merge from 3.0

include/common/tmsg.h

@@ -207,7 +207,7 @@ typedef struct {
 // Submit message for one table
 typedef struct SSubmitBlk {
   int64_t uid;        // table unique id
-  int32_t tid;        // table id
+  int64_t suid;        // stable id
   int32_t padding;    // TODO just for padding here
   int32_t sversion;   // data schema version
   int32_t dataLen;    // data part length, not including the SSubmitBlk head
@@ -232,12 +232,12 @@ typedef struct {
 } SSubmitBlkIter;
 
 typedef struct {
-  int32_t totalLen;
-  int32_t len;
-  void*   pMsg;
+  int32_t     totalLen;
+  int32_t     len;
+  const void* pMsg;
 } SSubmitMsgIter;
 
-int32_t tInitSubmitMsgIter(SSubmitReq* pMsg, SSubmitMsgIter* pIter);
+int32_t tInitSubmitMsgIter(const SSubmitReq* pMsg, SSubmitMsgIter* pIter);
 int32_t tGetSubmitMsgNext(SSubmitMsgIter* pIter, SSubmitBlk** pPBlock);
 int32_t tInitSubmitBlkIter(SSubmitBlk* pBlock, SSubmitBlkIter* pIter);
 STSRow* tGetSubmitBlkNext(SSubmitBlkIter* pIter);
@@ -2093,6 +2093,11 @@ static FORCE_INLINE void tdDestroyTSmaWrapper(STSmaWrapper* pSW) {
   }
 }
 
+static FORCE_INLINE void tdFreeTSmaWrapper(STSmaWrapper* pSW) {
+  tdDestroyTSmaWrapper(pSW);
+  tfree(pSW);
+}
+
 static FORCE_INLINE int32_t tEncodeTSma(void** buf, const STSma* pSma) {
   int32_t tlen = 0;
 

include/common/ttypes.h

@@ -159,6 +159,7 @@ typedef struct {
   (IS_SIGNED_NUMERIC_TYPE(_t) || (_t) == (TSDB_DATA_TYPE_BOOL) || (_t) == (TSDB_DATA_TYPE_TIMESTAMP))
 #define IS_CONVERT_AS_UNSIGNED(_t) (IS_UNSIGNED_NUMERIC_TYPE(_t) || (_t) == (TSDB_DATA_TYPE_BOOL))
 
+// TODO remove this function
 static FORCE_INLINE bool isNull(const void *val, int32_t type) {
   switch (type) {
     case TSDB_DATA_TYPE_BOOL:

include/libs/nodes/cmdnodes.h

@@ -199,6 +199,7 @@ typedef struct SIndexOptions {
 typedef struct SCreateIndexStmt {
   ENodeType type;
   EIndexType indexType;
+  bool ignoreExists;
   char indexName[TSDB_INDEX_NAME_LEN];
   char tableName[TSDB_TABLE_NAME_LEN];
   SNodeList* pCols;
@@ -207,6 +208,7 @@ typedef struct SCreateIndexStmt {
 
 typedef struct SDropIndexStmt {
   ENodeType type;
+  bool ignoreNotExists;
   char indexName[TSDB_INDEX_NAME_LEN];
   char tableName[TSDB_TABLE_NAME_LEN];
 } SDropIndexStmt;

include/libs/parser/parser.h

@@ -60,6 +60,8 @@ int32_t qParseQuerySql(SParseContext* pCxt, SQuery** pQuery);
 
 void qDestroyQuery(SQuery* pQueryNode);
 
+int32_t qExtractResultSchema(const SNode* pRoot, int32_t* numOfCols, SSchema** pSchema);
+
 #ifdef __cplusplus
 }
 #endif

include/util/tdef.h

@@ -453,6 +453,10 @@ enum {
   SND_WORKER_TYPE__UNIQUE,
 };
 
+#define MND_VGID -1
+#define QND_VGID 1
+#define VND_VGID 0
+
 #ifdef __cplusplus
 }
 #endif

include/util/types.h

@@ -84,6 +84,8 @@ typedef uint16_t VarDataLenT;  // maxVarDataLen: 32767
 #define varDataLen(v) ((VarDataLenT *)(v))[0]
 #define varDataVal(v) ((char *)(v) + VARSTR_HEADER_SIZE)
 
+#define NCHAR_WIDTH_TO_BYTES(n)  ((n) * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE)
+
 typedef int32_t VarDataOffsetT;
 
 typedef struct tstr {

source/client/inc/clientInt.h

@@ -155,6 +155,7 @@ typedef struct SReqResultInfo {
   TAOS_FIELD*    fields;
   uint32_t       numOfCols;
   int32_t*       length;
+  char**         convertBuf;
   TAOS_ROW       row;
   SResultColumn* pCol;
   uint32_t       numOfRows;

source/client/src/clientEnv.c

@@ -169,6 +169,13 @@ static void doFreeReqResultInfo(SReqResultInfo *pResInfo) {
   tfree(pResInfo->row);
   tfree(pResInfo->pCol);
   tfree(pResInfo->fields);
+
+  if (pResInfo->convertBuf != NULL) {
+    for (int32_t i = 0; i < pResInfo->numOfCols; ++i) {
+      tfree(pResInfo->convertBuf[i]);
+    }
+    tfree(pResInfo->convertBuf);
+  }
 }
 
 static void doDestroyRequest(void *p) {

source/client/src/clientImpl.c

@@ -634,18 +634,30 @@ _return:
   for (int32_t i = 0; i < pResultInfo->numOfCols; ++i) {
     SResultColumn* pCol = &pResultInfo->pCol[i];
 
-    if (IS_VAR_DATA_TYPE(pResultInfo->fields[i].type)) {
+    int32_t type = pResultInfo->fields[i].type;
+    int32_t bytes = pResultInfo->fields[i].bytes;
+
+    if (IS_VAR_DATA_TYPE(type)) {
       if (pCol->offset[pResultInfo->current] != -1) {
         char* pStart = pResultInfo->pCol[i].offset[pResultInfo->current] + pResultInfo->pCol[i].pData;
 
         pResultInfo->length[i] = varDataLen(pStart);
         pResultInfo->row[i] = varDataVal(pStart);
+
+        if (type == TSDB_DATA_TYPE_NCHAR) {
+          int32_t len = taosUcs4ToMbs((TdUcs4*)varDataVal(pStart), varDataLen(pStart), varDataVal(pResultInfo->convertBuf[i]));
+          ASSERT(len <= bytes);
+
+          pResultInfo->row[i] = varDataVal(pResultInfo->convertBuf[i]);
+          varDataSetLen(pResultInfo->convertBuf[i], len);
+          pResultInfo->length[i] = len;
+        }
       } else {
         pResultInfo->row[i] = NULL;
       }
     } else {
       if (!colDataIsNull_f(pCol->nullbitmap, pResultInfo->current)) {
-        pResultInfo->row[i] = pResultInfo->pCol[i].pData + pResultInfo->fields[i].bytes * pResultInfo->current;
+        pResultInfo->row[i] = pResultInfo->pCol[i].pData + bytes * pResultInfo->current;
       } else {
         pResultInfo->row[i] = NULL;
       }
@@ -661,13 +673,20 @@ static int32_t doPrepareResPtr(SReqResultInfo* pResInfo) {
     pResInfo->row    = calloc(pResInfo->numOfCols, POINTER_BYTES);
     pResInfo->pCol   = calloc(pResInfo->numOfCols, sizeof(SResultColumn));
     pResInfo->length = calloc(pResInfo->numOfCols, sizeof(int32_t));
-  }
+    pResInfo->convertBuf = calloc(pResInfo->numOfCols, POINTER_BYTES);
 
-  if (pResInfo->row == NULL || pResInfo->pCol == NULL || pResInfo->length == NULL) {
-    return TSDB_CODE_OUT_OF_MEMORY;
-  } else {
-    return TSDB_CODE_SUCCESS;
+    if (pResInfo->row == NULL || pResInfo->pCol == NULL || pResInfo->length == NULL || pResInfo->convertBuf == NULL) {
+      return TSDB_CODE_OUT_OF_MEMORY;
+    }
+
+    for(int32_t i = 0; i < pResInfo->numOfCols; ++i) {
+      if(pResInfo->fields[i].type == TSDB_DATA_TYPE_NCHAR) {
+        pResInfo->convertBuf[i] = calloc(1, NCHAR_WIDTH_TO_BYTES(pResInfo->fields[i].bytes));
+      }
+    }
   }
+
+  return TSDB_CODE_SUCCESS;
 }
 
 int32_t setResultDataPtr(SReqResultInfo* pResultInfo, TAOS_FIELD* pFields, int32_t numOfCols, int32_t numOfRows) {

source/client/test/clientTests.cpp

@@ -52,7 +52,7 @@ TEST(testCase, driverInit_Test) {
   // taosInitGlobalCfg();
 //  taos_init();
 }
-
+#if 0
 TEST(testCase, connect_Test) {
 //  taos_options(TSDB_OPTION_CONFIGDIR, "/home/ubuntu/first/cfg");
 
@@ -652,6 +652,7 @@ TEST(testCase, projection_query_stables) {
   taos_free_result(pRes);
   taos_close(pConn);
 }
+#endif
 
 TEST(testCase, agg_query_tables) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
@@ -660,7 +661,7 @@ TEST(testCase, agg_query_tables) {
   TAOS_RES* pRes = taos_query(pConn, "use abc1");
   taos_free_result(pRes);
 
-  pRes = taos_query(pConn, "select count(*), sum(k),min(k),max(k) from tu");
+  pRes = taos_query(pConn, "select count(*) from tu");
   if (taos_errno(pRes) != 0) {
     printf("failed to select from table, reason:%s\n", taos_errstr(pRes));
     taos_free_result(pRes);

source/common/src/tglobal.c

@@ -180,6 +180,10 @@ static int32_t taosSetTfsCfg(SConfig *pCfg) {
       memcpy(&tsDiskCfg[index], pCfg, sizeof(SDiskCfg));
       if (pCfg->level == 0 && pCfg->primary == 1) {
         tstrncpy(tsDataDir, pCfg->dir, PATH_MAX);
+        if (taosMkDir(tsDataDir) != 0) {
+          uError("failed to create dataDir:%s since %s", tsDataDir, terrstr());
+          return -1;
+        }
       }
       if (taosMkDir(pCfg->dir) != 0) {
         uError("failed to create tfsDir:%s since %s", tsDataDir, terrstr());

source/common/src/tmsg.c

@@ -28,7 +28,7 @@
 #undef TD_MSG_SEG_CODE_
 #include "tmsgdef.h"
 
-int32_t tInitSubmitMsgIter(SSubmitReq *pMsg, SSubmitMsgIter *pIter) {
+int32_t tInitSubmitMsgIter(const SSubmitReq *pMsg, SSubmitMsgIter *pIter) {
   if (pMsg == NULL) {
     terrno = TSDB_CODE_TDB_SUBMIT_MSG_MSSED_UP;
     return -1;

source/common/src/ttime.c

@@ -458,16 +458,21 @@ int64_t taosTimeAdd(int64_t t, int64_t duration, char unit, int32_t precision) {
   if (duration == 0) {
     return t;
   }
-  if (unit == 'y') {
-    duration *= 12;
-  } else if (unit != 'n') {
+
+  if (unit != 'n' && unit != 'y') {
     return t + duration;
   }
 
+  // The following code handles the y/n time duration
+  int64_t numOfMonth = duration;
+  if (unit == 'y') {
+    numOfMonth *= 12;
+  }
+
   struct tm tm;
   time_t    tt = (time_t)(t / TSDB_TICK_PER_SECOND(precision));
   taosLocalTime(&tt, &tm);
-  int32_t mon = tm.tm_year * 12 + tm.tm_mon + (int32_t)duration;
+  int32_t mon = tm.tm_year * 12 + tm.tm_mon + (int32_t)numOfMonth;
   tm.tm_year = mon / 12;
   tm.tm_mon = mon % 12;
 
@@ -557,8 +562,7 @@ int64_t taosTimeTruncate(int64_t t, const SInterval* pInterval, int32_t precisio
 
     // not enough time range
     if (start < 0 || INT64_MAX - start > pInterval->interval - 1) {
-      end = start + pInterval->interval - 1;
-
+      end = taosTimeAdd(start, pInterval->interval, pInterval->intervalUnit, precision) - 1;
       while (end < t && ((start + pInterval->sliding) <= INT64_MAX)) {  // move forward to the correct time window
         start += pInterval->sliding;
 
@@ -574,12 +578,23 @@ int64_t taosTimeTruncate(int64_t t, const SInterval* pInterval, int32_t precisio
     }
   }
 
+  ASSERT(pInterval->offset >= 0);
+
   if (pInterval->offset > 0) {
     start = taosTimeAdd(start, pInterval->offset, pInterval->offsetUnit, precision);
     if (start > t) {
       start = taosTimeAdd(start, -pInterval->interval, pInterval->intervalUnit, precision);
+    } else {
+      // try to move current window to the left-hande-side, due to the offset effect.
+      int64_t end = taosTimeAdd(start, pInterval->interval, pInterval->intervalUnit, precision) - 1;
+      ASSERT(end >= t);
+      end = taosTimeAdd(end, -pInterval->sliding, pInterval->slidingUnit, precision);
+      if (end >= t) {
+        start = taosTimeAdd(start, -pInterval->sliding, pInterval->slidingUnit, precision);
+      }
     }
   }
+
   return start;
 }
 

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

@@ -77,7 +77,11 @@ int32_t bmStartWorker(SBnodeMgmt *pMgmt) {
     return -1;
   }
 
+  dDebug("bnode workers are initialized");
   return 0;
 }
 
-void bmStopWorker(SBnodeMgmt *pMgmt) { tMultiWorkerCleanup(&pMgmt->writeWorker); }
+void bmStopWorker(SBnodeMgmt *pMgmt) {
+  tMultiWorkerCleanup(&pMgmt->writeWorker);
+  dDebug("bnode workers are closed");
+}

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

@@ -19,7 +19,6 @@
 #include "os.h"
 
 #include "cJSON.h"
-#include "monitor.h"
 #include "tcache.h"
 #include "tcrc32c.h"
 #include "tdatablock.h"
@@ -36,8 +35,7 @@
 #include "tworker.h"
 
 #include "dnode.h"
-#include "tfs.h"
-#include "wal.h"
+#include "monitor.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -53,7 +51,6 @@ extern "C" {
 typedef enum { DNODE, VNODES, QNODE, SNODE, MNODE, BNODE, NODE_MAX } ENodeType;
 typedef enum { DND_STAT_INIT, DND_STAT_RUNNING, DND_STAT_STOPPED } EDndStatus;
 typedef enum { DND_ENV_INIT, DND_ENV_READY, DND_ENV_CLEANUP } EEnvStatus;
-typedef enum { DND_WORKER_SINGLE, DND_WORKER_MULTI } EWorkerType;
 typedef enum { PROC_SINGLE, PROC_CHILD, PROC_PARENT } EProcType;
 
 typedef struct SMgmtFp      SMgmtFp;
@@ -75,10 +72,8 @@ typedef int32_t (*DropNodeFp)(SMgmtWrapper *pWrapper, SNodeMsg *pMsg);
 typedef int32_t (*RequireNodeFp)(SMgmtWrapper *pWrapper, bool *required);
 
 typedef struct SMsgHandle {
-  int32_t       vgId;
-  NodeMsgFp     vgIdMsgFp;
-  SMgmtWrapper *pVgIdWrapper;  // Handle the case where the same message type is distributed to qnode or vnode
-  NodeMsgFp     msgFp;
+  SMgmtWrapper *pQndWrapper; 
+  SMgmtWrapper *pMndWrapper;
   SMgmtWrapper *pWrapper;
 } SMsgHandle;
 
@@ -129,29 +124,30 @@ typedef struct SDnode {
   bool         dropped;
   EDndStatus   status;
   EDndEvent    event;
-  EProcType    procType;
   SStartupReq  startup;
   TdFilePtr    pLockFile;
   STransMgmt   trans;
   SMgmtWrapper wrappers[NODE_MAX];
 } SDnode;
 
-EDndStatus    dndGetStatus(SDnode *pDnode);
-void          dndSetStatus(SDnode *pDnode, EDndStatus stat);
-SMgmtWrapper *dndAcquireWrapper(SDnode *pDnode, ENodeType nodeType);
-void          dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId);
-void          dndReportStartup(SDnode *pDnode, char *pName, char *pDesc);
-void          dndSendMonitorReport(SDnode *pDnode);
+EDndStatus dndGetStatus(SDnode *pDnode);
+void       dndSetStatus(SDnode *pDnode, EDndStatus stat);
+void       dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId);
+void       dndReportStartup(SDnode *pDnode, const char *pName, const char *pDesc);
+void       dndSendMonitorReport(SDnode *pDnode);
 
 int32_t dndSendReqToMnode(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
 int32_t dndSendReqToDnode(SMgmtWrapper *pWrapper, SEpSet *pEpSet, SRpcMsg *pMsg);
 void    dndSendRsp(SMgmtWrapper *pWrapper, SRpcMsg *pRsp);
 
 int32_t dndProcessNodeMsg(SDnode *pDnode, SNodeMsg *pMsg);
-
 int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed);
 int32_t dndWriteFile(SMgmtWrapper *pWrapper, bool deployed);
 
+SMgmtWrapper *dndAcquireWrapper(SDnode *pDnode, ENodeType nodeType);
+int32_t       dndMarkWrapper(SMgmtWrapper *pWrapper);
+void          dndReleaseWrapper(SMgmtWrapper *pWrapper);
+
 #ifdef __cplusplus
 }
 #endif

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

@@ -34,7 +34,7 @@ int32_t     dndInit();
 void        dndCleanup();
 const char *dndStatStr(EDndStatus stat);
 void        dndGetStartup(SDnode *pDnode, SStartupReq *pStartup);
-TdFilePtr   dndCheckRunning(char *dataDir);
+TdFilePtr   dndCheckRunning(const char *dataDir);
 void        dndProcessStartupReq(SDnode *pDnode, SRpcMsg *pMsg);
 
 // dndMsg.c
@@ -50,10 +50,6 @@ SDnode *dndCreate(const SDnodeOpt *pOption);
 void    dndClose(SDnode *pDnode);
 void    dndHandleEvent(SDnode *pDnode, EDndEvent event);
 
-SMgmtWrapper *dndAcquireWrapper(SDnode *pDnode, ENodeType nodeType);
-int32_t       dndMarkWrapper(SMgmtWrapper *pWrapper);
-void          dndReleaseWrapper(SMgmtWrapper *pWrapper);
-
 // dndTransport.c
 int32_t dndInitServer(SDnode *pDnode);
 void    dndCleanupServer(SDnode *pDnode);

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

@@ -50,16 +50,23 @@ int32_t dndOpenNode(SMgmtWrapper *pWrapper) {
 }
 
 void dndCloseNode(SMgmtWrapper *pWrapper) {
+  dDebug("node:%s, start to close", pWrapper->name);
   taosWLockLatch(&pWrapper->latch);
   if (pWrapper->deployed) {
     (*pWrapper->fp.closeFp)(pWrapper);
     pWrapper->deployed = false;
   }
+  taosWUnLockLatch(&pWrapper->latch);
+
+  while (pWrapper->refCount > 0) {
+    taosMsleep(10);
+  }
+
   if (pWrapper->pProc) {
     taosProcCleanup(pWrapper->pProc);
     pWrapper->pProc = NULL;
   }
-  taosWUnLockLatch(&pWrapper->latch);
+  dDebug("node:%s, has been closed", pWrapper->name);
 }
 
 static int32_t dndRunInSingleProcess(SDnode *pDnode) {

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

@@ -15,11 +15,12 @@
 
 #define _DEFAULT_SOURCE
 #include "dndInt.h"
+#include "wal.h"
 
 static int8_t once = DND_ENV_INIT;
 
 int32_t dndInit() {
-  dDebug("start to init dnode env");
+  dInfo("start to init dnode env");
   if (atomic_val_compare_exchange_8(&once, DND_ENV_INIT, DND_ENV_READY) != DND_ENV_INIT) {
     terrno = TSDB_CODE_REPEAT_INIT;
     dError("failed to init dnode env since %s", terrstr());
@@ -52,7 +53,7 @@ int32_t dndInit() {
 }
 
 void dndCleanup() {
-  dDebug("start to cleanup dnode env");
+  dInfo("start to cleanup dnode env");
   if (atomic_val_compare_exchange_8(&once, DND_ENV_READY, DND_ENV_CLEANUP) != DND_ENV_READY) {
     dError("dnode env is already cleaned up");
     return;
@@ -92,7 +93,7 @@ const char *dndStatStr(EDndStatus status) {
   }
 }
 
-void dndReportStartup(SDnode *pDnode, char *pName, char *pDesc) {
+void dndReportStartup(SDnode *pDnode, const char *pName, const char *pDesc) {
   SStartupReq *pStartup = &pDnode->startup;
   tstrncpy(pStartup->name, pName, TSDB_STEP_NAME_LEN);
   tstrncpy(pStartup->desc, pDesc, TSDB_STEP_DESC_LEN);
@@ -104,21 +105,21 @@ void dndGetStartup(SDnode *pDnode, SStartupReq *pStartup) {
   pStartup->finished = (dndGetStatus(pDnode) == DND_STAT_RUNNING);
 }
 
-TdFilePtr dndCheckRunning(char *dataDir) {
+TdFilePtr dndCheckRunning(const char *dataDir) {
   char filepath[PATH_MAX] = {0};
   snprintf(filepath, sizeof(filepath), "%s/.running", dataDir);
 
   TdFilePtr pFile = taosOpenFile(filepath, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
   if (pFile == NULL) {
     terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to lock file:%s since %s, quit", filepath, terrstr());
+    dError("failed to lock file:%s since %s", filepath, terrstr());
     return NULL;
   }
 
   int32_t ret = taosLockFile(pFile);
   if (ret != 0) {
     terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to lock file:%s since %s, quit", filepath, terrstr());
+    dError("failed to lock file:%s since %s", filepath, terrstr());
     taosCloseFile(&pFile);
     return NULL;
   }
@@ -129,12 +130,10 @@ TdFilePtr dndCheckRunning(char *dataDir) {
 
 void dndProcessStartupReq(SDnode *pDnode, SRpcMsg *pReq) {
   dDebug("startup req is received");
-
   SStartupReq *pStartup = rpcMallocCont(sizeof(SStartupReq));
   dndGetStartup(pDnode, pStartup);
 
   dDebug("startup req is sent, step:%s desc:%s finished:%d", pStartup->name, pStartup->desc, pStartup->finished);
-
   SRpcMsg rpcRsp = {.handle = pReq->handle, .pCont = pStartup, .contLen = sizeof(SStartupReq)};
   rpcSendResponse(&rpcRsp);
 }

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

@@ -22,7 +22,12 @@ static int32_t dndGetMonitorDiskInfo(SDnode *pDnode, SMonDiskInfo *pInfo) {
   tstrncpy(pInfo->tempdir.name, tsTempDir, sizeof(pInfo->tempdir.name));
   pInfo->tempdir.size = tsTempSpace.size;
 
-  return vmMonitorTfsInfo(dndAcquireWrapper(pDnode, VNODES), pInfo);
+  SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, VNODES);
+  if (pWrapper != NULL) {
+    vmMonitorTfsInfo(pWrapper, pInfo);
+    dndReleaseWrapper(pWrapper);
+  }
+  return 0;
 }
 
 static void dndGetMonitorBasicInfo(SDnode *pDnode, SMonBasicInfo *pInfo) {
@@ -45,8 +50,17 @@ static void dndGetMonitorDnodeInfo(SDnode *pDnode, SMonDnodeInfo *pInfo) {
   taosGetCardInfo(&pInfo->net_in, &pInfo->net_out);
   taosGetProcIO(&pInfo->io_read, &pInfo->io_write, &pInfo->io_read_disk, &pInfo->io_write_disk);
 
-  vmMonitorVnodeReqs(dndAcquireWrapper(pDnode, VNODES), pInfo);
-  pInfo->has_mnode = (dndAcquireWrapper(pDnode, MNODE)->required);
+  SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, VNODES);
+  if (pWrapper != NULL) {
+    vmMonitorVnodeReqs(pWrapper, pInfo);
+    dndReleaseWrapper(pWrapper);
+  }
+
+  pWrapper = dndAcquireWrapper(pDnode, MNODE);
+  if (pWrapper != NULL) {
+    pInfo->has_mnode = pWrapper->required;
+    dndReleaseWrapper(pWrapper);
+  }
 }
 
 void dndSendMonitorReport(SDnode *pDnode) {
@@ -63,10 +77,15 @@ void dndSendMonitorReport(SDnode *pDnode) {
   SMonClusterInfo clusterInfo = {0};
   SMonVgroupInfo  vgroupInfo = {0};
   SMonGrantInfo   grantInfo = {0};
-  if (mmMonitorMnodeInfo(dndAcquireWrapper(pDnode, MNODE), &clusterInfo, &vgroupInfo, &grantInfo) == 0) {
-    monSetClusterInfo(pMonitor, &clusterInfo);
-    monSetVgroupInfo(pMonitor, &vgroupInfo);
-    monSetGrantInfo(pMonitor, &grantInfo);
+
+  SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, MNODE);
+  if (pWrapper != NULL) {
+    if (mmMonitorMnodeInfo(pWrapper, &clusterInfo, &vgroupInfo, &grantInfo) == 0) {
+      monSetClusterInfo(pMonitor, &clusterInfo);
+      monSetVgroupInfo(pMonitor, &vgroupInfo);
+      monSetGrantInfo(pMonitor, &grantInfo);
+    }
+    dndReleaseWrapper(pWrapper);
   }
 
   SMonDnodeInfo dnodeInfo = {0};

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

@@ -20,8 +20,8 @@ static void dndUpdateMnodeEpSet(SDnode *pDnode, SEpSet *pEpSet) {
   SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, DNODE);
   if (pWrapper != NULL) {
     dmUpdateMnodeEpSet(pWrapper->pMgmt, pEpSet);
+    dndReleaseWrapper(pWrapper);
   }
-  dndReleaseWrapper(pWrapper);
 }
 
 static inline NodeMsgFp dndGetMsgFp(SMgmtWrapper *pWrapper, SRpcMsg *pRpc) {

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

@@ -58,7 +58,7 @@ static void dndClearMemory(SDnode *pDnode) {
 SDnode *dndCreate(const SDnodeOpt *pOption) {
   dInfo("start to create dnode object");
   int32_t code = -1;
-  char    path[PATH_MAX];
+  char    path[PATH_MAX] = {0};
   SDnode *pDnode = NULL;
 
   pDnode = calloc(1, sizeof(SDnode));
@@ -117,6 +117,7 @@ SDnode *dndCreate(const SDnodeOpt *pOption) {
 _OVER:
   if (code != 0 && pDnode) {
     dndClearMemory(pDnode);
+    pDnode = NULL;
     dError("failed to create dnode object since %s", terrstr());
   } else {
     dInfo("dnode object is created, data:%p", pDnode);

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

@@ -20,13 +20,15 @@
 #define INTERNAL_CKEY   "_key"
 #define INTERNAL_SECRET "_pwd"
 
-static inline void dndProcessQVnodeRpcMsg(SMsgHandle *pHandle, SRpcMsg *pMsg, SEpSet *pEpSet) {
+static inline void dndProcessQMVnodeRpcMsg(SMsgHandle *pHandle, SRpcMsg *pMsg, SEpSet *pEpSet) {
   SMsgHead *pHead = pMsg->pCont;
   int32_t   vgId = htonl(pHead->vgId);
 
   SMgmtWrapper *pWrapper = pHandle->pWrapper;
-  if (vgId == pHandle->vgId && pHandle->pVgIdWrapper != NULL) {
-    pWrapper = pHandle->pVgIdWrapper;
+  if (vgId == QND_VGID) {
+    pWrapper = pHandle->pQndWrapper;
+  } else if (vgId == MND_VGID) {
+    pWrapper = pHandle->pMndWrapper;
   }
 
   dTrace("msg:%s will be processed by %s, handle:%p app:%p vgId:%d", TMSG_INFO(pMsg->msgType), pWrapper->name,
@@ -46,13 +48,13 @@ static void dndProcessResponse(void *parent, SRpcMsg *pRsp, SEpSet *pEpSet) {
   }
 
   SMsgHandle *pHandle = &pMgmt->msgHandles[TMSG_INDEX(msgType)];
-  if (pHandle->msgFp != NULL) {
-    if (pHandle->vgId == 0) {
+  if (pHandle->pWrapper != NULL) {
+    if (pHandle->pMndWrapper == NULL && pHandle->pQndWrapper == NULL) {
       dTrace("rsp:%s will be processed by %s, handle:%p app:%p code:0x%04x:%s", TMSG_INFO(msgType),
              pHandle->pWrapper->name, pRsp->handle, pRsp->ahandle, pRsp->code & 0XFFFF, tstrerror(pRsp->code));
       dndProcessRpcMsg(pHandle->pWrapper, pRsp, pEpSet);
     } else {
-      dndProcessQVnodeRpcMsg(pHandle, pRsp, pEpSet);
+      dndProcessQMVnodeRpcMsg(pHandle, pRsp, pEpSet);
     }
   } else {
     dError("rsp:%s not processed since no handle, handle:%p app:%p", TMSG_INFO(msgType), pRsp->handle, pRsp->ahandle);
@@ -126,13 +128,13 @@ static void dndProcessRequest(void *param, SRpcMsg *pReq, SEpSet *pEpSet) {
   }
 
   SMsgHandle *pHandle = &pMgmt->msgHandles[TMSG_INDEX(msgType)];
-  if (pHandle->msgFp != NULL) {
-    if (pHandle->vgId == 0) {
+  if (pHandle->pWrapper != NULL) {
+    if (pHandle->pMndWrapper == NULL && pHandle->pQndWrapper == NULL) {
       dTrace("req:%s will be processed by %s, handle:%p app:%p", TMSG_INFO(msgType), pHandle->pWrapper->name,
              pReq->handle, pReq->ahandle);
       dndProcessRpcMsg(pHandle->pWrapper, pReq, pEpSet);
     } else {
-      dndProcessQVnodeRpcMsg(pHandle, pReq, pEpSet);
+      dndProcessQMVnodeRpcMsg(pHandle, pReq, pEpSet);
     }
   } else {
     dError("req:%s not processed since no handle, handle:%p app:%p", TMSG_INFO(msgType), pReq->handle, pReq->ahandle);
@@ -144,9 +146,14 @@ static void dndProcessRequest(void *param, SRpcMsg *pReq, SEpSet *pEpSet) {
 
 static void dndSendMsgToMnodeRecv(SDnode *pDnode, SRpcMsg *pRpcMsg, SRpcMsg *pRpcRsp) {
   STransMgmt *pMgmt = &pDnode->trans;
+  SEpSet      epSet = {0};
+
+  SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, DNODE);
+  if (pWrapper != NULL) {
+    dmGetMnodeEpSet(pWrapper->pMgmt, &epSet);
+    dndReleaseWrapper(pWrapper);
+  }
 
-  SEpSet epSet = {0};
-  dmGetMnodeEpSet(dndAcquireWrapper(pDnode, DNODE)->pMgmt, &epSet);
   rpcSendRecv(pMgmt->clientRpc, &epSet, pRpcMsg, pRpcRsp);
 }
 
@@ -180,9 +187,14 @@ static int32_t dndRetrieveUserAuthInfo(void *parent, char *user, char *spi, char
     return 0;
   }
 
-  if (mmGetUserAuth(dndAcquireWrapper(pDnode, MNODE), user, spi, encrypt, secret, ckey) == 0) {
-    dTrace("user:%s, get auth from mnode, spi:%d encrypt:%d", user, *spi, *encrypt);
-    return 0;
+  SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, MNODE);
+  if (pWrapper != NULL) {
+    if (mmGetUserAuth(pWrapper, user, spi, encrypt, secret, ckey) == 0) {
+      dndReleaseWrapper(pWrapper);
+      dTrace("user:%s, get auth from mnode, spi:%d encrypt:%d", user, *spi, *encrypt);
+      return 0;
+    }
+    dndReleaseWrapper(pWrapper);
   }
 
   if (terrno != TSDB_CODE_APP_NOT_READY) {
@@ -269,21 +281,27 @@ int32_t dndInitMsgHandle(SDnode *pDnode) {
       int32_t   vgId = pWrapper->msgVgIds[msgIndex];
       if (msgFp == NULL) continue;
 
+      // dTrace("msg:%s will be processed by %s, vgId:%d", tMsgInfo[msgIndex], pWrapper->name, vgId);
+
       SMsgHandle *pHandle = &pMgmt->msgHandles[msgIndex];
-      if (pHandle->msgFp != NULL && pHandle->vgId == vgId) {
-        dError("msg:%s has multiple process nodes, prev node:%s:%d, curr node:%s:%d", tMsgInfo[msgIndex],
-               pHandle->pWrapper->name, pHandle->pWrapper->msgVgIds[msgIndex], pWrapper->name, vgId);
-        return -1;
+      if (vgId == QND_VGID) {
+        if (pHandle->pQndWrapper != NULL) {
+          dError("msg:%s has multiple process nodes", tMsgInfo[msgIndex]);
+          return -1;
+        }
+        pHandle->pQndWrapper = pWrapper;
+      } else if (vgId == MND_VGID) {
+        if (pHandle->pMndWrapper != NULL) {
+          dError("msg:%s has multiple process nodes", tMsgInfo[msgIndex]);
+          return -1;
+        }
+        pHandle->pMndWrapper = pWrapper;
       } else {
-        dTrace("msg:%s will be processed by %s, vgId:%d", tMsgInfo[msgIndex], pWrapper->name, vgId);
-        if (vgId == 0) {
-          pHandle->msgFp = msgFp;
-          pHandle->pWrapper = pWrapper;
-        } else {
-          pHandle->vgId = vgId;
-          pHandle->vgIdMsgFp = msgFp;
-          pHandle->pVgIdWrapper = pWrapper;
+        if (pHandle->pWrapper != NULL) {
+          dError("msg:%s has multiple process nodes", tMsgInfo[msgIndex]);
+          return -1;
         }
+        pHandle->pWrapper = pWrapper;
       }
     }
   }
@@ -320,7 +338,12 @@ int32_t dndSendReqToMnode(SMgmtWrapper *pWrapper, SRpcMsg *pReq) {
     SDnode     *pDnode = pWrapper->pDnode;
     STransMgmt *pTrans = &pDnode->trans;
     SEpSet      epSet = {0};
-    dmGetMnodeEpSet(dndAcquireWrapper(pDnode, DNODE)->pMgmt, &epSet);
+
+    SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, DNODE);
+    if (pWrapper != NULL) {
+      dmGetMnodeEpSet(pWrapper->pMgmt, &epSet);
+      dndReleaseWrapper(pWrapper);
+    }
     return dndSendRpcReq(pTrans, &epSet, pReq);
   }
 }
@@ -328,7 +351,12 @@ int32_t dndSendReqToMnode(SMgmtWrapper *pWrapper, SRpcMsg *pReq) {
 void dndSendRpcRsp(SMgmtWrapper *pWrapper, SRpcMsg *pRsp) {
   if (pRsp->code == TSDB_CODE_APP_NOT_READY) {
     SMgmtWrapper *pDnodeWrapper = dndAcquireWrapper(pWrapper->pDnode, DNODE);
-    dmSendRedirectRsp(pDnodeWrapper->pMgmt, pRsp);
+    if (pDnodeWrapper != NULL) {
+      dmSendRedirectRsp(pDnodeWrapper->pMgmt, pRsp);
+      dndReleaseWrapper(pDnodeWrapper);
+    } else {
+      rpcSendResponse(pRsp);
+    }
   } else {
     rpcSendResponse(pRsp);
   }

source/dnode/mgmt/dnode/src/dmFile.c

@@ -209,7 +209,7 @@ int32_t dmWriteFile(SDnodeMgmt *pMgmt) {
   }
 
   pMgmt->updateTime = taosGetTimestampMs();
-  dDebug("successed to write %s", file);
+  dDebug("successed to write %s", realfile);
   return 0;
 }
 

source/dnode/mgmt/dnode/src/dmInt.c

@@ -74,14 +74,14 @@ void dmSendRedirectRsp(SDnodeMgmt *pMgmt, SRpcMsg *pReq) {
 }
 
 static int32_t dmStart(SMgmtWrapper *pWrapper) {
-  dDebug("dnode mgmt start to run");
+  dDebug("dnode-mgmt start to run");
   return dmStartThread(pWrapper->pMgmt);
 }
 
 int32_t dmInit(SMgmtWrapper *pWrapper) {
   SDnode     *pDnode = pWrapper->pDnode;
   SDnodeMgmt *pMgmt = calloc(1, sizeof(SDnodeMgmt));
-  dInfo("dnode-mgmt is initialized");
+  dInfo("dnode-mgmt start to init");
 
   pDnode->dnodeId = 0;
   pDnode->dropped = 0;

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

@@ -41,8 +41,12 @@ void dmSendStatusReq(SDnodeMgmt *pMgmt) {
   memcpy(req.clusterCfg.charset, tsCharset, TD_LOCALE_LEN);
   taosRUnLockLatch(&pMgmt->latch);
 
-  req.pVloads = taosArrayInit(TSDB_MAX_VNODES, sizeof(SVnodeLoad));
-  vmMonitorVnodeLoads(dndAcquireWrapper(pDnode, VNODES), req.pVloads);
+  SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, VNODES);
+  if (pWrapper != NULL) {
+    req.pVloads = taosArrayInit(TSDB_MAX_VNODES, sizeof(SVnodeLoad));
+    vmMonitorVnodeLoads(pWrapper, req.pVloads);
+    dndReleaseWrapper(pWrapper);
+  }
 
   int32_t contLen = tSerializeSStatusReq(NULL, 0, &req);
   void   *pHead = rpcMallocCont(contLen);
@@ -114,19 +118,19 @@ int32_t dmProcessConfigReq(SDnodeMgmt *pMgmt, SNodeMsg *pMsg) {
 
 void dmInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by DNODE
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_NETWORK_TEST, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_NETWORK_TEST, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
 
   // Requests handled by MNODE
-  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS_RSP, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT_RSP, (NodeMsgFp)dmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH_RSP, (NodeMsgFp)dmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS_RSP, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT_RSP, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH_RSP, (NodeMsgFp)dmProcessMgmtMsg, VND_VGID);
 }

source/dnode/mgmt/dnode/src/dmWorker.c

@@ -114,6 +114,7 @@ int32_t dmStartWorker(SDnodeMgmt *pMgmt) {
     return -1;
   }
 
+  dDebug("dnode workers are initialized");
   return 0;
 }
 
@@ -136,6 +137,7 @@ void dmStopWorker(SDnodeMgmt *pMgmt) {
     taosDestoryThread(pMgmt->threadId);
     pMgmt->threadId = NULL;
   }
+  dDebug("dnode workers are closed");
 }
 
 int32_t dmProcessMgmtMsg(SDnodeMgmt *pMgmt, SNodeMsg *pMsg) {
@@ -144,6 +146,6 @@ int32_t dmProcessMgmtMsg(SDnodeMgmt *pMgmt, SNodeMsg *pMsg) {
     pWorker = &pMgmt->statusWorker;
   }
 
-  dTrace("msg:%p, will be written to worker %s", pMsg, pWorker->name);
+  dTrace("msg:%p, put into worker %s", pMsg, pWorker->name);
   return taosWriteQitem(pWorker->queue, pMsg);
 }

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

@@ -15,6 +15,7 @@
 
 #define _DEFAULT_SOURCE
 #include "mmInt.h"
+#include "wal.h"
 
 static bool mmDeployRequired(SDnode *pDnode) {
   if (pDnode->dnodeId > 0) return false;
@@ -226,7 +227,7 @@ static int32_t mmOpen(SMgmtWrapper *pWrapper) {
 }
 
 static int32_t mmStart(SMgmtWrapper *pWrapper) {
-  dDebug("mnode mgmt start to run");
+  dDebug("mnode-mgmt start to run");
   SMnodeMgmt *pMgmt = pWrapper->pMgmt;
   return mndStart(pMgmt->pMnode);
 }

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

@@ -75,84 +75,90 @@ int32_t mmProcessAlterReq(SMnodeMgmt *pMgmt, SNodeMsg *pMsg) {
 
 void mmInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by DNODE
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_MNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_QNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_SNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_BNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_CONFIG_DNODE_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
 
   // Requests handled by MNODE
-  dndSetMsgHandle(pWrapper, TDMT_MND_CONNECT, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_ACCT, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_ACCT, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_ACCT, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_USER, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_USER, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_USER, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GET_USER_AUTH, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CONFIG_DNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_MNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_MNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_QNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_QNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_SNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_SNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_BNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_BNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_USE_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SYNC_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_COMPACT_DB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_FUNC, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_RETRIEVE_FUNC, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_FUNC, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_STB, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_SMA, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_SMA, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_TABLE_META, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_VGROUP_LIST, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_QUERY, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_CONN, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_HEARTBEAT, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW_RETRIEVE, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SYSTABLE_RETRIEVE, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_TRANS, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_TOPIC, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_TOPIC, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_TOPIC, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_SUBSCRIBE, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_MQ_COMMIT_OFFSET, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_GET_SUB_EP, (NodeMsgFp)mmProcessReadMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STREAM, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CONNECT, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_ACCT, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_ACCT, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_ACCT, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_USER, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_USER, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_USER, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GET_USER_AUTH, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CONFIG_DNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_MNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_MNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_QNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_QNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_SNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_SNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_BNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_BNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_DB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_DB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_USE_DB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_DB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SYNC_DB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_COMPACT_DB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_FUNC, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_RETRIEVE_FUNC, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_FUNC, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_STB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_STB, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_SMA, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_SMA, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_TABLE_META, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_VGROUP_LIST, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_QUERY, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_CONN, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_HEARTBEAT, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SHOW_RETRIEVE, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SYSTABLE_RETRIEVE, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_STATUS, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_KILL_TRANS, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GRANT, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_AUTH, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_MNODE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_TOPIC, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_ALTER_TOPIC, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_DROP_TOPIC, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_SUBSCRIBE, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_MQ_COMMIT_OFFSET, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_GET_SUB_EP, (NodeMsgFp)mmProcessReadMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_MND_CREATE_STREAM, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
 
   // Requests handled by VNODE
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_SMA_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_SMA_RSP, (NodeMsgFp)mmProcessWriteMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_SMA_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_SMA_RSP, (NodeMsgFp)mmProcessWriteMsg, VND_VGID);
+
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)mmProcessReadMsg, MND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)mmProcessReadMsg, MND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)mmProcessReadMsg, MND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)mmProcessReadMsg, MND_VGID);
+
 }

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

@@ -108,6 +108,7 @@ int32_t mmStartWorker(SMnodeMgmt *pMgmt) {
     return -1;
   }
 
+  dDebug("mnode workers are initialized");
   return 0;
 }
 
@@ -115,4 +116,5 @@ void mmStopWorker(SMnodeMgmt *pMgmt) {
   tSingleWorkerCleanup(&pMgmt->readWorker);
   tSingleWorkerCleanup(&pMgmt->writeWorker);
   tSingleWorkerCleanup(&pMgmt->syncWorker);
+  dDebug("mnode workers are closed");
 }

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

@@ -56,14 +56,14 @@ int32_t qmProcessDropReq(SMgmtWrapper *pWrapper, SNodeMsg *pMsg) {
 
 void qmInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by VNODE
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)qmProcessQueryMsg, 1);
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)qmProcessQueryMsg, 1);
-  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)qmProcessFetchMsg, 1);
-  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)qmProcessFetchMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)qmProcessQueryMsg, QND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)qmProcessQueryMsg, QND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)qmProcessFetchMsg, QND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)qmProcessFetchMsg, QND_VGID);
 
-  dndSetMsgHandle(pWrapper, TDMT_VND_RES_READY, (NodeMsgFp)qmProcessFetchMsg, 1);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASKS_STATUS, (NodeMsgFp)qmProcessFetchMsg, 1);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_TASK, (NodeMsgFp)qmProcessFetchMsg, 1);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TASK, (NodeMsgFp)qmProcessFetchMsg, 1);
-  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)qmProcessFetchMsg, 1);
+  dndSetMsgHandle(pWrapper, TDMT_VND_RES_READY, (NodeMsgFp)qmProcessFetchMsg, QND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASKS_STATUS, (NodeMsgFp)qmProcessFetchMsg, QND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_TASK, (NodeMsgFp)qmProcessFetchMsg, QND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TASK, (NodeMsgFp)qmProcessFetchMsg, QND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)qmProcessFetchMsg, QND_VGID);
 }

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

@@ -132,10 +132,12 @@ int32_t qmStartWorker(SQnodeMgmt *pMgmt) {
     return -1;
   }
 
+  dDebug("qnode workers are initialized");
   return 0;
 }
 
 void qmStopWorker(SQnodeMgmt *pMgmt) {
   tSingleWorkerCleanup(&pMgmt->queryWorker);
   tSingleWorkerCleanup(&pMgmt->fetchWorker);
+  dDebug("qnode workers are closed");
 }

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

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

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

@@ -80,6 +80,7 @@ int32_t smStartWorker(SSnodeMgmt *pMgmt) {
     return -1;
   }
 
+  dDebug("snode workers are initialized");
   return 0;
 }
 
@@ -90,6 +91,7 @@ void smStopWorker(SSnodeMgmt *pMgmt) {
   }
   taosArrayDestroy(pMgmt->uniqueWorkers);
   tSingleWorkerCleanup(&pMgmt->sharedWorker);
+  dDebug("snode workers are closed");
 }
 
 static FORCE_INLINE int32_t smGetSWIdFromMsg(SRpcMsg *pMsg) {

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

@@ -257,14 +257,14 @@ static void vmCleanup(SMgmtWrapper *pWrapper) {
   SVnodesMgmt *pMgmt = pWrapper->pMgmt;
   if (pMgmt == NULL) return;
 
-  dInfo("vnodes-mgmt start to cleanup");
+  dInfo("vnode-mgmt start to cleanup");
   vmCloseVnodes(pMgmt);
   vmStopWorker(pMgmt);
   vnodeCleanup();
   // walCleanUp();
   free(pMgmt);
   pWrapper->pMgmt = NULL;
-  dInfo("vnodes-mgmt is cleaned up");
+  dInfo("vnode-mgmt is cleaned up");
 }
 
 static int32_t vmInit(SMgmtWrapper *pWrapper) {
@@ -272,7 +272,7 @@ static int32_t vmInit(SMgmtWrapper *pWrapper) {
   SVnodesMgmt *pMgmt = calloc(1, sizeof(SVnodesMgmt));
   int32_t      code = -1;
 
-  dInfo("vnodes-mgmt start to init");
+  dInfo("vnode-mgmt start to init");
   if (pMgmt == NULL) goto _OVER;
 
   pMgmt->path = pWrapper->path;
@@ -312,7 +312,7 @@ static int32_t vmInit(SMgmtWrapper *pWrapper) {
   }
 
   if (vmOpenVnodes(pMgmt) != 0) {
-    dError("failed to open vnodes since %s", terrstr());
+    dError("failed to open vnode since %s", terrstr());
     return -1;
   }
 

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

@@ -244,49 +244,49 @@ int32_t vmProcessCompactVnodeReq(SVnodesMgmt *pMgmt, SNodeMsg *pMsg) {
 
 void vmInitMsgHandles(SMgmtWrapper *pWrapper) {
   // Requests handled by VNODE
-  dndSetMsgHandle(pWrapper, TDMT_VND_SUBMIT, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)vmProcessQueryMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)vmProcessQueryMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_UPDATE_TAG_VAL, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TABLE_META, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TABLES_META, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONSUME, (NodeMsgFp)vmProcessQueryMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_QUERY, (NodeMsgFp)vmProcessQueryMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONNECT, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_DISCONNECT, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_RES_READY, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASKS_STATUS, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_TASK, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TASK, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TABLE, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_SMA, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_SMA, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_SMA, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES_FETCH, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_CONSUME, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY, (NodeMsgFp)vmProcessWriteMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_HEARTBEAT, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_EXEC, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_PIPE_EXEC, (NodeMsgFp)vmProcessFetchMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_MERGE_EXEC, (NodeMsgFp)vmProcessMergeMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_VND_STREAM_TRIGGER, (NodeMsgFp)vmProcessFetchMsg, 0);
-
-  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
-  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE, (NodeMsgFp)vmProcessMgmtMsg, 0);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SUBMIT, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY, (NodeMsgFp)vmProcessQueryMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_CONTINUE, (NodeMsgFp)vmProcessQueryMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_FETCH_RSP, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_UPDATE_TAG_VAL, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TABLE_META, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TABLES_META, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONSUME, (NodeMsgFp)vmProcessQueryMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_QUERY, (NodeMsgFp)vmProcessQueryMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_CONNECT, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_DISCONNECT, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_RES_READY, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASKS_STATUS, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_TASK, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TASK, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_STB, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_STB, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_STB, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_TABLE, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_ALTER_TABLE, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_TABLE, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CREATE_SMA, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CANCEL_SMA, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_DROP_SMA, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_SHOW_TABLES_FETCH, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CONN, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_REB, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_MQ_SET_CUR, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_CONSUME, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_DEPLOY, (NodeMsgFp)vmProcessWriteMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_QUERY_HEARTBEAT, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_EXEC, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_PIPE_EXEC, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_TASK_MERGE_EXEC, (NodeMsgFp)vmProcessMergeMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_VND_STREAM_TRIGGER, (NodeMsgFp)vmProcessFetchMsg, VND_VGID);
+
+  dndSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE, (NodeMsgFp)vmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_ALTER_VNODE, (NodeMsgFp)vmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE, (NodeMsgFp)vmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE, (NodeMsgFp)vmProcessMgmtMsg, VND_VGID);
+  dndSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE, (NodeMsgFp)vmProcessMgmtMsg, VND_VGID);
 }

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

@@ -372,7 +372,7 @@ int32_t vmStartWorker(SVnodesMgmt *pMgmt) {
     return -1;
   }
 
-  dDebug("vnode workers is initialized");
+  dDebug("vnode workers are initialized");
   return 0;
 }
 
@@ -382,5 +382,5 @@ void vmStopWorker(SVnodesMgmt *pMgmt) {
   tQWorkerCleanup(&pMgmt->queryPool);
   tWWorkerCleanup(&pMgmt->writePool);
   tWWorkerCleanup(&pMgmt->syncPool);
-  dDebug("vnode workers is closed");
+  dDebug("vnode workers are closed");
 }

source/dnode/mnode/impl/CMakeLists.txt

@@ -6,7 +6,7 @@ target_include_directories(
     PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/inc"
 )
 target_link_libraries(
-    mnode scheduler sdb wal transport cjson sync monitor
+    mnode scheduler sdb wal transport cjson sync monitor parser
 )
 
 if(${BUILD_TEST})

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

@@ -69,7 +69,7 @@ void mndCleanupSma(SMnode *pMnode) {}
 static SSdbRaw *mndSmaActionEncode(SSmaObj *pSma) {
   terrno = TSDB_CODE_OUT_OF_MEMORY;
 
-  int32_t  size = sizeof(SSmaObj) + pSma->exprLen + pSma->tagsFilterLen + TSDB_SMA_RESERVE_SIZE;
+  int32_t  size = sizeof(SSmaObj) + pSma->exprLen + pSma->tagsFilterLen + pSma->sqlLen + pSma->astLen + TSDB_SMA_RESERVE_SIZE;
   SSdbRaw *pRaw = sdbAllocRaw(SDB_SMA, TSDB_SMA_VER_NUMBER, size);
   if (pRaw == NULL) goto _OVER;
 

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

@@ -14,6 +14,7 @@
  */
 
 #include "mndStream.h"
+#include "parser.h"
 #include "mndAuth.h"
 #include "mndDb.h"
 #include "mndDnode.h"
@@ -218,28 +219,6 @@ static int32_t mndCheckCreateStreamReq(SCMCreateStreamReq *pCreate) {
   return 0;
 }
 
-static SArray *mndExtractNamesFromAst(const SNode *pAst) {
-  if (pAst->type != QUERY_NODE_SELECT_STMT) return NULL;
-
-  SArray *names = taosArrayInit(0, sizeof(void *));
-  if (names == NULL) {
-    return NULL;
-  }
-  SSelectStmt *pSelect = (SSelectStmt *)pAst;
-  SNodeList   *pNodes = pSelect->pProjectionList;
-  SListCell   *pCell = pNodes->pHead;
-  while (pCell != NULL) {
-    if (pCell->pNode->type != QUERY_NODE_FUNCTION) {
-      continue;
-    }
-    SFunctionNode *pFunction = (SFunctionNode *)pCell->pNode;
-    char          *name = strdup(pFunction->node.aliasName);
-    taosArrayPush(names, &name);
-    pCell = pCell->pNext;
-  }
-  return names;
-}
-
 static int32_t mndStreamGetPlanString(const char *ast, char **pStr) {
   if (NULL == ast) {
     return TSDB_CODE_SUCCESS;
@@ -269,18 +248,23 @@ static int32_t mndStreamGetPlanString(const char *ast, char **pStr) {
 
 int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast, STrans *pTrans) {
   SNode *pAst = NULL;
+#if 1 // TODO: remove debug info later
+  printf("ast = %s\n", ast); 
+#endif
   if (nodesStringToNode(ast, &pAst) < 0) {
     return -1;
   }
 #if 1
-  SArray *names = mndExtractNamesFromAst(pAst);
+  SSchemaWrapper sw = {0};
+  qExtractResultSchema(pAst, (int32_t*)&sw.nCols, &sw.pSchema);
+
   printf("|");
-  for (int i = 0; i < taosArrayGetSize(names); i++) {
-    printf(" %15s |", (char *)taosArrayGetP(names, i));
+  for (int i = 0; i < sw.nCols; i++) {
+    printf(" %15s |", (char *)sw.pSchema[i].name);
   }
   printf("\n=======================================================\n");
 
-  pStream->ColAlias = names;
+  pStream->ColAlias = NULL;
 #endif
 
   if (TSDB_CODE_SUCCESS != mndStreamGetPlanString(ast, &pStream->physicalPlan)) {

source/dnode/vnode/inc/tsdb.h

@@ -87,6 +87,15 @@ int    tsdbInsertData(STsdb *pTsdb, SSubmitReq *pMsg, SSubmitRsp *pRsp);
 int    tsdbPrepareCommit(STsdb *pTsdb);
 int    tsdbCommit(STsdb *pTsdb);
 
+/**
+ * @brief When submit msg received, update the relative expired window synchronously.
+ * 
+ * @param pTsdb 
+ * @param msg 
+ * @return int32_t 
+ */
+int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, const char *msg);
+
 /**
  * @brief Insert tSma(Time-range-wise SMA) data from stream computing engine
  *
@@ -95,11 +104,18 @@ int    tsdbCommit(STsdb *pTsdb);
  * @return int32_t
  */
 int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg);
-int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, int8_t smaType, char *msg);
+
+/**
+ * @brief Drop tSma data and local cache.
+ * 
+ * @param pTsdb 
+ * @param indexUid 
+ * @return int32_t 
+ */
 int32_t tsdbDropTSmaData(STsdb *pTsdb, int64_t indexUid);
 
 /**
- * @brief Insert RSma(Time-range-wise Rollup SMA) data.
+ * @brief Insert RSma(Rollup SMA) data.
  *
  * @param pTsdb
  * @param msg
@@ -108,6 +124,20 @@ int32_t tsdbDropTSmaData(STsdb *pTsdb, int64_t indexUid);
 int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg);
 
 // TODO: This is the basic params, and should wrap the params to a queryHandle.
+/**
+ * @brief Get tSma(Time-range-wise SMA) data.
+ * 
+ * @param pTsdb 
+ * @param pData 
+ * @param indexUid 
+ * @param interval 
+ * @param intervalUnit 
+ * @param tableUid 
+ * @param colId 
+ * @param querySKey 
+ * @param nMaxResult 
+ * @return int32_t 
+ */
 int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
                         tb_uid_t tableUid, col_id_t colId, TSKEY querySKey, int32_t nMaxResult);
 

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

@@ -63,6 +63,7 @@ struct STsdb {
 #define REPO_ID(r)         ((r)->vgId)
 #define REPO_CFG(r)        (&(r)->config)
 #define REPO_FS(r)         (r)->fs
+#define REPO_META(r)       (r)->pMeta
 #define REPO_TFS(r)        (r)->pTfs
 #define IS_REPO_LOCKED(r)  (r)->repoLocked
 #define REPO_SMA_ENV(r, t) ((TSDB_SMA_TYPE_ROLLUP == (t)) ? (r)->pRSmaEnv : (r)->pTSmaEnv)

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

@@ -16,6 +16,8 @@
 #ifndef _TD_TSDB_SMA_H_
 #define _TD_TSDB_SMA_H_
 
+#define TSDB_SMA_TEST // remove after test finished
+
 typedef struct SSmaStat     SSmaStat;
 typedef struct SSmaEnv      SSmaEnv;
 

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

@@ -44,7 +44,7 @@ int32_t tqReadHandleSetMsg(STqReadHandle* pReadHandle, SSubmitReq* pMsg, int64_t
     if (pReadHandle->pBlock == NULL) break;
 
     pReadHandle->pBlock->uid = htobe64(pReadHandle->pBlock->uid);
-    pReadHandle->pBlock->tid = htonl(pReadHandle->pBlock->tid);
+    pReadHandle->pBlock->suid = htobe64(pReadHandle->pBlock->suid);
     pReadHandle->pBlock->sversion = htonl(pReadHandle->pBlock->sversion);
     pReadHandle->pBlock->dataLen = htonl(pReadHandle->pBlock->dataLen);
     pReadHandle->pBlock->schemaLen = htonl(pReadHandle->pBlock->schemaLen);

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

@@ -248,7 +248,7 @@ static int tsdbScanAndConvertSubmitMsg(STsdb *pTsdb, SSubmitReq *pMsg) {
     if (pBlock == NULL) break;
 
     pBlock->uid = htobe64(pBlock->uid);
-    pBlock->tid = htonl(pBlock->tid);
+    pBlock->suid = htobe64(pBlock->suid);
     pBlock->sversion = htonl(pBlock->sversion);
     pBlock->dataLen = htonl(pBlock->dataLen);
     pBlock->schemaLen = htonl(pBlock->schemaLen);

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

@@ -38,7 +38,7 @@ typedef enum {
 } ESmaStorageLevel;
 
 typedef struct {
-  STsdb * pTsdb;
+  STsdb  *pTsdb;
   SDBFile dFile;
   int32_t interval;  // interval with the precision of DB
 } STSmaWriteH;
@@ -49,7 +49,7 @@ typedef struct {
 } SmaFsIter;
 
 typedef struct {
-  STsdb *   pTsdb;
+  STsdb    *pTsdb;
   SDBFile   dFile;
   int32_t   interval;   // interval with the precision of DB
   int32_t   blockSize;  // size of SMA block item
@@ -69,7 +69,7 @@ typedef struct {
    */
   int8_t    state;           // ETsdbSmaStat
   SHashObj *expiredWindows;  // key: skey of time window, value: N/A
-  STSma *   pSma;            // cache schema
+  STSma    *pSma;            // cache schema
 } SSmaStatItem;
 
 struct SSmaStat {
@@ -80,9 +80,9 @@ struct SSmaStat {
 // declaration of static functions
 
 // expired window
-static int32_t  tsdbUpdateExpiredWindow(STsdb *pTsdb, ETsdbSmaType smaType, char *msg);
+static int32_t  tsdbUpdateExpiredWindowImpl(STsdb *pTsdb, const char *msg);
 static int32_t  tsdbInitSmaStat(SSmaStat **pSmaStat);
-static void *   tsdbFreeSmaStatItem(SSmaStatItem *pSmaStatItem);
+static void    *tsdbFreeSmaStatItem(SSmaStatItem *pSmaStatItem);
 static int32_t  tsdbDestroySmaState(SSmaStat *pSmaStat);
 static SSmaEnv *tsdbNewSmaEnv(const STsdb *pTsdb, const char *path, SDiskID did);
 static int32_t  tsdbInitSmaEnv(STsdb *pTsdb, const char *path, SDiskID did, SSmaEnv **pEnv);
@@ -124,7 +124,7 @@ static FORCE_INLINE int8_t tsdbSmaStat(SSmaStatItem *pStatItem) {
 }
 
 static FORCE_INLINE bool tsdbSmaStatIsOK(SSmaStatItem *pStatItem, int8_t *state) {
-  if(!pStatItem) {
+  if (!pStatItem) {
     return false;
   }
 
@@ -384,49 +384,20 @@ static int32_t tsdbCheckAndInitSmaEnv(STsdb *pTsdb, int8_t smaType) {
   return TSDB_CODE_SUCCESS;
 };
 
-/**
- * @brief Update expired window according to msg from stream computing module.
- *
- * @param pTsdb
- * @param smaType ETsdbSmaType
- * @param msg
- * @return int32_t
- */
-int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, ETsdbSmaType smaType, char *msg) {
-  if (!msg || !pTsdb->pMeta) {
-    terrno = TSDB_CODE_INVALID_PTR;
-    return TSDB_CODE_FAILED;
-  }
-
-  // TODO: decode the msg from Stream Computing module => start
-  int64_t       indexUid = SMA_TEST_INDEX_UID;
-  const int32_t SMA_TEST_EXPIRED_WINDOW_SIZE = 10;
-  TSKEY         expiredWindows[SMA_TEST_EXPIRED_WINDOW_SIZE];
-  TSKEY         skey1 = 1646987196 * 1e3;
-  for (int32_t i = 0; i < SMA_TEST_EXPIRED_WINDOW_SIZE; ++i) {
-    expiredWindows[i] = skey1 + i;
-  }
-  // TODO: decode the msg <= end
-
-  if (tsdbCheckAndInitSmaEnv(pTsdb, smaType) != TSDB_CODE_SUCCESS) {
-    terrno = TSDB_CODE_TDB_INIT_FAILED;
-    return TSDB_CODE_FAILED;
-  }
-
-  SSmaEnv * pEnv = REPO_SMA_ENV(pTsdb, smaType);
-  SSmaStat *pStat = SMA_ENV_STAT(pEnv);
-  SHashObj *pItemsHash = SMA_ENV_STAT_ITEMS(pEnv);
-
-  TASSERT(pEnv != NULL && pStat != NULL && pItemsHash != NULL);
+static STimeWindow getActiveTimeWindowX(int64_t ts, SInterval* pInterval) {
+  STimeWindow tw = {0};
+  tw.skey = 100;
+  tw.ekey = 1000;
+  return tw;
+}
 
-  tsdbRefSmaStat(pTsdb, pStat);
+static int32_t tsdbSetExpiredWindow(STsdb *pTsdb, SHashObj *pItemsHash, int64_t indexUid, int64_t winSKey) {
   SSmaStatItem *pItem = taosHashGet(pItemsHash, &indexUid, sizeof(indexUid));
   if (pItem == NULL) {
     pItem = tsdbNewSmaStatItem(TSDB_SMA_STAT_EXPIRED);  // TODO use the real state
     if (pItem == NULL) {
       // Response to stream computing: OOM
       // For query, if the indexUid not found, the TSDB should tell query module to query raw TS data.
-      tsdbUnRefSmaStat(pTsdb, pStat);
       return TSDB_CODE_FAILED;
     }
 
@@ -436,7 +407,6 @@ int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, ETsdbSmaType smaType, char *msg) {
       terrno = TSDB_CODE_TDB_NO_SMA_INDEX_IN_META;
       taosHashCleanup(pItem->expiredWindows);
       free(pItem);
-      tsdbUnRefSmaStat(pTsdb, pStat);
       tsdbWarn("vgId:%d update expired window failed for smaIndex %" PRIi64 " since %s", REPO_ID(pTsdb), indexUid,
                tstrerror(terrno));
       return TSDB_CODE_FAILED;
@@ -447,34 +417,150 @@ int32_t tsdbUpdateExpiredWindow(STsdb *pTsdb, ETsdbSmaType smaType, char *msg) {
       // If error occurs during put smaStatItem, free the resources of pItem
       taosHashCleanup(pItem->expiredWindows);
       free(pItem);
-      tsdbUnRefSmaStat(pTsdb, pStat);
       return TSDB_CODE_FAILED;
     }
   }
 
   int8_t state = TSDB_SMA_STAT_EXPIRED;
+  if (taosHashPut(pItem->expiredWindows, &winSKey, sizeof(TSKEY), &state, sizeof(state)) != 0) {
+    // If error occurs during taosHashPut expired windows, remove the smaIndex from pTsdb->pSmaStat, thus TSDB would
+    // tell query module to query raw TS data.
+    // N.B.
+    //  1) It is assumed to be extemely little probability event of fail to taosHashPut.
+    //  2) This would solve the inconsistency to some extent, but not completely, unless we record all expired
+    // windows failed to put into hash table.
+    taosHashCleanup(pItem->expiredWindows);
+    tfree(pItem->pSma);
+    taosHashRemove(pItemsHash, &indexUid, sizeof(indexUid));
+    return TSDB_CODE_FAILED;
+  }
+  tsdbDebug("vgId:%d smaIndex %" PRIi64 " tsKey %" PRIi64 " is put to hash", REPO_ID(pTsdb), indexUid, winSKey);
+}
+
+/**
+ * @brief Update expired window according to msg from stream computing module.
+ *
+ * @param pTsdb
+ * @param msg SSubmitReq
+ * @return int32_t
+ */
+int32_t tsdbUpdateExpiredWindowImpl(STsdb *pTsdb, const char *msg) {
+  const SSubmitReq *pMsg = (const SSubmitReq *)msg;
+
+  if (pMsg->length <= sizeof(SSubmitReq)) {
+    terrno = TSDB_CODE_TDB_SUBMIT_MSG_MSSED_UP;
+    return TSDB_CODE_FAILED;
+  }
+  if (!pTsdb->pMeta) {
+    terrno = TSDB_CODE_INVALID_PTR;
+    return TSDB_CODE_FAILED;
+  }
+
+// TODO: decode the msg from Stream Computing module => start
+#ifdef TSDB_SMA_TESTx
+  int64_t       indexUid = SMA_TEST_INDEX_UID;
+  const int32_t SMA_TEST_EXPIRED_WINDOW_SIZE = 10;
+  TSKEY         expiredWindows[SMA_TEST_EXPIRED_WINDOW_SIZE];
+  TSKEY         skey1 = 1646987196 * 1e3;
   for (int32_t i = 0; i < SMA_TEST_EXPIRED_WINDOW_SIZE; ++i) {
-    if (taosHashPut(pItem->expiredWindows, expiredWindows + i, sizeof(TSKEY), &state, sizeof(state)) != 0) {
-      // If error occurs during taosHashPut expired windows, remove the smaIndex from pTsdb->pSmaStat, thus TSDB would
-      // tell query module to query raw TS data.
-      // N.B.
-      //  1) It is assumed to be extemely little probability event of fail to taosHashPut.
-      //  2) This would solve the inconsistency to some extent, but not completely, unless we record all expired
-      // windows failed to put into hash table.
-      taosHashCleanup(pItem->expiredWindows);
-      tfree(pItem->pSma);
-      taosHashRemove(pItemsHash, &indexUid, sizeof(indexUid));
-      tsdbUnRefSmaStat(pTsdb, pStat);
-      return TSDB_CODE_FAILED;
+    expiredWindows[i] = skey1 + i;
+  }
+#else
+
+#endif
+  // TODO: decode the msg <= end
+
+  if (tsdbCheckAndInitSmaEnv(pTsdb, TSDB_SMA_TYPE_TIME_RANGE) != TSDB_CODE_SUCCESS) {
+    terrno = TSDB_CODE_TDB_INIT_FAILED;
+    return TSDB_CODE_FAILED;
+  }
+
+#ifndef TSDB_SMA_TEST
+  TSKEY expiredWindows[SMA_TEST_EXPIRED_WINDOW_SIZE];
+#endif
+
+
+  // Firstly, assume that tSma can only be created on super table/normal table.
+  // getActiveTimeWindow
+
+
+  SSmaEnv *pEnv = REPO_SMA_ENV(pTsdb, TSDB_SMA_TYPE_TIME_RANGE);
+  SSmaStat *pStat = SMA_ENV_STAT(pEnv);
+  SHashObj *pItemsHash = SMA_ENV_STAT_ITEMS(pEnv);
+
+  TASSERT(pEnv != NULL && pStat != NULL && pItemsHash != NULL);
+
+
+
+  SSubmitMsgIter msgIter = {0};
+  SSubmitBlk    *pBlock = NULL;
+  SInterval      interval = {0};
+
+
+  if (tInitSubmitMsgIter(pMsg, &msgIter) != TSDB_CODE_SUCCESS) {
+    return TSDB_CODE_FAILED;
+  }
+
+  // basic procedure
+  // TODO: optimization
+  tsdbRefSmaStat(pTsdb, pStat);
+
+  while (true) {
+    tGetSubmitMsgNext(&msgIter, &pBlock);
+    if (pBlock == NULL) break;
+
+    int64_t suid = htobe64(pBlock->uid);
+    STSmaWrapper *pSW = NULL;
+    STSma        *pTSma = NULL;
+
+    while (true) {
+      SSubmitBlkIter blkIter = {0};
+      if (tInitSubmitBlkIter(pBlock, &blkIter) != TSDB_CODE_SUCCESS) {
+        tdFreeTSmaWrapper(pSW);
+        break;
+      }
+      STSRow *row = tGetSubmitBlkNext(&blkIter);
+      if (row == NULL) {
+        tdFreeTSmaWrapper(pSW);
+        break;
+      }
+      if(pSW == NULL) {
+        if((pSW =metaGetSmaInfoByTable(REPO_META(pTsdb), suid)) == NULL) {
+          break;
+        }
+        if((pSW->number) <= 0 || (pSW->tSma == NULL)) {
+          tdFreeTSmaWrapper(pSW);
+          break;
+        }
+        pTSma = pSW->tSma;
+      }
+
+      interval.interval = pTSma->interval;
+      interval.intervalUnit = pTSma->intervalUnit;
+      interval.offset = pTSma->offset;
+      interval.precision = REPO_CFG(pTsdb)->precision;
+      interval.sliding = pTSma->sliding;
+      interval.slidingUnit = pTSma->slidingUnit;
+
+      STimeWindow tw = getActiveTimeWindowX(TD_ROW_KEY(row), &interval);
+      tsdbSetExpiredWindow(pTsdb, pItemsHash, pTSma->indexUid, TD_ROW_KEY(row));
     }
-    tsdbDebug("vgId:%d smaIndex %" PRIi64 " tsKey %" PRIi64 " is put to hash", REPO_ID(pTsdb), indexUid,
-              expiredWindows[i]);
   }
 
   tsdbUnRefSmaStat(pTsdb, pStat);
+
   return TSDB_CODE_SUCCESS;
 }
 
+/**
+ * @brief When sma data received from stream computing, make the relative expired window valid.
+ *
+ * @param pTsdb
+ * @param pStat
+ * @param indexUid
+ * @param skey
+ * @return int32_t
+ */
 static int32_t tsdbResetExpiredWindow(STsdb *pTsdb, SSmaStat *pStat, int64_t indexUid, TSKEY skey) {
   SSmaStatItem *pItem = NULL;
 
@@ -582,7 +668,7 @@ static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t k
 
 #ifdef _TEST_SMA_PRINT_DEBUG_LOG_
   uint32_t valueSize = 0;
-  void *   data = tsdbGetSmaDataByKey(pDBFile, smaKey, keyLen, &valueSize);
+  void    *data = tsdbGetSmaDataByKey(pDBFile, smaKey, keyLen, &valueSize);
   ASSERT(data != NULL);
   for (uint32_t v = 0; v < valueSize; v += 8) {
     tsdbWarn("vgId:%d insert sma data val[%d] %" PRIi64, REPO_ID(pSmaH->pTsdb), v, *(int64_t *)POINTER_SHIFT(data, v));
@@ -699,7 +785,7 @@ static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *p
       assert(pTbData->dataLen > 0);
       STSmaColData *pColData = (STSmaColData *)POINTER_SHIFT(pTbData->data, tbLen);
       char          smaKey[SMA_KEY_LEN] = {0};
-      void *        pSmaKey = &smaKey;
+      void         *pSmaKey = &smaKey;
 #if 0
       printf("tsdbInsertTSmaDataSection: index %" PRIi64 ", skey %" PRIi64 " table[%" PRIi64 "]col[%" PRIu16 "]\n",
              pData->indexUid, pData->skey, pTbData->tableUid, pColData->colId);
@@ -773,11 +859,10 @@ static int32_t tsdbGetTSmaDays(STsdb *pTsdb, int64_t interval, int32_t storageLe
  * @return int32_t
  */
 static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
-  STsdbCfg *        pCfg = REPO_CFG(pTsdb);
+  STsdbCfg         *pCfg = REPO_CFG(pTsdb);
   STSmaDataWrapper *pData = (STSmaDataWrapper *)msg;
-  SSmaEnv *         pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
-  int64_t indexUid = SMA_TEST_INDEX_UID;
-
+  SSmaEnv          *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
+  int64_t           indexUid = SMA_TEST_INDEX_UID;
 
   if (pEnv == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
@@ -797,7 +882,7 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
     return TSDB_CODE_FAILED;
   }
 
-  SSmaStat *pStat = SMA_ENV_STAT(pTsdb->pTSmaEnv);
+  SSmaStat     *pStat = SMA_ENV_STAT(pTsdb->pTSmaEnv);
   SSmaStatItem *pItem = NULL;
 
   tsdbRefSmaStat(pTsdb, pStat);
@@ -812,8 +897,8 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
     return TSDB_CODE_FAILED;
   }
 
-  char    rPath[TSDB_FILENAME_LEN] = {0};
-  char    aPath[TSDB_FILENAME_LEN] = {0};
+  char rPath[TSDB_FILENAME_LEN] = {0};
+  char aPath[TSDB_FILENAME_LEN] = {0};
   snprintf(rPath, TSDB_FILENAME_LEN, "%s%s%" PRIi64, SMA_ENV_PATH(pEnv), TD_DIRSEP, indexUid);
   tfsAbsoluteName(REPO_TFS(pTsdb), SMA_ENV_DID(pEnv), rPath, aPath);
   if (!taosCheckExistFile(aPath)) {
@@ -859,9 +944,9 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
 /**
  * @brief Drop tSma data and local cache
  *        - insert/query reference
- * @param pTsdb 
- * @param msg 
- * @return int32_t 
+ * @param pTsdb
+ * @param msg
+ * @return int32_t
  */
 static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid) {
   SSmaEnv *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
@@ -902,8 +987,7 @@ static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid) {
     }
   }
   // clear sma data files
-  // TODO: 
-
+  // TODO:
 }
 
 static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int32_t fid) {
@@ -917,9 +1001,9 @@ static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData,
 }
 
 static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
-  STsdbCfg *        pCfg = REPO_CFG(pTsdb);
+  STsdbCfg         *pCfg = REPO_CFG(pTsdb);
   STSmaDataWrapper *pData = (STSmaDataWrapper *)msg;
-  SSmaEnv *         pEnv = atomic_load_ptr(&pTsdb->pRSmaEnv);
+  SSmaEnv          *pEnv = atomic_load_ptr(&pTsdb->pRSmaEnv);
 
   if (pEnv == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
@@ -1137,7 +1221,7 @@ static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_
             tReadH.dFile.path, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
             *(int64_t *)POINTER_SHIFT(smaKey, 10), SMA_KEY_LEN);
 
-  void *   result = NULL;
+  void    *result = NULL;
   uint32_t valueSize = 0;
   if ((result = tsdbGetSmaDataByKey(&tReadH.dFile, smaKey, SMA_KEY_LEN, &valueSize)) == NULL) {
     tsdbWarn("vgId:%d get sma data failed from smaIndex %" PRIi64 ", smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64
@@ -1219,15 +1303,8 @@ int32_t tsdbRemoveTSmaData(STsdb *pTsdb, void *smaIndex, STimeWindow *pWin) {
 }
 #endif
 
-/**
- * @brief Insert/Update tSma(Time-range-wise SMA) data from stream computing engine
- *
- * @param pTsdb
- * @param param
- * @param msg
- * @return int32_t
- * TODO: Who is responsible for resource allocate and release?
- */
+
+// TODO: Who is responsible for resource allocate and release?
 int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg) {
   int32_t code = TSDB_CODE_SUCCESS;
   if ((code = tsdbInsertTSmaDataImpl(pTsdb, msg)) < 0) {
@@ -1236,21 +1313,14 @@ int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg) {
   return code;
 }
 
-int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, int8_t smaType, char *msg) {
+int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, const char *msg) {
   int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbUpdateExpiredWindow(pTsdb, smaType, msg)) < 0) {
+  if ((code = tsdbUpdateExpiredWindowImpl(pTsdb, msg)) < 0) {
     tsdbWarn("vgId:%d update expired sma window failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
   }
   return code;
 }
 
-/**
- * @brief Insert Time-range-wise Rollup Sma(RSma) data
- *
- * @param pTsdb
- * @param msg
- * @return int32_t
- */
 int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
   int32_t code = TSDB_CODE_SUCCESS;
   if ((code = tsdbInsertRSmaDataImpl(pTsdb, msg)) < 0) {
@@ -1259,20 +1329,7 @@ int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
   return code;
 }
 
-/**
- * @brief Get tSma data
- * 
- * @param pTsdb 
- * @param pData 
- * @param indexUid 
- * @param interval 
- * @param intervalUnit 
- * @param tableUid 
- * @param colId 
- * @param querySKey 
- * @param nMaxResult 
- * @return int32_t 
- */
+
 int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
                         tb_uid_t tableUid, col_id_t colId, TSKEY querySKey, int32_t nMaxResult) {
   int32_t code = TSDB_CODE_SUCCESS;
@@ -1283,13 +1340,7 @@ int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid,
   return code;
 }
 
-/**
- * @brief Drop tSma Data and caches
- * 
- * @param pTsdb 
- * @param msg 
- * @return int32_t 
- */
+
 int32_t tsdbDropTSmaData(STsdb *pTsdb, int64_t indexUid) {
   int32_t code = TSDB_CODE_SUCCESS;
   if ((code = tsdbDropTSmaDataImpl(pTsdb, indexUid)) < 0) {

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

@@ -32,48 +32,4 @@ int tsdbInsertData(STsdb *pTsdb, SSubmitReq *pMsg, SSubmitRsp *pRsp) {
     }
   }
   return tsdbMemTableInsert(pTsdb, pTsdb->mem, pMsg, NULL);
-}
-
+}
\ No newline at end of file
-#if 0
-/**
- * @brief Insert/Update tSma(Time-range-wise SMA) data from stream computing engine
- * 
- * @param pTsdb 
- * @param param 
- * @param msg 
- * @return int32_t 
- * TODO: Who is responsible for resource allocate and release?
- */
-int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg) {
-  int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbInsertTSmaDataImpl(pTsdb, msg)) < 0) {
-    tsdbWarn("vgId:%d insert tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
-  }
-  return code;
-}
-
-int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, int8_t smaType, char *msg) {
-  int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbUpdateExpiredWindow(pTsdb, smaType, msg)) < 0) {
-    tsdbWarn("vgId:%d update expired sma window failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
-  }
-  return code;
-}
-
-/**
- * @brief Insert Time-range-wise Rollup Sma(RSma) data
- *
- * @param pTsdb
- * @param param
- * @param msg
- * @return int32_t
- */
-int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
-  int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbInsertRSmaDataImpl(pTsdb, msg)) < 0) {
-    tsdbWarn("vgId:%d insert rSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
-  }
-  return code;
-}
-
-#endif
\ No newline at end of file

source/dnode/vnode/test/tsdbSmaTest.cpp

@@ -98,7 +98,7 @@ TEST(testCase, tSma_Meta_Encode_Decode_Test) {
   tSma.slidingUnit = TIME_UNIT_HOUR;
   tSma.sliding = 0;
   tstrncpy(tSma.indexName, "sma_index_test", TSDB_INDEX_NAME_LEN);
-  tstrncpy(tSma.timezone, "Asia/Shanghai", TD_TIMEZONE_LEN);
+  tSma.timezoneInt = 8;
   tSma.indexUid = 2345678910;
   tSma.tableUid = 1234567890;
 
@@ -128,7 +128,7 @@ TEST(testCase, tSma_Meta_Encode_Decode_Test) {
     ASSERT_EQ(pSma->intervalUnit, qSma->intervalUnit);
     ASSERT_EQ(pSma->slidingUnit, qSma->slidingUnit);
     ASSERT_STRCASEEQ(pSma->indexName, qSma->indexName);
-    ASSERT_STRCASEEQ(pSma->timezone, qSma->timezone);
+    ASSERT_EQ(pSma->timezoneInt, qSma->timezoneInt);
     ASSERT_EQ(pSma->indexUid, qSma->indexUid);
     ASSERT_EQ(pSma->tableUid, qSma->tableUid);
     ASSERT_EQ(pSma->interval, qSma->interval);
@@ -150,7 +150,7 @@ TEST(testCase, tSma_Meta_Encode_Decode_Test) {
 TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   const char *   smaIndexName1 = "sma_index_test_1";
   const char *   smaIndexName2 = "sma_index_test_2";
-  const char *   timezone = "Asia/Shanghai";
+  int8_t         timezone = 8;
   const char *   expr = "select count(a,b, top 20), from table interval 1d, sliding 1h;";
   const char *   tagsFilter = "I'm tags filter";
   const char *   smaTestDir = "./smaTest";
@@ -167,7 +167,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   tSma.sliding = 0;
   tSma.indexUid = indexUid1;
   tstrncpy(tSma.indexName, smaIndexName1, TSDB_INDEX_NAME_LEN);
-  tstrncpy(tSma.timezone, timezone, TD_TIMEZONE_LEN);
+  tSma.timezoneInt = 8;
   tSma.tableUid = tbUid;
 
   tSma.exprLen = strlen(expr);
@@ -207,7 +207,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   qSmaCfg = metaGetSmaInfoByIndex(pMeta, indexUid1);
   assert(qSmaCfg != NULL);
   printf("name1 = %s\n", qSmaCfg->indexName);
-  printf("timezone1 = %s\n", qSmaCfg->timezone);
+  printf("timezone1 = %" PRIi8 "\n", qSmaCfg->timezoneInt);
   printf("expr1 = %s\n", qSmaCfg->expr != NULL ? qSmaCfg->expr : "");
   printf("tagsFilter1 = %s\n", qSmaCfg->tagsFilter != NULL ? qSmaCfg->tagsFilter : "");
   ASSERT_STRCASEEQ(qSmaCfg->indexName, smaIndexName1);
@@ -218,7 +218,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   qSmaCfg = metaGetSmaInfoByIndex(pMeta, indexUid2);
   assert(qSmaCfg != NULL);
   printf("name2 = %s\n", qSmaCfg->indexName);
-  printf("timezone2 = %s\n", qSmaCfg->timezone);
+  printf("timezone2 = %" PRIi8 "\n", qSmaCfg->timezoneInt);
   printf("expr2 = %s\n", qSmaCfg->expr != NULL ? qSmaCfg->expr : "");
   printf("tagsFilter2 = %s\n", qSmaCfg->tagsFilter != NULL ? qSmaCfg->tagsFilter : "");
   ASSERT_STRCASEEQ(qSmaCfg->indexName, smaIndexName2);
@@ -246,13 +246,13 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   assert(pSW != NULL);
   ASSERT_EQ(pSW->number, nCntTSma);
   ASSERT_STRCASEEQ(pSW->tSma->indexName, smaIndexName1);
-  ASSERT_STRCASEEQ(pSW->tSma->timezone, timezone);
+  ASSERT_EQ(pSW->tSma->timezoneInt, timezone);
   ASSERT_STRCASEEQ(pSW->tSma->expr, expr);
   ASSERT_STRCASEEQ(pSW->tSma->tagsFilter, tagsFilter);
   ASSERT_EQ(pSW->tSma->indexUid, indexUid1);
   ASSERT_EQ(pSW->tSma->tableUid, tbUid);
   ASSERT_STRCASEEQ((pSW->tSma + 1)->indexName, smaIndexName2);
-  ASSERT_STRCASEEQ((pSW->tSma + 1)->timezone, timezone);
+  ASSERT_EQ((pSW->tSma + 1)->timezoneInt, timezone);
   ASSERT_STRCASEEQ((pSW->tSma + 1)->expr, expr);
   ASSERT_STRCASEEQ((pSW->tSma + 1)->tagsFilter, tagsFilter);
   ASSERT_EQ((pSW->tSma + 1)->indexUid, indexUid2);
@@ -284,7 +284,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
 TEST(testCase, tSma_Data_Insert_Query_Test) {
   // step 1: prepare meta
   const char *   smaIndexName1 = "sma_index_test_1";
-  const char *   timezone = "Asia/Shanghai";
+  const int8_t   timezone = 8;
   const char *   expr = "select count(a,b, top 20), from table interval 1d, sliding 1h;";
   const char *   tagsFilter = "where tags.location='Beijing' and tags.district='ChaoYang'";
   const char *   smaTestDir = "./smaTest";
@@ -305,7 +305,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   tSma.sliding = 0;
   tSma.indexUid = indexUid1;
   tstrncpy(tSma.indexName, smaIndexName1, TSDB_INDEX_NAME_LEN);
-  tstrncpy(tSma.timezone, timezone, TD_TIMEZONE_LEN);
+  tSma.timezoneInt = timezone;
   tSma.tableUid = tbUid;
 
   tSma.exprLen = strlen(expr);
@@ -369,7 +369,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
 
   char *msg = (char *)calloc(1, 100);
   ASSERT_NE(msg, nullptr);
-  ASSERT_EQ(tsdbUpdateSmaWindow(pTsdb, TSDB_SMA_TYPE_TIME_RANGE, msg), 0);
+  ASSERT_EQ(tsdbUpdateSmaWindow(pTsdb, msg), 0);
 
   // init
   int32_t allocCnt = 0;

source/libs/executor/inc/executorimpl.h

@@ -243,6 +243,8 @@ typedef struct STaskAttr {
 
 struct SOperatorInfo;
 
+typedef void (*__optr_encode_fn_t)(struct SOperatorInfo* pOperator, char **result, int32_t *length);
+typedef bool (*__optr_decode_fn_t)(struct SOperatorInfo* pOperator, char *result, int32_t length);
 typedef int32_t (*__optr_open_fn_t)(struct SOperatorInfo* param);
 typedef SSDataBlock* (*__optr_fn_t)(struct SOperatorInfo* param, bool* newgroup);
 typedef void (*__optr_close_fn_t)(void* param, int32_t num);
@@ -332,6 +334,8 @@ typedef struct SOperatorInfo {
   __optr_fn_t            cleanupFn;
   __optr_close_fn_t      closeFn;
   __optr_open_fn_t       _openFn;          // DO NOT invoke this function directly
+  __optr_encode_fn_t     encodeResultRow;           //
+  __optr_decode_fn_t     decodeResultRow;
 } SOperatorInfo;
 
 typedef struct {
@@ -661,6 +665,9 @@ SOperatorInfo* createGroupOperatorInfo(SOperatorInfo* downstream, SExprInfo* pEx
                                        SArray* pGroupColList, SExecTaskInfo* pTaskInfo, const STableGroupInfo* pTableGroupInfo);
 SOperatorInfo* createFillOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExpr, int32_t numOfCols, SInterval* pInterval, SSDataBlock* pResBlock,
                                       int32_t fillType, char* fillVal, bool multigroupResult, SExecTaskInfo* pTaskInfo);
+SOperatorInfo* createDistinctOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream, SExprInfo* pExpr,
+                                          int32_t numOfOutput);
+SOperatorInfo* createTableBlockInfoScanOperator(void* pTsdbReadHandle, STaskRuntimeEnv* pRuntimeEnv);
 
 SOperatorInfo* createTableSeqScanOperatorInfo(void* pTsdbReadHandle, STaskRuntimeEnv* pRuntimeEnv);
 SOperatorInfo* createAllTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,
@@ -671,9 +678,7 @@ SOperatorInfo* createMultiTableTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntim
 SOperatorInfo* createAllMultiTableTimeIntervalOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,
                                                            SExprInfo* pExpr, int32_t numOfOutput);
 SOperatorInfo* createTagScanOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SExprInfo* pExpr, int32_t numOfOutput);
-SOperatorInfo* createDistinctOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream, SExprInfo* pExpr,
-                                          int32_t numOfOutput);
-SOperatorInfo* createTableBlockInfoScanOperator(void* pTsdbReadHandle, STaskRuntimeEnv* pRuntimeEnv);
+
 SOperatorInfo* createMultiwaySortOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SExprInfo* pExpr, int32_t numOfOutput,
                                               int32_t numOfRows, void* merger);
 SOperatorInfo* createGlobalAggregateOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorInfo* downstream,

source/libs/executor/src/executorimpl.c

@@ -674,12 +674,7 @@ static void getInitialStartTimeWindow(SInterval* pInterval, int32_t precision, T
 
     int64_t key = w->skey;
     while(key < ts) { // moving towards end
-      if (pInterval->intervalUnit == 'n' || pInterval->intervalUnit == 'y') {
-        key = taosTimeAdd(key, pInterval->sliding, pInterval->slidingUnit, precision);
-      } else {
-        key += pInterval->sliding;
-      }
-
+      key = taosTimeAdd(key, pInterval->sliding, pInterval->slidingUnit, precision);
       if (key >= ts) {
         break;
       }
@@ -695,12 +690,7 @@ static STimeWindow getActiveTimeWindow(SResultRowInfo * pResultRowInfo, int64_t
 
  if (pResultRowInfo->curPos == -1) {  // the first window, from the previous stored value
     getInitialStartTimeWindow(pInterval, precision, ts, &w, win->ekey, true);
-
-    if (pInterval->intervalUnit == 'n' || pInterval->intervalUnit == 'y') {
-      w.ekey = taosTimeAdd(w.skey, pInterval->interval, pInterval->intervalUnit, precision) - 1;
-    } else {
-      w.ekey = w.skey + pInterval->interval - 1;
-    }
+    w.ekey = taosTimeAdd(w.skey, pInterval->interval, pInterval->intervalUnit, precision) - 1;
   } else {
     w = getResultRow(pResultRowInfo, pResultRowInfo->curPos)->win;
   }
@@ -722,7 +712,7 @@ static STimeWindow getActiveTimeWindow(SResultRowInfo * pResultRowInfo, int64_t
       }
 
       w.skey = st;
-      w.ekey = w.skey + pInterval->interval - 1;
+      w.ekey = taosTimeAdd(w.skey, pInterval->interval, pInterval->intervalUnit, precision) - 1;
     }
   }
   return w;
@@ -2208,7 +2198,7 @@ static SqlFunctionCtx* createSqlFunctionCtx_rv(SExprInfo* pExprInfo, int32_t num
   }
 
   for(int32_t i = 1; i < numOfOutput; ++i) {
-    (*rowCellInfoOffset)[i] = (int32_t)((*rowCellInfoOffset)[i - 1] + sizeof(SResultRowEntryInfo) + pFuncCtx[i].resDataInfo.interBufSize);
+    (*rowCellInfoOffset)[i] = (int32_t)((*rowCellInfoOffset)[i - 1] + sizeof(SResultRowEntryInfo) + pFuncCtx[i - 1].resDataInfo.interBufSize);
   }
 
   setCtxTagColumnInfo(pFuncCtx, numOfOutput);
@@ -2407,7 +2397,7 @@ static bool isCachedLastQuery(STaskAttr *pQueryAttr) {
 /////////////////////////////////////////////////////////////////////////////////////////////
 //todo refactor : return window
 void getAlignQueryTimeWindow(SInterval* pInterval, int32_t precision, int64_t key, int64_t keyFirst, int64_t keyLast, STimeWindow *win) {
-  assert(key >= keyFirst && key <= keyLast && pInterval->sliding <= pInterval->interval);
+  ASSERT(key >= keyFirst && key <= keyLast);
   win->skey = taosTimeTruncate(key, pInterval, precision);
 
   /*
@@ -2417,10 +2407,8 @@ void getAlignQueryTimeWindow(SInterval* pInterval, int32_t precision, int64_t ke
   if (keyFirst > (INT64_MAX - pInterval->interval)) {
     assert(keyLast - keyFirst < pInterval->interval);
     win->ekey = INT64_MAX;
-  } else if (pInterval->intervalUnit == 'n' || pInterval->intervalUnit == 'y') {
-    win->ekey = taosTimeAdd(win->skey, pInterval->interval, pInterval->intervalUnit, precision) - 1;
   } else {
-    win->ekey = win->skey + pInterval->interval - 1;
+    win->ekey = taosTimeAdd(win->skey, pInterval->interval, pInterval->intervalUnit, precision) - 1;
   }
 }
 
@@ -6391,6 +6379,111 @@ static SSDataBlock* getAggregateResult(SOperatorInfo *pOperator, bool* newgroup)
   return (blockDataGetNumOfRows(pInfo->pRes) != 0)? pInfo->pRes:NULL;
 }
 
+static void aggEncodeResultRow(SOperatorInfo* pOperator, char **result, int32_t *length) {
+  SAggOperatorInfo *pAggInfo = pOperator->info;
+  SAggSupporter   *pSup = &pAggInfo->aggSup;
+
+  int32_t size = taosHashGetSize(pSup->pResultRowHashTable);
+  size_t keyLen = POINTER_BYTES;    // estimate the key length
+  int32_t totalSize = sizeof(int32_t) + size * (sizeof(int32_t) + keyLen + sizeof(int32_t) + pSup->resultRowSize);
+  *result = calloc(1, totalSize);
+  if(*result == NULL){
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return;
+  }
+  *(int32_t*)(*result) = size;
+  int32_t offset = sizeof(int32_t);
+  void *pIter = taosHashIterate(pSup->pResultRowHashTable, NULL);
+  while (pIter) {
+    void *key = taosHashGetKey(pIter, &keyLen);
+    SResultRow **p1 = (SResultRow **)pIter;
+
+    // recalculate the result size
+    int32_t realTotalSize = offset + sizeof(int32_t) + keyLen + sizeof(int32_t) + pSup->resultRowSize;
+    if (realTotalSize > totalSize){
+      char *tmp = realloc(*result, realTotalSize);
+      if (tmp == NULL){
+        terrno = TSDB_CODE_OUT_OF_MEMORY;
+        free(*result);
+        *result = NULL;
+        return;
+      }else{
+        *result = tmp;
+      }
+    }
+    // save key
+    *(int32_t*)(*result + offset) = keyLen;
+    offset += sizeof(int32_t);
+    memcpy(*result + offset, key, keyLen);
+    offset += keyLen;
+
+    // save value
+    *(int32_t*)(*result + offset) = pSup->resultRowSize;
+    offset += sizeof(int32_t);
+    memcpy(*result + offset, *p1, pSup->resultRowSize);
+    offset += pSup->resultRowSize;
+
+    pIter = taosHashIterate(pSup->pResultRowHashTable, pIter);
+  }
+
+  if(length) {
+    *length = offset;
+  }
+  return;
+}
+
+static bool aggDecodeResultRow(SOperatorInfo* pOperator, char *result, int32_t length) {
+  if (!result || length <= 0){
+    return false;
+  }
+
+  SAggOperatorInfo *pAggInfo = pOperator->info;
+  SAggSupporter    *pSup     = &pAggInfo->aggSup;
+  SOptrBasicInfo   *pInfo    = &pAggInfo->binfo;
+
+  //  int32_t size = taosHashGetSize(pSup->pResultRowHashTable);
+  int32_t count = *(int32_t*)(result);
+
+  int32_t offset = sizeof(int32_t);
+  while(count-- > 0 && length > offset){
+    int32_t keyLen = *(int32_t*)(result + offset);
+    offset += sizeof(int32_t);
+
+    uint64_t tableGroupId = *(uint64_t *)(result + offset);
+    SResultRow *resultRow = getNewResultRow_rv(pSup->pResultBuf, tableGroupId, pSup->resultRowSize);
+    if (!resultRow){
+      terrno = TSDB_CODE_TSC_INVALID_INPUT;
+      return false;
+    }
+    // add a new result set for a new group
+    taosHashPut(pSup->pResultRowHashTable, result + offset, keyLen, &resultRow, POINTER_BYTES);
+
+    offset += keyLen;
+    int32_t valueLen = *(int32_t*)(result + offset);
+    if (valueLen != pSup->resultRowSize){
+      terrno = TSDB_CODE_TSC_INVALID_INPUT;
+      return false;
+    }
+    offset += sizeof(int32_t);
+    int32_t pageId = resultRow->pageId;
+    int32_t pOffset = resultRow->offset;
+    memcpy(resultRow, result + offset, valueLen);
+    resultRow->pageId = pageId;
+    resultRow->offset = pOffset;
+    offset += valueLen;
+
+    initResultRow(resultRow);
+
+    pInfo->resultRowInfo.pPosition[pInfo->resultRowInfo.size++] = (SResultRowPosition) {.pageId = resultRow->pageId, .offset = resultRow->offset};
+  }
+
+  if (offset != length){
+    terrno = TSDB_CODE_TSC_INVALID_INPUT;
+    return false;
+  }
+  return true;
+}
+
 static SSDataBlock* doMultiTableAggregate(SOperatorInfo *pOperator, bool* newgroup) {
   if (pOperator->status == OP_EXEC_DONE) {
     return NULL;
@@ -7086,10 +7179,10 @@ static SSDataBlock* hashGroupbyAggregate(SOperatorInfo *pOperator, bool* newgrou
 static void doHandleRemainBlockForNewGroupImpl(SFillOperatorInfo *pInfo, SResultInfo* pResultInfo, bool* newgroup, SExecTaskInfo* pTaskInfo) {
   pInfo->totalInputRows = pInfo->existNewGroupBlock->info.rows;
 
-//  int64_t ekey = Q_STATUS_EQUAL(pRuntimeEnv->status, TASK_COMPLETED)? pTaskInfo->window.ekey:pInfo->existNewGroupBlock->info.window.ekey;
+  int64_t ekey = Q_STATUS_EQUAL(pTaskInfo->status, TASK_COMPLETED)? pTaskInfo->window.ekey:pInfo->existNewGroupBlock->info.window.ekey;
   taosResetFillInfo(pInfo->pFillInfo, getFillInfoStart(pInfo->pFillInfo));
 
-//  taosFillSetStartInfo(pInfo->pFillInfo, pInfo->existNewGroupBlock->info.rows, ekey);
+  taosFillSetStartInfo(pInfo->pFillInfo, pInfo->existNewGroupBlock->info.rows, ekey);
   taosFillSetInputDataBlock(pInfo->pFillInfo, pInfo->existNewGroupBlock);
 
   doFillTimeIntervalGapsInResults(pInfo->pFillInfo, pInfo->pRes, pResultInfo->capacity, pInfo->p);
@@ -7097,7 +7190,7 @@ static void doHandleRemainBlockForNewGroupImpl(SFillOperatorInfo *pInfo, SResult
   *newgroup = true;
 }
 
-static void doHandleRemainBlockFromNewGroup(SFillOperatorInfo *pInfo, SResultInfo *pResultInfo, bool *newgroup) {
+static void doHandleRemainBlockFromNewGroup(SFillOperatorInfo *pInfo, SResultInfo *pResultInfo, bool *newgroup, SExecTaskInfo* pTaskInfo) {
   if (taosFillHasMoreResults(pInfo->pFillInfo)) {
     *newgroup = false;
     doFillTimeIntervalGapsInResults(pInfo->pFillInfo, pInfo->pRes, (int32_t)pResultInfo->capacity, pInfo->p);
@@ -7108,12 +7201,13 @@ static void doHandleRemainBlockFromNewGroup(SFillOperatorInfo *pInfo, SResultInf
 
   // handle the cached new group data block
   if (pInfo->existNewGroupBlock) {
-//    doHandleRemainBlockForNewGroupImpl(pInfo, pResultInfo, newgroup);
+    doHandleRemainBlockForNewGroupImpl(pInfo, pResultInfo, newgroup, pTaskInfo);
   }
 }
 
 static SSDataBlock* doFill(SOperatorInfo *pOperator, bool* newgroup) {
   SFillOperatorInfo *pInfo = pOperator->info;
+  SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
 
   SResultInfo* pResultInfo = &pOperator->resultInfo;
   blockDataCleanup(pInfo->pRes);
@@ -7121,7 +7215,7 @@ static SSDataBlock* doFill(SOperatorInfo *pOperator, bool* newgroup) {
     return NULL;
   }
 
-  doHandleRemainBlockFromNewGroup(pInfo, pResultInfo, newgroup);
+  doHandleRemainBlockFromNewGroup(pInfo, pResultInfo, newgroup, pTaskInfo);
   if (pInfo->pRes->info.rows > pResultInfo->threshold || (!pInfo->multigroupResult && pInfo->pRes->info.rows > 0)) {
     return pInfo->pRes;
   }
@@ -7142,7 +7236,7 @@ static SSDataBlock* doFill(SOperatorInfo *pOperator, bool* newgroup) {
 
       // Fill the previous group data block, before handle the data block of new group.
       // Close the fill operation for previous group data block
-//      taosFillSetStartInfo(pInfo->pFillInfo, 0, pRuntimeEnv->pQueryAttr->window.ekey);
+      taosFillSetStartInfo(pInfo->pFillInfo, 0, pTaskInfo->window.ekey);
     } else {
       if (pBlock == NULL) {
         if (pInfo->totalInputRows == 0) {
@@ -7150,7 +7244,7 @@ static SSDataBlock* doFill(SOperatorInfo *pOperator, bool* newgroup) {
           return NULL;
         }
 
-//        taosFillSetStartInfo(pInfo->pFillInfo, 0, pRuntimeEnv->pQueryAttr->window.ekey);
+        taosFillSetStartInfo(pInfo->pFillInfo, 0, pTaskInfo->window.ekey);
       } else {
         pInfo->totalInputRows += pBlock->info.rows;
         taosFillSetStartInfo(pInfo->pFillInfo, pBlock->info.rows, pBlock->info.window.ekey);
@@ -7168,14 +7262,13 @@ static SSDataBlock* doFill(SOperatorInfo *pOperator, bool* newgroup) {
         return pInfo->pRes;
       }
 
-//      doHandleRemainBlockFromNewGroup(pInfo, pRuntimeEnv, newgroup);
+      doHandleRemainBlockFromNewGroup(pInfo, pResultInfo, newgroup, pTaskInfo);
       if (pInfo->pRes->info.rows > pOperator->resultInfo.threshold || pBlock == NULL) {
         return pInfo->pRes;
       }
     } else if (pInfo->existNewGroupBlock) {  // try next group
       assert(pBlock != NULL);
-//      doHandleRemainBlockForNewGroupImpl(pInfo, pRuntimeEnv, newgroup);
-
+      doHandleRemainBlockForNewGroupImpl(pInfo, pResultInfo, newgroup, pTaskInfo);
       if (pInfo->pRes->info.rows > pResultInfo->threshold) {
         return pInfo->pRes;
       }
@@ -7312,6 +7405,8 @@ SOperatorInfo* createAggregateOperatorInfo(SOperatorInfo* downstream, SExprInfo*
   pOperator->_openFn      = doOpenAggregateOptr;
   pOperator->getNextFn    = getAggregateResult;
   pOperator->closeFn      = destroyAggOperatorInfo;
+  pOperator->encodeResultRow = aggEncodeResultRow;
+  pOperator->decodeResultRow = aggDecodeResultRow;
 
   code = appendDownstream(pOperator, &downstream, 1);
   if (code != TSDB_CODE_SUCCESS) {
@@ -7576,7 +7671,7 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo*
   }
 
   pInfo->order     = TSDB_ORDER_ASC;
-  pInfo->precision = TSDB_TIME_PRECISION_MICRO;
+  pInfo->precision = TSDB_TIME_PRECISION_MILLI;
   pInfo->win       = pTaskInfo->window;
   pInfo->interval  = *pInterval;
 
@@ -7863,10 +7958,10 @@ SOperatorInfo* createFillOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExp
   pInfo->intervalInfo     = *pInterval;
 
   SResultInfo* pResultInfo = &pOperator->resultInfo;
-//  int32_t code = initFillInfo(pInfo, pExpr, numOfCols, fillVal, , pResultInfo->capacity, pTaskInfo->id.str, pInterval, fillType);
-//  if (code != TSDB_CODE_SUCCESS) {
-//    goto _error;
-//  }
+  int32_t code = initFillInfo(pInfo, pExpr, numOfCols, (int64_t*) fillVal, pTaskInfo->window, pResultInfo->capacity, pTaskInfo->id.str, pInterval, fillType);
+  if (code != TSDB_CODE_SUCCESS) {
+    goto _error;
+  }
 
   pOperator->name         = "FillOperator";
   pOperator->blockingOptr = false;
@@ -7881,7 +7976,7 @@ SOperatorInfo* createFillOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExp
 
   pOperator->closeFn      = destroySFillOperatorInfo;
 
-  int32_t code = appendDownstream(pOperator, &downstream, 1);
+  code = appendDownstream(pOperator, &downstream, 1);
   return pOperator;
 
   _error:

source/libs/function/CMakeLists.txt

 aux_source_directory(src FUNCTION_SRC)
+list(REMOVE_ITEM FUNCTION_SRC src/udfd.c)
 add_library(function STATIC ${FUNCTION_SRC})
 target_include_directories(
     function
-    PUBLIC "${CMAKE_SOURCE_DIR}/include/libs/function"
+    PUBLIC
+        "${CMAKE_SOURCE_DIR}/include/libs/function"
+        "${CMAKE_SOURCE_DIR}/contrib/libuv/include"
     PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/inc"
 )
 
 target_link_libraries(
     function
+    PUBLIC uv_a
     PRIVATE os util common nodes
-)
+)
\ No newline at end of file
+
+add_executable(runUdf test/runUdf.c)
+target_include_directories(
+        runUdf
+        PUBLIC
+            "${CMAKE_SOURCE_DIR}/include/libs/function"
+            "${CMAKE_SOURCE_DIR}/contrib/libuv/include"
+        PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/inc"
+)
+target_link_libraries(
+        runUdf
+        PUBLIC uv_a
+        PRIVATE os util common nodes function
+)
+
+add_library(udf1 MODULE test/udf1.c)
+target_include_directories(
+        udf1
+        PUBLIC
+            "${CMAKE_SOURCE_DIR}/include/libs/function"
+        PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/inc"
+)
+
+#SET(EXECUTABLE_OUTPUT_PATH ${CMAKE_BINARY_DIR}/build/bin)
+add_executable(udfd src/udfd.c)
+target_include_directories(
+        udfd
+        PUBLIC
+            "${CMAKE_SOURCE_DIR}/include/libs/function"
+            "${CMAKE_SOURCE_DIR}/contrib/libuv/include"
+        PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/inc"
+)
+
+target_link_libraries(
+        udfd
+        PUBLIC uv_a
+        PRIVATE os util common nodes function
+)
+

source/libs/function/inc/tudf.h

@@ -20,68 +20,116 @@
 extern "C" {
 #endif
 
-#include "os.h"
-#include "taoserror.h"
+//======================================================================================
+//begin API to taosd and qworker
+/**
+ * start udf dameon service
+ * @return error code
+ */
+int32_t startUdfService();
+
+/**
+ * stop udf dameon service
+ * @return error code
+ */
+int32_t stopUdfService();
 
 enum {
-  TSDB_UDF_FUNC_NORMAL = 0,
-  TSDB_UDF_FUNC_INIT,
-  TSDB_UDF_FUNC_FINALIZE,
-  TSDB_UDF_FUNC_MERGE,
-  TSDB_UDF_FUNC_DESTROY,
-  TSDB_UDF_FUNC_MAX_NUM
+  TSDB_UDF_TYPE_SCALAR = 0,
+  TSDB_UDF_TYPE_AGGREGATE = 1
 };
 
-typedef struct SUdfInit {
-  int32_t  maybe_null; /* 1 if function can return NULL */
-  uint32_t decimals;   /* for real functions */
-  uint64_t length;     /* For string functions */
-  char*    ptr;        /* free pointer for function data */
-  int32_t  const_item; /* 0 if result is independent of arguments */
-
-  // script like lua/javascript
-  void* script_ctx;
-  void (*destroyCtxFunc)(void* script_ctx);
-} SUdfInit;
+enum {
+  TSDB_UDF_SCRIPT_BIN_LIB = 0,
+  TSDB_UDF_SCRIPT_LUA = 1,
+};
 
 typedef struct SUdfInfo {
-  int32_t functionId;                    // system assigned function id
-  int32_t funcType;                      // scalar function or aggregate function
-  int8_t  resType;                       // result type
-  int16_t resBytes;                      // result byte
-  int32_t contLen;                       // content length
-  int32_t bufSize;                       // interbuf size
-  char*   name;                          // function name
-  void*   handle;                        // handle loaded in mem
-  void*   funcs[TSDB_UDF_FUNC_MAX_NUM];  // function ptr
-
-  // for script like lua/javascript only
-  int   isScript;
-  void* pScriptCtx;
-
-  SUdfInit init;
-  char*    content;
-  char*    path;
+  char   *udfName;        // function name
+  int32_t udfType;    // scalar function or aggregate function
+  int8_t    scriptType;
+  char *path;
+
+  int8_t  resType;     // result type
+  int16_t resBytes;    // result byte
+  int32_t bufSize;     //interbuf size
+
 } SUdfInfo;
 
+typedef void *UdfHandle;
+
+/**
+ * setup udf
+ * @param udf, in
+ * @param handle, out
+ * @return error code
+ */
+int32_t setupUdf(SUdfInfo* udf, UdfHandle *handle);
+
+
+enum {
+  TSDB_UDF_STEP_NORMAL = 0,
+  TSDB_UDF_STEP_MERGE,
+  TSDb_UDF_STEP_FINALIZE,
+  TSDB_UDF_STEP_MAX_NUM
+};
+/**
+ * call udf
+ * @param handle udf handle
+ * @param step
+ * @param state
+ * @param stateSize
+ * @param input
+ * @param newstate
+ * @param newStateSize
+ * @param output
+ * @return error code
+ */
+
+//TODO: must change the following after metadata flow and data flow between qworker and udfd is well defined
+typedef struct SUdfDataBlock {
+  char* data;
+  int32_t size;
+} SUdfDataBlock;
+
+int32_t callUdf(UdfHandle handle, int8_t step, char *state, int32_t stateSize, SUdfDataBlock input, char **newstate,
+                int32_t *newStateSize, SUdfDataBlock *output);
+
+/**
+ * tearn down udf
+ * @param handle
+ * @return
+ */
+int32_t teardownUdf(UdfHandle handle);
+
+// end API to taosd and qworker
+//=============================================================================================================================
+// TODO: Must change
+// begin API to UDF writer.
+
 // script
 
-typedef int32_t (*scriptInitFunc)(void* pCtx);
-typedef void (*scriptNormalFunc)(void* pCtx, char* data, int16_t iType, int16_t iBytes, int32_t numOfRows,
-                                 int64_t* ptList, int64_t key, char* dataOutput, char* tsOutput, int32_t* numOfOutput,
-                                 int16_t oType, int16_t oBytes);
-typedef void (*scriptFinalizeFunc)(void* pCtx, int64_t key, char* dataOutput, int32_t* numOfOutput);
-typedef void (*scriptMergeFunc)(void* pCtx, char* data, int32_t numOfRows, char* dataOutput, int32_t* numOfOutput);
-typedef void (*scriptDestroyFunc)(void* pCtx);
+//typedef int32_t (*scriptInitFunc)(void* pCtx);
+//typedef void (*scriptNormalFunc)(void* pCtx, char* data, int16_t iType, int16_t iBytes, int32_t numOfRows,
+//                                 int64_t* ptList, int64_t key, char* dataOutput, char* tsOutput, int32_t* numOfOutput,
+//                                 int16_t oType, int16_t oBytes);
+//typedef void (*scriptFinalizeFunc)(void* pCtx, int64_t key, char* dataOutput, int32_t* numOfOutput);
+//typedef void (*scriptMergeFunc)(void* pCtx, char* data, int32_t numOfRows, char* dataOutput, int32_t* numOfOutput);
+//typedef void (*scriptDestroyFunc)(void* pCtx);
 
 // dynamic lib
-typedef void (*udfNormalFunc)(char* data, int16_t itype, int16_t iBytes, int32_t numOfRows, int64_t* ts,
-                              char* dataOutput, char* interBuf, char* tsOutput, int32_t* numOfOutput, int16_t oType,
-                              int16_t oBytes, SUdfInit* buf);
-typedef int32_t (*udfInitFunc)(SUdfInit* data);
-typedef void (*udfFinalizeFunc)(char* dataOutput, char* interBuf, int32_t* numOfOutput, SUdfInit* buf);
-typedef void (*udfMergeFunc)(char* data, int32_t numOfRows, char* dataOutput, int32_t* numOfOutput, SUdfInit* buf);
-typedef void (*udfDestroyFunc)(SUdfInit* buf);
+typedef int32_t (*TUdfInitFunc)();
+typedef void (*TUdfDestroyFunc)();
+
+typedef void (*TUdfFunc)(int8_t step,
+                         char *state, int32_t stateSize, SUdfDataBlock input,
+                         char **newstate, int32_t *newStateSize, SUdfDataBlock *output);
+
+//typedef void (*udfMergeFunc)(char *data, int32_t numOfRows, char *dataOutput, int32_t* numOfOutput);
+//typedef void (*udfFinalizeFunc)(char* state, int32_t stateSize, SUdfDataBlock *output);
+
+// end API to UDF writer
+//=======================================================================================================================
 
 #ifdef __cplusplus
 }

source/libs/function/inc/tudfInt.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 TDENGINE_TUDF_INT_H
+#define TDENGINE_TUDF_INT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+//TODO replaces them with fnDebug
+//#define debugPrint(...) taosPrintLog("Function", DEBUG_INFO,  135, __VA_ARGS__)
+#define debugPrint(...) {fprintf(stderr, __VA_ARGS__);fprintf(stderr, "\n");}
+enum {
+  UDF_TASK_SETUP = 0,
+  UDF_TASK_CALL = 1,
+  UDF_TASK_TEARDOWN = 2
+
+};
+
+typedef struct SUdfSetupRequest {
+  char udfName[16]; //
+  int8_t scriptType; // 0:c, 1: lua, 2:js
+  int8_t udfType; //udaf, udf
+  int16_t pathSize;
+  char *path;
+} SUdfSetupRequest;
+
+typedef struct SUdfSetupResponse {
+  int64_t udfHandle;
+} SUdfSetupResponse;
+
+
+typedef struct SUdfCallRequest {
+  int64_t udfHandle;
+  int8_t step;
+
+  int32_t inputBytes;
+  char *input;
+
+  int32_t stateBytes;
+  char *state;
+} SUdfCallRequest;
+
+
+typedef struct SUdfCallResponse {
+  int32_t outputBytes;
+  char *output;
+  int32_t newStateBytes;
+  char *newState;
+} SUdfCallResponse;
+
+
+typedef struct SUdfTeardownRequest {
+  int64_t udfHandle;
+} SUdfTeardownRequest;
+
+
+typedef struct SUdfTeardownResponse {
+} SUdfTeardownResponse;
+
+typedef struct SUdfRequest {
+  int32_t msgLen;
+  int64_t seqNum;
+
+  int8_t type;
+  void *subReq;
+} SUdfRequest;
+
+typedef struct SUdfResponse {
+  int32_t msgLen;
+  int64_t seqNum;
+
+  int8_t type;
+  int32_t code;
+  void *subRsp;
+} SUdfResponse;
+
+int32_t decodeRequest(char *buf, int32_t bufLen, SUdfRequest **pRequest);
+int32_t encodeResponse(char **buf, int32_t *bufLen, SUdfResponse *response);
+int32_t encodeRequest(char **buf, int32_t *bufLen, SUdfRequest *request);
+int32_t decodeResponse(char *buf, int32_t bufLen, SUdfResponse **pResponse);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TDENGINE_TUDF_INT_H

source/libs/function/inc/udfc.h

+//
+// Created by shenglian on 28/02/22.
+//
+
+#ifndef UDF_UDF_H
+#define UDF_UDF_H
+
+#include <stdlib.h>
+#define DEBUG
+#ifdef DEBUG
+#define debugPrint(...) fprintf(__VA_ARGS__)
+#else
+#define debugPrint(...) /**/
+#endif
+
+enum {
+    UDF_TASK_SETUP = 0,
+    UDF_TASK_CALL = 1,
+    UDF_TASK_TEARDOWN = 2
+
+};
+
+typedef struct SSDataBlock{
+    char *data;
+    int32_t size;
+} SSDataBlock;
+
+typedef struct SUdfInfo {
+    char *udfName;
+    char *path;
+} SUdfInfo;
+
+typedef void *UdfHandle;
+
+int32_t startUdfService();
+
+int32_t stopUdfService();
+
+//int32_t setupUdf(SUdfInfo *udf, int32_t numOfUdfs, UdfHandle *handles);
+
+int32_t setupUdf(SUdfInfo* udf, UdfHandle* handle);
+
+int32_t callUdf(UdfHandle handle, int8_t step, char *state, int32_t stateSize, SSDataBlock input, char **newstate,
+                int32_t *newStateSize, SSDataBlock *output);
+
+int32_t teardownUdf(UdfHandle handle);
+
+typedef struct SUdfSetupRequest {
+    char udfName[16]; //
+    int8_t scriptType; // 0:c, 1: lua, 2:js
+    int8_t udfType; //udaf, udf, udtf
+    int16_t pathSize;
+    char *path;
+} SUdfSetupRequest;
+
+typedef struct SUdfSetupResponse {
+    int64_t udfHandle;
+} SUdfSetupResponse;
+
+
+typedef struct SUdfCallRequest {
+    int64_t udfHandle;
+    int8_t step;
+
+    int32_t inputBytes;
+    char *input;
+
+    int32_t stateBytes;
+    char *state;
+} SUdfCallRequest;
+
+
+typedef struct SUdfCallResponse {
+    int32_t outputBytes;
+    char *output;
+    int32_t newStateBytes;
+    char *newState;
+} SUdfCallResponse;
+
+
+typedef struct SUdfTeardownRequest {
+    int64_t udfHandle;
+} SUdfTeardownRequest;
+
+
+typedef struct SUdfTeardownResponse {
+} SUdfTeardownResponse;
+
+typedef struct SUdfRequest {
+    int32_t msgLen;
+    int64_t seqNum;
+
+    int8_t type;
+    void *subReq;
+} SUdfRequest;
+
+typedef struct SUdfResponse {
+    int32_t msgLen;
+    int64_t seqNum;
+
+    int8_t type;
+    int32_t code;
+    void *subRsp;
+} SUdfResponse;
+
+int32_t decodeRequest(char *buf, int32_t bufLen, SUdfRequest **pRequest);
+int32_t encodeResponse(char **buf, int32_t *bufLen, SUdfResponse *response);
+int32_t encodeRequest(char **buf, int32_t *bufLen, SUdfRequest *request);
+int32_t decodeResponse(char *buf, int32_t bufLen, SUdfResponse **pResponse);
+#endif //UDF_UDF_H

source/libs/function/src/builtinsimpl.c

@@ -528,11 +528,10 @@ void firstFunction(SqlFunctionCtx *pCtx) {
   char* buf = GET_ROWCELL_INTERBUF(pResInfo);
 
   SInputColumnInfoData* pInput = &pCtx->input;
-
   SColumnInfoData* pInputCol = pInput->pData[0];
 
   // All null data column, return directly.
-  if (pInput->pColumnDataAgg[0]->numOfNull == pInput->totalRows) {
+  if (pInput->colDataAggIsSet && (pInput->pColumnDataAgg[0]->numOfNull == pInput->totalRows)) {
     ASSERT(pInputCol->hasNull == true);
     return;
   }

source/libs/function/src/udfd.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/>.
+ */
+
+#include "uv.h"
+#include "os.h"
+#include "tlog.h"
+
+#include "tudf.h"
+#include "tudfInt.h"
+
+
+static uv_loop_t *loop;
+
+typedef struct SUdfdUvConn {
+    uv_stream_t *client;
+    char *inputBuf;
+    int32_t inputLen;
+    int32_t inputCap;
+    int32_t inputTotal;
+} SUdfdUvConn;
+
+typedef struct SUvUdfWork {
+    uv_stream_t *client;
+    uv_buf_t input;
+    uv_buf_t output;
+} SUvUdfWork;
+
+typedef struct SUdf {
+    int32_t refCount;
+
+    char name[16];
+    int8_t type;
+
+    uv_lib_t lib;
+    TUdfFunc normalFunc;
+} SUdf;
+
+//TODO: low priority: change name onxxx to xxxCb, and udfc or udfd as prefix
+//TODO: add private udf structure.
+typedef struct SUdfHandle {
+    SUdf *udf;
+} SUdfHandle;
+
+
+void udfdProcessRequest(uv_work_t *req) {
+    SUvUdfWork *uvUdf = (SUvUdfWork *) (req->data);
+    SUdfRequest *request = NULL;
+    decodeRequest(uvUdf->input.base, uvUdf->input.len, &request);
+
+    switch (request->type) {
+        case UDF_TASK_SETUP: {
+            debugPrint("%s", "process setup request");
+            SUdf *udf = malloc(sizeof(SUdf));
+            udf->refCount = 0;
+            SUdfSetupRequest *setup = request->subReq;
+            strcpy(udf->name, setup->udfName);
+            int err = uv_dlopen(setup->path, &udf->lib);
+            if (err != 0) {
+                debugPrint("can not load library %s. error: %s", setup->path, uv_strerror(err));
+                //TODO set error
+            }
+
+            char normalFuncName[32] = {0};
+            strcpy(normalFuncName, setup->udfName);
+	    //TODO error,
+	    //TODO find all functions normal, init, destroy, normal, merge, finalize
+            uv_dlsym(&udf->lib, normalFuncName, (void **) (&udf->normalFunc));
+
+            SUdfHandle *handle = malloc(sizeof(SUdfHandle));
+            handle->udf = udf;
+            udf->refCount++;
+            //TODO: allocate private structure and call init function and set it to handle
+            SUdfResponse *rsp = malloc(sizeof(SUdfResponse));
+            rsp->seqNum = request->seqNum;
+            rsp->type = request->type;
+            rsp->code = 0;
+            SUdfSetupResponse *subRsp = malloc(sizeof(SUdfSetupResponse));
+            subRsp->udfHandle = (int64_t) (handle);
+            rsp->subRsp = subRsp;
+            char *buf;
+            int32_t len;
+            encodeResponse(&buf, &len, rsp);
+
+            uvUdf->output = uv_buf_init(buf, len);
+
+            free(rsp->subRsp);
+            free(rsp);
+            free(request->subReq);
+            free(request);
+            free(uvUdf->input.base);
+            break;
+        }
+
+        case UDF_TASK_CALL: {
+            debugPrint("%s", "process call request");
+            SUdfCallRequest *call = request->subReq;
+            SUdfHandle *handle = (SUdfHandle *) (call->udfHandle);
+            SUdf *udf = handle->udf;
+            char *newState;
+            int32_t newStateSize;
+            SUdfDataBlock input = {.data = call->input, .size= call->inputBytes};
+            SUdfDataBlock output;
+	    //TODO: call different functions according to the step 
+            udf->normalFunc(call->step, call->state, call->stateBytes, input, &newState, &newStateSize, &output);
+
+            SUdfResponse *rsp = malloc(sizeof(SUdfResponse));
+            rsp->seqNum = request->seqNum;
+            rsp->type = request->type;
+            rsp->code = 0;
+            SUdfCallResponse *subRsp = malloc(sizeof(SUdfCallResponse));
+            subRsp->outputBytes = output.size;
+            subRsp->output = output.data;
+            subRsp->newStateBytes = newStateSize;
+            subRsp->newState = newState;
+            rsp->subRsp = subRsp;
+
+            char *buf;
+            int32_t len;
+            encodeResponse(&buf, &len, rsp);
+            uvUdf->output = uv_buf_init(buf, len);
+
+            free(rsp->subRsp);
+            free(rsp);
+            free(newState);
+            free(output.data);
+            free(request->subReq);
+            free(request);
+            free(uvUdf->input.base);
+            break;
+        }
+        case UDF_TASK_TEARDOWN: {
+            debugPrint("%s", "process teardown request");
+
+            SUdfTeardownRequest *teardown = request->subReq;
+            SUdfHandle *handle = (SUdfHandle *) (teardown->udfHandle);
+            SUdf *udf = handle->udf;
+            udf->refCount--;
+            if (udf->refCount == 0) {
+                uv_dlclose(&udf->lib);
+            }
+            free(udf);
+	    //TODO: call destroy and free udf private 
+            free(handle);
+
+            SUdfResponse *rsp = malloc(sizeof(SUdfResponse));
+            rsp->seqNum = request->seqNum;
+            rsp->type = request->type;
+            rsp->code = 0;
+            SUdfTeardownResponse *subRsp = malloc(sizeof(SUdfTeardownResponse));
+            rsp->subRsp = subRsp;
+            char *buf;
+            int32_t len;
+            encodeResponse(&buf, &len, rsp);
+            uvUdf->output = uv_buf_init(buf, len);
+
+            free(rsp->subRsp);
+            free(rsp);
+            free(request->subReq);
+            free(request);
+            free(uvUdf->input.base);
+            break;
+        }
+        default: {
+            break;
+        }
+
+    }
+
+}
+
+void udfdOnWrite(uv_write_t *req, int status) {
+    debugPrint("%s", "after writing to pipe");
+    if (status < 0) {
+        debugPrint("Write error %s", uv_err_name(status));
+    }
+    SUvUdfWork *work = (SUvUdfWork *) req->data;
+    debugPrint("\tlength: %zu", work->output.len);
+    free(work->output.base);
+    free(work);
+    free(req);
+}
+
+
+void udfdSendResponse(uv_work_t *work, int status) {
+    debugPrint("%s", "send response");
+    SUvUdfWork *udfWork = (SUvUdfWork *) (work->data);
+
+    uv_write_t *write_req = malloc(sizeof(uv_write_t));
+    write_req->data = udfWork;
+    uv_write(write_req, udfWork->client, &udfWork->output, 1, udfdOnWrite);
+
+    free(work);
+}
+
+void udfdAllocBuffer(uv_handle_t *handle, size_t suggestedSize, uv_buf_t *buf) {
+    debugPrint("%s", "allocate buffer for read");
+    SUdfdUvConn *ctx = handle->data;
+    int32_t msgHeadSize = sizeof(int32_t) + sizeof(int64_t);
+    if (ctx->inputCap == 0) {
+        ctx->inputBuf = malloc(msgHeadSize);
+        if (ctx->inputBuf) {
+            ctx->inputLen = 0;
+            ctx->inputCap = msgHeadSize;
+            ctx->inputTotal = -1;
+
+            buf->base = ctx->inputBuf;
+            buf->len = ctx->inputCap;
+        } else {
+            //TODO: log error
+            buf->base = NULL;
+            buf->len = 0;
+        }
+    } else {
+        ctx->inputCap = ctx->inputTotal > ctx->inputCap ? ctx->inputTotal : ctx->inputCap;
+        void *inputBuf = realloc(ctx->inputBuf, ctx->inputCap);
+        if (inputBuf) {
+            ctx->inputBuf = inputBuf;
+            buf->base = ctx->inputBuf + ctx->inputLen;
+            buf->len = ctx->inputCap - ctx->inputLen;
+        } else {
+            //TODO: log error
+            buf->base = NULL;
+            buf->len = 0;
+        }
+    }
+    debugPrint("\tinput buf cap - len - total : %d - %d - %d", ctx->inputCap, ctx->inputLen, ctx->inputTotal);
+
+}
+
+bool isUdfdUvMsgComplete(SUdfdUvConn *pipe) {
+    if (pipe->inputTotal == -1 && pipe->inputLen >= sizeof(int32_t)) {
+        pipe->inputTotal = *(int32_t *) (pipe->inputBuf);
+    }
+    if (pipe->inputLen == pipe->inputCap && pipe->inputTotal == pipe->inputCap) {
+        return true;
+    }
+    return false;
+}
+
+void udfdHandleRequest(SUdfdUvConn *conn) {
+    uv_work_t *work = malloc(sizeof(uv_work_t));
+    SUvUdfWork *udfWork = malloc(sizeof(SUvUdfWork));
+    udfWork->client = conn->client;
+    udfWork->input = uv_buf_init(conn->inputBuf, conn->inputLen);
+    conn->inputBuf = NULL;
+    conn->inputLen = 0;
+    conn->inputCap = 0;
+    conn->inputTotal = -1;
+    work->data = udfWork;
+    uv_queue_work(loop, work, udfdProcessRequest, udfdSendResponse);
+}
+
+void udfdPipeCloseCb(uv_handle_t *pipe) {
+    SUdfdUvConn *conn = pipe->data;
+    free(conn->client);
+    free(conn->inputBuf);
+    free(conn);
+}
+
+void udfdUvHandleError(SUdfdUvConn *conn) {
+    uv_close((uv_handle_t *) conn->client, udfdPipeCloseCb);
+}
+
+void udfdPipeRead(uv_stream_t *client, ssize_t nread, const uv_buf_t *buf) {
+    debugPrint("%s, nread: %zd", "read from pipe", nread);
+
+    if (nread == 0) return;
+
+    SUdfdUvConn *conn = client->data;
+
+    if (nread > 0) {
+        conn->inputLen += nread;
+        if (isUdfdUvMsgComplete(conn)) {
+            udfdHandleRequest(conn);
+        } else {
+            //log error or continue;
+        }
+        return;
+    }
+
+    if (nread < 0) {
+        debugPrint("Read error %s", uv_err_name(nread));
+        if (nread == UV_EOF) {
+            //TODO check more when close
+        } else {
+        }
+        udfdUvHandleError(conn);
+    }
+}
+
+void udfdOnNewConnection(uv_stream_t *server, int status) {
+    debugPrint("%s", "on new connection");
+    if (status < 0) {
+        // TODO
+        return;
+    }
+
+    uv_pipe_t *client = (uv_pipe_t *) malloc(sizeof(uv_pipe_t));
+    uv_pipe_init(loop, client, 0);
+    if (uv_accept(server, (uv_stream_t *) client) == 0) {
+        SUdfdUvConn *ctx = malloc(sizeof(SUdfdUvConn));
+        ctx->client = (uv_stream_t *) client;
+        ctx->inputBuf = 0;
+        ctx->inputLen = 0;
+        ctx->inputCap = 0;
+        client->data = ctx;
+        ctx->client = (uv_stream_t *) client;
+        uv_read_start((uv_stream_t *) client, udfdAllocBuffer, udfdPipeRead);
+    } else {
+        uv_close((uv_handle_t *) client, NULL);
+    }
+}
+
+void removeListeningPipe(int sig) {
+    uv_fs_t req;
+    uv_fs_unlink(loop, &req, "udf.sock", NULL);
+    exit(0);
+}
+
+int main() {
+    debugPrint("libuv version: %x", UV_VERSION_HEX);
+
+    loop = uv_default_loop();
+    uv_fs_t req;
+    uv_fs_unlink(loop, &req, "udf.sock", NULL);
+
+    uv_pipe_t server;
+    uv_pipe_init(loop, &server, 0);
+
+    signal(SIGINT, removeListeningPipe);
+
+    int r;
+    if ((r = uv_pipe_bind(&server, "udf.sock"))) {
+        debugPrint("Bind error %s\n", uv_err_name(r));
+        removeListeningPipe(0);
+        return 1;
+    }
+    if ((r = uv_listen((uv_stream_t *) &server, 128, udfdOnNewConnection))) {
+        debugPrint("Listen error %s", uv_err_name(r));
+        return 2;
+    }
+    uv_run(loop, UV_RUN_DEFAULT);
+    uv_loop_close(loop);
+}

source/libs/function/test/runUdf.c

+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "uv.h"
+#include "tudf.h"
+
+int main(int argc, char *argv[]) {
+    startUdfService();
+    uv_sleep(1000);
+    char path[256] = {0};
+    size_t cwdSize = 256;
+    int err = uv_cwd(path, &cwdSize);
+    if (err != 0) {
+        fprintf(stderr, "err cwd: %s\n", uv_strerror(err));
+	    return err;
+    }
+    fprintf(stdout, "current working directory:%s\n", path);
+    strcat(path, "/libudf1.so");
+    SUdfInfo udfInfo = {.udfName="udf1", .path=path};
+
+    UdfHandle handle;
+    setupUdf(&udfInfo, &handle);
+
+    //char state[5000000] = "state";
+    //char input[5000000] = "input";
+    int dataSize = 500;
+    int callCount = 2;
+    if (argc > 1) dataSize = atoi(argv[1]);
+    if (argc > 2) callCount = atoi(argv[2]);
+    char *state = malloc(dataSize);
+    char *input = malloc(dataSize);
+    SUdfDataBlock blockInput = {.data = input, .size = dataSize};
+    SUdfDataBlock blockOutput;
+    char* newState;
+    int32_t newStateSize;
+    for (int l = 0; l < callCount; ++l) {
+        callUdf(handle, 0, state, dataSize, blockInput, &newState, &newStateSize, &blockOutput);
+    }
+    free(state);
+    free(input);
+    teardownUdf(handle);
+
+    stopUdfService();
+}

source/libs/function/test/udf1.c

+#include <string.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+#include "tudf.h"
+
+void udf1(int8_t step, char *state, int32_t stateSize, SUdfDataBlock input,
+          char **newState, int32_t *newStateSize, SUdfDataBlock *output) {
+    fprintf(stdout, "%s, step:%d\n", "udf function called", step);
+    char *newStateBuf = malloc(stateSize);
+    memcpy(newStateBuf, state, stateSize);
+    *newState = newStateBuf;
+    *newStateSize = stateSize;
+
+    char *outputBuf = malloc(input.size);
+    memcpy(outputBuf, input.data, input.size);
+    output->data = outputBuf;
+    output->size = input.size;
+    return;
+}

source/libs/nodes/src/nodesCloneFuncs.c

@@ -141,12 +141,12 @@ static SNode* valueNodeCopy(const SValueNode* pSrc, SValueNode* pDst) {
     case TSDB_DATA_TYPE_NCHAR:
     case TSDB_DATA_TYPE_VARCHAR:
     case TSDB_DATA_TYPE_VARBINARY:
-      pDst->datum.p = malloc(pSrc->node.resType.bytes + VARSTR_HEADER_SIZE);
+      pDst->datum.p = malloc(pSrc->node.resType.bytes + VARSTR_HEADER_SIZE + 1);
       if (NULL == pDst->datum.p) {
         nodesDestroyNode(pDst);
         return NULL;
       }
-      memcpy(pDst->datum.p, pSrc->datum.p, pSrc->node.resType.bytes + VARSTR_HEADER_SIZE);
+      memcpy(pDst->datum.p, pSrc->datum.p, pSrc->node.resType.bytes + VARSTR_HEADER_SIZE + 1);
       break;
     case TSDB_DATA_TYPE_JSON:
     case TSDB_DATA_TYPE_DECIMAL:

source/libs/nodes/src/nodesCodeFuncs.c

@@ -1480,7 +1480,7 @@ static int32_t jsonToDatum(const SJson* pJson, void* pObj) {
     case TSDB_DATA_TYPE_NCHAR:
     case TSDB_DATA_TYPE_VARCHAR:
     case TSDB_DATA_TYPE_VARBINARY: {
-      pNode->datum.p = calloc(1, pNode->node.resType.bytes);
+      pNode->datum.p = calloc(1, pNode->node.resType.bytes + VARSTR_HEADER_SIZE + 1);
       if (NULL == pNode->datum.p) {
         code = TSDB_CODE_OUT_OF_MEMORY;
         break;
@@ -1807,6 +1807,45 @@ static int32_t groupingSetNodeToJson(const void* pObj, SJson* pJson) {
   return code;
 }
 
+static const char* jkIntervalWindowInterval = "Interval";
+static const char* jkIntervalWindowOffset = "Offset";
+static const char* jkIntervalWindowSliding = "Sliding";
+static const char* jkIntervalWindowFill = "Fill";
+
+static int32_t intervalWindowNodeToJson(const void* pObj, SJson* pJson) {
+  const SIntervalWindowNode* pNode = (const SIntervalWindowNode*)pObj;
+
+  int32_t code = tjsonAddObject(pJson, jkIntervalWindowInterval, nodeToJson, pNode->pInterval);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddObject(pJson, jkIntervalWindowOffset, nodeToJson, pNode->pOffset);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddObject(pJson, jkIntervalWindowSliding, nodeToJson, pNode->pSliding);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddObject(pJson, jkIntervalWindowFill, nodeToJson, pNode->pFill);
+  }
+
+  return code;
+}
+
+static int32_t jsonToIntervalWindowNode(const SJson* pJson, void* pObj) {
+  SIntervalWindowNode* pNode = (SIntervalWindowNode*)pObj;
+
+  int32_t code = jsonToNodeObject(pJson, jkIntervalWindowInterval, &pNode->pInterval);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeObject(pJson, jkIntervalWindowOffset, &pNode->pOffset);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeObject(pJson, jkIntervalWindowSliding, &pNode->pSliding);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeObject(pJson, jkIntervalWindowFill, &pNode->pFill);
+  }
+
+  return code;
+}
+
 static const char* jkNodeListDataType = "DataType";
 static const char* jkNodeListNodeList = "NodeList";
 
@@ -2119,8 +2158,9 @@ static int32_t specificNodeToJson(const void* pObj, SJson* pJson) {
     case QUERY_NODE_LIMIT:
     case QUERY_NODE_STATE_WINDOW:
     case QUERY_NODE_SESSION_WINDOW:
-    case QUERY_NODE_INTERVAL_WINDOW:
       break;
+    case QUERY_NODE_INTERVAL_WINDOW:
+      return intervalWindowNodeToJson(pObj, pJson);
     case QUERY_NODE_NODE_LIST:
       return nodeListNodeToJson(pObj, pJson);
     case QUERY_NODE_FILL:
@@ -2222,7 +2262,8 @@ static int32_t jsonToSpecificNode(const SJson* pJson, void* pObj) {
     // case QUERY_NODE_LIMIT:
     // case QUERY_NODE_STATE_WINDOW:
     // case QUERY_NODE_SESSION_WINDOW:
-    // case QUERY_NODE_INTERVAL_WINDOW:
+    case QUERY_NODE_INTERVAL_WINDOW:
+      return jsonToIntervalWindowNode(pJson, pObj);
     case QUERY_NODE_NODE_LIST:
       return jsonToNodeListNode(pJson, pObj);
     // case QUERY_NODE_FILL:

source/libs/parser/inc/parAst.h

@@ -150,9 +150,9 @@ SNode* createDropUserStmt(SAstCreateContext* pCxt, SToken* pUserName);
 SNode* createCreateDnodeStmt(SAstCreateContext* pCxt, const SToken* pFqdn, const SToken* pPort);
 SNode* createDropDnodeStmt(SAstCreateContext* pCxt, const SToken* pDnode);
 SNode* createAlterDnodeStmt(SAstCreateContext* pCxt, const SToken* pDnode, const SToken* pConfig, const SToken* pValue);
-SNode* createCreateIndexStmt(SAstCreateContext* pCxt, EIndexType type, SToken* pIndexName, SToken* pTableName, SNodeList* pCols, SNode* pOptions);
+SNode* createCreateIndexStmt(SAstCreateContext* pCxt, EIndexType type, bool ignoreExists, SToken* pIndexName, SToken* pTableName, SNodeList* pCols, SNode* pOptions);
 SNode* createIndexOption(SAstCreateContext* pCxt, SNodeList* pFuncs, SNode* pInterval, SNode* pOffset, SNode* pSliding);
-SNode* createDropIndexStmt(SAstCreateContext* pCxt, SToken* pIndexName, SToken* pTableName);
+SNode* createDropIndexStmt(SAstCreateContext* pCxt, bool ignoreNotExists, SToken* pIndexName, SToken* pTableName);
 SNode* createCreateQnodeStmt(SAstCreateContext* pCxt, const SToken* pDnodeId);
 SNode* createDropQnodeStmt(SAstCreateContext* pCxt, const SToken* pDnodeId);
 SNode* createCreateTopicStmt(SAstCreateContext* pCxt, bool ignoreExists, const SToken* pTopicName, SNode* pQuery, const SToken* pSubscribeDbName);

source/libs/parser/inc/parInsertData.h

@@ -118,7 +118,7 @@ static FORCE_INLINE void getMemRowAppendInfo(SSchema *pSchema, uint8_t rowType,
 }
 
 static FORCE_INLINE int32_t setBlockInfo(SSubmitBlk *pBlocks, STableDataBlocks* dataBuf, int32_t numOfRows) {
-  pBlocks->tid = dataBuf->pTableMeta->suid;
+  pBlocks->suid = (TSDB_NORMAL_TABLE == dataBuf->pTableMeta->tableType ? dataBuf->pTableMeta->uid : dataBuf->pTableMeta->suid);
   pBlocks->uid = dataBuf->pTableMeta->uid;
   pBlocks->sversion = dataBuf->pTableMeta->sversion;
 

source/libs/parser/inc/sql.y

@@ -313,10 +313,11 @@ func_name_list(A) ::= func_name_list(B) NK_COMMA col_name(C).
 func_name(A) ::= function_name(B).                                                { A = createFunctionNode(pCxt, &B, NULL); }
 
 /************************************************ create index ********************************************************/
-cmd ::= CREATE SMA INDEX index_name(A) ON table_name(B) index_options(C).         { pCxt->pRootNode = createCreateIndexStmt(pCxt, INDEX_TYPE_SMA, &A, &B, NULL, C); }
-cmd ::= CREATE FULLTEXT INDEX
-  index_name(A) ON table_name(B) NK_LP col_name_list(C) NK_RP.                    { pCxt->pRootNode = createCreateIndexStmt(pCxt, INDEX_TYPE_FULLTEXT, &A, &B, C, NULL); }
-cmd ::= DROP INDEX index_name(A) ON table_name(B).                                { pCxt->pRootNode = createDropIndexStmt(pCxt, &A, &B); }
+cmd ::= CREATE SMA INDEX not_exists_opt(D) 
+  index_name(A) ON table_name(B) index_options(C).                                { pCxt->pRootNode = createCreateIndexStmt(pCxt, INDEX_TYPE_SMA, D, &A, &B, NULL, C); }
+cmd ::= CREATE FULLTEXT INDEX not_exists_opt(D)
+  index_name(A) ON table_name(B) NK_LP col_name_list(C) NK_RP.                    { pCxt->pRootNode = createCreateIndexStmt(pCxt, INDEX_TYPE_FULLTEXT, D, &A, &B, C, NULL); }
+cmd ::= DROP INDEX exists_opt(C) index_name(A) ON table_name(B).                  { pCxt->pRootNode = createDropIndexStmt(pCxt, C, &A, &B); }
 
 index_options(A) ::= .                                                            { A = NULL; }
 index_options(A) ::= FUNCTION NK_LP func_list(B) NK_RP INTERVAL 

source/libs/parser/src/parAstCreater.c

@@ -1160,13 +1160,14 @@ SNode* createAlterDnodeStmt(SAstCreateContext* pCxt, const SToken* pDnode, const
   return (SNode*)pStmt;
 }
 
-SNode* createCreateIndexStmt(SAstCreateContext* pCxt, EIndexType type, SToken* pIndexName, SToken* pTableName, SNodeList* pCols, SNode* pOptions) {
+SNode* createCreateIndexStmt(SAstCreateContext* pCxt, EIndexType type, bool ignoreExists, SToken* pIndexName, SToken* pTableName, SNodeList* pCols, SNode* pOptions) {
   if (!checkIndexName(pCxt, pIndexName) || !checkTableName(pCxt, pTableName)) {
     return NULL;
   }
   SCreateIndexStmt* pStmt = nodesMakeNode(QUERY_NODE_CREATE_INDEX_STMT);
   CHECK_OUT_OF_MEM(pStmt);
   pStmt->indexType = type;
+  pStmt->ignoreExists = ignoreExists;
   strncpy(pStmt->indexName, pIndexName->z, pIndexName->n);
   strncpy(pStmt->tableName, pTableName->z, pTableName->n);
   pStmt->pCols = pCols;
@@ -1184,12 +1185,13 @@ SNode* createIndexOption(SAstCreateContext* pCxt, SNodeList* pFuncs, SNode* pInt
   return (SNode*)pOptions;
 }
 
-SNode* createDropIndexStmt(SAstCreateContext* pCxt, SToken* pIndexName, SToken* pTableName) {
+SNode* createDropIndexStmt(SAstCreateContext* pCxt, bool ignoreNotExists, SToken* pIndexName, SToken* pTableName) {
   if (!checkIndexName(pCxt, pIndexName) || !checkTableName(pCxt, pTableName)) {
     return NULL;
   }
   SDropIndexStmt* pStmt = nodesMakeNode(QUERY_NODE_DROP_INDEX_STMT);
   CHECK_OUT_OF_MEM(pStmt);
+  pStmt->ignoreNotExists = ignoreNotExists;
   strncpy(pStmt->indexName, pIndexName->z, pIndexName->n);
   strncpy(pStmt->tableName, pTableName->z, pTableName->n);
   return (SNode*)pStmt;

source/libs/parser/src/parInsert.c

@@ -48,11 +48,6 @@
     } \
   } while (0)
 
-enum {
-  TSDB_USE_SERVER_TS = 0,
-  TSDB_USE_CLI_TS = 1,
-};
-
 typedef struct SInsertParseContext {
   SParseContext* pComCxt;       // input
   char          *pSql;          // input
@@ -264,7 +259,7 @@ static void buildMsgHeader(STableDataBlocks* src, SVgDataBlocks* blocks) {
     while (numOfBlocks--) {
       int32_t dataLen = blk->dataLen;
       blk->uid = htobe64(blk->uid);
-      blk->tid = htonl(blk->tid);
+      blk->suid = htobe64(blk->suid);
       blk->padding = htonl(blk->padding);
       blk->sversion = htonl(blk->sversion);
       blk->dataLen = htonl(blk->dataLen);
@@ -303,20 +298,7 @@ static int32_t checkTimestamp(STableDataBlocks *pDataBlocks, const char *start)
   }
 
   TSKEY k = *(TSKEY *)start;
-
-  if (k == INT64_MIN) {
-    if (pDataBlocks->tsSource == TSDB_USE_CLI_TS) {
-      return TSDB_CODE_FAILED; // client time/server time can not be mixed
-    }
-    pDataBlocks->tsSource = TSDB_USE_SERVER_TS;
-  } else {
-    if (pDataBlocks->tsSource == TSDB_USE_SERVER_TS) {
-      return TSDB_CODE_FAILED;  // client time/server time can not be mixed
-    }
-    pDataBlocks->tsSource = TSDB_USE_CLI_TS;
-  }
-
-  if (k <= pDataBlocks->prevTS && (pDataBlocks->tsSource == TSDB_USE_CLI_TS)) {
+  if (k <= pDataBlocks->prevTS) {
     pDataBlocks->ordered = false;
   }
 

source/libs/parser/src/parInsertData.c

@@ -141,7 +141,6 @@ static int32_t createDataBlock(size_t defaultSize, int32_t rowSize, int32_t star
   dataBuf->prevTS   = INT64_MIN;
   dataBuf->rowSize  = rowSize;
   dataBuf->size     = startOffset;
-  dataBuf->tsSource = -1;
   dataBuf->vgId     = dataBuf->pTableMeta->vgId;
 
   assert(defaultSize > 0 && pTableMeta != NULL && dataBuf->pTableMeta != NULL);

source/libs/parser/src/parTranslater.c

@@ -383,7 +383,7 @@ static EDealRes translateValue(STranslateContext* pCxt, SValueNode* pVal) {
       case TSDB_DATA_TYPE_NCHAR:
       case TSDB_DATA_TYPE_VARCHAR:
       case TSDB_DATA_TYPE_VARBINARY: {
-        pVal->datum.p = calloc(1, pVal->node.resType.bytes + VARSTR_HEADER_SIZE);
+        pVal->datum.p = calloc(1, pVal->node.resType.bytes + VARSTR_HEADER_SIZE + 1);
         if (NULL == pVal->datum.p) {
           return generateDealNodeErrMsg(pCxt, TSDB_CODE_OUT_OF_MEMORY);
         }
@@ -1406,34 +1406,110 @@ static int32_t translateShowTables(STranslateContext* pCxt) {
   return TSDB_CODE_SUCCESS;
 }
 
-static int32_t translateCreateSmaIndex(STranslateContext* pCxt, SCreateIndexStmt* pStmt) {
-  SVCreateTSmaReq createSmaReq = {0};
+static int32_t getSmaIndexDstVgId(STranslateContext* pCxt, char* pTableName, int32_t* pVgId) {
+  SVgroupInfo vg = {0};
+  int32_t code = getTableHashVgroup(pCxt, pCxt->pParseCxt->db, pTableName, &vg);
+  if (TSDB_CODE_SUCCESS == code) {
+    *pVgId = vg.vgId;
+  }
+  return code;
+}
 
-  if (DEAL_RES_ERROR == translateValue(pCxt, (SValueNode*)pStmt->pOptions->pInterval) ||
-      (NULL != pStmt->pOptions->pOffset && DEAL_RES_ERROR == translateValue(pCxt, (SValueNode*)pStmt->pOptions->pOffset)) ||
-      (NULL != pStmt->pOptions->pSliding && DEAL_RES_ERROR == translateValue(pCxt, (SValueNode*)pStmt->pOptions->pSliding))) {
-    return pCxt->errCode;
+static int32_t getSmaIndexSql(STranslateContext* pCxt, char** pSql, int32_t* pLen) {
+  *pSql = strdup(pCxt->pParseCxt->pSql);
+  if (NULL == *pSql) {
+    return TSDB_CODE_OUT_OF_MEMORY;
   }
+  *pLen = pCxt->pParseCxt->sqlLen + 1;
+  return TSDB_CODE_SUCCESS;
+}
 
-  createSmaReq.tSma.intervalUnit = ((SValueNode*)pStmt->pOptions->pInterval)->unit;
-  createSmaReq.tSma.slidingUnit = (NULL != pStmt->pOptions->pSliding ? ((SValueNode*)pStmt->pOptions->pSliding)->unit : 0);
-  strcpy(createSmaReq.tSma.indexName, pStmt->indexName);
+static int32_t getSmaIndexExpr(STranslateContext* pCxt, SCreateIndexStmt* pStmt, char** pExpr, int32_t* pLen) {
+  return nodesListToString(pStmt->pOptions->pFuncs, false, pExpr, pLen);
+}
 
-  SName name;
-  name.type = TSDB_TABLE_NAME_T;
-  name.acctId = pCxt->pParseCxt->acctId;
+static int32_t getSmaIndexBuildAst(STranslateContext* pCxt, SCreateIndexStmt* pStmt, char** pAst, int32_t* pLen) {
+  SSelectStmt* pSelect = nodesMakeNode(QUERY_NODE_SELECT_STMT);
+  if (NULL == pSelect) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  SRealTableNode* pTable = nodesMakeNode(QUERY_NODE_REAL_TABLE);
+  if (NULL == pTable) {
+    nodesDestroyNode(pSelect);
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+  strcpy(pTable->table.dbName, pCxt->pParseCxt->db);
+  strcpy(pTable->table.tableName, pStmt->tableName);
+  pSelect->pFromTable = (SNode*)pTable;
+
+  pSelect->pProjectionList = nodesCloneList(pStmt->pOptions->pFuncs);
+  if (NULL == pTable) {
+    nodesDestroyNode(pSelect);
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  SIntervalWindowNode* pInterval = nodesMakeNode(QUERY_NODE_INTERVAL_WINDOW);
+  if (NULL == pInterval) {
+    nodesDestroyNode(pSelect);
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+  pSelect->pWindow = (SNode*)pInterval;
+  pInterval->pInterval = nodesCloneNode(pStmt->pOptions->pInterval);
+  pInterval->pOffset = nodesCloneNode(pStmt->pOptions->pOffset);
+  pInterval->pSliding = nodesCloneNode(pStmt->pOptions->pSliding);
+  if (NULL == pInterval->pInterval || (NULL != pStmt->pOptions->pOffset && NULL == pInterval->pOffset) ||
+      (NULL != pStmt->pOptions->pSliding && NULL == pInterval->pSliding)) {
+    nodesDestroyNode(pSelect);
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  int32_t code = translateQuery(pCxt, (SNode*)pSelect);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = nodesNodeToString(pSelect, false, pAst, pLen);
+  }
+  nodesDestroyNode(pSelect);
+  return code;
+}
+
+static int32_t buildCreateSmaReq(STranslateContext* pCxt, SCreateIndexStmt* pStmt, SMCreateSmaReq* pReq) {
+  SName name = { .type = TSDB_TABLE_NAME_T, .acctId = pCxt->pParseCxt->acctId };
   strcpy(name.dbname, pCxt->pParseCxt->db);
+  strcpy(name.tname, pStmt->indexName);
+  tNameExtractFullName(&name, pReq->name);
   strcpy(name.tname, pStmt->tableName);
-  STableMeta* pMeta = NULL;
-  int32_t code = catalogGetTableMeta(pCxt->pParseCxt->pCatalog, pCxt->pParseCxt->pTransporter, &pCxt->pParseCxt->mgmtEpSet, &name, &pMeta);
-  if (TSDB_CODE_SUCCESS != code) {
-    return code;
+  name.tname[strlen(pStmt->tableName)] = '\0';
+  tNameExtractFullName(&name, pReq->stb);
+  pReq->igExists = pStmt->ignoreExists;
+  pReq->interval = ((SValueNode*)pStmt->pOptions->pInterval)->datum.i;
+  pReq->intervalUnit = ((SValueNode*)pStmt->pOptions->pInterval)->unit;
+  pReq->offset = (NULL != pStmt->pOptions->pOffset ? ((SValueNode*)pStmt->pOptions->pOffset)->datum.i : 0);
+  pReq->sliding = (NULL != pStmt->pOptions->pSliding ? ((SValueNode*)pStmt->pOptions->pSliding)->datum.i : pReq->interval);
+  pReq->slidingUnit = (NULL != pStmt->pOptions->pSliding ? ((SValueNode*)pStmt->pOptions->pSliding)->unit : pReq->intervalUnit);
+
+  int32_t code = getSmaIndexDstVgId(pCxt, pStmt->tableName, &pReq->dstVgId);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = getSmaIndexSql(pCxt, &pReq->sql, &pReq->sqlLen);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = getSmaIndexExpr(pCxt, pStmt, &pReq->expr, &pReq->exprLen);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = getSmaIndexBuildAst(pCxt, pStmt, &pReq->ast, &pReq->astLen);
+  }
+
+  return code;
+}
+
+static int32_t translateCreateSmaIndex(STranslateContext* pCxt, SCreateIndexStmt* pStmt) {
+  if (DEAL_RES_ERROR == translateValue(pCxt, (SValueNode*)pStmt->pOptions->pInterval) ||
+      (NULL != pStmt->pOptions->pOffset && DEAL_RES_ERROR == translateValue(pCxt, (SValueNode*)pStmt->pOptions->pOffset)) ||
+      (NULL != pStmt->pOptions->pSliding && DEAL_RES_ERROR == translateValue(pCxt, (SValueNode*)pStmt->pOptions->pSliding))) {
+    return pCxt->errCode;
   }
 
-  createSmaReq.tSma.tableUid = pMeta->uid;
-  createSmaReq.tSma.interval = ((SValueNode*)pStmt->pOptions->pInterval)->datum.i;
-  createSmaReq.tSma.sliding = (NULL != pStmt->pOptions->pSliding ? ((SValueNode*)pStmt->pOptions->pSliding)->datum.i : 0);
-  code = nodesListToString(pStmt->pOptions->pFuncs, false, &createSmaReq.tSma.expr, &createSmaReq.tSma.exprLen);
+  SMCreateSmaReq createSmaReq = {0};
+  int32_t code = buildCreateSmaReq(pCxt, pStmt, &createSmaReq);
   if (TSDB_CODE_SUCCESS != code) {
     return code;
   }
@@ -1443,15 +1519,14 @@ static int32_t translateCreateSmaIndex(STranslateContext* pCxt, SCreateIndexStmt
     return TSDB_CODE_OUT_OF_MEMORY;
   }
   pCxt->pCmdMsg->epSet = pCxt->pParseCxt->mgmtEpSet;
-  pCxt->pCmdMsg->msgType = TDMT_VND_CREATE_SMA;
-  pCxt->pCmdMsg->msgLen = tSerializeSVCreateTSmaReq(NULL, &createSmaReq);
+  pCxt->pCmdMsg->msgType = TDMT_MND_CREATE_SMA;
+  pCxt->pCmdMsg->msgLen = tSerializeSMCreateSmaReq(NULL, 0, &createSmaReq);
   pCxt->pCmdMsg->pMsg = malloc(pCxt->pCmdMsg->msgLen);
   if (NULL == pCxt->pCmdMsg->pMsg) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
-  void* pBuf = pCxt->pCmdMsg->pMsg;
-  tSerializeSVCreateTSmaReq(&pBuf, &createSmaReq);
-  tdDestroyTSma(&createSmaReq.tSma);
+  tSerializeSMCreateSmaReq(pCxt->pCmdMsg->pMsg, pCxt->pCmdMsg->msgLen, &createSmaReq);
+  tFreeSMCreateSmaReq(&createSmaReq);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -1696,24 +1771,27 @@ static int32_t translateSubquery(STranslateContext* pCxt, SNode* pNode) {
   return code;
 }
 
-static int32_t setReslutSchema(STranslateContext* pCxt, SQuery* pQuery) {
-  if (QUERY_NODE_SELECT_STMT == nodeType(pQuery->pRoot)) {
-    SSelectStmt* pSelect = (SSelectStmt*)pQuery->pRoot;
-    pQuery->numOfResCols = LIST_LENGTH(pSelect->pProjectionList);
-    pQuery->pResSchema = calloc(pQuery->numOfResCols, sizeof(SSchema));
-    if (NULL == pQuery->pResSchema) {
-      return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_OUT_OF_MEMORY);
+int32_t qExtractResultSchema(const SNode* pRoot, int32_t* numOfCols, SSchema** pSchema) {
+  if (QUERY_NODE_SELECT_STMT == nodeType(pRoot)) {
+    SSelectStmt* pSelect = (SSelectStmt*) pRoot;
+    *numOfCols = LIST_LENGTH(pSelect->pProjectionList);
+    *pSchema = calloc((*numOfCols), sizeof(SSchema));
+    if (NULL == (*pSchema)) {
+      return TSDB_CODE_OUT_OF_MEMORY;
     }
+
     SNode* pNode;
     int32_t index = 0;
     FOREACH(pNode, pSelect->pProjectionList) {
       SExprNode* pExpr = (SExprNode*)pNode;
-      pQuery->pResSchema[index].type = pExpr->resType.type;
-      pQuery->pResSchema[index].bytes = pExpr->resType.bytes;
-      strcpy(pQuery->pResSchema[index].name, pExpr->aliasName);
+      (*pSchema)[index].type = pExpr->resType.type;
+      (*pSchema)[index].bytes = pExpr->resType.bytes;
+      (*pSchema)[index].colId = index + 1;
+      strcpy((*pSchema)[index].name, pExpr->aliasName);
       index +=1;
     }
   }
+
   return TSDB_CODE_SUCCESS;
 }
 
@@ -2297,7 +2375,7 @@ static int32_t setQuery(STranslateContext* pCxt, SQuery* pQuery) {
       pQuery->haveResultSet = true;
       pQuery->directRpc = false;
       pQuery->msgType = TDMT_VND_QUERY;
-      code = setReslutSchema(pCxt, pQuery);
+      code = qExtractResultSchema(pQuery->pRoot, &pQuery->numOfResCols, &pQuery->pResSchema);
       break;
     case QUERY_NODE_VNODE_MODIF_STMT:
       pQuery->haveResultSet = false;

source/libs/parser/test/parserAstTest.cpp

@@ -358,7 +358,6 @@ TEST_F(ParserTest, selectSemanticError) {
   ASSERT_TRUE(run(TSDB_CODE_SUCCESS, TSDB_CODE_PAR_NOT_SELECTED_EXPRESSION));
 }
 
-
 TEST_F(ParserTest, showUsers) {
   setDatabase("root", "test");
 
@@ -366,8 +365,6 @@ TEST_F(ParserTest, showUsers) {
   ASSERT_TRUE(run());
 }
 
-
-
 TEST_F(ParserTest, createDnode) {
   setDatabase("root", "test");
 

source/libs/parser/test/parserInsertTest.cpp

@@ -72,7 +72,7 @@ protected:
       SSubmitBlk* blk = (SSubmitBlk*)(submit + 1);
       for (int32_t i = 0; i < numOfBlocks; ++i) {
         cout << "Block:" << i << endl;
-        cout << "\tuid:" << be64toh(blk->uid) << ", tid:" << ntohl(blk->tid) << ", padding:" << ntohl(blk->padding) << ", sversion:" << ntohl(blk->sversion)
+        cout << "\tuid:" << be64toh(blk->uid) << ", tid:" << be64toh(blk->suid) << ", padding:" << ntohl(blk->padding) << ", sversion:" << ntohl(blk->sversion)
             << ", dataLen:" << ntohl(blk->dataLen) << ", schemaLen:" << ntohl(blk->schemaLen) << ", numOfRows:" << ntohs(blk->numOfRows) << endl;
         blk = (SSubmitBlk*)(blk->data + ntohl(blk->dataLen));
       }

source/libs/planner/src/planPhysiCreater.c

@@ -213,6 +213,33 @@ static SNodeptr createPrimaryKeyCol(SPhysiPlanContext* pCxt, uint64_t tableId) {
   return pCol;
 }
 
+static int32_t colIdCompare(const void* pLeft, const void* pRight) {
+  SColumnNode* pLeftCol = *(SColumnNode**)pLeft;
+  SColumnNode* pRightCol = *(SColumnNode**)pRight;
+  return pLeftCol->colId > pRightCol->colId ? 1 : -1;
+}
+
+static int32_t sortScanCols(SNodeList* pScanCols) {
+  SArray* pArray = taosArrayInit(LIST_LENGTH(pScanCols), POINTER_BYTES);
+  if (NULL == pArray) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  SNode* pCol = NULL;
+  FOREACH(pCol, pScanCols) {
+    taosArrayPush(pArray, &pCol);
+  }
+  taosArraySort(pArray, colIdCompare);
+
+  int32_t index = 0;
+  FOREACH(pCol, pScanCols) {
+    REPLACE_NODE(taosArrayGetP(pArray, index++));
+  }
+  taosArrayDestroy(pArray);
+
+  return TSDB_CODE_SUCCESS;
+}
+
 static int32_t createScanCols(SPhysiPlanContext* pCxt, SScanPhysiNode* pScanPhysiNode, SNodeList* pScanCols) {
   if (QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN == nodeType(pScanPhysiNode)
       || QUERY_NODE_PHYSICAL_PLAN_TABLE_SEQ_SCAN == nodeType(pScanPhysiNode)) {
@@ -235,6 +262,7 @@ static int32_t createScanCols(SPhysiPlanContext* pCxt, SScanPhysiNode* pScanPhys
     CHECK_ALLOC(pScanPhysiNode->pScanCols, TSDB_CODE_OUT_OF_MEMORY);
   }
 
+  // return sortScanCols(pScanPhysiNode->pScanCols);
   return TSDB_CODE_SUCCESS;
 }
 

source/libs/planner/test/plannerTest.cpp

@@ -52,7 +52,11 @@ protected:
     const string syntaxTreeStr = toString(query_->pRoot, false);
   
     SLogicNode* pLogicNode = nullptr;
-    SPlanContext cxt = { .queryId = 1, .acctId = 0, .streamQuery = streamQuery };
+    SPlanContext cxt = {0};
+    cxt.queryId = 1;
+    cxt.acctId = 0;
+    cxt.streamQuery = streamQuery;
+
     setPlanContext(query_, &cxt);
     code = createLogicPlan(&cxt, &pLogicNode);
     if (code != TSDB_CODE_SUCCESS) {
@@ -108,6 +112,11 @@ private:
     if (QUERY_NODE_CREATE_TOPIC_STMT == nodeType(pQuery->pRoot)) {
       pCxt->pAstRoot = ((SCreateTopicStmt*)pQuery->pRoot)->pQuery;
       pCxt->topicQuery = true;
+    } else if (QUERY_NODE_CREATE_INDEX_STMT == nodeType(pQuery->pRoot)) {
+      SMCreateSmaReq req = {0};
+      tDeserializeSMCreateSmaReq(pQuery->pCmdMsg->pMsg, pQuery->pCmdMsg->msgLen, &req);
+      nodesStringToNode(req.ast, &pCxt->pAstRoot);
+      pCxt->streamQuery = true;
     } else {
       pCxt->pAstRoot = pQuery->pRoot;
     }
@@ -211,3 +220,10 @@ TEST_F(PlannerTest, stream) {
   bind("SELECT sum(c1) FROM st1");
   ASSERT_TRUE(run(true));
 }
+
+TEST_F(PlannerTest, createSmaIndex) {
+  setDatabase("root", "test");
+
+  bind("create sma index index1 on t1 function(max(c1), min(c3 + 10), sum(c4)) INTERVAL(10s)");
+  ASSERT_TRUE(run());
+}

source/libs/sync/inc/syncRaftEntry.h

@@ -27,6 +27,12 @@ extern "C" {
 #include "syncMessage.h"
 #include "taosdef.h"
 
+typedef enum EntryType {
+  SYNC_RAFT_ENTRY_NOOP = 0,
+  SYNC_RAFT_ENTRY_DATA = 1,
+  SYNC_RAFT_ENTRY_CONFIG = 2,
+} EntryType;
+
 typedef struct SSyncRaftEntry {
   uint32_t  bytes;
   uint32_t  msgType;
@@ -35,12 +41,15 @@ typedef struct SSyncRaftEntry {
   bool      isWeak;
   SyncTerm  term;
   SyncIndex index;
+  EntryType entryType;
   uint32_t  dataLen;
   char      data[];
 } SSyncRaftEntry;
 
 SSyncRaftEntry* syncEntryBuild(uint32_t dataLen);
 SSyncRaftEntry* syncEntryBuild2(SyncClientRequest* pMsg, SyncTerm term, SyncIndex index);  // step 4
+SSyncRaftEntry* syncEntryBuild3(SyncClientRequest* pMsg, SyncTerm term, SyncIndex index, EntryType entryType);
+SSyncRaftEntry* syncEntryBuildNoop(SyncTerm term, SyncIndex index);
 void            syncEntryDestory(SSyncRaftEntry* pEntry);
 char*           syncEntrySerialize(const SSyncRaftEntry* pEntry, uint32_t* len);  // step 5
 SSyncRaftEntry* syncEntryDeserialize(const char* buf, uint32_t len);              // step 6

source/libs/sync/src/syncAppendEntries.c

@@ -199,6 +199,9 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
             SSyncRaftEntry* pRollBackEntry = logStoreGetEntry(ths->pLogStore, index);
             assert(pRollBackEntry != NULL);
 
+            // maybe is a NOOP ENTRY
+            // assert(pRollBackEntry->entryType == SYNC_RAFT_ENTRY_DATA);
+
             SRpcMsg rpcMsg;
             syncEntry2OriginalRpc(pRollBackEntry, &rpcMsg);
             ths->pFsm->FpRollBackCb(ths->pFsm, &rpcMsg, pRollBackEntry->index, pRollBackEntry->isWeak, 0, ths->state);
@@ -217,7 +220,7 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
         SRpcMsg rpcMsg;
         syncEntry2OriginalRpc(pAppendEntry, &rpcMsg);
         if (ths->pFsm != NULL) {
-          if (ths->pFsm->FpPreCommitCb != NULL) {
+          if (ths->pFsm->FpPreCommitCb != NULL && pAppendEntry->entryType == SYNC_RAFT_ENTRY_DATA) {
             ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pAppendEntry->index, pAppendEntry->isWeak, 2, ths->state);
           }
         }
@@ -242,7 +245,7 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
       SRpcMsg rpcMsg;
       syncEntry2OriginalRpc(pAppendEntry, &rpcMsg);
       if (ths->pFsm != NULL) {
-        if (ths->pFsm->FpPreCommitCb != NULL) {
+        if (ths->pFsm->FpPreCommitCb != NULL && pAppendEntry->entryType == SYNC_RAFT_ENTRY_DATA) {
           ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pAppendEntry->index, pAppendEntry->isWeak, 3, ths->state);
         }
       }
@@ -298,7 +301,7 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
               SRpcMsg rpcMsg;
               syncEntry2OriginalRpc(pEntry, &rpcMsg);
 
-              if (ths->pFsm->FpCommitCb != NULL) {
+              if (ths->pFsm->FpCommitCb != NULL && pEntry->entryType == SYNC_RAFT_ENTRY_DATA) {
                 ths->pFsm->FpCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0, ths->state);
               }
 

source/libs/sync/src/syncCommit.c

@@ -97,7 +97,7 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
           SRpcMsg rpcMsg;
           syncEntry2OriginalRpc(pEntry, &rpcMsg);
 
-          if (pSyncNode->pFsm->FpCommitCb != NULL) {
+          if (pSyncNode->pFsm->FpCommitCb != NULL && pEntry->entryType == SYNC_RAFT_ENTRY_DATA) {
             pSyncNode->pFsm->FpCommitCb(pSyncNode->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0, pSyncNode->state);
           }
 

source/libs/sync/src/syncIO.c

@@ -29,7 +29,7 @@ static int32_t  syncIODestroy(SSyncIO *io);
 static int32_t  syncIOStartInternal(SSyncIO *io);
 static int32_t  syncIOStopInternal(SSyncIO *io);
 
-static void *  syncIOConsumerFunc(void *param);
+static void   *syncIOConsumerFunc(void *param);
 static void    syncIOProcessRequest(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet);
 static void    syncIOProcessReply(void *pParent, SRpcMsg *pMsg, SEpSet *pEpSet);
 static int32_t syncIOAuth(void *parent, char *meterId, char *spi, char *encrypt, char *secret, char *ckey);
@@ -75,6 +75,7 @@ int32_t syncIOSendMsg(void *clientRpc, const SEpSet *pEpSet, SRpcMsg *pMsg) {
   syncRpcMsgPrint2(logBuf, pMsg);
 
   pMsg->handle = NULL;
+  pMsg->noResp = 1;
   rpcSendRequest(clientRpc, pEpSet, pMsg, NULL);
   return ret;
 }
@@ -234,9 +235,9 @@ static int32_t syncIOStopInternal(SSyncIO *io) {
 }
 
 static void *syncIOConsumerFunc(void *param) {
-  SSyncIO *  io = param;
+  SSyncIO   *io = param;
   STaosQall *qall;
-  SRpcMsg *  pRpcMsg, rpcMsg;
+  SRpcMsg   *pRpcMsg, rpcMsg;
   qall = taosAllocateQall();
 
   while (1) {
@@ -324,19 +325,21 @@ static void *syncIOConsumerFunc(void *param) {
       taosGetQitem(qall, (void **)&pRpcMsg);
       rpcFreeCont(pRpcMsg->pCont);
 
-      if (pRpcMsg->handle != NULL) {
-        int msgSize = 32;
-        memset(&rpcMsg, 0, sizeof(rpcMsg));
-        rpcMsg.msgType = SYNC_RESPONSE;
-        rpcMsg.pCont = rpcMallocCont(msgSize);
-        rpcMsg.contLen = msgSize;
-        snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", "give a reply");
-        rpcMsg.handle = pRpcMsg->handle;
-        rpcMsg.code = 0;
-
-        syncRpcMsgPrint2((char *)"syncIOConsumerFunc rpcSendResponse --> ", &rpcMsg);
-        rpcSendResponse(&rpcMsg);
-      }
+      /*
+            if (pRpcMsg->handle != NULL) {
+              int msgSize = 32;
+              memset(&rpcMsg, 0, sizeof(rpcMsg));
+              rpcMsg.msgType = SYNC_RESPONSE;
+              rpcMsg.pCont = rpcMallocCont(msgSize);
+              rpcMsg.contLen = msgSize;
+              snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", "give a reply");
+              rpcMsg.handle = pRpcMsg->handle;
+              rpcMsg.code = 0;
+
+              syncRpcMsgPrint2((char *)"syncIOConsumerFunc rpcSendResponse --> ", &rpcMsg);
+              rpcSendResponse(&rpcMsg);
+            }
+      */
 
       taosFreeQitem(pRpcMsg);
     }

source/libs/sync/src/syncMain.c

@@ -37,11 +37,13 @@ static int32_t tsNodeRefId = -1;
 
 // ------ local funciton ---------
 // enqueue message ----
-static void syncNodeEqPingTimer(void* param, void* tmrId);
-static void syncNodeEqElectTimer(void* param, void* tmrId);
-static void syncNodeEqHeartbeatTimer(void* param, void* tmrId);
+static void    syncNodeEqPingTimer(void* param, void* tmrId);
+static void    syncNodeEqElectTimer(void* param, void* tmrId);
+static void    syncNodeEqHeartbeatTimer(void* param, void* tmrId);
+static int32_t syncNodeEqNoop(SSyncNode* ths);
+static int32_t syncNodeAppendNoop(SSyncNode* ths);
 
-// on message ----
+// process message ----
 static int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg);
 static int32_t syncNodeOnPingReplyCb(SSyncNode* ths, SyncPingReply* pMsg);
 static int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg);
@@ -669,6 +671,10 @@ void syncNodeCandidate2Leader(SSyncNode* pSyncNode) {
   assert(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
   assert(voteGrantedMajority(pSyncNode->pVotesGranted));
   syncNodeBecomeLeader(pSyncNode);
+
+  // Raft 3.6.2 Committing entries from previous terms
+  syncNodeAppendNoop(pSyncNode);
+  // syncNodeEqNoop(pSyncNode);
 }
 
 void syncNodeFollower2Candidate(SSyncNode* pSyncNode) {
@@ -803,6 +809,47 @@ static void syncNodeEqHeartbeatTimer(void* param, void* tmrId) {
   }
 }
 
+static int32_t syncNodeEqNoop(SSyncNode* ths) {
+  int32_t ret = 0;
+  assert(ths->state == TAOS_SYNC_STATE_LEADER);
+
+  SyncIndex       index = ths->pLogStore->getLastIndex(ths->pLogStore) + 1;
+  SyncTerm        term = ths->pRaftStore->currentTerm;
+  SSyncRaftEntry* pEntry = syncEntryBuildNoop(term, index);
+  assert(pEntry != NULL);
+
+  uint32_t           entryLen;
+  char*              serialized = syncEntrySerialize(pEntry, &entryLen);
+  SyncClientRequest* pSyncMsg = syncClientRequestBuild(entryLen);
+  assert(pSyncMsg->dataLen == entryLen);
+  memcpy(pSyncMsg->data, serialized, entryLen);
+
+  SRpcMsg rpcMsg;
+  syncClientRequest2RpcMsg(pSyncMsg, &rpcMsg);
+  ths->FpEqMsg(ths->queue, &rpcMsg);
+
+  free(serialized);
+  syncClientRequestDestroy(pSyncMsg);
+
+  return ret;
+}
+
+static int32_t syncNodeAppendNoop(SSyncNode* ths) {
+  int32_t ret = 0;
+
+  SyncIndex       index = ths->pLogStore->getLastIndex(ths->pLogStore) + 1;
+  SyncTerm        term = ths->pRaftStore->currentTerm;
+  SSyncRaftEntry* pEntry = syncEntryBuildNoop(term, index);
+  assert(pEntry != NULL);
+
+  if (ths->state == TAOS_SYNC_STATE_LEADER) {
+    ths->pLogStore->appendEntry(ths->pLogStore, pEntry);
+    syncNodeReplicate(ths);
+  }
+
+  return ret;
+}
+
 // on message ----
 static int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg) {
   int32_t ret = 0;
@@ -851,7 +898,7 @@ static int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg
     syncEntry2OriginalRpc(pEntry, &rpcMsg);
 
     if (ths->pFsm != NULL) {
-      if (ths->pFsm->FpPreCommitCb != NULL) {
+      if (ths->pFsm->FpPreCommitCb != NULL && pEntry->entryType == SYNC_RAFT_ENTRY_DATA) {
         ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 0, ths->state);
       }
     }
@@ -866,7 +913,7 @@ static int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg
     syncEntry2OriginalRpc(pEntry, &rpcMsg);
 
     if (ths->pFsm != NULL) {
-      if (ths->pFsm->FpPreCommitCb != NULL) {
+      if (ths->pFsm->FpPreCommitCb != NULL && pEntry->entryType == SYNC_RAFT_ENTRY_DATA) {
         ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, pEntry->index, pEntry->isWeak, 1, ths->state);
       }
     }

source/libs/sync/src/syncRaftEntry.c

@@ -28,6 +28,13 @@ SSyncRaftEntry* syncEntryBuild(uint32_t dataLen) {
 
 // step 4. SyncClientRequest => SSyncRaftEntry, add term, index
 SSyncRaftEntry* syncEntryBuild2(SyncClientRequest* pMsg, SyncTerm term, SyncIndex index) {
+  SSyncRaftEntry* pEntry = syncEntryBuild3(pMsg, term, index, SYNC_RAFT_ENTRY_DATA);
+  assert(pEntry != NULL);
+
+  return pEntry;
+}
+
+SSyncRaftEntry* syncEntryBuild3(SyncClientRequest* pMsg, SyncTerm term, SyncIndex index, EntryType entryType) {
   SSyncRaftEntry* pEntry = syncEntryBuild(pMsg->dataLen);
   assert(pEntry != NULL);
 
@@ -37,12 +44,23 @@ SSyncRaftEntry* syncEntryBuild2(SyncClientRequest* pMsg, SyncTerm term, SyncInde
   pEntry->isWeak = pMsg->isWeak;
   pEntry->term = term;
   pEntry->index = index;
+  pEntry->entryType = entryType;
   pEntry->dataLen = pMsg->dataLen;
   memcpy(pEntry->data, pMsg->data, pMsg->dataLen);
 
   return pEntry;
 }
 
+SSyncRaftEntry* syncEntryBuildNoop(SyncTerm term, SyncIndex index) {
+  SSyncRaftEntry* pEntry = syncEntryBuild(0);
+  assert(pEntry != NULL);
+  pEntry->term = term;
+  pEntry->index = index;
+  pEntry->entryType = SYNC_RAFT_ENTRY_NOOP;
+
+  return pEntry;
+}
+
 void syncEntryDestory(SSyncRaftEntry* pEntry) {
   if (pEntry != NULL) {
     free(pEntry);
@@ -85,6 +103,7 @@ cJSON* syncEntry2Json(const SSyncRaftEntry* pEntry) {
     cJSON_AddStringToObject(pRoot, "term", u64buf);
     snprintf(u64buf, sizeof(u64buf), "%lu", pEntry->index);
     cJSON_AddStringToObject(pRoot, "index", u64buf);
+    cJSON_AddNumberToObject(pRoot, "entryType", pEntry->entryType);
     cJSON_AddNumberToObject(pRoot, "dataLen", pEntry->dataLen);
 
     char* s;

source/libs/sync/src/syncRaftLog.c

@@ -34,7 +34,7 @@ SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
   pLogStore->getLastTerm = logStoreLastTerm;
   pLogStore->updateCommitIndex = logStoreUpdateCommitIndex;
   pLogStore->getCommitIndex = logStoreGetCommitIndex;
-  return pLogStore;  // to avoid compiler error
+  return pLogStore;
 }
 
 void logStoreDestory(SSyncLogStore* pLogStore) {
@@ -48,18 +48,22 @@ int32_t logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
   SSyncLogStoreData* pData = pLogStore->data;
   SWal*              pWal = pData->pWal;
 
-  assert(pEntry->index == logStoreLastIndex(pLogStore) + 1);
+  SyncIndex lastIndex = logStoreLastIndex(pLogStore);
+  assert(pEntry->index == lastIndex + 1);
   uint32_t len;
   char*    serialized = syncEntrySerialize(pEntry, &len);
   assert(serialized != NULL);
 
-  int code;
-  code = walWrite(pWal, pEntry->index, pEntry->msgType, serialized, len);
-  assert(code == 0);
+  int code = 0;
+  /*
+    code = walWrite(pWal, pEntry->index, pEntry->entryType, serialized, len);
+    assert(code == 0);
+  */
+  assert(walWrite(pWal, pEntry->index, pEntry->entryType, serialized, len) == 0);
 
   walFsync(pWal, true);
   free(serialized);
-  return code;  // to avoid compiler error
+  return code;
 }
 
 SSyncRaftEntry* logStoreGetEntry(SSyncLogStore* pLogStore, SyncIndex index) {
@@ -69,7 +73,7 @@ SSyncRaftEntry* logStoreGetEntry(SSyncLogStore* pLogStore, SyncIndex index) {
 
   if (index >= SYNC_INDEX_BEGIN && index <= logStoreLastIndex(pLogStore)) {
     SWalReadHandle* pWalHandle = walOpenReadHandle(pWal);
-    walReadWithHandle(pWalHandle, index);
+    assert(walReadWithHandle(pWalHandle, index) == 0);
     pEntry = syncEntryDeserialize(pWalHandle->pHead->head.body, pWalHandle->pHead->head.len);
     assert(pEntry != NULL);
 
@@ -83,7 +87,7 @@ SSyncRaftEntry* logStoreGetEntry(SSyncLogStore* pLogStore, SyncIndex index) {
 int32_t logStoreTruncate(SSyncLogStore* pLogStore, SyncIndex fromIndex) {
   SSyncLogStoreData* pData = pLogStore->data;
   SWal*              pWal = pData->pWal;
-  walRollback(pWal, fromIndex);
+  assert(walRollback(pWal, fromIndex) == 0);
   return 0;  // to avoid compiler error
 }
 
@@ -107,7 +111,7 @@ SyncTerm logStoreLastTerm(SSyncLogStore* pLogStore) {
 int32_t logStoreUpdateCommitIndex(SSyncLogStore* pLogStore, SyncIndex index) {
   SSyncLogStoreData* pData = pLogStore->data;
   SWal*              pWal = pData->pWal;
-  walCommit(pWal, index);
+  assert(walCommit(pWal, index) == 0);
   return 0;  // to avoid compiler error
 }
 

source/libs/sync/test/CMakeLists.txt

@@ -34,6 +34,7 @@ add_executable(syncEncodeTest "")
 add_executable(syncWriteTest "")
 add_executable(syncReplicateTest "")
 add_executable(syncReplicateTest2 "")
+add_executable(syncReplicateTest3 "")
 add_executable(syncReplicateLoadTest "")
 add_executable(syncRefTest "")
 add_executable(syncLogStoreCheck "")
@@ -183,6 +184,10 @@ target_sources(syncReplicateTest2
     PRIVATE
     "syncReplicateTest2.cpp"
 )
+target_sources(syncReplicateTest3
+    PRIVATE
+    "syncReplicateTest3.cpp"
+)
 target_sources(syncReplicateLoadTest
     PRIVATE
     "syncReplicateLoadTest.cpp"
@@ -377,6 +382,11 @@ target_include_directories(syncReplicateTest2
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
 )
+target_include_directories(syncReplicateTest3
+    PUBLIC
+    "${CMAKE_SOURCE_DIR}/include/libs/sync"
+    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+)
 target_include_directories(syncReplicateLoadTest
     PUBLIC
     "${CMAKE_SOURCE_DIR}/include/libs/sync"
@@ -538,6 +548,10 @@ target_link_libraries(syncReplicateTest2
     sync
     gtest_main
 )
+target_link_libraries(syncReplicateTest3
+    sync
+    gtest_main
+)
 target_link_libraries(syncReplicateLoadTest
     sync
     gtest_main

source/libs/sync/test/syncReplicateLoadTest.cpp

@@ -178,9 +178,10 @@ int main(int argc, char **argv) {
   while (1) {
     sTrace(
         "replicate sleep, state: %d, %s, term:%lu electTimerLogicClock:%lu, electTimerLogicClockUser:%lu, "
-        "electTimerMS:%d",
+        "electTimerMS:%d, commitIndex:%ld",
         pSyncNode->state, syncUtilState2String(pSyncNode->state), pSyncNode->pRaftStore->currentTerm,
-        pSyncNode->electTimerLogicClock, pSyncNode->electTimerLogicClockUser, pSyncNode->electTimerMS);
+        pSyncNode->electTimerLogicClock, pSyncNode->electTimerLogicClockUser, pSyncNode->electTimerMS,
+        pSyncNode->commitIndex);
     taosMsleep(1000);
   }
 

taos-tools @ f36b07f7

-Subproject commit 8145dd1713ab9e7652ea621ca7e6895fd0b21d65
+Subproject commit f36b07f710d661dca88fdd70e73b5e3e16a960e0