Merge from develop

src/client/inc/tscSecondaryMerge.h

@@ -63,19 +63,21 @@ typedef struct SLocalReducer {
 //  char *                 pBufForInterpo;  // intermediate buffer for interpolation
   tFilePage *            pTempBuffer;
   struct SQLFunctionCtx *pCtx;
-  int32_t                rowSize;     // size of each intermediate result.
+  int32_t                rowSize;      // size of each intermediate result.
-  int32_t                status;      // denote it is in reduce process, in reduce process, it
+  int32_t                finalRowSize; // final result row size
-  bool                   hasPrevRow;  // cannot be released
+  int32_t                status;       // denote it is in reduce process, in reduce process, it
+  bool                   hasPrevRow;   // cannot be released
   bool                   hasUnprocessedRow;
   tOrderDescriptor *     pDesc;
   SColumnModel *         resColModel;
   tExtMemBuffer **       pExtMemBuffer;      // disk-based buffer
-  SFillInfo*             pFillInfo;         // interpolation support structure
+  SFillInfo*             pFillInfo;          // interpolation support structure
   char *                 pFinalRes;          // result data after interpo
   tFilePage *            discardData;
   SResultInfo *          pResInfo;
   bool                   discard;
   int32_t                offset;             // limit offset value
+  bool                   orderPrjOnSTable;   // projection query on stable
 } SLocalReducer;
 
 typedef struct SSubqueryState {

src/client/inc/tsclient.h

@@ -285,8 +285,6 @@ typedef struct {
 typedef struct STscObj {
   void *             signature;
   void *             pTimer;
-  char               mnodeIp[TSDB_USER_LEN];
-  uint16_t           mnodePort;
   char               user[TSDB_USER_LEN];
   char               pass[TSDB_KEY_LEN];
   char               acctId[TSDB_DB_NAME_LEN];

src/client/src/tscFunctionImpl.c

@@ -341,16 +341,6 @@ bool stableQueryFunctChanged(int32_t funcId) {
  */
 void resetResultInfo(SResultInfo *pResInfo) { pResInfo->initialized = false; }
 
-void initResultInfo(SResultInfo *pResInfo) {
-  pResInfo->initialized = true;  // the this struct has been initialized flag
-  
-  pResInfo->complete = false;
-  pResInfo->hasResult = false;
-  pResInfo->numOfRes = 0;
-  
-  memset(pResInfo->interResultBuf, 0, (size_t)pResInfo->bufLen);
-}
-
 void setResultInfoBuf(SResultInfo *pResInfo, int32_t size, bool superTable) {
   assert(pResInfo->interResultBuf == NULL);
   
@@ -387,9 +377,7 @@ static bool function_setup(SQLFunctionCtx *pCtx) {
  */
 static void function_finalizer(SQLFunctionCtx *pCtx) {
   SResultInfo *pResInfo = GET_RES_INFO(pCtx);
-  
   if (pResInfo->hasResult != DATA_SET_FLAG) {
-    tscTrace("no result generated, result is set to NULL");
     if (pCtx->outputType == TSDB_DATA_TYPE_BINARY || pCtx->outputType == TSDB_DATA_TYPE_NCHAR) {
       setVardataNull(pCtx->aOutputBuf, pCtx->outputType);
     } else {

src/client/src/tscLocal.c

@@ -26,7 +26,7 @@
 #include "tschemautil.h"
 #include "tname.h"
 
-static void tscSetLocalQueryResult(SSqlObj *pSql, const char *val, const char *columnName, size_t valueLength);
+static void tscSetLocalQueryResult(SSqlObj *pSql, const char *val, const char *columnName, int16_t type, size_t valueLength);
 
 static int32_t getToStringLength(const char *pData, int32_t length, int32_t type) {
   char buf[512] = {0};
@@ -275,22 +275,37 @@ static void tscProcessCurrentUser(SSqlObj *pSql) {
   SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
   
   SSqlExpr* pExpr = taosArrayGetP(pQueryInfo->exprList, 0);
-  tscSetLocalQueryResult(pSql, pSql->pTscObj->user, pExpr->aliasName, TSDB_USER_LEN);
+  pExpr->resBytes = TSDB_USER_LEN + TSDB_DATA_TYPE_BINARY;
+  pExpr->resType = TSDB_DATA_TYPE_BINARY;
+  
+  char* vx = calloc(1, pExpr->resBytes);
+  STR_WITH_MAXSIZE_TO_VARSTR(vx, pSql->pTscObj->user, TSDB_USER_LEN);
+  
+  tscSetLocalQueryResult(pSql, vx, pExpr->aliasName, pExpr->resType, pExpr->resBytes);
+  free(vx);
 }
 
 static void tscProcessCurrentDB(SSqlObj *pSql) {
   char db[TSDB_DB_NAME_LEN + 1] = {0};
   extractDBName(pSql->pTscObj->db, db);
   
-  // no use db is invoked before.
-  if (strlen(db) == 0) {
-    setNull(db, TSDB_DATA_TYPE_BINARY, TSDB_DB_NAME_LEN);
-  }
-  
   SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
   
   SSqlExpr* pExpr = taosArrayGetP(pQueryInfo->exprList, 0);
-  tscSetLocalQueryResult(pSql, db, pExpr->aliasName, TSDB_DB_NAME_LEN);
+  pExpr->resType = TSDB_DATA_TYPE_BINARY;
+  
+  size_t t = strlen(db);
+  pExpr->resBytes = TSDB_DB_NAME_LEN + VARSTR_HEADER_SIZE;
+  
+  char* vx = calloc(1, pExpr->resBytes);
+  if (t == 0) {
+    setVardataNull(vx, TSDB_DATA_TYPE_BINARY);
+  } else {
+    STR_WITH_SIZE_TO_VARSTR(vx, db, t);
+  }
+  
+  tscSetLocalQueryResult(pSql, vx, pExpr->aliasName, pExpr->resType, pExpr->resBytes);
+  free(vx);
 }
 
 static void tscProcessServerVer(SSqlObj *pSql) {
@@ -298,14 +313,32 @@ static void tscProcessServerVer(SSqlObj *pSql) {
   SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
   
   SSqlExpr* pExpr = taosArrayGetP(pQueryInfo->exprList, 0);
-  tscSetLocalQueryResult(pSql, v, pExpr->aliasName, tListLen(pSql->pTscObj->sversion));
+  pExpr->resType = TSDB_DATA_TYPE_BINARY;
+  
+  size_t t = strlen(v);
+  pExpr->resBytes = t + VARSTR_HEADER_SIZE;
+  
+  char* vx = calloc(1, pExpr->resBytes);
+  STR_WITH_SIZE_TO_VARSTR(vx, v, t);
+  tscSetLocalQueryResult(pSql, vx, pExpr->aliasName, pExpr->resType, pExpr->resBytes);
+  
+  tfree(vx);
 }
 
 static void tscProcessClientVer(SSqlObj *pSql) {
   SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
   
   SSqlExpr* pExpr = taosArrayGetP(pQueryInfo->exprList, 0);
-  tscSetLocalQueryResult(pSql, version, pExpr->aliasName, strlen(version));
+  pExpr->resType = TSDB_DATA_TYPE_BINARY;
+  
+  size_t t = strlen(version);
+  pExpr->resBytes = t + VARSTR_HEADER_SIZE;
+  
+  char* v = calloc(1, pExpr->resBytes);
+  STR_WITH_SIZE_TO_VARSTR(v, version, t);
+  tscSetLocalQueryResult(pSql, v, pExpr->aliasName, pExpr->resType, pExpr->resBytes);
+  
+  tfree(v);
 }
 
 static void tscProcessServStatus(SSqlObj *pSql) {
@@ -325,10 +358,11 @@ static void tscProcessServStatus(SSqlObj *pSql) {
   SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
   
   SSqlExpr* pExpr = taosArrayGetP(pQueryInfo->exprList, 0);
-  tscSetLocalQueryResult(pSql, "1", pExpr->aliasName, 2);
+  int32_t val = 1;
+  tscSetLocalQueryResult(pSql, (char*) &val, pExpr->aliasName, TSDB_DATA_TYPE_INT, sizeof(int32_t));
 }
 
-void tscSetLocalQueryResult(SSqlObj *pSql, const char *val, const char *columnName, size_t valueLength) {
+void tscSetLocalQueryResult(SSqlObj *pSql, const char *val, const char *columnName, int16_t type, size_t valueLength) {
   SSqlCmd *pCmd = &pSql->cmd;
   SSqlRes *pRes = &pSql->res;
 
@@ -338,8 +372,10 @@ void tscSetLocalQueryResult(SSqlObj *pSql, const char *val, const char *columnNa
   pQueryInfo->order.order = TSDB_ORDER_ASC;
   
   tscFieldInfoClear(&pQueryInfo->fieldsInfo);
+  pQueryInfo->fieldsInfo.pFields = taosArrayInit(1, sizeof(TAOS_FIELD));
+  pQueryInfo->fieldsInfo.pSupportInfo = taosArrayInit(1, sizeof(SFieldSupInfo));
   
-  TAOS_FIELD f = tscCreateField(TSDB_DATA_TYPE_BINARY, columnName, valueLength);
+  TAOS_FIELD f = tscCreateField(type, columnName, valueLength);
   tscFieldInfoAppend(&pQueryInfo->fieldsInfo, &f);
   
   tscInitResObjForLocalQuery(pSql, 1, valueLength);
@@ -348,7 +384,7 @@ void tscSetLocalQueryResult(SSqlObj *pSql, const char *val, const char *columnNa
   SFieldSupInfo* pInfo = tscFieldInfoGetSupp(&pQueryInfo->fieldsInfo, 0);
   pInfo->pSqlExpr = taosArrayGetP(pQueryInfo->exprList, 0);
   
-  strncpy(pRes->data, val, pField->bytes);
+  memcpy(pRes->data, val, pField->bytes);
 }
 
 int tscProcessLocalCmd(SSqlObj *pSql) {
@@ -385,7 +421,7 @@ int tscProcessLocalCmd(SSqlObj *pSql) {
   // keep the code in local variable in order to avoid invalid read in case of async query
   int32_t code = pSql->res.code;
   if (code == TSDB_CODE_SUCCESS) {
-    (*pSql->fp)(pSql->param, pSql, 0);
+    (*pSql->fp)(pSql->param, pSql, code);
   } else {
     tscQueueAsyncRes(pSql);
   }

src/client/src/tscParseInsert.c

@@ -48,7 +48,7 @@ static int32_t tscToInteger(SSQLToken *pToken, int64_t *value, char **endPtr) {
   
   int32_t radix = 10;
   
-  int32_t radixList[3] = {16, 8, 2};
+  int32_t radixList[3] = {16, 8, 2}; // the integer number with different radix: hex, oct, bin
   if (pToken->type == TK_HEX || pToken->type == TK_OCT || pToken->type == TK_BIN) {
     radix = radixList[pToken->type - TK_HEX];
   }

src/client/src/tscPrepare.c

@@ -494,7 +494,6 @@ TAOS_STMT* taos_stmt_init(TAOS* taos) {
   tsem_init(&pSql->rspSem, 0, 0);
   pSql->signature     = pSql;
   pSql->pTscObj       = pObj;
-  //pSql->pTscObj->pSql = pSql;
   pSql->maxRetry      = TSDB_MAX_REPLICA_NUM;
 
   pStmt->pSql = pSql;
@@ -515,7 +514,7 @@ int taos_stmt_prepare(TAOS_STMT* stmt, const char* sql, unsigned long length) {
   //doAsyncQuery(pObj, pSql, waitForQueryRsp, taos, sqlstr, sqlLen);
   SSqlCmd *pCmd    = &pSql->cmd;
   SSqlRes *pRes    = &pSql->res;
-  pSql->param      = (void*)pSql;
+  pSql->param      = (void*) pSql;
   pSql->fp         = waitForQueryRsp;
   pSql->insertType = TSDB_QUERY_TYPE_STMT_INSERT;
   

src/client/src/tscSQLParser.c

@@ -1881,22 +1881,38 @@ int32_t addExprAndResultField(SQueryInfo* pQueryInfo, int32_t colIndex, tSQLExpr
   
       // functions can not be applied to normal columns
       int32_t numOfCols = tscGetNumOfColumns(pTableMetaInfo->pTableMeta);
-      if (index.columnIndex < numOfCols) {
+      if (index.columnIndex < numOfCols && index.columnIndex != TSDB_TBNAME_COLUMN_INDEX) {
         return invalidSqlErrMsg(pQueryInfo->msg, msg6);
       }
     
-      index.columnIndex -= numOfCols;
+      if (index.columnIndex > 0) {
+        index.columnIndex -= numOfCols;
+      }
       
       // 2. valid the column type
-      int16_t colType = pSchema[index.columnIndex].type;
+      int16_t colType = 0;
-      if (colType == TSDB_DATA_TYPE_BOOL || colType >= TSDB_DATA_TYPE_BINARY) {
+      if (index.columnIndex == TSDB_TBNAME_COLUMN_INDEX) {
+        colType = TSDB_DATA_TYPE_BINARY;
+      } else {
+        colType = pSchema[index.columnIndex].type;
+      }
+      
+      if (colType == TSDB_DATA_TYPE_BOOL) {
         return invalidSqlErrMsg(pQueryInfo->msg, msg1);
       }
 
       tscColumnListInsert(pTableMetaInfo->tagColList, &index);
       SSchema* pTagSchema = tscGetTableTagSchema(pTableMetaInfo->pTableMeta);
-      SSchema s = pTagSchema[index.columnIndex];
+      
-
+      SSchema s = {0};
+      if (index.columnIndex == TSDB_TBNAME_COLUMN_INDEX) {
+        s.bytes = TSDB_TABLE_NAME_LEN + VARSTR_HEADER_SIZE;
+        s.type  = TSDB_DATA_TYPE_BINARY;
+        s.colId = TSDB_TBNAME_COLUMN_INDEX;
+      } else {
+        s = pTagSchema[index.columnIndex];
+      }
+      
       int16_t bytes = 0;
       int16_t type  = 0;
       int32_t inter = 0;
@@ -5379,20 +5395,15 @@ int32_t doLocalQueryProcess(SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql) {
 
   switch (index) {
     case 0:
-      pQueryInfo->command = TSDB_SQL_CURRENT_DB;
+      pQueryInfo->command = TSDB_SQL_CURRENT_DB;break;
-      return TSDB_CODE_SUCCESS;
     case 1:
-      pQueryInfo->command = TSDB_SQL_SERV_VERSION;
+      pQueryInfo->command = TSDB_SQL_SERV_VERSION;break;
-      return TSDB_CODE_SUCCESS;
+      case 2:
-    case 2:
+      pQueryInfo->command = TSDB_SQL_SERV_STATUS;break;
-      pQueryInfo->command = TSDB_SQL_SERV_STATUS;
-      return TSDB_CODE_SUCCESS;
     case 3:
-      pQueryInfo->command = TSDB_SQL_CLI_VERSION;
+      pQueryInfo->command = TSDB_SQL_CLI_VERSION;break;
-      return TSDB_CODE_SUCCESS;
     case 4:
-      pQueryInfo->command = TSDB_SQL_CURRENT_USER;
+      pQueryInfo->command = TSDB_SQL_CURRENT_USER;break;
-      return TSDB_CODE_SUCCESS;
     default: { return invalidSqlErrMsg(pQueryInfo->msg, msg3); }
   }
   
@@ -5402,6 +5413,8 @@ int32_t doLocalQueryProcess(SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql) {
   
   const char* name = (pExprList->a[0].aliasName != NULL)? pExprList->a[0].aliasName:functionsInfo[index].name;
   strncpy(pExpr1->aliasName, name, tListLen(pExpr1->aliasName));
+  
+  return TSDB_CODE_SUCCESS;
 }
 
 // can only perform the parameters based on the macro definitation

src/client/src/tscSchemaUtil.c

@@ -131,13 +131,6 @@ SSchema* tscGetTableColumnSchema(const STableMeta* pTableMeta, int32_t startCol)
   assert(pTableMeta != NULL);
   
   SSchema* pSchema = (SSchema*) pTableMeta->schema;
-#if 0
-  if (pTableMeta->tableType == TSDB_CHILD_TABLE) {
-    assert (pTableMeta->pSTable != NULL);
-    pSchema = pTableMeta->pSTable->schema;
-  }
-#endif
-
   return &pSchema[startCol];
 }
 

src/client/src/tscSecondaryMerge.c

@@ -217,7 +217,7 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
 
   pReducer->numOfBuffer = numOfFlush;
   pReducer->numOfVnode = numOfBuffer;
-
+  
   pReducer->pDesc = pDesc;
   tscTrace("%p the number of merged leaves is: %d", pSql, pReducer->numOfBuffer);
 
@@ -278,6 +278,7 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
 
   param->groupOrderType = pQueryInfo->groupbyExpr.orderType;
+  pReducer->orderPrjOnSTable = tscOrderedProjectionQueryOnSTable(pQueryInfo, 0);
 
   pRes->code = tLoserTreeCreate(&pReducer->pLoserTree, pReducer->numOfBuffer, param, treeComparator);
   if (pReducer->pLoserTree == NULL || pRes->code != 0) {
@@ -309,10 +310,10 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
   pReducer->nResultBufSize = pMemBuffer[0]->pageSize * 16;
   pReducer->pResultBuf = (tFilePage *)calloc(1, pReducer->nResultBufSize + sizeof(tFilePage));
 
-  int32_t finalRowLength = tscGetResRowLength(pQueryInfo->exprList);
+  pReducer->finalRowSize = tscGetResRowLength(pQueryInfo->exprList);
   pReducer->resColModel = finalmodel;
-  pReducer->resColModel->capacity = pReducer->nResultBufSize / finalRowLength;
+  pReducer->resColModel->capacity = pReducer->nResultBufSize / pReducer->finalRowSize;
-  assert(finalRowLength <= pReducer->rowSize);
+  assert(pReducer->finalRowSize <= pReducer->rowSize);
 
   pReducer->pFinalRes = calloc(1, pReducer->rowSize * pReducer->resColModel->capacity);
 //  pReducer->pBufForInterpo = calloc(1, pReducer->nResultBufSize);
@@ -389,7 +390,7 @@ static int32_t tscFlushTmpBufferImpl(tExtMemBuffer *pMemoryBuf, tOrderDescriptor
   assert(pPage->num <= pDesc->pColumnModel->capacity);
 
   // sort before flush to disk, the data must be consecutively put on tFilePage.
-  if (pDesc->orderIdx.numOfCols > 0) {
+  if (pDesc->orderInfo.numOfCols > 0) {
     tColDataQSort(pDesc, pPage->num, 0, pPage->num - 1, pPage->data, orderType);
   }
 
@@ -590,12 +591,10 @@ static int32_t createOrderDescriptor(tOrderDescriptor **pOrderDesc, SSqlCmd *pCm
 bool isSameGroup(SSqlCmd *pCmd, SLocalReducer *pReducer, char *pPrev, tFilePage *tmpBuffer) {
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
 
-  int16_t functionId = tscSqlExprGet(pQueryInfo, 0)->functionId;
-
   // disable merge procedure for column projection query
+  int16_t functionId = pReducer->pCtx[0].functionId;
   assert(functionId != TSDB_FUNC_ARITHM);
-
+  if (pReducer->orderPrjOnSTable) {
-  if (tscOrderedProjectionQueryOnSTable(pQueryInfo, 0)) {
     return true;
   }
 
@@ -604,26 +603,33 @@ bool isSameGroup(SSqlCmd *pCmd, SLocalReducer *pReducer, char *pPrev, tFilePage
   }
 
   tOrderDescriptor *pOrderDesc = pReducer->pDesc;
-  int32_t           numOfCols = pOrderDesc->orderIdx.numOfCols;
+  SColumnOrderInfo* orderInfo = &pOrderDesc->orderInfo;
-
+  
   // no group by columns, all data belongs to one group
+  int32_t numOfCols = orderInfo->numOfCols;
   if (numOfCols <= 0) {
     return true;
   }
 
-  if (pOrderDesc->orderIdx.pData[numOfCols - 1] == PRIMARYKEY_TIMESTAMP_COL_INDEX) {  //<= 0
+  if (orderInfo->pData[numOfCols - 1] == PRIMARYKEY_TIMESTAMP_COL_INDEX) {
-    // super table interval query
+    /*
+     * super table interval query
+     * if the order columns is the primary timestamp, all result data belongs to one group
+     */
     assert(pQueryInfo->intervalTime > 0);
-    pOrderDesc->orderIdx.numOfCols -= 1;
+    if (numOfCols == 1) {
+      return true;
+    }
   } else {  // simple group by query
     assert(pQueryInfo->intervalTime == 0);
   }
 
   // only one row exists
-  int32_t ret = compare_a(pOrderDesc, 1, 0, pPrev, 1, 0, tmpBuffer->data);
+  int32_t index = orderInfo->pData[0];
-  pOrderDesc->orderIdx.numOfCols = numOfCols;
+  int32_t offset = (pOrderDesc->pColumnModel)->pFields[index].offset;
-
+  
-  return (ret == 0);
+  int32_t ret = memcmp(pPrev + offset, tmpBuffer->data + offset, pOrderDesc->pColumnModel->rowSize - offset);
+  return ret == 0;
 }
 
 int32_t tscLocalReducerEnvCreate(SSqlObj *pSql, tExtMemBuffer ***pMemBuffer, tOrderDescriptor **pOrderDesc,
@@ -873,24 +879,24 @@ static void reversedCopyFromInterpolationToDstBuf(SQueryInfo *pQueryInfo, SSqlRe
  * Note: pRes->pLocalReducer may be null, due to the fact that "tscDestroyLocalReducer" is called
  * by "interuptHandler" function in shell
  */
-static void doInterpolateResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneOutput) {
+static void doFillResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneOutput) {
   SSqlCmd *   pCmd = &pSql->cmd;
   SSqlRes *   pRes = &pSql->res;
   
   tFilePage * pFinalDataPage = pLocalReducer->pResultBuf;
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
 
-  if (pRes->pLocalReducer != pLocalReducer) {
+//  if (pRes->pLocalReducer != pLocalReducer) {
-    /*
+//    /*
-     * Release the SSqlObj is called, and it is int destroying function invoked by other thread.
+//     * Release the SSqlObj is called, and it is int destroying function invoked by other thread.
-     * However, the other thread will WAIT until current process fully completes.
+//     * However, the other thread will WAIT until current process fully completes.
-     * Since the flag of release struct is set by doLocalReduce function
+//     * Since the flag of release struct is set by doLocalReduce function
-     */
+//     */
-    assert(pRes->pLocalReducer == NULL);
+//    assert(pRes->pLocalReducer == NULL);
-  }
+//  }
 
+  // no interval query, no fill operation
   if (pQueryInfo->intervalTime == 0 || pQueryInfo->fillType == TSDB_FILL_NONE) {
-    // no interval query, no fill operation
     pRes->data = pLocalReducer->pFinalRes;
     pRes->numOfRows = pFinalDataPage->num;
     pRes->numOfClauseTotal += pRes->numOfRows;
@@ -929,9 +935,7 @@ static void doInterpolateResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, boo
       savePrevRecordAndSetupInterpoInfo(pLocalReducer, pQueryInfo, pLocalReducer->pFillInfo);
     }
 
-    int32_t rowSize = tscGetResRowLength(pQueryInfo->exprList);
+    memcpy(pRes->data, pFinalDataPage->data, pRes->numOfRows * pLocalReducer->finalRowSize);
-    memcpy(pRes->data, pFinalDataPage->data, pRes->numOfRows * rowSize);
-
     pFinalDataPage->num = 0;
     return;
   }
@@ -1037,16 +1041,13 @@ static void savePreviousRow(SLocalReducer *pLocalReducer, tFilePage *tmpBuffer)
 static void doExecuteSecondaryMerge(SSqlCmd *pCmd, SLocalReducer *pLocalReducer, bool needInit) {
   // the tag columns need to be set before all functions execution
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
+  
   size_t size = tscSqlExprNumOfExprs(pQueryInfo);
-
   for (int32_t j = 0; j < size; ++j) {
-    SSqlExpr *      pExpr = tscSqlExprGet(pQueryInfo, j);
     SQLFunctionCtx *pCtx = &pLocalReducer->pCtx[j];
 
-    tVariantAssign(&pCtx->param[0], &pExpr->param[0]);
-
     // tags/tags_dummy function, the tag field of SQLFunctionCtx is from the input buffer
-    int32_t functionId = pExpr->functionId;
+    int32_t functionId = pCtx->functionId;
     if (functionId == TSDB_FUNC_TAG_DUMMY || functionId == TSDB_FUNC_TAG || functionId == TSDB_FUNC_TS_DUMMY) {
       tVariantDestroy(&pCtx->tag);
       char* input = pCtx->aInputElemBuf;
@@ -1057,17 +1058,20 @@ static void doExecuteSecondaryMerge(SSqlCmd *pCmd, SLocalReducer *pLocalReducer,
       } else {
         tVariantCreateFromBinary(&pCtx->tag, input, pCtx->inputBytes, pCtx->inputType);
       }
+    } else if (functionId == TSDB_FUNC_TOP || functionId == TSDB_FUNC_BOTTOM) {
+      SSqlExpr *pExpr = tscSqlExprGet(pQueryInfo, j);
+      pCtx->param[0].i64Key = pExpr->param[0].i64Key;
     }
 
     pCtx->currentStage = SECONDARY_STAGE_MERGE;
 
     if (needInit) {
-      aAggs[pExpr->functionId].init(pCtx);
+      aAggs[pCtx->functionId].init(pCtx);
     }
   }
 
   for (int32_t j = 0; j < size; ++j) {
-    int32_t functionId = tscSqlExprGet(pQueryInfo, j)->functionId;
+    int32_t functionId = pLocalReducer->pCtx[j].functionId;
     if (functionId == TSDB_FUNC_TAG_DUMMY || functionId == TSDB_FUNC_TS_DUMMY) {
       continue;
     }
@@ -1101,8 +1105,7 @@ static int64_t getNumOfResultLocal(SQueryInfo *pQueryInfo, SQLFunctionCtx *pCtx)
      * ts, tag, tagprj function can not decide the output number of current query
      * the number of output result is decided by main output
      */
-    SSqlExpr *pExpr = tscSqlExprGet(pQueryInfo, j);
+    int32_t functionId = pCtx[j].functionId;
-    int32_t   functionId = pExpr->functionId;
     if (functionId == TSDB_FUNC_TS || functionId == TSDB_FUNC_TAG || functionId == TSDB_FUNC_TAGPRJ) {
       continue;
     }
@@ -1136,15 +1139,13 @@ static void fillMultiRowsOfTagsVal(SQueryInfo *pQueryInfo, int32_t numOfRes, SLo
 
   char *buf = malloc((size_t)maxBufSize);
   for (int32_t k = 0; k < size; ++k) {
-    SSqlExpr *pExpr = tscSqlExprGet(pQueryInfo, k);
+    SQLFunctionCtx *pCtx = &pLocalReducer->pCtx[k];
-    if (pExpr->functionId != TSDB_FUNC_TAG) {
+    if (pCtx->functionId != TSDB_FUNC_TAG) {
       continue;
     }
 
     int32_t inc = numOfRes - 1;  // tsdb_func_tag function only produce one row of result
     memset(buf, 0, (size_t)maxBufSize);
-
-    SQLFunctionCtx *pCtx = &pLocalReducer->pCtx[k];
     memcpy(buf, pCtx->aOutputBuf, (size_t)pCtx->outputBytes);
 
     for (int32_t i = 0; i < inc; ++i) {
@@ -1160,8 +1161,8 @@ int32_t finalizeRes(SQueryInfo *pQueryInfo, SLocalReducer *pLocalReducer) {
   size_t size = tscSqlExprNumOfExprs(pQueryInfo);
   
   for (int32_t k = 0; k < size; ++k) {
-    SSqlExpr *pExpr = tscSqlExprGet(pQueryInfo, k);
+    SQLFunctionCtx* pCtx = &pLocalReducer->pCtx[k];
-    aAggs[pExpr->functionId].xFinalize(&pLocalReducer->pCtx[k]);
+    aAggs[pCtx->functionId].xFinalize(pCtx);
   }
 
   pLocalReducer->hasPrevRow = false;
@@ -1182,13 +1183,13 @@ int32_t finalizeRes(SQueryInfo *pQueryInfo, SLocalReducer *pLocalReducer) {
  */
 bool needToMerge(SQueryInfo *pQueryInfo, SLocalReducer *pLocalReducer, tFilePage *tmpBuffer) {
   int32_t ret = 0;  // merge all result by default
-  int16_t functionId = tscSqlExprGet(pQueryInfo, 0)->functionId;
+  
-
+  int16_t functionId = pLocalReducer->pCtx[0].functionId;
   if (functionId == TSDB_FUNC_PRJ || functionId == TSDB_FUNC_ARITHM) {  // column projection query
     ret = 1;                                                            // disable merge procedure
   } else {
     tOrderDescriptor *pDesc = pLocalReducer->pDesc;
-    if (pDesc->orderIdx.numOfCols > 0) {
+    if (pDesc->orderInfo.numOfCols > 0) {
       if (pDesc->tsOrder == TSDB_ORDER_ASC) {  // asc
         // todo refactor comparator
         ret = compare_a(pLocalReducer->pDesc, 1, 0, pLocalReducer->prevRowOfInput, 1, 0, tmpBuffer->data);
@@ -1274,7 +1275,7 @@ bool doGenerateFinalResults(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool no
     taosFillCopyInputDataFromOneFilePage(pFillInfo, pResBuf);
   }
   
-  doInterpolateResult(pSql, pLocalReducer, noMoreCurrentGroupRes);
+  doFillResult(pSql, pLocalReducer, noMoreCurrentGroupRes);
   return true;
 }
 
@@ -1341,7 +1342,7 @@ static bool doBuildFilledResultForGroup(SSqlObj *pSql) {
     // the first column must be the timestamp column
     int32_t rows = taosGetNumOfResultWithFill(pFillInfo, remain, ekey, pLocalReducer->resColModel->capacity);
     if (rows > 0) {  // do interpo
-      doInterpolateResult(pSql, pLocalReducer, false);
+      doFillResult(pSql, pLocalReducer, false);
     }
 
     return true;
@@ -1374,7 +1375,7 @@ static bool doHandleLastRemainData(SSqlObj *pSql) {
                                    pQueryInfo->slidingTimeUnit, tinfo.precision);
       int32_t rows = taosGetNumOfResultWithFill(pFillInfo, 0, etime, pLocalReducer->resColModel->capacity);
       if (rows > 0) {  // do interpo
-        doInterpolateResult(pSql, pLocalReducer, true);
+        doFillResult(pSql, pLocalReducer, true);
       }
     }
 
@@ -1408,13 +1409,11 @@ static void doProcessResultInNextWindow(SSqlObj *pSql, int32_t numOfRes) {
   size_t size = tscSqlExprNumOfExprs(pQueryInfo);
 
   for (int32_t k = 0; k < size; ++k) {
-    SSqlExpr *      pExpr = tscSqlExprGet(pQueryInfo, k);
     SQLFunctionCtx *pCtx = &pLocalReducer->pCtx[k];
-
     pCtx->aOutputBuf += pCtx->outputBytes * numOfRes;
 
     // set the correct output timestamp column position
-    if (pExpr->functionId == TSDB_FUNC_TOP || pExpr->functionId == TSDB_FUNC_BOTTOM) {
+    if (pCtx->functionId == TSDB_FUNC_TOP || pCtx->functionId == TSDB_FUNC_BOTTOM) {
       pCtx->ptsOutputBuf = ((char *)pCtx->ptsOutputBuf + TSDB_KEYSIZE * numOfRes);
     }
   }

src/client/src/tscServer.c

@@ -333,10 +333,10 @@ void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcIpSet *pIpSet) {
   if (pRes->code == TSDB_CODE_SUCCESS && tscProcessMsgRsp[pCmd->command])
     rpcMsg->code = (*tscProcessMsgRsp[pCmd->command])(pSql);
   
+
   if (rpcMsg->code != TSDB_CODE_TSC_ACTION_IN_PROGRESS) {
-    rpcMsg->code = pRes->code ? pRes->code : pRes->numOfRows;
+    rpcMsg->code = (pRes->code == TSDB_CODE_SUCCESS) ? pRes->numOfRows: pRes->code;
     
-    tscTrace("%p SQL result:%s res:%p", pSql, tstrerror(pRes->code), pSql);
     bool shouldFree = tscShouldBeFreed(pSql);
     (*pSql->fp)(pSql->param, pSql, rpcMsg->code);
 
@@ -1392,22 +1392,9 @@ int tscProcessDescribeTableRsp(SSqlObj *pSql) {
   return tscLocalResultCommonBuilder(pSql, numOfRes);
 }
 
-int tscProcessTagRetrieveRsp(SSqlObj *pSql) {
+int tscProcessLocalRetrieveRsp(SSqlObj *pSql) {
-//  SSqlCmd *pCmd = &pSql->cmd;
+  int32_t numOfRes = 1;
-
+  pSql->res.completed = true;
-//  SQueryInfo *    pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
-//  STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-
-  int32_t numOfRes = 0;
-#if 0
-  if (tscSqlExprGet(pQueryInfo, 0)->functionId == TSDB_FUNC_TAGPRJ) {
-    numOfRes = pTableMetaInfo->pMetricMeta->numOfTables;
-  } else {
-    numOfRes = 1;  // for count function, there is only one output.
-  }
-  
-#endif
-
   return tscLocalResultCommonBuilder(pSql, numOfRes);
 }
 
@@ -2312,7 +2299,7 @@ int tscProcessRetrieveRspFromNode(SSqlObj *pSql) {
   }
 
   pRes->row = 0;
-  tscTrace("%p numOfRows:%d, offset:%d", pSql, pRes->numOfRows, pRes->offset);
+  tscTrace("%p numOfRows:%d, offset:%d, complete:%d", pSql, pRes->numOfRows, pRes->offset, pRes->completed);
 
   return 0;
 }
@@ -2563,11 +2550,11 @@ void tscInitMsgsFp() {
   tscProcessMsgRsp[TSDB_SQL_RETRIEVE] = tscProcessRetrieveRspFromNode;  // rsp handled by same function.
   tscProcessMsgRsp[TSDB_SQL_DESCRIBE_TABLE] = tscProcessDescribeTableRsp;
 
-  tscProcessMsgRsp[TSDB_SQL_CURRENT_DB] = tscProcessTagRetrieveRsp;
+  tscProcessMsgRsp[TSDB_SQL_CURRENT_DB]   = tscProcessLocalRetrieveRsp;
-  tscProcessMsgRsp[TSDB_SQL_CURRENT_USER] = tscProcessTagRetrieveRsp;
+  tscProcessMsgRsp[TSDB_SQL_CURRENT_USER] = tscProcessLocalRetrieveRsp;
-  tscProcessMsgRsp[TSDB_SQL_SERV_VERSION] = tscProcessTagRetrieveRsp;
+  tscProcessMsgRsp[TSDB_SQL_SERV_VERSION] = tscProcessLocalRetrieveRsp;
-  tscProcessMsgRsp[TSDB_SQL_CLI_VERSION] = tscProcessTagRetrieveRsp;
+  tscProcessMsgRsp[TSDB_SQL_CLI_VERSION]  = tscProcessLocalRetrieveRsp;
-  tscProcessMsgRsp[TSDB_SQL_SERV_STATUS] = tscProcessTagRetrieveRsp;
+  tscProcessMsgRsp[TSDB_SQL_SERV_STATUS]  = tscProcessLocalRetrieveRsp;
 
   tscProcessMsgRsp[TSDB_SQL_RETRIEVE_EMPTY_RESULT] = tscProcessEmptyResultRsp;
 

src/client/src/tscSql.c

@@ -88,7 +88,6 @@ SSqlObj *taosConnectImpl(const char *ip, const char *user, const char *pass, con
 
   strncpy(pObj->user, user, TSDB_USER_LEN);
   taosEncryptPass((uint8_t *)pass, strlen(pass), pObj->pass);
-  pObj->mnodePort = port ? port : tsDnodeShellPort;
 
   if (db) {
     int32_t len = strlen(db);
@@ -414,10 +413,6 @@ int taos_fetch_block_impl(TAOS_RES *res, TAOS_ROW *rows) {
 
 static void waitForRetrieveRsp(void *param, TAOS_RES *tres, int numOfRows) {
   SSqlObj* pSql = (SSqlObj*) tres;
-  
-  if (numOfRows < 0) { // set the error code
-    pSql->res.code = -numOfRows;
-  }
   sem_post(&pSql->rspSem);
 }
 
@@ -445,7 +440,12 @@ TAOS_ROW taos_fetch_row(TAOS_RES *res) {
        pCmd->command == TSDB_SQL_FETCH ||
        pCmd->command == TSDB_SQL_SHOW ||
        pCmd->command == TSDB_SQL_SELECT ||
-       pCmd->command == TSDB_SQL_DESCRIBE_TABLE)) {
+       pCmd->command == TSDB_SQL_DESCRIBE_TABLE ||
+       pCmd->command == TSDB_SQL_SERV_STATUS ||
+       pCmd->command == TSDB_SQL_CURRENT_DB ||
+       pCmd->command == TSDB_SQL_SERV_VERSION ||
+       pCmd->command == TSDB_SQL_CLI_VERSION ||
+       pCmd->command == TSDB_SQL_CURRENT_USER )) {
     taos_fetch_rows_a(res, waitForRetrieveRsp, pSql->pTscObj);
     sem_wait(&pSql->rspSem);
   }

src/client/src/tscSubquery.c

@@ -1557,8 +1557,8 @@ static void tscRetrieveFromDnodeCallBack(void *param, TAOS_RES *tres, int numOfR
     assert(pRes->numOfRows == numOfRows);
     int64_t num = atomic_add_fetch_64(&pState->numOfRetrievedRows, numOfRows);
     
-//    tscTrace("%p sub:%p retrieve numOfRows:%d totalNumOfRows:%d from ip:%u,vid:%d,orderOfSub:%d", pPObj, pSql,
+    tscTrace("%p sub:%p retrieve numOfRows:%d totalNumOfRows:%d from ip:%s, orderOfSub:%d", pPObj, pSql,
-//             pRes->numOfRows, pState->numOfRetrievedRows, pSvd->ip, pSvd->vnode, idx);
+             pRes->numOfRows, pState->numOfRetrievedRows, pSql->ipList.fqdn[pSql->ipList.inUse], idx);
     
     if (num > tsMaxNumOfOrderedResults && tscIsProjectionQueryOnSTable(pQueryInfo, 0)) {
       tscError("%p sub:%p num of OrderedRes is too many, max allowed:%" PRId64 " , current:%" PRId64,
@@ -1713,6 +1713,11 @@ static void multiVnodeInsertMerge(void* param, TAOS_RES* tres, int numOfRows) {
   // increase the total inserted rows
   if (numOfRows > 0) {
     pParentObj->res.numOfRows += numOfRows;
+  } else {
+    SSqlObj* pSql = (SSqlObj*) tres;
+    assert(pSql != NULL && pSql->res.code == numOfRows);
+    
+    pParentObj->res.code = pSql->res.code;
   }
   
   taos_free_result(tres);
@@ -1947,7 +1952,8 @@ void **doSetResultRowData(SSqlObj *pSql, bool finalResult) {
   
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
   
-  for (int i = 0; i < tscNumOfFields(pQueryInfo); ++i) {
+  size_t size = tscNumOfFields(pQueryInfo);
+  for (int i = 0; i < size; ++i) {
     SFieldSupInfo* pSup = tscFieldInfoGetSupp(&pQueryInfo->fieldsInfo, i);
     if (pSup->pSqlExpr != NULL) {
       tscGetResultColumnChr(pRes, &pQueryInfo->fieldsInfo, i);

src/client/src/tscUtil.c

@@ -404,8 +404,6 @@ void tscPartiallyFreeSqlObj(SSqlObj* pSql) {
   pSql->numOfSubs = 0;
   
   tscResetSqlCmdObj(pCmd);
-  
-  tscTrace("%p partially free sqlObj completed", pSql);
 }
 
 void tscFreeSqlObj(SSqlObj* pSql) {
@@ -967,7 +965,8 @@ static SSqlExpr* doBuildSqlExpr(SQueryInfo* pQueryInfo, int16_t functionId, SCol
       SSchema* pSchema = tscGetTableTagSchema(pTableMetaInfo->pTableMeta);
       pExpr->colInfo.colId = pSchema[pColIndex->columnIndex].colId;
       strncpy(pExpr->colInfo.name, pSchema[pColIndex->columnIndex].name, TSDB_COL_NAME_LEN);
-    } else {
+    } else if (pTableMetaInfo->pTableMeta != NULL) {
+      // in handling select database/version/server_status(), the pTableMeta is NULL
       SSchema* pSchema = tscGetTableColumnSchema(pTableMetaInfo->pTableMeta, pColIndex->columnIndex);
       pExpr->colInfo.colId = pSchema->colId;
       strncpy(pExpr->colInfo.name, pSchema->name, TSDB_COL_NAME_LEN);
@@ -979,8 +978,12 @@ static SSqlExpr* doBuildSqlExpr(SQueryInfo* pQueryInfo, int16_t functionId, SCol
   pExpr->colInfo.colIndex = pColIndex->columnIndex;
   pExpr->resType       = type;
   pExpr->resBytes      = size;
-  pExpr->interBytes = interSize;
+  pExpr->interBytes    = interSize;
-  pExpr->uid           = pTableMetaInfo->pTableMeta->uid;
+  
+  if (pTableMetaInfo->pTableMeta) {
+    pExpr->uid = pTableMetaInfo->pTableMeta->uid;
+  }
+  
   
   return pExpr;
 }
@@ -2104,7 +2107,7 @@ void tscTryQueryNextClause(SSqlObj* pSql, void (*queryFp)()) {
 }
 
 void tscGetResultColumnChr(SSqlRes* pRes, SFieldInfo* pFieldInfo, int32_t columnIndex) {
-  SFieldSupInfo* pInfo = tscFieldInfoGetSupp(pFieldInfo, columnIndex);
+  SFieldSupInfo* pInfo = taosArrayGet(pFieldInfo->pSupportInfo, columnIndex);//tscFieldInfoGetSupp(pFieldInfo, columnIndex);
   assert(pInfo->pSqlExpr != NULL);
 
   int32_t type = pInfo->pSqlExpr->resType;

src/common/src/tglobal.c

@@ -112,7 +112,7 @@ int32_t tsMaxShellConns = 5000;
 char    tsDefaultDB[TSDB_DB_NAME_LEN] = {0};
 char    tsDefaultUser[64] = "root";
 char    tsDefaultPass[64] = "taosdata";
-int32_t tsMaxConnections = 50;
+int32_t tsMaxConnections = 5000;
 
 int32_t tsBalanceInterval = 300;  // seconds
 int32_t tsOfflineThreshold = 86400*100;   // seconds 10days
@@ -840,7 +840,7 @@ static void doInitGlobalConfig() {
   cfg.valType = TAOS_CFG_VTYPE_INT32;
   cfg.cfgType = TSDB_CFG_CTYPE_B_CONFIG | TSDB_CFG_CTYPE_B_SHOW;
   cfg.minValue = 1;
-  cfg.maxValue = 100;
+  cfg.maxValue = 100000;
   cfg.ptrLength = 0;
   cfg.unitType = TAOS_CFG_UTYPE_NONE;
   taosInitConfigOption(cfg);

src/connector/python/linux/python2/taos/cinterface.py

@@ -150,6 +150,7 @@ class CTaosInterface(object):
     libtaos.taos_fetch_lengths.restype = ctypes.c_void_p
     libtaos.taos_free_result.restype = None
     libtaos.taos_errno.restype = ctypes.c_int
+    libtaos.taos_query.restype = ctypes.POINTER(ctypes.c_void_p)
 
     def __init__(self, config=None):
         '''

src/connector/python/linux/python3/taos/cinterface.py

@@ -150,6 +150,7 @@ class CTaosInterface(object):
     libtaos.taos_fetch_lengths.restype = ctypes.c_void_p
     libtaos.taos_free_result.restype = None
     libtaos.taos_errno.restype = ctypes.c_int
+    libtaos.taos_query.restype = ctypes.POINTER(ctypes.c_void_p)
 
     def __init__(self, config=None):
         '''

src/connector/python/windows/python2/taos/cinterface.py

@@ -150,6 +150,7 @@ class CTaosInterface(object):
     libtaos.taos_fetch_lengths.restype = ctypes.c_void_p
     libtaos.taos_free_result.restype = None
     libtaos.taos_errno.restype = ctypes.c_int
+    libtaos.taos_query.restype = ctypes.POINTER(ctypes.c_void_p)
 
     def __init__(self, config=None):
         '''

src/connector/python/windows/python3/taos/cinterface.py

@@ -150,6 +150,7 @@ class CTaosInterface(object):
     libtaos.taos_fetch_lengths.restype = ctypes.c_void_p
     libtaos.taos_free_result.restype = None
     libtaos.taos_errno.restype = ctypes.c_int
+    libtaos.taos_query.restype = ctypes.POINTER(ctypes.c_void_p)
 
     def __init__(self, config=None):
         '''

src/inc/tsdb.h

@@ -188,9 +188,10 @@ typedef void *TsdbPosT;
  * @param tsdb       tsdb handle
  * @param pCond      query condition, including time window, result set order, and basic required columns for each block
  * @param groupInfo  tableId list in the form of set, seperated into different groups according to group by condition
+ * @param qinfo      query info handle from query processor
  * @return
  */
-TsdbQueryHandleT *tsdbQueryTables(TsdbRepoT *tsdb, STsdbQueryCond *pCond, STableGroupInfo *groupInfo);
+TsdbQueryHandleT *tsdbQueryTables(TsdbRepoT *tsdb, STsdbQueryCond *pCond, STableGroupInfo *groupInfo, void* qinfo);
 
 /**
  * Get the last row of the given query time window for all the tables in STableGroupInfo object.
@@ -202,11 +203,11 @@ TsdbQueryHandleT *tsdbQueryTables(TsdbRepoT *tsdb, STsdbQueryCond *pCond, STable
  * @param groupInfo   tableId list.
  * @return
  */
-TsdbQueryHandleT tsdbQueryLastRow(TsdbRepoT *tsdb, STsdbQueryCond *pCond, STableGroupInfo *groupInfo);
+TsdbQueryHandleT tsdbQueryLastRow(TsdbRepoT *tsdb, STsdbQueryCond *pCond, STableGroupInfo *groupInfo, void* qinfo);
 
 SArray* tsdbGetQueriedTableIdList(TsdbQueryHandleT *pHandle);
 
-TsdbQueryHandleT tsdbQueryRowsInExternalWindow(TsdbRepoT *tsdb, STsdbQueryCond* pCond, STableGroupInfo *groupList);
+TsdbQueryHandleT tsdbQueryRowsInExternalWindow(TsdbRepoT *tsdb, STsdbQueryCond* pCond, STableGroupInfo *groupList, void* qinfo);
 
 /**
  * move to next block if exists

src/mnode/src/mnodeTable.c

@@ -1284,7 +1284,7 @@ static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) {
   char *msg = (char *)pRsp + sizeof(SCMSTableVgroupRspMsg);
 
   for (int32_t i = 0; i < numOfTable; ++i) {
-    char *stableName = (char*)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_ID_LEN) * i;
+    char *          stableName = (char *)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_ID_LEN)*i;
     SSuperTableObj *pTable = mnodeGetSuperTable(stableName);
     if (pTable == NULL) {
       mError("stable:%s, not exist while get stable vgroup info", stableName);
@@ -1294,41 +1294,48 @@ static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) {
     if (pTable->vgHash == NULL) {
       mError("stable:%s, not vgroup exist while get stable vgroup info", stableName);
       mnodeDecTableRef(pTable);
-      continue;
-    }
 
-    SVgroupsInfo *pVgroupInfo = (SVgroupsInfo *)msg;
+      // even this super table has no corresponding table, still return
+      pRsp->numOfTables++;
 
-    SHashMutableIterator *pIter = taosHashCreateIter(pTable->vgHash);
+      SVgroupsInfo *pVgroupInfo = (SVgroupsInfo *)msg;
-    int32_t vgSize = 0;
+      pVgroupInfo->numOfVgroups = 0;
-    while (taosHashIterNext(pIter)) {
+      
-      int32_t *pVgId = taosHashIterGet(pIter);
+      msg += sizeof(SVgroupsInfo);
-      SVgObj * pVgroup = mnodeGetVgroup(*pVgId);
+    } else {
-      if (pVgroup == NULL) continue;
+      SVgroupsInfo *pVgroupInfo = (SVgroupsInfo *)msg;
 
-      pVgroupInfo->vgroups[vgSize].vgId = htonl(pVgroup->vgId);
+      SHashMutableIterator *pIter = taosHashCreateIter(pTable->vgHash);
-      for (int32_t vn = 0; vn < pVgroup->numOfVnodes; ++vn) {
+      int32_t               vgSize = 0;
-        SDnodeObj *pDnode = pVgroup->vnodeGid[vn].pDnode;
+      while (taosHashIterNext(pIter)) {
-        if (pDnode == NULL) break;
+        int32_t *pVgId = taosHashIterGet(pIter);
+        SVgObj * pVgroup = mnodeGetVgroup(*pVgId);
+        if (pVgroup == NULL) continue;
 
-        strncpy(pVgroupInfo->vgroups[vgSize].ipAddr[vn].fqdn, pDnode->dnodeFqdn, tListLen(pDnode->dnodeFqdn));
+        pVgroupInfo->vgroups[vgSize].vgId = htonl(pVgroup->vgId);
-        pVgroupInfo->vgroups[vgSize].ipAddr[vn].port = htons(pDnode->dnodePort);
+        for (int32_t vn = 0; vn < pVgroup->numOfVnodes; ++vn) {
+          SDnodeObj *pDnode = pVgroup->vnodeGid[vn].pDnode;
+          if (pDnode == NULL) break;
 
-        pVgroupInfo->vgroups[vgSize].numOfIps++;
+          strncpy(pVgroupInfo->vgroups[vgSize].ipAddr[vn].fqdn, pDnode->dnodeFqdn, tListLen(pDnode->dnodeFqdn));
+          pVgroupInfo->vgroups[vgSize].ipAddr[vn].port = htons(pDnode->dnodePort);
+
+          pVgroupInfo->vgroups[vgSize].numOfIps++;
+        }
+
+        vgSize++;
+        mnodeDecVgroupRef(pVgroup);
       }
 
-      vgSize++;
+      taosHashDestroyIter(pIter);
-      mnodeDecVgroupRef(pVgroup);
+      mnodeDecTableRef(pTable);
+
+      pVgroupInfo->numOfVgroups = htonl(vgSize);
+
+      // one table is done, try the next table
+      msg += sizeof(SVgroupsInfo) + vgSize * sizeof(SCMVgroupInfo);
+      pRsp->numOfTables++;
     }
-
-    taosHashDestroyIter(pIter);
-    mnodeDecTableRef(pTable);
-
-    pVgroupInfo->numOfVgroups = htonl(vgSize);
-
-    // one table is done, try the next table
-    msg += sizeof(SVgroupsInfo) + vgSize * sizeof(SCMVgroupInfo);
-    pRsp->numOfTables++;
   }
 
   if (pRsp->numOfTables != numOfTable) {

src/query/inc/qExecutor.h

@@ -172,10 +172,11 @@ typedef struct SQueryRuntimeEnv {
   STSBuf*              pTSBuf;
   STSCursor            cur;
   SQueryCostInfo       summary;
-  bool                 stableQuery;  // super table query or not
+  bool                 stableQuery;      // super table query or not
   void*                pQueryHandle;
   void*                pSecQueryHandle;  // another thread for
   SDiskbasedResultBuf* pResultBuf;       // query result buffer based on blocked-wised disk file
+  bool                 topBotQuery;      // false;
 } SQueryRuntimeEnv;
 
 typedef struct SQInfo {

src/query/inc/qUtil.h

@@ -31,7 +31,8 @@ void    closeTimeWindow(SWindowResInfo* pWindowResInfo, int32_t slot);
 void    closeAllTimeWindow(SWindowResInfo* pWindowResInfo);
 void    removeRedundantWindow(SWindowResInfo *pWindowResInfo, TSKEY lastKey, int32_t order);
 SWindowResult *getWindowResult(SWindowResInfo *pWindowResInfo, int32_t slot);
-int32_t curTimeWindow(SWindowResInfo *pWindowResInfo);
+
+#define curTimeWindow(_winres)  ((_winres)->curIndex)
 bool isWindowResClosed(SWindowResInfo *pWindowResInfo, int32_t slot);
 
 void createQueryResultInfo(SQuery *pQuery, SWindowResult *pResultRow, bool isSTableQuery, SPosInfo *posInfo);

src/query/inc/qextbuffer.h

@@ -28,7 +28,7 @@ extern "C" {
 #include "tdataformat.h"
 #include "talgo.h"
 
-#define DEFAULT_PAGE_SIZE (1024L*56)  // 16k larger than the SHistoInfo
+#define DEFAULT_PAGE_SIZE (1024L*64)  // 16k larger than the SHistoInfo
 #define MAX_TMPFILE_PATH_LENGTH PATH_MAX
 #define INITIAL_ALLOCATION_BUFFER_SIZE 64
 
@@ -96,7 +96,7 @@ typedef struct SColumnOrderInfo {
 typedef struct tOrderDescriptor {
   SColumnModel *   pColumnModel;
   int32_t          tsOrder;  // timestamp order type if exists
-  SColumnOrderInfo orderIdx;
+  SColumnOrderInfo orderInfo;
 } tOrderDescriptor;
 
 typedef struct tExtMemBuffer {

src/query/inc/qresultBuf.h

@@ -85,7 +85,7 @@ SIDList getDataBufPagesIdList(SDiskbasedResultBuf* pResultBuf, int32_t groupId);
  * @param id
  * @return
  */
-tFilePage* getResultBufferPageById(SDiskbasedResultBuf* pResultBuf, int32_t id);
+#define GET_RES_BUF_PAGE_BY_ID(buf, id)  ((tFilePage*)((buf)->pBuf + DEFAULT_INTERN_BUF_PAGE_SIZE*(id)))
 
 /**
  * get the total buffer size in the format of disk file

src/query/inc/sql.y

@@ -647,9 +647,9 @@ cmd ::= ALTER TABLE ids(X) cpxName(F) SET TAG ids(Y) EQ tagitem(Z).     {
 }
 
 ////////////////////////////////////////kill statement///////////////////////////////////////
-cmd ::= KILL CONNECTION IPTOKEN(X) COLON(Z) INTEGER(Y).   {X.n += (Z.n + Y.n); setKillSQL(pInfo, TSDB_SQL_KILL_CONNECTION, &X);}
+cmd ::= KILL CONNECTION INTEGER(Y).   {setKillSQL(pInfo, TSDB_SQL_KILL_CONNECTION, &Y);}
-cmd ::= KILL STREAM IPTOKEN(X) COLON(Z) INTEGER(Y) COLON(K) INTEGER(F).       {X.n += (Z.n + Y.n + K.n + F.n); setKillSQL(pInfo, TSDB_SQL_KILL_STREAM, &X);}
+cmd ::= KILL STREAM INTEGER(X) COLON(Z) INTEGER(Y).       {X.n += (Z.n + Y.n); setKillSQL(pInfo, TSDB_SQL_KILL_STREAM, &X);}
-cmd ::= KILL QUERY IPTOKEN(X) COLON(Z) INTEGER(Y) COLON(K) INTEGER(F).        {X.n += (Z.n + Y.n + K.n + F.n); setKillSQL(pInfo, TSDB_SQL_KILL_QUERY, &X);}
+cmd ::= KILL QUERY INTEGER(X) COLON(Z) INTEGER(Y).        {X.n += (Z.n + Y.n); setKillSQL(pInfo, TSDB_SQL_KILL_QUERY, &X);}
 
 %fallback ID ABORT AFTER ASC ATTACH BEFORE BEGIN CASCADE CLUSTER CONFLICT COPY DATABASE DEFERRED
   DELIMITERS DESC DETACH EACH END EXPLAIN FAIL FOR GLOB IGNORE IMMEDIATE INITIALLY INSTEAD

src/query/inc/tsqlfunction.h

@@ -272,9 +272,18 @@ bool top_bot_datablock_filter(SQLFunctionCtx *pCtx, int32_t functionId, char *mi
 bool stableQueryFunctChanged(int32_t funcId);
 
 void resetResultInfo(SResultInfo *pResInfo);
-void initResultInfo(SResultInfo *pResInfo);
 void setResultInfoBuf(SResultInfo *pResInfo, int32_t size, bool superTable);
 
+static FORCE_INLINE void initResultInfo(SResultInfo *pResInfo) {
+  pResInfo->initialized = true;  // the this struct has been initialized flag
+  
+  pResInfo->complete = false;
+  pResInfo->hasResult = false;
+  pResInfo->numOfRes = 0;
+  
+  memset(pResInfo->interResultBuf, 0, (size_t)pResInfo->bufLen);
+}
+
 #ifdef __cplusplus
 }
 #endif

src/query/src/qExecutor.c

@@ -98,6 +98,7 @@ static bool isIntervalQuery(SQuery *pQuery) { return pQuery->intervalTime > 0; }
 static int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *group);
 
 static void setWindowResOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult);
+static void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult);
 static void resetMergeResultBuf(SQuery *pQuery, SQLFunctionCtx *pCtx, SResultInfo *pResultInfo);
 static bool functionNeedToExecute(SQueryRuntimeEnv *pRuntimeEnv, SQLFunctionCtx *pCtx, int32_t functionId);
 static void getNextTimeWindow(SQuery *pQuery, STimeWindow *pTimeWindow);
@@ -441,7 +442,7 @@ static int32_t addNewWindowResultBuf(SWindowResult *pWindowRes, SDiskbasedResult
     pData = getNewDataBuf(pResultBuf, sid, &pageId);
   } else {
     pageId = getLastPageId(&list);
-    pData = getResultBufferPageById(pResultBuf, pageId);
+    pData = GET_RES_BUF_PAGE_BY_ID(pResultBuf, pageId);
 
     if (pData->num >= numOfRowsPerPage) {
       pData = getNewDataBuf(pResultBuf, sid, &pageId);
@@ -484,10 +485,8 @@ static int32_t setWindowOutputBufByKey(SQueryRuntimeEnv *pRuntimeEnv, SWindowRes
 
   // set time window for current result
   pWindowRes->window = *win;
-
+  
-  setWindowResOutputBuf(pRuntimeEnv, pWindowRes);
+  setWindowResOutputBufInitCtx(pRuntimeEnv, pWindowRes);
-  initCtxOutputBuf(pRuntimeEnv);
-
   return TSDB_CODE_SUCCESS;
 }
 
@@ -651,7 +650,7 @@ static void doBlockwiseApplyFunctions(SQueryRuntimeEnv *pRuntimeEnv, SWindowStat
       pCtx[k].startOffset = (QUERY_IS_ASC_QUERY(pQuery)) ? offset : offset - (forwardStep - 1);
 
       if ((aAggs[functionId].nStatus & TSDB_FUNCSTATE_SELECTIVITY) != 0) {
-        pCtx[k].ptsList = tsBuf;
+        pCtx[k].ptsList = &tsBuf[offset];
       }
 
       // not a whole block involved in query processing, statistics data can not be used
@@ -687,47 +686,52 @@ static int32_t getNextQualifiedWindow(SQueryRuntimeEnv *pRuntimeEnv, STimeWindow
                                       SDataBlockInfo *pDataBlockInfo, TSKEY *primaryKeys,
                                       __block_search_fn_t searchFn) {
   SQuery *pQuery = pRuntimeEnv->pQuery;
-
+  
-  while (1) {
+  // tumbling time window query, a special case of sliding time window query
-    if ((pNextWin->ekey > pQuery->window.ekey && QUERY_IS_ASC_QUERY(pQuery)) ||
+  if (pQuery->slidingTime == pQuery->intervalTime) {
-        (pNextWin->skey < pQuery->window.ekey && !QUERY_IS_ASC_QUERY(pQuery))) {
+    // todo opt
-      return -1;
-    }
-
-    getNextTimeWindow(pQuery, pNextWin);
-
-    // next time window is not in current block
-    if ((pNextWin->skey > pDataBlockInfo->window.ekey && QUERY_IS_ASC_QUERY(pQuery)) ||
-        (pNextWin->ekey < pDataBlockInfo->window.skey && !QUERY_IS_ASC_QUERY(pQuery))) {
-      return -1;
-    }
-
-    TSKEY startKey = -1;
-    if (QUERY_IS_ASC_QUERY(pQuery)) {
-      startKey = pNextWin->skey;
-      if (startKey < pQuery->window.skey) {
-        startKey = pQuery->window.skey;
-      }
-    } else {
-      startKey = pNextWin->ekey;
-      if (startKey > pQuery->window.skey) {
-        startKey = pQuery->window.skey;
-      }
-    }
-
-    int32_t startPos = searchFn((char *)primaryKeys, pDataBlockInfo->rows, startKey, pQuery->order.order);
-
-    /*
-     * This time window does not cover any data, try next time window,
-     * this case may happen when the time window is too small
-     */
-    if ((primaryKeys[startPos] > pNextWin->ekey && QUERY_IS_ASC_QUERY(pQuery)) ||
-        (primaryKeys[startPos] < pNextWin->skey && !QUERY_IS_ASC_QUERY(pQuery))) {
-      continue;
-    }
-
-    return startPos;
   }
+  
+  getNextTimeWindow(pQuery, pNextWin);
+  
+  // next time window is not in current block
+  if ((pNextWin->skey > pDataBlockInfo->window.ekey && QUERY_IS_ASC_QUERY(pQuery)) ||
+      (pNextWin->ekey < pDataBlockInfo->window.skey && !QUERY_IS_ASC_QUERY(pQuery))) {
+    return -1;
+  }
+
+  TSKEY startKey = -1;
+  if (QUERY_IS_ASC_QUERY(pQuery)) {
+    startKey = pNextWin->skey;
+    if (startKey < pQuery->window.skey) {
+      startKey = pQuery->window.skey;
+    }
+  } else {
+    startKey = pNextWin->ekey;
+    if (startKey > pQuery->window.skey) {
+      startKey = pQuery->window.skey;
+    }
+  }
+
+  int32_t startPos = searchFn((char *)primaryKeys, pDataBlockInfo->rows, startKey, pQuery->order.order);
+
+  /*
+   * This time window does not cover any data, try next time window,
+   * this case may happen when the time window is too small
+   */
+  if (QUERY_IS_ASC_QUERY(pQuery) && primaryKeys[startPos] > pNextWin->ekey) {
+    TSKEY next = primaryKeys[startPos];
+    
+    pNextWin->ekey += ((next - pNextWin->ekey + pQuery->slidingTime - 1)/pQuery->slidingTime) * pQuery->slidingTime;
+    pNextWin->skey = pNextWin->ekey - pQuery->intervalTime + 1;
+  } else if ((!QUERY_IS_ASC_QUERY(pQuery)) && primaryKeys[startPos] < pNextWin->skey) {
+    TSKEY next = primaryKeys[startPos];
+
+    pNextWin->skey -= ((pNextWin->skey - next + pQuery->slidingTime - 1) / pQuery->slidingTime) * pQuery->slidingTime;
+    pNextWin->ekey = pNextWin->skey + pQuery->intervalTime - 1;
+  }
+  
+  return startPos;
 }
 
 static TSKEY reviseWindowEkey(SQuery *pQuery, STimeWindow *pWindow) {
@@ -1027,7 +1031,7 @@ static bool functionNeedToExecute(SQueryRuntimeEnv *pRuntimeEnv, SQLFunctionCtx
   }
   
   // in the supplementary scan, only the following functions need to be executed
-  if (IS_REVERSE_SCAN(pRuntimeEnv)) {// && (functionId == TSDB_FUNC_TAG || functionId == TSDB_FUNC_TS)) {
+  if (IS_REVERSE_SCAN(pRuntimeEnv)) {
     return false;
   }
 
@@ -1183,7 +1187,6 @@ static int32_t tableApplyFunctionsOnBlock(SQueryRuntimeEnv *pRuntimeEnv, SDataBl
   
   STableQueryInfo* pTableQInfo = pQuery->current;
   SWindowResInfo*  pWindowResInfo = &pRuntimeEnv->windowResInfo;
-  pQuery->pos = QUERY_IS_ASC_QUERY(pQuery)? 0 : pDataBlockInfo->rows - 1;
   
   if (pQuery->numOfFilterCols > 0 || pRuntimeEnv->pTSBuf != NULL || isGroupbyNormalCol(pQuery->pGroupbyExpr)) {
     rowwiseApplyFunctions(pRuntimeEnv, pStatis, pDataBlockInfo, pWindowResInfo, pDataBlock);
@@ -1598,8 +1601,7 @@ static bool onlyQueryTags(SQuery* pQuery) {
 
 /////////////////////////////////////////////////////////////////////////////////////////////
 
-void getAlignQueryTimeWindow(SQuery *pQuery, int64_t key, int64_t keyFirst, int64_t keyLast, int64_t *realSkey,
+void getAlignQueryTimeWindow(SQuery *pQuery, int64_t key, int64_t keyFirst, int64_t keyLast, STimeWindow *realWin, STimeWindow *win) {
-                             int64_t *realEkey, STimeWindow *win) {
   assert(key >= keyFirst && key <= keyLast && pQuery->slidingTime <= pQuery->intervalTime);
 
   win->skey = taosGetIntervalStartTimestamp(key, pQuery->slidingTime, pQuery->slidingTimeUnit, pQuery->precision);
@@ -1611,8 +1613,8 @@ void getAlignQueryTimeWindow(SQuery *pQuery, int64_t key, int64_t keyFirst, int6
      */
     assert(keyLast - keyFirst < pQuery->intervalTime);
 
-    *realSkey = keyFirst;
+    realWin->skey = keyFirst;
-    *realEkey = keyLast;
+    realWin->ekey = keyLast;
 
     win->ekey = INT64_MAX;
     return;
@@ -1620,17 +1622,8 @@ void getAlignQueryTimeWindow(SQuery *pQuery, int64_t key, int64_t keyFirst, int6
 
   win->ekey = win->skey + pQuery->intervalTime - 1;
 
-  if (win->skey < keyFirst) {
+  realWin->skey = (win->skey < keyFirst)? keyFirst : win->skey;
-    *realSkey = keyFirst;
+  realWin->ekey = (win->ekey < keyLast) ? win->ekey : keyLast;
-  } else {
-    *realSkey = win->skey;
-  }
-
-  if (win->ekey < keyLast) {
-    *realEkey = win->ekey;
-  } else {
-    *realEkey = keyLast;
-  }
 }
 
 static void setScanLimitationByResultBuffer(SQuery *pQuery) {
@@ -1847,31 +1840,21 @@ static int32_t getInitialPageNum(SQInfo *pQInfo) {
   return num;
 }
 
-static int32_t getRowParamForMultiRowsOutput(SQuery *pQuery, bool isSTableQuery) {
+#define GET_ROW_PARAM_FOR_MULTIOUTPUT(_q, tbq, sq) (((tbq) && (!sq))? (_q)->pSelectExpr[1].base.arg->argValue.i64:1)
-  int32_t rowparam = 1;
 
-  if (isTopBottomQuery(pQuery) && (!isSTableQuery)) {
+static FORCE_INLINE int32_t getNumOfRowsInResultPage(SQuery *pQuery, bool topBotQuery, bool isSTableQuery) {
-    rowparam = pQuery->pSelectExpr[1].base.arg->argValue.i64;
+  int32_t rowSize = pQuery->rowSize * GET_ROW_PARAM_FOR_MULTIOUTPUT(pQuery, topBotQuery, isSTableQuery);
-  }
-
-  return rowparam;
-}
-
-static int32_t getNumOfRowsInResultPage(SQuery *pQuery, bool isSTableQuery) {
-  int32_t rowSize = pQuery->rowSize * getRowParamForMultiRowsOutput(pQuery, isSTableQuery);
   return (DEFAULT_INTERN_BUF_PAGE_SIZE - sizeof(tFilePage)) / rowSize;
 }
 
 char *getPosInResultPage(SQueryRuntimeEnv *pRuntimeEnv, int32_t columnIndex, SWindowResult *pResult) {
   assert(pResult != NULL && pRuntimeEnv != NULL);
 
-  SQuery *   pQuery = pRuntimeEnv->pQuery;
+  SQuery    *pQuery = pRuntimeEnv->pQuery;
-  tFilePage *page = getResultBufferPageById(pRuntimeEnv->pResultBuf, pResult->pos.pageId);
+  tFilePage *page = GET_RES_BUF_PAGE_BY_ID(pRuntimeEnv->pResultBuf, pResult->pos.pageId);
+  int32_t realRowId = pResult->pos.rowId * GET_ROW_PARAM_FOR_MULTIOUTPUT(pQuery, pRuntimeEnv->topBotQuery, pRuntimeEnv->stableQuery);
 
-  int32_t numOfRows = getNumOfRowsInResultPage(pQuery, pRuntimeEnv->stableQuery);
+  return ((char *)page->data) + pRuntimeEnv->offset[columnIndex] * pRuntimeEnv->numOfRowsPerPage +
-  int32_t realRowId = pResult->pos.rowId * getRowParamForMultiRowsOutput(pQuery, pRuntimeEnv->stableQuery);
-
-  return ((char *)page->data) + pRuntimeEnv->offset[columnIndex] * numOfRows +
          pQuery->pSelectExpr[columnIndex].bytes * realRowId;
 }
 
@@ -2034,6 +2017,7 @@ SArray *loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, void* pQueryHandle,
     }
   
     pRuntimeEnv->summary.totalCheckedRows += pBlockInfo->rows;
+    pRuntimeEnv->summary.loadBlocks += 1;
     pDataBlock = tsdbRetrieveDataBlock(pQueryHandle, NULL);
   }
 
@@ -2143,6 +2127,7 @@ static void ensureOutputBuffer(SQueryRuntimeEnv* pRuntimeEnv, SDataBlockInfo* pB
 static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
   SQuery *pQuery = pRuntimeEnv->pQuery;
   STableQueryInfo* pTableQueryInfo = pQuery->current;
+  SQueryCostInfo*  summary  = &pRuntimeEnv->summary;
   
   qTrace("QInfo:%p query start, qrange:%" PRId64 "-%" PRId64 ", lastkey:%" PRId64 ", order:%d",
          GET_QINFO_ADDR(pRuntimeEnv), pTableQueryInfo->win.skey, pTableQueryInfo->win.ekey, pTableQueryInfo->lastKey,
@@ -2150,7 +2135,7 @@ static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
 
   TsdbQueryHandleT pQueryHandle = IS_MASTER_SCAN(pRuntimeEnv)? pRuntimeEnv->pQueryHandle : pRuntimeEnv->pSecQueryHandle;
   while (tsdbNextDataBlock(pQueryHandle)) {
-    pRuntimeEnv->summary.totalBlocks += 1;
+    summary->totalBlocks += 1;
     if (isQueryKilled(GET_QINFO_ADDR(pRuntimeEnv))) {
       return 0;
     }
@@ -2159,19 +2144,16 @@ static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
 
     // todo extract methods
     if (isIntervalQuery(pQuery) && pRuntimeEnv->windowResInfo.prevSKey == TSKEY_INITIAL_VAL) {
-      TSKEY           skey1, ekey1;
+      STimeWindow realWin = TSWINDOW_INITIALIZER, w = TSWINDOW_INITIALIZER;
-      STimeWindow     w = TSWINDOW_INITIALIZER;
       SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo;
 
       if (QUERY_IS_ASC_QUERY(pQuery)) {
-        getAlignQueryTimeWindow(pQuery, blockInfo.window.skey, blockInfo.window.skey, pQuery->window.ekey, &skey1,
+        getAlignQueryTimeWindow(pQuery, blockInfo.window.skey, blockInfo.window.skey, pQuery->window.ekey, &realWin, &w);
-                                &ekey1, &w);
         pWindowResInfo->startTime = w.skey;
         pWindowResInfo->prevSKey = w.skey;
       } else {
         // the start position of the first time window in the endpoint that spreads beyond the queried last timestamp
-        getAlignQueryTimeWindow(pQuery, blockInfo.window.ekey, pQuery->window.ekey, blockInfo.window.ekey, &skey1,
+        getAlignQueryTimeWindow(pQuery, blockInfo.window.ekey, pQuery->window.ekey, blockInfo.window.ekey, &realWin, &w);
-                                &ekey1, &w);
 
         pWindowResInfo->startTime = pQuery->window.skey;
         pWindowResInfo->prevSKey = w.skey;
@@ -2187,9 +2169,12 @@ static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
 
     SDataStatis *pStatis = NULL;
     SArray *pDataBlock = loadDataBlockOnDemand(pRuntimeEnv, pQueryHandle, &blockInfo, &pStatis);
+  
+    // query start position can not move into tableApplyFunctionsOnBlock due to limit/offset condition
+    pQuery->pos = QUERY_IS_ASC_QUERY(pQuery)? 0 : blockInfo.rows - 1;
     int32_t numOfRes = tableApplyFunctionsOnBlock(pRuntimeEnv, &blockInfo, pStatis, binarySearchForKey, pDataBlock);
 
-    pRuntimeEnv->summary.totalRows += blockInfo.rows;
+    summary->totalRows += blockInfo.rows;
     qTrace("QInfo:%p check data block, brange:%" PRId64 "-%" PRId64 ", numOfRows:%d, numOfRes:%d, lastKey:%"PRId64, GET_QINFO_ADDR(pRuntimeEnv),
            blockInfo.window.skey, blockInfo.window.ekey, blockInfo.rows, numOfRes, pQuery->current->lastKey);
 
@@ -2531,7 +2516,7 @@ void copyResToQueryResultBuf(SQInfo *pQInfo, SQuery *pQuery) {
 
   int32_t total = 0;
   for (int32_t i = 0; i < list.size; ++i) {
-    tFilePage *pData = getResultBufferPageById(pResultBuf, list.pData[i]);
+    tFilePage *pData = GET_RES_BUF_PAGE_BY_ID(pResultBuf, list.pData[i]);
     total += pData->num;
   }
 
@@ -2539,7 +2524,7 @@ void copyResToQueryResultBuf(SQInfo *pQInfo, SQuery *pQuery) {
 
   int32_t offset = 0;
   for (int32_t num = 0; num < list.size; ++num) {
-    tFilePage *pData = getResultBufferPageById(pResultBuf, list.pData[num]);
+    tFilePage *pData = GET_RES_BUF_PAGE_BY_ID(pResultBuf, list.pData[num]);
 
     for (int32_t i = 0; i < pQuery->numOfOutput; ++i) {
       int32_t bytes = pRuntimeEnv->pCtx[i].outputBytes;
@@ -2558,10 +2543,8 @@ void copyResToQueryResultBuf(SQInfo *pQInfo, SQuery *pQuery) {
   pQInfo->offset += 1;
 }
 
-int64_t getNumOfResultWindowRes(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pWindowRes) {
+int64_t getNumOfResultWindowRes(SQuery *pQuery, SWindowResult *pWindowRes) {
-  SQuery *pQuery = pRuntimeEnv->pQuery;
+//  int64_t maxOutput = 0;
-
-  int64_t maxOutput = 0;
   for (int32_t j = 0; j < pQuery->numOfOutput; ++j) {
     int32_t functionId = pQuery->pSelectExpr[j].base.functionId;
 
@@ -2574,12 +2557,23 @@ int64_t getNumOfResultWindowRes(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pW
     }
 
     SResultInfo *pResultInfo = &pWindowRes->resultInfo[j];
-    if (pResultInfo != NULL && maxOutput < pResultInfo->numOfRes) {
+    assert(pResultInfo != NULL);
-      maxOutput = pResultInfo->numOfRes;
+    
+    if (pResultInfo->numOfRes > 0) {
+      return pResultInfo->numOfRes;
     }
+//    if (pResultInfo != NULL && maxOutput < pResultInfo->numOfRes) {
+//      maxOutput = pResultInfo->numOfRes;
+//
+//      if (maxOutput > 0) {
+//        break;
+//      }
+//    }
+//
+//    assert(pResultInfo != NULL);
   }
 
-  return maxOutput;
+  return 0;
 }
 
 int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) {
@@ -2612,6 +2606,8 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) {
 
     assert(pQInfo->numOfGroupResultPages == 0);
     return 0;
+  } else if (numOfTables == 1) { // no need to merge results since only one table in each group
+  
   }
 
   SCompSupporter cs = {pTableList, posList, pQInfo};
@@ -2636,7 +2632,7 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) {
     TSKEY ts = GET_INT64_VAL(b);
 
     assert(ts == pWindowRes->window.skey);
-    int64_t num = getNumOfResultWindowRes(pRuntimeEnv, pWindowRes);
+    int64_t num = getNumOfResultWindowRes(pQuery, pWindowRes);
     if (num <= 0) {
       cs.position[pos] += 1;
 
@@ -2699,10 +2695,11 @@ int32_t mergeIntoGroupResultImpl(SQInfo *pQInfo, SArray *pGroup) {
   displayInterResult(pQuery->sdata, pRuntimeEnv, pQuery->sdata[0]->num);
 #endif
 
-  qTrace("QInfo:%p result merge completed, elapsed time:%" PRId64 " ms", GET_QINFO_ADDR(pQuery), endt - startt);
+  qTrace("QInfo:%p result merge completed for group:%d, elapsed time:%" PRId64 " ms", pQInfo, pQInfo->groupIndex, endt - startt);
-  tfree(pTree);
+  
   tfree(pTableList);
   tfree(posList);
+  tfree(pTree);
 
   pQInfo->offset = 0;
   for (int32_t i = 0; i < pQuery->numOfOutput; ++i) {
@@ -2928,8 +2925,13 @@ void initCtxOutputBuf(SQueryRuntimeEnv *pRuntimeEnv) {
 
   for (int32_t j = 0; j < pQuery->numOfOutput; ++j) {
     int32_t functionId = pQuery->pSelectExpr[j].base.functionId;
-
     pRuntimeEnv->pCtx[j].currentStage = 0;
+    
+    SResultInfo* pResInfo = GET_RES_INFO(&pRuntimeEnv->pCtx[j]);
+    if (pResInfo->initialized) {
+      continue;
+    }
+    
     aAggs[functionId].init(&pRuntimeEnv->pCtx[j]);
   }
 }
@@ -3078,7 +3080,7 @@ static void setEnvBeforeReverseScan(SQueryRuntimeEnv *pRuntimeEnv, SQueryStatusI
     tsdbCleanupQueryHandle(pRuntimeEnv->pSecQueryHandle);
   }
 
-  pRuntimeEnv->pSecQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &pQInfo->tableIdGroupInfo);
+  pRuntimeEnv->pSecQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &pQInfo->tableIdGroupInfo, pQInfo);
 
   setQueryStatus(pQuery, QUERY_NOT_COMPLETED);
   switchCtxOrder(pRuntimeEnv);
@@ -3150,7 +3152,7 @@ void scanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv, TSKEY start) {
       tsdbCleanupQueryHandle(pRuntimeEnv->pSecQueryHandle);
     }
 
-    pRuntimeEnv->pSecQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &pQInfo->tableIdGroupInfo);
+    pRuntimeEnv->pSecQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &pQInfo->tableIdGroupInfo, pQInfo);
     pRuntimeEnv->windowResInfo.curIndex = qstatus.windowIndex;
 
     setQueryStatus(pQuery, QUERY_NOT_COMPLETED);
@@ -3295,13 +3297,14 @@ void setExecutionContext(SQInfo *pQInfo, STableId* pTableId, int32_t groupIndex,
   setAdditionalInfo(pQInfo, pTableId, pTableQueryInfo);
 }
 
-static void setWindowResOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult) {
+void setWindowResOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult) {
   SQuery *pQuery = pRuntimeEnv->pQuery;
 
   // Note: pResult->pos[i]->num == 0, there is only fixed number of results for each group
   for (int32_t i = 0; i < pQuery->numOfOutput; ++i) {
     SQLFunctionCtx *pCtx = &pRuntimeEnv->pCtx[i];
     pCtx->aOutputBuf = getPosInResultPage(pRuntimeEnv, i, pResult);
+    
     int32_t functionId = pQuery->pSelectExpr[i].base.functionId;
     if (functionId == TSDB_FUNC_TOP || functionId == TSDB_FUNC_BOTTOM || functionId == TSDB_FUNC_DIFF) {
       pCtx->ptsOutputBuf = pRuntimeEnv->pCtx[0].aOutputBuf;
@@ -3319,6 +3322,38 @@ static void setWindowResOutputBuf(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *
   }
 }
 
+void setWindowResOutputBufInitCtx(SQueryRuntimeEnv *pRuntimeEnv, SWindowResult *pResult) {
+  SQuery *pQuery = pRuntimeEnv->pQuery;
+  
+  // Note: pResult->pos[i]->num == 0, there is only fixed number of results for each group
+  for (int32_t i = 0; i < pQuery->numOfOutput; ++i) {
+    SQLFunctionCtx *pCtx = &pRuntimeEnv->pCtx[i];
+
+    pCtx->resultInfo = &pResult->resultInfo[i];
+    if (pCtx->resultInfo->complete) {
+      continue;
+    }
+    
+    pCtx->aOutputBuf = getPosInResultPage(pRuntimeEnv, i, pResult);
+    pCtx->currentStage = 0;
+  
+    int32_t functionId = pCtx->functionId;
+    if (functionId == TSDB_FUNC_TOP || functionId == TSDB_FUNC_BOTTOM || functionId == TSDB_FUNC_DIFF) {
+      pCtx->ptsOutputBuf = pRuntimeEnv->pCtx[0].aOutputBuf;
+    }
+    
+    /*
+     * set the output buffer information and intermediate buffer
+     * not all queries require the interResultBuf, such as COUNT
+     */
+    pCtx->resultInfo->superTableQ = pRuntimeEnv->stableQuery;     // set super table query flag
+  
+    if (!pCtx->resultInfo->initialized) {
+      aAggs[functionId].init(pCtx);
+    }
+  }
+}
+
 int32_t setAdditionalInfo(SQInfo *pQInfo, STableId* pTableId, STableQueryInfo *pTableQueryInfo) {
   SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv;
   assert(pTableQueryInfo->lastKey >= TSKEY_INITIAL_VAL);
@@ -3374,13 +3409,12 @@ void setIntervalQueryRange(SQInfo *pQInfo, TSKEY key) {
      * In ascending query, key is the first qualified timestamp. However, in the descending order query, additional
      * operations involve.
      */
-    TSKEY           skey1, ekey1;
+    STimeWindow     w = TSWINDOW_INITIALIZER, realWin = TSWINDOW_INITIALIZER;
-    STimeWindow     w = {0};
     SWindowResInfo *pWindowResInfo = &pTableQueryInfo->windowResInfo;
 
     TSKEY sk = MIN(win.skey, win.ekey);
     TSKEY ek = MAX(win.skey, win.ekey);
-    getAlignQueryTimeWindow(pQuery, win.skey, sk, ek, &skey1, &ekey1, &w);
+    getAlignQueryTimeWindow(pQuery, win.skey, sk, ek, &realWin, &w);
     pWindowResInfo->startTime = pTableQueryInfo->win.skey;  // windowSKey may be 0 in case of 1970 timestamp
 
     if (pWindowResInfo->prevSKey == TSKEY_INITIAL_VAL) {
@@ -3738,7 +3772,7 @@ static void updateOffsetVal(SQueryRuntimeEnv *pRuntimeEnv, SDataBlockInfo *pBloc
   SColumnInfoData *pColInfoData = taosArrayGet(pDataBlock, 0);
 
   // update the pQuery->limit.offset value, and pQuery->pos value
-  TSKEY *keys = (TSKEY *)pColInfoData->pData;
+  TSKEY *keys = (TSKEY *) pColInfoData->pData;
 
   // update the offset value
   pTableQueryInfo->lastKey = keys[pQuery->pos];
@@ -3800,8 +3834,7 @@ static bool skipTimeInterval(SQueryRuntimeEnv *pRuntimeEnv, TSKEY* start) {
    */
   assert(pRuntimeEnv->windowResInfo.prevSKey == TSKEY_INITIAL_VAL);
 
-  TSKEY       skey1, ekey1;
+  STimeWindow w = TSWINDOW_INITIALIZER, realWin = TSWINDOW_INITIALIZER;
-  STimeWindow w = TSWINDOW_INITIALIZER;
   
   SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo;
   STableQueryInfo *pTableQueryInfo = pQuery->current;
@@ -3811,14 +3844,12 @@ static bool skipTimeInterval(SQueryRuntimeEnv *pRuntimeEnv, TSKEY* start) {
 
     if (QUERY_IS_ASC_QUERY(pQuery)) {
       if (pWindowResInfo->prevSKey == TSKEY_INITIAL_VAL) {
-        getAlignQueryTimeWindow(pQuery, blockInfo.window.skey, blockInfo.window.skey, pQuery->window.ekey, &skey1,
+        getAlignQueryTimeWindow(pQuery, blockInfo.window.skey, blockInfo.window.skey, pQuery->window.ekey, &realWin, &w);
-                                &ekey1, &w);
         pWindowResInfo->startTime = w.skey;
         pWindowResInfo->prevSKey = w.skey;
       }
     } else {
-      getAlignQueryTimeWindow(pQuery, blockInfo.window.ekey, pQuery->window.ekey, blockInfo.window.ekey, &skey1, &ekey1,
+      getAlignQueryTimeWindow(pQuery, blockInfo.window.ekey, pQuery->window.ekey, blockInfo.window.ekey, &realWin, &w);
-                              &w);
 
       pWindowResInfo->startTime = pQuery->window.skey;
       pWindowResInfo->prevSKey = w.skey;
@@ -3939,11 +3970,11 @@ static void setupQueryHandle(void* tsdb, SQInfo* pQInfo, bool isSTableQuery) {
   }
 
   if (isFirstLastRowQuery(pQuery)) {
-    pRuntimeEnv->pQueryHandle = tsdbQueryLastRow(tsdb, &cond, &pQInfo->tableIdGroupInfo);
+    pRuntimeEnv->pQueryHandle = tsdbQueryLastRow(tsdb, &cond, &pQInfo->tableIdGroupInfo, pQInfo);
   } else if (isPointInterpoQuery(pQuery)) {
-    pRuntimeEnv->pQueryHandle = tsdbQueryRowsInExternalWindow(tsdb, &cond, &pQInfo->tableIdGroupInfo);
+    pRuntimeEnv->pQueryHandle = tsdbQueryRowsInExternalWindow(tsdb, &cond, &pQInfo->tableIdGroupInfo, pQInfo);
   } else {
-    pRuntimeEnv->pQueryHandle = tsdbQueryTables(tsdb, &cond, &pQInfo->tableIdGroupInfo);
+    pRuntimeEnv->pQueryHandle = tsdbQueryTables(tsdb, &cond, &pQInfo->tableIdGroupInfo, pQInfo);
   }
 }
 
@@ -3999,7 +4030,7 @@ int32_t doInitQInfo(SQInfo *pQInfo, void *param, void *tsdb, int32_t vgId, bool
     return code;
   }
 
-  pRuntimeEnv->numOfRowsPerPage = getNumOfRowsInResultPage(pQuery, isSTableQuery);
+  pRuntimeEnv->numOfRowsPerPage = getNumOfRowsInResultPage(pQuery, pRuntimeEnv->topBotQuery, isSTableQuery);
 
   if (isSTableQuery) {
     int32_t rows = getInitialPageNum(pQInfo);
@@ -4057,6 +4088,7 @@ int32_t doInitQInfo(SQInfo *pQInfo, void *param, void *tsdb, int32_t vgId, bool
                                               pQuery->slidingTime, pQuery->fillType, pColInfo);
   }
   
+  pRuntimeEnv->topBotQuery = isTopBottomQuery(pQuery);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -4086,10 +4118,10 @@ static void enableExecutionForNextTable(SQueryRuntimeEnv *pRuntimeEnv) {
   }
 }
 
-static int64_t queryOnDataBlocks(SQInfo *pQInfo) {
+static int64_t scanMultiTableDataBlocks(SQInfo *pQInfo) {
   SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv;
-  SQuery *          pQuery = pRuntimeEnv->pQuery;
+  SQuery*           pQuery = pRuntimeEnv->pQuery;
-  SQueryCostInfo* summary  = &pRuntimeEnv->summary;
+  SQueryCostInfo*   summary  = &pRuntimeEnv->summary;
   
   int64_t st = taosGetTimestampMs();
 
@@ -4144,8 +4176,9 @@ static int64_t queryOnDataBlocks(SQInfo *pQInfo) {
     summary->totalRows += blockInfo.rows;
     stableApplyFunctionsOnBlock(pRuntimeEnv, &blockInfo, pStatis, pDataBlock, binarySearchForKey);
   
-    qTrace("QInfo:%p check data block, brange:%" PRId64 "-%" PRId64 ", numOfRows:%d, lastKey:%"PRId64, GET_QINFO_ADDR(pRuntimeEnv),
+    qTrace("QInfo:%p check data block, uid:%"PRId64", tid:%d, brange:%" PRId64 "-%" PRId64 ", numOfRows:%d, lastKey:%" PRId64,
-           blockInfo.window.skey, blockInfo.window.ekey, blockInfo.rows, pQuery->current->lastKey);
+           GET_QINFO_ADDR(pRuntimeEnv), blockInfo.uid, blockInfo.tid, blockInfo.window.skey, blockInfo.window.ekey,
+           blockInfo.rows, pQuery->current->lastKey);
   }
 
   int64_t et = taosGetTimestampMs();
@@ -4186,7 +4219,7 @@ static bool multiTableMultioutputHelper(SQInfo *pQInfo, int32_t index) {
     pRuntimeEnv->pQueryHandle = NULL;
   }
 
-  pRuntimeEnv->pQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &gp);
+  pRuntimeEnv->pQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &gp, pQInfo);
   taosArrayDestroy(tx);
   taosArrayDestroy(g1);
 
@@ -4252,9 +4285,9 @@ static void sequentialTableProcess(SQInfo *pQInfo) {
       }
       
       if (isFirstLastRowQuery(pQuery)) {
-        pRuntimeEnv->pQueryHandle = tsdbQueryLastRow(pQInfo->tsdb, &cond, &gp);
+        pRuntimeEnv->pQueryHandle = tsdbQueryLastRow(pQInfo->tsdb, &cond, &gp, pQInfo);
       } else {
-        pRuntimeEnv->pQueryHandle = tsdbQueryRowsInExternalWindow(pQInfo->tsdb, &cond, &gp);
+        pRuntimeEnv->pQueryHandle = tsdbQueryRowsInExternalWindow(pQInfo->tsdb, &cond, &gp, pQInfo);
       }
       
       initCtxOutputBuf(pRuntimeEnv);
@@ -4448,7 +4481,7 @@ static void doSaveContext(SQInfo *pQInfo) {
     tsdbCleanupQueryHandle(pRuntimeEnv->pSecQueryHandle);
   }
   
-  pRuntimeEnv->pSecQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &pQInfo->tableIdGroupInfo);
+  pRuntimeEnv->pSecQueryHandle = tsdbQueryTables(pQInfo->tsdb, &cond, &pQInfo->tableIdGroupInfo, pQInfo);
   
   setQueryStatus(pQuery, QUERY_NOT_COMPLETED);
   switchCtxOrder(pRuntimeEnv);
@@ -4515,7 +4548,7 @@ static void multiTableQueryProcess(SQInfo *pQInfo) {
          pQuery->window.skey, pQuery->window.ekey, pQuery->order.order);
 
   // do check all qualified data blocks
-  int64_t el = queryOnDataBlocks(pQInfo);
+  int64_t el = scanMultiTableDataBlocks(pQInfo);
   qTrace("QInfo:%p master scan completed, elapsed time: %lldms, reverse scan start", pQInfo, el);
 
   // query error occurred or query is killed, abort current execution
@@ -4530,7 +4563,7 @@ static void multiTableQueryProcess(SQInfo *pQInfo) {
   if (needReverseScan(pQuery)) {
     doSaveContext(pQInfo);
 
-    el = queryOnDataBlocks(pQInfo);
+    el = scanMultiTableDataBlocks(pQInfo);
     qTrace("QInfo:%p reversed scan completed, elapsed time: %lldms", pQInfo, el);
 
     doRestoreContext(pQInfo);
@@ -4817,7 +4850,7 @@ static void tableQueryImpl(SQInfo *pQInfo) {
   /* check if query is killed or not */
   if (isQueryKilled(pQInfo)) {
     qTrace("QInfo:%p query is killed", pQInfo);
-  } else {// todo set the table uid and tid in log
+  } else {
     qTrace("QInfo:%p query paused, %" PRId64 " rows returned, numOfTotal:%" PRId64 " rows",
         pQInfo, pQuery->rec.rows, pQuery->rec.total + pQuery->rec.rows);
   }
@@ -4909,9 +4942,12 @@ static bool validateQuerySourceCols(SQueryTableMsg *pQueryMsg, SSqlFuncMsg** pEx
     return false;
   } else if (numOfTotal == 0) {
     for(int32_t i = 0; i < pQueryMsg->numOfOutput; ++i) {
-      if (pExprMsg[i]->functionId != TSDB_FUNC_TAGPRJ) {
+      if ((pExprMsg[i]->functionId == TSDB_FUNC_TAGPRJ) ||
-        return false;
+          (pExprMsg[i]->functionId == TSDB_FUNC_TID_TAG && pExprMsg[i]->colInfo.colId == TSDB_TBNAME_COLUMN_INDEX)) {
+        continue;
       }
+      
+      return false;
     }
   }
 
@@ -5996,53 +6032,58 @@ static void buildTagQueryResult(SQInfo* pQInfo) {
   num = taosArrayGetSize(pa);
 
   assert(num == pQInfo->groupInfo.numOfTables);
-//  int16_t type, bytes;
-
   int32_t functionId = pQuery->pSelectExpr[0].base.functionId;
   if (functionId == TSDB_FUNC_TID_TAG) { // return the tags & table Id
     assert(pQuery->numOfOutput == 1);
+    
     SExprInfo* pExprInfo = &pQuery->pSelectExpr[0];
-
     int32_t rsize = pExprInfo->bytes;
 
     for(int32_t i = 0; i < num; ++i) {
-      SGroupItem* item = taosArrayGet(pa, i);
+      SGroupItem *item = taosArrayGet(pa, i);
 
-      char* output = pQuery->sdata[0]->data + i * rsize;
+      char *output = pQuery->sdata[0]->data + i * rsize;
       varDataSetLen(output, rsize - VARSTR_HEADER_SIZE);
-      
+
       output = varDataVal(output);
-      *(int64_t*) output = item->id.uid;  // memory align problem, todo serialize
+      *(int64_t *)output = item->id.uid;  // memory align problem, todo serialize
       output += sizeof(item->id.uid);
 
-      *(int32_t*) output = item->id.tid;
+      *(int32_t *)output = item->id.tid;
       output += sizeof(item->id.tid);
 
-      *(int32_t*) output = pQInfo->vgId;
+      *(int32_t *)output = pQInfo->vgId;
       output += sizeof(pQInfo->vgId);
 
       int16_t bytes = pExprInfo->bytes;
       int16_t type = pExprInfo->type;
-      
+
-      char* val = tsdbGetTableTagVal(pQInfo->tsdb, &item->id, pExprInfo->base.colInfo.colId, type, bytes);
+      if (pExprInfo->base.colInfo.colId == TSDB_TBNAME_COLUMN_INDEX) {
-      
+        char *data = tsdbGetTableName(pQInfo->tsdb, &item->id);
-      // todo refactor
+        memcpy(output, data, varDataTLen(data));
-      if (type == TSDB_DATA_TYPE_BINARY || type == TSDB_DATA_TYPE_NCHAR) {
-        if (val == NULL) {
-          setVardataNull(output, type);
-        } else {
-          memcpy(output, val, varDataTLen(val));
-        }
       } else {
-        if (val == NULL) {
+        char *val = tsdbGetTableTagVal(pQInfo->tsdb, &item->id, pExprInfo->base.colInfo.colId, type, bytes);
-          setNull(output, type, bytes);
+
+        // todo refactor
+        if (type == TSDB_DATA_TYPE_BINARY || type == TSDB_DATA_TYPE_NCHAR) {
+          if (val == NULL) {
+            setVardataNull(output, type);
+          } else {
+            memcpy(output, val, varDataTLen(val));
+          }
         } else {
-          memcpy(output, val, bytes);
+          if (val == NULL) {
+            setNull(output, type, bytes);
+          } else {
+            memcpy(output, val, bytes);
+          }
         }
       }
     }
-
+  
+    pQInfo->tableIndex = pQInfo->groupInfo.numOfTables;
     qTrace("QInfo:%p create (tableId, tag) info completed, rows:%d", pQInfo, num);
+    
   } else {  // return only the tags|table name etc.
     for(int32_t i = 0; i < num; ++i) {
       SExprInfo* pExprInfo = pQuery->pSelectExpr;

src/query/src/qUtil.c

@@ -206,11 +206,6 @@ bool isWindowResClosed(SWindowResInfo *pWindowResInfo, int32_t slot) {
   return (getWindowResult(pWindowResInfo, slot)->status.closed == true);
 }
 
-int32_t curTimeWindow(SWindowResInfo *pWindowResInfo) {
-  assert(pWindowResInfo->curIndex >= 0 && pWindowResInfo->curIndex < pWindowResInfo->size);
-  return pWindowResInfo->curIndex;
-}
-
 void closeTimeWindow(SWindowResInfo *pWindowResInfo, int32_t slot) {
   getWindowResult(pWindowResInfo, slot)->status.closed = true;
 }

src/query/src/qextbuffer.c

@@ -356,17 +356,15 @@ static FORCE_INLINE int32_t primaryKeyComparator(int64_t f1, int64_t f2, int32_t
 static FORCE_INLINE int32_t columnValueAscendingComparator(char *f1, char *f2, int32_t type, int32_t bytes) {
   switch (type) {
     case TSDB_DATA_TYPE_INT: {
-      int32_t first = *(int32_t *)f1;
+      int32_t first  = *(int32_t *) f1;
-      int32_t second = *(int32_t *)f2;
+      int32_t second = *(int32_t *) f2;
       if (first == second) {
         return 0;
       }
       return (first < second) ? -1 : 1;
     };
     case TSDB_DATA_TYPE_DOUBLE: {
-      //double first = *(double *)f1;
+      double first  = GET_DOUBLE_VAL(f1);
-      double first = GET_DOUBLE_VAL(f1);
-      //double second = *(double *)f2;
       double second = GET_DOUBLE_VAL(f2);
       if (first == second) {
         return 0;
@@ -374,9 +372,7 @@ static FORCE_INLINE int32_t columnValueAscendingComparator(char *f1, char *f2, i
       return (first < second) ? -1 : 1;
     };
     case TSDB_DATA_TYPE_FLOAT: {
-      //float first = *(float *)f1;
+      float first  = GET_FLOAT_VAL(f1);
-      //float second = *(float *)f2;
-      float first = GET_FLOAT_VAL(f1);
       float second = GET_FLOAT_VAL(f2);
       if (first == second) {
         return 0;
@@ -439,9 +435,9 @@ int32_t compare_a(tOrderDescriptor *pDescriptor, int32_t numOfRows1, int32_t s1,
                   int32_t s2, char *data2) {
   assert(numOfRows1 == numOfRows2);
 
-  int32_t cmpCnt = pDescriptor->orderIdx.numOfCols;
+  int32_t cmpCnt = pDescriptor->orderInfo.numOfCols;
   for (int32_t i = 0; i < cmpCnt; ++i) {
-    int32_t colIdx = pDescriptor->orderIdx.pData[i];
+    int32_t colIdx = pDescriptor->orderInfo.pData[i];
 
     char *f1 = COLMODEL_GET_VAL(data1, pDescriptor->pColumnModel, numOfRows1, s1, colIdx);
     char *f2 = COLMODEL_GET_VAL(data2, pDescriptor->pColumnModel, numOfRows2, s2, colIdx);
@@ -471,9 +467,9 @@ int32_t compare_d(tOrderDescriptor *pDescriptor, int32_t numOfRows1, int32_t s1,
                   int32_t s2, char *data2) {
   assert(numOfRows1 == numOfRows2);
 
-  int32_t cmpCnt = pDescriptor->orderIdx.numOfCols;
+  int32_t cmpCnt = pDescriptor->orderInfo.numOfCols;
   for (int32_t i = 0; i < cmpCnt; ++i) {
-    int32_t colIdx = pDescriptor->orderIdx.pData[i];
+    int32_t colIdx = pDescriptor->orderInfo.pData[i];
 
     char *f1 = COLMODEL_GET_VAL(data1, pDescriptor->pColumnModel, numOfRows1, s1, colIdx);
     char *f2 = COLMODEL_GET_VAL(data2, pDescriptor->pColumnModel, numOfRows2, s2, colIdx);
@@ -563,13 +559,13 @@ static void median(tOrderDescriptor *pDescriptor, int32_t numOfRows, int32_t sta
   int32_t midIdx = ((end - start) >> 1) + start;
 
 #if defined(_DEBUG_VIEW)
-  int32_t f = pDescriptor->orderIdx.pData[0];
+  int32_t f = pDescriptor->orderInfo.pData[0];
 
   char *midx = COLMODEL_GET_VAL(data, pDescriptor->pColumnModel, numOfRows, midIdx, f);
   char *startx = COLMODEL_GET_VAL(data, pDescriptor->pColumnModel, numOfRows, start, f);
   char *endx = COLMODEL_GET_VAL(data, pDescriptor->pColumnModel, numOfRows, end, f);
 
-  int32_t colIdx = pDescriptor->orderIdx.pData[0];
+  int32_t colIdx = pDescriptor->orderInfo.pData[0];
   tSortDataPrint(pDescriptor->pColumnModel->pFields[colIdx].field.type, "before", startx, midx, endx);
 #endif
 
@@ -596,7 +592,7 @@ static void median(tOrderDescriptor *pDescriptor, int32_t numOfRows, int32_t sta
 }
 
 static UNUSED_FUNC void tRowModelDisplay(tOrderDescriptor *pDescriptor, int32_t numOfRows, char *d, int32_t len) {
-  int32_t colIdx = pDescriptor->orderIdx.pData[0];
+  int32_t colIdx = pDescriptor->orderInfo.pData[0];
 
   for (int32_t i = 0; i < len; ++i) {
     char *startx = COLMODEL_GET_VAL(d, pDescriptor->pColumnModel, numOfRows, i, colIdx);
@@ -1062,9 +1058,9 @@ tOrderDescriptor *tOrderDesCreate(const int32_t *orderColIdx, int32_t numOfOrder
   desc->pColumnModel = pModel;
   desc->tsOrder = tsOrderType;
 
-  desc->orderIdx.numOfCols = numOfOrderCols;
+  desc->orderInfo.numOfCols = numOfOrderCols;
   for (int32_t i = 0; i < numOfOrderCols; ++i) {
-    desc->orderIdx.pData[i] = orderColIdx[i];
+    desc->orderInfo.pData[i] = orderColIdx[i];
   }
 
   return desc;

src/query/src/qresultBuf.c

@@ -50,12 +50,6 @@ int32_t createDiskbasedResultBuffer(SDiskbasedResultBuf** pResultBuf, int32_t si
   return TSDB_CODE_SUCCESS;
 }
 
-tFilePage* getResultBufferPageById(SDiskbasedResultBuf* pResultBuf, int32_t id) {
-  assert(id < pResultBuf->numOfPages && id >= 0);
-
-  return (tFilePage*)(pResultBuf->pBuf + DEFAULT_INTERN_BUF_PAGE_SIZE * id);
-}
-
 int32_t getNumOfResultBufGroupId(SDiskbasedResultBuf* pResultBuf) { return taosHashGetSize(pResultBuf->idsTable); }
 
 int32_t getResBufSize(SDiskbasedResultBuf* pResultBuf) { return pResultBuf->totalBufSize; }
@@ -169,7 +163,7 @@ tFilePage* getNewDataBuf(SDiskbasedResultBuf* pResultBuf, int32_t groupId, int32
   *pageId = (pResultBuf->allocateId++);
   registerPageId(pResultBuf, groupId, *pageId);
 
-  tFilePage* page = getResultBufferPageById(pResultBuf, *pageId);
+  tFilePage* page = GET_RES_BUF_PAGE_BY_ID(pResultBuf, *pageId);
   
   // clear memory for the new page
   memset(page, 0, DEFAULT_INTERN_BUF_PAGE_SIZE);

src/rpc/src/rpcCache.c

@@ -127,7 +127,7 @@ void rpcAddConnIntoCache(void *handle, void *data, char *fqdn, uint16_t port, in
 
   hash = rpcHashConn(pCache, fqdn, port, connType);
   pNode = (SConnHash *)taosMemPoolMalloc(pCache->connHashMemPool);
-  strcpy(pNode->fqdn, fqdn);
+  tstrncpy(pNode->fqdn, fqdn, sizeof(pNode->fqdn));
   pNode->port = port;
   pNode->connType = connType;
   pNode->data = data;

src/rpc/src/rpcMain.c

@@ -60,7 +60,7 @@ typedef struct {
 
   void     *idPool;   // handle to ID pool
   void     *tmrCtrl;  // handle to timer
-  void     *hash;     // handle returned by hash utility
+  SHashObj *hash;     // handle returned by hash utility
   void     *tcphandle;// returned handle from TCP initialization
   void     *udphandle;// returned handle from UDP initialization
   void     *pCache;   // connection cache
@@ -211,7 +211,7 @@ void *rpcOpen(const SRpcInit *pInit) {
   pRpc = (SRpcInfo *)calloc(1, sizeof(SRpcInfo));
   if (pRpc == NULL) return NULL;
 
-  if(pInit->label) strcpy(pRpc->label, pInit->label);
+  if(pInit->label) tstrncpy(pRpc->label, pInit->label, sizeof(pRpc->label));
   pRpc->connType = pInit->connType;
   pRpc->idleTime = pInit->idleTime;
   pRpc->numOfThreads = pInit->numOfThreads>TSDB_MAX_RPC_THREADS ? TSDB_MAX_RPC_THREADS:pInit->numOfThreads;
@@ -228,7 +228,7 @@ void *rpcOpen(const SRpcInit *pInit) {
   size_t size = sizeof(SRpcConn) * pRpc->sessions;
   pRpc->connList = (SRpcConn *)calloc(1, size);
   if (pRpc->connList == NULL) {
-    tError("%s failed to allocate memory for taos connections, size:%d", pRpc->label, size);
+    tError("%s failed to allocate memory for taos connections, size:%ld", pRpc->label, size);
     rpcClose(pRpc);
     return NULL;
   }
@@ -459,7 +459,7 @@ int rpcGetConnInfo(void *thandle, SRpcConnInfo *pInfo) {
   pInfo->clientPort = pConn->peerPort;
   // pInfo->serverIp = pConn->destIp;
 
-  strcpy(pInfo->user, pConn->user);
+  strncpy(pInfo->user, pConn->user, sizeof(pInfo->user));
   return 0;
 }
 
@@ -503,10 +503,10 @@ static SRpcConn *rpcOpenConn(SRpcInfo *pRpc, char *peerFqdn, uint16_t peerPort,
   pConn = rpcAllocateClientConn(pRpc); 
 
   if (pConn) { 
-    strcpy(pConn->peerFqdn, peerFqdn);
+    tstrncpy(pConn->peerFqdn, peerFqdn, sizeof(pConn->peerFqdn));
     pConn->peerIp = peerIp;
     pConn->peerPort = peerPort;
-    strcpy(pConn->user, pRpc->user);
+    tstrncpy(pConn->user, pRpc->user, sizeof(pConn->user));
     pConn->connType = connType;
 
     if (taosOpenConn[connType]) {
@@ -804,7 +804,7 @@ static SRpcConn *rpcProcessMsgHead(SRpcInfo *pRpc, SRecvInfo *pRecv) {
 
   pConn = rpcGetConnObj(pRpc, sid, pRecv);
   if (pConn == NULL) {
-    tTrace("%s %p, failed to get connection obj(%s)", pRpc->label, pHead->ahandle, tstrerror(terrno)); 
+    tTrace("%s %p, failed to get connection obj(%s)", pRpc->label, (void *)pHead->ahandle, tstrerror(terrno)); 
     return NULL;
   } else {
     if (rpcIsReq(pHead->msgType)) {

src/rpc/src/rpcTcp.c

@@ -73,7 +73,7 @@ void *taosInitTcpServer(uint32_t ip, uint16_t port, char *label, int numOfThread
   pServerObj = (SServerObj *)calloc(sizeof(SServerObj), 1);
   pServerObj->ip = ip;
   pServerObj->port = port;
-  strcpy(pServerObj->label, label);
+  tstrncpy(pServerObj->label, label, sizeof(pServerObj->label));
   pServerObj->numOfThreads = numOfThreads;
 
   pServerObj->pThreadObj = (SThreadObj *)calloc(sizeof(SThreadObj), numOfThreads);
@@ -87,7 +87,7 @@ void *taosInitTcpServer(uint32_t ip, uint16_t port, char *label, int numOfThread
   pThreadObj = pServerObj->pThreadObj;
   for (int i = 0; i < numOfThreads; ++i) {
     pThreadObj->processData = fp;
-    strcpy(pThreadObj->label, label);
+    tstrncpy(pThreadObj->label, label, sizeof(pThreadObj->label));
     pThreadObj->shandle = shandle;
 
     code = pthread_mutex_init(&(pThreadObj->mutex), NULL);
@@ -247,7 +247,7 @@ void *taosInitTcpClient(uint32_t ip, uint16_t port, char *label, int num, void *
 
   pThreadObj = (SThreadObj *)malloc(sizeof(SThreadObj));
   memset(pThreadObj, 0, sizeof(SThreadObj));
-  strcpy(pThreadObj->label, label);
+  tstrncpy(pThreadObj->label, label, sizeof(pThreadObj->label));
   pThreadObj->ip = ip;
   pThreadObj->shandle = shandle;
 

src/rpc/src/rpcUdp.c

@@ -72,7 +72,7 @@ void *taosInitUdpConnection(uint32_t ip, uint16_t port, char *label, int threads
   pSet->port = port;
   pSet->shandle = shandle;
   pSet->fp = fp;
-  strcpy(pSet->label, label);
+  tstrncpy(pSet->label, label, sizeof(pSet->label));
 
   uint16_t ownPort;
   for (int i = 0; i < threads; ++i) {
@@ -99,7 +99,7 @@ void *taosInitUdpConnection(uint32_t ip, uint16_t port, char *label, int threads
       pConn->localPort = (uint16_t)ntohs(sin.sin_port);
     }
 
-    strcpy(pConn->label, label);
+    tstrncpy(pConn->label, label, sizeof(pConn->label));
     pConn->shandle = shandle;
     pConn->processData = fp;
     pConn->index = i;

src/rpc/test/rclient.c

@@ -76,6 +76,7 @@ int main(int argc, char *argv[]) {
   int      numOfReqs = 0;
   int      appThreads = 1;
   char     serverIp[40] = "127.0.0.1";
+  char     secret[TSDB_KEY_LEN] = "mypassword";
   struct   timeval systemTime;
   int64_t  startTime, endTime;
   pthread_attr_t thattr;
@@ -97,7 +98,7 @@ int main(int argc, char *argv[]) {
   rpcInit.sessions     = 100;
   rpcInit.idleTime     = tsShellActivityTimer*1000;
   rpcInit.user         = "michael";
-  rpcInit.secret       = "mypassword";
+  rpcInit.secret       = secret;
   rpcInit.ckey         = "key";
   rpcInit.spi          = 1;
   rpcInit.connType     = TAOS_CONN_CLIENT;
@@ -106,7 +107,7 @@ int main(int argc, char *argv[]) {
     if (strcmp(argv[i], "-p")==0 && i < argc-1) {
       ipSet.port[0] = atoi(argv[++i]);
     } else if (strcmp(argv[i], "-i") ==0 && i < argc-1) {
-      tstrncpy(ipSet.fqdn[0], argv[++i], sizeof(ipSet.fqdn)); 
+      tstrncpy(ipSet.fqdn[0], argv[++i], sizeof(ipSet.fqdn[0])); 
     } else if (strcmp(argv[i], "-t")==0 && i < argc-1) {
       rpcInit.numOfThreads = atoi(argv[++i]);
     } else if (strcmp(argv[i], "-m")==0 && i < argc-1) {

src/rpc/test/rsclient.c

@@ -77,6 +77,7 @@ int main(int argc, char *argv[]) {
   int      numOfReqs = 0;
   int      appThreads = 1;
   char     serverIp[40] = "127.0.0.1";
+  char     secret[TSDB_KEY_LEN] = "mypassword";
   struct   timeval systemTime;
   int64_t  startTime, endTime;
   pthread_attr_t thattr;
@@ -98,7 +99,7 @@ int main(int argc, char *argv[]) {
   rpcInit.sessions     = 100;
   rpcInit.idleTime     = tsShellActivityTimer*1000;
   rpcInit.user         = "michael";
-  rpcInit.secret       = "mypassword";
+  rpcInit.secret       = secret;
   rpcInit.ckey         = "key";
   rpcInit.spi          = 1;
   rpcInit.connType     = TAOS_CONN_CLIENT;

src/tsdb/src/tsdbRead.c

@@ -17,6 +17,7 @@
 #include "tulog.h"
 #include "talgo.h"
 #include "tutil.h"
+#include "ttime.h"
 #include "tcompare.h"
 #include "exception.h"
 
@@ -71,6 +72,8 @@ typedef struct STableCheckInfo {
   int32_t       compSize;
   int32_t       numOfBlocks;    // number of qualified data blocks not the original blocks
   SDataCols*    pDataCols;
+  
+  int32_t       chosen;         // indicate which iterator should move forward
   bool          initBuf;        // whether to initialize the in-memory skip list iterator or not
   SSkipListIterator* iter;      // mem buffer skip list iterator
   SSkipListIterator* iiter;     // imem buffer skip list iterator
@@ -79,8 +82,6 @@ typedef struct STableCheckInfo {
 typedef struct STableBlockInfo {
   SCompBlock*        compBlock;
   STableCheckInfo*   pTableCheckInfo;
-//  int32_t            blockIndex;
-//  int32_t            groupIdx; /* number of group is less than the total number of tables */
 } STableBlockInfo;
 
 typedef struct SBlockOrderSupporter {
@@ -120,7 +121,7 @@ static void changeQueryHandleForInterpQuery(TsdbQueryHandleT pHandle);
 static void doMergeTwoLevelData(STsdbQueryHandle* pQueryHandle, STableCheckInfo* pCheckInfo, SCompBlock* pBlock,
                                    SArray* sa);
 static int32_t binarySearchForKey(char* pValue, int num, TSKEY key, int order);
-static int tsdbReadRowsFromCache(SSkipListIterator* pIter, STable* pTable, TSKEY maxKey, int maxRowsToRead, TSKEY* skey, TSKEY* ekey,
+static int tsdbReadRowsFromCache(STableCheckInfo* pCheckInfo, TSKEY maxKey, int maxRowsToRead, TSKEY* skey, TSKEY* ekey,
                                  STsdbQueryHandle* pQueryHandle);
 
 static void tsdbInitDataBlockLoadInfo(SDataBlockLoadInfo* pBlockLoadInfo) {
@@ -134,7 +135,7 @@ static void tsdbInitCompBlockLoadInfo(SLoadCompBlockInfo* pCompBlockLoadInfo) {
   pCompBlockLoadInfo->fileId = -1;
 }
 
-TsdbQueryHandleT* tsdbQueryTables(TsdbRepoT* tsdb, STsdbQueryCond* pCond, STableGroupInfo* groupList) {
+TsdbQueryHandleT* tsdbQueryTables(TsdbRepoT* tsdb, STsdbQueryCond* pCond, STableGroupInfo* groupList, void* qinfo) {
   // todo 1. filter not exist table
   // todo 2. add the reference count for each table that is involved in query
 
@@ -147,6 +148,7 @@ TsdbQueryHandleT* tsdbQueryTables(TsdbRepoT* tsdb, STsdbQueryCond* pCond, STable
   pQueryHandle->cur.win     = TSWINDOW_INITIALIZER;
   pQueryHandle->checkFiles  = true;//ASCENDING_TRAVERSE(pQueryHandle->order);
   pQueryHandle->activeIndex = 0;   // current active table index
+  pQueryHandle->qinfo       = qinfo;
   pQueryHandle->outputCapacity = ((STsdbRepo*)tsdb)->config.maxRowsPerFileBlock;
   
   tsdbInitReadHelper(&pQueryHandle->rhelper, (STsdbRepo*) tsdb);
@@ -201,8 +203,8 @@ TsdbQueryHandleT* tsdbQueryTables(TsdbRepoT* tsdb, STsdbQueryCond* pCond, STable
   return (TsdbQueryHandleT) pQueryHandle;
 }
 
-TsdbQueryHandleT tsdbQueryLastRow(TsdbRepoT *tsdb, STsdbQueryCond *pCond, STableGroupInfo *groupList) {
+TsdbQueryHandleT tsdbQueryLastRow(TsdbRepoT *tsdb, STsdbQueryCond *pCond, STableGroupInfo *groupList, void* qinfo) {
-  STsdbQueryHandle *pQueryHandle = (STsdbQueryHandle*) tsdbQueryTables(tsdb, pCond, groupList);
+  STsdbQueryHandle *pQueryHandle = (STsdbQueryHandle*) tsdbQueryTables(tsdb, pCond, groupList, qinfo);
   
   pQueryHandle->type = TSDB_QUERY_TYPE_LAST;
   pQueryHandle->order = TSDB_ORDER_DESC;
@@ -227,8 +229,8 @@ SArray* tsdbGetQueriedTableIdList(TsdbQueryHandleT *pHandle) {
   return res;
 }
 
-TsdbQueryHandleT tsdbQueryRowsInExternalWindow(TsdbRepoT *tsdb, STsdbQueryCond* pCond, STableGroupInfo *groupList) {
+TsdbQueryHandleT tsdbQueryRowsInExternalWindow(TsdbRepoT *tsdb, STsdbQueryCond* pCond, STableGroupInfo *groupList, void* qinfo) {
-  STsdbQueryHandle *pQueryHandle = (STsdbQueryHandle*) tsdbQueryTables(tsdb, pCond, groupList);
+  STsdbQueryHandle *pQueryHandle = (STsdbQueryHandle*) tsdbQueryTables(tsdb, pCond, groupList, qinfo);
   
   pQueryHandle->type = TSDB_QUERY_TYPE_EXTERNAL;
 //  pQueryHandle->outputCapacity = 2;  // only allowed two rows to be loaded
@@ -303,6 +305,83 @@ static bool initTableMemIterator(STsdbQueryHandle* pHandle, STableCheckInfo* pCh
   return true;
 }
 
+SDataRow getSDataRowInTableMem(STableCheckInfo* pCheckInfo) {
+  SDataRow rmem = NULL, rimem = NULL;
+  if (pCheckInfo->iter) {
+    SSkipListNode* node = tSkipListIterGet(pCheckInfo->iter);
+    if (node != NULL) {
+      rmem = SL_GET_NODE_DATA(node);
+    }
+  }
+  
+  if (pCheckInfo->iiter) {
+    SSkipListNode* node = tSkipListIterGet(pCheckInfo->iiter);
+    if (node != NULL) {
+      rimem = SL_GET_NODE_DATA(node);
+    }
+  }
+  
+  if (rmem != NULL && rimem != NULL) {
+    if (dataRowKey(rmem) < dataRowKey(rimem)) {
+      pCheckInfo->chosen = 0;
+      return rmem;
+    } else if (dataRowKey(rmem) == dataRowKey(rimem)) {
+      // data ts are duplicated, ignore the data in mem
+      tSkipListIterNext(pCheckInfo->iter);
+      pCheckInfo->chosen = 1;
+      return rimem;
+    } else {
+      pCheckInfo->chosen = 1;
+      return rimem;
+    }
+  }
+  
+  if (rmem != NULL) {
+    pCheckInfo->chosen = 0;
+    return rmem;
+  }
+  
+  if (rimem != NULL) {
+    pCheckInfo->chosen = 1;
+    return rimem;
+  }
+  
+  return NULL;
+}
+
+bool moveToNextRow(STableCheckInfo* pCheckInfo) {
+  bool hasNext = false;
+  if (pCheckInfo->chosen == 0) {
+    if (pCheckInfo->iter != NULL) {
+      hasNext = tSkipListIterNext(pCheckInfo->iter);
+    }
+    
+    if (hasNext) {
+      return hasNext;
+    }
+    
+    if (pCheckInfo->iiter != NULL) {
+      return tSkipListIterGet(pCheckInfo->iiter) != NULL;
+    }
+  } else {
+    if (pCheckInfo->chosen == 1) {
+      if (pCheckInfo->iiter != NULL) {
+        hasNext = tSkipListIterNext(pCheckInfo->iiter);
+      }
+  
+      if (hasNext) {
+        return hasNext;
+      }
+  
+      if (pCheckInfo->iter != NULL) {
+        return tSkipListIterGet(pCheckInfo->iter) != NULL;
+      }
+    }
+  }
+  
+  return hasNext;
+}
+
 static bool hasMoreDataInCache(STsdbQueryHandle* pHandle) {
   size_t size = taosArrayGetSize(pHandle->pTableCheckInfo);
   assert(pHandle->activeIndex < size && pHandle->activeIndex >= 0 && size >= 1);
@@ -312,31 +391,13 @@ static bool hasMoreDataInCache(STsdbQueryHandle* pHandle) {
 
   STable* pTable = pCheckInfo->pTableObj;
   assert(pTable != NULL);
-
+  
-  // no data in cache, abort
+  initTableMemIterator(pHandle, pCheckInfo);
-  if (pTable->mem == NULL && pTable->imem == NULL) {
+  SDataRow row = getSDataRowInTableMem(pCheckInfo);
+  if (row == NULL) {
     return false;
   }
   
-  if (pCheckInfo->iter == NULL && pTable->mem) {
-    pCheckInfo->iter = tSkipListCreateIterFromVal(pTable->mem->pData, (const char*) &pCheckInfo->lastKey,
-        TSDB_DATA_TYPE_TIMESTAMP, pHandle->order);
-    
-    if (pCheckInfo->iter == NULL) {
-      return false;
-    }
-  
-    if (!tSkipListIterNext(pCheckInfo->iter)) {  // buffer is empty
-      return false;
-    }
-  }
-  
-  SSkipListNode* node = tSkipListIterGet(pCheckInfo->iter);
-  if (node == NULL) {
-    return false;
-  }
-
-  SDataRow row = SL_GET_NODE_DATA(node);
   pCheckInfo->lastKey = dataRowKey(row);  // first timestamp in buffer
   tsdbTrace("%p uid:%" PRId64", tid:%d check data in buffer from skey:%" PRId64 ", order:%d, %p", pHandle,
       pCheckInfo->tableId.uid, pCheckInfo->tableId.tid, pCheckInfo->lastKey, pHandle->order, pHandle->qinfo);
@@ -349,7 +410,7 @@ static bool hasMoreDataInCache(STsdbQueryHandle* pHandle) {
   
   int32_t step = ASCENDING_TRAVERSE(pHandle->order)? 1:-1;
   STimeWindow* win = &pHandle->cur.win;
-  pHandle->cur.rows = tsdbReadRowsFromCache(pCheckInfo->iter, pCheckInfo->pTableObj, pHandle->window.ekey,
+  pHandle->cur.rows = tsdbReadRowsFromCache(pCheckInfo, pHandle->window.ekey,
                                     pHandle->outputCapacity, &win->skey, &win->ekey, pHandle);  // todo refactor API
   
   // update the last key value
@@ -382,7 +443,7 @@ static int32_t getFileIdFromKey(TSKEY key, int32_t daysPerFile, int32_t precisio
   return fid;
 }
 
-static int32_t binarySearchForBlockImpl(SCompBlock* pBlock, int32_t numOfBlocks, TSKEY skey, int32_t order) {
+static int32_t binarySearchForBlock(SCompBlock* pBlock, int32_t numOfBlocks, TSKEY skey, int32_t order) {
   int32_t firstSlot = 0;
   int32_t lastSlot = numOfBlocks - 1;
   
@@ -448,7 +509,7 @@ static int32_t getFileCompInfo(STsdbQueryHandle* pQueryHandle, int32_t* numOfBlo
       TSKEY e = MAX(pCheckInfo->lastKey, pQueryHandle->window.ekey);
       
       // discard the unqualified data block based on the query time window
-      int32_t start = binarySearchForBlockImpl(pCompInfo->blocks, compIndex->numOfBlocks, s, TSDB_ORDER_ASC);
+      int32_t start = binarySearchForBlock(pCompInfo->blocks, compIndex->numOfBlocks, s, TSDB_ORDER_ASC);
       int32_t end = start;
       
       if (s > pCompInfo->blocks[start].keyLast) {
@@ -522,7 +583,9 @@ static bool doLoadFileDataBlock(STsdbQueryHandle* pQueryHandle, SCompBlock* pBlo
 
   bool    blockLoaded = false;
   SArray* sa = getDefaultLoadColumns(pQueryHandle, true);
-
+  
+  int64_t st = taosGetTimestampUs();
+  
   if (pCheckInfo->pDataCols == NULL) {
     STsdbMeta* pMeta = tsdbGetMeta(pRepo);
     pCheckInfo->pDataCols = tdNewDataCols(pMeta->maxRowBytes, pMeta->maxCols, pRepo->config.maxRowsPerFileBlock);
@@ -540,9 +603,15 @@ static bool doLoadFileDataBlock(STsdbQueryHandle* pQueryHandle, SCompBlock* pBlo
     blockLoaded = true;
   }
 
+  SDataCols* pCols = pQueryHandle->rhelper.pDataCols[0];
+  assert(pCols->numOfRows != 0);
+  
   taosArrayDestroy(sa);
   tfree(data);
-
+  
+  int64_t et = taosGetTimestampUs() - st;
+  tsdbTrace("%p load file block into buffer, elapsed time:%"PRId64 " us", pQueryHandle, et);
+  
   return blockLoaded;
 }
 
@@ -600,7 +669,7 @@ static void handleDataMergeIfNeeded(STsdbQueryHandle* pQueryHandle, SCompBlock*
       // do not load file block into buffer
       int32_t step = ASCENDING_TRAVERSE(pQueryHandle->order) ? 1 : -1;
 
-      cur->rows = tsdbReadRowsFromCache(pCheckInfo->iter, pCheckInfo->pTableObj, binfo.window.skey - step,
+      cur->rows = tsdbReadRowsFromCache(pCheckInfo, binfo.window.skey - step,
                                         pQueryHandle->outputCapacity, &cur->win.skey, &cur->win.ekey, pQueryHandle);
       pQueryHandle->realNumOfRows = cur->rows;
 
@@ -649,7 +718,7 @@ static bool loadFileDataBlock(STsdbQueryHandle* pQueryHandle, SCompBlock* pBlock
       if (!doLoadFileDataBlock(pQueryHandle, pBlock, pCheckInfo)) {
         return false;
       }
-  
+      
       SDataCols* pTSCol = pQueryHandle->rhelper.pDataCols[0];
       assert(pTSCol->cols->type == TSDB_DATA_TYPE_TIMESTAMP && pTSCol->numOfRows == pBlock->numOfRows);
       
@@ -1155,7 +1224,7 @@ static int32_t createDataBlocksInfo(STsdbQueryHandle* pQueryHandle, int32_t numO
   *numOfAllocBlocks = numOfBlocks;
 
   int32_t numOfTables = taosArrayGetSize(pQueryHandle->pTableCheckInfo);
-
+  
   SBlockOrderSupporter sup = {0};
   sup.numOfTables = numOfTables;
   sup.numOfBlocksPerTable = calloc(1, sizeof(int32_t) * numOfTables);
@@ -1192,15 +1261,26 @@ static int32_t createDataBlocksInfo(STsdbQueryHandle* pQueryHandle, int32_t numO
 
       pBlockInfo->compBlock = &pBlock[k];
       pBlockInfo->pTableCheckInfo = pTableCheck;
-      //      pBlockInfo->groupIdx = pTableCheckInfo[j]->groupIdx;     // set the group index
-      //      pBlockInfo->blockIndex = pTableCheckInfo[j]->start + k;    // set the block index in original table
       cnt++;
     }
 
     numOfQualTables++;
   }
 
-  tsdbTrace("%p create data blocks info struct completed, %d blocks in %d tables", pQueryHandle, cnt, numOfQualTables);
+  assert(numOfBlocks == cnt);
+  
+  // since there is only one table qualified, blocks are not sorted
+  if (numOfQualTables == 1) {
+    memcpy(pQueryHandle->pDataBlockInfo, sup.pDataBlockInfo[0], sizeof(STableBlockInfo) * numOfBlocks);
+    cleanBlockOrderSupporter(&sup, numOfQualTables);
+  
+    tsdbTrace("%p create data blocks info struct completed for 1 table, %d blocks not sorted %p ", pQueryHandle, cnt,
+        pQueryHandle->qinfo);
+    return TSDB_CODE_SUCCESS;
+  }
+  
+  tsdbTrace("%p create data blocks info struct completed, %d blocks in %d tables %p", pQueryHandle, cnt,
+      numOfQualTables, pQueryHandle->qinfo);
 
   assert(cnt <= numOfBlocks && numOfQualTables <= numOfTables);  // the pTableQueryInfo[j]->numOfBlocks may be 0
   sup.numOfTables = numOfQualTables;
@@ -1257,8 +1337,8 @@ static bool getDataBlocksInFilesImpl(STsdbQueryHandle* pQueryHandle) {
       break;
     }
     
-    tsdbTrace("%p %d blocks found in file for %d table(s), fid:%d", pQueryHandle, numOfBlocks,
+    tsdbTrace("%p %d blocks found in file for %d table(s), fid:%d, %p", pQueryHandle, numOfBlocks,
-           numOfTables, pQueryHandle->pFileGroup->fileId);
+           numOfTables, pQueryHandle->pFileGroup->fileId, pQueryHandle->qinfo);
     
     assert(numOfBlocks >= 0);
     if (numOfBlocks == 0) {
@@ -1565,19 +1645,22 @@ static void changeQueryHandleForInterpQuery(TsdbQueryHandleT pHandle) {
   pQueryHandle->window = (STimeWindow) {info.lastKey, TSKEY_INITIAL_VAL};
 }
 
-static int tsdbReadRowsFromCache(SSkipListIterator* pIter, STable* pTable, TSKEY maxKey, int maxRowsToRead, TSKEY* skey, TSKEY* ekey,
+static int tsdbReadRowsFromCache(STableCheckInfo* pCheckInfo, TSKEY maxKey, int maxRowsToRead, TSKEY* skey, TSKEY* ekey,
                                  STsdbQueryHandle* pQueryHandle) {
   int     numOfRows = 0;
   int32_t numOfCols = taosArrayGetSize(pQueryHandle->pColumns);
   *skey = TSKEY_INITIAL_VAL;
 
+  int64_t st = taosGetTimestampUs();
+  STSchema* pSchema = tsdbGetTableSchema(tsdbGetMeta(pQueryHandle->pTsdb), pCheckInfo->pTableObj);
+  int32_t numOfTableCols = schemaNCols(pSchema);
+  
   do {
-    SSkipListNode* node = tSkipListIterGet(pIter);
+    SDataRow row = getSDataRowInTableMem(pCheckInfo);
-    if (node == NULL) {
+    if (row == NULL) {
       break;
     }
 
-    SDataRow row = SL_GET_NODE_DATA(node);
     TSKEY key = dataRowKey(row);
     
     if ((key > maxKey && ASCENDING_TRAVERSE(pQueryHandle->order)) ||
@@ -1590,49 +1673,69 @@ static int tsdbReadRowsFromCache(SSkipListIterator* pIter, STable* pTable, TSKEY
     }
 
     if (*skey == INT64_MIN) {
-      *skey = dataRowKey(row);
+      *skey = key;
     }
 
-    *ekey = dataRowKey(row);
+    *ekey = key;
-
-    int32_t offset = -1;
     char* pData = NULL;
-  
-    STSchema* pSchema = tsdbGetTableSchema(tsdbGetMeta(pQueryHandle->pTsdb), pTable);
-    int32_t numOfTableCols = schemaNCols(pSchema);
     
-    for (int32_t i = 0; i < numOfCols; ++i) {
+    int32_t i = 0, j = 0;
+    while(i < numOfCols && j < numOfTableCols) {
       SColumnInfoData* pColInfo = taosArrayGet(pQueryHandle->pColumns, i);
-      
+      if (pSchema->columns[j].colId < pColInfo->info.colId) {
+        j++;
+        continue;
+      }
+  
       if (ASCENDING_TRAVERSE(pQueryHandle->order)) {
         pData = pColInfo->pData + numOfRows * pColInfo->info.bytes;
       } else {
         pData = pColInfo->pData + (maxRowsToRead - numOfRows - 1) * pColInfo->info.bytes;
       }
       
-      for(int32_t j = 0; j < numOfTableCols; ++j) {
+      if (pSchema->columns[j].colId == pColInfo->info.colId) {
-        if (pColInfo->info.colId == pSchema->columns[j].colId) {
+        void* value = tdGetRowDataOfCol(row, pColInfo->info.type, TD_DATA_ROW_HEAD_SIZE + pSchema->columns[j].offset);
-          offset = pSchema->columns[j].offset;
+        if (pColInfo->info.type == TSDB_DATA_TYPE_BINARY || pColInfo->info.type == TSDB_DATA_TYPE_NCHAR) {
-          break;
+          memcpy(pData, value, varDataTLen(value));
+        } else {
+          memcpy(pData, value, pColInfo->info.bytes);
         }
-      }
+        
-      
+        j++;
-      assert(offset != -1); // todo handle error
+        i++;
-      void *value = tdGetRowDataOfCol(row, pColInfo->info.type, TD_DATA_ROW_HEAD_SIZE + offset);
+      } else { // pColInfo->info.colId < pSchema->columns[j].colId, it is a NULL data
-      
+        if (pColInfo->info.type == TSDB_DATA_TYPE_BINARY || pColInfo->info.type == TSDB_DATA_TYPE_NCHAR) {
-      if (pColInfo->info.type == TSDB_DATA_TYPE_BINARY || pColInfo->info.type == TSDB_DATA_TYPE_NCHAR) {
+          setVardataNull(pData, pColInfo->info.type);
-        memcpy(pData, value, varDataTLen(value));
+        } else {
-      } else {
+          setNull(pData, pColInfo->info.type, pColInfo->info.bytes);
-        memcpy(pData, value, pColInfo->info.bytes);
+        }
+        i++;
       }
     }
-
+    
+    while (i < numOfCols) { // the remain columns are all null data
+      SColumnInfoData* pColInfo = taosArrayGet(pQueryHandle->pColumns, i);
+      if (ASCENDING_TRAVERSE(pQueryHandle->order)) {
+        pData = pColInfo->pData + numOfRows * pColInfo->info.bytes;
+      } else {
+        pData = pColInfo->pData + (maxRowsToRead - numOfRows - 1) * pColInfo->info.bytes;
+      }
+  
+      if (pColInfo->info.type == TSDB_DATA_TYPE_BINARY || pColInfo->info.type == TSDB_DATA_TYPE_NCHAR) {
+        setVardataNull(pData, pColInfo->info.type);
+      } else {
+        setNull(pData, pColInfo->info.type, pColInfo->info.bytes);
+      }
+      
+      i++;
+    }
+    
     if (++numOfRows >= maxRowsToRead) {
-      tSkipListIterNext(pIter);
+      moveToNextRow(pCheckInfo);
       break;
     }
     
-  } while(tSkipListIterNext(pIter));
+  } while(moveToNextRow(pCheckInfo));
 
   assert(numOfRows <= maxRowsToRead);
   
@@ -1646,6 +1749,10 @@ static int tsdbReadRowsFromCache(SSkipListIterator* pIter, STable* pTable, TSKEY
     }
   }
   
+  int64_t elapsedTime = taosGetTimestampUs() - st;
+  tsdbTrace("%p build data block from cache completed, elapsed time:%"PRId64" us, numOfRows:%d, numOfCols:%d", pQueryHandle,
+            elapsedTime, numOfRows, numOfCols);
+  
   return numOfRows;
 }
 

src/util/inc/tutil.h

@@ -153,6 +153,8 @@ bool taosMbsToUcs4(char *mbs, size_t mbs_len, char *ucs4, int32_t ucs4_max_len,
 
 int tasoUcs4Compare(void* f1_ucs4, void *f2_ucs4, int bytes);
 
+void taosRandStr(char* str, int32_t size);
+
 int32_t taosUcs4ToMbs(void *ucs4, int32_t ucs4_max_len, char *mbs);
 
 bool taosValidateEncodec(const char *encodec);

src/util/src/tarray.c

@@ -92,12 +92,14 @@ void* taosArrayGet(const SArray* pArray, size_t index) {
 }
 
 void* taosArrayGetP(const SArray* pArray, size_t index) {
-  void* ret = taosArrayGet(pArray, index);
+  assert(index < pArray->size);
-  if (ret == NULL) {
+  
+  void* d = TARRAY_GET_ELEM(pArray, index);
+  if (d == NULL) {
     return NULL;
   }
   
-  return *(void**)ret;
+  return *(void**)d;
 }
 
 size_t taosArrayGetSize(const SArray* pArray) { return pArray->size; }

src/util/src/tsocket.c

@@ -19,6 +19,7 @@
 #include "tutil.h"
 
 int taosGetFqdn(char *fqdn) {
+  int  code = 0;
   char hostname[1024];
   hostname[1023] = '\0';
   gethostname(hostname, 1023);
@@ -27,13 +28,15 @@ int taosGetFqdn(char *fqdn) {
   h = gethostbyname(hostname);
   if (h != NULL) {
     strcpy(fqdn, h->h_name);
-    return 0;
   } else {
-    uError("failed to get host name");
+    uError("failed to get host name(%s)", strerror(errno));
-    return -1;
+    code = -1;
   }
 
-  free(h);
+  // to do: free the resources
+  // free(h);
+
+  return code;
 }
 
 uint32_t taosGetIpFromFqdn(const char *fqdn) {
@@ -47,7 +50,7 @@ uint32_t ip2uint(const char *const ip_addr) {
   char ip_addr_cpy[20];
   char ip[5];
 
-  strcpy(ip_addr_cpy, ip_addr);
+  tstrncpy(ip_addr_cpy, ip_addr, sizeof(ip_addr_cpy));
 
   char *s_start, *s_end;
   s_start = ip_addr_cpy;
@@ -206,7 +209,7 @@ int taosOpenUdpSocket(uint32_t ip, uint16_t port) {
   int                reuse, nocheck;
   int                bufSize = 8192000;
 
-  uTrace("open udp socket:%s:%hu", ip, port);
+  uTrace("open udp socket:0x%x:%hu", ip, port);
 
   memset((char *)&localAddr, 0, sizeof(localAddr));
   localAddr.sin_family = AF_INET;
@@ -257,7 +260,7 @@ int taosOpenUdpSocket(uint32_t ip, uint16_t port) {
 
   /* bind socket to local address */
   if (bind(sockFd, (struct sockaddr *)&localAddr, sizeof(localAddr)) < 0) {
-    uError("failed to bind udp socket: %d (%s), %s:%hu", errno, strerror(errno), ip, port);
+    uError("failed to bind udp socket: %d (%s), 0x%x:%hu", errno, strerror(errno), ip, port);
     taosCloseSocket(sockFd);
     return -1;
   }
@@ -363,7 +366,7 @@ int taosOpenTcpServerSocket(uint32_t ip, uint16_t port) {
   int                sockFd;
   int                reuse;
 
-  uTrace("open tcp server socket:%s:%hu", ip, port);
+  uTrace("open tcp server socket:0x%x:%hu", ip, port);
 
   bzero((char *)&serverAdd, sizeof(serverAdd));
   serverAdd.sin_family = AF_INET;

src/util/src/tutil.c

@@ -27,8 +27,6 @@
 #include "tulog.h"
 #include "taoserror.h"
 
-int32_t tmpFileSerialNum = 0;
-
 int32_t strdequote(char *z) {
   if (z == NULL) {
     return 0;
@@ -433,12 +431,24 @@ void getTmpfilePath(const char *fileNamePrefix, char *dstPath) {
 #else
   char *tmpDir = "/tmp/";
 #endif
-  int64_t ts = taosGetTimestampUs();
+  
   strcpy(tmpPath, tmpDir);
   strcat(tmpPath, tdengineTmpFileNamePrefix);
   strcat(tmpPath, fileNamePrefix);
-  strcat(tmpPath, "-%d-%"PRIu64"-%u-%"PRIu64);
+  strcat(tmpPath, "-%d-%s");
-  snprintf(dstPath, PATH_MAX, tmpPath, getpid(), taosGetPthreadId(), atomic_add_fetch_32(&tmpFileSerialNum, 1), ts);
+  
+  char rand[8] = {0};
+  taosRandStr(rand, tListLen(rand) - 1);
+  snprintf(dstPath, PATH_MAX, tmpPath, getpid(), rand);
+}
+
+void taosRandStr(char* str, int32_t size) {
+  const char* set = "abcdefghijklmnopqrstuvwxyz0123456789-_.";
+  int32_t len = 39;
+  
+  for(int32_t i = 0; i < size; ++i) {
+    str[i] = set[rand()%len];
+  }
 }
 
 int tasoUcs4Compare(void* f1_ucs4, void *f2_ucs4, int bytes) {

src/wal/src/walMain.c

@@ -68,11 +68,19 @@ void *walOpen(const char *path, const SWalCfg *pCfg) {
   pWal->num = 0;
   pWal->level = pCfg->walLevel;
   pWal->keep = pCfg->keep;
-  strcpy(pWal->path, path);
+  tstrncpy(pWal->path, path, sizeof(pWal->path));
   pthread_mutex_init(&pWal->mutex, NULL);
 
-  if (access(path, F_OK) != 0) mkdir(path, 0755);
+  if (access(path, F_OK) != 0) {
-  
+    if (mkdir(path, 0755) != 0) {
+      terrno = TAOS_SYSTEM_ERROR(errno);
+      wError("wal:%s, failed to create directory(%s)", path, strerror(errno));
+      pthread_mutex_destroy(&pWal->mutex);
+      free(pWal);
+      pWal = NULL;
+    }
+  }
+     
   if (pCfg->keep == 1) return pWal;
 
   if (walHandleExistingFiles(path) == 0) 
@@ -80,7 +88,7 @@ void *walOpen(const char *path, const SWalCfg *pCfg) {
 
   if (pWal->fd <0) {
     terrno = TAOS_SYSTEM_ERROR(errno);
-    wError("wal:%s, failed to open", path);
+    wError("wal:%s, failed to open(%s)", path, strerror(errno));
     pthread_mutex_destroy(&pWal->mutex);
     free(pWal);
     pWal = NULL;
@@ -119,7 +127,8 @@ void walClose(void *handle) {
 int walRenew(void *handle) {
   if (handle == NULL) return 0;
   SWal *pWal = handle;
-  int   code = 0;
+
+  terrno = 0;
 
   pthread_mutex_lock(&pWal->mutex);
 
@@ -135,8 +144,8 @@ int walRenew(void *handle) {
   pWal->fd = open(pWal->name, O_WRONLY | O_CREAT, S_IRWXU | S_IRWXG | S_IRWXO);
 
   if (pWal->fd < 0) {
-    wError("wal:%d, failed to open(%s)", pWal->name, strerror(errno));
+    wError("wal:%s, failed to open(%s)", pWal->name, strerror(errno));
-    code = -1;
+    terrno = TAOS_SYSTEM_ERROR(errno);
   } else {
     wTrace("wal:%s, it is created", pWal->name);
 
@@ -156,14 +165,15 @@ int walRenew(void *handle) {
   
   pthread_mutex_unlock(&pWal->mutex);
 
-  return code;
+  return terrno;
 }
 
 int walWrite(void *handle, SWalHead *pHead) {
   SWal *pWal = handle;
-  int   code = 0;
   if (pWal == NULL) return -1;
 
+  terrno = 0;
+
   // no wal  
   if (pWal->level == TAOS_WAL_NOLOG) return 0;
   if (pHead->version <= pWal->version) return 0;
@@ -174,12 +184,12 @@ int walWrite(void *handle, SWalHead *pHead) {
 
   if(write(pWal->fd, pHead, contLen) != contLen) {
     wError("wal:%s, failed to write(%s)", pWal->name, strerror(errno));
-    code = -1;
+    terrno = TAOS_SYSTEM_ERROR(errno);
   } else {
     pWal->version = pHead->version;
   }
 
-  return code;
+  return terrno;
 }
 
 void walFsync(void *handle) {
@@ -196,11 +206,11 @@ void walFsync(void *handle) {
 
 int walRestore(void *handle, void *pVnode, int (*writeFp)(void *, void *, int)) {
   SWal    *pWal = handle;
-  int      code = 0;
   struct   dirent *ent;
   int      count = 0;
   uint32_t maxId = 0, minId = -1, index =0;
 
+  terrno = 0;
   int   plen = strlen(walPrefix);
   char  opath[TSDB_FILENAME_LEN+5];
    
@@ -224,30 +234,30 @@ int walRestore(void *handle, void *pVnode, int (*writeFp)(void *, void *, int))
   closedir(dir);
 
   if (count == 0) {
-    if (pWal->keep) code = walRenew(pWal);
+    if (pWal->keep) terrno = walRenew(pWal);
-    return code;
+    return terrno;
   }
 
   if ( count != (maxId-minId+1) ) {
     wError("wal:%s, messed up, count:%d max:%d min:%d", opath, count, maxId, minId);
-    code = -1;
+    terrno = TAOS_SYSTEM_ERROR(TSDB_CODE_APP_ERROR);
   } else {
     wTrace("wal:%s, %d files will be restored", opath, count);
 
     for (index = minId; index<=maxId; ++index) {
       sprintf(pWal->name, "%s/%s%d", opath, walPrefix, index);
-      code = walRestoreWalFile(pWal, pVnode, writeFp);
+      terrno = walRestoreWalFile(pWal, pVnode, writeFp);
-      if (code < 0) break;
+      if (terrno < 0) break;
     }
   }
 
-  if (code == 0) {
+  if (terrno == 0) {
     if (pWal->keep == 0) {
-      code = walRemoveWalFiles(opath);
+      terrno = walRemoveWalFiles(opath);
-      if (code == 0) {
+      if (terrno == 0) {
         if (remove(opath) < 0) {
           wError("wal:%s, failed to remove directory(%s)", opath, strerror(errno));
-          code = -1;
+          terrno = TAOS_SYSTEM_ERROR(errno);
         }
       }
     } else { 
@@ -258,12 +268,12 @@ int walRestore(void *handle, void *pVnode, int (*writeFp)(void *, void *, int))
       pWal->fd = open(pWal->name, O_WRONLY | O_CREAT | O_APPEND, S_IRWXU | S_IRWXG | S_IRWXO);
       if (pWal->fd < 0) {
         wError("wal:%s, failed to open file(%s)", pWal->name, strerror(errno));
-        code = -1;
+        terrno = TAOS_SYSTEM_ERROR(errno);
       }
     }
   }
 
-  return code;
+  return terrno;
 }
 
 int walGetWalFile(void *handle, char *name, uint32_t *index) {
@@ -292,40 +302,47 @@ int walGetWalFile(void *handle, char *name, uint32_t *index) {
 }  
 
 static int walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp) {
-  int   code = 0;
   char *name = pWal->name;
 
+  terrno = 0;
   char *buffer = malloc(1024000);  // size for one record
-  if (buffer == NULL) return -1;
+  if (buffer == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);   
+    return terrno;
+  }
 
   SWalHead *pHead = (SWalHead *)buffer;
 
   int fd = open(name, O_RDONLY);
   if (fd < 0) {
     wError("wal:%s, failed to open for restore(%s)", name, strerror(errno));
+    terrno = TAOS_SYSTEM_ERROR(errno);
     free(buffer);
-    return -1;
+    return terrno;
   }
 
   wTrace("wal:%s, start to restore", name);
 
   while (1) {
     int ret = read(fd, pHead, sizeof(SWalHead));
-    if ( ret == 0) { code = 0; break;}  
+    if ( ret == 0)  break;  
 
     if (ret != sizeof(SWalHead)) {
       wWarn("wal:%s, failed to read head, skip, ret:%d(%s)", name, ret, strerror(errno));
+      terrno = TAOS_SYSTEM_ERROR(errno);
       break;
     }
 
     if (!taosCheckChecksumWhole((uint8_t *)pHead, sizeof(SWalHead))) {
       wWarn("wal:%s, cksum is messed up, skip the rest of file", name);
+      terrno = TAOS_SYSTEM_ERROR(errno);
       break;
     } 
 
     ret = read(fd, pHead->cont, pHead->len);
     if ( ret != pHead->len) {
       wWarn("wal:%s, failed to read body, skip, len:%d ret:%d", name, pHead->len, ret);
+      terrno = TAOS_SYSTEM_ERROR(errno);
       break;
     }
 
@@ -336,11 +353,10 @@ static int walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp) {
   close(fd);
   free(buffer);
 
-  return code;
+  return terrno;
 }
 
 int walHandleExistingFiles(const char *path) {
-  int    code = 0;
   char   oname[TSDB_FILENAME_LEN * 3];
   char   nname[TSDB_FILENAME_LEN * 3];
   char   opath[TSDB_FILENAME_LEN];
@@ -350,6 +366,7 @@ int walHandleExistingFiles(const char *path) {
   struct dirent *ent;
   DIR   *dir = opendir(path);
   int    plen = strlen(walPrefix);
+  terrno = 0;
 
   if (access(opath, F_OK) == 0) {
     // old directory is there, it means restore process is not finished
@@ -360,13 +377,19 @@ int walHandleExistingFiles(const char *path) {
     int count = 0;
     while ((ent = readdir(dir))!= NULL) {  
       if ( strncmp(ent->d_name, walPrefix, plen) == 0) {
-        if (access(opath, F_OK) != 0) mkdir(opath, 0755);
-
         sprintf(oname, "%s/%s", path, ent->d_name);
         sprintf(nname, "%s/old/%s", path, ent->d_name);
+        if (access(opath, F_OK) != 0) {
+          if (mkdir(opath, 0755) != 0) {
+            wError("wal:%s, failed to create directory:%s(%s)", oname, opath, strerror(errno));
+            terrno = TAOS_SYSTEM_ERROR(errno);
+            break;
+          } 
+        }
+
         if (rename(oname, nname) < 0) {
           wError("wal:%s, failed to move to new:%s", oname, nname);
-          code = -1;
+          terrno = TAOS_SYSTEM_ERROR(errno);
           break;
         } 
 
@@ -378,14 +401,14 @@ int walHandleExistingFiles(const char *path) {
   }
   
   closedir(dir);
-  return code;
+  return terrno;
 }
 
 static int walRemoveWalFiles(const char *path) {
   int    plen = strlen(walPrefix);
   char   name[TSDB_FILENAME_LEN * 3];
-  int    code = 0;
+ 
-
+  terrno = 0;
   if (access(path, F_OK) != 0) return 0;
 
   struct dirent *ent;
@@ -396,13 +419,13 @@ static int walRemoveWalFiles(const char *path) {
       sprintf(name, "%s/%s", path, ent->d_name);
       if (remove(name) <0) {
         wError("wal:%s, failed to remove(%s)", name, strerror(errno));
-        code = -1; break;
+        terrno = TAOS_SYSTEM_ERROR(errno);
       }
     }
   } 
 
   closedir(dir);
 
-  return code;
+  return terrno;
 }
 

tests/pytest/valgrind-test.sh

@@ -1,42 +1,42 @@
 #!/bin/bash
 # insert
-python3 ./test.py -g -f insert/basic.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/basic.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/int.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/int.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/float.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/float.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/bigint.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/bigint.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/bool.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/bool.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/double.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/double.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/smallint.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/smallint.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/tinyint.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/tinyint.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/binary.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/binary.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/date.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/date.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/nchar.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/nchar.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f insert/multi.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f insert/multi.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
 
 # table
-python3 ./test.py -g -f table/column_name.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f table/column_name.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f table/column_num.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f table/column_num.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
-python3 ./test.py -g -f table/db_table.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f table/db_table.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
 
 # import
-python3 ./test.py -g -f import_merge/importDataLastSub.py
+PYTHONMALLOC=malloc python3 ./test.py -g -f import_merge/importDataLastSub.py
-python3 ./test.py -g -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py -g -s && sleep 1
 
 #tag 
-python3 ./test.py $1 -f tag_lite/filter.py
+PYTHONMALLOC=malloc python3 ./test.py $1 -f tag_lite/filter.py
-python3 ./test.py $1 -s && sleep 1
+PYTHONMALLOC=malloc python3 ./test.py $1 -s && sleep 1

tests/script/unique/arbitrator/dn2_mn1_cache_file_sync.sim

@@ -2,11 +2,11 @@
 # step 1: start dnode1
 # step 2: start dnode2 and dnode3, and all added into cluster (Suppose dnode2 is master-vnode)
 # step 3: create db, table, insert data, and Falling disc into file (control only one file, e.g. 1841)
-# step 4: insert old data(now-15d) and new data(now+15d), control data rows  in order to save in cache, not falling disc 
+# step 4: insert old data(now-20d) and new data(now-40d), control data rows  in order to save in cache, not falling disc 
 # step 5: stop dnode2, so date rows falling disc, generate two new files 1840, 1842 in dnode2
-# step 6: insert two data rows: now-16d, now+16d
+# step 6: insert two data rows: now-21d, now-41d
 # step 7: restart dnode2, waiting sync end
-# expect: in dnode2, the files 1840 and 1842 will be removed
+# expect: in dnode2, the files 1837 and 1839 will be removed
 
 system sh/stop_dnodes.sh
 system sh/deploy.sh -n dnode1 -i 1
@@ -14,10 +14,10 @@ system sh/deploy.sh -n dnode2 -i 2
 system sh/deploy.sh -n dnode3 -i 3
 system sh/deploy.sh -n dnode4 -i 4
 
-system sh/cfg.sh -n dnode1 -c numOfMPeers -v 1
+system sh/cfg.sh -n dnode1 -c numOfMnodes -v 1
-system sh/cfg.sh -n dnode2 -c numOfMPeers -v 1
+system sh/cfg.sh -n dnode2 -c numOfMnodes -v 1
-system sh/cfg.sh -n dnode3 -c numOfMPeers -v 1
+system sh/cfg.sh -n dnode3 -c numOfMnodes -v 1
-system sh/cfg.sh -n dnode4 -c numOfMPeers -v 1
+system sh/cfg.sh -n dnode4 -c numOfMnodes -v 1
 
 system sh/cfg.sh -n dnode1 -c walLevel -v 2
 system sh/cfg.sh -n dnode2 -c walLevel -v 2
@@ -39,11 +39,6 @@ system sh/cfg.sh -n dnode2 -c alternativeRole -v 2
 system sh/cfg.sh -n dnode3 -c alternativeRole -v 2
 system sh/cfg.sh -n dnode4 -c alternativeRole -v 2
 
-system sh/cfg.sh -n dnode1 -c numOfTotalVnodes -v 4
-system sh/cfg.sh -n dnode2 -c numOfTotalVnodes -v 4
-system sh/cfg.sh -n dnode3 -c numOfTotalVnodes -v 4
-system sh/cfg.sh -n dnode4 -c numOfTotalVnodes -v 4
-
 system sh/cfg.sh -n dnode1 -c maxtablesPerVnode -v 4
 system sh/cfg.sh -n dnode2 -c maxtablesPerVnode -v 4
 system sh/cfg.sh -n dnode3 -c maxtablesPerVnode -v 4
@@ -64,16 +59,16 @@ sql connect
 
 print ============== step2: start dnode2/dnode3 and add into cluster , then create database with replica 2, and create table, insert data
 system sh/exec.sh -n dnode2 -s start
-#system sh/exec.sh -n dnode3 -s start
+system sh/exec.sh -n dnode3 -s start
 sql create dnode $hostname2
-#sql create dnode $hostname3
+sql create dnode $hostname3
 sleep 3000
 
 $totalTableNum = 1
 $sleepTimer = 3000
 
 $db = db
-sql create database $db replica 1 cache 1
+sql create database $db replica 2 cache 1
 sql use $db
 
 # create table , insert data
@@ -82,7 +77,7 @@ sql create table $stb (ts timestamp, c1 double) tags(t1 int)
 $rowNum = 130000
 $tblNum = $totalTableNum
 $totalRows = 0
-#$tsStart = 1420041600000
+$tsStart = 1577808000000  # 2020-01-01 00:00:00.000
 
 # insert over 2M data in order to falling disc, generate one file
 $i = 0
@@ -102,19 +97,24 @@ while $i < $tblNum
 endw 
 
 sql select count(*) from $stb
-sleep 1000
+print rows:$rows data00:$data00
-print data00 $data00
+if $rows != 1 then
-if $data00 != $totalRows then
+  return -1
-	return -1
 endi
 
+if $data00 == 0 then
+  return -1
+endi
+
+$totalRows = $data00
+
 print ============== step3: insert old data(now-15d) and new data(now+15d), control data rows  in order to save in cache, not falling disc
-sql insert into $tb values ( now - 15d , -15 )
+sql insert into $tb values ( now - 20d , -20 )
-sql insert into $tb values ( now + 15d ,  15 )
+sql insert into $tb values ( now - 40d , -40 )
 $totalRows = $totalRows + 2
 
 print ============== step4: stop dnode2, so date rows falling disc, generate two new files in dnode2
-system sh/exec.sh -n dnode2 -s stop
+system sh/exec.sh -n dnode2 -s stop -x SIGINT
 sleep $sleepTimer
 
 wait_dnode2_offline:
@@ -153,48 +153,14 @@ if $data00 != $totalRows then
 endi
 
 print ============== step5: insert two data rows: now-16d, now+16d, 
-sql insert into $tb values ( now - 16d , -16 )
+sql insert into $tb values ( now - 21d , -21 )
-sql insert into $tb values ( now + 16d ,  16 )
+sql insert into $tb values ( now - 41d , -41 )
 $totalRows = $totalRows + 2
 
-return 1
-
-
-
-print ============== step5: restart dnode2, waiting sync end
-system sh/exec.sh -n dnode2 -s start
-sleep 3000
-
-wait_dnode2_ready:
-sql show dnodes
-if $rows != 3 then
-  sleep 2000
-  goto wait_dnode2_ready
-endi
-print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
-print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
-print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
-#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
-#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
-#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
-#$dnode1Status = $data4_1  
-$dnode2Status = $data4_2
-$dnode3Status = $data4_3  
-#$dnode4Status = $data4_4 
-#$dnode5Status = $data4_5
-
-if $dnode2Status != ready then
-  sleep 2000
-  goto wait_dnode2_ready
-endi
-
-sleep $sleepTimer
-# check using select
 sql select count(*) from $stb
 print data00 $data00
 if $data00 != $totalRows then
 	return -1
 endi
 
-print ============== step6: in dnode2, the files 1840 and 1842 will be removed
+print ============== step6: please check there should be 3 file in sim/dnode2/data/vnode/vnode2/tsdb/data/, and 1 file sim/dnode3/data/vnode/vnode2/tsdb/data/
-# how check in script ???

tests/script/unique/arbitrator/dn2_mn1_cache_file_sync_second.sim

@@ -0,0 +1,52 @@
+# Test case describe: dnode1 is only mnode, dnode2/dnode3 are only vnode
+# step 1: start dnode1
+# step 2: start dnode2 and dnode3, and all added into cluster (Suppose dnode2 is master-vnode)
+# step 3: create db, table, insert data, and Falling disc into file (control only one file, e.g. 1841)
+# step 4: insert old data(now-20d) and new data(now-40d), control data rows  in order to save in cache, not falling disc 
+# step 5: stop dnode2, so date rows falling disc, generate two new files 1840, 1842 in dnode2
+# step 6: insert two data rows: now-21d, now-41d
+# step 7: restart dnode2, waiting sync end
+# expect: in dnode2, the files 1837 and 1839 will be removed
+
+sql connect
+sleep 3000
+
+print ============== step7: restart dnode2, waiting sync end
+system sh/exec.sh -n dnode2 -s start
+sleep 3000
+
+wait_dnode2_ready:
+sql show dnodes
+if $rows != 3 then
+  sleep 2000
+  goto wait_dnode2_ready
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+#$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+$dnode3Status = $data4_3  
+#$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode2Status != ready then
+  sleep 2000
+  goto wait_dnode2_ready
+endi
+
+sleep 1000
+# check using select
+$db = db
+$stb = stb
+sql use $db
+sql select count(*) from $stb
+print data00 $data00, should equal to dn2_mn1_cache_file_sync.sim output
+#if $data00 != $totalRows then
+#	return -1
+#endi
+
+print ============== step8: please check there should be 1 file in sim/dnode2/data/vnode/vnode2/tsdb/data/, and 1 file sim/dnode3/data/vnode/vnode2/tsdb/data/

tests/script/unique/arbitrator/dn3_mn1_replica2_wal1_AddDelDnode.sim

@@ -23,28 +23,39 @@ system sh/cfg.sh -n dnode3 -c balanceInterval -v 10
 system sh/cfg.sh -n dnode4 -c balanceInterval -v 10
 system sh/cfg.sh -n dnode5 -c balanceInterval -v 10
 
-system sh/cfg.sh -n dnode1 -c numOfTotalVnodes -v 200
+system sh/cfg.sh -n dnode1 -c numOfTotalVnodes -v 4
 system sh/cfg.sh -n dnode2 -c numOfTotalVnodes -v 4
 system sh/cfg.sh -n dnode3 -c numOfTotalVnodes -v 4
 system sh/cfg.sh -n dnode4 -c numOfTotalVnodes -v 4
 system sh/cfg.sh -n dnode5 -c numOfTotalVnodes -v 4
 
+system sh/cfg.sh -n dnode1 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode2 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode3 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode4 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode5 -c maxtablesPerVnode -v 4
+
 system sh/cfg.sh -n dnode1 -c mnodeEqualVnodeNum -v 200
 #system sh/cfg.sh -n dnode2 -c mnodeEqualVnodeNum -v 200
 #system sh/cfg.sh -n dnode3 -c mnodeEqualVnodeNum -v 200
 #system sh/cfg.sh -n dnode4 -c mnodeEqualVnodeNum -v 200
 
+system sh/cfg.sh -n dnode1 -c alternativeRole -v 1
+system sh/cfg.sh -n dnode2 -c alternativeRole -v 2
+system sh/cfg.sh -n dnode3 -c alternativeRole -v 2
+system sh/cfg.sh -n dnode4 -c alternativeRole -v 2
+
 system sh/cfg.sh -n dnode1 -c arbitrator -v $arbitrator
 system sh/cfg.sh -n dnode2 -c arbitrator -v $arbitrator
 system sh/cfg.sh -n dnode3 -c arbitrator -v $arbitrator
 system sh/cfg.sh -n dnode4 -c arbitrator -v $arbitrator
 system sh/cfg.sh -n dnode5 -c arbitrator -v $arbitrator
 
-system sh/cfg.sh -n dnode1 -c offlineThreshold -v 20
+system sh/cfg.sh -n dnode1 -c offlineThreshold -v 10
-system sh/cfg.sh -n dnode2 -c offlineThreshold -v 20
+system sh/cfg.sh -n dnode2 -c offlineThreshold -v 10
-system sh/cfg.sh -n dnode3 -c offlineThreshold -v 20
+system sh/cfg.sh -n dnode3 -c offlineThreshold -v 10
-system sh/cfg.sh -n dnode4 -c offlineThreshold -v 20
+system sh/cfg.sh -n dnode4 -c offlineThreshold -v 10
-system sh/cfg.sh -n dnode5 -c offlineThreshold -v 20
+system sh/cfg.sh -n dnode5 -c offlineThreshold -v 10
 
 system sh/cfg.sh -n dnode1 -c enableCoreFile -v 1
 system sh/cfg.sh -n dnode2 -c enableCoreFile -v 1
@@ -63,9 +74,9 @@ sql connect
 print ============== step2: start dnode2/dnode3 and add into cluster, then create database, create table , and insert data
 system sh/exec.sh -n dnode2 -s start
 system sh/exec.sh -n dnode3 -s start
-sleep 1000
 sql create dnode $hostname2
 sql create dnode $hostname3
+sleep 3000
 
 $rowNum = 100
 $tblNum = 16
@@ -151,7 +162,7 @@ endi
 #  return -1
 #endi
 
-sleep 30000
+sleep 15000
 
 wait_drop:
 sql show dnodes
@@ -174,22 +185,36 @@ endi
 if $dnode2Status != ready then
   return -1
 endi
+if $dnode3Status != null then
+  return -1
+endi
 if $dnode4Status != ready then
   return -1
 endi
 
-
+print ============== step4-1: restart dnode3, adn add into cluster
-print ============== step5: start dnode5 and add into cluster , drop database
+system rm -rf ../../sim/dnode3
-sql drop database $db
-sleep 1000
-system sh/exec.sh -n dnode5 -s start
-sql create dnode $hostname5
 sleep 3000
-wait_dnode5:
+
+system sh/deploy.sh -n dnode3 -i 3
+system sh/cfg.sh -n dnode3 -c numOfMnodes -v 1
+system sh/cfg.sh -n dnode3 -c walLevel -v 1
+system sh/cfg.sh -n dnode3 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode3 -c numOfTotalVnodes -v 4
+system sh/cfg.sh -n dnode3 -c arbitrator -v $arbitrator
+system sh/cfg.sh -n dnode3 -c offlineThreshold -v 10
+system sh/cfg.sh -n dnode3 -c enableCoreFile -v 1
+
+system sh/exec.sh -n dnode3 -s start
+sql create dnode $hostname3
+sleep 3000
+
+wait_dnode3_ready:
 sql show dnodes
+print rows: $rows
 if $rows != 4 then
   sleep 3000
-  goto wait_dnode5
+  goto wait_dnode3_ready
 endi
 print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
 print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
@@ -198,6 +223,7 @@ print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
 print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
 $dnode1Status = $data4_1  
 $dnode2Status = $data4_2
+$dnode3Status = $data4_3  
 $dnode4Status = $data4_4  
 $dnode5Status = $data4_5
 
@@ -207,6 +233,9 @@ endi
 if $dnode2Status != ready then
   return -1
 endi
+if $dnode3Status != null then
+  return -1
+endi
 if $dnode4Status != ready then
   return -1
 endi
@@ -214,10 +243,58 @@ if $dnode5Status != ready then
   return -1
 endi
 
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+
+print ============== step5: start dnode5 and add into cluster , drop database
+sql drop database $db
+sleep 1000
+system sh/exec.sh -n dnode5 -s start
+sql create dnode $hostname5
+sleep 3000
+wait_dnode5:
+sql show dnodes
+if $rows != 5 then
+  sleep 3000
+  goto wait_dnode5
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+$dnode4Status = $data4_4  
+$dnode5Status = $data4_5 
+$dnode6Status = $data4_6
+
+if $dnode1Status != ready then
+  return -1
+endi
+if $dnode2Status != ready then
+  return -1
+endi
+if $dnode3Status != null then
+  return -1
+endi
+if $dnode4Status != ready then
+  return -1
+endi
+if $dnode5Status != ready then
+  return -1
+endi
+if $dnode6Status != ready then
+  return -1
+endi
 
 print ============== step6: create database and table until not free vnodes 
 $rowNum = 100
-$tblNum = 24
+$tblNum = 32
 $totalRows = 0
 $tsStart = 1420041600000
 
@@ -259,8 +336,9 @@ if $data00 != $totalRows then
 	return -1
 endi
 
-print ============== step7: drop dnode3, and system should prompt cannot drop dnodes 
+print ============== step7: drop dnode $hostname5, system should prompt "DB error: no enough dnodes" 
-sql_error drop dnode $hostname3
+sql_error drop dnode $hostname5
-print ============== step8: add one new table, and system should prompt 'need more dnode' 
+print error: $error 
+print ============== step8: create table tb_more using $stb tags( 1000 ), system should prompt 'DB error: no enough dnodes' 
 sql_error create table tb_more using $stb tags( 1000 )
-
+print error: $error 

tests/script/unique/arbitrator/dn3_mn1_replica_change.sim

@@ -89,11 +89,8 @@ if $data00 != $totalRows then
 	return -1
 endi
 
-
+print ============== step2-1: stop dnode2 for falling disc, then restart dnode2, and check rows
-
+system sh/exec.sh -n dnode2 -s stop -x SIGINT
-
-print ============== step2-1: start dnode2 for falling disc, then restart dnode2, and check rows
-system sh/exec.sh -n dnode2 -s stop
 sleep $sleepTimer
 wait_dnode2_offline_0:
 sql show dnodes
@@ -151,10 +148,6 @@ if $data00 != $totalRows then
 endi
 
 
-
-
-
-
 print ============== step3: start dnode3 and add into cluster , then alter replica from 1 to 2, and waiting sync
 system sh/exec.sh -n dnode3 -s start
 sql create dnode $hostname3

tests/script/unique/arbitrator/dn3_mn1_vnode_change.sim

@@ -0,0 +1,273 @@
+system sh/stop_dnodes.sh
+system sh/deploy.sh -n dnode1 -i 1
+system sh/deploy.sh -n dnode2 -i 2
+system sh/deploy.sh -n dnode3 -i 3
+system sh/deploy.sh -n dnode4 -i 4
+
+system sh/cfg.sh -n dnode1 -c numOfMnodes -v 1
+system sh/cfg.sh -n dnode2 -c numOfMnodes -v 1
+system sh/cfg.sh -n dnode3 -c numOfMnodes -v 1
+
+system sh/cfg.sh -n dnode1 -c walLevel -v 2
+system sh/cfg.sh -n dnode2 -c walLevel -v 2
+system sh/cfg.sh -n dnode3 -c walLevel -v 2
+system sh/cfg.sh -n dnode4 -c walLevel -v 2
+
+system sh/cfg.sh -n dnode1 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode2 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode3 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode4 -c balanceInterval -v 10
+
+system sh/cfg.sh -n dnode1 -c numOfTotalVnodes -v 4
+system sh/cfg.sh -n dnode2 -c numOfTotalVnodes -v 4
+system sh/cfg.sh -n dnode3 -c numOfTotalVnodes -v 4
+system sh/cfg.sh -n dnode4 -c numOfTotalVnodes -v 4
+
+system sh/cfg.sh -n dnode1 -c alternativeRole -v 1
+system sh/cfg.sh -n dnode2 -c alternativeRole -v 2
+system sh/cfg.sh -n dnode3 -c alternativeRole -v 2
+system sh/cfg.sh -n dnode4 -c alternativeRole -v 2
+
+system sh/cfg.sh -n dnode1 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode2 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode3 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode4 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode5 -c maxtablesPerVnode -v 4
+
+system sh/cfg.sh -n dnode1 -c arbitrator -v $arbitrator
+system sh/cfg.sh -n dnode2 -c arbitrator -v $arbitrator
+system sh/cfg.sh -n dnode3 -c arbitrator -v $arbitrator
+
+print ============== step0: start tarbitrator
+system sh/exec_tarbitrator.sh -s start
+
+print ============== step1: start dnode1, only deploy mnode
+system sh/exec.sh -n dnode1 -s start
+sleep 3000
+sql connect
+
+print ============== step2: start dnode2/dnode3/dnode4 and add into cluster , then create database with replica 3, and create table, insert data
+system sh/exec.sh -n dnode2 -s start
+system sh/exec.sh -n dnode3 -s start
+system sh/exec.sh -n dnode4 -s start
+sql create dnode $hostname2
+sql create dnode $hostname3
+sql create dnode $hostname4
+sleep 3000
+
+$totalTableNum = 100
+$sleepTimer = 3000
+
+$db = db
+sql create database $db replica 3 maxTables $totalTableNum
+sql use $db
+
+# create table , insert data
+$stb = stb
+sql create table $stb (ts timestamp, c1 int) tags(t1 int)
+$rowNum = 100
+$tblNum = $totalTableNum
+$totalRows = 0
+$tsStart = 1420041600000
+
+$i = 0
+while $i < $tblNum
+  $tb = tb . $i
+  sql create table $tb using $stb tags( $i )
+ 
+  $x = 0
+  while $x < $rowNum
+    $ts = $tsStart + $x
+    sql insert into $tb values ( $ts + 0a , $x ) ( $ts + 1a , $x ) ( $ts + 2a , $x ) ( $ts + 3a , $x ) ( $ts + 4a , $x ) ( $ts + 5a , $x ) ( $ts + 6a , $x ) ( $ts + 7a , $x ) ( $ts + 8a , $x ) ( $ts + 9a , $x ) ( $ts + 10a , $x ) ( $ts + 11a , $x ) ( $ts + 12a , $x ) ( $ts + 13a , $x ) ( $ts + 14a , $x ) ( $ts + 15a , $x ) ( $ts + 16a , $x ) ( $ts + 17a , $x ) ( $ts + 18a , $x ) ( $ts + 19a , $x ) ( $ts + 20a , $x ) ( $ts + 21a , $x ) ( $ts + 22a , $x ) ( $ts + 23a , $x ) ( $ts + 24a , $x ) ( $ts + 25a , $x ) ( $ts + 26a , $x ) ( $ts + 27a , $x ) ( $ts + 28a , $x ) ( $ts + 29a , $x ) ( $ts + 30a , $x ) ( $ts + 31a , $x ) ( $ts + 32a , $x ) ( $ts + 33a , $x ) ( $ts + 34a , $x ) ( $ts + 25a , $x ) ( $ts + 26a , $x ) ( $ts + 27a , $x ) ( $ts + 28a , $x ) ( $ts + 29a , $x ) ( $ts + 30a , $x ) ( $ts + 31a , $x ) ( $ts + 32a , $x ) ( $ts + 33a , $x ) ( $ts + 34a , $x ) ( $ts + 35a , $x ) ( $ts + 36a , $x ) ( $ts + 37a , $x ) ( $ts + 38a , $x ) ( $ts + 39a , $x ) ( $ts + 40a , $x ) ( $ts + 41a , $x ) ( $ts + 42a , $x ) ( $ts + 43a , $x ) ( $ts + 44a , $x ) ( $ts + 45a , $x ) ( $ts + 46a , $x ) ( $ts + 47a , $x ) ( $ts + 48a , $x ) ( $ts + 49a , $x ) ( $ts + 50a , $x ) ( $ts + 51a , $x ) ( $ts + 52a , $x ) ( $ts + 53a , $x ) ( $ts + 54a , $x ) ( $ts + 55a , $x ) ( $ts + 56a , $x ) ( $ts + 57a , $x ) ( $ts + 58a , $x ) ( $ts + 59a , $x ) 
+    $x = $x + 60
+  endw 
+  $totalRows = $totalRows + $x
+  print       info: inserted $x rows into $tb and totalRows: $totalRows
+  $i = $i + 1
+endw 
+
+sql select count(*) from $stb
+sleep 1000
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+
+print ============== step3: stop dnode4, and remove its vnodeX subdirector 
+system sh/exec.sh -n dnode4 -s stop -x SIGINT
+sleep $sleepTimer
+wait_dnode4_offline_0:
+sql show dnodes
+if $rows != 4 then
+  sleep 2000
+  goto wait_dnode4_offline_0
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+#$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+$dnode3Status = $data4_3  
+$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode4Status != offline then
+  sleep 2000
+  goto wait_dnode4_offline_0
+endi
+
+wait_dnode4_vgroup_offline:
+sql show vgroups
+if $rows != 1 then
+  sleep 2000
+  goto wait_dnode4_vgroup_offline
+endi
+print show vgroups:
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1  $data5_1  $data6_1  $data7_1  $data8_1  $data9_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2  $data5_2  $data6_2  $data7_2  $data8_2  $data9_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3  $data5_3  $data6_3  $data7_3  $data8_3  $data9_3
+$dnode4Vtatus = $data4_2
+$dnode3Vtatus = $data7_2
+
+if $dnode4Vtatus != offline then
+  sleep 2000
+  goto wait_dnode4_vgroup_offline
+endi
+if $dnode3Vtatus != master then
+  sleep 2000
+  goto wait_dnode4_vgroup_offline
+endi
+
+
+system rm -rf ../../../sim/dnode4/data/vnode/*
+sleep 1000
+
+print ============== step4: restart dnode4, waiting sync end 
+system sh/exec.sh -n dnode4 -s start
+sleep $sleepTimer
+wait_dnode4_reready:
+sql show dnodes
+if $rows != 4 then
+  sleep 2000
+  goto wait_dnode4_reready
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+#$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+$dnode3Status = $data4_3  
+$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode4Status != ready then
+  sleep 2000
+  goto wait_dnode4_reready
+endi
+
+wait_dnode4_vgroup_slave:
+sql show vgroups
+if $rows != 1 then
+  sleep 2000
+  goto wait_dnode4_vgroup_slave
+endi
+print show vgroups:
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1  $data5_1  $data6_1  $data7_1  $data8_1  $data9_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2  $data5_2  $data6_2  $data7_2  $data8_2  $data9_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3  $data5_3  $data6_3  $data7_3  $data8_3  $data9_3
+$dnode4Vtatus = $data4_2
+$dnode3Vtatus = $data7_2
+
+print dnode4Vtatus: $dnode4Vtatus
+print dnode3Vtatus: $dnode3Vtatus 
+if $dnode4Vtatus != slave then
+  sleep 2000
+  goto wait_dnode4_vgroup_slave
+endi
+if $dnode3Vtatus != master then
+  sleep 2000
+  goto wait_dnode4_vgroup_slave
+endi
+
+print ============== step5: stop dnode3/dnode2, and check rows
+system sh/exec.sh -n dnode2 -s stop
+system sh/exec.sh -n dnode3 -s stop
+sleep $sleepTimer
+
+wait_dnode23_offline:
+sql show dnodes
+if $rows != 4 then
+  sleep 2000
+  goto wait_dnode23_offline
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+#$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+$dnode3Status = $data4_3  
+$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode2Status != offline then
+  sleep 2000
+  goto wait_dnode23_offline
+endi
+if $dnode3Status != offline then
+  sleep 2000
+  goto wait_dnode23_offline
+endi
+if $dnode4Status != ready then
+  sleep 2000
+  goto wait_dnode23_offline
+endi
+
+wait_dnode4_vgroup_master:
+sql show vgroups
+if $rows != 1 then
+  sleep 2000
+  goto wait_dnode4_vgroup_master
+endi
+print show vgroups:
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1  $data5_1  $data6_1  $data7_1  $data8_1  $data9_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2  $data5_2  $data6_2  $data7_2  $data8_2  $data9_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3  $data5_3  $data6_3  $data7_3  $data8_3  $data9_3
+$dnode4Vtatus = $data4_2
+$dnode3Vtatus = $data7_2
+
+print dnode4Vtatus: $dnode4Vtatus
+print dnode3Vtatus: $dnode3Vtatus 
+if $dnode4Vtatus != master then
+  sleep 2000
+  goto wait_dnode4_vgroup_master
+endi
+if $dnode3Vtatus != offline then
+  sleep 2000
+  goto wait_dnode4_vgroup_master
+endi
+
+# check using select
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+
+sql insert into $tb values ( now , 20000 ) ( now + 1a, 20001 ) ( now + 2a, 20002 )
+$totalRows = $totalRows + 3
+
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+

tests/script/unique/arbitrator/insert_duplicationTs.sim

@@ -0,0 +1,194 @@
+# Test case describe: dnode1 is only mnode, dnode2/dnode3 are only vnode
+# step 1: start dnode1
+# step 2: start dnode2 and dnode3, and all added into cluster (Suppose dnode2 is master-vnode)
+# step 3: create db, table, insert data, and Falling disc into file (control only one file, e.g. 1841)
+# step 4: insert old data(now-15d) and new data(now+15d), control data rows  in order to save in cache, not falling disc 
+# step 5: stop dnode2, so date rows falling disc, generate two new files 1840, 1842 in dnode2
+# step 6: insert two data rows: now-16d, now+16d
+# step 7: restart dnode2, waiting sync end
+# expect: in dnode2, the files 1840 and 1842 will be removed
+
+system sh/stop_dnodes.sh
+system sh/deploy.sh -n dnode1 -i 1
+system sh/deploy.sh -n dnode2 -i 2
+system sh/deploy.sh -n dnode3 -i 3
+system sh/deploy.sh -n dnode4 -i 4
+
+system sh/cfg.sh -n dnode1 -c numOfMnodes -v 1
+system sh/cfg.sh -n dnode2 -c numOfMnodes -v 1
+system sh/cfg.sh -n dnode3 -c numOfMnodes -v 1
+system sh/cfg.sh -n dnode4 -c numOfMnodes -v 1
+
+system sh/cfg.sh -n dnode1 -c walLevel -v 2
+system sh/cfg.sh -n dnode2 -c walLevel -v 2
+system sh/cfg.sh -n dnode3 -c walLevel -v 2
+system sh/cfg.sh -n dnode4 -c walLevel -v 2
+
+system sh/cfg.sh -n dnode1 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode2 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode3 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode4 -c balanceInterval -v 10
+
+system sh/cfg.sh -n dnode1 -c numOfTotalVnodes -v 4
+system sh/cfg.sh -n dnode2 -c numOfTotalVnodes -v 4
+system sh/cfg.sh -n dnode3 -c numOfTotalVnodes -v 4
+system sh/cfg.sh -n dnode4 -c numOfTotalVnodes -v 4
+
+system sh/cfg.sh -n dnode1 -c alternativeRole -v 1
+system sh/cfg.sh -n dnode2 -c alternativeRole -v 2
+system sh/cfg.sh -n dnode3 -c alternativeRole -v 2
+system sh/cfg.sh -n dnode4 -c alternativeRole -v 2
+
+system sh/cfg.sh -n dnode1 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode2 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode3 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode4 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode5 -c maxtablesPerVnode -v 4
+
+system sh/cfg.sh -n dnode1 -c arbitrator -v $arbitrator
+system sh/cfg.sh -n dnode2 -c arbitrator -v $arbitrator
+system sh/cfg.sh -n dnode3 -c arbitrator -v $arbitrator
+
+print ============== step0: start tarbitrator
+system sh/exec_tarbitrator.sh -s start
+
+print ============== step1: start dnode1, only deploy mnode
+system sh/exec.sh -n dnode1 -s start
+sleep 3000
+sql connect
+
+print ============== step2: start dnode2/dnode3 and add into cluster , then create database with replica 2, and create table, insert data
+system sh/exec.sh -n dnode2 -s start
+system sh/exec.sh -n dnode3 -s start
+sql create dnode $hostname2
+sql create dnode $hostname3
+sleep 3000
+
+$totalTableNum = 1
+$sleepTimer = 3000
+
+$db = db
+sql create database $db replica 2 cache 1
+sql use $db
+
+# create table , insert data
+$stb = stb
+sql create table $stb (ts timestamp, c1 double) tags(t1 int)
+$rowNum = 130000
+$tblNum = $totalTableNum
+$totalRows = 0
+$tsStart = 1577808000000  # 2020-01-01 00:00:00.000
+
+# insert over 2M data in order to falling disc, generate one file
+$i = 0
+while $i < $tblNum
+  $tb = tb . $i
+  sql create table $tb using $stb tags( $i )
+ 
+  $x = 0
+  while $x < $rowNum
+#    $ts = $tsStart + $x
+    sql insert into $tb values ( now + 0s , $x ) ( now + 1s , $x ) ( now + 2s , $x ) ( now + 3s , $x ) ( now + 4s , $x ) ( now + 5s , $x ) ( now + 6s , $x ) ( now + 7s , $x ) ( now + 8s , $x ) ( now + 9s , $x )
+    $x = $x + 10
+  endw 
+  $totalRows = $totalRows + $x
+  print       info: inserted $x rows into $tb and totalRows: $totalRows
+  $i = $i + 1
+endw 
+
+sql select count(*) from $stb
+print rows:$rows data00:$data00
+if $rows != 1 then
+  return -1
+endi
+
+if $data00 == 0 then
+  return -1
+endi
+
+$totalRows = $data00
+
+print ============== step3: insert old data(now-15d) and new data(now+15d), control data rows  in order to save in cache, not falling disc
+sql insert into $tb values ( now - 20d , -20 )
+sql insert into $tb values ( now - 40d , -40 )
+$totalRows = $totalRows + 2
+
+print ============== step4: stop dnode2, so date rows falling disc, generate two new files in dnode2
+system sh/exec.sh -n dnode2 -s stop -x SIGINT
+sleep $sleepTimer
+
+wait_dnode2_offline:
+sql show dnodes
+if $rows != 3 then
+  sleep 2000
+  goto wait_dnode2_offline
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+#$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+$dnode3Status = $data4_3  
+#$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode2Status != offline then
+  sleep 2000
+  goto wait_dnode2_offline
+endi
+if $dnode3Status != ready then
+  sleep 2000
+  goto wait_dnode2_offline
+endi
+
+sleep $sleepTimer   # waitting for move master vnode of dnode2 to dnode3
+# check using select
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+
+print ============== step5: insert two data rows: now-16d, now+16d, 
+sql insert into $tb values ( now - 21d , -21 )
+sql insert into $tb values ( now - 41d , -41 )
+$totalRows = $totalRows + 2
+
+print ============== step5: restart dnode2, waiting sync end
+system sh/exec.sh -n dnode2 -s start
+sleep 3000
+
+wait_dnode2_ready:
+sql show dnodes
+if $rows != 3 then
+  sleep 2000
+  goto wait_dnode2_ready
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+#$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+$dnode3Status = $data4_3  
+#$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode2Status != ready then
+  sleep 2000
+  goto wait_dnode2_ready
+endi
+
+sleep $sleepTimer
+# check using select
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+

tests/script/unique/arbitrator/replica_changeWithArbitrator.sim

@@ -2,18 +2,42 @@ system sh/stop_dnodes.sh
 system sh/deploy.sh -n dnode1 -i 1
 system sh/deploy.sh -n dnode2 -i 2
 system sh/deploy.sh -n dnode3 -i 3
+system sh/deploy.sh -n dnode4 -i 4
 
-system sh/cfg.sh -n dnode1 -c numOfMnodes -v 2
+system sh/cfg.sh -n dnode1 -c numOfMnodes -v 1
-system sh/cfg.sh -n dnode2 -c numOfMnodes -v 2
+system sh/cfg.sh -n dnode2 -c numOfMnodes -v 1
-system sh/cfg.sh -n dnode3 -c numOfMnodes -v 2
+system sh/cfg.sh -n dnode3 -c numOfMnodes -v 1
 
-system sh/cfg.sh -n dnode1 -c walLevel -v 1
+system sh/cfg.sh -n dnode1 -c walLevel -v 2
-system sh/cfg.sh -n dnode2 -c walLevel -v 1
+system sh/cfg.sh -n dnode2 -c walLevel -v 2
-system sh/cfg.sh -n dnode3 -c walLevel -v 1
+system sh/cfg.sh -n dnode3 -c walLevel -v 2
+system sh/cfg.sh -n dnode4 -c walLevel -v 2
 
 system sh/cfg.sh -n dnode1 -c balanceInterval -v 10
 system sh/cfg.sh -n dnode2 -c balanceInterval -v 10
 system sh/cfg.sh -n dnode3 -c balanceInterval -v 10
+system sh/cfg.sh -n dnode4 -c balanceInterval -v 10
+
+system sh/cfg.sh -n dnode1 -c numOfTotalVnodes -v 8
+system sh/cfg.sh -n dnode2 -c numOfTotalVnodes -v 8
+system sh/cfg.sh -n dnode3 -c numOfTotalVnodes -v 8
+system sh/cfg.sh -n dnode4 -c numOfTotalVnodes -v 8
+
+system sh/cfg.sh -n dnode1 -c mnodeEqualVnodeNum -v 4
+#system sh/cfg.sh -n dnode2 -c mnodeEqualVnodeNum -v 4
+#system sh/cfg.sh -n dnode3 -c mnodeEqualVnodeNum -v 4
+#system sh/cfg.sh -n dnode4 -c mnodeEqualVnodeNum -v 4
+
+system sh/cfg.sh -n dnode1 -c alternativeRole -v 0
+system sh/cfg.sh -n dnode2 -c alternativeRole -v 0
+system sh/cfg.sh -n dnode3 -c alternativeRole -v 0
+system sh/cfg.sh -n dnode4 -c alternativeRole -v 0
+
+system sh/cfg.sh -n dnode1 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode2 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode3 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode4 -c maxtablesPerVnode -v 4
+system sh/cfg.sh -n dnode5 -c maxtablesPerVnode -v 4
 
 system sh/cfg.sh -n dnode1 -c arbitrator -v $arbitrator
 system sh/cfg.sh -n dnode2 -c arbitrator -v $arbitrator
@@ -22,21 +46,23 @@ system sh/cfg.sh -n dnode3 -c arbitrator -v $arbitrator
 print ============== step0: start tarbitrator
 system sh/exec_tarbitrator.sh -s start
 
-
+print ============== step1: replica is 1, and start 1 dnode, then create tables and insert data
-print ============== step1: replica is 1, and start 1 dnode
 system sh/exec.sh -n dnode1 -s start
 sleep 3000
 sql connect
 
-$db = replica_db1
+$totalTableNum = 12
-sql create database $db replica 1 maxTables 4
+
+$db = db
+sql create database $db replica 1 maxTables  $totalTableNum
 sql use $db
 
 # create table , insert data
-$stb = repl_stb
+$stb = stb
 sql create table $stb (ts timestamp, c1 int) tags(t1 int)
 $rowNum = 10
-$tblNum = 12
+$tblNum = $totalTableNum
+$totalRows = $rowNum * $tblNum
 
 $ts0 = 1420041600000
 $ts = $ts0
@@ -55,46 +81,136 @@ while $i < $tblNum
     $x = $x + 1
   endw 
   $i = $i + 1
+  print       $tb inserted rows: $x
 endw 
 
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+
 print ============== step2: add 1 new dnode, expect balanced
 system sh/exec.sh -n dnode2 -s start
 sql create dnode $hostname2
+sleep 3000
 
 # expect after balanced, 2 vondes in dnode1, 1 vonde in dnode2
-$x = 0
+wait_dnode2_ready:
-show2: 
-	$x = $x + 1
-	sleep 2000
-	if $x == 10 then
-		return -1
-	endi
-	
 sql show dnodes
-print dnode1 openVnodes $data2_1
+if $rows != 2 then
-print dnode2 openVnodes $data2_2
+  sleep 2000
-if $data2_1 != 2 then
+  goto wait_dnode2_ready
-	goto show2
 endi
-if $data2_2 != 1 then
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
-	goto show2
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+#$dnode3Status = $data4_3  
+#$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode1Status != ready then
+  sleep 2000
+  goto wait_dnode2_ready
+endi
+if $dnode2Status != ready then
+  sleep 2000
+  goto wait_dnode2_ready
 endi
 
-print ============== step4: stop dnode1, and wait dnode2 master
+print ============== step3: stop dnode1/dnode2, modify cfg mpeers to 2, and restart dnode1/dnode2
 system sh/exec.sh -n dnode1 -s stop
+system sh/exec.sh -n dnode2 -s stop
+sleep 3000
 
-$x = 0
+system sh/cfg.sh -n dnode1 -c numOfMnodes -v 2
-loop_wait:
+system sh/cfg.sh -n dnode2 -c numOfMnodes -v 2
-	$x = $x + 1
-	sleep 2000
-	if $x == 10 then
-	  print ERROR: after dnode1 stop, dnode2 didn't become a master! 
-		return -1
-	endi
-sql show mnodes
-$dnodeRole = $data2_1
-print dnodeRole ==> $dnodeRole
 
-if $dnodeRole != master then
+system sh/cfg.sh -n dnode1 -c mnodeEqualVnodeNum -v 4
-  goto loop_wait
+system sh/cfg.sh -n dnode2 -c mnodeEqualVnodeNum -v 4
+
+system sh/cfg.sh -n dnode1 -c alternativeRole -v 1
+system sh/cfg.sh -n dnode2 -c alternativeRole -v 0
+
+system sh/exec.sh -n dnode1 -s start
+system sh/exec.sh -n dnode2 -s start
+sleep 5000
+
+
+print ============= step4: wait dnode ready
+wait_dnode_ready:
+sql show dnodes
+if $rows != 2 then
+  sleep 2000
+  goto wait_dnode_ready
 endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+#print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+#$dnode3Status = $data4_3  
+#$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $dnode1Status != ready then
+  sleep 2000
+  goto wait_dnode_ready
+endi
+if $dnode2Status != ready then
+  sleep 2000
+  goto wait_dnode_ready
+endi
+
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+
+print ============== step5: stop dnode1
+system sh/exec.sh -n dnode1 -s stop
+sleep 3000
+
+wait_dnode2_master:
+sql show mnodes
+if $rows != 2 then
+  sleep 2000
+  goto wait_dnode2_master
+endi
+print $data0_1  $data1_1  $data2_1  $data3_1  $data4_1
+print $data0_2  $data1_2  $data2_2  $data3_2  $data4_2
+#print $data0_3  $data1_3  $data2_3  $data3_3  $data4_3
+#print $data0_4  $data1_4  $data2_4  $data3_4  $data4_4
+#print $data0_5  $data1_5  $data2_5  $data3_5  $data4_5
+#print $data0_6  $data1_6  $data2_6  $data3_6  $data4_6
+$dnode1Status = $data4_1  
+$dnode2Status = $data4_2
+#$dnode3Status = $data4_3  
+#$dnode4Status = $data4_4 
+#$dnode5Status = $data4_5
+
+if $data2_1 != offline then
+  sleep 2000
+  goto wait_dnode2_master
+endi
+if $data2_2 != master then
+  sleep 2000
+  goto wait_dnode2_master
+endi
+
+sql select count(*) from $stb
+print data00 $data00
+if $data00 != $totalRows then
+	return -1
+endi
+
+