessionNode(SQueryStmtInfo *pQueryInfo, SSessionWindowVal* pSession, int32_t precision, SMsgBuf* pMsgBuf) {
+  const char* msg1 = "gap should be fixed time window";
+  const char* msg2 = "only one type time window allowed";
+  const char* msg3 = "invalid column name";
+  const char* msg4 = "invalid time window";
+
+  // no session window
+  if (!TPARSER_HAS_TOKEN(pSession->gap)) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  SToken* col = &pSession->col;
+  SToken* gap = &pSession->gap;
+
+  char timeUnit = 0;
+  if (parseNatualDuration(gap->z, gap->n, &pQueryInfo->sessionWindow.gap, &timeUnit, precision) != TSDB_CODE_SUCCESS) {
+    return buildInvalidOperationMsg(pMsgBuf, msg4);
+  }
+
+  if (TIME_IS_VAR_DURATION(timeUnit)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg1);
+  }
+
+  if (pQueryInfo->sessionWindow.gap != 0 && pQueryInfo->interval.interval != 0) {
+    return buildInvalidOperationMsg(pMsgBuf, msg2);
+  }
+
+  if (pQueryInfo->sessionWindow.gap == 0) {
+    return buildInvalidOperationMsg(pMsgBuf, msg4);
+  }
+
+  SColumnIndex index = COLUMN_INDEX_INITIALIZER;
+  if ((getColumnIndexByName(col, pQueryInfo, &index, pMsgBuf) != TSDB_CODE_SUCCESS)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  }
+
+  if (index.columnIndex != PRIMARYKEY_TIMESTAMP_COL_ID) {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  }
+
+  pQueryInfo->sessionWindow.primaryColId = PRIMARYKEY_TIMESTAMP_COL_ID;
+  return TSDB_CODE_SUCCESS;
+  // The following part is used to check for the invalid query expression.
+//  return checkInvalidExprForTimeWindow(pCmd, pQueryInfo);
 }
 
 // parse the window_state
-int32_t validateStateWindowNode(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgBuf* pMsgBuf) {
-return 0;
+int32_t validateStateWindowNode(SQueryStmtInfo *pQueryInfo, SWindowStateVal* pWindowState, SMsgBuf* pMsgBuf) {
+  const char* msg1 = "invalid column name";
+  const char* msg2 = "invalid column type";
+  const char* msg3 = "not support state_window with group by ";
+  const char* msg4 = "function not support for super table query";
+  const char* msg5 = "not support state_window on tag column";
+
+  SToken *col = &(pWindowState->col) ;
+  if (!TPARSER_HAS_TOKEN(*col)) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  SGroupbyExpr* pGroupExpr = &pQueryInfo->groupbyExpr;
+  if (taosArrayGetSize(pGroupExpr->columnInfo) > 0) {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  }
+
+  SColumnIndex index = COLUMN_INDEX_INITIALIZER;
+  if (getColumnIndexByName(col, pQueryInfo, &index, pMsgBuf) !=  TSDB_CODE_SUCCESS) {
+    return buildInvalidOperationMsg(pMsgBuf, msg1);
+  }
+
+  STableMetaInfo *pTableMetaInfo = getMetaInfo(pQueryInfo, index.tableIndex);
+  STableMeta* pTableMeta = pTableMetaInfo->pTableMeta;
+
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg4);
+  }
+
+  if (TSDB_COL_IS_TAG(index.type)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg5);
+  }
+
+  if (pGroupExpr->columnInfo == NULL) {
+    pGroupExpr->columnInfo = taosArrayInit(4, sizeof(SColIndex));
+  }
+
+  SSchema* pSchema = getOneColumnSchema(pTableMeta, index.columnIndex);
+  if (pSchema->type == TSDB_DATA_TYPE_TIMESTAMP || IS_FLOAT_TYPE(pSchema->type)) {
+    return buildInvalidOperationMsg(pMsgBuf, msg2);
+  }
+
+  columnListInsert(pQueryInfo->colList, index.columnIndex, pTableMeta->uid, pSchema);
+  SColIndex colIndex = { .colIndex = index.columnIndex, .flag = TSDB_COL_NORMAL, .colId = pSchema->colId };
+
+  //TODO use group by routine? state window query not support stable query.
+  taosArrayPush(pGroupExpr->columnInfo, &colIndex);
+  pGroupExpr->orderType = TSDB_ORDER_ASC;
+  pQueryInfo->info.stateWindow = true;
+
+  return TSDB_CODE_SUCCESS;
 }
 
 // parse the having clause in the first place
@@ -609,16 +889,533 @@ int32_t validateHavingNode(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgB
 }
 
 int32_t validateLimitNode(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgBuf* pMsgBuf) {
-  return 0;
+  STableMetaInfo* pTableMetaInfo = getMetaInfo(pQueryInfo, 0);
+
+  const char* msg1 = "slimit/soffset only available for STable query";
+  const char* msg2 = "slimit/soffset can not apply to projection query";
+  const char* msg3 = "soffset/offset can not be less than 0";
+
+  // handle the limit offset value, validate the limit
+  pQueryInfo->limit = pSqlNode->limit;
+  pQueryInfo->slimit = pSqlNode->slimit;
+
+//  tscDebug("0x%"PRIx64" limit:%" PRId64 ", offset:%" PRId64 " slimit:%" PRId64 ", soffset:%" PRId64, pSql->self,
+//           pQueryInfo->limit.limit, pQueryInfo->limit.offset, pQueryInfo->slimit.limit, pQueryInfo->slimit.offset);
+
+  if (pQueryInfo->slimit.offset < 0 || pQueryInfo->limit.offset < 0) {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  }
+
+  if (pQueryInfo->limit.limit == 0) {
+//    tscDebug("0x%"PRIx64" limit 0, no output result", pSql->self);
+    pQueryInfo->command = TSDB_SQL_RETRIEVE_EMPTY_RESULT;
+    return TSDB_CODE_SUCCESS;
+  }
+
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
+//    if (!tscQueryTags(pQueryInfo)) {  // local handle the super table tag query
+//      if (tscIsProjectionQueryOnSTable(pQueryInfo, 0)) {
+//        if (pQueryInfo->slimit.limit > 0 || pQueryInfo->slimit.offset > 0) {
+//          return buildInvalidOperationMsg(pMsgBuf, msg2);
+//        }
+//
+//        // for projection query on super table, all queries are subqueries
+//        if (tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0) &&
+//            !TSDB_QUERY_HAS_TYPE(pQueryInfo->type, TSDB_QUERY_TYPE_JOIN_QUERY)) {
+//          pQueryInfo->type |= TSDB_QUERY_TYPE_SUBQUERY;
+//        }
+//      }
+//    }
+
+    if (pQueryInfo->slimit.limit == 0) {
+//      tscDebug("0x%"PRIx64" slimit 0, no output result", pSql->self);
+      pQueryInfo->command = TSDB_SQL_RETRIEVE_EMPTY_RESULT;
+      return TSDB_CODE_SUCCESS;
+    }
+    
+    // No tables included. No results generated. Query results are empty.
+    if (pTableMetaInfo->vgroupList->numOfVgroups == 0) {
+//      tscDebug("0x%"PRIx64" no table in super table, no output result", pSql->self);
+      pQueryInfo->command = TSDB_SQL_RETRIEVE_EMPTY_RESULT;
+      return TSDB_CODE_SUCCESS;
+    }
+  } else {
+    if (pQueryInfo->slimit.limit != -1 || pQueryInfo->slimit.offset != 0) {
+      return buildInvalidOperationMsg(pMsgBuf, msg1);
+    }
+  }
 }
 
-// set order by info
+static void setTsOutputExprInfo(SQueryStmtInfo* pQueryInfo, STableMetaInfo* pTableMetaInfo, int32_t outputIndex, int32_t tableIndex);
+
+
 int32_t validateOrderbyNode(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgBuf* pMsgBuf) {
-  return 0;
+  const char* msg1 = "invalid column name in orderby clause";
+  const char* msg2 = "too many order by columns";
+  const char* msg3 = "only one column allowed in orderby";
+  const char* msg4 = "invalid order by column index";
+
+  if (pSqlNode->pSortOrder == NULL) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  STableMetaInfo* pTableMetaInfo = getMetaInfo(pQueryInfo, 0);
+  SArray* pSortOrder = pSqlNode->pSortOrder;
+
+  /*
+   * for table query, there is only one or none order option is allowed, which is the
+   * ts or values(top/bottom) order is supported.
+   *
+   * for super table query, the order option must be less than 3.
+   */
+  size_t size = taosArrayGetSize(pSortOrder);
+  if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo) || UTIL_TABLE_IS_TMP_TABLE(pTableMetaInfo)) {
+    if (size > 1) {
+      return buildInvalidOperationMsg(pMsgBuf, msg3);
+    }
+  } else {
+    if (size > 2) {
+      return buildInvalidOperationMsg(pMsgBuf, msg2);
+    }
+  }
+
+  // handle the first part of order by
+  SVariant* pVar = taosArrayGet(pSortOrder, 0);
+  SSchema s = {0};
+  if (pVar->nType == TSDB_DATA_TYPE_BINARY) {
+    SColumnIndex index = COLUMN_INDEX_INITIALIZER;
+    SToken columnName = {pVar->nLen, pVar->nType, pVar->pz};
+    if (getColumnIndexByName(&columnName, pQueryInfo, &index, pMsgBuf) != TSDB_CODE_SUCCESS) {
+      return buildInvalidOperationMsg(pMsgBuf, msg1);
+    }
+
+    s = *(SSchema*) getOneColumnSchema(pTableMetaInfo->pTableMeta, index.columnIndex);
+  } else { // order by [1|2|3]
+    if (pVar->i64 > getNumOfExprs(pQueryInfo)) {
+      return buildInvalidOperationMsg(pMsgBuf, msg4);
+    }
+
+    SExprInfo* pExprInfo = getExprInfo(pQueryInfo, pVar->i64);
+    s = pExprInfo->base.resSchema;
+  }
+
+  SListItem* pItem = taosArrayGet(pSqlNode->pSortOrder, 0);
+  pQueryInfo->order.order = pItem->sortOrder;
+  pQueryInfo->order.orderColId = s.colId;
+
+  return TSDB_CODE_SUCCESS;
+}
+
+#if 0
+// set order by info
+int32_t checkForInvalidOrderby(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgBuf* pMsgBuf) {
+  const char* msg0 = "only one column allowed in orderby";
+  const char* msg1 = "invalid column name in orderby clause";
+  const char* msg2 = "too many order by columns";
+  const char* msg3 = "only primary timestamp/tbname/first tag in groupby clause allowed";
+  const char* msg4 = "only tag in groupby clause allowed in order clause";
+  const char* msg5 = "only primary timestamp/column in top/bottom function allowed as order column";
+  const char* msg6 = "only primary timestamp allowed as the second order column";
+  const char* msg7 = "only primary timestamp/column in groupby clause allowed as order column";
+  const char* msg8 = "only column in groupby clause allowed as order column";
+  const char* msg9 = "orderby column must projected in subquery";
+  const char* msg10 = "not support distinct mixed with order by";
+
+//  setDefaultOrderInfo(pQueryInfo);
+  STableMetaInfo* pTableMetaInfo = getMetaInfo(pQueryInfo, 0);
+  SSchema* pSchema = getTableColumnSchema(pTableMetaInfo->pTableMeta);
+  int32_t numOfCols = getNumOfColumns(pTableMetaInfo->pTableMeta);
+
+  if (pSqlNode->pSortOrder == NULL) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  SArray* pSortOrder = pSqlNode->pSortOrder;
+
+  /*
+   * for table query, there is only one or none order option is allowed, which is the
+   * ts or values(top/bottom) order is supported.
+   *
+   * for super table query, the order option must be less than 3.
+   */
+  size_t size = taosArrayGetSize(pSortOrder);
+  if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo) || UTIL_TABLE_IS_TMP_TABLE(pTableMetaInfo)) {
+    if (size > 1) {
+      return buildInvalidOperationMsg(pMsgBuf, msg0);
+    }
+  } else {
+    if (size > 2) {
+      return buildInvalidOperationMsg(pMsgBuf, msg2);
+    }
+  }
+
+#if 0
+  if (size > 0 && pQueryInfo->distinct) {
+    return buildInvalidOperationMsg(pMsgBuf, msg10);
+  }
+#endif
+
+  // handle the first part of order by
+  SVariant* pVar = taosArrayGet(pSortOrder, 0);
+
+#if 0
+  // e.g., order by 1 asc, return directly with out further check.
+  if (pVar->nType >= TSDB_DATA_TYPE_TINYINT && pVar->nType <= TSDB_DATA_TYPE_BIGINT) {
+    return TSDB_CODE_SUCCESS;
+  }
+#endif
+
+  SToken columnName = {pVar->nLen, pVar->nType, pVar->pz};
+
+  SColumnIndex index = COLUMN_INDEX_INITIALIZER;
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {  // super table query
+    if (getColumnIndexByName(&columnName, pQueryInfo, &index, pMsgBuf) != TSDB_CODE_SUCCESS) {
+      return buildInvalidOperationMsg(pMsgBuf, msg1);
+    }
+
+    bool orderByTags = false;
+    bool orderByTS = false;
+    bool orderByGroupbyCol = false;
+
+    if (TSDB_COL_IS_TAG(index.type) && index.columnIndex != TSDB_TBNAME_COLUMN_INDEX) {
+      // it is a tag column
+      if (pQueryInfo->groupbyExpr.columnInfo == NULL) {
+        return buildInvalidOperationMsg(pMsgBuf, msg4);
+      }
+
+      int32_t relTagIndex = index.columnIndex - numOfCols;
+      SColIndex* pColIndex = taosArrayGet(pQueryInfo->groupbyExpr.columnInfo, 0);
+      if (relTagIndex == pColIndex->colIndex) {
+        orderByTags = true;
+      }
+    } else if (index.columnIndex == TSDB_TBNAME_COLUMN_INDEX) {
+      orderByTags = true;
+    }
+
+    if (PRIMARYKEY_TIMESTAMP_COL_ID == index.columnIndex) {
+      orderByTS = true;
+    }
+
+    SArray *columnInfo = pQueryInfo->groupbyExpr.columnInfo;
+    if (columnInfo != NULL && taosArrayGetSize(columnInfo) > 0) {
+      SColIndex* pColIndex = taosArrayGet(columnInfo, 0);
+      if (PRIMARYKEY_TIMESTAMP_COL_ID != index.columnIndex && pColIndex->colIndex == index.columnIndex) {
+        orderByGroupbyCol = true;
+      }
+    }
+
+    if (!(orderByTags || orderByTS || orderByGroupbyCol) /*&& !isTopBottomQuery(pQueryInfo)*/) {
+      return buildInvalidOperationMsg(pMsgBuf, msg3);
+    } else {  // order by top/bottom result value column is not supported in case of interval query.
+      assert(!(orderByTags && orderByTS && orderByGroupbyCol));
+    }
+
+    size_t s = taosArrayGetSize(pSortOrder);
+    if (s == 1) {
+      if (orderByTags) {
+        pQueryInfo->groupbyExpr.orderIndex = index.columnIndex - numOfCols;
+
+        SListItem* p1 = taosArrayGet(pSqlNode->pSortOrder, 0);
+        pQueryInfo->groupbyExpr.orderType = p1->sortOrder;
+      } else if (orderByGroupbyCol) {
+        SListItem* p1 = taosArrayGet(pSqlNode->pSortOrder, 0);
+
+        pQueryInfo->groupbyExpr.orderType = p1->sortOrder;
+        pQueryInfo->order.orderColId = pSchema[index.columnIndex].colId;
+      } else if (isTopBottomQuery(pQueryInfo)) {
+        /* order of top/bottom query in interval is not valid  */
+        int32_t pos = tscExprTopBottomIndex(pQueryInfo);
+        assert(pos > 0);
+        
+        SExprInfo* pExpr = getExprInfo(pQueryInfo, pos - 1);
+//        assert(getExprFunctionId(pExpr) == FUNCTION_TS);
+
+        pExpr = getExprInfo(pQueryInfo, pos);
+
+        // other tag are not allowed
+        if (pExpr->base.colInfo.colIndex != index.columnIndex && index.columnIndex != PRIMARYKEY_TIMESTAMP_COL_ID) {
+          return buildInvalidOperationMsg(pMsgBuf, msg5);
+        }
+
+        SListItem* p1 = taosArrayGet(pSqlNode->pSortOrder, 0);
+        pQueryInfo->order.order = p1->sortOrder;
+        pQueryInfo->order.orderColId = pSchema[index.columnIndex].colId;
+        return TSDB_CODE_SUCCESS;
+      } else {
+        SListItem* p1 = taosArrayGet(pSqlNode->pSortOrder, 0);
+
+        pQueryInfo->order.order = p1->sortOrder;
+        pQueryInfo->order.orderColId = PRIMARYKEY_TIMESTAMP_COL_ID;
+
+        // orderby ts query on super table
+        if (tscOrderedProjectionQueryOnSTable(pQueryInfo, 0)) {
+          bool found = false;
+          for (int32_t i = 0; i < tscNumOfExprs(pQueryInfo); ++i) {
+            SExprInfo* pExpr = getExprInfo(pQueryInfo, i);
+            if (getExprFunctionId(pExpr) == FUNCTION_PRJ && pExpr->base.colInfo.colId == PRIMARYKEY_TIMESTAMP_COL_ID) {
+              found = true;
+              break;
+            }
+          }
+
+          if (!found && pQueryInfo->pDownstream) {
+            return buildInvalidOperationMsg(pMsgBuf, msg9);
+          }
+
+          // this is a invisible output column, in order to used to sort the result.
+          setTsOutputExprInfo(pQueryInfo, pTableMetaInfo, 0, index.tableIndex);
+        }
+      }
+    } else {
+      SListItem *pItem = taosArrayGet(pSqlNode->pSortOrder, 0);
+      if (orderByTags) {
+        pQueryInfo->groupbyExpr.orderIndex = index.columnIndex - numOfCols;
+        pQueryInfo->groupbyExpr.orderType = pItem->sortOrder;
+      } else if (orderByGroupbyCol) {
+        pQueryInfo->order.order = pItem->sortOrder;
+        pQueryInfo->order.orderColId = index.columnIndex;
+      } else {
+        pQueryInfo->order.order = pItem->sortOrder;
+        pQueryInfo->order.orderColId = PRIMARYKEY_TIMESTAMP_COL_ID;
+      }
+
+      pItem = taosArrayGet(pSqlNode->pSortOrder, 1);
+      SVariant* pVar2 = &pItem->pVar;
+      SToken cname = {pVar2->nLen, pVar2->nType, pVar2->pz};
+      if (getColumnIndexByName(&cname, pQueryInfo, &index, pMsgBuf) != TSDB_CODE_SUCCESS) {
+        return buildInvalidOperationMsg(pMsgBuf, msg1);
+      }
+
+      if (index.columnIndex != PRIMARYKEY_TIMESTAMP_COL_ID) {
+        return buildInvalidOperationMsg(pMsgBuf, msg6);
+      } else {
+        SListItem* p1 = taosArrayGet(pSortOrder, 1);
+        pQueryInfo->order.order = p1->sortOrder;
+        pQueryInfo->order.orderColId = PRIMARYKEY_TIMESTAMP_COL_ID;
+      }
+    }
+
+  } else if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo) || UTIL_TABLE_IS_CHILD_TABLE(pTableMetaInfo)) { // check order by clause for normal table & temp table
+    if (getColumnIndexByName(&columnName, pQueryInfo, &index, pMsgBuf) != TSDB_CODE_SUCCESS) {
+      return buildInvalidOperationMsg(pMsgBuf, msg1);
+    }
+
+    if (index.columnIndex != PRIMARYKEY_TIMESTAMP_COL_ID && !isTopBottomQuery(pQueryInfo)) {
+      bool validOrder = false;
+      SArray *columnInfo = pQueryInfo->groupbyExpr.columnInfo;
+      if (columnInfo != NULL && taosArrayGetSize(columnInfo) > 0) {
+        SColIndex* pColIndex = taosArrayGet(columnInfo, 0);
+        validOrder = (pColIndex->colIndex == index.columnIndex);
+      }
+
+      if (!validOrder) {
+        return buildInvalidOperationMsg(pMsgBuf, msg7);
+      }
+
+      SListItem* p1 = taosArrayGet(pSqlNode->pSortOrder, 0);
+      pQueryInfo->groupbyExpr.orderIndex = pSchema[index.columnIndex].colId;
+      pQueryInfo->groupbyExpr.orderType = p1->sortOrder;
+    }
+
+    if (isTopBottomQuery(pQueryInfo)) {
+      SArray *columnInfo = pQueryInfo->groupbyExpr.columnInfo;
+      if (columnInfo != NULL && taosArrayGetSize(columnInfo) > 0) {
+        SColIndex* pColIndex = taosArrayGet(columnInfo, 0);
+
+        if (pColIndex->colIndex == index.columnIndex) {
+          return buildInvalidOperationMsg(pMsgBuf, msg8);
+        }
+      } else {
+        int32_t pos = tscExprTopBottomIndex(pQueryInfo);
+        assert(pos > 0);
+        SExprInfo* pExpr = getExprInfo(pQueryInfo, pos - 1);
+        assert(getExprFunctionId(pExpr) == FUNCTION_TS);
+
+        pExpr = getExprInfo(pQueryInfo, pos);
+
+        if (pExpr->base.colInfo.colIndex != index.columnIndex && index.columnIndex != PRIMARYKEY_TIMESTAMP_COL_ID) {
+          return buildInvalidOperationMsg(pMsgBuf, msg5);
+        }
+      }
+
+      SListItem* pItem = taosArrayGet(pSqlNode->pSortOrder, 0);
+      pQueryInfo->order.order = pItem->sortOrder;
+
+      pQueryInfo->order.orderColId = pSchema[index.columnIndex].colId;
+      return TSDB_CODE_SUCCESS;
+    }
+
+    SListItem* pItem = taosArrayGet(pSqlNode->pSortOrder, 0);
+    pQueryInfo->order.order = pItem->sortOrder;
+    pQueryInfo->order.orderColId = pSchema[index.columnIndex].colId;
+  } else {
+    // handle the temp table order by clause. You can order by any single column in case of the temp table, created by
+    // inner subquery.
+    assert(UTIL_TABLE_IS_TMP_TABLE(pTableMetaInfo) && taosArrayGetSize(pSqlNode->pSortOrder) == 1);
+
+    if (getColumnIndexByName(&columnName, pQueryInfo, &index, pMsgBuf) != TSDB_CODE_SUCCESS) {
+      return buildInvalidOperationMsg(pMsgBuf, msg1);
+    }
+
+    SListItem* pItem = taosArrayGet(pSqlNode->pSortOrder, 0);
+    pQueryInfo->order.order = pItem->sortOrder;
+    pQueryInfo->order.orderColId = pSchema[index.columnIndex].colId;
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+#endif
+
+static int32_t checkQueryRangeForFill(SQueryStmtInfo* pQueryInfo, SMsgBuf* pMsgBuf) {
+  const char* msg3 = "start(end) time of time range required or time range too large";
+
+  if (pQueryInfo->interval.interval == 0) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  bool initialWindows = TSWINDOW_IS_EQUAL(pQueryInfo->window, TSWINDOW_INITIALIZER);
+  if (initialWindows) {
+    return buildInvalidOperationMsg(pMsgBuf, msg3);
+  }
+
+  int64_t timeRange = ABS(pQueryInfo->window.skey - pQueryInfo->window.ekey);
+
+  int64_t intervalRange = 0;
+  if (!TIME_IS_VAR_DURATION(pQueryInfo->interval.intervalUnit)) {
+    intervalRange = pQueryInfo->interval.interval;
+
+    // number of result is not greater than 10,000,000
+    if ((timeRange == 0) || (timeRange / intervalRange) >= MAX_INTERVAL_TIME_WINDOW) {
+      return buildInvalidOperationMsg(pMsgBuf, msg3);
+    }
+  }
+
+  return TSDB_CODE_SUCCESS;
 }
 
 int32_t validateFillNode(SQueryStmtInfo *pQueryInfo, SSqlNode* pSqlNode, SMsgBuf* pMsgBuf) {
-  return 0;
+  SArray* pFillToken = pSqlNode->fillType;
+  if (pSqlNode->fillType == NULL) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  SListItem* pItem = taosArrayGet(pFillToken, 0);
+
+  const int32_t START_INTERPO_COL_IDX = 1;
+
+  const char* msg1 = "value is expected";
+  const char* msg2 = "invalid fill option";
+  const char* msg3 = "top/bottom not support fill";
+  const char* msg4 = "illegal value or data overflow";
+  const char* msg5 = "fill only available for interval query";
+  const char* msg6 = "not supported function now";
+
+//  if ((!isTimeWindowQuery(pQueryInfo)) && (!tscIsPointInterpQuery(pQueryInfo))) {
+//    return buildInvalidOperationMsg(pMsgBuf, msg5);
+//  }
+
+  /*
+   * fill options are set at the end position, when all columns are set properly
+   * the columns may be increased due to group by operation
+   */
+  if (checkQueryRangeForFill(pQueryInfo, pMsgBuf) != TSDB_CODE_SUCCESS) {
+    return TSDB_CODE_TSC_INVALID_OPERATION;
+  }
+
+
+  if (pItem->pVar.nType != TSDB_DATA_TYPE_BINARY) {
+    return buildInvalidOperationMsg(pMsgBuf, msg2);
+  }
+
+  int32_t numOfFields = (int32_t) getNumOfFields(&pQueryInfo->fieldsInfo);
+
+  pQueryInfo->fillVal = calloc(numOfFields, sizeof(int64_t));
+  if (pQueryInfo->fillVal == NULL) {
+    return TSDB_CODE_TSC_OUT_OF_MEMORY;
+  }
+
+  pQueryInfo->numOfFillVal = (int32_t)numOfFields;
+  if (strncasecmp(pItem->pVar.pz, "none", 4) == 0 && pItem->pVar.nLen == 4) {
+    pQueryInfo->fillType = TSDB_FILL_NONE;
+  } else if (strncasecmp(pItem->pVar.pz, "null", 4) == 0 && pItem->pVar.nLen == 4) {
+    pQueryInfo->fillType = TSDB_FILL_NULL;
+    for (int32_t i = START_INTERPO_COL_IDX; i < numOfFields; ++i) {
+      TAOS_FIELD* pField = &getInternalField(&pQueryInfo->fieldsInfo, i)->field;
+      setNull((char*)&pQueryInfo->fillVal[i], pField->type, pField->bytes);
+    }
+  } else if (strncasecmp(pItem->pVar.pz, "prev", 4) == 0 && pItem->pVar.nLen == 4) {
+    pQueryInfo->fillType = TSDB_FILL_PREV;
+//    if (pQueryInfo->info.interpQuery && pQueryInfo->order.order == TSDB_ORDER_DESC) {
+//      return buildInvalidOperationMsg(pMsgBuf, msg6);
+//    }
+  } else if (strncasecmp(pItem->pVar.pz, "next", 4) == 0 && pItem->pVar.nLen == 4) {
+    pQueryInfo->fillType = TSDB_FILL_NEXT;
+  } else if (strncasecmp(pItem->pVar.pz, "linear", 6) == 0 && pItem->pVar.nLen == 6) {
+    pQueryInfo->fillType = TSDB_FILL_LINEAR;
+  } else if (strncasecmp(pItem->pVar.pz, "value", 5) == 0 && pItem->pVar.nLen == 5) {
+    pQueryInfo->fillType = TSDB_FILL_SET_VALUE;
+
+    size_t num = taosArrayGetSize(pFillToken);
+    if (num == 1) {  // no actual value, return with error code
+      return buildInvalidOperationMsg(pMsgBuf, msg1);
+    }
+
+    int32_t startPos = 1;
+    int32_t numOfFillVal = (int32_t)(num - 1);
+
+    // for point interpolation query, we do not have the timestamp column
+    if (pQueryInfo->info.interpQuery) {
+      startPos = 0;
+      if (numOfFillVal > numOfFields) {
+        numOfFillVal = numOfFields;
+      }
+    } else {
+      numOfFillVal = MIN(num, numOfFields);
+    }
+
+    int32_t j = 1;
+
+    for (int32_t i = startPos; i < numOfFillVal; ++i, ++j) {
+      TAOS_FIELD* pField = &getInternalField(&pQueryInfo->fieldsInfo, i)->field;
+      if (pField->type == TSDB_DATA_TYPE_BINARY || pField->type == TSDB_DATA_TYPE_NCHAR) {
+        setVardataNull((char*) &pQueryInfo->fillVal[i], pField->type);
+        continue;
+      }
+
+      SVariant* p = taosArrayGet(pFillToken, j);
+      int32_t ret = taosVariantDump(p, (char*)&pQueryInfo->fillVal[i], pField->type, true);
+      if (ret != TSDB_CODE_SUCCESS) {
+        return buildInvalidOperationMsg(pMsgBuf, msg4);
+      }
+    }
+
+    if ((num < numOfFields) || ((num - 1 < numOfFields) && (pQueryInfo->info.interpQuery))) {
+      SListItem* lastItem = taosArrayGetLast(pFillToken);
+
+      for (int32_t i = numOfFillVal; i < numOfFields; ++i) {
+        TAOS_FIELD* pField = &getInternalField(&pQueryInfo->fieldsInfo, i)->field;
+
+        if (pField->type == TSDB_DATA_TYPE_BINARY || pField->type == TSDB_DATA_TYPE_NCHAR) {
+          setVardataNull((char*) &pQueryInfo->fillVal[i], pField->type);
+        } else {
+          taosVariantDump(&lastItem->pVar, (char*)&pQueryInfo->fillVal[i], pField->type, true);
+        }
+      }
+    }
+  } else {
+    return buildInvalidOperationMsg(pMsgBuf, msg2);
+  }
+
+//  for(int32_t i = 0; i < tscNumOfExprs(pQueryInfo); ++i) {
+//    SExprInfo* pExpr = getExprInfo(pQueryInfo, i);
+//
+//    int32_t functionId = pExpr->pExpr->_node.functionId;
+//    if (functionId == FUNCTION_TOP || functionId == FUNCTION_BOTTOM) {
+//      return buildInvalidOperationMsg(pMsgBuf, msg3);
+//    }
+//  }
+
+  return TSDB_CODE_SUCCESS;
 }
 
 int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf* pMsgBuf) {
@@ -679,6 +1476,7 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
 
     STableMeta* pTableMeta = getMetaInfo(pQueryInfo, 0)->pTableMeta;
     SSchema*    pSchema = getOneColumnSchema(pTableMeta, 0);
+    int32_t precision = pTableMeta->tableInfo.precision;
 
     if (pSchema->type != TSDB_DATA_TYPE_TIMESTAMP) {
       int32_t numOfExprs = (int32_t)tscNumOfExprs(pQueryInfo);
@@ -708,32 +1506,19 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
     if (validateIntervalNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     } else {
-      if (validateSessionNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
+      if (validateSessionNode(pQueryInfo, &pSqlNode->sessionVal, precision, pMsgBuf) != TSDB_CODE_SUCCESS) {
         return TSDB_CODE_TSC_INVALID_OPERATION;
       }
 
       // parse the window_state
-      if (validateStateWindowNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
+      if (validateStateWindowNode(pQueryInfo, &pSqlNode->windowstateVal, pMsgBuf) != TSDB_CODE_SUCCESS) {
         return TSDB_CODE_TSC_INVALID_OPERATION;
       }
-
-//      if (isTimeWindowQuery(pQueryInfo)) {
-//        // check if the first column of the nest query result is timestamp column
-//        SColumn* pCol = taosArrayGetP(pQueryInfo->colList, 0);
-//        if (pCol->info.type != TSDB_DATA_TYPE_TIMESTAMP) {
-//          return buildInvalidOperationMsg(pMsgBuf, msg4);
-//        }
-//
-//        if (validateFunctionsInIntervalOrGroupbyQuery(pCmd, pQueryInfo) != TSDB_CODE_SUCCESS) {
-//          return TSDB_CODE_TSC_INVALID_OPERATION;
-//        }
-//      }
     }
 
     // parse the having clause in the first place
     int32_t joinQuery = (pSqlNode->from != NULL && taosArrayGetSize(pSqlNode->from->list) > 1);
-    if (validateHavingNode(pQueryInfo, pSqlNode, pMsgBuf) !=
-        TSDB_CODE_SUCCESS) {
+    if (validateHavingNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
 
@@ -758,6 +1543,8 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
     }
 
     STableMetaInfo* pTableMetaInfo = getMetaInfo(pQueryInfo, 0);
+    int32_t precision = pTableMetaInfo->pTableMeta->tableInfo.precision;
+    
     bool isSTable = UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo);
 
     int32_t type = isSTable? TSDB_QUERY_TYPE_STABLE_QUERY:TSDB_QUERY_TYPE_TABLE_QUERY;
@@ -767,14 +1554,12 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
     if (validateGroupbyNode(pQueryInfo, pSqlNode->pGroupby, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
-    pQueryInfo->onlyHasTagCond = true;
 
     // set where info
     if (pSqlNode->pWhere != NULL) {
       if (validateWhereNode(pQueryInfo, pSqlNode->pWhere, pMsgBuf) != TSDB_CODE_SUCCESS) {
         return TSDB_CODE_TSC_INVALID_OPERATION;
       }
-      pSqlNode->pWhere = NULL;
     } else {
       if (taosArrayGetSize(pSqlNode->from->list) > 1) { // Cross join not allowed yet
         return buildInvalidOperationMsg(pMsgBuf, "cross join not supported yet");
@@ -786,18 +1571,13 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
 
-    if (isSTable && (pQueryInfo) && pQueryInfo->distinct && !pQueryInfo->onlyHasTagCond) {
-      return TSDB_CODE_TSC_INVALID_OPERATION;
-    }
-
     // parse the window_state
-    if (validateStateWindowNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
+    if (validateStateWindowNode(pQueryInfo, &pSqlNode->windowstateVal, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
 
     // set order by info
-    if (validateOrderbyNode(pQueryInfo, pSqlNode, pMsgBuf) !=
-        TSDB_CODE_SUCCESS) {
+    if (validateOrderbyNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
 
@@ -807,8 +1587,7 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
     }
 
     // parse the having clause in the first place
-    if (validateHavingNode(pQueryInfo, pSqlNode, pMsgBuf) !=
-        TSDB_CODE_SUCCESS) {
+    if (validateHavingNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
 
@@ -816,7 +1595,7 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
      * transfer sql functions that need secondary merge into another format
      * in dealing with super table queries such as: count/first/last
      */
-    if (validateSessionNode(pQueryInfo, pSqlNode, pMsgBuf) != TSDB_CODE_SUCCESS) {
+    if (validateSessionNode(pQueryInfo, &pSqlNode->sessionVal, precision, pMsgBuf) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
 
@@ -839,18 +1618,6 @@ int32_t validateSqlNode(SSqlNode* pSqlNode, SQueryStmtInfo* pQueryInfo, SMsgBuf*
     }
   }
 
-  // set the query info
-  SExprInfo** p = NULL;
-  int32_t numOfExpr = 0;
-  STableMetaInfo* pTableMetaInfo = getMetaInfo(pQueryInfo, 0);
-  code = createProjectionExpr(pQueryInfo, pTableMetaInfo, &p, &numOfExpr);
-  if (pQueryInfo->exprList1 == NULL) {
-    pQueryInfo->exprList1 = taosArrayInit(4, POINTER_BYTES);
-  }
-
-  taosArrayAddBatch(pQueryInfo->exprList1, (void*) p, numOfExpr);
-  tfree(p);
-
   return TSDB_CODE_SUCCESS;  // Does not build query message here
 }
 
@@ -947,7 +1714,7 @@ void doAddSourceColumnAndResColumn(SQueryStmtInfo* pQueryInfo, SColumnIndex* ind
 }
 
 static int32_t addOneExprInfo(SQueryStmtInfo* pQueryInfo, tSqlExprItem* pItem, int32_t functionId,
-                              int32_t resColIdx, SColumnIndex* pColIndex, bool finalResult, SMsgBuf* pMsgBuf) {
+                              int32_t outputColIndex, SColumnIndex* pColIndex, tExprNode* pNode, bool finalResult, SMsgBuf* pMsgBuf) {
   const char* msg1 = "not support column types";
 
   STableMetaInfo* pTableMetaInfo = getMetaInfo(pQueryInfo, pColIndex->tableIndex);
@@ -967,7 +1734,7 @@ static int32_t addOneExprInfo(SQueryStmtInfo* pQueryInfo, tSqlExprItem* pItem, i
   getResultDataInfo(pSchema->type, pSchema->bytes, functionId, 0, &resInfo, 0, false);
 
   SSchema resultSchema = createSchema(resInfo.type, resInfo.bytes, getNewResColId(), name);
-  doAddOneExprInfo(pQueryInfo, resColIdx, functionId, pColIndex, pSchema, &resultSchema, NULL, resInfo.intermediateBytes, name);
+  doAddOneExprInfo(pQueryInfo, outputColIndex, functionId, pColIndex, pSchema, &resultSchema, pNode, resInfo.intermediateBytes, name);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -1005,7 +1772,7 @@ static int64_t getTickPerSecond(SVariant* pVariant, int32_t precision, int64_t*
 
 // set the first column ts for top/bottom query
 static void setTsOutputExprInfo(SQueryStmtInfo* pQueryInfo, STableMetaInfo* pTableMetaInfo, int32_t outputIndex, int32_t tableIndex) {
-  SColumnIndex indexTS = {.tableIndex = tableIndex, .columnIndex = PRIMARYKEY_TIMESTAMP_COL_INDEX, .type = TSDB_COL_NORMAL};
+  SColumnIndex indexTS = {.tableIndex = tableIndex, .columnIndex = PRIMARYKEY_TIMESTAMP_COL_ID, .type = TSDB_COL_NORMAL};
   SSchema s = createSchema(TSDB_DATA_TYPE_TIMESTAMP, TSDB_KEYSIZE, getNewResColId(), "ts");
 
   SExprInfo* pExpr = createExprInfo(pTableMetaInfo, FUNCTION_TS_DUMMY, &indexTS, NULL, &s, TSDB_KEYSIZE);
@@ -1038,7 +1805,7 @@ static int32_t setColumnIndex(SQueryStmtInfo* pQueryInfo, SArray* pParamList, SC
         return buildInvalidOperationMsg(pMsgBuf, msg4);
       }
 
-      index->columnIndex = PRIMARYKEY_TIMESTAMP_COL_INDEX;
+      index->columnIndex = PRIMARYKEY_TIMESTAMP_COL_ID;
     } else {
       // count the number of table created according to the super table
       if (getColumnIndexByName(pToken, pQueryInfo, index, pMsgBuf) != TSDB_CODE_SUCCESS) {
@@ -1046,7 +1813,7 @@ static int32_t setColumnIndex(SQueryStmtInfo* pQueryInfo, SArray* pParamList, SC
       }
     }
   } else {  // count(*) is equalled to count(primary_timestamp_key)
-    *index = (SColumnIndex) {0, PRIMARYKEY_TIMESTAMP_COL_INDEX, false};
+    *index = (SColumnIndex) {0, PRIMARYKEY_TIMESTAMP_COL_ID, false};
   }
 
   return TSDB_CODE_SUCCESS;
@@ -1057,7 +1824,7 @@ static int32_t doAddAllColumnExprInSelectClause(SQueryStmtInfo *pQueryInfo, STab
   for (int32_t i = 0; i < getNumOfColumns(pTableMetaInfo->pTableMeta); ++i) {
     SColumnIndex index = {.tableIndex = tableIndex, .columnIndex = i, .type = TSDB_COL_NORMAL};
 
-    if (addOneExprInfo(pQueryInfo, pItem, functionId, *colIndex, &index, finalResult, pMsgBuf) != 0) {
+    if (addOneExprInfo(pQueryInfo, pItem, functionId, *colIndex, &index, NULL, finalResult, pMsgBuf) != 0) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
 
@@ -1102,7 +1869,7 @@ static int32_t doHandleOneParam(SQueryStmtInfo *pQueryInfo, tSqlExprItem* pItem,
       return buildInvalidOperationMsg(pMsgBuf, msg6);
     }
 
-    if (addOneExprInfo(pQueryInfo, pItem, functionId, (*outputIndex)++, &index, finalResult, pMsgBuf) != 0) {
+    if (addOneExprInfo(pQueryInfo, pItem, functionId, (*outputIndex)++, &index, pNode, finalResult, pMsgBuf) != 0) {
       return TSDB_CODE_TSC_INVALID_OPERATION;
     }
   }
@@ -1156,6 +1923,8 @@ int32_t extractFunctionParameterInfo(SQueryStmtInfo* pQueryInfo, int32_t tokenId
 
     pIndex->tableIndex = 0;
     multiColumnListInsert(pQueryInfo, pColumnList, pMsgBuf);
+    taosArrayDestroy(colList);
+    taosArrayDestroy(pColumnList);
   } else {
     assert(0);
   }
@@ -1421,7 +2190,6 @@ int32_t addExprAndResColumn(SQueryStmtInfo* pQueryInfo, int32_t colIndex, tSqlEx
           return buildInvalidOperationMsg(pMsgBuf, msg5);
         }
 
-
         getResultDataInfo(pSchema->type, pSchema->bytes, functionId, 0, &resInfo, 0, false);
 
         /*
@@ -2129,13 +2897,11 @@ static int32_t createComplexExpr(SQueryStmtInfo* pQueryInfo, int32_t exprIndex,
     char* c = tbufGetData(&bw, false);
 
     // set the serialized binary string as the parameter of arithmetic expression
-    SColumnIndex* index1 = taosArrayGet(pColumnList, 0);
     addExprInfoParam(&pExpr->base, c, TSDB_DATA_TYPE_BINARY, (int32_t)len);
     addResColumnInfo(pQueryInfo, exprIndex, &pExpr->base.resSchema, pExpr);
 
     tbufCloseWriter(&bw);
     taosArrayDestroy(colList);
-    tExprTreeDestroy(pNode, NULL);
   } else {
     SColumnIndex  columnIndex = {0};
 
@@ -2174,6 +2940,7 @@ static int32_t createComplexExpr(SQueryStmtInfo* pQueryInfo, int32_t exprIndex,
 //    tbufCloseWriter(&bw); // TODO there is a memory leak
   }
 
+  taosArrayDestroy(pColumnList);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -2756,6 +3523,9 @@ int32_t qParserValidateSqlNode(struct SCatalog* pCatalog, SSqlInfo* pInfo, SQuer
     validateSqlNode(p, pQueryInfo, &buf);
   }
 
+  SArray* functionList = extractFunctionIdList(pQueryInfo->exprList);
+  extractFunctionDesc(functionList, &pQueryInfo->info);
+
   if ((code = checkForInvalidExpr(pQueryInfo, &buf)) != TSDB_CODE_SUCCESS) {
     return code;
   }

source/libs/parser/src/parser.c

@@ -185,4 +185,13 @@ int32_t qParserExtractRequestedMetaInfo(const SSqlInfo* pSqlInfo, SMetaReq* pMet
   }
 
   return code;
-}
+}
\ No newline at end of file
+
+void qParserClearupMetaRequestInfo(SMetaReq* pMetaReq) {
+  if (pMetaReq == NULL) {
+    return;
+  }
+
+  taosArrayDestroy(pMetaReq->pTableName);
+  taosArrayDestroy(pMetaReq->pUdf);
+}

source/libs/parser/src/parserUtil.c

@@ -576,13 +576,6 @@ TAOS_FIELD* getFieldInfo(SFieldInfo* pFieldInfo, int32_t index) {
   return &((SInternalField*)TARRAY_GET_ELEM(pFieldInfo->internalField, index))->field;
 }
 
-int16_t getFieldInfoOffset(SQueryStmtInfo* pQueryInfo, int32_t index) {
-  SInternalField* pInfo = getInternalField(&pQueryInfo->fieldsInfo, index);
-  assert(pInfo != NULL && pInfo->pExpr->pExpr == NULL);
-  return 0;
-//  return pInfo->pExpr->base.offset;
-}
-
 int32_t fieldInfoCompare(const SFieldInfo* pFieldInfo1, const SFieldInfo* pFieldInfo2, int32_t *diffSize) {
   assert(pFieldInfo1 != NULL && pFieldInfo2 != NULL);
 
@@ -780,8 +773,8 @@ SColumn* columnListInsert(SArray* pColumnList, int32_t columnIndex, uint64_t uid
 }
 
 SColumn* insertPrimaryTsColumn(SArray* pColumnList, uint64_t tableUid) {
-  SSchema s = {.type = TSDB_DATA_TYPE_TIMESTAMP, .bytes = TSDB_KEYSIZE, .colId = PRIMARYKEY_TIMESTAMP_COL_INDEX};
-  return columnListInsert(pColumnList, PRIMARYKEY_TIMESTAMP_COL_INDEX, tableUid, &s);
+  SSchema s = {.type = TSDB_DATA_TYPE_TIMESTAMP, .bytes = TSDB_KEYSIZE, .colId = PRIMARYKEY_TIMESTAMP_COL_ID};
+  return columnListInsert(pColumnList, PRIMARYKEY_TIMESTAMP_COL_ID, tableUid, &s);
 }
 
 void columnCopy(SColumn* pDest, const SColumn* pSrc);

source/libs/parser/src/queryInfoUtil.c

 #include "queryInfoUtil.h"
+#include <function.h>
 #include "astGenerator.h"
 #include "function.h"
 #include "os.h"
@@ -55,7 +56,6 @@ SSchema* getTableTagSchema(const STableMeta* pTableMeta) {
 }
 
 static tExprNode* createUnaryFunctionExprNode(int32_t functionId, SSchema* pSchema, tExprNode* pColumnNode) {
-
   if (pColumnNode == NULL) {
     pColumnNode = calloc(1, sizeof(tExprNode));
     pColumnNode->nodeType = TEXPR_COL_NODE;
@@ -167,6 +167,10 @@ SExprInfo* getExprInfo(SQueryStmtInfo* pQueryInfo, int32_t index) {
 
 void destroyExprInfo(SExprInfo* pExprInfo) {
   tExprTreeDestroy(pExprInfo->pExpr, NULL);
+
+  for(int32_t i = 0; i < pExprInfo->base.numOfParams; ++i) {
+    taosVariantDestroy(&pExprInfo->base.param[i]);
+  }
   tfree(pExprInfo);
 }
 
@@ -192,6 +196,11 @@ void addExprInfoParam(SSqlExpr* pExpr, char* argument, int32_t type, int32_t byt
   assert(pExpr->numOfParams <= 3);
 }
 
+int32_t getExprFunctionId(SExprInfo *pExprInfo) {
+  assert(pExprInfo != NULL && pExprInfo->pExpr != NULL && pExprInfo->pExpr->nodeType == TEXPR_UNARYEXPR_NODE);
+  return pExprInfo->pExpr->_node.functionId;
+}
+
 void assignExprInfo(SExprInfo* dst, const SExprInfo* src) {
   assert(dst != NULL && src != NULL);
 
@@ -284,62 +293,11 @@ int32_t getResRowLength(SArray* pExprList) {
   return size;
 }
 
-static void freeQueryInfoImpl(SQueryStmtInfo* pQueryInfo) {
-  cleanupTagCond(&pQueryInfo->tagCond);
-  cleanupColumnCond(&pQueryInfo->colCond);
-  cleanupFieldInfo(&pQueryInfo->fieldsInfo);
-
-  dropAllExprInfo(pQueryInfo->exprList);
-  pQueryInfo->exprList = NULL;
-
-  if (pQueryInfo->exprList1 != NULL) {
-    dropAllExprInfo(pQueryInfo->exprList1);
-    pQueryInfo->exprList1 = NULL;
-  }
-
-  columnListDestroy(pQueryInfo->colList);
-  pQueryInfo->colList = NULL;
-
-  if (pQueryInfo->groupbyExpr.columnInfo != NULL) {
-    taosArrayDestroy(pQueryInfo->groupbyExpr.columnInfo);
-    pQueryInfo->groupbyExpr.columnInfo = NULL;
-  }
-
-  pQueryInfo->fillType = 0;
-
-  tfree(pQueryInfo->fillVal);
-  tfree(pQueryInfo->buf);
-
-  taosArrayDestroy(pQueryInfo->pUpstream);
-  pQueryInfo->pUpstream = NULL;
-  pQueryInfo->bufLen = 0;
-}
-
-void freeQueryInfo(SQueryStmtInfo* pQueryInfo, bool removeCachedMeta, uint64_t id) {
-  while(pQueryInfo != NULL) {
-    SQueryStmtInfo* p = pQueryInfo->sibling;
-
-    size_t numOfUpstream = taosArrayGetSize(pQueryInfo->pUpstream);
-    for(int32_t i = 0; i < numOfUpstream; ++i) {
-      SQueryStmtInfo* pUpQueryInfo = taosArrayGetP(pQueryInfo->pUpstream, i);
-      freeQueryInfoImpl(pUpQueryInfo);
-      clearAllTableMetaInfo(pUpQueryInfo, removeCachedMeta, id);
-      tfree(pUpQueryInfo);
-    }
-
-    freeQueryInfoImpl(pQueryInfo);
-    clearAllTableMetaInfo(pQueryInfo, removeCachedMeta, id);
-
-    tfree(pQueryInfo);
-    pQueryInfo = p;
-  }
-}
-
 SArray* extractFunctionIdList(SArray* pExprInfoList) {
   assert(pExprInfoList != NULL);
 
   size_t len = taosArrayGetSize(pExprInfoList);
-  SArray* p = taosArrayInit(len, sizeof(int16_t));
+  SArray* p = taosArrayInit(len, sizeof(int32_t));
   for(int32_t i = 0; i < len; ++i) {
     SExprInfo* pExprInfo = taosArrayGetP(pExprInfoList, i);
     taosArrayPush(p, &pExprInfo->pExpr->_node.functionId);

source/libs/parser/test/CMakeLists.txt

@@ -5,14 +5,14 @@ MESSAGE(STATUS "build parser unit test")
 SET(CMAKE_CXX_STANDARD 11)
 AUX_SOURCE_DIRECTORY(${CMAKE_CURRENT_SOURCE_DIR} SOURCE_LIST)
 
-ADD_EXECUTABLE(astTest ${SOURCE_LIST})
+ADD_EXECUTABLE(parserTest ${SOURCE_LIST})
 TARGET_LINK_LIBRARIES(
-        astTest
-        PUBLIC os util common parser catalog transport gtest
+        parserTest
+        PUBLIC os util common parser catalog transport gtest function
 )
 
 TARGET_INCLUDE_DIRECTORIES(
-        astTest
+        parserTest
         PUBLIC "${CMAKE_SOURCE_DIR}/include/libs/parser/"
         PRIVATE "${CMAKE_SOURCE_DIR}/source/libs/parser/inc"
 )

source/libs/parser/test/parserTests.cpp

@@ -13,6 +13,7 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <function.h>
 #include <gtest/gtest.h>
 #include <iostream>
 #pragma GCC diagnostic ignored "-Wwrite-strings"
@@ -65,61 +66,64 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 }
 }
 
-//TEST(testCase, validateAST_test) {
-//  SSqlInfo info1 = doGenerateAST("select a a1111, a+b + 22, tbname from `t.1abc` where ts<now+2h and `col` < 20 + 99");
-//  ASSERT_EQ(info1.valid, true);
-//
-//  char    msg[128] = {0};
-//  SMsgBuf buf;
-//  buf.len = 128;
-//  buf.buf = msg;
-//
-//  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
-//  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &buf);
-//  ASSERT_EQ(code, 0);
-//
-//  SMetaReq req = {0};
-//  int32_t  ret = qParserExtractRequestedMetaInfo(&info1, &req, msg, 128);
-//  ASSERT_EQ(ret, 0);
-//  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
-//
-//  SQueryStmtInfo* pQueryInfo = (SQueryStmtInfo*)calloc(1, sizeof(SQueryStmtInfo));
-//  initQueryInfo(pQueryInfo);
-//  setTableMetaInfo(pQueryInfo, &req);
-//
-//  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
-//  ret = validateSqlNode(pSqlNode, pQueryInfo, &buf);
-//
-//  SArray* pExprList = pQueryInfo->exprList;
-//  ASSERT_EQ(taosArrayGetSize(pExprList), 3);
-//
-//  SExprInfo* p1 = (SExprInfo*) taosArrayGetP(pExprList, 0);
-//  ASSERT_EQ(p1->base.uid, 110);
-//  ASSERT_EQ(p1->base.numOfParams, 0);
-//  ASSERT_EQ(p1->base.resSchema.type, TSDB_DATA_TYPE_INT);
-//  ASSERT_STRCASEEQ(p1->base.resSchema.name, "a1111");
-//  ASSERT_STRCASEEQ(p1->base.colInfo.name, "t.1abc.a");
+TEST(testCase, validateAST_test) {
+  SSqlInfo info1 = doGenerateAST("select a a1111, a+b + 22, tbname from `t.1abc` where ts<now+2h and `col` < 20 + 99");
+  ASSERT_EQ(info1.valid, true);
+
+  char    msg[128] = {0};
+  SMsgBuf buf;
+  buf.len = 128;
+  buf.buf = msg;
+
+  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
+  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &buf);
+  ASSERT_EQ(code, 0);
+
+  SMetaReq req = {0};
+  int32_t  ret = qParserExtractRequestedMetaInfo(&info1, &req, msg, 128);
+  ASSERT_EQ(ret, 0);
+  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
+
+  SQueryStmtInfo* pQueryInfo = createQueryInfo();
+  setTableMetaInfo(pQueryInfo, &req);
+
+  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
+  ret = validateSqlNode(pSqlNode, pQueryInfo, &buf);
+
+  SArray* pExprList = pQueryInfo->exprList;
+  ASSERT_EQ(taosArrayGetSize(pExprList), 3);
+
+  SExprInfo* p1 = (SExprInfo*) taosArrayGetP(pExprList, 0);
+  ASSERT_EQ(p1->base.uid, 110);
+  ASSERT_EQ(p1->base.numOfParams, 0);
+  ASSERT_EQ(p1->base.resSchema.type, TSDB_DATA_TYPE_INT);
+  ASSERT_STRCASEEQ(p1->base.resSchema.name, "a1111");
+  ASSERT_STRCASEEQ(p1->base.colInfo.name, "t.1abc.a");
+  ASSERT_EQ(p1->base.colInfo.colId, 1);
+  ASSERT_EQ(p1->base.colInfo.flag, TSDB_COL_NORMAL);
+  ASSERT_STRCASEEQ(p1->base.token, "a");
+
+  ASSERT_EQ(taosArrayGetSize(pExprList), 3);
+
+  SExprInfo* p2 = (SExprInfo*) taosArrayGetP(pExprList, 1);
+  ASSERT_EQ(p2->base.uid, 0);
+  ASSERT_EQ(p2->base.numOfParams, 1);  // it is the serialized binary string of expression.
+  ASSERT_EQ(p2->base.resSchema.type, TSDB_DATA_TYPE_DOUBLE);
+  ASSERT_STRCASEEQ(p2->base.resSchema.name, "a+b + 22");
+
+//  ASSERT_STRCASEEQ(p2->base.colInfo.name, "t.1abc.a");
 //  ASSERT_EQ(p1->base.colInfo.colId, 1);
 //  ASSERT_EQ(p1->base.colInfo.flag, TSDB_COL_NORMAL);
-//  ASSERT_STRCASEEQ(p1->base.token, "a");
-//
-//  ASSERT_EQ(taosArrayGetSize(pExprList), 3);
-//
-//  SExprInfo* p2 = (SExprInfo*) taosArrayGetP(pExprList, 1);
-//  ASSERT_EQ(p2->base.uid, 0);
-//  ASSERT_EQ(p2->base.numOfParams, 1);  // it is the serialized binary string of expression.
-//  ASSERT_EQ(p2->base.resSchema.type, TSDB_DATA_TYPE_DOUBLE);
-//  ASSERT_STRCASEEQ(p2->base.resSchema.name, "a+b + 22");
-//
-////  ASSERT_STRCASEEQ(p2->base.colInfo.name, "t.1abc.a");
-////  ASSERT_EQ(p1->base.colInfo.colId, 1);
-////  ASSERT_EQ(p1->base.colInfo.flag, TSDB_COL_NORMAL);
-//  ASSERT_STRCASEEQ(p2->base.token, "a+b + 22");
-//
-//  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 3);
-//  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 3);
-//}
-//
+  ASSERT_STRCASEEQ(p2->base.token, "a+b + 22");
+
+  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 3);
+  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 3);
+
+  destroyQueryInfo(pQueryInfo);
+  qParserClearupMetaRequestInfo(&req);
+  destroySqlInfo(&info1);
+}
+
 //TEST(testCase, function_Test) {
 //  SSqlInfo info1 = doGenerateAST("select count(a) from `t.1abc`");
 //  ASSERT_EQ(info1.valid, true);
@@ -138,8 +142,7 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //  ASSERT_EQ(ret, 0);
 //  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
 //
-//  SQueryStmtInfo* pQueryInfo = (SQueryStmtInfo*)calloc(1, sizeof(SQueryStmtInfo));
-//  initQueryInfo(pQueryInfo);
+//  SQueryStmtInfo* pQueryInfo = createQueryInfo();
 //  setTableMetaInfo(pQueryInfo, &req);
 //
 //  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
@@ -161,6 +164,10 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //
 //  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 2);
 //  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 1);
+//
+//  destroyQueryInfo(pQueryInfo);
+//  qParserClearupMetaRequestInfo(&req);
+//  destroySqlInfo(&info1);
 //}
 //
 //TEST(testCase, function_Test2) {
@@ -181,8 +188,7 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //  ASSERT_EQ(ret, 0);
 //  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
 //
-//  SQueryStmtInfo* pQueryInfo = (SQueryStmtInfo*)calloc(1, sizeof(SQueryStmtInfo));
-//  initQueryInfo(pQueryInfo);
+//  SQueryStmtInfo* pQueryInfo = createQueryInfo();
 //  setTableMetaInfo(pQueryInfo, &req);
 //
 //  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
@@ -204,6 +210,10 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //
 //  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 2);
 //  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 1);
+//
+//  destroyQueryInfo(pQueryInfo);
+//  qParserClearupMetaRequestInfo(&req);
+//  destroySqlInfo(&info1);
 //}
 //
 //TEST(testCase, function_Test3) {
@@ -224,8 +234,7 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //  ASSERT_EQ(ret, 0);
 //  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
 //
-//  SQueryStmtInfo* pQueryInfo = (SQueryStmtInfo*)calloc(1, sizeof(SQueryStmtInfo));
-//  initQueryInfo(pQueryInfo);
+//  SQueryStmtInfo* pQueryInfo = createQueryInfo();
 //  setTableMetaInfo(pQueryInfo, &req);
 //
 //  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
@@ -246,6 +255,10 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //  ASSERT_EQ(p1->base.interBytes, 24);
 //
 //  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 4);
+//
+//  destroyQueryInfo(pQueryInfo);
+//  qParserClearupMetaRequestInfo(&req);
+//  destroySqlInfo(&info1);
 //}
 //
 //TEST(testCase, function_Test4) {
@@ -266,8 +279,7 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //  ASSERT_EQ(ret, 0);
 //  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
 //
-//  SQueryStmtInfo* pQueryInfo = (SQueryStmtInfo*)calloc(1, sizeof(SQueryStmtInfo));
-//  initQueryInfo(pQueryInfo);
+//  SQueryStmtInfo* pQueryInfo = createQueryInfo();
 //  setTableMetaInfo(pQueryInfo, &req);
 //
 //  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
@@ -289,6 +301,10 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //
 //  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 1);
 //  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 1);
+//
+//  destroyQueryInfo(pQueryInfo);
+//  qParserClearupMetaRequestInfo(&req);
+//  destroySqlInfo(&info1);
 //}
 //
 //TEST(testCase, function_Test5) {
@@ -309,8 +325,7 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //  ASSERT_EQ(ret, 0);
 //  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
 //
-//  SQueryStmtInfo* pQueryInfo = (SQueryStmtInfo*)calloc(1, sizeof(SQueryStmtInfo));
-//  initQueryInfo(pQueryInfo);
+//  SQueryStmtInfo* pQueryInfo = createQueryInfo();
 //  setTableMetaInfo(pQueryInfo, &req);
 //
 //  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
@@ -333,46 +348,63 @@ void setTableMetaInfo(SQueryStmtInfo* pQueryInfo, SMetaReq *req) {
 //
 //  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 3);
 //  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 1);
+//
+//  destroyQueryInfo(pQueryInfo);
+//  qParserClearupMetaRequestInfo(&req);
+//  destroySqlInfo(&info1);
 //}
-
-TEST(testCase, function_Test6) {
-  SSqlInfo info1 = doGenerateAST("select sum(a+b) as a1, first(b*a) from `t.1abc`");
-  ASSERT_EQ(info1.valid, true);
-
-  char    msg[128] = {0};
-  SMsgBuf buf;
-  buf.len = 128;
-  buf.buf = msg;
-
-  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
-  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &buf);
-  ASSERT_EQ(code, 0);
-
-  SMetaReq req = {0};
-  int32_t  ret = qParserExtractRequestedMetaInfo(&info1, &req, msg, 128);
-  ASSERT_EQ(ret, 0);
-  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
-
-  SQueryStmtInfo* pQueryInfo = (SQueryStmtInfo*)calloc(1, sizeof(SQueryStmtInfo));
-  initQueryInfo(pQueryInfo);
-  setTableMetaInfo(pQueryInfo, &req);
-
-  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
-  ret = validateSqlNode(pSqlNode, pQueryInfo, &buf);
-  ASSERT_EQ(ret, 0);
-
-  SArray* pExprList = pQueryInfo->exprList;
-  ASSERT_EQ(taosArrayGetSize(pExprList), 2);
-
-  SExprInfo* p1 = (SExprInfo*) taosArrayGetP(pExprList, 0);
-  ASSERT_EQ(p1->base.uid, 110);
-  ASSERT_EQ(p1->base.numOfParams, 0);
-  ASSERT_EQ(p1->base.resSchema.type, TSDB_DATA_TYPE_DOUBLE);
-  ASSERT_STRCASEEQ(p1->base.resSchema.name, "a1");
-  ASSERT_EQ(p1->base.colInfo.flag, TSDB_COL_NORMAL);
-  ASSERT_STRCASEEQ(p1->base.token, "sum(a+b)");
-  ASSERT_EQ(p1->base.interBytes, 16);
-
-  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 3);
-  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 2);
-}
+//
+//TEST(testCase, function_Test6) {
+//  SSqlInfo info1 = doGenerateAST("select sum(a+b) as a1, first(b*a) from `t.1abc` interval(10s, 1s)");
+//  ASSERT_EQ(info1.valid, true);
+//
+//  char    msg[128] = {0};
+//  SMsgBuf buf;
+//  buf.len = 128;
+//  buf.buf = msg;
+//
+//  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
+//  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &buf);
+//  ASSERT_EQ(code, 0);
+//
+//  SMetaReq req = {0};
+//  int32_t  ret = qParserExtractRequestedMetaInfo(&info1, &req, msg, 128);
+//  ASSERT_EQ(ret, 0);
+//  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
+//
+//  SQueryStmtInfo* pQueryInfo = createQueryInfo();
+//  setTableMetaInfo(pQueryInfo, &req);
+//
+//  SSqlNode* pSqlNode = (SSqlNode*)taosArrayGetP(info1.list, 0);
+//  ret = validateSqlNode(pSqlNode, pQueryInfo, &buf);
+//  ASSERT_EQ(ret, 0);
+//
+//  SArray* pExprList = pQueryInfo->exprList;
+//  ASSERT_EQ(taosArrayGetSize(pExprList), 2);
+//
+//  SExprInfo* p1 = (SExprInfo*) taosArrayGetP(pExprList, 0);
+//  ASSERT_EQ(p1->base.uid, 110);
+//  ASSERT_EQ(p1->base.numOfParams, 0);
+//  ASSERT_EQ(p1->base.resSchema.type, TSDB_DATA_TYPE_DOUBLE);
+//  ASSERT_STRCASEEQ(p1->base.resSchema.name, "a1");
+//  ASSERT_EQ(p1->base.colInfo.flag, TSDB_COL_NORMAL);
+//  ASSERT_STRCASEEQ(p1->base.token, "sum(a+b)");
+//  ASSERT_EQ(p1->base.interBytes, 16);
+//  ASSERT_EQ(p1->pExpr->nodeType, TEXPR_UNARYEXPR_NODE);
+//  ASSERT_EQ(p1->pExpr->_node.functionId, FUNCTION_SUM);
+//  ASSERT_TRUE(p1->pExpr->_node.pRight == NULL);
+//
+//  tExprNode* pParam = p1->pExpr->_node.pLeft;
\ No newline at end of file
+//
+//  ASSERT_EQ(pParam->nodeType, TEXPR_BINARYEXPR_NODE);
+//  ASSERT_EQ(pParam->_node.optr, TSDB_BINARY_OP_ADD);
+//  ASSERT_EQ(pParam->_node.pLeft->nodeType, TEXPR_COL_NODE);
+//  ASSERT_EQ(pParam->_node.pRight->nodeType, TEXPR_COL_NODE);
+//
+//  ASSERT_EQ(taosArrayGetSize(pQueryInfo->colList), 3);
+//  ASSERT_EQ(pQueryInfo->fieldsInfo.numOfOutput, 2);
+//
+//  destroyQueryInfo(pQueryInfo);
+//  qParserClearupMetaRequestInfo(&req);
+//  destroySqlInfo(&info1);
+//}
\ No newline at end of file

source/libs/parser/test/tokenizerTest.cpp

@@ -667,51 +667,59 @@ TEST(testCase, isValidNumber_test) {
   EXPECT_EQ(tGetNumericStringType(&t1), TK_FLOAT);
 }
 
-TEST(testCase, generateAST_test) {
-  SSqlInfo info = doGenerateAST("select * from t1 where ts < now");
-  ASSERT_EQ(info.valid, true);
-
-  SSqlInfo info1 = doGenerateAST("select * from `t.1abc` where ts<now+2h  and col < 20+99");
-  ASSERT_EQ(info1.valid, true);
-
-  char msg[128] = {0};
-
-  SMsgBuf msgBuf = {0};
-  msgBuf.buf = msg;
-  msgBuf.len = 128;
-
-  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
-  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &msgBuf);
-  ASSERT_EQ(code, 0);
-
-  SSqlInfo info2 = doGenerateAST("select * from abc where ts<now+2");
-  SSqlNode* pNode2 = (SSqlNode*) taosArrayGetP(((SArray*)info2.list), 0);
-  code = evaluateSqlNode(pNode2, TSDB_TIME_PRECISION_MILLI, &msgBuf);
-  ASSERT_NE(code, 0);
-}
-
-TEST(testCase, evaluateAST_test) {
-  SSqlInfo info1 = doGenerateAST("select a, b+22 from `t.1abc` where ts<now+2h and `col` < 20 + 99");
-  ASSERT_EQ(info1.valid, true);
-
-  char msg[128] = {0};
-  SMsgBuf msgBuf = {0};
-  msgBuf.buf = msg;
-  msgBuf.len = 128;
-
-  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
-  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &msgBuf);
-  ASSERT_EQ(code, 0);
-}
-
-TEST(testCase, extractMeta_test) {
-  SSqlInfo info1 = doGenerateAST("select a, b+22 from `t.1abc` where ts<now+2h and `col` < 20 + 99");
-  ASSERT_EQ(info1.valid, true);
-
-  char msg[128] = {0};
-  SMetaReq req  = {0};
-  int32_t ret = qParserExtractRequestedMetaInfo(&info1, &req, msg, 128);
-  ASSERT_EQ(ret, 0);
-  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
-}
+//TEST(testCase, generateAST_test) {
+//  SSqlInfo info = doGenerateAST("select * from t1 where ts < now");
+//  ASSERT_EQ(info.valid, true);
+//
+//  SSqlInfo info1 = doGenerateAST("select * from `t.1abc` where ts<now+2h  and col < 20+99");
+//  ASSERT_EQ(info1.valid, true);
+//
+//  char msg[128] = {0};
+//
+//  SMsgBuf msgBuf = {0};
+//  msgBuf.buf = msg;
+//  msgBuf.len = 128;
+//
+//  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
+//  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &msgBuf);
+//  ASSERT_EQ(code, 0);
+//
+//  SSqlInfo info2 = doGenerateAST("select * from abc where ts<now+2");
+//  SSqlNode* pNode2 = (SSqlNode*) taosArrayGetP(((SArray*)info2.list), 0);
+//  code = evaluateSqlNode(pNode2, TSDB_TIME_PRECISION_MILLI, &msgBuf);
+//  ASSERT_NE(code, 0);
+//
+//  destroySqlInfo(&info);
+//  destroySqlInfo(&info1);
+//  destroySqlInfo(&info2);
+//}
+//
+//TEST(testCase, evaluateAST_test) {
+//  SSqlInfo info1 = doGenerateAST("select a, b+22 from `t.1abc` where ts<now+2h and `col` < 20 + 99");
+//  ASSERT_EQ(info1.valid, true);
+//
+//  char msg[128] = {0};
+//  SMsgBuf msgBuf = {0};
+//  msgBuf.buf = msg;
+//  msgBuf.len = 128;
+//
+//  SSqlNode* pNode = (SSqlNode*) taosArrayGetP(((SArray*)info1.list), 0);
+//  int32_t code = evaluateSqlNode(pNode, TSDB_TIME_PRECISION_NANO, &msgBuf);
+//  ASSERT_EQ(code, 0);
+//  destroySqlInfo(&info1);
+//}
+//
+//TEST(testCase, extractMeta_test) {
+//  SSqlInfo info1 = doGenerateAST("select a, b+22 from `t.1abc` where ts<now+2h and `col` < 20 + 99");
+//  ASSERT_EQ(info1.valid, true);
+//
+//  char msg[128] = {0};
+//  SMetaReq req  = {0};
+//  int32_t ret = qParserExtractRequestedMetaInfo(&info1, &req, msg, 128);
+//  ASSERT_EQ(ret, 0);
+//  ASSERT_EQ(taosArrayGetSize(req.pTableName), 1);
+//
+//  qParserClearupMetaRequestInfo(&req);
+//  destroySqlInfo(&info1);
+//}
 

source/libs/sync/src/sync.c

@@ -15,5 +15,12 @@
 
 #include "sync.h"
 
-int32_t  syncInit() {return 0;}
-void syncCleanUp() {}
+int32_t syncInit() { return 0; }
+
\ No newline at end of file
+void syncCleanUp() {}
+
+SSyncNode* syncStart(const SSyncInfo* pInfo) { return NULL; }
+
+void syncStop(const SSyncNode* pNode) {}
+
+void syncReconfig(const SSyncNode* pNode, const SSyncCluster* pCfg) {}
\ No newline at end of file

source/libs/wal/src/wal.c

@@ -15,5 +15,22 @@
 
 #include "wal.h"
 
-int32_t  walInit() {return 0;}
-void walCleanUp() {}
+int32_t walInit() { return 0; }
+
\ No newline at end of file
+void walCleanUp() {}
+
+SWal *walOpen(char *path, SWalCfg *pCfg) { return NULL; }
+
+int32_t walAlter(SWal *pWal, SWalCfg *pCfg) { return 0; }
+
+void walClose(SWal *pWal) {}
+
+void walFsync(SWal *pWal, bool force) {}
+
+int64_t walWrite(SWal *pWal, int64_t index, void *body, int32_t bodyLen) {}
+
+int32_t walCommit(SWal *pWal, int64_t ver) { return 0; }
+
+int32_t walRollback(SWal *pWal, int64_t ver) { return 0; }
+
+int32_t walPrune(SWal *pWal, int64_t ver) { return 0; }
\ No newline at end of file

source/server/dnode/src/dnodeInt.c

@@ -37,7 +37,7 @@ EDnStat dnodeGetRunStat() { return tsDnode.runStatus; }
 
 void dnodeSetRunStat(EDnStat stat) { tsDnode.runStatus = stat; }
 
-static void dnodeReportStartup(char *name, char *desc) {
+void dnodeReportStartup(char *name, char *desc) {
   SStartupStep *startup = &tsDnode.startup;
   tstrncpy(startup->name, name, strlen(startup->name));
   tstrncpy(startup->desc, desc, strlen(startup->desc));
@@ -58,6 +58,7 @@ static int32_t dnodeInitVnode() {
   para.fp.GetDnodeEp = dnodeGetEp;
   para.fp.SendMsgToDnode = dnodeSendMsgToDnode;
   para.fp.SendMsgToMnode = dnodeSendMsgToMnode;
+  para.fp.ReportStartup = dnodeReportStartup;
 
   return vnodeInit(para);
 }

source/server/mnode/inc/mnodeInt.h

@@ -24,7 +24,7 @@ extern "C" {
 
 tmr_h     mnodeGetTimer();
 int32_t   mnodeGetDnodeId();
-char     *mnodeGetClusterId();
+int64_t   mnodeGetClusterId();
 EMnStatus mnodeGetStatus();
 
 void mnodeSendMsgToDnode(struct SRpcEpSet *epSet, struct SRpcMsg *rpcMsg);

source/server/mnode/src/mnodeTelem.c

@@ -202,12 +202,13 @@ static void mnodeSendTelemetryReport() {
     return;
   }
 
-  char clusterId[TSDB_CLUSTER_ID_LEN] = {0};
-  mnodeGetClusterId(clusterId);
+  int64_t clusterId = mnodeGetClusterId();
+  char    clusterIdStr[20] = {0};
+  snprintf(clusterIdStr, sizeof(clusterIdStr), "%" PRId64, clusterId);
 
   SBufferWriter bw = tbufInitWriter(NULL, false);
   mnodeBeginObject(&bw);
-  mnodeAddStringField(&bw, "instanceId", clusterId);
+  mnodeAddStringField(&bw, "instanceId", clusterIdStr);
   mnodeAddIntField(&bw, "reportVersion", 1);
   mnodeAddOsInfo(&bw);
   mnodeAddCpuInfo(&bw);

source/server/mnode/src/mondeInt.c

@@ -39,7 +39,7 @@
 static struct {
   int32_t  state;
   int32_t  dnodeId;
-  char     clusterId[TSDB_CLUSTER_ID_LEN];
+  int64_t  clusterId;
   tmr_h    timer;
   SMnodeFp fp;
   SSteps * steps1;
@@ -50,7 +50,7 @@ tmr_h mnodeGetTimer() { return tsMint.timer; }
 
 int32_t mnodeGetDnodeId() { return tsMint.dnodeId; }
 
-char *mnodeGetClusterId() { return tsMint.clusterId; }
+int64_t mnodeGetClusterId() { return tsMint.clusterId; }
 
 EMnStatus mnodeGetStatus() { return tsMint.state; }
 
@@ -71,12 +71,14 @@ int32_t mnodeGetStatistics(SMnodeStat *stat) { return 0; }
 static int32_t mnodeSetPara(SMnodePara para) {
   tsMint.fp = para.fp;
   tsMint.dnodeId = para.dnodeId;
-  strncpy(tsMint.clusterId, para.clusterId, TSDB_CLUSTER_ID_LEN);
+  tsMint.clusterId = para.clusterId;
 
   if (tsMint.fp.SendMsgToDnode == NULL) return -1;
   if (tsMint.fp.SendMsgToMnode == NULL) return -1;
   if (tsMint.fp.SendRedirectMsg == NULL) return -1;
+  if (tsMint.fp.GetDnodeEp == NULL) return -1;
   if (tsMint.dnodeId < 0) return -1;
+  if (tsMint.clusterId < 0) return -1;
 
   return 0;
 }
@@ -141,7 +143,7 @@ static void mnodeCleanupStep2() { taosStepCleanup(tsMint.steps2); }
 
 static bool mnodeNeedDeploy() {
   if (tsMint.dnodeId > 0) return false;
-  if (tsMint.clusterId[0] != 0) return false;
+  if (tsMint.clusterId > 0) return false;
   if (strcmp(tsFirst, tsLocalEp) != 0) return false;
   return true;
 }
@@ -154,7 +156,7 @@ int32_t mnodeDeploy() {
     tsMint.state = MN_STATUS_INIT;
   }
 
-  if (tsMint.dnodeId <= 0 || tsMint.clusterId[0] == 0) {
+  if (tsMint.dnodeId <= 0 || tsMint.clusterId <= 0) {
     mError("failed to deploy mnode since cluster not ready");
     return TSDB_CODE_MND_NOT_READY;
   }

source/server/vnode/inc/vnodeCfg.h → source/server/vnode/inc/vnodeFile.h

@@ -21,8 +21,10 @@ extern "C" {
 #endif
 #include "vnodeInt.h"
 
-int32_t vnodeReadCfg(SVnode *pVnode);
-int32_t vnodeWriteCfg(SCreateVnodeMsg *pVnodeCfg);
+int32_t vnodeReadCfg(int32_t vgId, SVnodeCfg *pCfg);
+int32_t vnodeWriteCfg(int32_t vgId, SVnodeCfg *pCfg);
+int32_t vnodeReadState(int32_t vgId, SSyncServerState *pState);
+int32_t vnodeSaveState(int32_t vgid, SSyncServerState *pState);
 
 #ifdef __cplusplus
 }

source/server/vnode/inc/vnodeInt.h

@@ -16,11 +16,12 @@
 #ifndef _TD_VNODE_INT_H_
 #define _TD_VNODE_INT_H_
 
+#include "os.h"
 #include "amalloc.h"
 #include "meta.h"
-#include "os.h"
 #include "sync.h"
 #include "taosmsg.h"
+#include "tglobal.h"
 #include "tlog.h"
 #include "tq.h"
 #include "tqueue.h"
@@ -43,57 +44,70 @@ extern int32_t vDebugFlag;
 #define vDebug(...) { if (vDebugFlag & DEBUG_DEBUG) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
 #define vTrace(...) { if (vDebugFlag & DEBUG_TRACE) { taosPrintLog("VND ", vDebugFlag, __VA_ARGS__); }}
 
+typedef struct STsdbCfg {
+  int32_t cacheBlockSize;  // MB
+  int32_t totalBlocks;
+  int32_t daysPerFile;
+  int32_t daysToKeep0;
+  int32_t daysToKeep1;
+  int32_t daysToKeep2;
+  int32_t minRowsPerFileBlock;
+  int32_t maxRowsPerFileBlock;
+  uint8_t precision;  // time resolution
+  int8_t  compression;
+  int8_t  cacheLastRow;
+  int8_t  update;
+} STsdbCfg;
+
+typedef struct SMetaCfg {
+} SMetaCfg;
+
+typedef struct SVnodeCfg {
+  char         db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN];
+  int8_t       dropped;
+  int8_t       quorum;
+  SWalCfg      wal;
+  STsdbCfg     tsdb;
+  SMetaCfg     meta;
+  SSyncCluster sync;
+} SVnodeCfg;
+
 typedef struct {
-  int32_t         vgId;      // global vnode group ID
-  int32_t         refCount;  // reference count
-  SMemAllocator  *allocator;
-  SMeta          *pMeta;
-  STsdb          *pTsdb;
-  STQ            *pTQ;
-  twalh           pWal;
-  SyncNodeId      syncNode;
-  taos_queue      pWriteQ;  // write queue
-  taos_queue      pQueryQ;  // read query queue
-  taos_queue      pFetchQ;  // read fetch/cancel queue
-  SWalCfg         walCfg;
-  SSyncCluster    syncCfg;
-  char            db[TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN];
-  int64_t         queuedWMsgSize;
-  int32_t         queuedWMsg;
-  int32_t         queuedRMsg;
-  int32_t         numOfQHandle;  // current initialized and existed query handle in current dnode
-  int8_t          status;
-  int8_t          role;
-  int8_t          accessState;
-  int8_t          dropped;
-  pthread_mutex_t statusMutex;
+  int32_t          vgId;      // global vnode group ID
+  int32_t          refCount;  // reference count
+  SMemAllocator   *allocator;
+  SMeta           *pMeta;
+  STsdb           *pTsdb;
+  STQ             *pTQ;
+  SWal            *pWal;
+  void            *pQuery;
+  SSyncNode       *pSync;
+  taos_queue       pWriteQ;  // write queue
+  taos_queue       pQueryQ;  // read query queue
+  taos_queue       pFetchQ;  // read fetch/cancel queue
+  SVnodeCfg        cfg;
+  SSyncServerState term;
+  int64_t          queuedWMsgSize;
+  int32_t          queuedWMsg;
+  int32_t          queuedRMsg;
+  int32_t          numOfQHandle;  // current initialized and existed query handle in current dnode
+  int8_t           role;
+  int8_t           accessState;
+  int8_t           dropped;
+  int8_t           status;
+  pthread_mutex_t  statusMutex;
 } SVnode;
 
 typedef struct {
   int32_t len;
-  void *  rsp;
-  void *  qhandle;  // used by query and retrieve msg
+  void   *rsp;
+  void   *qhandle;  // used by query and retrieve msg
 } SVnRsp;
 
 void vnodeSendMsgToDnode(struct SRpcEpSet *epSet, struct SRpcMsg *rpcMsg);
 void vnodeSendMsgToMnode(struct SRpcMsg *rpcMsg);
 void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port);
-
-int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg);
-int32_t vnodeDrop(int32_t vgId);
-int32_t vnodeOpen(int32_t vgId);
-int32_t vnodeAlter(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg);
-int32_t vnodeSync(int32_t vgId);
-int32_t vnodeClose(int32_t vgId);
-void    vnodeCleanUp(SVnode *pVnode);
-void    vnodeDestroy(SVnode *pVnode);
-int32_t vnodeCompact(int32_t vgId);
-void    vnodeBackup(int32_t vgId);
-void    vnodeGetStatus(struct SStatusMsg *status);
-
-SVnode *vnodeAcquire(int32_t vgId);
-SVnode *vnodeAcquireNotClose(int32_t vgId);
-void    vnodeRelease(SVnode *pVnode);
+void vnodeReportStartup(char *name, char *desc);
 
 #ifdef __cplusplus
 }

source/server/vnode/inc/vnodeMgmtMsg.h → source/server/vnode/inc/vnodeMain.h

@@ -13,23 +13,30 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef _TD_VNODE_MGMT_MSG_H_
-#define _TD_VNODE_MGMT_MSG_H_
+#ifndef _TD_VNODE_MAIN_H_
+#define _TD_VNODE_MAIN_H_
+
+#include "vnodeInt.h"
 
 #ifdef __cplusplus
 extern "C" {
 #endif
-#include "vnodeInt.h"
 
-int32_t vnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg);
-int32_t vnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg);
-int32_t vnodeProcessSyncVnodeMsg(SRpcMsg *rpcMsg);
-int32_t vnodeProcessCompactVnodeMsg(SRpcMsg *rpcMsg);
-int32_t vnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg);
-int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg);
+int32_t vnodeInitMain();
+void    vnodeCleanupMain();
+
+SVnode *vnodeAcquireInAllState(int32_t vgId);
+SVnode *vnodeAcquire(int32_t vgId);
+void    vnodeRelease(SVnode *pVnode);
+
+int32_t vnodeCreateVnode(int32_t vgId, SVnodeCfg *pCfg);
+int32_t vnodeAlterVnode(SVnode *pVnode, SVnodeCfg *pCfg);
+int32_t vnodeDropVnode(SVnode *pVnode);
+int32_t vnodeSyncVnode(SVnode *pVnode);
+int32_t vnodeCompactVnode(SVnode *pVnode);
 
 #ifdef __cplusplus
 }
 #endif
 
-#endif /*_TD_VNODE_MGMT_H_*/
+#endif /*_TD_VNODE_MAIN_H_*/

source/server/vnode/inc/vnodeMgmt.h

@@ -21,6 +21,14 @@ extern "C" {
 #endif
 #include "vnodeInt.h"
 
+
+typedef struct {
+  SVnode *pVnode;
+  SRpcMsg rpcMsg;
+  char    pCont[];
+} SVnMgmtMsg;
+
+
 int32_t vnodeInitMgmt();
 void    vnodeCleanupMgmt();
 void    vnodeProcessMgmtMsg(SRpcMsg *pMsg);

source/server/vnode/inc/vnodeStatus.h

-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef _TD_VNODE_STATUS_H_
-#define _TD_VNODE_STATUS_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-#include "vnodeInt.h"
-
-typedef enum _VN_STATUS {
-  TAOS_VN_STATUS_INIT = 0,
-  TAOS_VN_STATUS_READY = 1,
-  TAOS_VN_STATUS_CLOSING = 2,
-  TAOS_VN_STATUS_UPDATING = 3
-} EVnodeStatus;
-
-// vnodeStatus
-extern char* vnodeStatus[];
-
-bool vnodeSetInitStatus(SVnode* pVnode);
-bool vnodeSetReadyStatus(SVnode* pVnode);
-bool vnodeSetClosingStatus(SVnode* pVnode);
-bool vnodeSetUpdatingStatus(SVnode* pVnode);
-
-bool vnodeInInitStatus(SVnode* pVnode);
-bool vnodeInReadyStatus(SVnode* pVnode);
-bool vnodeInClosingStatus(SVnode* pVnode);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /*_TD_VNODE_STATUS_H_*/
\ No newline at end of file

source/server/vnode/inc/vnodeVersion.h

-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef _TD_VNODE_VERSION_H_
-#define _TD_VNODE_VERSION_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-#include "vnodeInt.h"
-
-int32_t vnodeReadVersion(SVnode *pVnode);
-int32_t vnodeSaveVersion(SVnode *pVnode);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /*_TD_VNODE_VERSION_H_*/

source/server/vnode/inc/vnodeWorker.h

-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef _TD_VNODE_WORKER_H_
-#define _TD_VNODE_WORKER_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-#include "vnodeInt.h"
-
-int32_t vnodeInitWorker();
-void    vnodeCleanupWorker();
-void    vnodeProcessCleanupTask(SVnode *pVnode);
-void    vnodeProcessDestroyTask(SVnode *pVnode);
-void    vnodeProcessBackupTask(SVnode *pVnode);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /*_TD_VNODE_WORKER_H_*/
\ No newline at end of file

source/server/vnode/src/vnodeCfg.c → source/server/vnode/src/vnodeFile.c

@@ -14,288 +14,210 @@
  */
 
 #define _DEFAULT_SOURCE
-#include "os.h"
 #include "cJSON.h"
-#include "tglobal.h"
-#include "vnodeCfg.h"
-
-static void vnodeLoadCfg(SVnode *pVnode, SCreateVnodeMsg *vnodeMsg) {
-#if 0
-  tstrncpy(pVnode->db, vnodeMsg->db, sizeof(pVnode->db));
-  pVnode->dbCfgVersion = vnodeMsg->cfg.dbCfgVersion;
-  pVnode->vgCfgVersion = vnodeMsg->cfg.vgCfgVersion;
-  pVnode->tsdbCfg.cacheBlockSize = vnodeMsg->cfg.cacheBlockSize;
-  pVnode->tsdbCfg.totalBlocks = vnodeMsg->cfg.totalBlocks;
-  pVnode->tsdbCfg.daysPerFile = vnodeMsg->cfg.daysPerFile;
-  pVnode->tsdbCfg.keep = vnodeMsg->cfg.daysToKeep;
-  pVnode->tsdbCfg.keep1 = vnodeMsg->cfg.daysToKeep1;
-  pVnode->tsdbCfg.keep2 = vnodeMsg->cfg.daysToKeep2;
-  pVnode->tsdbCfg.minRowsPerFileBlock = vnodeMsg->cfg.minRowsPerFileBlock;
-  pVnode->tsdbCfg.maxRowsPerFileBlock = vnodeMsg->cfg.maxRowsPerFileBlock;
-  pVnode->tsdbCfg.precision = vnodeMsg->cfg.precision;
-  pVnode->tsdbCfg.compression = vnodeMsg->cfg.compression;
-  pVnode->tsdbCfg.update = vnodeMsg->cfg.update;
-  pVnode->tsdbCfg.cacheLastRow = vnodeMsg->cfg.cacheLastRow;
-  pVnode->walCfg.walLevel = vnodeMsg->cfg.walLevel;
-  pVnode->walCfg.fsyncPeriod = vnodeMsg->cfg.fsyncPeriod;
-  pVnode->walCfg.keep = TAOS_WAL_NOT_KEEP;
-  pVnode->syncCfg.replica = vnodeMsg->cfg.vgReplica;
-  pVnode->syncCfg.quorum = vnodeMsg->cfg.quorum;
-  pVnode->dbReplica = vnodeMsg->cfg.dbReplica;
-  pVnode->dbType = vnodeMsg->cfg.dbType;
-
-  for (int i = 0; i < pVnode->syncCfg.replica; ++i) {
-    SVnodeDesc *node = &vnodeMsg->nodes[i];
-    pVnode->syncCfg.nodeInfo[i].nodeId = node->nodeId;
-    taosGetFqdnPortFromEp(node->nodeEp, pVnode->syncCfg.nodeInfo[i].nodeFqdn, &pVnode->syncCfg.nodeInfo[i].nodePort);
-    pVnode->syncCfg.nodeInfo[i].nodePort += TSDB_PORT_SYNC;
-  }
-
-  vInfo("vgId:%d, load vnode cfg successfully, replcia:%d", pVnode->vgId, pVnode->syncCfg.replica);
-  for (int32_t i = 0; i < pVnode->syncCfg.replica; i++) {
-    SNodeInfo *node = &pVnode->syncCfg.nodeInfo[i];
-    vInfo("vgId:%d, dnode:%d, %s:%u", pVnode->vgId, node->nodeId, node->nodeFqdn, node->nodePort);
-  }
-#endif  
-}
+#include "vnodeFile.h"
 
-int32_t vnodeReadCfg(SVnode *pVnode) {
-#if 0
+int32_t vnodeReadCfg(int32_t vgId, SVnodeCfg *pCfg) {
   int32_t ret = TSDB_CODE_VND_APP_ERROR;
   int32_t len = 0;
   int     maxLen = 1000;
-  char *  content = calloc(1, maxLen + 1);
-  cJSON * root = NULL;
-  FILE *  fp = NULL;
-  bool    nodeChanged = false;
-
-  SCreateVnodeMsg vnodeMsg;
-
-  char file[TSDB_FILENAME_LEN + 30] = {0};
-  sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, pVnode->vgId);
+  char   *content = calloc(1, maxLen + 1);
+  cJSON  *root = NULL;
+  FILE   *fp = NULL;
 
-  vnodeMsg.cfg.vgId = pVnode->vgId;
+  char file[PATH_MAX + 30] = {0};
+  sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, vgId);
 
   fp = fopen(file, "r");
   if (!fp) {
-    vError("vgId:%d, failed to open vnode cfg file:%s to read, error:%s", pVnode->vgId, file, strerror(errno));
+    vError("vgId:%d, failed to open vnode cfg file:%s to read since %s", vgId, file, strerror(errno));
     ret = TAOS_SYSTEM_ERROR(errno);
     goto PARSE_VCFG_ERROR;
   }
 
   len = (int32_t)fread(content, 1, maxLen, fp);
   if (len <= 0) {
-    vError("vgId:%d, failed to read %s, content is null", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since content is null", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
 
   content[len] = 0;
   root = cJSON_Parse(content);
   if (root == NULL) {
-    vError("vgId:%d, failed to read %s, invalid json format", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since invalid json format", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
 
   cJSON *db = cJSON_GetObjectItem(root, "db");
   if (!db || db->type != cJSON_String || db->valuestring == NULL) {
-    vError("vgId:%d, failed to read %s, db not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since db not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  tstrncpy(vnodeMsg.db, db->valuestring, sizeof(vnodeMsg.db));
+  tstrncpy(pCfg->db, db->valuestring, sizeof(pCfg->db));
 
-  cJSON *dbCfgVersion = cJSON_GetObjectItem(root, "cfgVersion");
-  if (!dbCfgVersion || dbCfgVersion->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, cfgVersion not found", pVnode->vgId, file);
+  cJSON *dropped = cJSON_GetObjectItem(root, "dropped");
+  if (!dropped || dropped->type != cJSON_Number) {
+    vError("vgId:%d, failed to read %s since dropped not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.dbCfgVersion = (int32_t)dbCfgVersion->valueint;
+  pCfg->dropped = (int32_t)dropped->valueint;
 
-  cJSON *vgCfgVersion = cJSON_GetObjectItem(root, "vgCfgVersion");
-  if (!vgCfgVersion || vgCfgVersion->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, vgCfgVersion not found", pVnode->vgId, file);
-    vnodeMsg.cfg.vgCfgVersion = 0;
-  } else {
-    vnodeMsg.cfg.vgCfgVersion = (int32_t)vgCfgVersion->valueint;
+  cJSON *quorum = cJSON_GetObjectItem(root, "quorum");
+  if (!quorum || quorum->type != cJSON_Number) {
+    vError("vgId: %d, failed to read %s, quorum not found", vgId, file);
+    goto PARSE_VCFG_ERROR;
   }
+  pCfg->quorum = (int8_t)quorum->valueint;
 
   cJSON *cacheBlockSize = cJSON_GetObjectItem(root, "cacheBlockSize");
   if (!cacheBlockSize || cacheBlockSize->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, cacheBlockSize not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since cacheBlockSize not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.cacheBlockSize = (int32_t)cacheBlockSize->valueint;
+  pCfg->tsdb.cacheBlockSize = (int32_t)cacheBlockSize->valueint;
 
   cJSON *totalBlocks = cJSON_GetObjectItem(root, "totalBlocks");
   if (!totalBlocks || totalBlocks->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, totalBlocks not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since totalBlocks not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.totalBlocks = (int32_t)totalBlocks->valueint;
+  pCfg->tsdb.totalBlocks = (int32_t)totalBlocks->valueint;
 
   cJSON *daysPerFile = cJSON_GetObjectItem(root, "daysPerFile");
   if (!daysPerFile || daysPerFile->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, daysPerFile not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since daysPerFile not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.daysPerFile = (int32_t)daysPerFile->valueint;
+  pCfg->tsdb.daysPerFile = (int32_t)daysPerFile->valueint;
 
-  cJSON *daysToKeep = cJSON_GetObjectItem(root, "daysToKeep");
-  if (!daysToKeep || daysToKeep->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, daysToKeep not found", pVnode->vgId, file);
+  cJSON *daysToKeep0 = cJSON_GetObjectItem(root, "daysToKeep0");
+  if (!daysToKeep0 || daysToKeep0->type != cJSON_Number) {
+    vError("vgId:%d, failed to read %s since daysToKeep0 not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.daysToKeep = (int32_t)daysToKeep->valueint;
+  pCfg->tsdb.daysToKeep0 = (int32_t)daysToKeep0->valueint;
 
   cJSON *daysToKeep1 = cJSON_GetObjectItem(root, "daysToKeep1");
   if (!daysToKeep1 || daysToKeep1->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, daysToKeep1 not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since daysToKeep1 not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.daysToKeep1 = (int32_t)daysToKeep1->valueint;
+  pCfg->tsdb.daysToKeep1 = (int32_t)daysToKeep1->valueint;
 
   cJSON *daysToKeep2 = cJSON_GetObjectItem(root, "daysToKeep2");
   if (!daysToKeep2 || daysToKeep2->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, daysToKeep2 not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since daysToKeep2 not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.daysToKeep2 = (int32_t)daysToKeep2->valueint;
+  pCfg->tsdb.daysToKeep2 = (int32_t)daysToKeep2->valueint;
 
   cJSON *minRowsPerFileBlock = cJSON_GetObjectItem(root, "minRowsPerFileBlock");
   if (!minRowsPerFileBlock || minRowsPerFileBlock->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, minRowsPerFileBlock not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since minRowsPerFileBlock not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.minRowsPerFileBlock = (int32_t)minRowsPerFileBlock->valueint;
+  pCfg->tsdb.minRowsPerFileBlock = (int32_t)minRowsPerFileBlock->valueint;
 
   cJSON *maxRowsPerFileBlock = cJSON_GetObjectItem(root, "maxRowsPerFileBlock");
   if (!maxRowsPerFileBlock || maxRowsPerFileBlock->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, maxRowsPerFileBlock not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since maxRowsPerFileBlock not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.maxRowsPerFileBlock = (int32_t)maxRowsPerFileBlock->valueint;
+  pCfg->tsdb.maxRowsPerFileBlock = (int32_t)maxRowsPerFileBlock->valueint;
 
   cJSON *precision = cJSON_GetObjectItem(root, "precision");
   if (!precision || precision->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, precision not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since precision not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.precision = (int8_t)precision->valueint;
+  pCfg->tsdb.precision = (int8_t)precision->valueint;
 
   cJSON *compression = cJSON_GetObjectItem(root, "compression");
   if (!compression || compression->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, compression not found", pVnode->vgId, file);
+    vError("vgId:%d, failed to read %s since compression not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.compression = (int8_t)compression->valueint;
+  pCfg->tsdb.compression = (int8_t)compression->valueint;
 
-  cJSON *walLevel = cJSON_GetObjectItem(root, "walLevel");
-  if (!walLevel || walLevel->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, walLevel not found", pVnode->vgId, file);
+  cJSON *update = cJSON_GetObjectItem(root, "update");
+  if (!update || update->type != cJSON_Number) {
+    vError("vgId: %d, failed to read %s since update not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.walLevel = (int8_t)walLevel->valueint;
+  pCfg->tsdb.update = (int8_t)update->valueint;
 
-  cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsync");
-  if (!walLevel || walLevel->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, fsyncPeriod not found", pVnode->vgId, file);
+  cJSON *cacheLastRow = cJSON_GetObjectItem(root, "cacheLastRow");
+  if (!cacheLastRow || cacheLastRow->type != cJSON_Number) {
+    vError("vgId: %d, failed to read %s since cacheLastRow not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.fsyncPeriod = (int32_t)fsyncPeriod->valueint;
+  pCfg->tsdb.cacheLastRow = (int8_t)cacheLastRow->valueint;
 
-  cJSON *wals = cJSON_GetObjectItem(root, "wals");
-  if (!wals || wals->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, wals not found", pVnode->vgId, file);
+  cJSON *walLevel = cJSON_GetObjectItem(root, "walLevel");
+  if (!walLevel || walLevel->type != cJSON_Number) {
+    vError("vgId:%d, failed to read %s since walLevel not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.wals = (int8_t)wals->valueint;
+  pCfg->wal.walLevel = (int8_t)walLevel->valueint;
 
-  cJSON *vgReplica = cJSON_GetObjectItem(root, "replica");
-  if (!vgReplica || vgReplica->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, replica not found", pVnode->vgId, file);
+  cJSON *fsyncPeriod = cJSON_GetObjectItem(root, "fsyncPeriod");
+  if (!walLevel || walLevel->type != cJSON_Number) {
+    vError("vgId:%d, failed to read %s since fsyncPeriod not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.vgReplica = (int8_t)vgReplica->valueint;
-
-  cJSON *dbReplica = cJSON_GetObjectItem(root, "dbReplica");
-  if (!dbReplica || dbReplica->type != cJSON_Number) {
-    vWarn("vgId:%d, failed to read %s, dbReplica not found", pVnode->vgId, file);
-    vnodeMsg.cfg.dbReplica = vnodeMsg.cfg.vgReplica;
-    vnodeMsg.cfg.vgCfgVersion = 0;
-  } else {
-    vnodeMsg.cfg.dbReplica = (int8_t)dbReplica->valueint;
-  }
+  pCfg->wal.fsyncPeriod = (int32_t)fsyncPeriod->valueint;
 
-  cJSON *quorum = cJSON_GetObjectItem(root, "quorum");
-  if (!quorum || quorum->type != cJSON_Number) {
-    vError("vgId: %d, failed to read %s, quorum not found", pVnode->vgId, file);
+  cJSON *selfIndex = cJSON_GetObjectItem(root, "selfIndex");
+  if (!selfIndex || selfIndex->type != cJSON_Number) {
+    vError("vgId:%d, failed to read %s since selfIndex not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
-  vnodeMsg.cfg.quorum = (int8_t)quorum->valueint;
-
-  cJSON *update = cJSON_GetObjectItem(root, "update");
-  if (!update || update->type != cJSON_Number) {
-    vWarn("vgId: %d, failed to read %s, update not found", pVnode->vgId, file);
-    vnodeMsg.cfg.update = 0;
-    vnodeMsg.cfg.vgCfgVersion = 0;
-  } else {
-    vnodeMsg.cfg.update = (int8_t)update->valueint;
-  }
+  pCfg->sync.selfIndex = selfIndex->valueint;
 
-  cJSON *cacheLastRow = cJSON_GetObjectItem(root, "cacheLastRow");
-  if (!cacheLastRow || cacheLastRow->type != cJSON_Number) {
-    vWarn("vgId: %d, failed to read %s, cacheLastRow not found", pVnode->vgId, file);
-    vnodeMsg.cfg.cacheLastRow = 0;
-    vnodeMsg.cfg.vgCfgVersion = 0;
-  } else {
-    vnodeMsg.cfg.cacheLastRow = (int8_t)cacheLastRow->valueint;
-  }
-
-  cJSON *dbType = cJSON_GetObjectItem(root, "dbType");
-  if (!dbType || dbType->type != cJSON_Number) {
-    vWarn("vgId: %d, failed to read %s, dbType not found", pVnode->vgId, file);
-    vnodeMsg.cfg.dbType = 0;
-  } else {
-    vnodeMsg.cfg.dbType = (int8_t)dbType->valueint;
+  cJSON *replica = cJSON_GetObjectItem(root, "replica");
+  if (!replica || replica->type != cJSON_Number) {
+    vError("vgId:%d, failed to read %s since replica not found", vgId, file);
+    goto PARSE_VCFG_ERROR;
   }
+  pCfg->sync.replica = replica->valueint;
 
-  cJSON *nodeInfos = cJSON_GetObjectItem(root, "nodeInfos");
-  if (!nodeInfos || nodeInfos->type != cJSON_Array) {
-    vError("vgId:%d, failed to read %s, nodeInfos not found", pVnode->vgId, file);
+  cJSON *nodes = cJSON_GetObjectItem(root, "nodes");
+  if (!nodes || nodes->type != cJSON_Array) {
+    vError("vgId:%d, failed to read %s, nodes not found", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
 
-  int size = cJSON_GetArraySize(nodeInfos);
-  if (size != vnodeMsg.cfg.vgReplica) {
-    vError("vgId:%d, failed to read %s, nodeInfos size not matched", pVnode->vgId, file);
+  int size = cJSON_GetArraySize(nodes);
+  if (size != pCfg->sync.replica) {
+    vError("vgId:%d, failed to read %s since nodes size not matched", vgId, file);
     goto PARSE_VCFG_ERROR;
   }
 
   for (int i = 0; i < size; ++i) {
-    cJSON *nodeInfo = cJSON_GetArrayItem(nodeInfos, i);
+    cJSON *nodeInfo = cJSON_GetArrayItem(nodes, i);
     if (nodeInfo == NULL) continue;
-    SVnodeDesc *node = &vnodeMsg.nodes[i];
+    SNodeInfo *node = &pCfg->sync.nodeInfo[i];
 
-    cJSON *nodeId = cJSON_GetObjectItem(nodeInfo, "nodeId");
+    cJSON *nodeId = cJSON_GetObjectItem(nodeInfo, "id");
     if (!nodeId || nodeId->type != cJSON_Number) {
-      vError("vgId:%d, failed to read %s, nodeId not found", pVnode->vgId, file);
+      vError("vgId:%d, failed to read %s since nodeId not found", vgId, file);
       goto PARSE_VCFG_ERROR;
     }
-    node->nodeId = (int32_t)nodeId->valueint;
+    node->nodeId = nodeId->valueint;
 
-    cJSON *nodeEp = cJSON_GetObjectItem(nodeInfo, "nodeEp");
-    if (!nodeEp || nodeEp->type != cJSON_String || nodeEp->valuestring == NULL) {
-      vError("vgId:%d, failed to read %s, nodeFqdn not found", pVnode->vgId, file);
+    cJSON *nodePort = cJSON_GetObjectItem(nodeInfo, "port");
+    if (!nodePort || nodePort->type != cJSON_Number) {
+      vError("vgId:%d, failed to read %s sincenodePort not found", vgId, file);
       goto PARSE_VCFG_ERROR;
     }
-    tstrncpy(node->nodeEp, nodeEp->valuestring, TSDB_EP_LEN);
-
-    char nodeEpStr[TSDB_EP_LEN];
-    vnodeGetDnodeEp(node->nodeId, nodeEpStr, NULL, NULL);
-    bool changed = (strcmp(node->nodeEp, nodeEpStr) != 0);
-    if (changed) {
-      tstrncpy(node->nodeEp, nodeEpStr, TSDB_EP_LEN);
-      nodeChanged = changed;
+    node->nodePort = (uint16_t)nodePort->valueint;
+
+    cJSON *nodeFqdn = cJSON_GetObjectItem(nodeInfo, "fqdn");
+    if (!nodeFqdn || nodeFqdn->type != cJSON_String || nodeFqdn->valuestring == NULL) {
+      vError("vgId:%d, failed to read %s since nodeFqdn not found", vgId, file);
+      goto PARSE_VCFG_ERROR;
     }
+    tstrncpy(node->nodeFqdn, nodeFqdn->valuestring, TSDB_FQDN_LEN);
   }
 
   ret = TSDB_CODE_SUCCESS;
@@ -305,63 +227,58 @@ PARSE_VCFG_ERROR:
   if (root != NULL) cJSON_Delete(root);
   if (fp != NULL) fclose(fp);
 
-  if (nodeChanged) {
-    vnodeWriteCfg(&vnodeMsg);
-  }
-
-  if (ret == TSDB_CODE_SUCCESS) {
-    vnodeLoadCfg(pVnode, &vnodeMsg);
-  }
-
   terrno = 0;
   return ret;
 }
 
-int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) {
-  char file[TSDB_FILENAME_LEN + 30] = {0};
-  sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, pMsg->cfg.vgId);
+int32_t vnodeWriteCfg(int32_t vgId, SVnodeCfg *pCfg) {
+  int32_t code = 0;
+  char    file[PATH_MAX + 30] = {0};
+  sprintf(file, "%s/vnode%d/config.json", tsVnodeDir, vgId);
 
   FILE *fp = fopen(file, "w");
   if (!fp) {
-    vError("vgId:%d, failed to write %s error:%s", pMsg->cfg.vgId, file, strerror(errno));
+    vError("vgId:%d, failed to write %s error:%s", vgId, file, strerror(errno));
     terrno = TAOS_SYSTEM_ERROR(errno);
     return terrno;
   }
 
   int32_t len = 0;
   int32_t maxLen = 1000;
-  char *  content = calloc(1, maxLen + 1);
+  char   *content = calloc(1, maxLen + 1);
 
   len += snprintf(content + len, maxLen - len, "{\n");
-  len += snprintf(content + len, maxLen - len, "  \"db\": \"%s\",\n", pMsg->db);
-  len += snprintf(content + len, maxLen - len, "  \"cfgVersion\": %d,\n", pMsg->cfg.dbCfgVersion);
-  len += snprintf(content + len, maxLen - len, "  \"vgCfgVersion\": %d,\n", pMsg->cfg.vgCfgVersion);
-  len += snprintf(content + len, maxLen - len, "  \"cacheBlockSize\": %d,\n", pMsg->cfg.cacheBlockSize);
-  len += snprintf(content + len, maxLen - len, "  \"totalBlocks\": %d,\n", pMsg->cfg.totalBlocks);
-  len += snprintf(content + len, maxLen - len, "  \"daysPerFile\": %d,\n", pMsg->cfg.daysPerFile);
-  len += snprintf(content + len, maxLen - len, "  \"daysToKeep\": %d,\n", pMsg->cfg.daysToKeep);
-  len += snprintf(content + len, maxLen - len, "  \"daysToKeep1\": %d,\n", pMsg->cfg.daysToKeep1);
-  len += snprintf(content + len, maxLen - len, "  \"daysToKeep2\": %d,\n", pMsg->cfg.daysToKeep2);
-  len += snprintf(content + len, maxLen - len, "  \"minRowsPerFileBlock\": %d,\n", pMsg->cfg.minRowsPerFileBlock);
-  len += snprintf(content + len, maxLen - len, "  \"maxRowsPerFileBlock\": %d,\n", pMsg->cfg.maxRowsPerFileBlock);
-  len += snprintf(content + len, maxLen - len, "  \"precision\": %d,\n", pMsg->cfg.precision);
-  len += snprintf(content + len, maxLen - len, "  \"compression\": %d,\n", pMsg->cfg.compression);
-  len += snprintf(content + len, maxLen - len, "  \"walLevel\": %d,\n", pMsg->cfg.walLevel);
-  len += snprintf(content + len, maxLen - len, "  \"fsync\": %d,\n", pMsg->cfg.fsyncPeriod);
-  len += snprintf(content + len, maxLen - len, "  \"replica\": %d,\n", pMsg->cfg.vgReplica);
-  len += snprintf(content + len, maxLen - len, "  \"dbReplica\": %d,\n", pMsg->cfg.dbReplica);
-  len += snprintf(content + len, maxLen - len, "  \"wals\": %d,\n", pMsg->cfg.wals);
-  len += snprintf(content + len, maxLen - len, "  \"quorum\": %d,\n", pMsg->cfg.quorum);
-  len += snprintf(content + len, maxLen - len, "  \"update\": %d,\n", pMsg->cfg.update);
-  len += snprintf(content + len, maxLen - len, "  \"cacheLastRow\": %d,\n", pMsg->cfg.cacheLastRow);
-  len += snprintf(content + len, maxLen - len, "  \"dbType\": %d,\n", pMsg->cfg.dbType);
-  len += snprintf(content + len, maxLen - len, "  \"nodeInfos\": [{\n");
-  for (int32_t i = 0; i < pMsg->cfg.vgReplica; i++) {
-    SVnodeDesc *node = &pMsg->nodes[i];
-    vnodeGetDnodeEp(node->nodeId, node->nodeEp, NULL, NULL);
-    len += snprintf(content + len, maxLen - len, "    \"nodeId\": %d,\n", node->nodeId);
-    len += snprintf(content + len, maxLen - len, "    \"nodeEp\": \"%s\"\n", node->nodeEp);
-    if (i < pMsg->cfg.vgReplica - 1) {
+  // vnode
+  len += snprintf(content + len, maxLen - len, "  \"vgId\": %d,\n", vgId);
+  len += snprintf(content + len, maxLen - len, "  \"db\": \"%s\",\n", pCfg->db);
+  len += snprintf(content + len, maxLen - len, "  \"dropped\": %d,\n", pCfg->dropped);
+  len += snprintf(content + len, maxLen - len, "  \"quorum\": %d,\n", pCfg->quorum);
+  // tsdb
+  len += snprintf(content + len, maxLen - len, "  \"cacheBlockSize\": %d,\n", pCfg->tsdb.cacheBlockSize);
+  len += snprintf(content + len, maxLen - len, "  \"totalBlocks\": %d,\n", pCfg->tsdb.totalBlocks);
+  len += snprintf(content + len, maxLen - len, "  \"daysPerFile\": %d,\n", pCfg->tsdb.daysPerFile);
+  len += snprintf(content + len, maxLen - len, "  \"daysToKeep0\": %d,\n", pCfg->tsdb.daysToKeep0);
+  len += snprintf(content + len, maxLen - len, "  \"daysToKeep1\": %d,\n", pCfg->tsdb.daysToKeep1);
+  len += snprintf(content + len, maxLen - len, "  \"daysToKeep2\": %d,\n", pCfg->tsdb.daysToKeep2);
+  len += snprintf(content + len, maxLen - len, "  \"minRowsPerFileBlock\": %d,\n", pCfg->tsdb.minRowsPerFileBlock);
+  len += snprintf(content + len, maxLen - len, "  \"maxRowsPerFileBlock\": %d,\n", pCfg->tsdb.maxRowsPerFileBlock);
+  len += snprintf(content + len, maxLen - len, "  \"precision\": %d,\n", pCfg->tsdb.precision);
+  len += snprintf(content + len, maxLen - len, "  \"compression\": %d,\n", pCfg->tsdb.compression);
+  len += snprintf(content + len, maxLen - len, "  \"cacheLastRow\": %d,\n", pCfg->tsdb.cacheLastRow);
+  len += snprintf(content + len, maxLen - len, "  \"update\": %d,\n", pCfg->tsdb.update);
+  // wal
+  len += snprintf(content + len, maxLen - len, "  \"walLevel\": %d,\n", pCfg->wal.walLevel);
+  len += snprintf(content + len, maxLen - len, "  \"fsyncPeriod\": %d,\n", pCfg->wal.fsyncPeriod);
+  // sync
+  len += snprintf(content + len, maxLen - len, "  \"replica\": %d,\n", pCfg->sync.replica);
+  len += snprintf(content + len, maxLen - len, "  \"selfIndex\": %d,\n", pCfg->sync.selfIndex);
+  len += snprintf(content + len, maxLen - len, "  \"nodes\": [{\n");
+  for (int32_t i = 0; i < pCfg->sync.replica; i++) {
+    SNodeInfo *node = &pCfg->sync.nodeInfo[i];
+    len += snprintf(content + len, maxLen - len, "    \"id\": %d,\n", node->nodeId);
+    len += snprintf(content + len, maxLen - len, "    \"port\": %u,\n", node->nodePort);
+    len += snprintf(content + len, maxLen - len, "    \"fqdn\": \"%s\"\n", node->nodeFqdn);
+    if (i < pCfg->sync.replica - 1) {
       len += snprintf(content + len, maxLen - len, "  },{\n");
     } else {
       len += snprintf(content + len, maxLen - len, "  }]\n");
@@ -370,12 +287,86 @@ int32_t vnodeWriteCfg(SCreateVnodeMsg *pMsg) {
   len += snprintf(content + len, maxLen - len, "}\n");
 
   fwrite(content, 1, len, fp);
-  taosFsync(fileno(fp));
+  taosFsyncFile(fileno(fp));
   fclose(fp);
   free(content);
   terrno = 0;
 
-  vInfo("vgId:%d, successed to write %s", pMsg->cfg.vgId, file);
-#endif  
+  vInfo("vgId:%d, successed to write %s", vgId, file);
   return TSDB_CODE_SUCCESS;
 }
+
+int32_t vnodeReadState(int32_t vgId, SSyncServerState *pState) {
+  int32_t ret = TSDB_CODE_VND_APP_ERROR;
+  int32_t len = 0;
+  int32_t maxLen = 100;
+  char   *content = calloc(1, maxLen + 1);
+  cJSON  *root = NULL;
+  FILE   *fp = NULL;
+
+  char file[PATH_MAX + 30] = {0};
+  sprintf(file, "%s/vnode%d/state.json", tsVnodeDir, vgId);
+
+  len = (int32_t)fread(content, 1, maxLen, fp);
+  if (len <= 0) {
+    vError("vgId:%d, failed to read %s since content is null", vgId, file);
+    goto PARSE_TERM_ERROR;
+  }
+
+  root = cJSON_Parse(content);
+  if (root == NULL) {
+    vError("vgId:%d, failed to read %s since invalid json format", vgId, file);
+    goto PARSE_TERM_ERROR;
+  }
+
+  cJSON *term = cJSON_GetObjectItem(root, "term");
+  if (!term || term->type != cJSON_String) {
+    vError("vgId:%d, failed to read %s since term not found", vgId, file);
+    goto PARSE_TERM_ERROR;
+  }
+  pState->term = atoll(term->valuestring);
+
+  cJSON *voteFor = cJSON_GetObjectItem(root, "voteFor");
+  if (!voteFor || voteFor->type != cJSON_String) {
+    vError("vgId:%d, failed to read %s since voteFor not found", vgId, file);
+    goto PARSE_TERM_ERROR;
+  }
+  pState->voteFor = atoi(voteFor->valuestring);
+
+  vInfo("vgId:%d, read %s success, voteFor:%d, term:%" PRIu64, vgId, file, pState->voteFor, pState->term);
+
+PARSE_TERM_ERROR:
+  if (content != NULL) free(content);
+  if (root != NULL) cJSON_Delete(root);
+  if (fp != NULL) fclose(fp);
+
+  return ret;
+}
+
+int32_t vnodeSaveState(int32_t vgId, SSyncServerState *pState) {
+  char file[PATH_MAX + 30] = {0};
+  sprintf(file, "%s/vnode%d/state.json", tsVnodeDir, vgId);
+
+  FILE *fp = fopen(file, "w");
+  if (!fp) {
+    vError("vgId:%d, failed to write %s since %s", vgId, file, strerror(errno));
+    return -1;
+  }
+
+  int32_t len = 0;
+  int32_t maxLen = 100;
+  char   *content = calloc(1, maxLen + 1);
+
+  len += snprintf(content + len, maxLen - len, "{\n");
+  len += snprintf(content + len, maxLen - len, "  \"term\": \"%" PRIu64 "\",\n", pState->term);
+  len += snprintf(content + len, maxLen - len, "  \"voteFor\": \"%d\"\n", pState->voteFor);
+  len += snprintf(content + len, maxLen - len, "}\n");
+
+  fwrite(content, 1, len, fp);
+  taosFsyncFile(fileno(fp));
+  fclose(fp);
+  free(content);
+
+  vInfo("vgId:%d, write %s success, voteFor:%d, term:%" PRIu64, vgId, file, pState->voteFor, pState->term);
+  return TSDB_CODE_SUCCESS;
+}
\ No newline at end of file

source/server/vnode/src/vnodeInt.c

@@ -14,935 +14,78 @@
  */
 
 #define _DEFAULT_SOURCE
-#include "tglobal.h"
-#include "ttimer.h"
-#include "thash.h"
-// #include "query.h"
-#include "vnodeCfg.h"
-#include "vnodeMgmt.h"
-#include "vnodeRead.h"
-#include "vnodeStatus.h"
-#include "vnodeVersion.h"
-#include "vnodeWorker.h"
-#include "vnodeWrite.h"
+#include "os.h"
 #include "tstep.h"
+#include "vnodeMain.h"
 #include "vnodeMgmt.h"
 #include "vnodeRead.h"
-#include "vnodeWorker.h"
 #include "vnodeWrite.h"
 
-typedef struct {
-  pthread_t thread;
-  int32_t   threadIndex;
-  int32_t   failed;
-  int32_t   opened;
-  int32_t   vnodeNum;
-  int32_t * vnodeList;
-} SOpenVnodeThread;
-
 static struct {
   SSteps  *steps;
   SVnodeFp fp;
-  void *    timer;
-  SHashObj *hash;
-  int32_t   openVnodes;
-  int32_t   totalVnodes;
   void (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
-} tsVnode;
+} tsVint;
 
 void vnodeGetDnodeEp(int32_t dnodeId, char *ep, char *fqdn, uint16_t *port) {
-  return (*tsVnode.fp.GetDnodeEp)(dnodeId, ep, fqdn, port);
+  return (*tsVint.fp.GetDnodeEp)(dnodeId, ep, fqdn, port);
 }
 
 void vnodeSendMsgToDnode(struct SRpcEpSet *epSet, struct SRpcMsg *rpcMsg) {
-  (*tsVnode.fp.SendMsgToDnode)(epSet, rpcMsg);
-}
-
-void vnodeSendMsgToMnode(struct SRpcMsg *rpcMsg) { return (*tsVnode.fp.SendMsgToMnode)(rpcMsg); }
-
-static void vnodeIncRef(void *ptNode) {
-  assert(ptNode != NULL);
-
-  SVnode **ppVnode = (SVnode **)ptNode;
-  assert(ppVnode);
-  assert(*ppVnode);
-
-  SVnode *pVnode = *ppVnode;
-  atomic_add_fetch_32(&pVnode->refCount, 1);
-  vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
-}
-
-SVnode *vnodeAcquire(int32_t vgId) {
-  SVnode *pVnode = NULL;
-
-  // taosHashGetClone(tsVnode.hash, &vgId, sizeof(int32_t), vnodeIncRef, (void*)&pVnode);
-  if (pVnode == NULL) {
-    terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
-    vDebug("vgId:%d, not exist", vgId);
-    return NULL;
-  }
-
-  return pVnode;
-}
-
-SVnode *vnodeAcquireNotClose(int32_t vgId) {
-  // SVnode *pVnode = vnodeAcquire(vgId);
-  // if (pVnode != NULL && pVnode->preClose == 1) {
-  //   vnodeRelease(pVnode);
-  //   terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
-  //   vDebug("vgId:%d, not exist, pre closing", vgId);
-  //   return NULL;
-  // }
-
-  // return pVnode;
-  return NULL;
-}
-
-void vnodeRelease(SVnode *pVnode) {
-  if (pVnode == NULL) return;
-
-  int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1);
-  int32_t vgId = pVnode->vgId;
-
-  vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", vgId, refCount, pVnode);
-  assert(refCount >= 0);
-
-  if (refCount <= 0) {
-    vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", vgId, refCount, pVnode);
-    vnodeProcessDestroyTask(pVnode);
-    int32_t count = taosHashGetSize(tsVnode.hash);
-    vDebug("vgId:%d, vnode is destroyed, vnodes:%d", vgId, count);
-  }
-}
-
-static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno);
-
-int32_t vnodeCreate(SCreateVnodeMsg *pVnodeCfg) {
-  int32_t code;
-
-  SVnode *pVnode = vnodeAcquire(pVnodeCfg->cfg.vgId);
-  if (pVnode != NULL) {
-    vDebug("vgId:%d, vnode already exist, refCount:%d pVnode:%p", pVnodeCfg->cfg.vgId, pVnode->refCount, pVnode);
-    vnodeRelease(pVnode);
-    return TSDB_CODE_SUCCESS;
-  }
-
-#if 0
-  if (tfsMkdir("vnode") < 0) {
-    vError("vgId:%d, failed to create vnode dir, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno));
-    return terrno;
-  }
-
-  char vnodeDir[TSDB_FILENAME_LEN] = "\0";
-  snprintf(vnodeDir, TSDB_FILENAME_LEN, "/vnode/vnode%d", pVnodeCfg->cfg.vgId);
-  if (tfsMkdir(vnodeDir) < 0) {
-    vError("vgId:%d, failed to create vnode dir %s, reason:%s", pVnodeCfg->cfg.vgId, vnodeDir, strerror(errno));
-    return terrno;
-  }
-
-  code = vnodeWriteCfg(pVnodeCfg);
-  if (code != TSDB_CODE_SUCCESS) {
-    vError("vgId:%d, failed to save vnode cfg, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(code));
-    return code;
-  }
-
-  if (tsdbCreateRepo(pVnodeCfg->cfg.vgId) < 0) {
-    vError("vgId:%d, failed to create tsdb in vnode, reason:%s", pVnodeCfg->cfg.vgId, tstrerror(terrno));
-    return TSDB_CODE_VND_INIT_FAILED;
-  }
-#endif
-  vInfo("vgId:%d, vnode dir is created, walLevel:%d fsyncPeriod:%d", pVnodeCfg->cfg.vgId, pVnodeCfg->cfg.walLevel,
-        pVnodeCfg->cfg.fsyncPeriod);
-  code = vnodeOpen(pVnodeCfg->cfg.vgId);
-
-  return code;
-}
-
-int32_t vnodeSync(int32_t vgId) {
-#if 0  
-  SVnode *pVnode = vnodeAcquireNotClose(vgId);
-  if (pVnode == NULL) {
-    vDebug("vgId:%d, failed to sync, vnode not find", vgId);
-    return TSDB_CODE_VND_INVALID_VGROUP_ID;
-  }
-
-  if (pVnode->role == TAOS_SYNC_ROLE_SLAVE) {
-    vInfo("vgId:%d, vnode will sync, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
-
-    pVnode->version = 0;
-    pVnode->fversion = 0;
-    walResetVersion(pVnode->wal, pVnode->fversion);
-
-    syncRecover(pVnode->sync);
-  }
-
-  vnodeRelease(pVnode);
-#endif
-  return TSDB_CODE_SUCCESS;
-}
-
-int32_t vnodeDrop(int32_t vgId) {
-  SVnode *pVnode = vnodeAcquireNotClose(vgId);
-  if (pVnode == NULL) {
-    vDebug("vgId:%d, failed to drop, vnode not find", vgId);
-    return TSDB_CODE_VND_INVALID_VGROUP_ID;
-  }
-  if (pVnode->dropped) {
-    vnodeRelease(pVnode);
-    return TSDB_CODE_SUCCESS;
-  }
-
-  vInfo("vgId:%d, vnode will be dropped, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
-  pVnode->dropped = 1;
-
-  vnodeRelease(pVnode);
-  vnodeProcessCleanupTask(pVnode);
-
-  return TSDB_CODE_SUCCESS;
-}
-
-int32_t vnodeCompact(int32_t vgId) {
-#if 0  
-  SVnode *pVnode = vnodeAcquire(vgId);
-  if (pVnode != NULL) {
-    vDebug("vgId:%d, compact vnode msg is received", vgId);
-    // not care success or not
-    tsdbCompact(((SVnode *)pVnode)->tsdb);
-    vnodeRelease(pVnode);
-  } else {
-    vInfo("vgId:%d, vnode not exist, can't compact it", vgId);
-    return TSDB_CODE_VND_INVALID_VGROUP_ID;
-  }
-#endif   
-  return TSDB_CODE_SUCCESS;
-}
-
-static int32_t vnodeAlterImp(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg) {
-#if 0  
-  STsdbCfg tsdbCfg = pVnode->tsdbCfg;
-  SSyncCfg syncCfg = pVnode->syncCfg;
-  int32_t  dbCfgVersion = pVnode->dbCfgVersion;
-  int32_t  vgCfgVersion = pVnode->vgCfgVersion;
-
-  int32_t code = vnodeWriteCfg(pVnodeCfg);
-  if (code != TSDB_CODE_SUCCESS) {
-    pVnode->dbCfgVersion = dbCfgVersion;
-    pVnode->vgCfgVersion = vgCfgVersion;
-    pVnode->syncCfg = syncCfg;
-    pVnode->tsdbCfg = tsdbCfg;
-    return code;
-  }
-
-  code = vnodeReadCfg(pVnode);
-  if (code != TSDB_CODE_SUCCESS) {
-    pVnode->dbCfgVersion = dbCfgVersion;
-    pVnode->vgCfgVersion = vgCfgVersion;
-    pVnode->syncCfg = syncCfg;
-    pVnode->tsdbCfg = tsdbCfg;
-    return code;
-  }
-
-  code = walAlter(pVnode->wal, &pVnode->walCfg);
-  if (code != TSDB_CODE_SUCCESS) {
-    pVnode->dbCfgVersion = dbCfgVersion;
-    pVnode->vgCfgVersion = vgCfgVersion;
-    pVnode->syncCfg = syncCfg;
-    pVnode->tsdbCfg = tsdbCfg;
-    return code;
-  }
-
-  bool tsdbCfgChanged = (memcmp(&tsdbCfg, &pVnode->tsdbCfg, sizeof(STsdbCfg)) != 0);
-  bool syncCfgChanged = (memcmp(&syncCfg, &pVnode->syncCfg, sizeof(SSyncCfg)) != 0);
-
-  vDebug("vgId:%d, tsdbchanged:%d syncchanged:%d while alter vnode", pVnode->vgId, tsdbCfgChanged, syncCfgChanged);
-
-  if (tsdbCfgChanged || syncCfgChanged) {
-    // vnode in non-ready state and still needs to return success instead of TSDB_CODE_VND_INVALID_STATUS
-    // dbCfgVersion can be corrected by status msg
-    if (syncCfgChanged) {
-      if (!vnodeSetUpdatingStatus(pVnode)) {
-        vDebug("vgId:%d, vnode is not ready, do alter operation later", pVnode->vgId);
-        pVnode->dbCfgVersion = dbCfgVersion;
-        pVnode->vgCfgVersion = vgCfgVersion;
-        pVnode->syncCfg = syncCfg;
-        pVnode->tsdbCfg = tsdbCfg;
-        return TSDB_CODE_SUCCESS;
-      }
-
-      code = syncReconfig(pVnode->sync, &pVnode->syncCfg);
-      if (code != TSDB_CODE_SUCCESS) {
-        pVnode->dbCfgVersion = dbCfgVersion;
-        pVnode->vgCfgVersion = vgCfgVersion;
-        pVnode->syncCfg = syncCfg;
-        pVnode->tsdbCfg = tsdbCfg;
-        vnodeSetReadyStatus(pVnode);
-        return code;
-      }
-    }
-
-    if (tsdbCfgChanged && pVnode->tsdb) {
-      code = tsdbConfigRepo(pVnode->tsdb, &pVnode->tsdbCfg);
-      if (code != TSDB_CODE_SUCCESS) {
-        pVnode->dbCfgVersion = dbCfgVersion;
-        pVnode->vgCfgVersion = vgCfgVersion;
-        pVnode->syncCfg = syncCfg;
-        pVnode->tsdbCfg = tsdbCfg;
-        vnodeSetReadyStatus(pVnode);
-        return code;
-      }
-    }
-
-    vnodeSetReadyStatus(pVnode);
-  }
-#endif
-  return 0;
-}
-
-int32_t vnodeAlter(SVnode *pVnode, SCreateVnodeMsg *pVnodeCfg) {
-#if 0  
-  vDebug("vgId:%d, current dbCfgVersion:%d vgCfgVersion:%d, input dbCfgVersion:%d vgCfgVersion:%d", pVnode->vgId,
-         pVnode->dbCfgVersion, pVnode->vgCfgVersion, pVnodeCfg->cfg.dbCfgVersion, pVnodeCfg->cfg.vgCfgVersion);
-
-  if (pVnode->dbCfgVersion == pVnodeCfg->cfg.dbCfgVersion && pVnode->vgCfgVersion == pVnodeCfg->cfg.vgCfgVersion) {
-    vDebug("vgId:%d, cfg not change", pVnode->vgId);
-    return TSDB_CODE_SUCCESS;
-  }
-
-  int32_t code = vnodeAlterImp(pVnode, pVnodeCfg);
-
-  if (code != 0) {
-    vError("vgId:%d, failed to alter vnode, code:0x%x", pVnode->vgId, code);
-  } else {
-    vDebug("vgId:%d, vnode is altered", pVnode->vgId);
-  }
-
-  return code;
-#endif 
-  return 0;
-}
-
-static void vnodeFindWalRootDir(int32_t vgId, char *walRootDir) {
-#if 0  
-  char vnodeDir[TSDB_FILENAME_LEN] = "\0";
-  snprintf(vnodeDir, TSDB_FILENAME_LEN, "/vnode/vnode%d/wal", vgId);
-
-  TDIR *tdir = tfsOpendir(vnodeDir);
-  if (!tdir) return;
-
-  const TFILE *tfile = tfsReaddir(tdir);
-  if (!tfile) {
-    tfsClosedir(tdir);
-    return;
-  }
-
-  sprintf(walRootDir, "%s/vnode/vnode%d", TFS_DISK_PATH(tfile->level, tfile->id), vgId);
-
-  tfsClosedir(tdir);
-#endif  
-}
-
-int32_t vnodeOpen(int32_t vgId) {
-#if 0  
-  char temp[TSDB_FILENAME_LEN * 3];
-  char rootDir[TSDB_FILENAME_LEN * 2];
-  char walRootDir[TSDB_FILENAME_LEN * 2] = {0};
-  snprintf(rootDir, TSDB_FILENAME_LEN * 2, "%s/vnode%d", tsVnodeDir, vgId);
-
-  SVnode *pVnode = calloc(sizeof(SVnode), 1);
-  if (pVnode == NULL) {
-    vError("vgId:%d, failed to open vnode since no enough memory", vgId);
-    return TAOS_SYSTEM_ERROR(errno);
-  }
-
-  atomic_add_fetch_32(&pVnode->refCount, 1);
-
-  pVnode->vgId = vgId;
-  pVnode->fversion = 0;
-  pVnode->version = 0;
-  pVnode->tsdbCfg.tsdbId = pVnode->vgId;
-  pVnode->rootDir = strdup(rootDir);
-  pVnode->accessState = TSDB_VN_ALL_ACCCESS;
-  tsem_init(&pVnode->sem, 0, 0);
-  pthread_mutex_init(&pVnode->statusMutex, NULL);
-  vnodeSetInitStatus(pVnode);
-
-  tsdbIncCommitRef(pVnode->vgId);
-
-  int32_t code = vnodeReadCfg(pVnode);
-  if (code != TSDB_CODE_SUCCESS) {
-    vError("vgId:%d, failed to read config file, set cfgVersion to 0", pVnode->vgId);
-    vnodeCleanUp(pVnode);
-    return 0;
-  }
-
-  code = vnodeReadVersion(pVnode);
-  if (code != TSDB_CODE_SUCCESS) {
-    pVnode->version = 0;
-    vError("vgId:%d, failed to read file version, generate it from data file", pVnode->vgId);
-    // Allow vnode start even when read file version fails, set file version as wal version or zero
-    // vnodeCleanUp(pVnode);
-    // return code;
-  }
-
-  pVnode->fversion = pVnode->version;
-
-  pVnode->pWriteQ = vnodeAllocWriteQueue(pVnode);
-  pVnode->pQueryQ = vnodeAllocQueryQueue(pVnode);
-  pVnode->pFetchQ = vnodeAllocFetchQueue(pVnode);
-  if (pVnode->pWriteQ == NULL || pVnode->pQueryQ == NULL || pVnode->pFetchQ == NULL) {
-    vnodeCleanUp(pVnode);
-    return terrno;
-  }
-
-  STsdbAppH appH = {0};
-  appH.appH = (void *)pVnode;
-  appH.notifyStatus = vnodeProcessTsdbStatus;
-  appH.cqH = pVnode->cq;
-  appH.cqCreateFunc = cqCreate;
-  appH.cqDropFunc = cqDrop;
-
-  terrno = 0;
-  pVnode->tsdb = tsdbOpenRepo(&(pVnode->tsdbCfg), &appH);
-  if (pVnode->tsdb == NULL) {
-    vnodeCleanUp(pVnode);
-    return terrno;
-  } else if (tsdbGetState(pVnode->tsdb) != TSDB_STATE_OK) {
-    vError("vgId:%d, failed to open tsdb(state: %d), replica:%d reason:%s", pVnode->vgId, tsdbGetState(pVnode->tsdb),
-           pVnode->syncCfg.replica, tstrerror(terrno));
-    if (pVnode->syncCfg.replica <= 1) {
-      vnodeCleanUp(pVnode);
-      return TSDB_CODE_VND_INVALID_TSDB_STATE;
-    } else {
-      pVnode->fversion = 0;
-      pVnode->version = 0;
-    }
-  }
-
-  // walRootDir for wal & syncInfo.path (not empty dir of /vnode/vnode{pVnode->vgId}/wal)
-  vnodeFindWalRootDir(pVnode->vgId, walRootDir);
-  if (walRootDir[0] == 0) {
-    int level = -1, id = -1;
-
-    tfsAllocDisk(TFS_PRIMARY_LEVEL, &level, &id);
-    if (level < 0 || id < 0) {
-      vnodeCleanUp(pVnode);
-      return terrno;
-    }
-
-    sprintf(walRootDir, "%s/vnode/vnode%d", TFS_DISK_PATH(level, id), vgId);
-  }
-
-  sprintf(temp, "%s/wal", walRootDir);
-  pVnode->walCfg.vgId = pVnode->vgId;
-  pVnode->wal = walOpen(temp, &pVnode->walCfg);
-  if (pVnode->wal == NULL) {
-    vnodeCleanUp(pVnode);
-    return terrno;
-  }
-
-  walRestore(pVnode->wal, pVnode, (FWalWrite)vnodeProcessWalMsg);
-  if (pVnode->version == 0) {
-    pVnode->fversion = 0;
-    pVnode->version = walGetVersion(pVnode->wal);
-  }
-
-  code = tsdbSyncCommit(pVnode->tsdb);
-  if (code != 0) {
-    vError("vgId:%d, failed to commit after restore from wal since %s", pVnode->vgId, tstrerror(code));
-    vnodeCleanUp(pVnode);
-    return code;
-  }
-
-  walRemoveAllOldFiles(pVnode->wal);
-  walRenew(pVnode->wal);
-
-  pVnode->qMgmt = qOpenQueryMgmt(pVnode->vgId);
-  if (pVnode->qMgmt == NULL) {
-    vnodeCleanUp(pVnode);
-    return terrno;
-  }
-
-  pVnode->events = NULL;
-
-  vDebug("vgId:%d, vnode is opened in %s - %s, pVnode:%p", pVnode->vgId, rootDir, walRootDir, pVnode);
-
-  taosHashPut(tsVnode.hash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnode *));
-
-  vnodeSetReadyStatus(pVnode);
-  pVnode->role = TAOS_SYNC_ROLE_MASTER;
-#endif  
-  return TSDB_CODE_SUCCESS;
-}
-
-int32_t vnodeClose(int32_t vgId) {
-  SVnode *pVnode = vnodeAcquireNotClose(vgId);
-  if (pVnode == NULL) return 0;
-  if (pVnode->dropped) {
-    vnodeRelease(pVnode);
-    return 0;
-  }
-
-  // pVnode->preClose = 1;
-
-  vDebug("vgId:%d, vnode will be closed, pVnode:%p", pVnode->vgId, pVnode);
-  vnodeRelease(pVnode);
-  vnodeCleanUp(pVnode);
-
-  return 0;
-}
-
-void vnodeDestroy(SVnode *pVnode) {
-#if 0  
-  int32_t code = 0;
-  int32_t vgId = pVnode->vgId;
-
-  if (pVnode->qMgmt) {
-    qCleanupQueryMgmt(pVnode->qMgmt);
-    pVnode->qMgmt = NULL;
-  }
-
-  if (pVnode->wal) {
-    walStop(pVnode->wal);
-  }
-
-  if (pVnode->tsdb) {
-    // the deleted vnode does not need to commit, so as to speed up the deletion
-    int toCommit = 1;
-    if (pVnode->dropped) toCommit = 0;
-
-    code = tsdbCloseRepo(pVnode->tsdb, toCommit);
-    pVnode->tsdb = NULL;
-  }
-
-  // stop continuous query
-  if (pVnode->cq) {
-    void *cq = pVnode->cq;
-    pVnode->cq = NULL;
-    cqClose(cq);
-  }
-
-  if (pVnode->wal) {
-    if (code != 0) {
-      vError("vgId:%d, failed to commit while close tsdb repo, keep wal", pVnode->vgId);
-    } else {
-      walRemoveAllOldFiles(pVnode->wal);
-    }
-    walClose(pVnode->wal);
-    pVnode->wal = NULL;
-  }
-
-  if (pVnode->pWriteQ) {
-    vnodeFreeWriteQueue(pVnode->pWriteQ);
-    pVnode->pWriteQ = NULL;
-  }
-
-  if (pVnode->pQueryQ) {
-    vnodeFreeQueryQueue(pVnode->pQueryQ);
-    pVnode->pQueryQ = NULL;
-  }
-
-  if (pVnode->pFetchQ) {
-    vnodeFreeFetchQueue(pVnode->pFetchQ);
-    pVnode->pFetchQ = NULL;
-  }
-
-  tfree(pVnode->rootDir);
-
-  if (pVnode->dropped) {
-    char rootDir[TSDB_FILENAME_LEN] = {0};
-    char stagingDir[TSDB_FILENAME_LEN] = {0};
-    sprintf(rootDir, "%s/vnode%d", "vnode", vgId);
-    sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId);
-
-    tfsRename(rootDir, stagingDir);
-
-    vnodeProcessBackupTask(pVnode);
-
-    // dnodeSendStatusMsgToMnode();
-  }
-
-  tsem_destroy(&pVnode->sem);
-  pthread_mutex_destroy(&pVnode->statusMutex);
-  free(pVnode);
-  tsdbDecCommitRef(vgId);
-#endif  
-}
-
-void vnodeCleanUp(SVnode *pVnode) {
-#if 0  
-  vDebug("vgId:%d, vnode will cleanup, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
-
-  vnodeSetClosingStatus(pVnode);
-
-  taosHashRemove(tsVnode.hash, &pVnode->vgId, sizeof(int32_t));
-
-  // stop replication module
-  if (pVnode->sync > 0) {
-    int64_t sync = pVnode->sync;
-    pVnode->sync = -1;
-    syncStop(sync);
-  }
-
-  vDebug("vgId:%d, vnode is cleaned, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
-  vnodeRelease(pVnode);
-#endif  
-}
-
-#if 0
-static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) {
-  SVnode *pVnode = arg;
-
-  if (eno != TSDB_CODE_SUCCESS) {
-    vError("vgId:%d, failed to commit since %s, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, tstrerror(eno),
-           pVnode->fversion, pVnode->version);
-    pVnode->isCommiting = 0;
-    pVnode->isFull = 1;
-    return 0;
-  }
-
-  if (status == TSDB_STATUS_COMMIT_START) {
-    pVnode->isCommiting = 1;
-    pVnode->cversion = pVnode->version;
-    vInfo("vgId:%d, start commit, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, pVnode->fversion, pVnode->version);
-    if (!vnodeInInitStatus(pVnode)) {
-      return walRenew(pVnode->wal);
-    }
-    return 0;
-  }
-
-  if (status == TSDB_STATUS_COMMIT_OVER) {
-    pVnode->isCommiting = 0;
-    pVnode->isFull = 0;
-    pVnode->fversion = pVnode->cversion;
-    vInfo("vgId:%d, commit over, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, pVnode->fversion, pVnode->version);
-    if (!vnodeInInitStatus(pVnode)) {
-      walRemoveOneOldFile(pVnode->wal);
-    }
-    return vnodeSaveVersion(pVnode);
-  }
-
-  // timer thread callback
-  if (status == TSDB_STATUS_COMMIT_NOBLOCK) {
-    qSolveCommitNoBlock(pVnode->tsdb, pVnode->qMgmt);
-  }
-
-  return 0;
-}
-#endif
-
-static void *vnodeOpenVnode(void *param) {
-  SOpenVnodeThread *pThread = param;
-
-  vDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
-  setThreadName("vnodeOpenVnode");
-
-  for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
-    int32_t vgId = pThread->vnodeList[v];
-
-    char stepDesc[TSDB_STEP_DESC_LEN] = {0};
-    snprintf(stepDesc, TSDB_STEP_DESC_LEN, "vgId:%d, start to restore, %d of %d have been opened", vgId,
-             tsVnode.openVnodes, tsVnode.totalVnodes);
-    // (*vnodeInst()->fp.ReportStartup)("open-vnodes", stepDesc);
-
-    if (vnodeOpen(vgId) < 0) {
-      vError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex);
-      pThread->failed++;
-    } else {
-      vDebug("vgId:%d, is opened by thread:%d", vgId, pThread->threadIndex);
-      pThread->opened++;
-    }
-
-    atomic_add_fetch_32(&tsVnode.openVnodes, 1);
-  }
-
-  vDebug("thread:%d, total vnodes:%d, opened:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
-         pThread->failed);
-  return NULL;
-}
-
-static int32_t vnodeGetVnodeListFromDisk(int32_t vnodeList[], int32_t *numOfVnodes) {
-#if 0
-  DIR *dir = opendir(tsVnodeDir);
-  if (dir == NULL) return TSDB_CODE_DND_NO_WRITE_ACCESS;
-
-  *numOfVnodes = 0;
-  struct dirent *de = NULL;
-  while ((de = readdir(dir)) != NULL) {
-    if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
-    if (de->d_type & DT_DIR) {
-      if (strncmp("vnode", de->d_name, 5) != 0) continue;
-      int32_t vnode = atoi(de->d_name + 5);
-      if (vnode == 0) continue;
-
-      (*numOfVnodes)++;
-
-      if (*numOfVnodes >= TSDB_MAX_VNODES) {
-        vError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES);
-        closedir(dir);
-        return TSDB_CODE_DND_TOO_MANY_VNODES;
-      } else {
-        vnodeList[*numOfVnodes - 1] = vnode;
-      }
-    }
-  }
-  closedir(dir);
-#endif
-  return TSDB_CODE_SUCCESS;
-}
-
-static int32_t vnodeOpenVnodes() {
-  int32_t vnodeList[TSDB_MAX_VNODES] = {0};
-  int32_t numOfVnodes = 0;
-  int32_t status = vnodeGetVnodeListFromDisk(vnodeList, &numOfVnodes);
-
-  if (status != TSDB_CODE_SUCCESS) {
-    vInfo("failed to get vnode list from disk since code:%d", status);
-    return status;
-  }
-
-  tsVnode.totalVnodes = numOfVnodes;
-
-  int32_t threadNum = tsNumOfCores;
-  int32_t vnodesPerThread = numOfVnodes / threadNum + 1;
-
-  SOpenVnodeThread *threads = calloc(threadNum, sizeof(SOpenVnodeThread));
-  for (int32_t t = 0; t < threadNum; ++t) {
-    threads[t].threadIndex = t;
-    threads[t].vnodeList = calloc(vnodesPerThread, sizeof(int32_t));
-  }
-
-  for (int32_t v = 0; v < numOfVnodes; ++v) {
-    int32_t           t = v % threadNum;
-    SOpenVnodeThread *pThread = &threads[t];
-    pThread->vnodeList[pThread->vnodeNum++] = vnodeList[v];
-  }
-
-  vInfo("start %d threads to open %d vnodes", threadNum, numOfVnodes);
-
-  for (int32_t t = 0; t < threadNum; ++t) {
-    SOpenVnodeThread *pThread = &threads[t];
-    if (pThread->vnodeNum == 0) continue;
-
-    pthread_attr_t thAttr;
-    pthread_attr_init(&thAttr);
-    pthread_attr_setdetachstate(&thAttr, PTHREAD_CREATE_JOINABLE);
-    if (pthread_create(&pThread->thread, &thAttr, vnodeOpenVnode, pThread) != 0) {
-      vError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno));
-    }
-
-    pthread_attr_destroy(&thAttr);
-  }
-
-  int32_t openVnodes = 0;
-  int32_t failedVnodes = 0;
-  for (int32_t t = 0; t < threadNum; ++t) {
-    SOpenVnodeThread *pThread = &threads[t];
-    if (pThread->vnodeNum > 0 && taosCheckPthreadValid(pThread->thread)) {
-      pthread_join(pThread->thread, NULL);
-    }
-    openVnodes += pThread->opened;
-    failedVnodes += pThread->failed;
-    free(pThread->vnodeList);
-  }
-
-  free(threads);
-  vInfo("there are total vnodes:%d, opened:%d", numOfVnodes, openVnodes);
-
-  if (failedVnodes != 0) {
-    vError("there are total vnodes:%d, failed:%d", numOfVnodes, failedVnodes);
-    return -1;
-  }
-
-  return TSDB_CODE_SUCCESS;
-}
-
-static int32_t vnodeGetVnodeList(int32_t vnodeList[], int32_t *numOfVnodes) {
-  void *pIter = taosHashIterate(tsVnode.hash, NULL);
-  while (pIter) {
-    SVnode **pVnode = pIter;
-    if (*pVnode) {
-      (*numOfVnodes)++;
-      if (*numOfVnodes >= TSDB_MAX_VNODES) {
-        vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES);
-        continue;
-      } else {
-        vnodeList[*numOfVnodes - 1] = (*pVnode)->vgId;
-      }
-    }
-
-    pIter = taosHashIterate(tsVnode.hash, pIter);
-  }
-
-  return TSDB_CODE_SUCCESS;
+  (*tsVint.fp.SendMsgToDnode)(epSet, rpcMsg);
 }
 
-static void vnodeCleanupVnodes() {
-  int32_t vnodeList[TSDB_MAX_VNODES] = {0};
-  int32_t numOfVnodes = 0;
-
-  int32_t code = vnodeGetVnodeList(vnodeList, &numOfVnodes);
+void vnodeSendMsgToMnode(struct SRpcMsg *rpcMsg) { return (*tsVint.fp.SendMsgToMnode)(rpcMsg); }
 
-  if (code != TSDB_CODE_SUCCESS) {
-    vInfo("failed to get dnode list since code %d", code);
-    return;
-  }
-
-  for (int32_t i = 0; i < numOfVnodes; ++i) {
-    vnodeClose(vnodeList[i]);
-  }
-
-  vInfo("total vnodes:%d are all closed", numOfVnodes);
-}
-
-static void vnodeInitMsgFp() {
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE]  = vnodeProcessMgmtMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE]   = vnodeProcessMgmtMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE]    = vnodeProcessMgmtMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessMgmtMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE]    = vnodeProcessMgmtMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM]  = vnodeProcessMgmtMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE]  = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_TABLE]    = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE]   = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MD_DROP_STABLE]   = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_SUBMIT]           = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL]   = vnodeProcessWriteMsg;
-  //mq related
-  tsVnode.msgFp[TSDB_MSG_TYPE_MQ_CONNECT]       = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MQ_DISCONNECT]    = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MQ_ACK]           = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MQ_RESET]         = vnodeProcessWriteMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MQ_QUERY]         = vnodeProcessReadMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_MQ_CONSUME]       = vnodeProcessReadMsg;
-  //mq related end
-  tsVnode.msgFp[TSDB_MSG_TYPE_QUERY]            = vnodeProcessReadMsg;
-  tsVnode.msgFp[TSDB_MSG_TYPE_FETCH]            = vnodeProcessReadMsg;
-}
+void vnodeReportStartup(char *name, char *desc) { (*tsVint.fp.ReportStartup)(name, desc); }
 
 void vnodeProcessMsg(SRpcMsg *pMsg) {
-  if (tsVnode.msgFp[pMsg->msgType]) {
-    (*tsVnode.msgFp[pMsg->msgType])(pMsg);
+  if (tsVint.msgFp[pMsg->msgType]) {
+    (*tsVint.msgFp[pMsg->msgType])(pMsg);
   } else {
     assert(0);
   }
 }
 
-int32_t vnodeInitMain() {
-  vnodeInitMsgFp();
-
-  tsVnode.timer = taosTmrInit(100, 200, 60000, "VND-TIMER");
-  if (tsVnode.timer == NULL) {
-    vError("failed to init vnode timer");
-    return -1;
-  }
-
-  tsVnode.hash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK);
-  if (tsVnode.hash == NULL) {
-    taosTmrCleanUp(tsVnode.timer);
-    vError("failed to init vnode mgmt");
-    return -1;
-  }
-
-  vInfo("vnode main is initialized");
-  return vnodeOpenVnodes();
-}
-
-void vnodeCleanupMain() {
-  taosTmrCleanUp(tsVnode.timer);
-  tsVnode.timer = NULL;
-
-  vnodeCleanupVnodes();
-
-  taosHashCleanup(tsVnode.hash);
-  tsVnode.hash = NULL;
-}
-
-static void vnodeBuildVloadMsg(SVnode *pVnode, SStatusMsg *pStatus) {
-#if 0  
-  int64_t totalStorage = 0;
-  int64_t compStorage = 0;
-  int64_t pointsWritten = 0;
-
-  if (vnodeInClosingStatus(pVnode)) return;
-  if (pStatus->openVnodes >= TSDB_MAX_VNODES) return;
-
-  if (pVnode->tsdb) {
-    tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage);
-  }
-
-  SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++];
-  pLoad->vgId = htonl(pVnode->vgId);
-  pLoad->dbCfgVersion = htonl(pVnode->dbCfgVersion);
-  pLoad->vgCfgVersion = htonl(pVnode->vgCfgVersion);
-  pLoad->totalStorage = htobe64(totalStorage);
-  pLoad->compStorage = htobe64(compStorage);
-  pLoad->pointsWritten = htobe64(pointsWritten);
-  pLoad->vnodeVersion = htobe64(pVnode->version);
-  pLoad->status = pVnode->status;
-  pLoad->role = pVnode->role;
-  pLoad->replica = pVnode->syncCfg.replica;
-  pLoad->compact = (pVnode->tsdb != NULL) ? tsdbGetCompactState(pVnode->tsdb) : 0;
-#endif  
-}
-
-void vnodeGetStatus(struct SStatusMsg *pStatus) {
-  void *pIter = taosHashIterate(tsVnode.hash, NULL);
-  while (pIter) {
-    SVnode **pVnode = pIter;
-    if (*pVnode) {
-      vnodeBuildVloadMsg(*pVnode, pStatus);
-    }
-    pIter = taosHashIterate(tsVnode.hash, pIter);
-  }
-}
-
-void vnodeSetAccess(struct SVgroupAccess *pAccess, int32_t numOfVnodes) {
-  for (int32_t i = 0; i < numOfVnodes; ++i) {
-    pAccess[i].vgId = htonl(pAccess[i].vgId);
-    SVnode *pVnode = vnodeAcquireNotClose(pAccess[i].vgId);
-    if (pVnode != NULL) {
-      pVnode->accessState = pAccess[i].accessState;
-      if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) {
-        vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState);
-      }
-      vnodeRelease(pVnode);
-    }
-  }
-}
-
-void vnodeBackup(int32_t vgId) {
-  char newDir[TSDB_FILENAME_LEN] = {0};
-  char stagingDir[TSDB_FILENAME_LEN] = {0};
-
-  sprintf(newDir, "%s/vnode%d", "vnode_bak", vgId);
-  sprintf(stagingDir, "%s/.staging/vnode%d", "vnode_bak", vgId);
-
-#if 0
-  if (tsEnableVnodeBak) {
-    tfsRmdir(newDir);
-    tfsRename(stagingDir, newDir);
-  } else {
-    vInfo("vgId:%d, vnode backup not enabled", vgId);
-
-    tfsRmdir(stagingDir);
-  }
-#endif  
+static void vnodeInitMsgFp() {
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessMgmtMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessMgmtMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessMgmtMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessMgmtMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessMgmtMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessMgmtMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_CREATE_TABLE] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_DROP_TABLE] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_ALTER_TABLE] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MD_DROP_STABLE] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_SUBMIT] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_UPDATE_TAG_VAL] = vnodeProcessWriteMsg;
+  // mq related
+  tsVint.msgFp[TSDB_MSG_TYPE_MQ_CONNECT] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MQ_DISCONNECT] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MQ_ACK] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MQ_RESET] = vnodeProcessWriteMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MQ_QUERY] = vnodeProcessReadMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_MQ_CONSUME] = vnodeProcessReadMsg;
+  // mq related end
+  tsVint.msgFp[TSDB_MSG_TYPE_QUERY] = vnodeProcessReadMsg;
+  tsVint.msgFp[TSDB_MSG_TYPE_FETCH] = vnodeProcessReadMsg;
 }
 
 int32_t vnodeInit(SVnodePara para) {
-  tsVnode.fp = para.fp;
+  vnodeInitMsgFp();
+  tsVint.fp = para.fp;
 
   struct SSteps *steps = taosStepInit(8, NULL);
   if (steps == NULL) return -1;
 
   taosStepAdd(steps, "vnode-main", vnodeInitMain, vnodeCleanupMain);
-  taosStepAdd(steps, "vnode-worker", vnodeInitWorker, vnodeCleanupWorker);
   taosStepAdd(steps, "vnode-read", vnodeInitRead, vnodeCleanupRead);
   taosStepAdd(steps, "vnode-mgmt", vnodeInitMgmt, vnodeCleanupMgmt);
   taosStepAdd(steps, "vnode-write", vnodeInitWrite, vnodeCleanupWrite);
 
-  tsVnode.steps = steps;
-  return taosStepExec(tsVnode.steps);
+  tsVint.steps = steps;
+  return taosStepExec(tsVint.steps);
 }
 
-void vnodeCleanup() { taosStepCleanup(tsVnode.steps); }
+void vnodeCleanup() { taosStepCleanup(tsVint.steps); }

source/server/vnode/src/vnodeMain.c

+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define _DEFAULT_SOURCE
+#include "thash.h"
+#include "tthread.h"
+#include "vnodeFile.h"
+#include "vnodeMain.h"
+#include "vnodeMgmt.h"
+#include "vnodeRead.h"
+#include "vnodeWrite.h"
+
+typedef enum _VN_STATUS {
+  TAOS_VN_STATUS_INIT = 0,
+  TAOS_VN_STATUS_READY = 1,
+  TAOS_VN_STATUS_CLOSING = 2,
+  TAOS_VN_STATUS_UPDATING = 3
+} EVnodeStatus;
+
+char *vnodeStatus[] = {"init", "ready", "closing", "updating"};
+
+typedef struct {
+  pthread_t *threadId;
+  int32_t    threadIndex;
+  int32_t    failed;
+  int32_t    opened;
+  int32_t    vnodeNum;
+  int32_t   *vnodeList;
+} SOpenVnodeThread;
+
+static struct {
+  SHashObj *hash;
+  int32_t   openVnodes;
+  int32_t   totalVnodes;
+} tsVnode;
+
+static bool vnodeSetInitStatus(SVnode *pVnode) {
+  pthread_mutex_lock(&pVnode->statusMutex);
+  pVnode->status = TAOS_VN_STATUS_INIT;
+  pthread_mutex_unlock(&pVnode->statusMutex);
+  return true;
+}
+
+static bool vnodeSetReadyStatus(SVnode *pVnode) {
+  bool set = false;
+  pthread_mutex_lock(&pVnode->statusMutex);
+
+  if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_UPDATING) {
+    pVnode->status = TAOS_VN_STATUS_READY;
+    set = true;
+  }
+
+  pthread_mutex_unlock(&pVnode->statusMutex);
+  return set;
+}
+
+static bool vnodeSetUpdatingStatus(SVnode *pVnode) {
+  bool set = false;
+  pthread_mutex_lock(&pVnode->statusMutex);
+
+  if (pVnode->status == TAOS_VN_STATUS_READY) {
+    pVnode->status = TAOS_VN_STATUS_UPDATING;
+    set = true;
+  }
+
+  pthread_mutex_unlock(&pVnode->statusMutex);
+  return set;
+}
+
+static bool vnodeSetClosingStatus(SVnode *pVnode) {
+  bool set = false;
+  pthread_mutex_lock(&pVnode->statusMutex);
+
+  if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_READY) {
+    pVnode->status = TAOS_VN_STATUS_CLOSING;
+    set = true;
+  }
+
+  pthread_mutex_unlock(&pVnode->statusMutex);
+  return set;
+}
+
+static bool vnodeInStatus(SVnode *pVnode, EVnodeStatus status) {
+  bool in = false;
+  pthread_mutex_lock(&pVnode->statusMutex);
+
+  if (pVnode->status == status) {
+    in = true;
+  }
+
+  pthread_mutex_unlock(&pVnode->statusMutex);
+  return in;
+}
+
+static void vnodeDestroyVnode(SVnode *pVnode) {
+  int32_t code = 0;
+  int32_t vgId = pVnode->vgId;
+
+  if (pVnode->pSync != NULL) {
+    syncStop(pVnode->pSync);
+    pVnode->pSync = NULL;
+  }
+
+  if (pVnode->pQuery) {
+    // todo
+  }
+
+  if (pVnode->pMeta) {
+    // todo
+  }
+
+  if (pVnode->pTsdb) {
+    // todo
+  }
+
+  if (pVnode->pTQ) {
+    // todo
+  }
+
+  if (pVnode->pWal) {
+    walClose(pVnode->pWal);
+    pVnode->pWal = NULL;
+  }
+
+  if (pVnode->allocator) {
+    // todo
+  }
+
+  if (pVnode->pWriteQ) {
+    vnodeFreeWriteQueue(pVnode->pWriteQ);
+    pVnode->pWriteQ = NULL;
+  }
+
+  if (pVnode->pQueryQ) {
+    vnodeFreeQueryQueue(pVnode->pQueryQ);
+    pVnode->pQueryQ = NULL;
+  }
+
+  if (pVnode->pFetchQ) {
+    vnodeFreeFetchQueue(pVnode->pFetchQ);
+    pVnode->pFetchQ = NULL;
+  }
+
+  if (pVnode->dropped) {
+    // todo
+  }
+
+  pthread_mutex_destroy(&pVnode->statusMutex);
+  free(pVnode);
+}
+
+static void vnodeCleanupVnode(SVnode *pVnode) {
+  vnodeSetClosingStatus(pVnode);
+  taosHashRemove(tsVnode.hash, &pVnode->vgId, sizeof(int32_t));
+  vnodeRelease(pVnode);
+}
+
+static inline int32_t vnodeLogWrite(struct SSyncLogStore *logStore, SyncIndex index, SSyncBuffer *pBuf) {
+  SVnode *pVnode = logStore->pData;  // vnode status can be checked here
+  return walWrite(pVnode->pWal, index, pBuf->data, (int32_t)pBuf->len);
+}
+
+static inline int32_t vnodeLogCommit(struct SSyncLogStore *logStore, SyncIndex index) {
+  SVnode *pVnode = logStore->pData;  // vnode status can be checked here
+  return walCommit(pVnode->pWal, index);
+}
+
+static inline int32_t vnodeLogPrune(struct SSyncLogStore *logStore, SyncIndex index) {
+  SVnode *pVnode = logStore->pData;  // vnode status can be checked here
+  return walPrune(pVnode->pWal, index);
+}
+
+static inline int32_t vnodeLogRollback(struct SSyncLogStore *logStore, SyncIndex index) {
+  SVnode *pVnode = logStore->pData;  // vnode status can be checked here
+  return walRollback(pVnode->pWal, index);
+}
+
+static inline int32_t vnodeSaveServerState(struct SStateManager *stateMng, SSyncServerState *pState) {
+  SVnode *pVnode = stateMng->pData;
+  return vnodeSaveState(pVnode->vgId, pState);
+}
+
+static inline int32_t vnodeReadServerState(struct SStateManager *stateMng, SSyncServerState *pState) {
+  SVnode *pVnode = stateMng->pData;
+  return vnodeSaveState(pVnode->vgId, pState);
+}
+
+static inline int32_t vnodeApplyLog(struct SSyncFSM *fsm, SyncIndex index, const SSyncBuffer *buf, void *pData) {
+  return 0;
+}
+
+static inline int32_t vnodeOnClusterChanged(struct SSyncFSM *fsm, const SSyncCluster *cluster, void *pData) { return 0; }
+
+static inline int32_t vnodeGetSnapshot(struct SSyncFSM *fsm, SSyncBuffer **ppBuf, int32_t *objId, bool *isLast) {
+  return 0;
+}
+
+static inline int32_t vnodeApplySnapshot(struct SSyncFSM *fsm, SSyncBuffer *pBuf, int32_t objId, bool isLast) {
+  return 0;
+}
+
+static inline int32_t vnodeOnRestoreDone(struct SSyncFSM *fsm) { return 0; }
+
+static inline void vnodeOnRollback(struct SSyncFSM *fsm, SyncIndex index, const SSyncBuffer *buf) {}
+
+static inline void vnodeOnRoleChanged(struct SSyncFSM *fsm, const SNodesRole *pRole) {}
+
+static int32_t vnodeOpenVnode(int32_t vgId) {
+  int32_t code = 0;
+
+  SVnode *pVnode = calloc(sizeof(SVnode), 1);
+  if (pVnode == NULL) {
+    vError("vgId:%d, failed to open vnode since no enough memory", vgId);
+    return TAOS_SYSTEM_ERROR(errno);
+  }
+
+  pVnode->vgId = vgId;
+  pVnode->accessState = TAOS_VN_STATUS_INIT;
+  pVnode->status = TSDB_VN_ALL_ACCCESS;
+  pVnode->refCount = 1;
+  pVnode->role = TAOS_SYNC_ROLE_CANDIDATE;
+  pthread_mutex_init(&pVnode->statusMutex, NULL);
+
+  vDebug("vgId:%d, vnode is opened", pVnode->vgId);
+  taosHashPut(tsVnode.hash, &pVnode->vgId, sizeof(int32_t), &pVnode, sizeof(SVnode *));
+
+  code = vnodeReadCfg(vgId, &pVnode->cfg);
+  if (code != TSDB_CODE_SUCCESS) {
+    vError("vgId:%d, failed to read config file, set cfgVersion to 0", pVnode->vgId);
+    pVnode->cfg.dropped = 1;
+    vnodeCleanupVnode(pVnode);
+    return 0;
+  }
+
+  code = vnodeSaveState(vgId, &pVnode->term);
+  if (code != TSDB_CODE_SUCCESS) {
+    vError("vgId:%d, failed to read term file since %s", pVnode->vgId, tstrerror(code));
+    pVnode->cfg.dropped = 1;
+    vnodeCleanupVnode(pVnode);
+    return code;
+  }
+
+  pVnode->pWriteQ = vnodeAllocWriteQueue(pVnode);
+  pVnode->pQueryQ = vnodeAllocQueryQueue(pVnode);
+  pVnode->pFetchQ = vnodeAllocFetchQueue(pVnode);
+  if (pVnode->pWriteQ == NULL || pVnode->pQueryQ == NULL || pVnode->pFetchQ == NULL) {
+    vnodeCleanupVnode(pVnode);
+    return terrno;
+  }
+
+  char path[PATH_MAX + 20];
+  snprintf(path, sizeof(path), "%s/vnode%d/wal", tsVnodeDir, vgId);
+  pVnode->pWal = walOpen(path, &pVnode->cfg.wal);
+  if (pVnode->pWal == NULL) {
+    vnodeCleanupVnode(pVnode);
+    return terrno;
+  }
+
+  // create sync node
+  SSyncInfo syncInfo = {0};
+  syncInfo.vgId = vgId;
+  syncInfo.snapshotIndex = 0;  // todo, from tsdb
+  memcpy(&syncInfo.syncCfg, &pVnode->cfg.sync, sizeof(SSyncCluster));
+  syncInfo.fsm.pData = pVnode;
+  syncInfo.fsm.applyLog = vnodeApplyLog;
+  syncInfo.fsm.onClusterChanged = vnodeOnClusterChanged;
+  syncInfo.fsm.getSnapshot = vnodeGetSnapshot;
+  syncInfo.fsm.applySnapshot = vnodeApplySnapshot;
+  syncInfo.fsm.onRestoreDone = vnodeOnRestoreDone;
+  syncInfo.fsm.onRollback = vnodeOnRollback;
+  syncInfo.fsm.onRoleChanged = vnodeOnRoleChanged;
+  syncInfo.logStore.pData = pVnode;
+  syncInfo.logStore.logWrite = vnodeLogWrite;
+  syncInfo.logStore.logCommit = vnodeLogCommit;
+  syncInfo.logStore.logPrune = vnodeLogPrune;
+  syncInfo.logStore.logRollback = vnodeLogRollback;
+  syncInfo.stateManager.pData = pVnode;
+  syncInfo.stateManager.saveServerState = vnodeSaveServerState;
+  syncInfo.stateManager.readServerState = vnodeReadServerState;
+
+  pVnode->pSync = syncStart(&syncInfo);
+  if (pVnode->pSync == NULL) {
+    vnodeCleanupVnode(pVnode);
+    return terrno;
+  }
+
+  vnodeSetReadyStatus(pVnode);
+  return TSDB_CODE_SUCCESS;
+}
+
+int32_t vnodeCreateVnode(int32_t vgId, SVnodeCfg *pCfg) {
+  int32_t code = 0;
+  char    path[PATH_MAX + 20] = {0};
+
+  snprintf(path, sizeof(path), "%s/vnode%d", tsVnodeDir, vgId);
+  if (taosMkDir(path) < 0) {
+    code = TAOS_SYSTEM_ERROR(errno);
+    vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  snprintf(path, sizeof(path), "%s/vnode%d/cfg", tsVnodeDir, vgId);
+  if (taosMkDir(path) < 0) {
+    code = TAOS_SYSTEM_ERROR(errno);
+    vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  snprintf(path, sizeof(path), "%s/vnode%d/wal", tsVnodeDir, vgId);
+  if (taosMkDir(path) < 0) {
+    code = TAOS_SYSTEM_ERROR(errno);
+    vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  snprintf(path, sizeof(path), "%s/vnode%d/tq", tsVnodeDir, vgId);
+  if (taosMkDir(path) < 0) {
+    code = TAOS_SYSTEM_ERROR(errno);
+    vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  snprintf(path, sizeof(path), "%s/vnode%d/tsdb", tsVnodeDir, vgId);
+  if (taosMkDir(path) < 0) {
+    code = TAOS_SYSTEM_ERROR(errno);
+    vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  snprintf(path, sizeof(path), "%s/vnode%d/meta", tsVnodeDir, vgId);
+  if (taosMkDir(path) < 0) {
+    code = TAOS_SYSTEM_ERROR(errno);
+    vError("vgId:%d, failed to create since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  code = vnodeWriteCfg(vgId, pCfg);
+  if (code != 0) {
+    vError("vgId:%d, failed to save vnode cfg since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  return vnodeOpenVnode(vgId);
+}
+
+int32_t vnodeAlterVnode(SVnode * pVnode, SVnodeCfg *pCfg) {
+  int32_t code = 0;
+  int32_t vgId = pVnode->vgId;
+
+  bool walChanged = (memcmp(&pCfg->wal, &pVnode->cfg.wal, sizeof(SWalCfg)) != 0);
+  bool tsdbChanged = (memcmp(&pCfg->tsdb, &pVnode->cfg.tsdb, sizeof(STsdbCfg)) != 0);
+  bool metaChanged = (memcmp(&pCfg->meta, &pVnode->cfg.meta, sizeof(SMetaCfg)) != 0);
+  bool syncChanged = (memcmp(&pCfg->sync, &pVnode->cfg.sync, sizeof(SSyncCluster)) != 0);
+
+  if (!walChanged && !tsdbChanged && !metaChanged && !syncChanged) {
+    vDebug("vgId:%d, nothing changed", vgId);
+    vnodeRelease(pVnode);
+    return code;
+  }
+
+  code = vnodeWriteCfg(pVnode->vgId, pCfg);
+  if (code != 0) {
+    vError("vgId:%d, failed to write alter msg to file since %s", vgId, tstrerror(code));
+    vnodeRelease(pVnode);
+    return code;
+  }
+
+  pVnode->cfg = *pCfg;
+
+  if (walChanged) {
+    code = walAlter(pVnode->pWal, &pVnode->cfg.wal);
+    if (code != 0) {
+      vDebug("vgId:%d, failed to alter wal since %s", vgId, tstrerror(code));
+      vnodeRelease(pVnode);
+      return code;
+    }
+  }
+
+  if (tsdbChanged) {
+    // todo
+  }
+
+  if (metaChanged) {
+    // todo
+  }
+
+  if (syncChanged) {
+    syncReconfig(pVnode->pSync, &pVnode->cfg.sync);
+  }
+
+  vnodeRelease(pVnode);
+  return code;
+}
+
+int32_t vnodeDropVnode(SVnode *pVnode) {
+  if (pVnode->cfg.dropped) {
+    vInfo("vgId:%d, already set drop flag, ref:%d", pVnode->vgId, pVnode->refCount);
+    vnodeRelease(pVnode);
+    return TSDB_CODE_SUCCESS;
+  }
+
+  pVnode->cfg.dropped = 1;
+  int32_t code = vnodeWriteCfg(pVnode->vgId, &pVnode->cfg);
+  if (code == 0) {
+    vInfo("vgId:%d, set drop flag, ref:%d", pVnode->vgId, pVnode->refCount);
+    vnodeCleanupVnode(pVnode);
+  } else {
+    vError("vgId:%d, failed to set drop flag since %s", pVnode->vgId, tstrerror(code));
+    pVnode->cfg.dropped = 0;
+  }
+
+  vnodeRelease(pVnode);
+  return code;
+}
+
+int32_t vnodeSyncVnode(SVnode *pVnode) {
+  return TSDB_CODE_SUCCESS;
+}
+
+int32_t vnodeCompactVnode(SVnode *pVnode) {
+  return TSDB_CODE_SUCCESS;
+}
+
+static void *vnodeOpenVnodeFunc(void *param) {
+  SOpenVnodeThread *pThread = param;
+
+  vDebug("thread:%d, start to open %d vnodes", pThread->threadIndex, pThread->vnodeNum);
+  setThreadName("vnodeOpenVnode");
+
+  for (int32_t v = 0; v < pThread->vnodeNum; ++v) {
+    int32_t vgId = pThread->vnodeList[v];
+
+    char stepDesc[TSDB_STEP_DESC_LEN] = {0};
+    snprintf(stepDesc, TSDB_STEP_DESC_LEN, "vgId:%d, start to restore, %d of %d have been opened", vgId,
+             tsVnode.openVnodes, tsVnode.totalVnodes);
+    // (*vnodeInst()->fp.ReportStartup)("open-vnodes", stepDesc);
+
+    if (vnodeOpenVnode(vgId) < 0) {
+      vError("vgId:%d, failed to open vnode by thread:%d", vgId, pThread->threadIndex);
+      pThread->failed++;
+    } else {
+      vDebug("vgId:%d, is opened by thread:%d", vgId, pThread->threadIndex);
+      pThread->opened++;
+    }
+
+    atomic_add_fetch_32(&tsVnode.openVnodes, 1);
+  }
+
+  vDebug("thread:%d, total vnodes:%d, opened:%d failed:%d", pThread->threadIndex, pThread->vnodeNum, pThread->opened,
+         pThread->failed);
+  return NULL;
+}
+
+static int32_t vnodeGetVnodeListFromDisk(int32_t vnodeList[], int32_t *numOfVnodes) {
+#if 0
+  DIR *dir = opendir(tsVnodeDir);
+  if (dir == NULL) return TSDB_CODE_DND_NO_WRITE_ACCESS;
+
+  *numOfVnodes = 0;
+  struct dirent *de = NULL;
+  while ((de = readdir(dir)) != NULL) {
+    if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) continue;
+    if (de->d_type & DT_DIR) {
+      if (strncmp("vnode", de->d_name, 5) != 0) continue;
+      int32_t vnode = atoi(de->d_name + 5);
+      if (vnode == 0) continue;
+
+      (*numOfVnodes)++;
+
+      if (*numOfVnodes >= TSDB_MAX_VNODES) {
+        vError("vgId:%d, too many vnode directory in disk, exist:%d max:%d", vnode, *numOfVnodes, TSDB_MAX_VNODES);
+        closedir(dir);
+        return TSDB_CODE_DND_TOO_MANY_VNODES;
+      } else {
+        vnodeList[*numOfVnodes - 1] = vnode;
+      }
+    }
+  }
+  closedir(dir);
+#endif
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t vnodeOpenVnodes() {
+  int32_t vnodeList[TSDB_MAX_VNODES] = {0};
+  int32_t numOfVnodes = 0;
+  int32_t status = vnodeGetVnodeListFromDisk(vnodeList, &numOfVnodes);
+
+  if (status != TSDB_CODE_SUCCESS) {
+    vInfo("failed to get vnode list from disk since code:%d", status);
+    return status;
+  }
+
+  tsVnode.totalVnodes = numOfVnodes;
+
+  int32_t threadNum = tsNumOfCores;
+  int32_t vnodesPerThread = numOfVnodes / threadNum + 1;
+
+  SOpenVnodeThread *threads = calloc(threadNum, sizeof(SOpenVnodeThread));
+  for (int32_t t = 0; t < threadNum; ++t) {
+    threads[t].threadIndex = t;
+    threads[t].vnodeList = calloc(vnodesPerThread, sizeof(int32_t));
+  }
+
+  for (int32_t v = 0; v < numOfVnodes; ++v) {
+    int32_t           t = v % threadNum;
+    SOpenVnodeThread *pThread = &threads[t];
+    pThread->vnodeList[pThread->vnodeNum++] = vnodeList[v];
+  }
+
+  vInfo("start %d threads to open %d vnodes", threadNum, numOfVnodes);
+
+  for (int32_t t = 0; t < threadNum; ++t) {
+    SOpenVnodeThread *pThread = &threads[t];
+    if (pThread->vnodeNum == 0) continue;
+
+    pThread->threadId = taosCreateThread(vnodeOpenVnodeFunc, pThread);
+    if (pThread->threadId == NULL) {
+      vError("thread:%d, failed to create thread to open vnode, reason:%s", pThread->threadIndex, strerror(errno));
+    }
+  }
+
+  int32_t openVnodes = 0;
+  int32_t failedVnodes = 0;
+  for (int32_t t = 0; t < threadNum; ++t) {
+    SOpenVnodeThread *pThread = &threads[t];
+    taosDestoryThread(pThread->threadId);
+    pThread->threadId = NULL;
+
+    openVnodes += pThread->opened;
+    failedVnodes += pThread->failed;
+    free(pThread->vnodeList);
+  }
+
+  free(threads);
+  vInfo("there are total vnodes:%d, opened:%d", numOfVnodes, openVnodes);
+
+  if (failedVnodes != 0) {
+    vError("there are total vnodes:%d, failed:%d", numOfVnodes, failedVnodes);
+    return -1;
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t vnodeGetVnodeList(SVnode *vnodeList[], int32_t *numOfVnodes) {
+  void *pIter = taosHashIterate(tsVnode.hash, NULL);
+  while (pIter) {
+    SVnode **pVnode = pIter;
+    if (*pVnode) {
+      (*numOfVnodes)++;
+      if (*numOfVnodes >= TSDB_MAX_VNODES) {
+        vError("vgId:%d, too many open vnodes, exist:%d max:%d", (*pVnode)->vgId, *numOfVnodes, TSDB_MAX_VNODES);
+        continue;
+      } else {
+        vnodeList[*numOfVnodes - 1] = (*pVnode);
+      }
+    }
+
+    pIter = taosHashIterate(tsVnode.hash, pIter);
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+
+static void vnodeCleanupVnodes() {
+  SVnode* vnodeList[TSDB_MAX_VNODES] = {0};
+  int32_t numOfVnodes = 0;
+
+  int32_t code = vnodeGetVnodeList(vnodeList, &numOfVnodes);
+
+  if (code != TSDB_CODE_SUCCESS) {
+    vInfo("failed to get dnode list since code %d", code);
+    return;
+  }
+
+  for (int32_t i = 0; i < numOfVnodes; ++i) {
+    vnodeCleanupVnode(vnodeList[i]);
+  }
+
+  vInfo("total vnodes:%d are all closed", numOfVnodes);
+}
+
+static void vnodeIncRef(void *ptNode) {
+  assert(ptNode != NULL);
+
+  SVnode **ppVnode = (SVnode **)ptNode;
+  assert(ppVnode);
+  assert(*ppVnode);
+
+  SVnode *pVnode = *ppVnode;
+  atomic_add_fetch_32(&pVnode->refCount, 1);
+  vTrace("vgId:%d, get vnode, refCount:%d pVnode:%p", pVnode->vgId, pVnode->refCount, pVnode);
+}
+
+SVnode *vnodeAcquireInAllState(int32_t vgId) {
+  SVnode *pVnode = NULL;
+
+  // taosHashGetClone(tsVnode.hash, &vgId, sizeof(int32_t), vnodeIncRef, (void*)&pVnode);
+  if (pVnode == NULL) {
+    vDebug("vgId:%d, can't accquire since not exist", vgId);
+    terrno = TSDB_CODE_VND_INVALID_VGROUP_ID;
+    return NULL;
+  }
+
+  return pVnode;
+}
+
+SVnode *vnodeAcquire(int32_t vgId) {
+  SVnode *pVnode = vnodeAcquireInAllState(vgId);
+  if (pVnode == NULL) return NULL;
+
+  if (vnodeInStatus(pVnode, TAOS_VN_STATUS_READY)) {
+    return pVnode;
+  } else {
+    vDebug("vgId:%d, can't accquire since not in ready status", vgId);
+    vnodeRelease(pVnode);
+    terrno = TSDB_CODE_VND_INVALID_TSDB_STATE;
+    return NULL;
+  }
+}
+
+void vnodeRelease(SVnode *pVnode) {
+  if (pVnode == NULL) return;
+
+  int32_t refCount = atomic_sub_fetch_32(&pVnode->refCount, 1);
+  int32_t vgId = pVnode->vgId;
+
+  vTrace("vgId:%d, release vnode, refCount:%d pVnode:%p", vgId, refCount, pVnode);
+  assert(refCount >= 0);
+
+  if (refCount <= 0) {
+    vDebug("vgId:%d, vnode will be destroyed, refCount:%d pVnode:%p", vgId, refCount, pVnode);
+    vnodeDestroyVnode(pVnode);
+    int32_t count = taosHashGetSize(tsVnode.hash);
+    vDebug("vgId:%d, vnode is destroyed, vnodes:%d", vgId, count);
+  }
+}
+
+int32_t vnodeInitMain() {
+  tsVnode.hash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_ENTRY_LOCK);
+  if (tsVnode.hash == NULL) {
+    vError("failed to init vnode mgmt");
+    return -1;
+  }
+
+  vInfo("vnode main is initialized");
+  return vnodeOpenVnodes();
+}
+
+void vnodeCleanupMain() {
+  vnodeCleanupVnodes();
+  taosHashCleanup(tsVnode.hash);
+  tsVnode.hash = NULL;
+}
+
+static void vnodeBuildVloadMsg(SVnode *pVnode, SStatusMsg *pStatus) {
+  int64_t totalStorage = 0;
+  int64_t compStorage = 0;
+  int64_t pointsWritten = 0;
+
+  if (pStatus->openVnodes >= TSDB_MAX_VNODES) return;
+
+  // if (pVnode->tsdb) {
+  //   tsdbReportStat(pVnode->tsdb, &pointsWritten, &totalStorage, &compStorage);
+  // }
+
+  SVnodeLoad *pLoad = &pStatus->load[pStatus->openVnodes++];
+  pLoad->vgId = htonl(pVnode->vgId);
+  pLoad->totalStorage = htobe64(totalStorage);
+  pLoad->compStorage = htobe64(compStorage);
+  pLoad->pointsWritten = htobe64(pointsWritten);
+  pLoad->status = pVnode->status;
+  pLoad->role = pVnode->role;
+}
+
+void vnodeGetStatus(SStatusMsg *pStatus) {
+  void *pIter = taosHashIterate(tsVnode.hash, NULL);
+  while (pIter) {
+    SVnode **pVnode = pIter;
+    if (*pVnode) {
+      vnodeBuildVloadMsg(*pVnode, pStatus);
+    }
+    pIter = taosHashIterate(tsVnode.hash, pIter);
+  }
+}
+
+void vnodeSetAccess(SVgroupAccess *pAccess, int32_t numOfVnodes) {
+  for (int32_t i = 0; i < numOfVnodes; ++i) {
+    pAccess[i].vgId = htonl(pAccess[i].vgId);
+    SVnode *pVnode = vnodeAcquire(pAccess[i].vgId);
+    if (pVnode != NULL) {
+      pVnode->accessState = pAccess[i].accessState;
+      if (pVnode->accessState != TSDB_VN_ALL_ACCCESS) {
+        vDebug("vgId:%d, access state is set to %d", pAccess[i].vgId, pVnode->accessState);
+      }
+      vnodeRelease(pVnode);
+    }
+  }
+}

source/server/vnode/src/vnodeMgmt.c

@@ -15,21 +15,185 @@
 
 #define _DEFAULT_SOURCE
 #include "os.h"
-
+#include "vnodeMain.h"
 #include "vnodeMgmt.h"
-#include "vnodeMgmtMsg.h"
-
-typedef struct {
-  SRpcMsg rpcMsg;
-  char    pCont[];
-} SVnMgmtMsg;
 
 static struct {
-  SWorkerPool pool;
-  taos_queue  pQueue;
+  SWorkerPool createPool;
+  taos_queue  createQueue;
+  SWorkerPool workerPool;
+  taos_queue  workerQueue;
   int32_t (*msgFp[TSDB_MSG_TYPE_MAX])(SRpcMsg *);
 } tsVmgmt = {0};
 
+static int32_t vnodeParseCreateVnodeReq(SRpcMsg *rpcMsg, int32_t *vgId, SVnodeCfg *pCfg) {
+  SCreateVnodeMsg *pCreate = rpcMsg->pCont;
+  *vgId = htonl(pCreate->vgId);
+
+  pCfg->dropped = 0;
+  pCfg->quorum = pCreate->quorum;
+  tstrncpy(pCfg->db, pCreate->db, sizeof(pCfg->db));
+
+  pCfg->tsdb.cacheBlockSize = htonl(pCreate->cacheBlockSize);
+  pCfg->tsdb.totalBlocks = htonl(pCreate->totalBlocks);
+  pCfg->tsdb.daysPerFile = htonl(pCreate->daysPerFile);
+  pCfg->tsdb.daysToKeep1 = htonl(pCreate->daysToKeep1);
+  pCfg->tsdb.daysToKeep2 = htonl(pCreate->daysToKeep2);
+  pCfg->tsdb.daysToKeep0 = htonl(pCreate->daysToKeep0);
+  pCfg->tsdb.minRowsPerFileBlock = htonl(pCreate->minRowsPerFileBlock);
+  pCfg->tsdb.maxRowsPerFileBlock = htonl(pCreate->maxRowsPerFileBlock);
+  pCfg->tsdb.precision = pCreate->precision;
+  pCfg->tsdb.compression = pCreate->compression;
+  pCfg->tsdb.cacheLastRow = pCreate->cacheLastRow;
+  pCfg->tsdb.update = pCreate->update;
+
+  pCfg->wal.fsyncPeriod = htonl(pCreate->fsyncPeriod);
+  pCfg->wal.walLevel = pCreate->walLevel;
+
+  pCfg->sync.replica = pCreate->replica;
+  pCfg->sync.selfIndex = pCreate->selfIndex;
+  
+  for (int32_t j = 0; j < pCreate->replica; ++j) {
+    pCfg->sync.nodeInfo[j].nodePort = htons(pCreate->nodes[j].port);
+    tstrncpy(pCfg->sync.nodeInfo[j].nodeFqdn, pCreate->nodes[j].fqdn, TSDB_FQDN_LEN);
+  }
+
+  return 0;
+}
+
+static int32_t vnodeProcessCreateVnodeReq(SRpcMsg *rpcMsg) {
+  SVnodeCfg vnodeCfg = {0};
+  int32_t   vgId = 0;
+
+  int32_t code = vnodeParseCreateVnodeReq(rpcMsg, &vgId, &vnodeCfg);
+  if (code != 0) {
+    vError("failed to parse create vnode msg since %s", tstrerror(code));
+  }
+
+  vDebug("vgId:%d, create vnode req is received", vgId);
+
+  SVnode *pVnode = vnodeAcquireInAllState(vgId);
+  if (pVnode != NULL) {
+    vDebug("vgId:%d, already exist, return success", vgId);
+    vnodeRelease(pVnode);
+    return code;
+  }
+
+  code = vnodeCreateVnode(vgId, &vnodeCfg);
+  if (code != 0) {
+    vError("vgId:%d, failed to create vnode since %s", vgId, tstrerror(code));
+  }
+
+  return code;
+}
+
+static int32_t vnodeProcessAlterVnodeReq(SRpcMsg *rpcMsg) {
+  SVnodeCfg vnodeCfg = {0};
+  int32_t   vgId = 0;
+
+  int32_t code = vnodeParseCreateVnodeReq(rpcMsg, &vgId, &vnodeCfg);
+  if (code != 0) {
+    vError("failed to parse create vnode msg since %s", tstrerror(code));
+  }
+
+  vDebug("vgId:%d, alter vnode req is received", vgId);
+
+  SVnode *pVnode = vnodeAcquire(vgId);
+  if (pVnode == NULL) {
+    code = terrno;
+    vDebug("vgId:%d, failed to alter vnode since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  code = vnodeAlterVnode(pVnode, &vnodeCfg);
+  if (code != 0) {
+    vError("vgId:%d, failed to alter vnode since %s", vgId, tstrerror(code));
+  }
+
+  vnodeRelease(pVnode);
+  return code;
+}
+
+static SDropVnodeMsg *vnodeParseDropVnodeReq(SRpcMsg *rpcMsg) {
+  SDropVnodeMsg *pDrop = rpcMsg->pCont;
+  pDrop->vgId = htonl(pDrop->vgId);
+  return pDrop;
+}
+
+static int32_t vnodeProcessSyncVnodeReq(SRpcMsg *rpcMsg) {
+  SSyncVnodeMsg *pSync = (SSyncVnodeMsg *)vnodeParseDropVnodeReq(rpcMsg);
+
+  int32_t code = 0;
+  int32_t vgId = pSync->vgId;
+  vDebug("vgId:%d, sync vnode req is received", vgId);
+
+  SVnode *pVnode = vnodeAcquire(vgId);
+  if (pVnode == NULL) {
+    code = terrno;
+    vDebug("vgId:%d, failed to sync since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  code = vnodeSyncVnode(pVnode);
+  if (code != 0) {
+    vError("vgId:%d, failed to compact vnode since %s", vgId, tstrerror(code));
+  }
+
+  vnodeRelease(pVnode);
+  return code;
+}
+
+static int32_t vnodeProcessCompactVnodeReq(SRpcMsg *rpcMsg) {
+  SCompactVnodeMsg *pCompact = (SCompactVnodeMsg *)vnodeParseDropVnodeReq(rpcMsg);
+
+  int32_t code = 0;
+  int32_t vgId = pCompact->vgId;
+  vDebug("vgId:%d, compact vnode req is received", vgId);
+
+  SVnode *pVnode = vnodeAcquire(vgId);
+  if (pVnode == NULL) {
+    code = terrno;
+    vDebug("vgId:%d, failed to compact since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  code = vnodeCompactVnode(pVnode);
+  if (code != 0) {
+    vError("vgId:%d, failed to compact vnode since %s", vgId, tstrerror(code));
+  }
+
+  vnodeRelease(pVnode);
+  return code;
+}
+
+static int32_t vnodeProcessDropVnodeReq(SRpcMsg *rpcMsg) {
+  SDropVnodeMsg *pDrop = vnodeParseDropVnodeReq(rpcMsg);
+
+  int32_t code = 0;
+  int32_t vgId = pDrop->vgId;
+  vDebug("vgId:%d, drop vnode req is received", vgId);
+
+  SVnode *pVnode = vnodeAcquire(vgId);
+  if (pVnode == NULL) {
+    code = terrno;
+    vDebug("vgId:%d, failed to drop since %s", vgId, tstrerror(code));
+    return code;
+  }
+
+  code = vnodeDropVnode(pVnode);
+  if (code != 0) {
+    vError("vgId:%d, failed to drop vnode since %s", vgId, tstrerror(code));
+  }
+
+  vnodeRelease(pVnode);
+  return code;
+}
+
+static int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg) {
+  vError("alter stream msg not processed");
+  return TSDB_CODE_VND_MSG_NOT_PROCESSED;
+}
+
 static int32_t vnodeProcessMgmtStart(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype) {
   SRpcMsg *pMsg = &pMgmt->rpcMsg;
   int32_t  msgType = pMsg->msgType;
@@ -43,27 +207,21 @@ static int32_t vnodeProcessMgmtStart(void *unused, SVnMgmtMsg *pMgmt, int32_t qt
   }
 }
 
-static void vnodeSendMgmtEnd(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype, int32_t code) {
+static void vnodeProcessMgmtEnd(void *unused, SVnMgmtMsg *pMgmt, int32_t qtype, int32_t code) {
   SRpcMsg *pMsg = &pMgmt->rpcMsg;
-  SRpcMsg  rsp = {0};
+  vTrace("msg:%p, is processed, result:%s", pMgmt, tstrerror(code));
 
-  rsp.code = code;
-  vTrace("msg:%p, is processed, code:0x%x", pMgmt, rsp.code);
-  if (rsp.code != TSDB_CODE_DND_ACTION_IN_PROGRESS) {
-    rsp.handle = pMsg->handle;
-    rsp.pCont = NULL;
-    rpcSendResponse(&rsp);
-  }
-
-  taosFreeQitem(pMsg);
+  SRpcMsg rsp = {.code = code, .handle = pMsg->handle};
+  rpcSendResponse(&rsp);
+  taosFreeQitem(pMgmt);
 }
 
 static void vnodeInitMgmtReqFp() {
-  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessCreateVnodeMsg;
-  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE]  = vnodeProcessAlterVnodeMsg;
-  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE]   = vnodeProcessSyncVnodeMsg;
-  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE]= vnodeProcessCompactVnodeMsg;
-  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE]   = vnodeProcessDropVnodeMsg;
+  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_CREATE_VNODE] = vnodeProcessCreateVnodeReq;
+  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_VNODE] = vnodeProcessAlterVnodeReq;
+  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_SYNC_VNODE] = vnodeProcessSyncVnodeReq;
+  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_COMPACT_VNODE] = vnodeProcessCompactVnodeReq;
+  tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_DROP_VNODE] = vnodeProcessDropVnodeReq;
   tsVmgmt.msgFp[TSDB_MSG_TYPE_MD_ALTER_STREAM] = vnodeProcessAlterStreamReq;
 }
 
@@ -75,14 +233,18 @@ static int32_t vnodeWriteToMgmtQueue(SRpcMsg *pMsg) {
   pMgmt->rpcMsg = *pMsg;
   pMgmt->rpcMsg.pCont = pMgmt->pCont;
   memcpy(pMgmt->pCont, pMsg->pCont, pMsg->contLen);
-  taosWriteQitem(tsVmgmt.pQueue, TAOS_QTYPE_RPC, pMgmt);
 
-  return TSDB_CODE_SUCCESS;
+  if (pMsg->msgType == TSDB_MSG_TYPE_MD_CREATE_VNODE) {
+    return taosWriteQitem(tsVmgmt.createQueue, TAOS_QTYPE_RPC, pMgmt);
+  } else {
+    return taosWriteQitem(tsVmgmt.workerQueue, TAOS_QTYPE_RPC, pMgmt);
+  }
 }
 
 void vnodeProcessMgmtMsg(SRpcMsg *pMsg) {
   int32_t code = vnodeWriteToMgmtQueue(pMsg);
   if (code != TSDB_CODE_SUCCESS) {
+    vError("msg, ahandle:%p type:%s not processed since %s", pMsg->ahandle, taosMsg[pMsg->msgType], tstrerror(code));
     SRpcMsg rsp = {.handle = pMsg->handle, .code = code};
     rpcSendResponse(&rsp);
   }
@@ -93,25 +255,41 @@ void vnodeProcessMgmtMsg(SRpcMsg *pMsg) {
 int32_t vnodeInitMgmt() {
   vnodeInitMgmtReqFp();
 
-  SWorkerPool *pPool = &tsVmgmt.pool;
-  pPool->name = "vmgmt";
+  SWorkerPool *pPool = &tsVmgmt.createPool;
+  pPool->name = "vnode-mgmt-create";
   pPool->startFp = (ProcessStartFp)vnodeProcessMgmtStart;
-  pPool->endFp = (ProcessEndFp)vnodeSendMgmtEnd;
+  pPool->endFp = (ProcessEndFp)vnodeProcessMgmtEnd;
   pPool->min = 1;
   pPool->max = 1;
   if (tWorkerInit(pPool) != 0) {
     return TSDB_CODE_VND_OUT_OF_MEMORY;
   }
 
-  tsVmgmt.pQueue = tWorkerAllocQueue(pPool, NULL);
+  tsVmgmt.createQueue = tWorkerAllocQueue(pPool, NULL);
 
-  vInfo("vmgmt is initialized, max worker %d", pPool->max);
+  pPool = &tsVmgmt.workerPool;
+  pPool->name = "vnode-mgmt-worker";
+  pPool->startFp = (ProcessStartFp)vnodeProcessMgmtStart;
+  pPool->endFp = (ProcessEndFp)vnodeProcessMgmtEnd;
+  pPool->min = 1;
+  pPool->max = 1;
+  if (tWorkerInit(pPool) != 0) {
+    return TSDB_CODE_VND_OUT_OF_MEMORY;
+  }
+
+  tsVmgmt.workerQueue = tWorkerAllocQueue(pPool, NULL);
+
+  vInfo("vmgmt is initialized");
   return TSDB_CODE_SUCCESS;
 }
 
 void vnodeCleanupMgmt() {
-  tWorkerFreeQueue(&tsVmgmt.pool, tsVmgmt.pQueue);
-  tWorkerCleanup(&tsVmgmt.pool);
-  tsVmgmt.pQueue = NULL;
+  tWorkerFreeQueue(&tsVmgmt.createPool, tsVmgmt.createQueue);
+  tWorkerCleanup(&tsVmgmt.createPool);
+  tsVmgmt.createQueue = NULL;
+
+  tWorkerFreeQueue(&tsVmgmt.workerPool, tsVmgmt.workerQueue);
+  tWorkerCleanup(&tsVmgmt.workerPool);
+  tsVmgmt.createQueue = NULL;
   vInfo("vmgmt is closed");
 }

source/server/vnode/src/vnodeMgmtMsg.c

-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "os.h"
-
-#include "vnodeMgmtMsg.h"
-
-static SCreateVnodeMsg* vnodeParseVnodeMsg(SRpcMsg *rpcMsg) {
-  SCreateVnodeMsg *pCreate = rpcMsg->pCont;
-  pCreate->cfg.vgId                = htonl(pCreate->cfg.vgId);
-  pCreate->cfg.dbCfgVersion        = htonl(pCreate->cfg.dbCfgVersion);
-  pCreate->cfg.vgCfgVersion        = htonl(pCreate->cfg.vgCfgVersion);
-  pCreate->cfg.maxTables           = htonl(pCreate->cfg.maxTables);
-  pCreate->cfg.cacheBlockSize      = htonl(pCreate->cfg.cacheBlockSize);
-  pCreate->cfg.totalBlocks         = htonl(pCreate->cfg.totalBlocks);
-  pCreate->cfg.daysPerFile         = htonl(pCreate->cfg.daysPerFile);
-  pCreate->cfg.daysToKeep1         = htonl(pCreate->cfg.daysToKeep1);
-  pCreate->cfg.daysToKeep2         = htonl(pCreate->cfg.daysToKeep2);
-  pCreate->cfg.daysToKeep          = htonl(pCreate->cfg.daysToKeep);
-  pCreate->cfg.minRowsPerFileBlock = htonl(pCreate->cfg.minRowsPerFileBlock);
-  pCreate->cfg.maxRowsPerFileBlock = htonl(pCreate->cfg.maxRowsPerFileBlock);
-  pCreate->cfg.fsyncPeriod         = htonl(pCreate->cfg.fsyncPeriod);
-  pCreate->cfg.commitTime          = htonl(pCreate->cfg.commitTime);
-
-  for (int32_t j = 0; j < pCreate->cfg.vgReplica; ++j) {
-    pCreate->nodes[j].nodeId = htonl(pCreate->nodes[j].nodeId);
-  }
-
-  return pCreate;
-}
-
-int32_t vnodeProcessCreateVnodeMsg(SRpcMsg *rpcMsg) {
-  SCreateVnodeMsg *pCreate = vnodeParseVnodeMsg(rpcMsg);
-  SVnode *pVnode = vnodeAcquire(pCreate->cfg.vgId);
-  if (pVnode != NULL) {
-    vDebug("vgId:%d, already exist, return success", pCreate->cfg.vgId);
-    vnodeRelease(pVnode);
-    return TSDB_CODE_SUCCESS;
-  } else {
-    vDebug("vgId:%d, create vnode msg is received", pCreate->cfg.vgId);
-    return vnodeCreate(pCreate);
-  }
-}
-
-int32_t vnodeProcessAlterVnodeMsg(SRpcMsg *rpcMsg) {
-  SAlterVnodeMsg *pAlter = vnodeParseVnodeMsg(rpcMsg);
-
-  void *pVnode = vnodeAcquireNotClose(pAlter->cfg.vgId);
-  if (pVnode != NULL) {
-    vDebug("vgId:%d, alter vnode msg is received", pAlter->cfg.vgId);
-    int32_t code = vnodeAlter(pVnode, pAlter);
-    vnodeRelease(pVnode);
-    return code;
-  } else {
-    vInfo("vgId:%d, vnode not exist, can't alter it", pAlter->cfg.vgId);
-    return TSDB_CODE_VND_INVALID_VGROUP_ID;
-  }
-}
-
-int32_t vnodeProcessSyncVnodeMsg(SRpcMsg *rpcMsg) {
-  SSyncVnodeMsg *pSyncVnode = rpcMsg->pCont;
-  pSyncVnode->vgId = htonl(pSyncVnode->vgId);
-
-  return vnodeSync(pSyncVnode->vgId);
-}
-
-int32_t vnodeProcessCompactVnodeMsg(SRpcMsg *rpcMsg) {
-  SCompactVnodeMsg *pCompactVnode = rpcMsg->pCont;
-  pCompactVnode->vgId = htonl(pCompactVnode->vgId);
-  return vnodeCompact(pCompactVnode->vgId);
-}
-
-int32_t vnodeProcessDropVnodeMsg(SRpcMsg *rpcMsg) {
-  SDropVnodeMsg *pDrop = rpcMsg->pCont;
-  pDrop->vgId = htonl(pDrop->vgId);
-
-  return vnodeDrop(pDrop->vgId);
-}
-
-int32_t vnodeProcessAlterStreamReq(SRpcMsg *pMsg) { return 0; }

source/server/vnode/src/vnodeRead.c

@@ -14,14 +14,9 @@
  */
 
 #define _DEFAULT_SOURCE
-#include "os.h"
-#include "taosmsg.h"
-#include "tglobal.h"
-// #include "query.h"
-
+#include "vnodeMain.h"
 #include "vnodeRead.h"
 #include "vnodeReadMsg.h"
-#include "vnodeStatus.h"
 
 static struct {
   SWorkerPool query;
@@ -50,11 +45,6 @@ static int32_t vnodeWriteToRQueue(SVnode *pVnode, void *pCont, int32_t contLen,
   }
 #endif  
 
-  if (!vnodeInReadyStatus(pVnode)) {
-    vDebug("vgId:%d, failed to write into vread queue, vnode status is %s", pVnode->vgId, vnodeStatus[pVnode->status]);
-    return TSDB_CODE_APP_NOT_READY;
-  }
-
   int32_t   size = sizeof(SReadMsg) + contLen;
   SReadMsg *pRead = taosAllocateQitem(size);
   if (pRead == NULL) {
@@ -119,7 +109,7 @@ void vnodeProcessReadMsg(SRpcMsg *pMsg) {
     pHead->contLen = htonl(pHead->contLen);
 
     assert(pHead->contLen > 0);
-    SVnode *pVnode = vnodeAcquireNotClose(pHead->vgId);
+    SVnode *pVnode = vnodeAcquire(pHead->vgId);
     if (pVnode != NULL) {
       code = vnodeWriteToRQueue(pVnode, pCont, pHead->contLen, TAOS_QTYPE_RPC, pMsg);
       if (code == TSDB_CODE_SUCCESS) queuedMsgNum++;

source/server/vnode/src/vnodeReadMsg.c

@@ -14,11 +14,7 @@
  */
 
 #define _DEFAULT_SOURCE
-#include "os.h"
-#include "taosmsg.h"
-#include "tglobal.h"
-// #include "query.h"
-#include "vnodeStatus.h"
+#include "vnodeMain.h"
 #include "vnodeRead.h"
 #include "vnodeReadMsg.h"
 

source/server/vnode/src/vnodeStatus.c

-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "os.h"
-#include "taosmsg.h"
-// #include "query.h"
-#include "vnodeRead.h"
-#include "vnodeStatus.h"
-#include "vnodeWrite.h"
-
-char* vnodeStatus[] = {
-  "init",
-  "ready",
-  "closing",
-  "updating",
-  "reset"
-};
-
-bool vnodeSetInitStatus(SVnode* pVnode) {
-  pthread_mutex_lock(&pVnode->statusMutex);
-  pVnode->status = TAOS_VN_STATUS_INIT;
-  pthread_mutex_unlock(&pVnode->statusMutex);
-  return true;
-}
-
-bool vnodeSetReadyStatus(SVnode* pVnode) {
-  bool set = false;
-  pthread_mutex_lock(&pVnode->statusMutex);
-
-  if (pVnode->status == TAOS_VN_STATUS_INIT || pVnode->status == TAOS_VN_STATUS_READY ||
-      pVnode->status == TAOS_VN_STATUS_UPDATING) {
-    pVnode->status = TAOS_VN_STATUS_READY;
-    set = true;
-  }
-
-#if 0
-  qQueryMgmtReOpen(pVnode->qMgmt);
-#endif   
-
-  pthread_mutex_unlock(&pVnode->statusMutex);
-  return set;
-}
-
-static bool vnodeSetClosingStatusImp(SVnode* pVnode) {
-  bool set = false;
-  pthread_mutex_lock(&pVnode->statusMutex);
-
-  if (pVnode->status == TAOS_VN_STATUS_READY || pVnode->status == TAOS_VN_STATUS_INIT) {
-    pVnode->status = TAOS_VN_STATUS_CLOSING;
-    set = true;
-  }
-
-  pthread_mutex_unlock(&pVnode->statusMutex);
-  return set;
-}
-
-bool vnodeSetClosingStatus(SVnode* pVnode) {
-  if (pVnode->status == TAOS_VN_STATUS_CLOSING)
-    return true;
-
-  while (!vnodeSetClosingStatusImp(pVnode)) {
-    taosMsleep(1);
-  }
-
-#if 0
-  // release local resources only after cutting off outside connections
-  qQueryMgmtNotifyClosed(pVnode->qMgmt);
-#endif   
-  vnodeWaitReadCompleted(pVnode);
-  vnodeWaitWriteCompleted(pVnode);
-
-  return true;
-}
-
-bool vnodeSetUpdatingStatus(SVnode* pVnode) {
-  bool set = false;
-  pthread_mutex_lock(&pVnode->statusMutex);
-
-  if (pVnode->status == TAOS_VN_STATUS_READY) {
-    pVnode->status = TAOS_VN_STATUS_UPDATING;
-    set = true;
-  }
-
-  pthread_mutex_unlock(&pVnode->statusMutex);
-  return set;
-}
-
-bool vnodeInInitStatus(SVnode* pVnode) {
-  bool in = false;
-  pthread_mutex_lock(&pVnode->statusMutex);
-
-  if (pVnode->status == TAOS_VN_STATUS_INIT) {
-    in = true;
-  }
-
-  pthread_mutex_unlock(&pVnode->statusMutex);
-  return in;
-}
-
-bool vnodeInReadyStatus(SVnode* pVnode) {
-  bool in = false;
-  pthread_mutex_lock(&pVnode->statusMutex);
-
-  if (pVnode->status == TAOS_VN_STATUS_READY) {
-    in = true;
-  }
-
-  pthread_mutex_unlock(&pVnode->statusMutex);
-  return in;
-}
-
-bool vnodeInClosingStatus(SVnode* pVnode) {
-  bool in = false;
-  pthread_mutex_lock(&pVnode->statusMutex);
-
-  if (pVnode->status == TAOS_VN_STATUS_CLOSING) {
-    in = true;
-  }
-
-  pthread_mutex_unlock(&pVnode->statusMutex);
-  return in;
-}
-

source/server/vnode/src/vnodeVersion.c

-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "os.h"
-#include "cJSON.h"
-#include "tglobal.h"
-#include "vnodeVersion.h"
-
-int32_t vnodeReadVersion(SVnode *pVnode) {
-  int32_t len = 0;
-  int32_t maxLen = 100;
-  char *  content = calloc(1, maxLen + 1);
-  cJSON * root = NULL;
-  FILE *  fp = NULL;
-
-  terrno = TSDB_CODE_VND_INVALID_VRESION_FILE;
-  char file[TSDB_FILENAME_LEN + 30] = {0};
-  sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId);
-
-  fp = fopen(file, "r");
-  if (!fp) {
-    if (errno != ENOENT) {
-      vError("vgId:%d, failed to read %s, error:%s", pVnode->vgId, file, strerror(errno));
-      terrno = TAOS_SYSTEM_ERROR(errno);
-    } else {
-      terrno = TSDB_CODE_SUCCESS;
-    }
-    goto PARSE_VER_ERROR;
-  }
-
-  len = (int32_t)fread(content, 1, maxLen, fp);
-  if (len <= 0) {
-    vError("vgId:%d, failed to read %s, content is null", pVnode->vgId, file);
-    goto PARSE_VER_ERROR;
-  }
-
-  root = cJSON_Parse(content);
-  if (root == NULL) {
-    vError("vgId:%d, failed to read %s, invalid json format", pVnode->vgId, file);
-    goto PARSE_VER_ERROR;
-  }
-
-  cJSON *ver = cJSON_GetObjectItem(root, "version");
-  if (!ver || ver->type != cJSON_Number) {
-    vError("vgId:%d, failed to read %s, version not found", pVnode->vgId, file);
-    goto PARSE_VER_ERROR;
-  }
-#if 0  
-  pVnode->version = (uint64_t)ver->valueint;
-
-  terrno = TSDB_CODE_SUCCESS;
-  vInfo("vgId:%d, read %s successfully, fver:%" PRIu64, pVnode->vgId, file, pVnode->version);
-#endif
-
-PARSE_VER_ERROR:
-  if (content != NULL) free(content);
-  if (root != NULL) cJSON_Delete(root);
-  if (fp != NULL) fclose(fp);
-
-  return terrno;
-}
-
-int32_t vnodeSaveVersion(SVnode *pVnode) {
-  char file[TSDB_FILENAME_LEN + 30] = {0};
-  sprintf(file, "%s/vnode%d/version.json", tsVnodeDir, pVnode->vgId);
-
-  FILE *fp = fopen(file, "w");
-  if (!fp) {
-    vError("vgId:%d, failed to write %s, reason:%s", pVnode->vgId, file, strerror(errno));
-    return -1;
-  }
-
-  int32_t len = 0;
-  int32_t maxLen = 100;
-  char *  content = calloc(1, maxLen + 1);
-
-#if 0
-  len += snprintf(content + len, maxLen - len, "{\n");
-  len += snprintf(content + len, maxLen - len, "  \"version\": %" PRIu64 "\n", pVnode->fversion);
-  len += snprintf(content + len, maxLen - len, "}\n");
-#endif
-  fwrite(content, 1, len, fp);
-  taosFsyncFile(fileno(fp));
-  fclose(fp);
-  free(content);
-  terrno = 0;
-
-  // vInfo("vgId:%d, successed to write %s, fver:%" PRIu64, pVnode->vgId, file, pVnode->fversion);
-  return TSDB_CODE_SUCCESS;
-}
\ No newline at end of file

source/server/vnode/src/vnodeWorker.c

-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "os.h"
-
-#include "vnodeWorker.h"
-
-enum { CLEANUP_TASK = 0, DESTROY_TASK = 1, BACKUP_TASK = 2 };
-
-typedef struct {
-  int32_t vgId;
-  int32_t code;
-  int32_t type;
-  void *  rpcHandle;
-  SVnode *pVnode;
-} SVnTask;
-
-static struct {
-  SWorkerPool pool;
-  taos_queue  pQueue;
-} tsVworker = {0};
-
-static void vnodeProcessTaskStart(void *unused, SVnTask *pTask, int32_t qtype) {
-  pTask->code = 0;
-
-  switch (pTask->type) {
-    case CLEANUP_TASK:
-      vnodeCleanUp(pTask->pVnode);
-      break;
-    case DESTROY_TASK:
-      vnodeDestroy(pTask->pVnode);
-      break;
-    case BACKUP_TASK:
-      vnodeBackup(pTask->vgId);
-      break;
-    default:
-      break;
-  }
-}
-
-static void vnodeProcessTaskEnd(void *unused, SVnTask *pTask, int32_t qtype, int32_t code) {
-  if (pTask->rpcHandle != NULL) {
-    SRpcMsg rpcRsp = {.handle = pTask->rpcHandle, .code = pTask->code};
-    rpcSendResponse(&rpcRsp);
-  }
-
-  taosFreeQitem(pTask);
-}
-
-static int32_t vnodeWriteIntoTaskQueue(SVnode *pVnode, int32_t type, void *rpcHandle) {
-  SVnTask *pTask = taosAllocateQitem(sizeof(SVnTask));
-  if (pTask == NULL) return TSDB_CODE_VND_OUT_OF_MEMORY;
-
-  pTask->vgId = pVnode->vgId;
-  pTask->pVnode = pVnode;
-  pTask->rpcHandle = rpcHandle;
-  pTask->type = type;
-
-
-  return taosWriteQitem(tsVworker.pQueue, TAOS_QTYPE_RPC, pTask);
-}
-
-void vnodeProcessCleanupTask(SVnode *pVnode) {
-  vnodeWriteIntoTaskQueue(pVnode, CLEANUP_TASK, NULL);
-}
-
-void vnodeProcessDestroyTask(SVnode *pVnode) {
-  vnodeWriteIntoTaskQueue(pVnode, DESTROY_TASK, NULL);
-}
-
-void vnodeProcessBackupTask(SVnode *pVnode) {
-  vnodeWriteIntoTaskQueue(pVnode, BACKUP_TASK, NULL);
-}
-
-int32_t vnodeInitWorker() {
-  SWorkerPool *pPool = &tsVworker.pool;
-  pPool->name = "vworker";
-  pPool->startFp = (ProcessStartFp)vnodeProcessTaskStart;
-  pPool->endFp = (ProcessEndFp)vnodeProcessTaskEnd;
-  pPool->min = 0;
-  pPool->max = 1;
-  if (tWorkerInit(pPool) != 0) {
-    return TSDB_CODE_VND_OUT_OF_MEMORY;
-  }
-
-  tsVworker.pQueue = tWorkerAllocQueue(pPool, NULL);
-
-  vInfo("vworker is initialized, max worker %d", pPool->max);
-  return TSDB_CODE_SUCCESS;
-}
-
-void vnodeCleanupWorker() {
-  tWorkerFreeQueue(&tsVworker.pool, tsVworker.pQueue);
-  tWorkerCleanup(&tsVworker.pool);
-  tsVworker.pQueue = NULL;
-  vInfo("vworker is closed");
-}

source/server/vnode/src/vnodeWrite.c

@@ -15,12 +15,7 @@
 
 #define _DEFAULT_SOURCE
 #include "os.h"
-#include "tglobal.h"
-#include "tqueue.h"
-#include "tworker.h"
-#include "taosmsg.h"
-
-#include "vnodeStatus.h"
+#include "vnodeMain.h"
 #include "vnodeWrite.h"
 #include "vnodeWriteMsg.h"
 
@@ -68,11 +63,6 @@ static int32_t vnodeWriteToWQueue(SVnode *pVnode, SWalHead *pHead, int32_t qtype
     return TSDB_CODE_WAL_SIZE_LIMIT;
   }
 
-  if (!vnodeInReadyStatus(pVnode)) {
-    vError("vgId:%d, failed to write into vwqueue, vstatus is %s", pVnode->vgId, vnodeStatus[pVnode->status]);
-    return TSDB_CODE_APP_NOT_READY;
-  }
-
   if (tsVwrite.queuedBytes > tsMaxVnodeQueuedBytes) {
     vDebug("vgId:%d, too many bytes:%" PRId64 " in vwqueue, flow control", pVnode->vgId, tsVwrite.queuedBytes);
     return TSDB_CODE_VND_IS_FLOWCTRL;
@@ -122,7 +112,7 @@ void vnodeProcessWriteMsg(SRpcMsg *pRpcMsg) {
   pMsg->vgId = htonl(pMsg->vgId);
   pMsg->contLen = htonl(pMsg->contLen);
 
-  SVnode *pVnode = vnodeAcquireNotClose(pMsg->vgId);
+  SVnode *pVnode = vnodeAcquire(pMsg->vgId);
   if (pVnode == NULL) {
     code = TSDB_CODE_VND_INVALID_VGROUP_ID;
   } else {

source/util/src/terror.c

@@ -245,11 +245,11 @@ TAOS_DEFINE_ERROR(TSDB_CODE_VND_NO_DISK_PERMISSIONS,      "No write permission f
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_NO_SUCH_FILE_OR_DIR,      "Missing data file")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_OUT_OF_MEMORY,            "Out of memory")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_APP_ERROR,                "Unexpected generic error in vnode")
-TAOS_DEFINE_ERROR(TSDB_CODE_VND_INVALID_VRESION_FILE,     "Invalid version file")
-TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_FULL,                  "Database memory is full for commit failed")
-TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_FLOWCTRL,              "Database memory is full for waiting commit")
+TAOS_DEFINE_ERROR(TSDB_CODE_VND_INVALID_CFG_FILE,         "Invalid config file")
+TAOS_DEFINE_ERROR(TSDB_CODE_VND_INVALID_TERM_FILE,        "Invalid term file")
+TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_FLOWCTRL,              "Database memory is full")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_DROPPING,              "Database is dropping")
-TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_BALANCING,             "Database is balancing")
+TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_UPDATING,             "Database is updating")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_CLOSING,               "Database is closing")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_NOT_SYNCED,               "Database suspended")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_NO_WRITE_AUTH,            "Database write operation denied")