Merge branch '3.0' of https://github.com/taosdata/TDengine into dev

example/src/tmq.c

@@ -80,8 +80,9 @@ int32_t create_topic() {
   }
   taos_free_result(pRes);
 
-  const char* sql = "select * from tu1";
+  /*const char* sql = "select * from tu1";*/
-  pRes = tmq_create_topic(pConn, "test_stb_topic_1", sql, strlen(sql));
+  /*pRes = tmq_create_topic(pConn, "test_stb_topic_1", sql, strlen(sql));*/
+  pRes = taos_query(pConn, "create topic test_stb_topic_1 as select * from tu1");
   if (taos_errno(pRes) != 0) {
     printf("failed to create topic test_stb_topic_1, reason:%s\n", taos_errstr(pRes));
     return -1;

include/client/taos.h

@@ -26,9 +26,11 @@ extern "C" {
 typedef void   TAOS;
 typedef void   TAOS_STMT;
 typedef void   TAOS_RES;
-typedef void   TAOS_STREAM;
-typedef void   TAOS_SUB;
 typedef void **TAOS_ROW;
+#if 0
+typedef void   TAOS_STREAM;
+#endif
+typedef void   TAOS_SUB;
 
 // Data type definition
 #define TSDB_DATA_TYPE_NULL       0   // 1 bytes
@@ -122,8 +124,25 @@ typedef struct TAOS_MULTI_BIND {
   int       num;
 } TAOS_MULTI_BIND;
 
+typedef enum {
+  SET_CONF_RET_SUCC = 0,
+  SET_CONF_RET_ERR_PART = -1,
+  SET_CONF_RET_ERR_INNER = -2,
+  SET_CONF_RET_ERR_JSON_INVALID = -3,
+  SET_CONF_RET_ERR_JSON_PARSE = -4,
+  SET_CONF_RET_ERR_ONLY_ONCE = -5,
+  SET_CONF_RET_ERR_TOO_LONG = -6
+} SET_CONF_RET_CODE;
+
+#define RET_MSG_LENGTH 1024
+typedef struct setConfRet {
+  SET_CONF_RET_CODE retCode;
+  char   retMsg[RET_MSG_LENGTH];
+} setConfRet;
+
 DLL_EXPORT void  taos_cleanup(void);
 DLL_EXPORT int   taos_options(TSDB_OPTION option, const void *arg, ...);
+DLL_EXPORT setConfRet   taos_set_config(const char *config);
 DLL_EXPORT TAOS *taos_connect(const char *ip, const char *user, const char *pass, const char *db, uint16_t port);
 DLL_EXPORT TAOS *taos_connect_l(const char *ip, int ipLen, const char *user, int userLen, const char *pass, int passLen,
                                 const char *db, int dbLen, uint16_t port);
@@ -149,6 +168,7 @@ DLL_EXPORT int       taos_stmt_execute(TAOS_STMT *stmt);
 DLL_EXPORT TAOS_RES *taos_stmt_use_result(TAOS_STMT *stmt);
 DLL_EXPORT int       taos_stmt_close(TAOS_STMT *stmt);
 DLL_EXPORT char     *taos_stmt_errstr(TAOS_STMT *stmt);
+DLL_EXPORT int       taos_stmt_affected_rows(TAOS_STMT *stmt);
 
 DLL_EXPORT TAOS_RES *taos_query(TAOS *taos, const char *sql);
 DLL_EXPORT TAOS_RES *taos_query_l(TAOS *taos, const char *sql, int sqlLen);
@@ -165,10 +185,13 @@ DLL_EXPORT int         taos_select_db(TAOS *taos, const char *db);
 DLL_EXPORT int         taos_print_row(char *str, TAOS_ROW row, TAOS_FIELD *fields, int num_fields);
 DLL_EXPORT void        taos_stop_query(TAOS_RES *res);
 DLL_EXPORT bool        taos_is_null(TAOS_RES *res, int32_t row, int32_t col);
+DLL_EXPORT bool        taos_is_update_query(TAOS_RES *res);
 DLL_EXPORT int         taos_fetch_block(TAOS_RES *res, TAOS_ROW *rows);
 DLL_EXPORT int         taos_validate_sql(TAOS *taos, const char *sql);
+DLL_EXPORT void taos_reset_current_db(TAOS *taos);
 
 DLL_EXPORT int *taos_fetch_lengths(TAOS_RES *res);
+DLL_EXPORT TAOS_ROW *taos_result_block(TAOS_RES *res);
 
 DLL_EXPORT const char *taos_get_server_info(TAOS *taos);
 DLL_EXPORT const char *taos_get_client_info();
@@ -179,15 +202,20 @@ DLL_EXPORT int         taos_errno(TAOS_RES *tres);
 DLL_EXPORT void taos_query_a(TAOS *taos, const char *sql, __taos_async_fn_t fp, void *param);
 DLL_EXPORT void taos_fetch_rows_a(TAOS_RES *res, __taos_async_fn_t fp, void *param);
 
+// Shuduo: temporary enable for app build
+#if 1
 typedef void (*__taos_sub_fn_t)(TAOS_SUB *tsub, TAOS_RES *res, void *param, int code);
 DLL_EXPORT TAOS_SUB *taos_subscribe(TAOS *taos, int restart, const char *topic, const char *sql, __taos_sub_fn_t fp,
                                     void *param, int interval);
 DLL_EXPORT TAOS_RES *taos_consume(TAOS_SUB *tsub);
 DLL_EXPORT void      taos_unsubscribe(TAOS_SUB *tsub, int keepProgress);
+#endif
 
+#if 0
 DLL_EXPORT TAOS_STREAM *taos_open_stream(TAOS *taos, const char *sql, void (*fp)(void *param, TAOS_RES *, TAOS_ROW row),
                                          int64_t stime, void *param, void (*callback)(void *));
 DLL_EXPORT void         taos_close_stream(TAOS_STREAM *tstr);
+#endif
 
 DLL_EXPORT int       taos_load_table_info(TAOS *taos, const char *tableNameList);
 DLL_EXPORT TAOS_RES *taos_schemaless_insert(TAOS *taos, char *lines[], int numLines, int protocol, int precision);
@@ -215,8 +243,12 @@ DLL_EXPORT tmq_list_t *tmq_list_new();
 DLL_EXPORT int32_t     tmq_list_append(tmq_list_t *, const char *);
 DLL_EXPORT void        tmq_list_destroy(tmq_list_t *);
 
+// will be removed in 3.0
 DLL_EXPORT tmq_t *tmq_consumer_new(void *conn, tmq_conf_t *conf, char *errstr, int32_t errstrLen);
+
+// will replace last one
 DLL_EXPORT tmq_t *tmq_consumer_new1(tmq_conf_t *conf, char *errstr, int32_t errstrLen);
+
 DLL_EXPORT const char *tmq_err2str(tmq_resp_err_t);
 
 /* ------------------------TMQ CONSUMER INTERFACE------------------------ */
@@ -257,18 +289,24 @@ int32_t tmqGetSkipLogNum(tmq_message_t *tmq_message);
 
 DLL_EXPORT TAOS_ROW    tmq_get_row(tmq_message_t *message);
 DLL_EXPORT char       *tmq_get_topic_name(tmq_message_t *message);
-DLL_EXPORT void    *tmq_get_topic_schema(tmq_t *tmq, const char *topic);
+DLL_EXPORT int32_t     tmq_get_vgroup_id(tmq_message_t *message);
+DLL_EXPORT int64_t     tmq_get_request_offset(tmq_message_t *message);
+DLL_EXPORT int64_t     tmq_get_response_offset(tmq_message_t *message);
+DLL_EXPORT TAOS_FIELD *tmq_get_fields(tmq_t *tmq, const char *topic);
+DLL_EXPORT int32_t     tmq_field_count(tmq_t *tmq, const char *topic);
 DLL_EXPORT void        tmq_message_destroy(tmq_message_t *tmq_message);
 
 /* --------------------TMPORARY INTERFACE FOR TESTING--------------------- */
+#if 0
 DLL_EXPORT TAOS_RES *tmq_create_topic(TAOS *taos, const char *name, const char *sql, int sqlLen);
+#endif
 
 DLL_EXPORT TAOS_RES *tmq_create_stream(TAOS *taos, const char *streamName, const char *tbName, const char *sql);
 
-/* -------------------------------- OTHER -------------------------------- */
+/* ------------------------------ TMQ END -------------------------------- */
+#if 1 // Shuduo: temporary enable for app build
 typedef void (*TAOS_SUBSCRIBE_CALLBACK)(TAOS_SUB *tsub, TAOS_RES *res, void *param, int code);
-
+#endif
-DLL_EXPORT int taos_stmt_affected_rows(TAOS_STMT *stmt);
 
 #ifdef __cplusplus
 }

include/common/tcommon.h

@@ -197,6 +197,11 @@ typedef struct SGroupbyExpr {
   bool    groupbyTag;  // group by tag or column
 } SGroupbyExpr;
 
+enum {
+  FUNC_PARAM_TYPE_VALUE = 0,
+  FUNC_PARAM_TYPE_COLUMN,
+};
+
 typedef struct SFunctParam {
   int32_t  type;
   SColumn* pCol;

include/common/tdatablock.h

@@ -133,7 +133,8 @@ static FORCE_INLINE int32_t colDataAppendInt32(SColumnInfoData* pColumnInfoData,
 }
 
 static FORCE_INLINE int32_t colDataAppendInt64(SColumnInfoData* pColumnInfoData, uint32_t currentRow, int64_t* v) {
-  ASSERT(pColumnInfoData->info.type == TSDB_DATA_TYPE_BIGINT || pColumnInfoData->info.type == TSDB_DATA_TYPE_UBIGINT);
+  int32_t type = pColumnInfoData->info.type;
+  ASSERT(type == TSDB_DATA_TYPE_BIGINT || type == TSDB_DATA_TYPE_UBIGINT || type == TSDB_DATA_TYPE_TIMESTAMP);
   char* p = pColumnInfoData->pData + pColumnInfoData->info.bytes * currentRow;
   *(int64_t*)p = *(int64_t*)v;
 }
@@ -175,18 +176,17 @@ size_t blockDataGetRowSize(SSDataBlock* pBlock);
 double blockDataGetSerialRowSize(const SSDataBlock* pBlock);
 size_t blockDataGetSerialMetaSize(const SSDataBlock* pBlock);
 
-SSchema* blockDataExtractSchema(const SSDataBlock* pBlock, int32_t* numOfCols);
-
 int32_t blockDataSort(SSDataBlock* pDataBlock, SArray* pOrderInfo);
 int32_t blockDataSort_rv(SSDataBlock* pDataBlock, SArray* pOrderInfo, bool nullFirst);
 
 int32_t blockDataEnsureColumnCapacity(SColumnInfoData* pColumn, uint32_t numOfRows);
 int32_t blockDataEnsureCapacity(SSDataBlock* pDataBlock, uint32_t numOfRows);
 void    blockDataCleanup(SSDataBlock* pDataBlock);
-SSDataBlock* createOneDataBlock(const SSDataBlock* pDataBlock);
 size_t  blockDataGetCapacityInRow(const SSDataBlock* pBlock, size_t pageSize);
 void*   blockDataDestroy(SSDataBlock* pBlock);
 
+SSDataBlock* createOneDataBlock(const SSDataBlock* pDataBlock);
+
 void blockDebugShowData(const SArray* dataBlocks);
 
 #ifdef __cplusplus

include/common/tglobal.h

@@ -51,7 +51,7 @@ extern int32_t tsCompatibleModel;
 extern bool    tsEnableSlaveQuery;
 extern bool    tsPrintAuth;
 extern int64_t tsTickPerDay[3];
-extern int32_t tsMultiProcess;
+extern bool    tsMultiProcess;
 
 // monitor
 extern bool     tsEnableMonitor;

include/common/tmsg.h

@@ -469,8 +469,7 @@ typedef struct {
   int32_t tz;  // query client timezone
   char    intervalUnit;
   char    slidingUnit;
-  char
+  char    offsetUnit;  // TODO Remove it, the offset is the number of precision tickle, and it must be a immutable duration.
-      offsetUnit;  // TODO Remove it, the offset is the number of precision tickle, and it must be a immutable duration.
   int8_t  precision;
   int64_t interval;
   int64_t sliding;
@@ -2017,7 +2016,6 @@ typedef struct {
   int8_t   slidingUnit;   // MACRO: TIME_UNIT_XXX
   int8_t   timezoneInt;   // sma data expired if timezone changes.
   char     indexName[TSDB_INDEX_NAME_LEN];
-  char     timezone[TD_TIMEZONE_LEN];
   int32_t  exprLen;
   int32_t  tagsFilterLen;
   int64_t  indexUid;
@@ -2055,32 +2053,6 @@ void*   tDeserializeSVCreateTSmaReq(void* buf, SVCreateTSmaReq* pReq);
 int32_t tSerializeSVDropTSmaReq(void** buf, SVDropTSmaReq* pReq);
 void*   tDeserializeSVDropTSmaReq(void* buf, SVDropTSmaReq* pReq);
 
-typedef struct {
-  col_id_t colId;
-  uint16_t blockSize;  // sma data block size
-  char     data[];
-} STSmaColData;
-
-typedef struct {
-  tb_uid_t tableUid;  // super/child/normal table uid
-  int32_t  dataLen;   // not including head
-  char     data[];
-} STSmaTbData;
-
-typedef struct {
-  int64_t indexUid;
-  TSKEY   skey;  // startKey of one interval/sliding window
-  int64_t interval;
-  int32_t dataLen;  // not including head
-  int8_t  intervalUnit;
-  char    data[];
-} STSmaDataWrapper;  // sma data for a interval/sliding window
-
-// interval/sliding => window
-
-// => window->table->colId
-// => 当一个window下所有的表均计算完成时，流计算告知tsdb清除window的过期标记
-
 // RSma: Rollup SMA
 typedef struct {
   int64_t  interval;
@@ -2215,23 +2187,6 @@ static FORCE_INLINE void* tDecodeTSmaWrapper(void* buf, STSmaWrapper* pSW) {
   return buf;
 }
 
-typedef struct {
-  int64_t uid;
-  int32_t numOfRows;
-  char*   colData;
-} SMqTbData;
-
-typedef struct {
-  char       topicName[TSDB_TOPIC_FNAME_LEN];
-  int64_t    committedOffset;
-  int64_t    reqOffset;
-  int64_t    rspOffset;
-  int32_t    skipLogNum;
-  int32_t    bodyLen;
-  int32_t    numOfTb;
-  SMqTbData* tbData;
-} SMqTopicData;
-
 typedef struct {
   int8_t  mqMsgType;
   int32_t code;
@@ -2260,7 +2215,10 @@ typedef struct {
 
 typedef struct {
   char        topic[TSDB_TOPIC_FNAME_LEN];
+  int8_t      isSchemaAdaptive;
   SArray*     vgs;  // SArray<SMqSubVgEp>
+  int32_t     numOfFields;
+  TAOS_FIELD* fields;
 } SMqSubTopicEp;
 
 typedef struct {
@@ -2281,32 +2239,6 @@ typedef struct {
   SArray*    topics;  // SArray<SMqSubTopicEp>
 } SMqCMGetSubEpRsp;
 
-typedef struct {
-  int32_t curBlock;
-  int32_t curRow;
-  void**  uData;
-} SMqRowIter;
-
-struct tmq_message_t {
-  SMqPollRsp msg;
-  void*      vg;
-  SMqRowIter iter;
-};
-
-#if 0
-struct tmq_message_t {
-  SMqRspHead head;
-  union {
-    SMqPollRsp       consumeRsp;
-    SMqCMGetSubEpRsp getEpRsp;
-  };
-  void*   extra;
-  int32_t curBlock;
-  int32_t curRow;
-  void**  uData;
-};
-#endif
-
 static FORCE_INLINE void tDeleteSMqSubTopicEp(SMqSubTopicEp* pSubTopicEp) { taosArrayDestroy(pSubTopicEp->vgs); }
 
 static FORCE_INLINE int32_t tEncodeSMqSubVgEp(void** buf, const SMqSubVgEp* pVgEp) {
@@ -2331,17 +2263,21 @@ static FORCE_INLINE void tDeleteSMqCMGetSubEpRsp(SMqCMGetSubEpRsp* pRsp) {
 static FORCE_INLINE int32_t tEncodeSMqSubTopicEp(void** buf, const SMqSubTopicEp* pTopicEp) {
   int32_t tlen = 0;
   tlen += taosEncodeString(buf, pTopicEp->topic);
+  tlen += taosEncodeFixedI8(buf, pTopicEp->isSchemaAdaptive);
   int32_t sz = taosArrayGetSize(pTopicEp->vgs);
   tlen += taosEncodeFixedI32(buf, sz);
   for (int32_t i = 0; i < sz; i++) {
     SMqSubVgEp* pVgEp = (SMqSubVgEp*)taosArrayGet(pTopicEp->vgs, i);
     tlen += tEncodeSMqSubVgEp(buf, pVgEp);
   }
+  tlen += taosEncodeFixedI32(buf, pTopicEp->numOfFields);
+  // tlen += taosEncodeBinary(buf, pTopicEp->fields, pTopicEp->numOfFields * sizeof(TAOS_FIELD));
   return tlen;
 }
 
 static FORCE_INLINE void* tDecodeSMqSubTopicEp(void* buf, SMqSubTopicEp* pTopicEp) {
   buf = taosDecodeStringTo(buf, pTopicEp->topic);
+  buf = taosDecodeFixedI8(buf, &pTopicEp->isSchemaAdaptive);
   int32_t sz;
   buf = taosDecodeFixedI32(buf, &sz);
   pTopicEp->vgs = taosArrayInit(sz, sizeof(SMqSubVgEp));
@@ -2353,6 +2289,8 @@ static FORCE_INLINE void* tDecodeSMqSubTopicEp(void* buf, SMqSubTopicEp* pTopicE
     buf = tDecodeSMqSubVgEp(buf, &vgEp);
     taosArrayPush(pTopicEp->vgs, &vgEp);
   }
+  buf = taosDecodeFixedI32(buf, &pTopicEp->numOfFields);
+  // buf = taosDecodeBinary(buf, (void**)&pTopicEp->fields, pTopicEp->numOfFields * sizeof(TAOS_FIELD));
   return buf;
 }
 

include/common/tmsgcb.h

@@ -56,13 +56,14 @@ typedef struct {
   ReleaseHandleFp         releaseHandleFp;
 } SMsgCb;
 
+void    tmsgSetDefaultMsgCb(const SMsgCb* pMsgCb);
 int32_t tmsgPutToQueue(const SMsgCb* pMsgCb, EQueueType qtype, SRpcMsg* pReq);
 int32_t tmsgGetQueueSize(const SMsgCb* pMsgCb, int32_t vgId, EQueueType qtype);
 int32_t tmsgSendReq(const SMsgCb* pMsgCb, const SEpSet* epSet, SRpcMsg* pReq);
 int32_t tmsgSendMnodeReq(const SMsgCb* pMsgCb, SRpcMsg* pReq);
-void    tmsgSendRsp(const SMsgCb* pMsgCb, const SRpcMsg* pRsp);
+void    tmsgSendRsp(const SRpcMsg* pRsp);
 void    tmsgRegisterBrokenLinkArg(const SMsgCb* pMsgCb, SRpcMsg* pMsg);
-void    tmsgReleaseHandle(const SMsgCb* pMsgCb, void* handle, int8_t type);
+void    tmsgReleaseHandle(void* handle, int8_t type);
 
 #ifdef __cplusplus
 }

include/common/trow.h

@@ -937,9 +937,8 @@ static FORCE_INLINE bool tdSTSRowIterNext(STSRowIter *pIter, col_id_t colId, col
     STColumn *pCol = NULL;
     STSchema *pSchema = pIter->pSchema;
     while (pIter->colIdx <= pSchema->numOfCols) {
-      pCol = &pSchema->columns[pIter->colIdx];
+      pCol = &pSchema->columns[pIter->colIdx];  // 1st column of schema is primary TS key
       if (colId == pCol->colId) {
-        ++pIter->colIdx;
         break;
       } else if (colId < pCol->colId) {
         ++pIter->colIdx;
@@ -948,7 +947,8 @@ static FORCE_INLINE bool tdSTSRowIterNext(STSRowIter *pIter, col_id_t colId, col
         return false;
       }
     }
-    return tdGetTpRowDataOfCol(pIter, pCol->type, pCol->offset - sizeof(TSKEY), pVal);
+    tdGetTpRowDataOfCol(pIter, pCol->type, pCol->offset - sizeof(TSKEY), pVal);
+    ++pIter->colIdx;
   } else if (TD_IS_KV_ROW(pIter->pRow)) {
     return tdGetKvRowValOfColEx(pIter, colId, colType, &pIter->kvIdx, pVal);
   } else {

include/common/ttokendef.h

@@ -138,55 +138,59 @@
 #define TK_INTERVAL                       120
 #define TK_TOPIC                          121
 #define TK_AS                             122
-#define TK_NK_BOOL                        123
+#define TK_EXPLAIN                        123
-#define TK_NK_VARIABLE                    124
+#define TK_ANALYZE                        124
-#define TK_NK_UNDERLINE                   125
+#define TK_VERBOSE                        125
-#define TK_ROWTS                          126
+#define TK_NK_BOOL                        126
-#define TK_TBNAME                         127
+#define TK_RATIO                          127
-#define TK_QSTARTTS                       128
+#define TK_NULL                           128
-#define TK_QENDTS                         129
+#define TK_NK_VARIABLE                    129
-#define TK_WSTARTTS                       130
+#define TK_NK_UNDERLINE                   130
-#define TK_WENDTS                         131
+#define TK_ROWTS                          131
-#define TK_WDURATION                      132
+#define TK_TBNAME                         132
-#define TK_BETWEEN                        133
+#define TK_QSTARTTS                       133
-#define TK_IS                             134
+#define TK_QENDTS                         134
-#define TK_NULL                           135
+#define TK_WSTARTTS                       135
-#define TK_NK_LT                          136
+#define TK_WENDTS                         136
-#define TK_NK_GT                          137
+#define TK_WDURATION                      137
-#define TK_NK_LE                          138
+#define TK_BETWEEN                        138
-#define TK_NK_GE                          139
+#define TK_IS                             139
-#define TK_NK_NE                          140
+#define TK_NK_LT                          140
-#define TK_MATCH                          141
+#define TK_NK_GT                          141
-#define TK_NMATCH                         142
+#define TK_NK_LE                          142
-#define TK_IN                             143
+#define TK_NK_GE                          143
-#define TK_JOIN                           144
+#define TK_NK_NE                          144
-#define TK_INNER                          145
+#define TK_MATCH                          145
-#define TK_SELECT                         146
+#define TK_NMATCH                         146
-#define TK_DISTINCT                       147
+#define TK_IN                             147
-#define TK_WHERE                          148
+#define TK_JOIN                           148
-#define TK_PARTITION                      149
+#define TK_INNER                          149
-#define TK_BY                             150
+#define TK_SELECT                         150
-#define TK_SESSION                        151
+#define TK_DISTINCT                       151
-#define TK_STATE_WINDOW                   152
+#define TK_WHERE                          152
-#define TK_SLIDING                        153
+#define TK_PARTITION                      153
-#define TK_FILL                           154
+#define TK_BY                             154
-#define TK_VALUE                          155
+#define TK_SESSION                        155
-#define TK_NONE                           156
+#define TK_STATE_WINDOW                   156
-#define TK_PREV                           157
+#define TK_SLIDING                        157
-#define TK_LINEAR                         158
+#define TK_FILL                           158
-#define TK_NEXT                           159
+#define TK_VALUE                          159
-#define TK_GROUP                          160
+#define TK_NONE                           160
-#define TK_HAVING                         161
+#define TK_PREV                           161
-#define TK_ORDER                          162
+#define TK_LINEAR                         162
-#define TK_SLIMIT                         163
+#define TK_NEXT                           163
-#define TK_SOFFSET                        164
+#define TK_GROUP                          164
-#define TK_LIMIT                          165
+#define TK_HAVING                         165
-#define TK_OFFSET                         166
+#define TK_ORDER                          166
-#define TK_ASC                            167
+#define TK_SLIMIT                         167
-#define TK_DESC                           168
+#define TK_SOFFSET                        168
-#define TK_NULLS                          169
+#define TK_LIMIT                          169
-#define TK_FIRST                          170
+#define TK_OFFSET                         170
-#define TK_LAST                           171
+#define TK_ASC                            171
+#define TK_DESC                           172
+#define TK_NULLS                          173
+#define TK_FIRST                          174
+#define TK_LAST                           175
 
 #define TK_NK_SPACE                       300
 #define TK_NK_COMMENT                     301

include/dnode/mgmt/dnode.h

@@ -46,11 +46,12 @@ typedef struct {
   char      localFqdn[TSDB_FQDN_LEN];
   char      firstEp[TSDB_EP_LEN];
   char      secondEp[TSDB_EP_LEN];
-  SDiskCfg *pDisks;
+  SDiskCfg *disks;
   int32_t   numOfDisks;
+  int8_t    ntype;
 } SDnodeOpt;
 
-typedef enum { DND_EVENT_START, DND_EVENT_STOP = 1, DND_EVENT_RELOAD } EDndEvent;
+typedef enum { DND_EVENT_START, DND_EVENT_STOP = 1, DND_EVENT_CHILD } EDndEvent;
 
 /**
  * @brief Initialize and start the dnode.

include/libs/executor/executor.h

@@ -20,8 +20,8 @@
 extern "C" {
 #endif
 
-#include "tcommon.h"
 #include "query.h"
+#include "tcommon.h"
 
 typedef void* qTaskInfo_t;
 typedef void* DataSinkHandle;
@@ -37,6 +37,11 @@ typedef struct SReadHandle {
 #define STREAM_DATA_TYPE_SUBMIT_BLOCK 0x1
 #define STREAM_DATA_TYPE_SSDATA_BLOCK 0x2
 
+typedef enum {
+  OPTR_EXEC_MODEL_BATCH  = 0x1,
+  OPTR_EXEC_MODEL_STREAM = 0x2,
+} EOPTR_EXEC_MODEL;
+
 /**
  * Create the exec task for streaming mode
  * @param pMsg
@@ -62,7 +67,7 @@ int32_t qSetStreamInput(qTaskInfo_t tinfo, const void* input, int32_t type);
  * @param type
  * @return
  */
-int32_t qSetMultiStreamInput(qTaskInfo_t tinfo, void** pBlocks, size_t numOfBlocks, int32_t type);
+int32_t qSetMultiStreamInput(qTaskInfo_t tinfo, const void* pBlocks, size_t numOfBlocks, int32_t type);
 
 /**
  * Update the table id list, add or remove.
@@ -84,7 +89,7 @@ int32_t qUpdateQualifiedTableId(qTaskInfo_t tinfo, SArray* tableIdList, bool isA
  * @return
  */
 int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, struct SSubplan* pPlan,
-                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle);
+                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, EOPTR_EXEC_MODEL model);
 
 /**
  * The main task execution function, including query on both table and multiple tables,
@@ -167,7 +172,6 @@ int32_t qUpdateQueriedTableIdList(qTaskInfo_t tinfo, int64_t uid, int32_t type);
  */
 void** qReleaseTask(void* pMgmt, void* pQInfo, bool freeHandle);
 
-
 void qProcessFetchRsp(void* parent, struct SRpcMsg* pMsg, struct SEpSet* pEpSet);
 
 #ifdef __cplusplus

include/libs/function/function.h

@@ -41,6 +41,7 @@ typedef void (*FExecFinalize)(struct SqlFunctionCtx *pCtx);
 typedef int32_t (*FScalarExecProcess)(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 
 typedef struct SScalarFuncExecFuncs {
+  FExecGetEnv getEnv;
   FScalarExecProcess process;
 } SScalarFuncExecFuncs;
 
@@ -241,7 +242,6 @@ typedef struct tExprNode {
   };
 } tExprNode;
 
-void exprTreeToBinary(SBufferWriter* bw, tExprNode* pExprTree);
 void tExprTreeDestroy(tExprNode *pNode, void (*fp)(void *));
 
 typedef struct SAggFunctionInfo {
@@ -267,28 +267,6 @@ struct SScalarParam {
   int32_t            numOfRows;
 };
 
-typedef struct SMultiFunctionsDesc {
-  bool stableQuery;
-  bool groupbyColumn;
-  bool agg;
-  bool arithmeticOnAgg;
-  bool projectionQuery;
-  bool hasFilter;
-  bool onlyTagQuery;
-  bool orderProjectQuery;
-  bool globalMerge;
-  bool multigroupResult;
-  bool blockDistribution;
-  bool stateWindow;
-  bool timewindow;
-  bool sessionWindow;
-  bool topbotQuery;
-  bool interpQuery;
-  bool distinct;
-  bool join;
-  bool continueQuery;
-} SMultiFunctionsDesc;
-
 int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionId, int32_t param, SResultDataInfo* pInfo, int16_t extLength,
                           bool isSuperTable);
 
@@ -296,8 +274,6 @@ bool qIsValidUdf(SArray* pUdfInfo, const char* name, int32_t len, int32_t* funct
 
 tExprNode* exprTreeFromBinary(const void* data, size_t size);
 
-void extractFunctionDesc(SArray* pFunctionIdList, SMultiFunctionsDesc* pDesc);
-
 tExprNode* exprdup(tExprNode* pTree);
 
 void resetResultRowEntryResult(SqlFunctionCtx* pCtx, int32_t num);

include/libs/nodes/nodes.h

@@ -70,6 +70,7 @@ typedef enum ENodeType {
   QUERY_NODE_DATABASE_OPTIONS,
   QUERY_NODE_TABLE_OPTIONS,
   QUERY_NODE_INDEX_OPTIONS,
+  QUERY_NODE_EXPLAIN_OPTIONS,
 
   // Statement nodes are used in parser and planner module.
   QUERY_NODE_SET_OPERATOR,
@@ -99,6 +100,7 @@ typedef enum ENodeType {
   QUERY_NODE_CREATE_TOPIC_STMT,
   QUERY_NODE_DROP_TOPIC_STMT,
   QUERY_NODE_ALTER_LOCAL_STMT,
+  QUERY_NODE_EXPLAIN_STMT,
   QUERY_NODE_SHOW_DATABASES_STMT,
   QUERY_NODE_SHOW_TABLES_STMT,
   QUERY_NODE_SHOW_STABLES_STMT,
@@ -121,6 +123,7 @@ typedef enum ENodeType {
   QUERY_NODE_LOGIC_PLAN_EXCHANGE,
   QUERY_NODE_LOGIC_PLAN_WINDOW,
   QUERY_NODE_LOGIC_PLAN_SORT,
+  QUERY_NODE_LOGIC_PLAN_PARTITION,
   QUERY_NODE_LOGIC_SUBPLAN,
   QUERY_NODE_LOGIC_PLAN,
 

include/libs/nodes/plannodes.h

@@ -49,6 +49,7 @@ typedef struct SScanLogicNode {
   STimeWindow scanRange;
   SName tableName;
   bool showRewrite;
+  double ratio;
 } SScanLogicNode;
 
 typedef struct SJoinLogicNode {
@@ -67,6 +68,10 @@ typedef struct SProjectLogicNode {
   SLogicNode node;
   SNodeList* pProjections;
   char stmtName[TSDB_TABLE_NAME_LEN];
+  int64_t limit;
+  int64_t offset;
+  int64_t slimit;
+  int64_t soffset;
 } SProjectLogicNode;
 
 typedef struct SVnodeModifLogicNode {
@@ -106,6 +111,11 @@ typedef struct SSortLogicNode {
   SNodeList* pSortKeys;
 } SSortLogicNode;
 
+typedef struct SPartitionLogicNode {
+  SLogicNode node;
+  SNodeList* pPartitionKeys;
+} SPartitionLogicNode;
+
 typedef enum ESubplanType {
   SUBPLAN_TYPE_MERGE = 1,
   SUBPLAN_TYPE_PARTIAL,
@@ -150,7 +160,8 @@ typedef struct SDataBlockDescNode {
   ENodeType type;
   int16_t dataBlockId;
   SNodeList* pSlots;
-  int32_t resultRowSize;
+  int32_t totalRowSize;
+  int32_t outputRowSize;
   int16_t precision;
 } SDataBlockDescNode;
 
@@ -187,7 +198,7 @@ typedef struct STableScanPhysiNode {
   SScanPhysiNode scan;
   uint8_t scanFlag;         // denotes reversed scan of data or not
   STimeWindow scanRange;
-  SNode* pScanConditions;
+  double ratio;
 } STableScanPhysiNode;
 
 typedef STableScanPhysiNode STableSeqScanPhysiNode;
@@ -195,6 +206,10 @@ typedef STableScanPhysiNode STableSeqScanPhysiNode;
 typedef struct SProjectPhysiNode {
   SPhysiNode node;
   SNodeList* pProjections;
+  int64_t limit;
+  int64_t offset;
+  int64_t slimit;
+  int64_t soffset;
 } SProjectPhysiNode;
 
 typedef struct SJoinPhysiNode {
@@ -238,6 +253,7 @@ typedef struct SIntervalPhysiNode {
   int64_t    sliding;
   int8_t     intervalUnit;
   int8_t     slidingUnit;
+  uint8_t precision;
   SFillNode* pFill;
 } SIntervalPhysiNode;
 
@@ -283,11 +299,23 @@ typedef struct SSubplan {
   SDataSinkNode* pDataSink;    // data of the subplan flow into the datasink
 } SSubplan;
 
+typedef enum EExplainMode {
+  EXPLAIN_MODE_DISABLE = 1,
+  EXPLAIN_MODE_STATIC,
+  EXPLAIN_MODE_ANALYZE
+} EExplainMode;
+
+typedef struct SExplainInfo {
+  EExplainMode mode;
+  bool verbose;
+} SExplainInfo;
+
 typedef struct SQueryPlan {
   ENodeType type;
   uint64_t queryId;
   int32_t numOfSubplans;
   SNodeList* pSubplans; // Element is SNodeListNode. The execution level of subplan, starting from 0.
+  SExplainInfo explainInfo;
 } SQueryPlan;
 
 #ifdef __cplusplus

include/libs/nodes/querynodes.h

@@ -131,6 +131,7 @@ typedef struct SRealTableNode {
   struct STableMeta* pMeta;
   SVgroupsInfo* pVgroupList;
   char useDbName[TSDB_DB_NAME_LEN];
+  double ratio;
 } SRealTableNode;
 
 typedef struct STempTableNode {
@@ -180,8 +181,8 @@ typedef struct SOrderByExprNode {
 
 typedef struct SLimitNode {
   ENodeType type; // QUERY_NODE_LIMIT
-  uint64_t limit;
+  int64_t limit;
-  uint64_t offset;
+  int64_t offset;
 } SLimitNode;
 
 typedef struct SStateWindowNode {
@@ -282,6 +283,19 @@ typedef struct SVnodeModifOpStmt {
   const char* sql;                 // current sql statement position
 } SVnodeModifOpStmt;
 
+typedef struct SExplainOptions {
+  ENodeType type;
+  bool verbose;
+  double ratio;
+} SExplainOptions;
+
+typedef struct SExplainStmt {
+  ENodeType type;
+  bool analyze;
+  SExplainOptions* pOptions;
+  SNode* pQuery;
+} SExplainStmt;
+
 void nodesWalkSelectStmt(SSelectStmt* pSelect, ESqlClause clause, FNodeWalker walker, void* pContext);
 void nodesRewriteSelectStmt(SSelectStmt* pSelect, ESqlClause clause, FNodeRewriter rewriter, void* pContext);
 

include/libs/scalar/scalar.h

@@ -58,6 +58,14 @@ int32_t ceilFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutp
 int32_t floorFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 int32_t roundFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 
+bool getTimePseudoFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
+
+int32_t winStartTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t winEndTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t winDurFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t qStartTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t qEndTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+
 #ifdef __cplusplus
 }
 #endif

include/os/os.h

@@ -42,6 +42,9 @@ extern "C" {
 #include <sys/types.h>
 #include <termios.h>
 #include <sys/statvfs.h>
+#include <sys/prctl.h>
+#include <sys/shm.h>
+#include <sys/wait.h>
 
 #if defined(DARWIN)
 #else
@@ -70,7 +73,6 @@ extern "C" {
 #include <wchar.h>
 #include <wctype.h>
 
-
 #include "osAtomic.h"
 #include "osDef.h"
 #include "osDir.h"
@@ -80,10 +82,12 @@ extern "C" {
 #include "osLz4.h"
 #include "osMath.h"
 #include "osMemory.h"
+#include "osProc.h"
 #include "osRand.h"
 #include "osThread.h"
 #include "osSemaphore.h"
 #include "osSignal.h"
+#include "osShm.h"
 #include "osSleep.h"
 #include "osSocket.h"
 #include "osString.h"

include/os/osProc.h

@@ -0,0 +1,33 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_OS_PROC_H_
+#define _TD_OS_PROC_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// start a copy of itself
+int32_t taosNewProc(const char *args);
+
+// the length of the new name must be less than the original name to take effect
+void taosSetProcName(char **argv, const char *name);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_OS_PROC_H_*/

include/os/osShm.h

@@ -13,4 +13,25 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "consumer.h"
+#ifndef _TD_OS_SHM_H_
+#define _TD_OS_SHM_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+  int32_t id;
+  int32_t size;
+  void*   ptr;
+} SShm;
+
+int32_t taosCreateShm(SShm *pShm, int32_t shmsize) ;
+void    taosDropShm(SShm *pShm);
+int32_t taosAttachShm(SShm *pShm);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_OS_SHM_H_*/

include/os/osSignal.h

@@ -49,6 +49,8 @@ void taosSetSignal(int32_t signum, FSignalHandler sigfp);
 void taosIgnSignal(int32_t signum);
 void taosDflSignal(int32_t signum);
 
+void taosKillChildOnSelfStopped();
+
 #ifdef __cplusplus
 }
 #endif

include/os/osThread.h

@@ -32,6 +32,9 @@ typedef pthread_once_t TdThreadOnce;
 typedef pthread_rwlockattr_t TdThreadRwlockAttr;
 typedef pthread_cond_t TdThreadCond;
 typedef pthread_condattr_t TdThreadCondAttr;
+typedef pthread_key_t TdThreadKey;
+typedef pthread_barrier_t TdThreadBarrier;
+typedef pthread_barrierattr_t TdThreadBarrierAttr;
 
 #define taosThreadCleanupPush pthread_cleanup_push
 #define taosThreadCleanupPop pthread_cleanup_pop
@@ -39,78 +42,190 @@ typedef pthread_condattr_t TdThreadCondAttr;
 // If the error is in a third-party library, place this header file under the third-party library header file.
 // When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC
-    #define pthread_t PTHREAD_T_TYPE_TAOS_FORBID
+    // #define pthread_t PTHREAD_T_TYPE_TAOS_FORBID
-    #define pthread_spinlock_t PTHREAD_SPINLOCK_T_TYPE_TAOS_FORBID
+    // #define pthread_spinlock_t PTHREAD_SPINLOCK_T_TYPE_TAOS_FORBID
-    #define pthread_mutex_t PTHREAD_MUTEX_T_TYPE_TAOS_FORBID
+    // #define pthread_mutex_t PTHREAD_MUTEX_T_TYPE_TAOS_FORBID
-    #define pthread_mutexattr_t PTHREAD_MUTEXATTR_T_TYPE_TAOS_FORBID
+    // #define pthread_mutexattr_t PTHREAD_MUTEXATTR_T_TYPE_TAOS_FORBID
-    #define pthread_rwlock_t PTHREAD_RWLOCK_T_TYPE_TAOS_FORBID
+    // #define pthread_rwlock_t PTHREAD_RWLOCK_T_TYPE_TAOS_FORBID
-    #define pthread_attr_t PTHREAD_ATTR_T_TYPE_TAOS_FORBID
+    // #define pthread_attr_t PTHREAD_ATTR_T_TYPE_TAOS_FORBID
-    #define pthread_once_t PTHREAD_ONCE_T_TYPE_TAOS_FORBID
+    // #define pthread_once_t PTHREAD_ONCE_T_TYPE_TAOS_FORBID
-    #define pthread_rwlockattr_t PTHREAD_RWLOCKATTR_T_TYPE_TAOS_FORBID
+    // #define pthread_rwlockattr_t PTHREAD_RWLOCKATTR_T_TYPE_TAOS_FORBID
-    #define pthread_cond_t PTHREAD_COND_T_TYPE_TAOS_FORBID
+    // #define pthread_cond_t PTHREAD_COND_T_TYPE_TAOS_FORBID
-    #define pthread_condattr_t PTHREAD_CONDATTR_T_TYPE_TAOS_FORBID
+    // #define pthread_condattr_t PTHREAD_CONDATTR_T_TYPE_TAOS_FORBID
-    #define pthread_spin_init PTHREAD_SPIN_INIT_FUNC_TAOS_FORBID
+    // #define pthread_key_t PTHREAD_KEY_T_TYPE_TAOS_FORBID
-    #define pthread_mutex_init PTHREAD_MUTEX_INIT_FUNC_TAOS_FORBID
+    // #define pthread_barrier_t PTHREAD_BARRIER_T_TYPE_TAOS_FORBID
-    #define pthread_spin_destroy PTHREAD_SPIN_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_barrierattr_t PTHREAD_BARRIERATTR_T_TYPE_TAOS_FORBID
-    #define pthread_mutex_destroy PTHREAD_MUTEX_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_create PTHREAD_CREATE_FUNC_TAOS_FORBID
-    #define pthread_spin_lock PTHREAD_SPIN_LOCK_FUNC_TAOS_FORBID
+    // #define pthread_attr_destroy PTHREAD_ATTR_DESTROY_FUNC_TAOS_FORBID
-    #define pthread_mutex_lock PTHREAD_MUTEX_LOCK_FUNC_TAOS_FORBID
+    // #define pthread_attr_getdetachstate PTHREAD_ATTR_GETDETACHSTATE_FUNC_TAOS_FORBID
-    #define pthread_spin_unlock PTHREAD_SPIN_UNLOCK_FUNC_TAOS_FORBID
+    // #define pthread_attr_getinheritsched PTHREAD_ATTR_GETINHERITSCHED_FUNC_TAOS_FORBID
-    #define pthread_mutex_unlock PTHREAD_MUTEX_UNLOCK_FUNC_TAOS_FORBID
+    // #define pthread_attr_getschedparam PTHREAD_ATTR_GETSCHEDPARAM_FUNC_TAOS_FORBID
-    #define pthread_rwlock_rdlock PTHREAD_RWLOCK_RDLOCK_FUNC_TAOS_FORBID
+    // #define pthread_attr_getschedpolicy PTHREAD_ATTR_GETSCHEDPOLICY_FUNC_TAOS_FORBID
-    #define pthread_rwlock_wrlock PTHREAD_RWLOCK_WRLOCK_FUNC_TAOS_FORBID
+    // #define pthread_attr_getscope PTHREAD_ATTR_GETSCOPE_FUNC_TAOS_FORBID
-    #define pthread_rwlock_unlock PTHREAD_RWLOCK_UNLOCK_FUNC_TAOS_FORBID
+    // #define pthread_attr_getstacksize PTHREAD_ATTR_GETSTACKSIZE_FUNC_TAOS_FORBID
-    #define pthread_testcancel PTHREAD_TESTCANCEL_FUNC_TAOS_FORBID
+    // #define pthread_attr_init PTHREAD_ATTR_INIT_FUNC_TAOS_FORBID
-    #define pthread_attr_init PTHREAD_ATTR_INIT_FUNC_TAOS_FORBID
+    // #define pthread_attr_setdetachstate PTHREAD_ATTR_SETDETACHSTATE_FUNC_TAOS_FORBID
-    #define pthread_create PTHREAD_CREATE_FUNC_TAOS_FORBID
+    // #define pthread_attr_setinheritsched PTHREAD_ATTR_SETINHERITSCHED_FUNC_TAOS_FORBID
-    #define pthread_once PTHREAD_ONCE_FUNC_TAOS_FORBID
+    // #define pthread_attr_setschedparam PTHREAD_ATTR_SETSCHEDPARAM_FUNC_TAOS_FORBID
-    #define pthread_attr_setdetachstate PTHREAD_ATTR_SETDETACHSTATE_FUNC_TAOS_FORBID
+    // #define pthread_attr_setschedpolicy PTHREAD_ATTR_SETSCHEDPOLICY_FUNC_TAOS_FORBID
-    #define pthread_attr_destroy PTHREAD_ATTR_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_attr_setscope PTHREAD_ATTR_SETSCOPE_FUNC_TAOS_FORBID
-    #define pthread_join PTHREAD_JOIN_FUNC_TAOS_FORBID
+    // #define pthread_attr_setstacksize PTHREAD_ATTR_SETSTACKSIZE_FUNC_TAOS_FORBID
-    #define pthread_rwlock_init PTHREAD_RWLOCK_INIT_FUNC_TAOS_FORBID
+    // #define pthread_barrier_destroy PTHREAD_BARRIER_DESTROY_FUNC_TAOS_FORBID
-    #define pthread_rwlock_destroy PTHREAD_RWLOCK_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_barrier_init PTHREAD_BARRIER_INIT_FUNC_TAOS_FORBID
-    #define pthread_cond_signal PTHREAD_COND_SIGNAL_FUNC_TAOS_FORBID
+    // #define pthread_barrier_wait PTHREAD_BARRIER_WAIT_FUNC_TAOS_FORBID
-    #define pthread_cond_init PTHREAD_COND_INIT_FUNC_TAOS_FORBID
+    // #define pthread_barrierattr_destroy PTHREAD_BARRIERATTR_DESTROY_FUNC_TAOS_FORBID
-    #define pthread_cond_broadcast PTHREAD_COND_BROADCAST_FUNC_TAOS_FORBID
+    // #define pthread_barrierattr_getpshared PTHREAD_BARRIERATTR_GETPSHARED_FUNC_TAOS_FORBID
-    #define pthread_cond_destroy PTHREAD_COND_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_barrierattr_init PTHREAD_BARRIERATTR_INIT_FUNC_TAOS_FORBID
-    #define pthread_cond_wait PTHREAD_COND_WAIT_FUNC_TAOS_FORBID
+    // #define pthread_barrierattr_setpshared PTHREAD_BARRIERATTR_SETPSHARED_FUNC_TAOS_FORBID
-    #define pthread_self PTHREAD_SELF_FUNC_TAOS_FORBID
+    // #define pthread_cancel PTHREAD_CANCEL_FUNC_TAOS_FORBID
-    #define pthread_equal PTHREAD_EQUAL_FUNC_TAOS_FORBID
+    // #define pthread_cond_destroy PTHREAD_COND_DESTROY_FUNC_TAOS_FORBID
-    #define pthread_sigmask PTHREAD_SIGMASK_FUNC_TAOS_FORBID
+    // #define pthread_cond_init PTHREAD_COND_INIT_FUNC_TAOS_FORBID
-    #define pthread_cancel PTHREAD_CANCEL_FUNC_TAOS_FORBID
+    // #define pthread_cond_signal PTHREAD_COND_SIGNAL_FUNC_TAOS_FORBID
-    #define pthread_kill PTHREAD_KILL_FUNC_TAOS_FORBID
+    // #define pthread_cond_broadcast PTHREAD_COND_BROADCAST_FUNC_TAOS_FORBID
+    // #define pthread_cond_wait PTHREAD_COND_WAIT_FUNC_TAOS_FORBID
+    // #define pthread_cond_timedwait PTHREAD_COND_TIMEDWAIT_FUNC_TAOS_FORBID
+    // #define pthread_condattr_destroy PTHREAD_CONDATTR_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_condattr_getpshared PTHREAD_CONDATTR_GETPSHARED_FUNC_TAOS_FORBID
+    // #define pthread_condattr_init PTHREAD_CONDATTR_INIT_FUNC_TAOS_FORBID
+    // #define pthread_condattr_setpshared PTHREAD_CONDATTR_SETPSHARED_FUNC_TAOS_FORBID
+    // #define pthread_detach PTHREAD_DETACH_FUNC_TAOS_FORBID
+    // #define pthread_equal PTHREAD_EQUAL_FUNC_TAOS_FORBID
+    // #define pthread_exit PTHREAD_EXIT_FUNC_TAOS_FORBID
+    // #define pthread_getschedparam PTHREAD_GETSCHEDPARAM_FUNC_TAOS_FORBID
+    // #define pthread_getspecific PTHREAD_GETSPECIFIC_FUNC_TAOS_FORBID
+    // #define pthread_join PTHREAD_JOIN_FUNC_TAOS_FORBID
+    // #define pthread_key_create PTHREAD_KEY_CREATE_FUNC_TAOS_FORBID
+    // #define pthread_key_delete PTHREAD_KEY_DELETE_FUNC_TAOS_FORBID
+    // #define pthread_kill PTHREAD_KILL_FUNC_TAOS_FORBID
+    // #define pthread_mutex_consistent PTHREAD_MUTEX_CONSISTENT_FUNC_TAOS_FORBID
+    // #define pthread_mutex_destroy PTHREAD_MUTEX_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_mutex_init PTHREAD_MUTEX_INIT_FUNC_TAOS_FORBID
+    // #define pthread_mutex_lock PTHREAD_MUTEX_LOCK_FUNC_TAOS_FORBID
+    // #define pthread_mutex_timedlock PTHREAD_MUTEX_TIMEDLOCK_FUNC_TAOS_FORBID
+    // #define pthread_mutex_trylock PTHREAD_MUTEX_TRYLOCK_FUNC_TAOS_FORBID
+    // #define pthread_mutex_unlock PTHREAD_MUTEX_UNLOCK_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_destroy PTHREAD_MUTEXATTR_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_getpshared PTHREAD_MUTEXATTR_GETPSHARED_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_getrobust PTHREAD_MUTEXATTR_GETROBUST_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_gettype PTHREAD_MUTEXATTR_GETTYPE_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_init PTHREAD_MUTEXATTR_INIT_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_setpshared PTHREAD_MUTEXATTR_SETPSHARED_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_setrobust PTHREAD_MUTEXATTR_SETROBUST_FUNC_TAOS_FORBID
+    // #define pthread_mutexattr_settype PTHREAD_MUTEXATTR_SETTYPE_FUNC_TAOS_FORBID
+    // #define pthread_once PTHREAD_ONCE_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_destroy PTHREAD_RWLOCK_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_init PTHREAD_RWLOCK_INIT_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_rdlock PTHREAD_RWLOCK_RDLOCK_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_timedrdlock PTHREAD_RWLOCK_TIMEDRDLOCK_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_timedwrlock PTHREAD_RWLOCK_TIMEDWRLOCK_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_tryrdlock PTHREAD_RWLOCK_TRYRDLOCK_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_trywrlock PTHREAD_RWLOCK_TRYWRLOCK_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_unlock PTHREAD_RWLOCK_UNLOCK_FUNC_TAOS_FORBID
+    // #define pthread_rwlock_wrlock PTHREAD_RWLOCK_WRLOCK_FUNC_TAOS_FORBID
+    // #define pthread_rwlockattr_destroy PTHREAD_RWLOCKATTR_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_rwlockattr_getpshared PTHREAD_RWLOCKATTR_GETPSHARED_FUNC_TAOS_FORBID
+    // #define pthread_rwlockattr_init PTHREAD_RWLOCKATTR_INIT_FUNC_TAOS_FORBID
+    // #define pthread_rwlockattr_setpshared PTHREAD_RWLOCKATTR_SETPSHARED_FUNC_TAOS_FORBID
+    // #define pthread_self PTHREAD_SELF_FUNC_TAOS_FORBID
+    // #define pthread_setcancelstate PTHREAD_SETCANCELSTATE_FUNC_TAOS_FORBID
+    // #define pthread_setcanceltype PTHREAD_SETCANCELTYPE_FUNC_TAOS_FORBID
+    // #define pthread_setschedparam PTHREAD_SETSCHEDPARAM_FUNC_TAOS_FORBID
+    // #define pthread_setspecific PTHREAD_SETSPECIFIC_FUNC_TAOS_FORBID
+    // #define pthread_spin_destroy PTHREAD_SPIN_DESTROY_FUNC_TAOS_FORBID
+    // #define pthread_spin_init PTHREAD_SPIN_INIT_FUNC_TAOS_FORBID
+    // #define pthread_spin_lock PTHREAD_SPIN_LOCK_FUNC_TAOS_FORBID
+    // #define pthread_spin_trylock PTHREAD_SPIN_TRYLOCK_FUNC_TAOS_FORBID
+    // #define pthread_spin_unlock PTHREAD_SPIN_UNLOCK_FUNC_TAOS_FORBID
+    // #define pthread_testcancel PTHREAD_TESTCANCEL_FUNC_TAOS_FORBID
+    // #define pthread_sigmask PTHREAD_SIGMASK_FUNC_TAOS_FORBID
+    // #define sigwait SIGWAIT_FUNC_TAOS_FORBID
 #endif
 
-int32_t taosThreadSpinInit(TdThreadSpinlock *lock, int pshared);
-int32_t taosThreadMutexInit(TdThreadMutex *mutex, const TdThreadMutexAttr *attr);
-int32_t taosThreadSpinDestroy(TdThreadSpinlock *lock);
-int32_t taosThreadMutexDestroy(TdThreadMutex * mutex);
-int32_t taosThreadSpinLock(TdThreadSpinlock *lock);
-int32_t taosThreadMutexLock(TdThreadMutex *mutex);
-int32_t taosThreadRwlockRdlock(TdThreadRwlock *rwlock);
-int32_t taosThreadSpinUnlock(TdThreadSpinlock *lock);
-int32_t taosThreadMutexUnlock(TdThreadMutex *mutex);
-int32_t taosThreadRwlockWrlock(TdThreadRwlock *rwlock);
-int32_t taosThreadRwlockUnlock(TdThreadRwlock *rwlock);
-void taosThreadTestCancel(void);
-int32_t taosThreadAttrInit(TdThreadAttr *attr);
 int32_t taosThreadCreate(TdThread * tid, const TdThreadAttr * attr, void *(*start)(void *), void *arg);
-int32_t taosThreadOnce(TdThreadOnce *onceControl, void(*initRoutine)(void));
-int32_t taosThreadAttrSetDetachState(TdThreadAttr *attr, int32_t detachState);
 int32_t taosThreadAttrDestroy(TdThreadAttr * attr);
-int32_t taosThreadJoin(TdThread thread, void **pValue);
+int32_t taosThreadAttrGetDetachState(const TdThreadAttr * attr, int32_t *detachstate);
-int32_t taosThreadRwlockInit(TdThreadRwlock *rwlock, const TdThreadRwlockAttr *attr);
+int32_t taosThreadAttrGetInheritSched(const TdThreadAttr * attr, int32_t *inheritsched);
-int32_t taosThreadRwlockDestroy(TdThreadRwlock *rwlock);
+int32_t taosThreadAttrGetSchedParam(const TdThreadAttr * attr, struct sched_param *param);
-int32_t taosThreadCondSignal(TdThreadCond *cond);
+int32_t taosThreadAttrGetSchedPolicy(const TdThreadAttr * attr, int32_t *policy);
-int32_t taosThreadCondInit(TdThreadCond *cond, const TdThreadCondAttr *attr);
+int32_t taosThreadAttrGetScope(const TdThreadAttr * attr, int32_t *contentionscope);
-int32_t taosThreadCondBroadcast(TdThreadCond *cond);
+int32_t taosThreadAttrGetStackSize(const TdThreadAttr * attr, size_t * stacksize);
-int32_t taosThreadCondDestroy(TdThreadCond *cond);
+int32_t taosThreadAttrInit(TdThreadAttr * attr);
-int32_t taosThreadCondWait(TdThreadCond *cond, TdThreadMutex *mutex);
+int32_t taosThreadAttrSetDetachState(TdThreadAttr * attr, int32_t detachstate);
-TdThread taosThreadSelf(void);
+int32_t taosThreadAttrSetInheritSched(TdThreadAttr * attr, int32_t inheritsched);
-int32_t taosThreadEqual(TdThread t1, TdThread t2);
+int32_t taosThreadAttrSetSchedParam(TdThreadAttr * attr, const struct sched_param *param);
-int32_t taosThreadSigmask(int how, sigset_t const *set, sigset_t *oset);
+int32_t taosThreadAttrSetSchedPolicy(TdThreadAttr * attr, int32_t policy);
+int32_t taosThreadAttrSetScope(TdThreadAttr * attr, int32_t contentionscope);
+int32_t taosThreadAttrSetStackSize(TdThreadAttr * attr, size_t stacksize);
+int32_t taosThreadBarrierDestroy(TdThreadBarrier * barrier);
+int32_t taosThreadBarrierInit(TdThreadBarrier * barrier, const TdThreadBarrierAttr * attr, uint32_t count);
+int32_t taosThreadBarrierWait(TdThreadBarrier * barrier);
+int32_t taosThreadBarrierAttrDestroy(TdThreadBarrierAttr * attr);
+int32_t taosThreadBarrierAttrGetPshared(const TdThreadBarrierAttr * attr, int32_t *pshared);
+int32_t taosThreadBarrierAttrInit(TdThreadBarrierAttr * attr);
+int32_t taosThreadBarrierAttrSetPshared(TdThreadBarrierAttr * attr, int32_t pshared);
 int32_t taosThreadCancel(TdThread thread);
-int32_t taosThreadKill(TdThread thread, int sig);
+int32_t taosThreadCondDestroy(TdThreadCond * cond);
+int32_t taosThreadCondInit(TdThreadCond * cond, const TdThreadCondAttr * attr);
+int32_t taosThreadCondSignal(TdThreadCond * cond);
+int32_t taosThreadCondBroadcast(TdThreadCond * cond);
+int32_t taosThreadCondWait(TdThreadCond * cond, TdThreadMutex * mutex);
+int32_t taosThreadCondTimedWait(TdThreadCond * cond, TdThreadMutex * mutex, const struct timespec *abstime);
+int32_t taosThreadCondAttrDestroy(TdThreadCondAttr * attr);
+int32_t taosThreadCondAttrGetPshared(const TdThreadCondAttr * attr, int32_t *pshared);
+int32_t taosThreadCondAttrInit(TdThreadCondAttr * attr);
+int32_t taosThreadCondAttrSetPshared(TdThreadCondAttr * attr, int32_t pshared);
+int32_t taosThreadDetach(TdThread thread);
+int32_t taosThreadEqual(TdThread t1, TdThread t2);
+void taosThreadExit(void *valuePtr);
+int32_t taosThreadGetSchedParam(TdThread thread, int32_t *policy, struct sched_param *param);
+void *taosThreadGetSpecific(TdThreadKey key);
+int32_t taosThreadJoin(TdThread thread, void **valuePtr);
+int32_t taosThreadKeyCreate(TdThreadKey * key, void(*destructor)(void *));
+int32_t taosThreadKeyDelete(TdThreadKey key);
+int32_t taosThreadKill(TdThread thread, int32_t sig);
+int32_t taosThreadMutexConsistent(TdThreadMutex* mutex);
+int32_t taosThreadMutexDestroy(TdThreadMutex * mutex);
+int32_t taosThreadMutexInit(TdThreadMutex * mutex, const TdThreadMutexAttr * attr);
+int32_t taosThreadMutexLock(TdThreadMutex * mutex);
+int32_t taosThreadMutexTimedLock(TdThreadMutex * mutex, const struct timespec *abstime);
+int32_t taosThreadMutexTryLock(TdThreadMutex * mutex);
+int32_t taosThreadMutexUnlock(TdThreadMutex * mutex);
+int32_t taosThreadMutexAttrDestroy(TdThreadMutexAttr * attr);
+int32_t taosThreadMutexAttrGetPshared(const TdThreadMutexAttr * attr, int32_t *pshared);
+int32_t taosThreadMutexAttrGetRobust(const TdThreadMutexAttr * attr, int32_t * robust);
+int32_t taosThreadMutexAttrGetType(const TdThreadMutexAttr * attr, int32_t *kind);
+int32_t taosThreadMutexAttrInit(TdThreadMutexAttr * attr);
+int32_t taosThreadMutexAttrSetPshared(TdThreadMutexAttr * attr, int32_t pshared);
+int32_t taosThreadMutexAttrSetRobust(TdThreadMutexAttr * attr, int32_t robust);
+int32_t taosThreadMutexAttrSetType(TdThreadMutexAttr * attr, int32_t kind);
+int32_t taosThreadOnce(TdThreadOnce * onceControl, void(*initRoutine)(void));
+int32_t taosThreadRwlockDestroy(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockInit(TdThreadRwlock * rwlock, const TdThreadRwlockAttr * attr);
+int32_t taosThreadRwlockRdlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockTimedRdlock(TdThreadRwlock * rwlock, const struct timespec *abstime);
+int32_t taosThreadRwlockTimedWrlock(TdThreadRwlock * rwlock, const struct timespec *abstime);
+int32_t taosThreadRwlockTryRdlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockTryWrlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockUnlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockWrlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockAttrDestroy(TdThreadRwlockAttr * attr);
+int32_t taosThreadRwlockAttrGetPshared(const TdThreadRwlockAttr * attr, int32_t *pshared);
+int32_t taosThreadRwlockAttrInit(TdThreadRwlockAttr * attr);
+int32_t taosThreadRwlockAttrSetPshared(TdThreadRwlockAttr * attr, int32_t pshared);
+TdThread taosThreadSelf(void);
+int32_t taosThreadSetCancelState(int32_t state, int32_t *oldstate);
+int32_t taosThreadSetCancelType(int32_t type, int32_t *oldtype);
+int32_t taosThreadSetSchedParam(TdThread thread, int32_t policy, const struct sched_param *param);
+int32_t taosThreadSetSpecific(TdThreadKey key, const void *value);
+int32_t taosThreadSpinDestroy(TdThreadSpinlock * lock);
+int32_t taosThreadSpinInit(TdThreadSpinlock * lock, int32_t pshared);
+int32_t taosThreadSpinLock(TdThreadSpinlock * lock);
+int32_t taosThreadSpinTrylock(TdThreadSpinlock * lock);
+int32_t taosThreadSpinUnlock(TdThreadSpinlock * lock);
+void taosThreadTestCancel(void);
+int32_t taosThreadSigMask(int32_t how, sigset_t const *set, sigset_t * oset);
+int32_t taosThreadSigWait(const sigset_t * set, int32_t *sig);
+
 #ifdef __cplusplus
 }
 #endif

include/os/osTime.h

@@ -20,6 +20,8 @@
 extern "C" {
 #endif
 
+#include <time.h>
+
 // If the error is in a third-party library, place this header file under the third-party library header file.
 // When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC

include/util/taoserror.h

@@ -16,7 +16,7 @@
 #ifndef _TD_UTIL_TAOS_ERROR_H_
 #define _TD_UTIL_TAOS_ERROR_H_
 
-#include "os.h"
+#include <stdint.h>
 
 #ifdef __cplusplus
 extern "C" {
@@ -78,6 +78,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_CFG_NOT_FOUND                 TAOS_DEF_ERROR_CODE(0, 0x010C)
 #define TSDB_CODE_INVALID_CFG                   TAOS_DEF_ERROR_CODE(0, 0x010D)
 #define TSDB_CODE_OUT_OF_SHM_MEM                TAOS_DEF_ERROR_CODE(0, 0x010E)
+#define TSDB_CODE_INVALID_SHM_ID                TAOS_DEF_ERROR_CODE(0, 0x010F)
 #define TSDB_CODE_REF_NO_MEMORY                 TAOS_DEF_ERROR_CODE(0, 0x0110)
 #define TSDB_CODE_REF_FULL                      TAOS_DEF_ERROR_CODE(0, 0x0111)
 #define TSDB_CODE_REF_ID_REMOVED                TAOS_DEF_ERROR_CODE(0, 0x0112)

include/util/tdef.h

@@ -381,6 +381,14 @@ typedef enum ELogicConditionType {
 #define TSDB_MAX_DB_DELAY     10
 #define TSDB_DEFAULT_DB_DELAY 2
 
+#define TSDB_DEFAULT_EXPLAIN_VERBOSE false
+
+#define TSDB_MIN_EXPLAIN_RATIO       0
+#define TSDB_MAX_EXPLAIN_RATIO       1
+#define TSDB_DEFAULT_EXPLAIN_RATIO   0.001
+
+#define TSDB_EXPLAIN_RESULT_ROW_SIZE 1024
+
 #define TSDB_MAX_JOIN_TABLE_NUM 10
 #define TSDB_MAX_UNION_CLAUSE   5
 

include/util/tprocess.h

@@ -22,7 +22,7 @@
 extern "C" {
 #endif
 
-typedef enum { PROC_REQ, PROC_RSP, PROC_REGISTER } ProcFuncType;
+typedef enum { PROC_REQ, PROC_RSP, PROC_REG, PROC_RELEASE } ProcFuncType;
 
 typedef struct SProcQueue SProcQueue;
 typedef struct SProcObj   SProcObj;
@@ -32,29 +32,25 @@ typedef void *(*ProcConsumeFp)(void *pParent, void *pHead, int16_t headLen, void
                                ProcFuncType ftype);
 
 typedef struct {
-  int32_t       childQueueSize;
   ProcConsumeFp childConsumeFp;
   ProcMallocFp  childMallocHeadFp;
   ProcFreeFp    childFreeHeadFp;
   ProcMallocFp  childMallocBodyFp;
   ProcFreeFp    childFreeBodyFp;
-  int32_t       parentQueueSize;
   ProcConsumeFp parentConsumeFp;
-  ProcMallocFp  parentdMallocHeadFp;
+  ProcMallocFp  parentMallocHeadFp;
   ProcFreeFp    parentFreeHeadFp;
   ProcMallocFp  parentMallocBodyFp;
   ProcFreeFp    parentFreeBodyFp;
-  bool          testFlag;
+  SShm          shm;
   void         *pParent;
   const char   *name;
+  bool          isChild;
 } SProcCfg;
 
 SProcObj *taosProcInit(const SProcCfg *pCfg);
 void      taosProcCleanup(SProcObj *pProc);
 int32_t   taosProcRun(SProcObj *pProc);
-void      taosProcStop(SProcObj *pProc);
-bool      taosProcIsChild(SProcObj *pProc);
-int32_t   taosProcChildId(SProcObj *pProc);
 int32_t   taosProcPutToChildQ(SProcObj *pProc, const void *pHead, int16_t headLen, const void *pBody, int32_t bodyLen,
                               ProcFuncType ftype);
 int32_t   taosProcPutToParentQ(SProcObj *pProc, const void *pHead, int16_t headLen, const void *pBody, int32_t bodyLen,

source/client/src/clientEnv.c

@@ -200,6 +200,9 @@ static void doDestroyRequest(void *p) {
     taosArrayDestroy(pRequest->body.showInfo.pArray);
   }
 
+  taosArrayDestroy(pRequest->tableList);
+  taosArrayDestroy(pRequest->dbList);
+
   deregisterRequest(pRequest);
   taosMemoryFreeClear(pRequest);
 }

source/client/src/clientMain.c

@@ -1,12 +1,27 @@
-#include "os.h"
+/*
-#include "tref.h"
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
-#include "trpc.h"
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "catalog.h"
 #include "clientInt.h"
 #include "clientLog.h"
+#include "os.h"
 #include "query.h"
-#include "tmsg.h"
 #include "tglobal.h"
-#include "catalog.h"
+#include "tmsg.h"
+#include "tref.h"
+#include "trpc.h"
 #include "version.h"
 
 #define TSC_VAR_NOT_RELEASE 1
@@ -56,6 +71,12 @@ void taos_cleanup(void) {
   tscInfo("all local resources released");
 }
 
+setConfRet   taos_set_config(const char *config) {
+  // TODO
+  setConfRet ret = {SET_CONF_RET_SUCC, {0}};
+  return ret;
+}
+
 TAOS *taos_connect(const char *ip, const char *user, const char *pass, const char *db, uint16_t port) {
   tscDebug("try to connect to %s:%u, user:%s db:%s", ip, port, user, db);
   if (user == NULL) {
@@ -117,9 +138,7 @@ int  taos_field_count(TAOS_RES *res) {
   return pResInfo->numOfCols;
 }
 
-int  taos_num_fields(TAOS_RES *res) {
+int taos_num_fields(TAOS_RES *res) { return taos_field_count(res); }
-  return taos_field_count(res);
-}
 
 TAOS_FIELD *taos_fetch_fields(TAOS_RES *res) {
   if (taos_num_fields(res) == 0) {
@@ -144,10 +163,8 @@ TAOS_ROW taos_fetch_row(TAOS_RES *res) {
   }
 
   SRequestObj *pRequest = (SRequestObj *)res;
-  if (pRequest->type == TSDB_SQL_RETRIEVE_EMPTY_RESULT ||
+  if (pRequest->type == TSDB_SQL_RETRIEVE_EMPTY_RESULT || pRequest->type == TSDB_SQL_INSERT ||
-      pRequest->type == TSDB_SQL_INSERT ||
+      pRequest->code != TSDB_CODE_SUCCESS || taos_num_fields(res) == 0) {
-      pRequest->code != TSDB_CODE_SUCCESS ||
-      taos_num_fields(res) == 0) {
     return NULL;
   }
 
@@ -246,23 +263,41 @@ int* taos_fetch_lengths(TAOS_RES *res) {
   return ((SRequestObj *)res)->body.resInfo.length;
 }
 
+TAOS_ROW *taos_result_block(TAOS_RES *res) {
+  // TODO
+  return NULL;
+}
+
 // todo intergrate with tDataTypes
 const char *taos_data_type(int type) {
   switch (type) {
-    case TSDB_DATA_TYPE_NULL:            return "TSDB_DATA_TYPE_NULL";
+    case TSDB_DATA_TYPE_NULL:
-    case TSDB_DATA_TYPE_BOOL:            return "TSDB_DATA_TYPE_BOOL";
+      return "TSDB_DATA_TYPE_NULL";
-    case TSDB_DATA_TYPE_TINYINT:         return "TSDB_DATA_TYPE_TINYINT";
+    case TSDB_DATA_TYPE_BOOL:
-    case TSDB_DATA_TYPE_SMALLINT:        return "TSDB_DATA_TYPE_SMALLINT";
+      return "TSDB_DATA_TYPE_BOOL";
-    case TSDB_DATA_TYPE_INT:             return "TSDB_DATA_TYPE_INT";
+    case TSDB_DATA_TYPE_TINYINT:
-    case TSDB_DATA_TYPE_BIGINT:          return "TSDB_DATA_TYPE_BIGINT";
+      return "TSDB_DATA_TYPE_TINYINT";
-    case TSDB_DATA_TYPE_FLOAT:           return "TSDB_DATA_TYPE_FLOAT";
+    case TSDB_DATA_TYPE_SMALLINT:
-    case TSDB_DATA_TYPE_DOUBLE:          return "TSDB_DATA_TYPE_DOUBLE";
+      return "TSDB_DATA_TYPE_SMALLINT";
-    case TSDB_DATA_TYPE_VARCHAR:         return "TSDB_DATA_TYPE_VARCHAR";
+    case TSDB_DATA_TYPE_INT:
+      return "TSDB_DATA_TYPE_INT";
+    case TSDB_DATA_TYPE_BIGINT:
+      return "TSDB_DATA_TYPE_BIGINT";
+    case TSDB_DATA_TYPE_FLOAT:
+      return "TSDB_DATA_TYPE_FLOAT";
+    case TSDB_DATA_TYPE_DOUBLE:
+      return "TSDB_DATA_TYPE_DOUBLE";
+    case TSDB_DATA_TYPE_VARCHAR:
+      return "TSDB_DATA_TYPE_VARCHAR";
       //    case TSDB_DATA_TYPE_BINARY:          return "TSDB_DATA_TYPE_VARCHAR";
-    case TSDB_DATA_TYPE_TIMESTAMP:       return "TSDB_DATA_TYPE_TIMESTAMP";
+    case TSDB_DATA_TYPE_TIMESTAMP:
-    case TSDB_DATA_TYPE_NCHAR:           return "TSDB_DATA_TYPE_NCHAR";
+      return "TSDB_DATA_TYPE_TIMESTAMP";
-    case TSDB_DATA_TYPE_JSON:            return "TSDB_DATA_TYPE_JSON";
+    case TSDB_DATA_TYPE_NCHAR:
-    default: return "UNKNOWN";
+      return "TSDB_DATA_TYPE_NCHAR";
+    case TSDB_DATA_TYPE_JSON:
+      return "TSDB_DATA_TYPE_JSON";
+    default:
+      return "UNKNOWN";
   }
 }
 
@@ -329,16 +364,19 @@ bool taos_is_null(TAOS_RES *res, int32_t row, int32_t col) {
   return colDataIsNull_f(pCol->nullbitmap, row);
 }
 
+bool taos_is_update_query(TAOS_RES *res) {
+  // TODO
+  return true;
+}
+
 int taos_fetch_block(TAOS_RES *res, TAOS_ROW *rows) {
   if (res == NULL) {
     return 0;
   }
 
   SRequestObj *pRequest = (SRequestObj *)res;
-  if (pRequest->type == TSDB_SQL_RETRIEVE_EMPTY_RESULT ||
+  if (pRequest->type == TSDB_SQL_RETRIEVE_EMPTY_RESULT || pRequest->type == TSDB_SQL_INSERT ||
-      pRequest->type == TSDB_SQL_INSERT ||
+      pRequest->code != TSDB_CODE_SUCCESS || taos_num_fields(res) == 0) {
-      pRequest->code != TSDB_CODE_SUCCESS ||
-      taos_num_fields(res) == 0) {
     return 0;
   }
 
@@ -352,8 +390,11 @@ int  taos_fetch_block(TAOS_RES *res, TAOS_ROW *rows) {
   return pResultInfo->numOfRows;
 }
 
-int  taos_validate_sql(TAOS *taos, const char *sql) {
+int taos_validate_sql(TAOS *taos, const char *sql) { return true; }
-  return true;
+
+void taos_reset_current_db(TAOS *taos) {
+  // TODO
+  return;
 }
 
 const char *taos_get_server_info(TAOS *taos) {
@@ -387,6 +428,11 @@ void taos_unsubscribe(TAOS_SUB *tsub, int keepProgress) {
     // TODO
 }
 
+int taos_load_table_info(TAOS *taos, const char *tableNameList) {
+  // TODO
+  return -1;
+}
+
 TAOS_STMT *taos_stmt_init(TAOS *taos) {
   // TODO
   return NULL;
@@ -437,11 +483,26 @@ int taos_stmt_set_tbname(TAOS_STMT* stmt, const char* name) {
   return -1;
 }
 
+int taos_stmt_is_insert(TAOS_STMT *stmt, int *insert) {
+    // TODO
+    return -1;
+}
+
+int taos_stmt_num_params(TAOS_STMT *stmt, int *nums) {
+    // TODO
+    return -1;
+}
+
 int taos_stmt_add_batch(TAOS_STMT* stmt) {
     // TODO
     return -1;
 }
 
+TAOS_RES *taos_stmt_use_result(TAOS_STMT *stmt) {
+    // TODO
+    return NULL;
+}
+
 int taos_stmt_bind_param_batch(TAOS_STMT* stmt, TAOS_MULTI_BIND* bind) {
     // TODO
     return -1;

source/client/src/tmq.c

@@ -26,6 +26,18 @@
 #include "tqueue.h"
 #include "tref.h"
 
+typedef struct {
+  int32_t curBlock;
+  int32_t curRow;
+  void**  uData;
+} SMqRowIter;
+
+struct tmq_message_t {
+  SMqPollRsp msg;
+  void*      vg;
+  SMqRowIter iter;
+};
+
 struct tmq_list_t {
   SArray container;
 };
@@ -103,9 +115,10 @@ typedef struct {
   char*       sql;
   char*       topicName;
   int64_t     topicId;
-  int32_t        nextVgIdx;
   SArray*     vgs;  // SArray<SMqClientVg>
-  SSchemaWrapper schema;
+  int8_t      isSchemaAdaptive;
+  int32_t     numOfFields;
+  TAOS_FIELD* fields;
 } SMqClientTopic;
 
 typedef struct {
@@ -131,10 +144,10 @@ typedef struct {
 
 typedef struct {
   tmq_t*         tmq;
-  /*SMqClientVg* pVg;*/
   int32_t        async;
   tsem_t         rspSem;
   tmq_resp_err_t rspErr;
+  /*SMqClientVg* pVg;*/
 } SMqCommitCbParam;
 
 tmq_conf_t* tmq_conf_new() {
@@ -228,6 +241,10 @@ void tmq_list_destroy(tmq_list_t* list) {
   taosArrayDestroyEx(container, (void (*)(void*))taosMemoryFree);
 }
 
+static int32_t tmqMakeTopicVgKey(char* dst, const char* topicName, int32_t vg) {
+  return sprintf(dst, "%s:%d", topicName, vg);
+}
+
 void tmqClearUnhandleMsg(tmq_t* tmq) {
   tmq_message_t* msg = NULL;
   while (1) {
@@ -471,7 +488,12 @@ tmq_resp_err_t tmq_subscribe(tmq_t* tmq, tmq_list_t* topic_list) {
     tNameExtractFullName(&name, topicFname);
     tscDebug("subscribe topic: %s", topicFname);
     SMqClientTopic topic = {
-        .nextVgIdx = 0, .sql = NULL, .sqlLen = 0, .topicId = 0, .topicName = topicFname, .vgs = NULL};
+        .sql = NULL,
+        .sqlLen = 0,
+        .topicId = 0,
+        .topicName = topicFname,
+        .vgs = NULL,
+    };
     topic.vgs = taosArrayInit(0, sizeof(SMqClientVg));
     taosArrayPush(tmq->clientTopics, &topic);
     taosArrayPush(req.topicNames, &topicFname);
@@ -615,6 +637,7 @@ _return:
   return pRequest;
 }
 
+#if 0
 TAOS_RES* tmq_create_topic(TAOS* taos, const char* topicName, const char* sql, int sqlLen) {
   STscObj*     pTscObj = (STscObj*)taos;
   SRequestObj* pRequest = NULL;
@@ -700,6 +723,7 @@ _return:
 
   return pRequest;
 }
+#endif
 
 static char* formatTimestamp(char* buf, int64_t val, int precision) {
   time_t  tt;
@@ -807,7 +831,7 @@ int32_t tmqPollCb(void* param, const SDataBuf* pMsg, int32_t code) {
   SMqClientVg*    pVg = pParam->pVg;
   tmq_t*          tmq = pParam->tmq;
   if (code != 0) {
-    printf("msg discard, code:%x\n", code);
+    tscWarn("msg discard, code:%x", code);
     goto WRITE_QUEUE_FAIL;
   }
 
@@ -815,12 +839,12 @@ int32_t tmqPollCb(void* param, const SDataBuf* pMsg, int32_t code) {
   int32_t tmqEpoch = atomic_load_32(&tmq->epoch);
   if (msgEpoch < tmqEpoch) {
     tsem_post(&tmq->rspSem);
-    printf("discard rsp epoch %d, current epoch %d\n", msgEpoch, tmqEpoch);
+    tscWarn("discard rsp epoch %d, current epoch %d", msgEpoch, tmqEpoch);
     return 0;
   }
 
   if (msgEpoch != tmqEpoch) {
-    printf("mismatch rsp epoch %d, current epoch %d\n", msgEpoch, tmqEpoch);
+    tscWarn("mismatch rsp epoch %d, current epoch %d", msgEpoch, tmqEpoch);
   } else {
     atomic_sub_fetch_32(&tmq->waitingRequest, 1);
   }
@@ -879,19 +903,54 @@ WRITE_QUEUE_FAIL:
 bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, SMqCMGetSubEpRsp* pRsp) {
   /*printf("call update ep %d\n", epoch);*/
   bool    set = false;
-  int32_t sz = taosArrayGetSize(pRsp->topics);
+  int32_t topicNumGet = taosArrayGetSize(pRsp->topics);
-  SArray* newTopics = taosArrayInit(sz, sizeof(SMqClientTopic));
+  char    vgKey[TSDB_TOPIC_FNAME_LEN + 22];
-  for (int32_t i = 0; i < sz; i++) {
+  SArray* newTopics = taosArrayInit(topicNumGet, sizeof(SMqClientTopic));
+  if (newTopics == NULL) {
+    return false;
+  }
+  SHashObj* pHash = taosHashInit(64, MurmurHash3_32, false, HASH_NO_LOCK);
+  if (pHash == NULL) {
+    taosArrayDestroy(newTopics);
+    return false;
+  }
+
+  // find topic, build hash
+  for (int32_t i = 0; i < topicNumGet; i++) {
     SMqClientTopic topic = {0};
     SMqSubTopicEp* pTopicEp = taosArrayGet(pRsp->topics, i);
+    taosHashClear(pHash);
     topic.topicName = strdup(pTopicEp->topic);
-    int32_t vgSz = taosArrayGetSize(pTopicEp->vgs);
+
-    topic.vgs = taosArrayInit(vgSz, sizeof(SMqClientVg));
+    int32_t topicNumCur = taosArrayGetSize(tmq->clientTopics);
-    for (int32_t j = 0; j < vgSz; j++) {
+    for (int32_t j = 0; j < topicNumCur; j++) {
+      // find old topic
+      SMqClientTopic* pTopicCur = taosArrayGet(tmq->clientTopics, j);
+      if (pTopicCur->vgs && strcmp(pTopicCur->topicName, pTopicEp->topic) == 0) {
+        int32_t vgNumCur = taosArrayGetSize(pTopicCur->vgs);
+        if (vgNumCur == 0) break;
+        for (int32_t k = 0; k < vgNumCur; k++) {
+          SMqClientVg* pVgCur = taosArrayGet(pTopicCur->vgs, k);
+          sprintf(vgKey, "%s:%d", topic.topicName, pVgCur->vgId);
+          taosHashPut(pHash, vgKey, strlen(vgKey), &pVgCur->currentOffset, sizeof(int64_t));
+        }
+        break;
+      }
+    }
+
+    int32_t vgNumGet = taosArrayGetSize(pTopicEp->vgs);
+    topic.vgs = taosArrayInit(vgNumGet, sizeof(SMqClientVg));
+    for (int32_t j = 0; j < vgNumGet; j++) {
       SMqSubVgEp* pVgEp = taosArrayGet(pTopicEp->vgs, j);
+      sprintf(vgKey, "%s:%d", topic.topicName, pVgEp->vgId);
+      int64_t* pOffset = taosHashGet(pHash, vgKey, strlen(vgKey));
+      int64_t  offset = pVgEp->offset;
+      if (pOffset != NULL) {
+        offset = *pOffset;
+      }
       SMqClientVg clientVg = {
           .pollCnt = 0,
-          .currentOffset = pVgEp->offset,
+          .currentOffset = offset,
           .vgId = pVgEp->vgId,
           .epSet = pVgEp->epSet,
           .vgStatus = TMQ_VG_STATUS__IDLE,
@@ -902,6 +961,7 @@ bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, SMqCMGetSubEpRsp* pRsp) {
     taosArrayPush(newTopics, &topic);
   }
   if (tmq->clientTopics) taosArrayDestroy(tmq->clientTopics);
+  taosHashCleanup(pHash);
   tmq->clientTopics = newTopics;
   atomic_store_32(&tmq->epoch, epoch);
   return set;
@@ -911,7 +971,7 @@ int32_t tmqAskEpCb(void* param, const SDataBuf* pMsg, int32_t code) {
   SMqAskEpCbParam* pParam = (SMqAskEpCbParam*)param;
   tmq_t*           tmq = pParam->tmq;
   if (code != 0) {
-    printf("get topic endpoint error, not ready, wait:%d\n", pParam->sync);
+    tscError("get topic endpoint error, not ready, wait:%d\n", pParam->sync);
     goto END;
   }
 
@@ -1282,6 +1342,7 @@ tmq_message_t* tmq_consumer_poll(tmq_t* tmq, int64_t blocking_time) {
 
   while (1) {
     /*printf("cycle\n");*/
+    tmqAskEp(tmq, false);
     tmqPollImpl(tmq, blocking_time);
 
     tsem_wait(&tmq->rspSem);

source/client/test/tmqTest.cpp

@@ -55,8 +55,9 @@ TEST(testCase, create_topic_ctb_Test) {
 
   taos_free_result(pRes);
 
-  char* sql = "select * from tu";
+  // char* sql = "select * from tu";
-  pRes = tmq_create_topic(pConn, "test_ctb_topic_1", sql, strlen(sql));
+  // pRes = tmq_create_topic(pConn, "test_ctb_topic_1", sql, strlen(sql));
+  pRes = taos_query(pConn, "create test_ctb_topic_1 as select * from tu");
   taos_free_result(pRes);
   taos_close(pConn);
 }
@@ -79,8 +80,9 @@ TEST(testCase, create_topic_stb_Test) {
 
   taos_free_result(pRes);
 
-  char* sql = "select * from st1";
+  // char* sql = "select * from st1";
-  pRes = tmq_create_topic(pConn, "test_stb_topic_1", sql, strlen(sql));
+  // pRes = tmq_create_topic(pConn, "test_stb_topic_1", sql, strlen(sql));
+  pRes = taos_query(pConn, "create test_ctb_topic_1 as select * from st1");
   taos_free_result(pRes);
   taos_close(pConn);
 }

source/common/src/tdatablock.c

@@ -331,7 +331,6 @@ int32_t blockDataUpdateTsWindow(SSDataBlock* pDataBlock) {
     return 0;
   }
 
-  ASSERT(pColInfoData->nullbitmap == NULL);
   pDataBlock->info.window.skey = *(TSKEY*)colDataGetData(pColInfoData, 0);
   pDataBlock->info.window.ekey = *(TSKEY*)colDataGetData(pColInfoData, (pDataBlock->info.rows - 1));
   return 0;
@@ -609,22 +608,6 @@ size_t blockDataGetSerialMetaSize(const SSDataBlock* pBlock) {
   return sizeof(int32_t) + pBlock->info.numOfCols * sizeof(int32_t);
 }
 
-SSchema* blockDataExtractSchema(const SSDataBlock* pBlock, int32_t* numOfCols) {
-  SSchema* pSchema = taosMemoryCalloc(pBlock->info.numOfCols, sizeof(SSchema));
-  for (int32_t i = 0; i < pBlock->info.numOfCols; ++i) {
-    SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, i);
-    pSchema[i].bytes = pColInfoData->info.bytes;
-    pSchema[i].type = pColInfoData->info.type;
-    pSchema[i].colId = pColInfoData->info.colId;
-  }
-
-  if (numOfCols != NULL) {
-    *numOfCols = pBlock->info.numOfCols;
-  }
-
-  return pSchema;
-}
-
 double blockDataGetSerialRowSize(const SSDataBlock* pBlock) {
   ASSERT(pBlock != NULL);
   double rowSize = 0;

source/common/src/tglobal.c

@@ -45,7 +45,7 @@ float   tsRatioOfQueryCores = 1.0f;
 int32_t tsMaxBinaryDisplayWidth = 30;
 bool    tsEnableSlaveQuery = 1;
 bool    tsPrintAuth = 0;
-int32_t tsMultiProcess = 0;
+bool    tsMultiProcess = 0;
 
 // monitor
 bool     tsEnableMonitor = 1;
@@ -347,7 +347,7 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddBool(pCfg, "printAuth", tsPrintAuth, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "slaveQuery", tsEnableSlaveQuery, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "deadLockKillQuery", tsDeadLockKillQuery, 0) != 0) return -1;
-  if (cfgAddInt32(pCfg, "multiProcess", tsMultiProcess, 0, 2, 0) != 0) return -1;
+  if (cfgAddBool(pCfg, "multiProcess", tsMultiProcess, 0) != 0) return -1;
 
   if (cfgAddBool(pCfg, "monitor", tsEnableMonitor, 0) != 0) return -1;
   if (cfgAddInt32(pCfg, "monitorInterval", tsMonitorInterval, 1, 360000, 0) != 0) return -1;
@@ -466,7 +466,7 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsPrintAuth = cfgGetItem(pCfg, "printAuth")->bval;
   tsEnableSlaveQuery = cfgGetItem(pCfg, "slaveQuery")->bval;
   tsDeadLockKillQuery = cfgGetItem(pCfg, "deadLockKillQuery")->bval;
-  tsMultiProcess = cfgGetItem(pCfg, "multiProcess")->i32;
+  tsMultiProcess = cfgGetItem(pCfg, "multiProcess")->bval;
 
   tsEnableMonitor = cfgGetItem(pCfg, "monitor")->bval;
   tsMonitorInterval = cfgGetItem(pCfg, "monitorInterval")->i32;

source/common/src/tmsgcb.c

@@ -16,6 +16,10 @@
 #define _DEFAULT_SOURCE
 #include "tmsgcb.h"
 
+static SMsgCb tsDefaultMsgCb;
+
+void tmsgSetDefaultMsgCb(const SMsgCb* pMsgCb) { tsDefaultMsgCb = *pMsgCb; }
+
 int32_t tmsgPutToQueue(const SMsgCb* pMsgCb, EQueueType qtype, SRpcMsg* pReq) {
   return (*pMsgCb->queueFps[qtype])(pMsgCb->pWrapper, pReq);
 }
@@ -32,12 +36,12 @@ int32_t tmsgSendMnodeReq(const SMsgCb* pMsgCb, SRpcMsg* pReq) {
   return (*pMsgCb->sendMnodeReqFp)(pMsgCb->pWrapper, pReq);
 }
 
-void tmsgSendRsp(const SMsgCb* pMsgCb, const SRpcMsg* pRsp) { return (*pMsgCb->sendRspFp)(pMsgCb->pWrapper, pRsp); }
+void tmsgSendRsp(const SRpcMsg* pRsp) { return (*tsDefaultMsgCb.sendRspFp)(tsDefaultMsgCb.pWrapper, pRsp); }
 
 void tmsgRegisterBrokenLinkArg(const SMsgCb* pMsgCb, SRpcMsg* pMsg) {
   (*pMsgCb->registerBrokenLinkArgFp)(pMsgCb->pWrapper, pMsg);
 }
 
-void tmsgReleaseHandle(const SMsgCb* pMsgCb, void* handle, int8_t type) {
+void tmsgReleaseHandle(void* handle, int8_t type) {
-  (*pMsgCb->releaseHandleFp)(pMsgCb->pWrapper, handle, type);
+  (*tsDefaultMsgCb.releaseHandleFp)(tsDefaultMsgCb.pWrapper, handle, type);
 }

source/dnode/mgmt/CMakeLists.txt

@@ -1,12 +1,10 @@
-aux_source_directory(src DNODE_SRC)
+aux_source_directory(dm/src DNODE_SRC)
-aux_source_directory(dnode/src DNODE_SRC)
+aux_source_directory(qm/src DNODE_SRC)
-aux_source_directory(qnode/src DNODE_SRC)
+aux_source_directory(bm/src DNODE_SRC)
-aux_source_directory(bnode/src DNODE_SRC)
+aux_source_directory(sm/src DNODE_SRC)
-aux_source_directory(snode/src DNODE_SRC)
+aux_source_directory(vm/src DNODE_SRC)
-aux_source_directory(vnode/src DNODE_SRC)
+aux_source_directory(mm/src DNODE_SRC)
-aux_source_directory(mnode/src DNODE_SRC)
+aux_source_directory(main/src DNODE_SRC)
-aux_source_directory(container/src DNODE_SRC)
-
 add_library(dnode STATIC ${DNODE_SRC})
 target_link_libraries(
     dnode cjson mnode vnode qnode snode bnode wal sync taos tfs monitor
@@ -14,16 +12,22 @@ target_link_libraries(
 target_include_directories(
     dnode
     PUBLIC "${CMAKE_SOURCE_DIR}/include/dnode/mgmt"
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/dnode/inc"
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/dm/inc"
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/qnode/inc"
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/qm/inc"
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/bnode/inc"
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/bm/inc"
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/snode/inc"
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/sm/inc"
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/vnode/inc"
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/vm/inc"
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/mnode/inc"
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/mm/inc"
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/container/inc"
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/main/inc"
 )
 
-add_subdirectory(main)
+aux_source_directory(main/exe EXEC_SRC)
+add_executable(taosd ${EXEC_SRC})
+target_include_directories(
+    taosd
+    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/main/inc"
+)
+target_link_libraries(taosd dnode)
 
 if(${BUILD_TEST})
     add_subdirectory(test)

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

@@ -1,267 +0,0 @@
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "dndInt.h"
-
-static void dndResetLog(SMgmtWrapper *pMgmt) {
-  char logname[24] = {0};
-  snprintf(logname, sizeof(logname), "%slog", pMgmt->name);
-
-  dInfo("node:%s, reset log to %s in child process", pMgmt->name, logname);
-  taosCloseLog();
-  taosInitLog(logname, 1);
-}
-
-static bool dndRequireNode(SMgmtWrapper *pWrapper) {
-  bool required = false;
-  int32_t code =(*pWrapper->fp.requiredFp)(pWrapper, &required);
-  if (!required) {
-    dDebug("node:%s, no need to start", pWrapper->name);
-  } else {
-    dDebug("node:%s, need to start", pWrapper->name);
-  }
-  return required;
-}
-
-int32_t dndOpenNode(SMgmtWrapper *pWrapper) {
-  int32_t code = (*pWrapper->fp.openFp)(pWrapper);
-  if (code != 0) {
-    dError("node:%s, failed to open since %s", pWrapper->name, terrstr());
-    return -1;
-  } else {
-    dDebug("node:%s, has been opened", pWrapper->name);
-  }
-
-  pWrapper->deployed = true;
-  return 0;
-}
-
-void dndCloseNode(SMgmtWrapper *pWrapper) {
-  dDebug("node:%s, start to close", pWrapper->name);
-  pWrapper->required = false;
-  taosWLockLatch(&pWrapper->latch);
-  if (pWrapper->deployed) {
-    (*pWrapper->fp.closeFp)(pWrapper);
-    pWrapper->deployed = false;
-  }
-  taosWUnLockLatch(&pWrapper->latch);
-
-  while (pWrapper->refCount > 0) {
-    taosMsleep(10);
-  }
-
-  if (pWrapper->pProc) {
-    taosProcCleanup(pWrapper->pProc);
-    pWrapper->pProc = NULL;
-  }
-  dDebug("node:%s, has been closed", pWrapper->name);
-}
-
-static int32_t dndRunInSingleProcess(SDnode *pDnode) {
-  dInfo("dnode run in single process mode");
-
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
-    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
-    pWrapper->required = dndRequireNode(pWrapper);
-    if (!pWrapper->required) continue;
-
-    if (taosMkDir(pWrapper->path) != 0) {
-      terrno = TAOS_SYSTEM_ERROR(errno);
-      dError("failed to create dir:%s since %s", pWrapper->path, terrstr());
-      return -1;
-    }
-
-    dInfo("node:%s, will start in single process", pWrapper->name);
-    pWrapper->procType = PROC_SINGLE;
-    if (dndOpenNode(pWrapper) != 0) {
-      dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
-      return -1;
-    }
-  }
-
-  dndSetStatus(pDnode, DND_STAT_RUNNING);
-
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
-    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
-    if (!pWrapper->required) continue;
-    if (pWrapper->fp.startFp == NULL) continue;
-    if ((*pWrapper->fp.startFp)(pWrapper) != 0) {
-      dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
-      return -1;
-    }
-  }
-
-  return 0;
-}
-
-static void dndClearNodesExecpt(SDnode *pDnode, ENodeType except) {
-  // dndCleanupServer(pDnode);
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
-    if (except == n) continue;
-    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
-    pWrapper->required = false;
-  }
-}
-
-static void dndConsumeChildQueue(SMgmtWrapper *pWrapper, SNodeMsg *pMsg, int16_t msgLen, void *pCont, int32_t contLen,
-                                 ProcFuncType ftype) {
-  SRpcMsg *pRpc = &pMsg->rpcMsg;
-  pRpc->pCont = pCont;
-  dTrace("msg:%p, get from child queue, type:%s handle:%p app:%p", pMsg, TMSG_INFO(pRpc->msgType), pRpc->handle,
-         pRpc->ahandle);
-
-  NodeMsgFp msgFp = pWrapper->msgFps[TMSG_INDEX(pRpc->msgType)];
-  int32_t   code = (*msgFp)(pWrapper, pMsg);
-
-  if (code != 0) {
-    dError("msg:%p, failed to process since code:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
-    if (pRpc->msgType & 1U) {
-      SRpcMsg rsp = {.handle = pRpc->handle, .ahandle = pRpc->ahandle, .code = terrno};
-      dndSendRsp(pWrapper, &rsp);
-    }
-
-    dTrace("msg:%p, is freed", pMsg);
-    taosFreeQitem(pMsg);
-    rpcFreeCont(pCont);
-  }
-}
-
-static void dndConsumeParentQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg, int16_t msgLen, void *pCont, int32_t contLen,
-                                  ProcFuncType ftype) {
-  pMsg->pCont = pCont;
-  dTrace("msg:%p, get from parent queue, type:%s handle:%p app:%p", pMsg, TMSG_INFO(pMsg->msgType), pMsg->handle,
-         pMsg->ahandle);
-
-  switch (ftype) {
-    case PROC_REGISTER:
-      rpcRegisterBrokenLinkArg(pMsg);
-      break;
-    default:
-      dndSendRpcRsp(pWrapper, pMsg);
-      break;
-  }
-  taosMemoryFree(pMsg);
-}
-
-static int32_t dndRunInMultiProcess(SDnode *pDnode) {
-  dInfo("dnode run in multi process mode");
-
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
-    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
-    pWrapper->required = dndRequireNode(pWrapper);
-    if (!pWrapper->required) continue;
-
-    if (taosMkDir(pWrapper->path) != 0) {
-      terrno = TAOS_SYSTEM_ERROR(errno);
-      dError("failed to create dir:%s since %s", pWrapper->path, terrstr());
-      return -1;
-    }
-
-    if (n == DNODE) {
-      dInfo("node:%s, will start in parent process", pWrapper->name);
-      pWrapper->procType = PROC_SINGLE;
-      if (dndOpenNode(pWrapper) != 0) {
-        dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
-        return -1;
-      }
-      continue;
-    }
-
-    SProcCfg  cfg = {.childQueueSize = 1024 * 1024 * 2,  // size will be a configuration item
-                     .childConsumeFp = (ProcConsumeFp)dndConsumeChildQueue,
-                     .childMallocHeadFp = (ProcMallocFp)taosAllocateQitem,
-                     .childFreeHeadFp = (ProcFreeFp)taosFreeQitem,
-                     .childMallocBodyFp = (ProcMallocFp)rpcMallocCont,
-                     .childFreeBodyFp = (ProcFreeFp)rpcFreeCont,
-                     .parentQueueSize = 1024 * 1024 * 2,  // size will be a configuration item
-                     .parentConsumeFp = (ProcConsumeFp)dndConsumeParentQueue,
-                     .parentdMallocHeadFp = (ProcMallocFp)taosMemoryMalloc,
-                     .parentFreeHeadFp = (ProcFreeFp)taosMemoryFree,
-                     .parentMallocBodyFp = (ProcMallocFp)rpcMallocCont,
-                     .parentFreeBodyFp = (ProcFreeFp)rpcFreeCont,
-                     .pParent = pWrapper,
-                     .name = pWrapper->name};
-    SProcObj *pProc = taosProcInit(&cfg);
-    if (pProc == NULL) {
-      dError("node:%s, failed to fork since %s", pWrapper->name, terrstr());
-      return -1;
-    }
-
-    pWrapper->pProc = pProc;
-
-    if (taosProcIsChild(pProc)) {
-      dInfo("node:%s, will start in child process", pWrapper->name);
-      pWrapper->procType = PROC_CHILD;
-      dndResetLog(pWrapper);
-
-      dInfo("node:%s, clean up resources inherited from parent", pWrapper->name);
-      dndClearNodesExecpt(pDnode, n);
-
-      dInfo("node:%s, will be initialized in child process", pWrapper->name);
-      dndOpenNode(pWrapper);
-    } else {
-      dInfo("node:%s, will not start in parent process, child pid:%d", pWrapper->name, taosProcChildId(pProc));
-      pWrapper->procType = PROC_PARENT;
-    }
-
-    if (taosProcRun(pProc) != 0) {
-      dError("node:%s, failed to run proc since %s", pWrapper->name, terrstr());
-      return -1;
-    }
-  }
-
-  dndSetStatus(pDnode, DND_STAT_RUNNING);
-
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
-    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
-    if (!pWrapper->required) continue;
-    if (pWrapper->fp.startFp == NULL) continue;
-    if (pWrapper->procType == PROC_PARENT && n != DNODE) continue;
-    if (pWrapper->procType == PROC_CHILD && n == DNODE) continue;
-    if ((*pWrapper->fp.startFp)(pWrapper) != 0) {
-      dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
-      return -1;
-    }
-  }
-
-  return 0;
-}
-
-int32_t dndRun(SDnode *pDnode) {
-  if (tsMultiProcess == 0) {
-    if (dndRunInSingleProcess(pDnode) != 0) {
-      dError("failed to run dnode in single process mode since %s", terrstr());
-      return -1;
-    }
-  } else {
-    if (dndRunInMultiProcess(pDnode) != 0) {
-      dError("failed to run dnode in multi process mode since %s", terrstr());
-      return -1;
-    }
-  }
-
-  dndReportStartup(pDnode, "TDengine", "initialized successfully");
-
-  while (1) {
-    if (pDnode->event == DND_EVENT_STOP) {
-      dInfo("dnode is about to stop");
-      break;
-    }
-    taosMsleep(100);
-  }
-
-  return 0;
-}

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

@@ -1,103 +0,0 @@
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http:www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "dndInt.h"
-
-#define MAXLEN 1024
-
-int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed) {
-  int32_t       code = TSDB_CODE_NODE_PARSE_FILE_ERROR;
-  int32_t       len = 0;
-  const int32_t maxLen = MAXLEN;
-  char          content[MAXLEN + 1] = {0};
-  cJSON        *root = NULL;
-  char          file[PATH_MAX];
-  TdFilePtr     pFile = NULL;
-
-  snprintf(file, sizeof(file), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
-  pFile = taosOpenFile(file, TD_FILE_READ);
-  if (pFile == NULL) {
-    dDebug("file %s not exist", file);
-    code = 0;
-    goto _OVER;
-  }
-
-  len = (int32_t)taosReadFile(pFile, content, maxLen);
-  if (len <= 0) {
-    dError("failed to read %s since content is null", file);
-    goto _OVER;
-  }
-
-  content[len] = 0;
-  root = cJSON_Parse(content);
-  if (root == NULL) {
-    dError("failed to read %s since invalid json format", file);
-    goto _OVER;
-  }
-
-  cJSON *deployed = cJSON_GetObjectItem(root, "deployed");
-  if (!deployed || deployed->type != cJSON_Number) {
-    dError("failed to read %s since deployed not found", file);
-    goto _OVER;
-  }
-  *pDeployed = deployed->valueint != 0;
-
-  code = 0;
-  dDebug("succcessed to read file %s, deployed:%d", file, *pDeployed);
-
-_OVER:
-  if (root != NULL) cJSON_Delete(root);
-  if (pFile != NULL) taosCloseFile(&pFile);
-
-  terrno = code;
-  return code;
-}
-
-int32_t dndWriteFile(SMgmtWrapper *pWrapper, bool deployed) {
-  char file[PATH_MAX] = {0};
-  snprintf(file, sizeof(file), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
-
-  TdFilePtr pFile = taosOpenFile(file, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  if (pFile == NULL) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to write %s since %s", file, terrstr());
-    return -1;
-  }
-
-  int32_t       len = 0;
-  const int32_t maxLen = MAXLEN;
-  char          content[MAXLEN + 1] = {0};
-
-  len += snprintf(content + len, maxLen - len, "{\n");
-  len += snprintf(content + len, maxLen - len, "  \"deployed\": %d\n", deployed);
-  len += snprintf(content + len, maxLen - len, "}\n");
-
-  taosWriteFile(pFile, content, len);
-  taosFsyncFile(pFile);
-  taosCloseFile(&pFile);
-
-  char realfile[PATH_MAX] = {0};
-  snprintf(realfile, sizeof(realfile), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
-
-  if (taosRenameFile(file, realfile) != 0) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to rename %s since %s", file, terrstr());
-    return -1;
-  }
-
-  dInfo("successed to write %s, deployed:%d", realfile, deployed);
-  return 0;
-}

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

@@ -130,7 +130,7 @@ int32_t dmReadFile(SDnodeMgmt *pMgmt) {
   }
 
   code = 0;
-  dInfo("succcessed to read file %s", file);
+  dDebug("succcessed to read file %s", file);
   dmPrintDnodes(pMgmt);
 
 PRASE_DNODE_OVER:

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

@@ -112,6 +112,16 @@ int32_t dmInit(SMgmtWrapper *pWrapper) {
     return -1;
   }
 
+  if (dndInitServer(pDnode) != 0) {
+    dError("failed to init trans server since %s", terrstr());
+    return -1;
+  }
+
+  if (dndInitClient(pDnode) != 0) {
+    dError("failed to init trans client since %s", terrstr());
+    return -1;
+  }
+
   pWrapper->pMgmt = pMgmt;
   dInfo("dnode-mgmt is initialized");
   return 0;
@@ -122,6 +132,7 @@ void dmCleanup(SMgmtWrapper *pWrapper) {
   if (pMgmt == NULL) return;
 
   dInfo("dnode-mgmt start to clean up");
+  SDnode *pDnode = pMgmt->pDnode;
   dmStopWorker(pMgmt);
 
   taosWLockLatch(&pMgmt->latch);
@@ -140,6 +151,9 @@ void dmCleanup(SMgmtWrapper *pWrapper) {
 
   taosMemoryFree(pMgmt);
   pWrapper->pMgmt = NULL;
+  dndCleanupServer(pDnode);
+  dndCleanupClient(pDnode);
+
   dInfo("dnode-mgmt is cleaned up");
 }
 

source/dnode/mgmt/main/CMakeLists.txt

@@ -1,9 +0,0 @@
-aux_source_directory(src EXEC_SRC)
-add_executable(taosd ${EXEC_SRC})
-
-target_include_directories(
-    taosd
-    PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/inc"
-)
-
-target_link_libraries(taosd dnode)

source/dnode/mgmt/main/exe/dndMain.c

@@ -0,0 +1,218 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define _DEFAULT_SOURCE
+#include "dnd.h"
+#include "tconfig.h"
+
+static struct {
+  bool      dumpConfig;
+  bool      generateGrant;
+  bool      printAuth;
+  bool      printVersion;
+  char      envFile[PATH_MAX];
+  char      apolloUrl[PATH_MAX];
+  SArray   *pArgs;  // SConfigPair
+  SDnode   *pDnode;
+  ENodeType ntype;
+} global = {0};
+
+static void dndStopDnode(int signum, void *info, void *ctx) {
+  dInfo("signal:%d is received", signum);
+  SDnode *pDnode = atomic_val_compare_exchange_ptr(&global.pDnode, 0, global.pDnode);
+  if (pDnode != NULL) {
+    dndHandleEvent(pDnode, DND_EVENT_STOP);
+  }
+}
+
+static void dndHandleChild(int signum, void *info, void *ctx) {
+  dInfo("signal:%d is received", signum);
+  dndHandleEvent(global.pDnode, DND_EVENT_CHILD);
+}
+
+static void dndSetSignalHandle() {
+  taosSetSignal(SIGTERM, dndStopDnode);
+  taosSetSignal(SIGHUP, dndStopDnode);
+  taosSetSignal(SIGINT, dndStopDnode);
+  taosSetSignal(SIGABRT, dndStopDnode);
+  taosSetSignal(SIGBREAK, dndStopDnode);
+
+  if (!tsMultiProcess) {
+    // Set the single process signal
+  } else if (global.ntype == DNODE) {
+    // When the child process exits, the parent process receives a signal
+    taosSetSignal(SIGCHLD, dndHandleChild);
+  } else {
+    // When the parent process exits, the child process will receive the SIGKILL signal
+    taosKillChildOnSelfStopped();
+  }
+}
+
+static int32_t dndParseArgs(int32_t argc, char const *argv[]) {
+  for (int32_t i = 1; i < argc; ++i) {
+    if (strcmp(argv[i], "-c") == 0) {
+      if (i < argc - 1) {
+        if (strlen(argv[++i]) >= PATH_MAX) {
+          printf("config file path overflow");
+          return -1;
+        }
+        tstrncpy(configDir, argv[i], PATH_MAX);
+      } else {
+        printf("'-c' requires a parameter, default is %s\n", configDir);
+        return -1;
+      }
+    } else if (strcmp(argv[i], "-a") == 0) {
+      tstrncpy(global.apolloUrl, argv[++i], PATH_MAX);
+    } else if (strcmp(argv[i], "-e") == 0) {
+      tstrncpy(global.envFile, argv[++i], PATH_MAX);
+    } else if (strcmp(argv[i], "-k") == 0) {
+      global.generateGrant = true;
+    } else if (strcmp(argv[i], "-n") == 0) {
+      global.ntype = atoi(argv[++i]);
+      if (global.ntype <= DNODE || global.ntype > NODE_MAX) {
+        printf("'-n' range is [1-5], default is 0\n");
+        return -1;
+      }
+    } else if (strcmp(argv[i], "-C") == 0) {
+      global.dumpConfig = true;
+    } else if (strcmp(argv[i], "-V") == 0) {
+      global.printVersion = true;
+    } else {
+    }
+  }
+
+  return 0;
+}
+
+static void dndGenerateGrant() {
+  // grantParseParameter();
+  printf("this feature is not implemented yet\n");
+}
+
+static void dndPrintVersion() {
+#ifdef TD_ENTERPRISE
+  char *releaseName = "enterprise";
+#else
+  char *releaseName = "community";
+#endif
+  printf("%s version: %s compatible_version: %s\n", releaseName, version, compatible_version);
+  printf("gitinfo: %s\n", gitinfo);
+  printf("buildInfo: %s\n", buildinfo);
+}
+
+static void dndDumpCfg() {
+  SConfig *pCfg = taosGetCfg();
+  cfgDumpCfg(pCfg, 0, 1);
+}
+
+static SDnodeOpt dndGetOpt() {
+  SConfig  *pCfg = taosGetCfg();
+  SDnodeOpt option = {0};
+
+  option.numOfSupportVnodes = cfgGetItem(pCfg, "supportVnodes")->i32;
+  tstrncpy(option.dataDir, tsDataDir, sizeof(option.dataDir));
+  tstrncpy(option.firstEp, tsFirst, sizeof(option.firstEp));
+  tstrncpy(option.secondEp, tsSecond, sizeof(option.firstEp));
+  option.serverPort = tsServerPort;
+  tstrncpy(option.localFqdn, tsLocalFqdn, sizeof(option.localFqdn));
+  snprintf(option.localEp, sizeof(option.localEp), "%s:%u", option.localFqdn, option.serverPort);
+  option.disks = tsDiskCfg;
+  option.numOfDisks = tsDiskCfgNum;
+  option.ntype = global.ntype;
+  return option;
+}
+
+static int32_t dndInitLog() {
+  char logName[12] = {0};
+  snprintf(logName, sizeof(logName), "%slog", dndNodeLogStr(global.ntype));
+  return taosCreateLog(logName, 1, configDir, global.envFile, global.apolloUrl, global.pArgs, 0);
+}
+
+static void dndSetProcName(char **argv) {
+  if (global.ntype != DNODE) {
+    const char *name = dndNodeProcStr(global.ntype);
+    taosSetProcName(argv, name);
+  }
+}
+
+static int32_t dndRunDnode() {
+  if (dndInit() != 0) {
+    dError("failed to initialize environment since %s", terrstr());
+    return -1;
+  }
+
+  SDnodeOpt option = dndGetOpt();
+  SDnode   *pDnode = dndCreate(&option);
+  if (pDnode == NULL) {
+    dError("failed to to create dnode object since %s", terrstr());
+    return -1;
+  } else {
+    global.pDnode = pDnode;
+    dndSetSignalHandle();
+  }
+
+  dInfo("start the service");
+  int32_t code = dndRun(pDnode);
+  dInfo("start shutting down the service");
+
+  global.pDnode = NULL;
+  dndClose(pDnode);
+  dndCleanup();
+  taosCloseLog();
+  taosCleanupCfg();
+  return code;
+}
+
+int main(int argc, char const *argv[]) {
+  if (!taosCheckSystemIsSmallEnd()) {
+    printf("failed to start since on non-small-end machines\n");
+    return -1;
+  }
+
+  if (dndParseArgs(argc, argv) != 0) {
+    printf("failed to start since parse args error\n");
+    return -1;
+  }
+
+  if (global.generateGrant) {
+    dndGenerateGrant();
+    return 0;
+  }
+
+  if (global.printVersion) {
+    dndPrintVersion();
+    return 0;
+  }
+
+  if (dndInitLog() != 0) {
+    printf("failed to start since init log error\n");
+    return -1;
+  }
+
+  if (taosInitCfg(configDir, global.envFile, global.apolloUrl, global.pArgs, 0) != 0) {
+    dError("failed to start since read config error");
+    return -1;
+  }
+
+  if (global.dumpConfig) {
+    dndDumpCfg();
+    taosCleanupCfg();
+    taosCloseLog();
+    return 0;
+  }
+
+  dndSetProcName((char **)argv);
+  return dndRunDnode();
+}

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

@@ -33,6 +33,7 @@
 #include "tthread.h"
 #include "ttime.h"
 #include "tworker.h"
+#include "tmsgcb.h"
 
 #include "dnode.h"
 #include "monitor.h"
@@ -94,12 +95,14 @@ typedef struct SMgmtWrapper {
   bool        deployed;
   bool        required;
   EProcType   procType;
+  int32_t     procId;
   SProcObj   *pProc;
+  SShm        shm;
   void       *pMgmt;
   SDnode     *pDnode;
-  NodeMsgFp   msgFps[TDMT_MAX];
-  int32_t     msgVgIds[TDMT_MAX];  // Handle the case where the same message type is distributed to qnode or vnode
   SMgmtFp     fp;
+  int8_t      msgVgIds[TDMT_MAX];  // Handle the case where the same message type is distributed to qnode or vnode
+  NodeMsgFp   msgFps[TDMT_MAX];
 } SMgmtWrapper;
 
 typedef struct {
@@ -118,30 +121,38 @@ typedef struct SDnode {
   char        *firstEp;
   char        *secondEp;
   char        *dataDir;
-  SDiskCfg    *pDisks;
+  SDiskCfg    *disks;
   int32_t      numOfDisks;
   uint16_t     serverPort;
   bool         dropped;
+  ENodeType    ntype;
   EDndStatus   status;
   EDndEvent    event;
   SStartupReq  startup;
-  TdFilePtr    pLockFile;
+  TdFilePtr    lockfile;
   STransMgmt   trans;
   SMgmtWrapper wrappers[NODE_MAX];
 } SDnode;
 
+const char *dndNodeLogStr(ENodeType ntype);
+const char *dndNodeProcStr(ENodeType ntype);
 EDndStatus  dndGetStatus(SDnode *pDnode);
 void        dndSetStatus(SDnode *pDnode, EDndStatus stat);
-void       dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId);
+void        dndSetMsgHandle(SMgmtWrapper *pWrapper, tmsg_t msgType, NodeMsgFp nodeMsgFp, int8_t vgId);
 void        dndReportStartup(SDnode *pDnode, const char *pName, const char *pDesc);
 void        dndSendMonitorReport(SDnode *pDnode);
 
+int32_t dndInitServer(SDnode *pDnode);
+void    dndCleanupServer(SDnode *pDnode);
+int32_t dndInitClient(SDnode *pDnode);
+void    dndCleanupClient(SDnode *pDnode);
+int32_t dndProcessNodeMsg(SDnode *pDnode, SNodeMsg *pMsg);
 int32_t dndSendReqToMnode(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
 int32_t dndSendReqToDnode(SMgmtWrapper *pWrapper, const SEpSet *pEpSet, SRpcMsg *pMsg);
 void    dndSendRsp(SMgmtWrapper *pWrapper, const SRpcMsg *pRsp);
 void    dndRegisterBrokenLinkArg(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
+SMsgCb  dndCreateMsgcb(SMgmtWrapper *pWrapper);
 
-int32_t dndProcessNodeMsg(SDnode *pDnode, SNodeMsg *pMsg);
 int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed);
 int32_t dndWriteFile(SMgmtWrapper *pWrapper, bool deployed);
 

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

@@ -34,7 +34,6 @@ int32_t     dndInit();
 void        dndCleanup();
 const char *dndStatStr(EDndStatus stat);
 void        dndGetStartup(SDnode *pDnode, SStartupReq *pStartup);
-TdFilePtr   dndCheckRunning(const char *dataDir);
 void        dndProcessStartupReq(SDnode *pDnode, SRpcMsg *pMsg);
 
 // dndMsg.c
@@ -51,13 +50,14 @@ void    dndClose(SDnode *pDnode);
 void    dndHandleEvent(SDnode *pDnode, EDndEvent event);
 
 // dndTransport.c
-int32_t dndInitServer(SDnode *pDnode);
-void    dndCleanupServer(SDnode *pDnode);
-int32_t dndInitClient(SDnode *pDnode);
-void    dndCleanupClient(SDnode *pDnode);
 int32_t dndInitMsgHandle(SDnode *pDnode);
 void    dndSendRpcRsp(SMgmtWrapper *pWrapper, const SRpcMsg *pRsp);
 
+// dndFile.c
+TdFilePtr dndCheckRunning(const char *dataDir);
+int32_t   dndReadShmFile(SDnode *pDnode);
+int32_t   dndWriteShmFile(SDnode *pDnode);
+
 #ifdef __cplusplus
 }
 #endif

source/dnode/mgmt/main/inc/dndMain.h

@@ -1,48 +0,0 @@
-
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef _TD_DND_MAIN_H_
-#define _TD_DND_MAIN_H_
-
-#include "dnode.h"
-
-#include "taoserror.h"
-#include "tconfig.h"
-#include "tglobal.h"
-#include "tlog.h"
-#include "version.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define dFatal(...) { if (dDebugFlag & DEBUG_FATAL) { taosPrintLog("DND FATAL ", DEBUG_FATAL, 255, __VA_ARGS__); }}
-#define dError(...) { if (dDebugFlag & DEBUG_ERROR) { taosPrintLog("DND ERROR ", DEBUG_ERROR, 255, __VA_ARGS__); }}
-#define dWarn(...)  { if (dDebugFlag & DEBUG_WARN)  { taosPrintLog("DND WARN ", DEBUG_WARN, 255, __VA_ARGS__); }}
-#define dInfo(...)  { if (dDebugFlag & DEBUG_INFO)  { taosPrintLog("DND ", DEBUG_INFO, 255, __VA_ARGS__); }}
-#define dDebug(...) { if (dDebugFlag & DEBUG_DEBUG) { taosPrintLog("DND ", DEBUG_DEBUG, dDebugFlag, __VA_ARGS__); }}
-#define dTrace(...) { if (dDebugFlag & DEBUG_TRACE) { taosPrintLog("DND ", DEBUG_TRACE, dDebugFlag, __VA_ARGS__); }}
-
-void      dndDumpCfg();
-void      dndPrintVersion();
-void      dndGenerateGrant();
-SDnodeOpt dndGetOpt();
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /*_TD_DND_MAIN_H_*/

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

@@ -0,0 +1,306 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define _DEFAULT_SOURCE
+#include "dndInt.h"
+
+static bool dndRequireNode(SMgmtWrapper *pWrapper) {
+  bool required = false;
+  int32_t code =(*pWrapper->fp.requiredFp)(pWrapper, &required);
+  if (!required) {
+    dDebug("node:%s, no need to start", pWrapper->name);
+  } else {
+    dDebug("node:%s, need to start", pWrapper->name);
+  }
+  return required;
+}
+
+int32_t dndOpenNode(SMgmtWrapper *pWrapper) {
+  if (taosMkDir(pWrapper->path) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("node:%s, failed to create dir:%s since %s", pWrapper->name, pWrapper->path, terrstr());
+    return -1;
+  }
+
+  if ((*pWrapper->fp.openFp)(pWrapper) != 0) {
+    dError("node:%s, failed to open since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  dDebug("node:%s, has been opened", pWrapper->name);
+  pWrapper->deployed = true;
+  return 0;
+}
+
+void dndCloseNode(SMgmtWrapper *pWrapper) {
+  dDebug("node:%s, start to close", pWrapper->name);
+  pWrapper->required = false;
+  taosWLockLatch(&pWrapper->latch);
+  if (pWrapper->deployed) {
+    (*pWrapper->fp.closeFp)(pWrapper);
+    pWrapper->deployed = false;
+  }
+  taosWUnLockLatch(&pWrapper->latch);
+
+  while (pWrapper->refCount > 0) {
+    taosMsleep(10);
+  }
+
+  if (pWrapper->pProc) {
+    taosProcCleanup(pWrapper->pProc);
+    pWrapper->pProc = NULL;
+  }
+  dDebug("node:%s, has been closed", pWrapper->name);
+}
+
+static int32_t dndRunInSingleProcess(SDnode *pDnode) {
+  dInfo("dnode start to run in single process");
+
+  for (ENodeType n = DNODE; n < NODE_MAX; ++n) {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
+    pWrapper->required = dndRequireNode(pWrapper);
+    if (!pWrapper->required) continue;
+
+    if (dndOpenNode(pWrapper) != 0) {
+      dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
+      return -1;
+    }
+  }
+
+  dndSetStatus(pDnode, DND_STAT_RUNNING);
+
+  for (ENodeType n = 0; n < NODE_MAX; ++n) {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
+    if (!pWrapper->required) continue;
+    if (pWrapper->fp.startFp == NULL) continue;
+    if ((*pWrapper->fp.startFp)(pWrapper) != 0) {
+      dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
+      return -1;
+    }
+  }
+
+  dInfo("dnode running in single process");
+  return 0;
+}
+
+static void dndConsumeChildQueue(SMgmtWrapper *pWrapper, SNodeMsg *pMsg, int16_t msgLen, void *pCont, int32_t contLen,
+                                 ProcFuncType ftype) {
+  SRpcMsg *pRpc = &pMsg->rpcMsg;
+  pRpc->pCont = pCont;
+  dTrace("msg:%p, get from child queue, handle:%p app:%p", pMsg, pRpc->handle, pRpc->ahandle);
+
+  NodeMsgFp msgFp = pWrapper->msgFps[TMSG_INDEX(pRpc->msgType)];
+  int32_t   code = (*msgFp)(pWrapper, pMsg);
+
+  if (code != 0) {
+    dError("msg:%p, failed to process since code:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
+    if (pRpc->msgType & 1U) {
+      SRpcMsg rsp = {.handle = pRpc->handle, .ahandle = pRpc->ahandle, .code = terrno};
+      dndSendRsp(pWrapper, &rsp);
+    }
+
+    dTrace("msg:%p, is freed", pMsg);
+    taosFreeQitem(pMsg);
+    rpcFreeCont(pCont);
+  }
+}
+
+static void dndConsumeParentQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg, int16_t msgLen, void *pCont, int32_t contLen,
+                                  ProcFuncType ftype) {
+  pMsg->pCont = pCont;
+  dTrace("msg:%p, get from parent queue, handle:%p app:%p", pMsg, pMsg->handle, pMsg->ahandle);
+
+  switch (ftype) {
+    case PROC_REG:
+      rpcRegisterBrokenLinkArg(pMsg);
+      break;
+    case PROC_RELEASE:
+      rpcReleaseHandle(pMsg->handle, (int8_t)pMsg->code);
+      rpcFreeCont(pCont);
+      break;
+    case PROC_REQ:
+      // todo send to dnode
+      dndSendReqToMnode(pWrapper, pMsg);
+    default:
+      dndSendRpcRsp(pWrapper, pMsg);
+      break;
+  }
+  taosMemoryFree(pMsg);
+}
+
+static int32_t dndRunInParentProcess(SDnode *pDnode) {
+  dInfo("dnode start to run in parent process");
+  SMgmtWrapper *pDWrapper = &pDnode->wrappers[DNODE];
+  if (dndOpenNode(pDWrapper) != 0) {
+    dError("node:%s, failed to start since %s", pDWrapper->name, terrstr());
+    return -1;
+  }
+
+  for (ENodeType n = DNODE + 1; n < NODE_MAX; ++n) {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
+    pWrapper->required = dndRequireNode(pWrapper);
+    if (!pWrapper->required) continue;
+
+    int64_t shmsize = 1024 * 1024 * 2;  // size will be a configuration item
+    if (taosCreateShm(&pWrapper->shm, shmsize) != 0) {
+      terrno = TAOS_SYSTEM_ERROR(terrno);
+      dError("node:%s, failed to create shm size:%" PRId64 " since %s", pWrapper->name, shmsize, terrstr());
+      return -1;
+    }
+
+    SProcCfg cfg = {.childConsumeFp = (ProcConsumeFp)dndConsumeChildQueue,
+                    .childMallocHeadFp = (ProcMallocFp)taosAllocateQitem,
+                    .childFreeHeadFp = (ProcFreeFp)taosFreeQitem,
+                    .childMallocBodyFp = (ProcMallocFp)rpcMallocCont,
+                    .childFreeBodyFp = (ProcFreeFp)rpcFreeCont,
+                    .parentConsumeFp = (ProcConsumeFp)dndConsumeParentQueue,
+                    .parentMallocHeadFp = (ProcMallocFp)taosMemoryMalloc,
+                    .parentFreeHeadFp = (ProcFreeFp)taosMemoryFree,
+                    .parentMallocBodyFp = (ProcMallocFp)rpcMallocCont,
+                    .parentFreeBodyFp = (ProcFreeFp)rpcFreeCont,
+                    .shm = pWrapper->shm,
+                    .pParent = pWrapper,
+                    .isChild = false,
+                    .name = pWrapper->name};
+
+    pWrapper->procType = PROC_PARENT;
+    pWrapper->pProc = taosProcInit(&cfg);
+    if (pWrapper->pProc == NULL) {
+      dError("node:%s, failed to create proc since %s", pWrapper->name, terrstr());
+      return -1;
+    }
+  }
+
+  if (dndWriteShmFile(pDnode) != 0) {
+    dError("failed to write runtime file since %s", terrstr());
+    return -1;
+  }
+
+  for (ENodeType n = DNODE + 1; n < NODE_MAX; ++n) {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
+    if (!pWrapper->required) continue;
+
+    if (pDnode->ntype == NODE_MAX) {
+      dInfo("node:%s, should be started manually", pWrapper->name);
+    } else {
+      char    args[PATH_MAX];
+      int32_t pid = taosNewProc(args);
+      if (pid <= 0) {
+        terrno = TAOS_SYSTEM_ERROR(errno);
+        dError("node:%s, failed to exec in new process since %s", pWrapper->name, terrstr());
+        return -1;
+      }
+      pWrapper->procId = pid;
+      dInfo("node:%s, run in new process, pid:%d", pWrapper->name, pid);
+    }
+
+    if (taosProcRun(pWrapper->pProc) != 0) {
+      dError("node:%s, failed to run proc since %s", pWrapper->name, terrstr());
+      return -1;
+    }
+  }
+
+  dndSetStatus(pDnode, DND_STAT_RUNNING);
+
+  if ((*pDWrapper->fp.startFp)(pDWrapper) != 0) {
+    dError("node:%s, failed to start since %s", pDWrapper->name, terrstr());
+    return -1;
+  }
+
+  dInfo("dnode running in parent process");
+  return 0;
+}
+
+static int32_t dndRunInChildProcess(SDnode *pDnode) {
+  dInfo("dnode start to run in child process");
+  SMgmtWrapper *pWrapper = &pDnode->wrappers[pDnode->ntype];
+
+  SMsgCb msgCb = dndCreateMsgcb(pWrapper);
+  tmsgSetDefaultMsgCb(&msgCb);
+  pWrapper->procType = PROC_CHILD;
+
+  if (dndOpenNode(pWrapper) != 0) {
+    dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  SProcCfg cfg = {.childConsumeFp = (ProcConsumeFp)dndConsumeChildQueue,
+                  .childMallocHeadFp = (ProcMallocFp)taosAllocateQitem,
+                  .childFreeHeadFp = (ProcFreeFp)taosFreeQitem,
+                  .childMallocBodyFp = (ProcMallocFp)rpcMallocCont,
+                  .childFreeBodyFp = (ProcFreeFp)rpcFreeCont,
+                  .parentConsumeFp = (ProcConsumeFp)dndConsumeParentQueue,
+                  .parentMallocHeadFp = (ProcMallocFp)taosMemoryMalloc,
+                  .parentFreeHeadFp = (ProcFreeFp)taosMemoryFree,
+                  .parentMallocBodyFp = (ProcMallocFp)rpcMallocCont,
+                  .parentFreeBodyFp = (ProcFreeFp)rpcFreeCont,
+                  .shm = pWrapper->shm,
+                  .pParent = pWrapper,
+                  .isChild = true,
+                  .name = pWrapper->name};
+
+  pWrapper->pProc = taosProcInit(&cfg);
+  if (pWrapper->pProc == NULL) {
+    dError("node:%s, failed to create proc since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  if (pWrapper->fp.startFp != NULL) {
+    if ((*pWrapper->fp.startFp)(pWrapper) != 0) {
+      dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
+      return -1;
+    }
+  }
+
+  if (taosProcRun(pWrapper->pProc) != 0) {
+    dError("node:%s, failed to run proc since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  dInfo("dnode running in child process");
+  return 0;
+}
+
+int32_t dndRun(SDnode * pDnode) {
+  if (!tsMultiProcess) {
+    if (dndRunInSingleProcess(pDnode) != 0) {
+      dError("failed to run dnode since %s", terrstr());
+      return -1;
+    }
+  } else if (pDnode->ntype == DNODE || pDnode->ntype == NODE_MAX) {
+    if (dndRunInParentProcess(pDnode) != 0) {
+      dError("failed to run dnode in parent process since %s", terrstr());
+      return -1;
+    }
+  } else {
+    if (dndRunInChildProcess(pDnode) != 0) {
+      dError("failed to run dnode in child process since %s", terrstr());
+      return -1;
+    }
+  }
+
+  dndReportStartup(pDnode, "TDengine", "initialized successfully");
+  dInfo("TDengine initialized successfully");
+
+  while (1) {
+    if (pDnode->event == DND_EVENT_STOP) {
+      dInfo("dnode is about to stop");
+      break;
+    }
+    taosMsleep(100);
+  }
+
+  return 0;
+}

source/dnode/mgmt/main/src/dndFile.c

@@ -0,0 +1,269 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http:www.gnu.org/licenses/>.
+ */
+
+#define _DEFAULT_SOURCE
+#include "dndInt.h"
+
+#define MAXLEN 1024
+
+int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed) {
+  int32_t   code = TSDB_CODE_NODE_PARSE_FILE_ERROR;
+  int64_t   len = 0;
+  char      content[MAXLEN + 1] = {0};
+  cJSON    *root = NULL;
+  char      file[PATH_MAX];
+  TdFilePtr pFile = NULL;
+
+  snprintf(file, sizeof(file), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
+  pFile = taosOpenFile(file, TD_FILE_READ);
+  if (pFile == NULL) {
+    dDebug("file %s not exist", file);
+    code = 0;
+    goto _OVER;
+  }
+
+  len = taosReadFile(pFile, content, MAXLEN);
+  if (len <= 0) {
+    dError("failed to read %s since content is null", file);
+    goto _OVER;
+  }
+
+  root = cJSON_Parse(content);
+  if (root == NULL) {
+    dError("failed to read %s since invalid json format", file);
+    goto _OVER;
+  }
+
+  cJSON *deployed = cJSON_GetObjectItem(root, "deployed");
+  if (!deployed || deployed->type != cJSON_Number) {
+    dError("failed to read %s since deployed not found", file);
+    goto _OVER;
+  }
+  *pDeployed = deployed->valueint != 0;
+
+  dDebug("succcessed to read file %s, deployed:%d", file, *pDeployed);
+  code = 0;
+
+_OVER:
+  if (root != NULL) cJSON_Delete(root);
+  if (pFile != NULL) taosCloseFile(&pFile);
+
+  terrno = code;
+  return code;
+}
+
+int32_t dndWriteFile(SMgmtWrapper *pWrapper, bool deployed) {
+  int32_t   code = -1;
+  int32_t   len = 0;
+  char      content[MAXLEN + 1] = {0};
+  char      file[PATH_MAX] = {0};
+  char      realfile[PATH_MAX] = {0};
+  TdFilePtr pFile = NULL;
+
+  snprintf(file, sizeof(file), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
+  snprintf(realfile, sizeof(realfile), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
+
+  pFile = taosOpenFile(file, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  if (pFile == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write %s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  len += snprintf(content + len, MAXLEN - len, "{\n");
+  len += snprintf(content + len, MAXLEN - len, "  \"deployed\": %d\n", deployed);
+  len += snprintf(content + len, MAXLEN - len, "}\n");
+
+  if (taosWriteFile(pFile, content, len) != len) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  if (taosFsyncFile(pFile) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to fsync file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  taosCloseFile(&pFile);
+
+  if (taosRenameFile(file, realfile) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to rename %s since %s", file, terrstr());
+    return -1;
+  }
+
+  dInfo("successed to write %s, deployed:%d", realfile, deployed);
+  code = 0;
+
+_OVER:
+  if (pFile != NULL) {
+    taosCloseFile(&pFile);
+  }
+
+  return code;
+}
+
+TdFilePtr dndCheckRunning(const char *dataDir) {
+  char filepath[PATH_MAX] = {0};
+  snprintf(filepath, sizeof(filepath), "%s%s.running", dataDir, TD_DIRSEP);
+
+  TdFilePtr pFile = taosOpenFile(filepath, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  if (pFile == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to lock file:%s since %s", filepath, terrstr());
+    return NULL;
+  }
+
+  int32_t ret = taosLockFile(pFile);
+  if (ret != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to lock file:%s since %s", filepath, terrstr());
+    taosCloseFile(&pFile);
+    return NULL;
+  }
+
+  dDebug("file:%s is locked", filepath);
+  return pFile;
+}
+
+int32_t dndReadShmFile(SDnode *pDnode) {
+  int32_t   code = -1;
+  char      itemName[24] = {0};
+  char      content[MAXLEN + 1] = {0};
+  char      file[PATH_MAX] = {0};
+  cJSON    *root = NULL;
+  TdFilePtr pFile = NULL;
+
+  snprintf(file, sizeof(file), "%s%s.shmfile", pDnode->dataDir, TD_DIRSEP);
+  pFile = taosOpenFile(file, TD_FILE_READ);
+  if (pFile == NULL) {
+    dDebug("file %s not exist", file);
+    code = 0;
+    goto _OVER;
+  }
+
+  if (taosReadFile(pFile, content, MAXLEN) > 0) {
+    root = cJSON_Parse(content);
+    if (root == NULL) {
+      terrno = TSDB_CODE_NODE_PARSE_FILE_ERROR;
+      dError("failed to read %s since invalid json format", file);
+      goto _OVER;
+    }
+
+    for (ENodeType ntype = DNODE + 1; ntype < NODE_MAX; ++ntype) {
+      snprintf(itemName, sizeof(itemName), "%s_shmid", dndNodeProcStr(ntype));
+      cJSON *shmid = cJSON_GetObjectItem(root, itemName);
+      if (shmid && shmid->type == cJSON_Number) {
+        pDnode->wrappers[ntype].shm.id = shmid->valueint;
+      }
+
+      snprintf(itemName, sizeof(itemName), "%s_shmsize", dndNodeProcStr(ntype));
+      cJSON *shmsize = cJSON_GetObjectItem(root, itemName);
+      if (shmsize && shmsize->type == cJSON_Number) {
+        pDnode->wrappers[ntype].shm.size = shmsize->valueint;
+      }
+    }
+  }
+
+  if (!tsMultiProcess || pDnode->ntype == DNODE || pDnode->ntype == DNODE) {
+    for (ENodeType ntype = DNODE; ntype < NODE_MAX; ++ntype) {
+      SMgmtWrapper *pWrapper = &pDnode->wrappers[ntype];
+      if (pWrapper->shm.id >= 0) {
+        dDebug("shmid:%d, is closed, size:%d", pWrapper->shm.id, pWrapper->shm.size);
+        taosDropShm(&pWrapper->shm);
+      }
+    }
+  } else {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[pDnode->ntype];
+    if (taosAttachShm(&pWrapper->shm) != 0) {
+      terrno = TAOS_SYSTEM_ERROR(errno);
+      dError("shmid:%d, failed to attach shm since %s", pWrapper->shm.id, terrstr());
+      goto _OVER;
+    }
+    dDebug("shmid:%d, is attached, size:%d", pWrapper->shm.id, pWrapper->shm.size);
+  }
+
+  dDebug("successed to open %s", file);
+  code = 0;
+
+_OVER:
+  if (root != NULL) cJSON_Delete(root);
+  if (pFile != NULL) taosCloseFile(&pFile);
+
+  return code;
+}
+
+int32_t dndWriteShmFile(SDnode *pDnode) {
+  int32_t   code = -1;
+  int32_t   len = 0;
+  char      content[MAXLEN + 1] = {0};
+  char      file[PATH_MAX] = {0};
+  char      realfile[PATH_MAX] = {0};
+  TdFilePtr pFile = NULL;
+
+  snprintf(file, sizeof(file), "%s%s.shmfile.bak", pDnode->dataDir, TD_DIRSEP);
+  snprintf(realfile, sizeof(realfile), "%s%s.shmfile", pDnode->dataDir, TD_DIRSEP);
+
+  pFile = taosOpenFile(file, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  if (pFile == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to open file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  len += snprintf(content + len, MAXLEN - len, "{\n");
+  for (ENodeType ntype = DNODE + 1; ntype < NODE_MAX; ++ntype) {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[ntype];
+    len += snprintf(content + len, MAXLEN - len, "  \"%s_shmid\":%d,\n", dndNodeProcStr(ntype), pWrapper->shm.id);
+    if (ntype == NODE_MAX - 1) {
+      len += snprintf(content + len, MAXLEN - len, "  \"%s_shmsize\":%d\n", dndNodeProcStr(ntype), pWrapper->shm.size);
+    } else {
+      len += snprintf(content + len, MAXLEN - len, "  \"%s_shmsize\":%d,\n", dndNodeProcStr(ntype), pWrapper->shm.size);
+    }
+  }
+  len += snprintf(content + len, MAXLEN - len, "}\n");
+
+  if (taosWriteFile(pFile, content, len) != len) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  if (taosFsyncFile(pFile) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to fsync file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  taosCloseFile(&pFile);
+
+  if (taosRenameFile(file, realfile) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to rename %s to %s since %s", file, realfile, terrstr());
+    return -1;
+  }
+
+  dInfo("successed to write %s", realfile);
+  code = 0;
+
+_OVER:
+  if (pFile != NULL) {
+    taosCloseFile(&pFile);
+  }
+
+  return code;
+}

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

@@ -20,7 +20,7 @@
 static int8_t once = DND_ENV_INIT;
 
 int32_t dndInit() {
-  dInfo("start to init dnode env");
+  dDebug("start to init dnode env");
   if (atomic_val_compare_exchange_8(&once, DND_ENV_INIT, DND_ENV_READY) != DND_ENV_INIT) {
     terrno = TSDB_CODE_REPEAT_INIT;
     dError("failed to init dnode env since %s", terrstr());
@@ -31,12 +31,6 @@ int32_t dndInit() {
   taosBlockSIGPIPE();
   taosResolveCRC();
 
-  if (rpcInit() != 0) {
-    dError("failed to init rpc since %s", terrstr());
-    dndCleanup();
-    return -1;
-  }
-
   SMonCfg monCfg = {0};
   monCfg.maxLogs = tsMonitorMaxLogs;
   monCfg.port = tsMonitorPort;
@@ -44,7 +38,6 @@ int32_t dndInit() {
   monCfg.comp = tsMonitorComp;
   if (monInit(&monCfg) != 0) {
     dError("failed to init monitor since %s", terrstr());
-    dndCleanup();
     return -1;
   }
 
@@ -53,20 +46,19 @@ int32_t dndInit() {
 }
 
 void dndCleanup() {
-  dInfo("start to cleanup dnode env");
+  dDebug("start to cleanup dnode env");
   if (atomic_val_compare_exchange_8(&once, DND_ENV_READY, DND_ENV_CLEANUP) != DND_ENV_READY) {
     dError("dnode env is already cleaned up");
     return;
   }
 
   monCleanup();
-  rpcCleanup();
   walCleanUp();
   taosStopCacheRefreshWorker();
   dInfo("dnode env is cleaned up");
 }
 
-void dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId) {
+void dndSetMsgHandle(SMgmtWrapper *pWrapper, tmsg_t msgType, NodeMsgFp nodeMsgFp, int8_t vgId) {
   pWrapper->msgFps[TMSG_INDEX(msgType)] = nodeMsgFp;
   pWrapper->msgVgIds[TMSG_INDEX(msgType)] = vgId;
 }
@@ -80,19 +72,6 @@ void dndSetStatus(SDnode *pDnode, EDndStatus status) {
   }
 }
 
-const char *dndStatStr(EDndStatus status) {
-  switch (status) {
-    case DND_STAT_INIT:
-      return "init";
-    case DND_STAT_RUNNING:
-      return "running";
-    case DND_STAT_STOPPED:
-      return "stopped";
-    default:
-      return "unknown";
-  }
-}
-
 void dndReportStartup(SDnode *pDnode, const char *pName, const char *pDesc) {
   SStartupReq *pStartup = &pDnode->startup;
   tstrncpy(pStartup->name, pName, TSDB_STEP_NAME_LEN);
@@ -105,29 +84,6 @@ void dndGetStartup(SDnode *pDnode, SStartupReq *pStartup) {
   pStartup->finished = (dndGetStatus(pDnode) == DND_STAT_RUNNING);
 }
 
-TdFilePtr dndCheckRunning(const char *dataDir) {
-  char filepath[PATH_MAX] = {0};
-  snprintf(filepath, sizeof(filepath), "%s/.running", dataDir);
-
-  TdFilePtr pFile = taosOpenFile(filepath, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  if (pFile == NULL) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to lock file:%s since %s", filepath, terrstr());
-    return NULL;
-  }
-
-  int32_t ret = taosLockFile(pFile);
-  if (ret != 0) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to lock file:%s since %s", filepath, terrstr());
-    taosCloseFile(&pFile);
-    return NULL;
-  }
-
-  dDebug("file:%s is locked", filepath);
-  return pFile;
-}
-
 void dndProcessStartupReq(SDnode *pDnode, SRpcMsg *pReq) {
   dDebug("startup req is received");
   SStartupReq *pStartup = rpcMallocCont(sizeof(SStartupReq));

source/dnode/mgmt/main/src/dndMain.c

@@ -1,138 +0,0 @@
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "dndMain.h"
-
-static struct {
-  bool    dumpConfig;
-  bool    generateGrant;
-  bool    printAuth;
-  bool    printVersion;
-  char    envFile[PATH_MAX];
-  char    apolloUrl[PATH_MAX];
-  SDnode *pDnode;
-} global = {0};
-
-static void dndSigintHandle(int signum, void *info, void *ctx) {
-  dInfo("signal:%d is received", signum);
-  SDnode *pDnode = atomic_val_compare_exchange_ptr(&global.pDnode, 0, global.pDnode);
-  if (pDnode != NULL) {
-    dndHandleEvent(pDnode, DND_EVENT_STOP);
-  }
-}
-
-static void dndSetSignalHandle() {
-  taosSetSignal(SIGTERM, dndSigintHandle);
-  taosSetSignal(SIGHUP, dndSigintHandle);
-  taosSetSignal(SIGINT, dndSigintHandle);
-  taosSetSignal(SIGABRT, dndSigintHandle);
-  taosSetSignal(SIGBREAK, dndSigintHandle);
-}
-
-static int32_t dndParseOption(int32_t argc, char const *argv[]) {
-  for (int32_t i = 1; i < argc; ++i) {
-    if (strcmp(argv[i], "-c") == 0) {
-      if (i < argc - 1) {
-        if (strlen(argv[++i]) >= PATH_MAX) {
-          printf("config file path overflow");
-          return -1;
-        }
-        tstrncpy(configDir, argv[i], PATH_MAX);
-      } else {
-        printf("'-c' requires a parameter, default is %s\n", configDir);
-        return -1;
-      }
-    } else if (strcmp(argv[i], "-C") == 0) {
-      global.dumpConfig = true;
-    } else if (strcmp(argv[i], "-k") == 0) {
-      global.generateGrant = true;
-    } else if (strcmp(argv[i], "-V") == 0) {
-      global.printVersion = true;
-    } else {
-    }
-  }
-
-  return 0;
-}
-
-static int32_t dndRunDnode() {
-  if (dndInit() != 0) {
-    dInfo("failed to initialize dnode environment since %s", terrstr());
-    return -1;
-  }
-
-  SDnodeOpt option = dndGetOpt();
-
-  SDnode *pDnode = dndCreate(&option);
-  if (pDnode == NULL) {
-    dError("failed to to create dnode object since %s", terrstr());
-    return -1;
-  } else {
-    global.pDnode = pDnode;
-    dndSetSignalHandle();
-  }
-
-  dInfo("start the TDengine service");
-  int32_t code = dndRun(pDnode);
-  dInfo("start shutting down the TDengine service");
-
-  global.pDnode = NULL;
-  dndClose(pDnode);
-  dndCleanup();
-  taosCloseLog();
-  taosCleanupCfg();
-  return code;
-}
-
-int main(int argc, char const *argv[]) {
-  if (!taosCheckSystemIsSmallEnd()) {
-    dError("failed to start TDengine since on non-small-end machines");
-    return -1;
-  }
-
-  if (dndParseOption(argc, argv) != 0) {
-    return -1;
-  }
-
-  if (global.generateGrant) {
-    dndGenerateGrant();
-    return 0;
-  }
-
-  if (global.printVersion) {
-    dndPrintVersion();
-    return 0;
-  }
-
-  if (taosCreateLog("taosdlog", 1, configDir, global.envFile, global.apolloUrl, NULL, 0) != 0) {
-    dError("failed to start TDengine since read log config error");
-    return -1;
-  }
-
-  if (taosInitCfg(configDir, global.envFile, global.apolloUrl, NULL, 0) != 0) {
-    dError("failed to start TDengine since read config error");
-    return -1;
-  }
-
-  if (global.dumpConfig) {
-    dndDumpCfg();
-    taosCleanupCfg();
-    taosCloseLog();
-    return 0;
-  }
-
-  return dndRunDnode();
-}

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

@@ -16,7 +16,7 @@
 #define _DEFAULT_SOURCE
 #include "dndInt.h"
 
-static int32_t dndInitMemory(SDnode *pDnode, const SDnodeOpt *pOption) {
+static int32_t dndInitVars(SDnode *pDnode, const SDnodeOpt *pOption) {
   pDnode->numOfSupportVnodes = pOption->numOfSupportVnodes;
   pDnode->serverPort = pOption->serverPort;
   pDnode->dataDir = strdup(pOption->dataDir);
@@ -24,8 +24,9 @@ static int32_t dndInitMemory(SDnode *pDnode, const SDnodeOpt *pOption) {
   pDnode->localFqdn = strdup(pOption->localFqdn);
   pDnode->firstEp = strdup(pOption->firstEp);
   pDnode->secondEp = strdup(pOption->secondEp);
-  pDnode->pDisks = pOption->pDisks;
+  pDnode->disks = pOption->disks;
   pDnode->numOfDisks = pOption->numOfDisks;
+  pDnode->ntype = pOption->ntype;
   pDnode->rebootTime = taosGetTimestampMs();
 
   if (pDnode->dataDir == NULL || pDnode->localEp == NULL || pDnode->localFqdn == NULL || pDnode->firstEp == NULL ||
@@ -33,18 +34,26 @@ static int32_t dndInitMemory(SDnode *pDnode, const SDnodeOpt *pOption) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
+
+  if (!tsMultiProcess || pDnode->ntype == DNODE || pDnode->ntype == NODE_MAX) {
+    pDnode->lockfile = dndCheckRunning(pDnode->dataDir);
+    if (pDnode->lockfile == NULL) {
+      return -1;
+    }
+  }
+
   return 0;
 }
 
-static void dndClearMemory(SDnode *pDnode) {
+static void dndClearVars(SDnode *pDnode) {
   for (ENodeType n = 0; n < NODE_MAX; ++n) {
     SMgmtWrapper *pMgmt = &pDnode->wrappers[n];
     taosMemoryFreeClear(pMgmt->path);
   }
-  if (pDnode->pLockFile != NULL) {
+  if (pDnode->lockfile != NULL) {
-    taosUnLockFile(pDnode->pLockFile);
+    taosUnLockFile(pDnode->lockfile);
-    taosCloseFile(&pDnode->pLockFile);
+    taosCloseFile(&pDnode->lockfile);
-    pDnode->pLockFile = NULL;
+    pDnode->lockfile = NULL;
   }
   taosMemoryFreeClear(pDnode->localEp);
   taosMemoryFreeClear(pDnode->localFqdn);
@@ -56,7 +65,7 @@ static void dndClearMemory(SDnode *pDnode) {
 }
 
 SDnode *dndCreate(const SDnodeOpt *pOption) {
-  dInfo("start to create dnode object");
+  dDebug("start to create dnode object");
   int32_t code = -1;
   char    path[PATH_MAX] = {0};
   SDnode *pDnode = NULL;
@@ -67,26 +76,12 @@ SDnode *dndCreate(const SDnodeOpt *pOption) {
     goto _OVER;
   }
 
-  if (dndInitMemory(pDnode, pOption) != 0) {
+  if (dndInitVars(pDnode, pOption) != 0) {
+    dError("failed to init variables since %s", terrstr());
     goto _OVER;
   }
 
   dndSetStatus(pDnode, DND_STAT_INIT);
-  pDnode->pLockFile = dndCheckRunning(pDnode->dataDir);
-  if (pDnode->pLockFile == NULL) {
-    goto _OVER;
-  }
-
-  if (dndInitServer(pDnode) != 0) {
-    dError("failed to init trans server since %s", terrstr());
-    goto _OVER;
-  }
-
-  if (dndInitClient(pDnode) != 0) {
-    dError("failed to init trans client since %s", terrstr());
-    goto _OVER;
-  }
-
   dmGetMgmtFp(&pDnode->wrappers[DNODE]);
   mmGetMgmtFp(&pDnode->wrappers[MNODE]);
   vmGetMgmtFp(&pDnode->wrappers[VNODES]);
@@ -94,14 +89,11 @@ SDnode *dndCreate(const SDnodeOpt *pOption) {
   smGetMgmtFp(&pDnode->wrappers[SNODE]);
   bmGetMgmtFp(&pDnode->wrappers[BNODE]);
 
-  if (dndInitMsgHandle(pDnode) != 0) {
-    goto _OVER;
-  }
-
   for (ENodeType n = 0; n < NODE_MAX; ++n) {
     SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
     snprintf(path, sizeof(path), "%s%s%s", pDnode->dataDir, TD_DIRSEP, pWrapper->name);
     pWrapper->path = strdup(path);
+    pWrapper->shm.id = -1;
     pWrapper->pDnode = pDnode;
     if (pWrapper->path == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
@@ -112,15 +104,27 @@ SDnode *dndCreate(const SDnodeOpt *pOption) {
     taosInitRWLatch(&pWrapper->latch);
   }
 
+  if (dndInitMsgHandle(pDnode) != 0) {
+    dError("failed to msg handles since %s", terrstr());
+    goto _OVER;
+  }
+
+  if (dndReadShmFile(pDnode) != 0) {
+    dError("failed to read shm file since %s", terrstr());
+    goto _OVER;
+  }
+
+  SMsgCb msgCb = dndCreateMsgcb(&pDnode->wrappers[0]);
+  tmsgSetDefaultMsgCb(&msgCb);
+
+  dInfo("dnode object is created, data:%p", pDnode);
   code = 0;
 
 _OVER:
   if (code != 0 && pDnode) {
-    dndClearMemory(pDnode);
+    dndClearVars(pDnode);
     pDnode = NULL;
     dError("failed to create dnode object since %s", terrstr());
-  } else {
-    dInfo("dnode object is created, data:%p", pDnode);
   }
 
   return pDnode;
@@ -137,15 +141,12 @@ void dndClose(SDnode *pDnode) {
   dInfo("start to close dnode, data:%p", pDnode);
   dndSetStatus(pDnode, DND_STAT_STOPPED);
 
-  dndCleanupServer(pDnode);
-  dndCleanupClient(pDnode);
-
   for (ENodeType n = 0; n < NODE_MAX; ++n) {
     SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
     dndCloseNode(pWrapper);
   }
 
-  dndClearMemory(pDnode);
+  dndClearVars(pDnode);
   dInfo("dnode object is closed, data:%p", pDnode);
 }
 

source/dnode/mgmt/main/src/dndStr.c

@@ -0,0 +1,64 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define _DEFAULT_SOURCE
+#include "dndInt.h"
+
+const char *dndStatStr(EDndStatus status) {
+  switch (status) {
+    case DND_STAT_INIT:
+      return "init";
+    case DND_STAT_RUNNING:
+      return "running";
+    case DND_STAT_STOPPED:
+      return "stopped";
+    default:
+      return "unknown";
+  }
+}
+
+const char *dndNodeLogStr(ENodeType ntype) {
+  switch (ntype) {
+    case VNODES:
+      return "vnode";
+    case QNODE:
+      return "qnode";
+    case SNODE:
+      return "snode";
+    case MNODE:
+      return "mnode";
+    case BNODE:
+      return "bnode";
+    default:
+      return "taosd";
+  }
+}
+
+const char *dndNodeProcStr(ENodeType ntype) {
+  switch (ntype) {
+    case VNODES:
+      return "taosv";
+    case QNODE:
+      return "taosq";
+    case SNODE:
+      return "taoss";
+    case MNODE:
+      return "taosm";
+    case BNODE:
+      return "taosb";
+    default:
+      return "taosd";
+  }
+}

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

@@ -320,8 +320,7 @@ static int32_t dndSendRpcReq(STransMgmt *pMgmt, const SEpSet *pEpSet, SRpcMsg *p
 }
 
 int32_t dndSendReqToDnode(SMgmtWrapper *pWrapper, const SEpSet *pEpSet, SRpcMsg *pReq) {
-  if (pWrapper->procType == PROC_CHILD) {
+  if (pWrapper->procType != PROC_CHILD) {
-  } else {
     SDnode *pDnode = pWrapper->pDnode;
     if (dndGetStatus(pDnode) != DND_STAT_RUNNING) {
       terrno = TSDB_CODE_DND_OFFLINE;
@@ -329,24 +328,25 @@ int32_t dndSendReqToDnode(SMgmtWrapper *pWrapper, const SEpSet *pEpSet, SRpcMsg
       return -1;
     }
     return dndSendRpcReq(&pDnode->trans, pEpSet, pReq);
+  } else {
+    while (taosProcPutToParentQ(pWrapper->pProc, pReq, sizeof(SRpcMsg), pReq->pCont, pReq->contLen, PROC_REQ) != 0) {
+      taosMsleep(1);
+    }
   }
 }
 
 int32_t dndSendReqToMnode(SMgmtWrapper *pWrapper, SRpcMsg *pReq) {
-  if (pWrapper->procType == PROC_CHILD) {
-  } else {
   SDnode     *pDnode = pWrapper->pDnode;
   STransMgmt *pTrans = &pDnode->trans;
   SEpSet      epSet = {0};
 
-    SMgmtWrapper *pWrapper = dndAcquireWrapper(pDnode, DNODE);
+  SMgmtWrapper *pWrapper2 = dndAcquireWrapper(pDnode, DNODE);
-    if (pWrapper != NULL) {
+  if (pWrapper2 != NULL) {
-      dmGetMnodeEpSet(pWrapper->pMgmt, &epSet);
+    dmGetMnodeEpSet(pWrapper2->pMgmt, &epSet);
-      dndReleaseWrapper(pWrapper);
+    dndReleaseWrapper(pWrapper2);
   }
   return dndSendRpcReq(pTrans, &epSet, pReq);
 }
-}
 
 void dndSendRpcRsp(SMgmtWrapper *pWrapper, const SRpcMsg *pRsp) {
   if (pRsp->code == TSDB_CODE_APP_NOT_READY) {
@@ -363,29 +363,44 @@ void dndSendRpcRsp(SMgmtWrapper *pWrapper, const SRpcMsg *pRsp) {
 }
 
 void dndSendRsp(SMgmtWrapper *pWrapper, const SRpcMsg *pRsp) {
-  if (pWrapper->procType == PROC_CHILD) {
+  if (pWrapper->procType != PROC_CHILD) {
-    int32_t code = -1;
-    do {
-      code = taosProcPutToParentQ(pWrapper->pProc, pRsp, sizeof(SRpcMsg), pRsp->pCont, pRsp->contLen, PROC_RSP);
-      if (code != 0) {
-        taosMsleep(10);
-      }
-    } while (code != 0);
-  } else {
     dndSendRpcRsp(pWrapper, pRsp);
+  } else {
+    while (taosProcPutToParentQ(pWrapper->pProc, pRsp, sizeof(SRpcMsg), pRsp->pCont, pRsp->contLen, PROC_RSP) != 0) {
+      taosMsleep(1);
+    }
   }
 }
 
 void dndRegisterBrokenLinkArg(SMgmtWrapper *pWrapper, SRpcMsg *pMsg) {
-  if (pWrapper->procType == PROC_CHILD) {
+  if (pWrapper->procType != PROC_CHILD) {
-    int32_t code = -1;
-    do {
-      code = taosProcPutToParentQ(pWrapper->pProc, pMsg, sizeof(SRpcMsg), pMsg->pCont, pMsg->contLen, PROC_REGISTER);
-      if (code != 0) {
-        taosMsleep(10);
-      }
-    } while (code != 0);
-  } else {
     rpcRegisterBrokenLinkArg(pMsg);
+  } else {
+    while (taosProcPutToParentQ(pWrapper->pProc, pMsg, sizeof(SRpcMsg), pMsg->pCont, pMsg->contLen, PROC_REG) != 0) {
+      taosMsleep(1);
     }
   }
+}
+
+static void dndReleaseHandle(SMgmtWrapper *pWrapper, void *handle, int8_t type) {
+  if (pWrapper->procType != PROC_CHILD) {
+    rpcReleaseHandle(handle, type);
+  } else {
+    SRpcMsg msg = {.handle = handle, .code = type};
+    while (taosProcPutToParentQ(pWrapper->pProc, &msg, sizeof(SRpcMsg), NULL, 0, PROC_RELEASE) != 0) {
+      taosMsleep(1);
+    }
+  }
+}
+
+SMsgCb dndCreateMsgcb(SMgmtWrapper *pWrapper) {
+  SMsgCb msgCb = {
+      .pWrapper = pWrapper,
+      .registerBrokenLinkArgFp = dndRegisterBrokenLinkArg,
+      .releaseHandleFp = dndReleaseHandle,
+      .sendMnodeReqFp = dndSendReqToMnode,
+      .sendReqFp = dndSendReqToDnode,
+      .sendRspFp = dndSendRsp,
+  };
+  return msgCb;
+}

source/dnode/mgmt/main/src/dndUtil.c

@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#define _DEFAULT_SOURCE
-#include "dndMain.h"
-
-void dndGenerateGrant() {
-#if 0
-  grantParseParameter();
-#endif
-}
-
-void dndPrintVersion() {
-#ifdef TD_ENTERPRISE
-  char *releaseName = "enterprise";
-#else
-  char *releaseName = "community";
-#endif
-  printf("%s version: %s compatible_version: %s\n", releaseName, version, compatible_version);
-  printf("gitinfo: %s\n", gitinfo);
-  printf("builuInfo: %s\n", buildinfo);
-}
-
-void dndDumpCfg() {
-  SConfig *pCfg = taosGetCfg();
-  cfgDumpCfg(pCfg, 0, 1);
-}
-
-SDnodeOpt dndGetOpt() {
-  SConfig  *pCfg = taosGetCfg();
-  SDnodeOpt option = {0};
-
-  option.numOfSupportVnodes = cfgGetItem(pCfg, "supportVnodes")->i32;
-  tstrncpy(option.dataDir, tsDataDir, sizeof(option.dataDir));
-  tstrncpy(option.firstEp, tsFirst, sizeof(option.firstEp));
-  tstrncpy(option.secondEp, tsSecond, sizeof(option.firstEp));
-  option.serverPort = tsServerPort;
-  tstrncpy(option.localFqdn, tsLocalFqdn, sizeof(option.localFqdn));
-  snprintf(option.localEp, sizeof(option.localEp), "%s:%u", option.localFqdn, option.serverPort);
-  option.pDisks = tsDiskCfg;
-  option.numOfDisks = tsDiskCfgNum;
-  return option;
-}

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

@@ -285,7 +285,7 @@ static int32_t vmInit(SMgmtWrapper *pWrapper) {
   tstrncpy(dCfg.dir, pDnode->dataDir, TSDB_FILENAME_LEN);
   dCfg.level = 0;
   dCfg.primary = 1;
-  SDiskCfg *pDisks = pDnode->pDisks;
+  SDiskCfg *pDisks = pDnode->disks;
   int32_t   numOfDisks = pDnode->numOfDisks;
   if (numOfDisks <= 0 || pDisks == NULL) {
     pDisks = &dCfg;

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

@@ -76,7 +76,7 @@ static void vmProcessFetchQueue(SQueueInfo *pInfo, SNodeMsg *pMsg) {
   SVnodeObj *pVnode = pInfo->ahandle;
 
   dTrace("msg:%p, will be processed in vnode-fetch queue", pMsg);
-  int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg);
+  int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg, pInfo);
   if (code != 0) {
     vmSendRsp(pVnode->pWrapper, pMsg, code);
     dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
@@ -168,7 +168,7 @@ static void vmProcessMergeQueue(SQueueInfo *pInfo, STaosQall *qall, int32_t numO
     taosGetQitem(qall, (void **)&pMsg);
 
     dTrace("msg:%p, will be processed in vnode-merge queue", pMsg);
-    int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg);
+    int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg, pInfo);
     if (code != 0) {
       vmSendRsp(pVnode->pWrapper, pMsg, code);
       dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
@@ -414,8 +414,7 @@ int32_t vmStartWorker(SVnodesMgmt *pMgmt) {
   pWPool->max = maxMergeThreads;
   if (tWWorkerInit(pWPool) != 0) return -1;
 
-  SSingleWorkerCfg cfg = {
+  SSingleWorkerCfg cfg = {.min = 1, .max = 1, .name = "vnode-mgmt", .fp = (FItem)vmProcessMgmtQueue, .param = pMgmt};
-      .min = 1, .max = 1, .name = "vnode-mgmt", .fp = (FItem)vmProcessMgmtQueue, .param = pMgmt};
   if (tSingleWorkerInit(&pMgmt->mgmtWorker, &cfg) != 0) {
     dError("failed to start vnode-mgmt worker since %s", terrstr());
     return -1;

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

@@ -20,7 +20,8 @@
 #define SYSTABLE_SCH_DB_NAME_LEN    ((TSDB_DB_NAME_LEN - 1) + VARSTR_HEADER_SIZE)
 
 //!!!! Note: only APPEND columns in below tables, NO insert !!!!
-static const SInfosTableSchema dnodesSchema[] = {{.name = "id",             .bytes = 2,   .type = TSDB_DATA_TYPE_SMALLINT},
+static const SInfosTableSchema dnodesSchema[] = {
+    {.name = "id", .bytes = 2, .type = TSDB_DATA_TYPE_SMALLINT},
     {.name = "endpoint", .bytes = TSDB_EP_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "vnodes", .bytes = 2, .type = TSDB_DATA_TYPE_SMALLINT},
     {.name = "max_vnodes", .bytes = 2, .type = TSDB_DATA_TYPE_SMALLINT},
@@ -28,21 +29,25 @@ static const SInfosTableSchema dnodesSchema[] = {{.name = "id",             .byt
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
     {.name = "note", .bytes = 256 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
 };
-static const SInfosTableSchema mnodesSchema[] = {{.name = "id",             .bytes = 4,   .type = TSDB_DATA_TYPE_INT},
+static const SInfosTableSchema mnodesSchema[] = {
+    {.name = "id", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "endpoint", .bytes = TSDB_EP_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "role", .bytes = 12 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "role_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
 };
-static const SInfosTableSchema modulesSchema[] = {{.name = "id",             .bytes = 4,   .type = TSDB_DATA_TYPE_INT},
+static const SInfosTableSchema modulesSchema[] = {
+    {.name = "id", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "endpoint", .bytes = 134, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "module", .bytes = 10, .type = TSDB_DATA_TYPE_BINARY},
 };
-static const SInfosTableSchema qnodesSchema[] = {{.name = "id",             .bytes = 4,   .type = TSDB_DATA_TYPE_INT},
+static const SInfosTableSchema qnodesSchema[] = {
+    {.name = "id", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "endpoint", .bytes = 134, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
 };
-static const SInfosTableSchema userDBSchema[] = {{.name = "name",             .bytes = SYSTABLE_SCH_DB_NAME_LEN,   .type = TSDB_DATA_TYPE_BINARY},
+static const SInfosTableSchema userDBSchema[] = {
+    {.name = "name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
     {.name = "vgroups", .bytes = 2, .type = TSDB_DATA_TYPE_SMALLINT},
     {.name = "ntables", .bytes = 8, .type = TSDB_DATA_TYPE_BIGINT},
@@ -59,15 +64,18 @@ static const SInfosTableSchema userDBSchema[] = {{.name = "name",             .b
     {.name = "comp", .bytes = 1, .type = TSDB_DATA_TYPE_TINYINT},
     {.name = "cachelast", .bytes = 1, .type = TSDB_DATA_TYPE_TINYINT},
     {.name = "precision", .bytes = 3 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
-//                                                 {.name = "update",           .bytes = 1,   .type = TSDB_DATA_TYPE_TINYINT},  // disable update
+    //                                                 {.name = "update",           .bytes = 1,   .type =
+    //                                                 TSDB_DATA_TYPE_TINYINT},  // disable update
 };
-static const SInfosTableSchema userFuncSchema[] = {{.name = "name",           .bytes = 32,   .type = TSDB_DATA_TYPE_BINARY},
+static const SInfosTableSchema userFuncSchema[] = {
+    {.name = "name", .bytes = 32, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
     {.name = "ntables", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "precision", .bytes = 2, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "status", .bytes = 10, .type = TSDB_DATA_TYPE_BINARY},
 };
-static const SInfosTableSchema userIdxSchema[] = {{.name = "db_name",          .bytes = 32,   .type = TSDB_DATA_TYPE_BINARY},
+static const SInfosTableSchema userIdxSchema[] = {
+    {.name = "db_name", .bytes = 32, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "table_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "index_database", .bytes = 32, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "index_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
@@ -75,7 +83,8 @@ static const SInfosTableSchema userIdxSchema[] = {{.name = "db_name",          .
     {.name = "index_type", .bytes = 10, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "index_extensions", .bytes = 256, .type = TSDB_DATA_TYPE_BINARY},
 };
-static const SInfosTableSchema userStbsSchema[] = {{.name = "stable_name",     .bytes = SYSTABLE_SCH_TABLE_NAME_LEN,  .type = TSDB_DATA_TYPE_BINARY},
+static const SInfosTableSchema userStbsSchema[] = {
+    {.name = "stable_name", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "db_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
     {.name = "columns", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
@@ -84,7 +93,8 @@ static const SInfosTableSchema userStbsSchema[] = {{.name = "stable_name",     .
     {.name = "last_update", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
     {.name = "table_comment", .bytes = 1024 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_INT},
 };
-static const SInfosTableSchema userStreamsSchema[] = {{.name = "stream_name",  .bytes = SYSTABLE_SCH_DB_NAME_LEN,  .type = TSDB_DATA_TYPE_BINARY},
+static const SInfosTableSchema userStreamsSchema[] = {
+    {.name = "stream_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "user_name", .bytes = 23, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "dest_table", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
@@ -101,7 +111,8 @@ static const SInfosTableSchema userTblsSchema[] = {
     {.name = "ttl", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "table_comment", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
 };
-static const SInfosTableSchema userTblDistSchema[] = {{.name = "db_name",                .bytes = 32,   .type = TSDB_DATA_TYPE_BINARY},
+static const SInfosTableSchema userTblDistSchema[] = {
+    {.name = "db_name", .bytes = 32, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "table_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "distributed_histogram", .bytes = 500, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "min_of_rows", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
@@ -115,12 +126,14 @@ static const SInfosTableSchema userTblDistSchema[] = {{.name = "db_name",
     {.name = "rows_in_mem", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "seek_header_time", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
 };
-static const SInfosTableSchema userUsersSchema[] = {{.name = "name",           .bytes = TSDB_USER_LEN + VARSTR_HEADER_SIZE,   .type = TSDB_DATA_TYPE_BINARY},
+static const SInfosTableSchema userUsersSchema[] = {
+    {.name = "name", .bytes = TSDB_USER_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "privilege", .bytes = 10 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
     {.name = "account", .bytes = TSDB_USER_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
 };
-static const SInfosTableSchema vgroupsSchema[] = {{.name = "vgroup_id",        .bytes = 4,   .type = TSDB_DATA_TYPE_INT},
+static const SInfosTableSchema vgroupsSchema[] = {
+    {.name = "vgroup_id", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "db_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
     {.name = "tables", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
     {.name = "status", .bytes = 12 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
@@ -136,7 +149,34 @@ static const SInfosTableSchema vgroupsSchema[] = {{.name = "vgroup_id",        .
     {.name = "file_size", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
 };
 
-static const SInfosTableMeta infosMeta[] = {{TSDB_INS_TABLE_DNODES, dnodesSchema, tListLen(dnodesSchema)},
+// TODO put into perf schema
+static const SInfosTableSchema topicSchema[] = {
+    {.name = "topic_name", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
+    {.name = "db_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
+    {.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
+    {.name = "sql", .bytes = 1024, .type = TSDB_DATA_TYPE_BINARY},
+    {.name = "row_len", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
+};
+
+static const SInfosTableSchema consumerSchema[] = {
+    {.name = "client_id", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
+    {.name = "group_id", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
+    {.name = "pid", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
+    {.name = "status", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
+    // ep
+    // up time
+    // topics
+};
+
+static const SInfosTableSchema subscribeSchema[] = {
+    {.name = "topic_name", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
+    {.name = "group_id", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
+    {.name = "vgroup_id", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
+    {.name = "client_id", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_BINARY},
+};
+
+static const SInfosTableMeta infosMeta[] = {
+    {TSDB_INS_TABLE_DNODES, dnodesSchema, tListLen(dnodesSchema)},
     {TSDB_INS_TABLE_MNODES, mnodesSchema, tListLen(mnodesSchema)},
     {TSDB_INS_TABLE_MODULES, modulesSchema, tListLen(modulesSchema)},
     {TSDB_INS_TABLE_QNODES, qnodesSchema, tListLen(qnodesSchema)},
@@ -251,5 +291,3 @@ void mndCleanupInfos(SMnode *pMnode) {
   pMnode->infosMeta = NULL;
 }
 
-
-

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

@@ -62,8 +62,6 @@ static int32_t mndPersistMqSetConnReq(SMnode *pMnode, STrans *pTrans, const SMqT
 
 static int32_t mndPersistRebalanceMsg(SMnode *pMnode, STrans *pTrans, const SMqConsumerEp *pConsumerEp);
 
-static int32_t mndInitUnassignedVg(SMnode *pMnode, const SMqTopicObj *pTopic, SMqSubscribeObj *pSub);
-
 int32_t mndInitSubscribe(SMnode *pMnode) {
   SSdbTable table = {.sdbType = SDB_SUBSCRIBE,
                      .keyType = SDB_KEY_BINARY,
@@ -98,14 +96,6 @@ static SMqSubscribeObj *mndCreateSubscription(SMnode *pMnode, const SMqTopicObj
     return NULL;
   }
 
-#if 0
-  if (mndInitUnassignedVg(pMnode, pTopic, pSub) < 0) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    tDeleteSMqSubscribeObj(pSub);
-    taosMemoryFree(pSub);
-    return NULL;
-  }
-#endif
   // TODO: disable alter subscribed table
   return pSub;
 }
@@ -210,45 +200,6 @@ static int32_t mndPersistCancelConnReq(SMnode *pMnode, STrans *pTrans, const SMq
   return 0;
 }
 
-#if 0
-static int32_t mndProcessResetOffsetReq(SNodeMsg *pMsg) {
-  SMnode             *pMnode = pMsg->pNode;
-  uint8_t            *str = pMsg->rpcMsg.pCont;
-  SMqCMResetOffsetReq req;
-
-  SCoder decoder;
-  tCoderInit(&decoder, TD_LITTLE_ENDIAN, str, pMsg->rpcMsg.contLen, TD_DECODER);
-  tDecodeSMqCMResetOffsetReq(&decoder, &req);
-
-  SHashObj *pHash = taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), false, HASH_NO_LOCK);
-  if (pHash == NULL) {
-    return -1;
-  }
-
-  for (int32_t i = 0; i < req.num; i++) {
-    SMqOffset    *pOffset = &req.offsets[i];
-    SMqVgOffsets *pVgOffset = taosHashGet(pHash, &pOffset->vgId, sizeof(int32_t));
-    if (pVgOffset == NULL) {
-      pVgOffset = taosMemoryMalloc(sizeof(SMqVgOffsets));
-      if (pVgOffset == NULL) {
-        return -1;
-      }
-      pVgOffset->offsets = taosArrayInit(0, sizeof(void *));
-      taosArrayPush(pVgOffset->offsets, &pOffset);
-    }
-    taosHashPut(pHash, &pOffset->vgId, sizeof(int32_t), &pVgOffset, sizeof(void *));
-  }
-
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, &pMsg->rpcMsg);
-  if (pTrans == NULL) {
-    mError("mq-reset-offset: failed since %s", terrstr());
-    return -1;
-  }
-
-  return 0;
-}
-#endif
-
 static int32_t mndProcessGetSubEpReq(SNodeMsg *pMsg) {
   SMnode           *pMnode = pMsg->pNode;
   SMqCMGetSubEpReq *pReq = (SMqCMGetSubEpReq *)pMsg->rpcMsg.pCont;
@@ -574,251 +525,6 @@ static int32_t mndProcessDoRebalanceMsg(SNodeMsg *pMsg) {
   return 0;
 }
 
-#if 0
-      for (int32_t j = 0; j < consumerNum; j++) {
-        bool            changed = false;
-        bool            unfished = false;
-
-        bool            canUseLeft = imbalanceSolved < imbalanceVg;
-        bool            mustUseLeft = canUseLeft && (imbalanceVg - imbalanceSolved >= consumerNum - j);
-        ASSERT(imbalanceVg - imbalanceSolved <= consumerNum - j);
-
-        int32_t maxVg = vgEachConsumer + canUseLeft;
-        int32_t minVg = vgEachConsumer + mustUseLeft;
-
-        SMqSubConsumer *pSubConsumer = taosArrayGet(pSub->consumers, j);
-        int32_t         vgThisConsumerBeforeRb = taosArrayGetSize(pSubConsumer->vgInfo);
-        int32_t         vgThisConsumerAfterRb;
-        if (vgThisConsumerBeforeRb > maxVg) {
-          vgThisConsumerAfterRb = maxVg; 
-          imbalanceSolved++;
-          changed = true;
-        } else if (vgThisConsumerBeforeRb < minVg) {
-          vgThisConsumerAfterRb = minVg;
-          if (mustUseLeft) imbalanceSolved++;
-          changed = true;
-        } else {
-          vgThisConsumerAfterRb = vgThisConsumerBeforeRb;
-        }
-
-        if (vgThisConsumerBeforeRb > vgThisConsumerAfterRb) {
-          while (taosArrayGetSize(pSubConsumer->vgInfo) > vgThisConsumerAfterRb) {
-            // put into unassigned
-            SMqConsumerEp *pConsumerEp = taosArrayPop(pSubConsumer->vgInfo);
-            ASSERT(pConsumerEp != NULL);
-            ASSERT(pConsumerEp->consumerId == pSubConsumer->consumerId);
-            taosArrayPush(unassignedVgStash, pConsumerEp);
-          }
-
-        } else if (vgThisConsumerBeforeRb < vgThisConsumerAfterRb) {
-          // assign from unassigned
-          while (taosArrayGetSize(pSubConsumer->vgInfo) < vgThisConsumerAfterRb) {
-            // if no unassgined, save j
-            if (taosArrayGetSize(unassignedVgStash) == 0) {
-              taosArrayPush(unassignedConsumerIdx, &j);
-              unfished = true;
-              break;
-            }
-            // assign vg to consumer
-            SMqConsumerEp *pConsumerEp = taosArrayPop(unassignedVgStash);
-            ASSERT(pConsumerEp != NULL);
-            pConsumerEp->oldConsumerId = pConsumerEp->consumerId;
-            pConsumerEp->consumerId = pSubConsumer->consumerId;
-            taosArrayPush(pSubConsumer->vgInfo, pConsumerEp);
-            // build msg and persist into trans
-          }
-        }
-
-        if (changed && !unfished) {
-          SMqConsumerObj *pRebConsumer = mndAcquireConsumer(pMnode, pSubConsumer->consumerId);
-          pRebConsumer->epoch++;
-          if (vgThisConsumerAfterRb != 0) {
-            atomic_store_32(&pRebConsumer->status, MQ_CONSUMER_STATUS__ACTIVE);
-          } else {
-            atomic_store_32(&pRebConsumer->status, MQ_CONSUMER_STATUS__IDLE);
-          }
-          SSdbRaw *pConsumerRaw = mndConsumerActionEncode(pRebConsumer);
-          sdbSetRawStatus(pConsumerRaw, SDB_STATUS_READY);
-          mndTransAppendRedolog(pTrans, pConsumerRaw);
-          mndReleaseConsumer(pMnode, pRebConsumer);
-          // TODO: save history
-        }
-      }
-
-      for (int32_t j = 0; j < taosArrayGetSize(unassignedConsumerIdx); j++) {
-        bool            canUseLeft = imbalanceSolved < imbalanceVg;
-        int32_t         consumerIdx = *(int32_t *)taosArrayGet(unassignedConsumerIdx, j);
-        SMqSubConsumer *pSubConsumer = taosArrayGet(pSub->consumers, consumerIdx);
-        if (canUseLeft) imbalanceSolved++;
-        // must use
-        int32_t vgThisConsumerAfterRb = taosArrayGetSize(pSubConsumer->vgInfo) + canUseLeft;
-        while (taosArrayGetSize(pSubConsumer->vgInfo) < vgEachConsumer + canUseLeft) {
-          // assign vg to consumer
-          SMqConsumerEp *pConsumerEp = taosArrayPop(unassignedVgStash);
-          ASSERT(pConsumerEp != NULL);
-          pConsumerEp->oldConsumerId = pConsumerEp->consumerId;
-          pConsumerEp->consumerId = pSubConsumer->consumerId;
-          taosArrayPush(pSubConsumer->vgInfo, pConsumerEp);
-          // build msg and persist into trans
-        }
-        SMqConsumerObj *pRebConsumer = mndAcquireConsumer(pMnode, pSubConsumer->consumerId);
-        pRebConsumer->epoch++;
-        atomic_store_32(&pRebConsumer->status, MQ_CONSUMER_STATUS__ACTIVE);
-        SSdbRaw *pConsumerRaw = mndConsumerActionEncode(pRebConsumer);
-        sdbSetRawStatus(pConsumerRaw, SDB_STATUS_READY);
-        mndTransAppendRedolog(pTrans, pConsumerRaw);
-        mndReleaseConsumer(pMnode, pRebConsumer);
-        // TODO: save history
-      }
-#endif
-
-#if 0
-    //update consumer status for the subscribption
-    for (int32_t i = 0; i < taosArrayGetSize(pSub->assigned); i++) {
-      SMqConsumerEp *pCEp = taosArrayGet(pSub->assigned, i);
-      int64_t        consumerId = pCEp->consumerId;
-      if (pCEp->status != -1) {
-        int32_t consumerHbStatus = atomic_fetch_add_32(&pCEp->consumerHbStatus, 1);
-        if (consumerHbStatus < MND_SUBSCRIBE_REBALANCE_CNT) {
-          continue;
-        }
-        // put consumer into lostConsumer
-        SMqConsumerEp* lostConsumer = taosArrayPush(pSub->lostConsumer, pCEp);
-        lostConsumer->qmsg = NULL;
-        // put vg into unassigned
-        taosArrayPush(pSub->unassignedVg, pCEp);
-        // remove from assigned
-        // TODO: swap with last one, reduce size and reset i
-        taosArrayRemove(pSub->assigned, i);
-        // remove from available consumer
-        for (int32_t j = 0; j < taosArrayGetSize(pSub->availConsumer); j++) {
-          if (*(int64_t *)taosArrayGet(pSub->availConsumer, i) == pCEp->consumerId) {
-            taosArrayRemove(pSub->availConsumer, j);
-            break;
-          }
-          // TODO: acquire consumer, set status to unavail
-        }
-#if 0
-        SMqConsumerObj* pConsumer = mndAcquireConsumer(pMnode, consumerId);
-        pConsumer->epoch++;
-        printf("current epoch %ld size %ld", pConsumer->epoch, pConsumer->topics->size);
-        SSdbRaw* pRaw = mndConsumerActionEncode(pConsumer);
-        sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-        sdbWriteNotFree(pMnode->pSdb, pRaw);
-        mndReleaseConsumer(pMnode, pConsumer);
-#endif
-      }
-    }
-    // no available consumer, skip rebalance
-    if (taosArrayGetSize(pSub->availConsumer) == 0) {
-      continue;
-    }
-    taosArrayGet(pSub->availConsumer, 0);
-    // rebalance condition1 : have unassigned vg
-    // assign vg to a consumer, trying to find the least assigned one
-    if ((sz = taosArrayGetSize(pSub->unassignedVg)) > 0) {
-      char *topic = NULL;
-      char *cgroup = NULL;
-      mndSplitSubscribeKey(pSub->key, &topic, &cgroup);
-
-      SMqTopicObj *pTopic = mndAcquireTopic(pMnode, topic);
-      STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, &pMsg->rpcMsg);
-      for (int32_t i = 0; i < sz; i++) {
-        int64_t        consumerId = *(int64_t *)taosArrayGet(pSub->availConsumer, pSub->nextConsumerIdx);
-        pSub->nextConsumerIdx = (pSub->nextConsumerIdx + 1) % taosArrayGetSize(pSub->availConsumer);
-
-        SMqConsumerEp *pCEp = taosArrayPop(pSub->unassignedVg);
-        pCEp->oldConsumerId = pCEp->consumerId;
-        pCEp->consumerId = consumerId;
-        taosArrayPush(pSub->assigned, pCEp);
-
-        SMqConsumerObj *pConsumer = mndAcquireConsumer(pMnode, consumerId);
-        pConsumer->epoch++;
-        SSdbRaw* pConsumerRaw = mndConsumerActionEncode(pConsumer);
-        sdbSetRawStatus(pConsumerRaw, SDB_STATUS_READY);
-        sdbWrite(pMnode->pSdb, pConsumerRaw);
-        mndReleaseConsumer(pMnode, pConsumer);
-
-        void* msg;
-        int32_t msgLen;
-        mndBuildRebalanceMsg(&msg, &msgLen, pTopic, pCEp, cgroup, topic);
-
-        // persist msg
-        STransAction action = {0};
-        action.epSet = pCEp->epSet;
-        action.pCont = msg;
-        action.contLen = sizeof(SMsgHead) + msgLen;
-        action.msgType = TDMT_VND_MQ_SET_CONN;
-        mndTransAppendRedoAction(pTrans, &action);
-
-        // persist data
-        SSdbRaw *pRaw = mndSubActionEncode(pSub);
-        sdbSetRawStatus(pRaw, SDB_STATUS_READY);
-        mndTransAppendRedolog(pTrans, pRaw);
-      }
-
-      if (mndTransPrepare(pMnode, pTrans) != 0) {
-        mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
-      }
-      mndReleaseTopic(pMnode, pTopic);
-      mndTransDrop(pTrans);
-      taosMemoryFreeClear(topic);
-      taosMemoryFreeClear(cgroup);
-    }
-    // rebalance condition2 : imbalance assignment
-  }
-  return 0;
-}
-#endif
-
-#if 0
-static int32_t mndInitUnassignedVg(SMnode *pMnode, const SMqTopicObj *pTopic, SMqSubscribeObj *pSub) {
-  SSdb      *pSdb = pMnode->pSdb;
-  SVgObj    *pVgroup = NULL;
-  SQueryPlan *pPlan = qStringToQueryPlan(pTopic->physicalPlan);
-  SArray    *pArray = NULL;
-  SNodeListNode    *inner = (SNodeListNode*)nodesListGetNode(pPlan->pSubplans, 0);
-  SSubplan  *plan = (SSubplan*)nodesListGetNode(inner->pNodeList, 0);
-  SArray    *unassignedVg = pSub->unassignedVg;
-
-  void *pIter = NULL;
-  while (1) {
-    pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void **)&pVgroup);
-    if (pIter == NULL) break;
-    if (pVgroup->dbUid != pTopic->dbUid) {
-      sdbRelease(pSdb, pVgroup);
-      continue;
-    }
-
-    pSub->vgNum++;
-    plan->execNode.nodeId = pVgroup->vgId;
-    plan->execNode.epset = mndGetVgroupEpset(pMnode, pVgroup);
-
-    if (schedulerConvertDagToTaskList(pPlan, &pArray) < 0) {
-      terrno = TSDB_CODE_MND_UNSUPPORTED_TOPIC;
-      mError("unsupport topic: %s, sql: %s", pTopic->name, pTopic->sql);
-      return -1;
-    }
-
-    SMqConsumerEp consumerEp = {0};
-    consumerEp.status = 0;
-    consumerEp.consumerId = -1;
-    STaskInfo *pTaskInfo = taosArrayGet(pArray, 0);
-    consumerEp.epSet = pTaskInfo->addr.epset;
-    consumerEp.vgId = pTaskInfo->addr.nodeId;
-
-    ASSERT(consumerEp.vgId == pVgroup->vgId);
-    consumerEp.qmsg = strdup(pTaskInfo->msg->msg);
-    taosArrayPush(unassignedVg, &consumerEp);
-    // TODO: free taskInfo
-    taosArrayDestroy(pArray);
-  }
-
-  /*qDestroyQueryDag(pDag);*/
-  return 0;
-}
-#endif
-
 static int32_t mndPersistMqSetConnReq(SMnode *pMnode, STrans *pTrans, const SMqTopicObj *pTopic, const char *cgroup,
                                       const SMqConsumerEp *pConsumerEp) {
   ASSERT(pConsumerEp->oldConsumerId == -1);

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

@@ -67,7 +67,9 @@ SSdbRaw *mndTopicActionEncode(SMqTopicObj *pTopic) {
 
   int32_t logicalPlanLen = strlen(pTopic->logicalPlan) + 1;
   int32_t physicalPlanLen = strlen(pTopic->physicalPlan) + 1;
-  int32_t  size = sizeof(SMqTopicObj) + logicalPlanLen + physicalPlanLen + pTopic->sqlLen + MND_TOPIC_RESERVE_SIZE;
+  int32_t swLen = taosEncodeSSchemaWrapper(NULL, &pTopic->schema);
+  int32_t size =
+      sizeof(SMqTopicObj) + logicalPlanLen + physicalPlanLen + pTopic->sqlLen + swLen + MND_TOPIC_RESERVE_SIZE;
   SSdbRaw *pRaw = sdbAllocRaw(SDB_TOPIC, MND_TOPIC_VER_NUMBER, size);
   if (pRaw == NULL) goto TOPIC_ENCODE_OVER;
 
@@ -86,7 +88,6 @@ SSdbRaw *mndTopicActionEncode(SMqTopicObj *pTopic) {
   SDB_SET_INT32(pRaw, dataPos, physicalPlanLen, TOPIC_ENCODE_OVER);
   SDB_SET_BINARY(pRaw, dataPos, pTopic->physicalPlan, physicalPlanLen, TOPIC_ENCODE_OVER);
 
-  int32_t swLen = taosEncodeSSchemaWrapper(NULL, &pTopic->schema);
   void *swBuf = taosMemoryMalloc(swLen);
   if (swBuf == NULL) {
     goto TOPIC_ENCODE_OVER;

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

@@ -771,7 +771,7 @@ static void mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans) {
                       .ahandle = pTrans->rpcAHandle,
                       .pCont = rpcCont,
                       .contLen = pTrans->rpcRspLen};
-    tmsgSendRsp(&pMnode->msgCb, &rspMsg);
+    tmsgSendRsp(&rspMsg);
     pTrans->rpcHandle = NULL;
     pTrans->rpcRsp = NULL;
     pTrans->rpcRspLen = 0;
@@ -898,7 +898,7 @@ static int32_t mndTransSendActionMsg(SMnode *pMnode, STrans *pTrans, SArray *pAr
       pAction->msgReceived = 0;
       pAction->errCode = 0;
     } else {
-      mDebug("trans:%d, action:%d not send since %s", pTrans->id, action, terrstr());
+      mError("trans:%d, action:%d not send since %s", pTrans->id, action, terrstr());
       return -1;
     }
   }
@@ -938,7 +938,7 @@ static int32_t mndTransExecuteActions(SMnode *pMnode, STrans *pTrans, SArray *pA
       return errCode;
     }
   } else {
-    mDebug("trans:%d, %d of %d actions executed, code:0x%04x", pTrans->id, numOfReceived, numOfActions, errCode & 0XFFFF);
+    mDebug("trans:%d, %d of %d actions executing", pTrans->id, numOfReceived, numOfActions);
     return TSDB_CODE_MND_ACTION_IN_PROGRESS;
   }
 }

source/dnode/vnode/CMakeLists.txt

@@ -46,14 +46,11 @@ target_link_libraries(
     PUBLIC stream
     PUBLIC qworker
     PUBLIC sync
+    # TODO: get rid of BDB
+    PUBLIC bdb
+    PUBLIC tdb
 )
 
-if(${META_DB_IMPL} STREQUAL "BDB")
-    target_link_libraries(vnode PUBLIC bdb)
-elseif(${META_DB_IMPL} STREQUAL "TDB")
-    target_link_libraries(vnode PUBLIC tdb)
-endif()
-
 if(${BUILD_TEST})
     add_subdirectory(test)
 endif(${BUILD_TEST})

source/dnode/vnode/inc/tsdb.h

@@ -100,10 +100,11 @@ int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, const char *msg);
  * @brief Insert tSma(Time-range-wise SMA) data from stream computing engine
  *
  * @param pTsdb
+ * @param indexUid
  * @param msg
  * @return int32_t
  */
-int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg);
+int32_t tsdbInsertTSmaData(STsdb *pTsdb, int64_t indexUid, const char *msg);
 
 /**
  * @brief Drop tSma data and local cache.
@@ -130,16 +131,11 @@ int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg);
  * @param pTsdb 
  * @param pData 
  * @param indexUid 
- * @param interval 
- * @param intervalUnit 
- * @param tableUid 
- * @param colId 
  * @param querySKey 
  * @param nMaxResult 
  * @return int32_t 
  */
-int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
+int32_t tsdbGetTSmaData(STsdb *pTsdb, char *pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult);
-                        tb_uid_t tableUid, col_id_t colId, TSKEY querySKey, int32_t nMaxResult);
 
 // STsdbCfg
 int  tsdbOptionsInit(STsdbCfg *);

source/dnode/vnode/inc/vnode.h

@@ -17,8 +17,9 @@
 #define _TD_VNODE_H_
 
 #include "os.h"
-#include "trpc.h"
 #include "tmsgcb.h"
+#include "tqueue.h"
+#include "trpc.h"
 
 #include "meta.h"
 #include "tarray.h"
@@ -166,7 +167,7 @@ int vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg);
  * @param pMsg The request message
  * @return int 0 for success, -1 for failure
  */
-int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg);
+int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo);
 
 /* ------------------------ SVnodeCfg ------------------------ */
 /**
@@ -185,7 +186,6 @@ void vnodeOptionsClear(SVnodeCfg *pOptions);
 
 int vnodeValidateTableHash(SVnodeCfg *pVnodeOptions, char *tableFName);
 
-
 /* ------------------------ FOR COMPILE ------------------------ */
 
 int32_t vnodeAlter(SVnode *pVnode, const SVnodeCfg *pCfg);

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

@@ -44,15 +44,14 @@ int32_t tsdbRemoveTSmaData(STsdb *pTsdb, STSma *param, STimeWindow *pWin);
 #endif
 
 // internal func
-static FORCE_INLINE int32_t tsdbEncodeTSmaKey(tb_uid_t tableUid, col_id_t colId, TSKEY tsKey, void **pData) {
+static FORCE_INLINE int32_t tsdbEncodeTSmaKey(int64_t groupId, TSKEY tsKey, void **pData) {
   int32_t len = 0;
-  len += taosEncodeFixedI64(pData, tableUid);
-  len += taosEncodeFixedU16(pData, colId);
   len += taosEncodeFixedI64(pData, tsKey);
+  len += taosEncodeFixedI64(pData, groupId);
   return len;
 }
 
-static FORCE_INLINE int tsdbRLockSma(SSmaEnv *pEnv) {
+static FORCE_INLINE int32_t tsdbRLockSma(SSmaEnv *pEnv) {
   int code = taosThreadRwlockRdlock(&(pEnv->lock));
   if (code != 0) {
     terrno = TAOS_SYSTEM_ERROR(code);
@@ -61,7 +60,7 @@ static FORCE_INLINE int tsdbRLockSma(SSmaEnv *pEnv) {
   return 0;
 }
 
-static FORCE_INLINE int tsdbWLockSma(SSmaEnv *pEnv) {
+static FORCE_INLINE int32_t tsdbWLockSma(SSmaEnv *pEnv) {
   int code = taosThreadRwlockWrlock(&(pEnv->lock));
   if (code != 0) {
     terrno = TAOS_SYSTEM_ERROR(code);
@@ -70,7 +69,7 @@ static FORCE_INLINE int tsdbWLockSma(SSmaEnv *pEnv) {
   return 0;
 }
 
-static FORCE_INLINE int tsdbUnLockSma(SSmaEnv *pEnv) {
+static FORCE_INLINE int32_t tsdbUnLockSma(SSmaEnv *pEnv) {
   int code = taosThreadRwlockUnlock(&(pEnv->lock));
   if (code != 0) {
     terrno = TAOS_SYSTEM_ERROR(code);

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

@@ -197,9 +197,9 @@ int tqCommit(STQ*);
 int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg);
 int32_t tqProcessSetConnReq(STQ* pTq, char* msg);
 int32_t tqProcessRebReq(STQ* pTq, char* msg);
-int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen);
+int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen, int32_t workerId);
 int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen);
-int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen);
+int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen, int32_t workerId);
 
 // sma
 void smaHandleRes(void* pVnode, int64_t smaId, const SArray* data);

source/dnode/vnode/src/meta/metaTDBImpl.c

@@ -15,7 +15,7 @@
 
 #include "metaDef.h"
 
-#include "tdb.h"
+#include "tdbInt.h"
 
 struct SMetaDB {
   TENV *pEnv;
@@ -25,121 +25,627 @@ struct SMetaDB {
   TDB  *pStbIdx;
   TDB  *pNtbIdx;
   TDB  *pCtbIdx;
-  // tag index hash table
-  // suid+colid --> TDB *
-  struct {
-  } tagIdxHt;
 };
 
-#define A(op, flag)                 \
+typedef struct __attribute__((__packed__)) {
-  do {                              \
+  tb_uid_t uid;
-    if ((ret = op) != 0) goto flag; \
+  int32_t  sver;
-  } while (0)
+} SSchemaDbKey;
+
+typedef struct {
+  char    *name;
+  tb_uid_t uid;
+} SNameIdxKey;
+
+typedef struct {
+  tb_uid_t suid;
+  tb_uid_t uid;
+} SCtbIdxKey;
+
+static int   metaEncodeTbInfo(void **buf, STbCfg *pTbCfg);
+static void *metaDecodeTbInfo(void *buf, STbCfg *pTbCfg);
+static int   metaEncodeSchema(void **buf, SSchemaWrapper *pSW);
+static void *metaDecodeSchema(void *buf, SSchemaWrapper *pSW);
+
+static inline int metaUidCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  tb_uid_t uid1, uid2;
+
+  ASSERT(len1 == sizeof(tb_uid_t));
+  ASSERT(len2 == sizeof(tb_uid_t));
+
+  uid1 = ((tb_uid_t *)arg1)[0];
+  uid2 = ((tb_uid_t *)arg2)[0];
+
+  if (uid1 < uid2) {
+    return -1;
+  }
+  if (uid1 == uid2) {
+    return 0;
+  } else {
+    return 1;
+  }
+}
+
+static inline int metaSchemaKeyCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  int           c;
+  SSchemaDbKey *pKey1 = (SSchemaDbKey *)arg1;
+  SSchemaDbKey *pKey2 = (SSchemaDbKey *)arg2;
+
+  c = metaUidCmpr(arg1, sizeof(tb_uid_t), arg2, sizeof(tb_uid_t));
+  if (c) return c;
+
+  if (pKey1->sver > pKey2->sver) {
+    return 1;
+  } else if (pKey1->sver == pKey2->sver) {
+    return 0;
+  } else {
+    return -1;
+  }
+}
+
+static inline int metaNameIdxCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  return strcmp((char *)arg1, (char *)arg2);
+}
+
+static inline int metaCtbIdxCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  int         c;
+  SCtbIdxKey *pKey1 = (SCtbIdxKey *)arg1;
+  SCtbIdxKey *pKey2 = (SCtbIdxKey *)arg2;
+
+  c = metaUidCmpr(arg1, sizeof(tb_uid_t), arg2, sizeof(tb_uid_t));
+  if (c) return c;
+
+  return metaUidCmpr(&pKey1->uid, sizeof(tb_uid_t), &pKey2->uid, sizeof(tb_uid_t));
+}
 
 int metaOpenDB(SMeta *pMeta) {
-  SMetaDB *pDb;
+  SMetaDB *pMetaDb;
-  TENV *   pEnv;
-  TDB *    pTbDB;
-  TDB *    pSchemaDB;
-  TDB *    pNameIdx;
-  TDB *    pStbIdx;
-  TDB *    pNtbIdx;
-  TDB *    pCtbIdx;
   int      ret;
 
-  pDb = (SMetaDB *)taosMemoryCalloc(1, sizeof(*pDb));
+  // allocate DB handle
-  if (pDb == NULL) {
+  pMetaDb = taosMemoryCalloc(1, sizeof(*pMetaDb));
+  if (pMetaDb == NULL) {
+    // TODO
+    ASSERT(0);
     return -1;
   }
 
-  // Create and open the ENV
+  // open the ENV
-  A((tdbEnvCreate(&pEnv)), _err);
+  ret = tdbEnvOpen(pMeta->path, 4096, 256, &(pMetaDb->pEnv));
-#if 0
+  if (ret < 0) {
-  // Set options of the environment
+    // TODO
-  A(tdbEnvSetPageSize(pEnv, 8192), _err);
+    ASSERT(0);
-  A(tdbEnvSetCacheSize(pEnv, 16 * 1024 * 1024), _err);
-#endif
-  A((tdbEnvOpen(&pEnv)), _err);
-
-  // Create and open each DB
-  A(tdbCreate(&pTbDB), _err);
-  A(tdbOpen(&pTbDB, "table.db", NULL, pEnv), _err);
-
-  A(tdbCreate(&pSchemaDB), _err);
-  A(tdbOpen(&pSchemaDB, "schema.db", NULL, pEnv), _err);
-
-  A(tdbCreate(&pNameIdx), _err);
-  A(tdbOpen(&pNameIdx, "name.db", NULL, pEnv), _err);
-  // tdbAssociate();
-
-  pDb->pEnv = pEnv;
-  pDb->pTbDB = pTbDB;
-  pDb->pSchemaDB = pSchemaDB;
-  pMeta->pDB = pDb;
-  return 0;
-
-_err:
     return -1;
   }
 
+  // open table DB
+  ret = tdbDbOpen("table.db", sizeof(tb_uid_t), TDB_VARIANT_LEN, metaUidCmpr, pMetaDb->pEnv, &(pMetaDb->pTbDB));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  // open schema DB
+  ret = tdbDbOpen("schema.db", sizeof(SSchemaDbKey), TDB_VARIANT_LEN, metaSchemaKeyCmpr, pMetaDb->pEnv,
+                  &(pMetaDb->pSchemaDB));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  ret = tdbDbOpen("name.idx", TDB_VARIANT_LEN, 0, metaNameIdxCmpr, pMetaDb->pEnv, &(pMetaDb->pNameIdx));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  ret = tdbDbOpen("stb.idx", sizeof(tb_uid_t), 0, metaUidCmpr, pMetaDb->pEnv, &(pMetaDb->pStbIdx));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  ret = tdbDbOpen("ntb.idx", sizeof(tb_uid_t), 0, metaUidCmpr, pMetaDb->pEnv, &(pMetaDb->pNtbIdx));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  ret = tdbDbOpen("ctb.idx", sizeof(SCtbIdxKey), 0, metaCtbIdxCmpr, pMetaDb->pEnv, &(pMetaDb->pCtbIdx));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  pMeta->pDB = pMetaDb;
+  return 0;
+}
+
 void metaCloseDB(SMeta *pMeta) {
-  // TODO
+  if (pMeta->pDB) {
+    tdbDbClose(pMeta->pDB->pCtbIdx);
+    tdbDbClose(pMeta->pDB->pNtbIdx);
+    tdbDbClose(pMeta->pDB->pStbIdx);
+    tdbDbClose(pMeta->pDB->pNameIdx);
+    tdbDbClose(pMeta->pDB->pSchemaDB);
+    tdbDbClose(pMeta->pDB->pTbDB);
+    taosMemoryFree(pMeta->pDB);
+  }
 }
 
 int metaSaveTableToDB(SMeta *pMeta, STbCfg *pTbCfg) {
-  // TODO
+  tb_uid_t       uid;
+  SMetaDB       *pMetaDb;
+  void          *pKey;
+  void          *pVal;
+  int            kLen;
+  int            vLen;
+  int            ret;
+  char           buf[512];
+  void          *pBuf;
+  SCtbIdxKey     ctbIdxKey;
+  SSchemaDbKey   schemaDbKey;
+  SSchemaWrapper schemaWrapper;
+
+  pMetaDb = pMeta->pDB;
+
+  // TODO: make this operation pre-process
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    uid = pTbCfg->stbCfg.suid;
+  } else {
+    uid = metaGenerateUid(pMeta);
+  }
+
+  // save to table.db
+  pKey = &uid;
+  kLen = sizeof(uid);
+  pVal = pBuf = buf;
+  metaEncodeTbInfo(&pBuf, pTbCfg);
+  vLen = POINTER_DISTANCE(pBuf, buf);
+  ret = tdbDbInsert(pMetaDb->pTbDB, pKey, kLen, pVal, vLen);
+  if (ret < 0) {
+    return -1;
+  }
+
+  // save to schema.db for META_SUPER_TABLE and META_NORMAL_TABLE
+  if (pTbCfg->type != META_CHILD_TABLE) {
+    schemaDbKey.uid = uid;
+    schemaDbKey.sver = 0;  // TODO
+    pKey = &schemaDbKey;
+    kLen = sizeof(schemaDbKey);
+
+    if (pTbCfg->type == META_SUPER_TABLE) {
+      schemaWrapper.nCols = pTbCfg->stbCfg.nCols;
+      schemaWrapper.pSchema = pTbCfg->stbCfg.pSchema;
+    } else {
+      schemaWrapper.nCols = pTbCfg->ntbCfg.nCols;
+      schemaWrapper.pSchema = pTbCfg->ntbCfg.pSchema;
+    }
+    pVal = pBuf = buf;
+    metaEncodeSchema(&pBuf, &schemaWrapper);
+    vLen = POINTER_DISTANCE(pBuf, buf);
+    ret = tdbDbInsert(pMetaDb->pSchemaDB, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  }
+
+  // update name.idx
+  int nameLen = strlen(pTbCfg->name);
+  memcpy(buf, pTbCfg->name, nameLen + 1);
+  ((tb_uid_t *)(buf + nameLen + 1))[0] = uid;
+  pKey = buf;
+  kLen = nameLen + 1 + sizeof(uid);
+  pVal = NULL;
+  vLen = 0;
+  ret = tdbDbInsert(pMetaDb->pNameIdx, pKey, kLen, pVal, vLen);
+  if (ret < 0) {
+    return -1;
+  }
+
+  // update other index
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    pKey = &uid;
+    kLen = sizeof(uid);
+    pVal = NULL;
+    vLen = 0;
+    ret = tdbDbInsert(pMetaDb->pStbIdx, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  } else if (pTbCfg->type == META_CHILD_TABLE) {
+    ctbIdxKey.suid = pTbCfg->ctbCfg.suid;
+    ctbIdxKey.uid = uid;
+    pKey = &ctbIdxKey;
+    kLen = sizeof(ctbIdxKey);
+    pVal = NULL;
+    vLen = 0;
+    ret = tdbDbInsert(pMetaDb->pCtbIdx, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  } else if (pTbCfg->type == META_NORMAL_TABLE) {
+    pKey = &uid;
+    kLen = sizeof(uid);
+    pVal = NULL;
+    vLen = 0;
+    ret = tdbDbInsert(pMetaDb->pNtbIdx, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  }
+
   return 0;
 }
 
 int metaRemoveTableFromDb(SMeta *pMeta, tb_uid_t uid) {
   // TODO
+  ASSERT(0);
   return 0;
 }
 
 STbCfg *metaGetTbInfoByUid(SMeta *pMeta, tb_uid_t uid) {
-  // TODO
+  int      ret;
+  SMetaDB *pMetaDb = pMeta->pDB;
+  void    *pKey;
+  void    *pVal;
+  int      kLen;
+  int      vLen;
+  STbCfg  *pTbCfg;
+
+  // Fetch
+  pKey = &uid;
+  kLen = sizeof(uid);
+  pVal = NULL;
+  ret = tdbDbGet(pMetaDb->pTbDB, pKey, kLen, &pVal, &vLen);
+  if (ret < 0) {
     return NULL;
   }
 
+  // Decode
+  pTbCfg = taosMemoryMalloc(sizeof(*pTbCfg));
+  metaDecodeTbInfo(pVal, pTbCfg);
+
+  TDB_FREE(pVal);
+
+  return pTbCfg;
+}
+
 STbCfg *metaGetTbInfoByName(SMeta *pMeta, char *tbname, tb_uid_t *uid) {
-  // TODO
+  void *pKey;
+  void *pVal;
+  void *ppKey;
+  int   pkLen;
+  int   kLen;
+  int   vLen;
+  int   ret;
+
+  pKey = tbname;
+  kLen = strlen(tbname) + 1;
+  pVal = NULL;
+  ppKey = NULL;
+  ret = tdbDbPGet(pMeta->pDB->pNameIdx, pKey, kLen, &ppKey, &pkLen, &pVal, &vLen);
+  if (ret < 0) {
     return NULL;
   }
 
+  ASSERT(pkLen == kLen + sizeof(uid));
+
+  *uid = *(tb_uid_t *)POINTER_SHIFT(ppKey, kLen);
+  TDB_FREE(ppKey);
+  TDB_FREE(pVal);
+
+  return metaGetTbInfoByUid(pMeta, *uid);
+}
+
 SSchemaWrapper *metaGetTableSchema(SMeta *pMeta, tb_uid_t uid, int32_t sver, bool isinline) {
-  // TODO
+  void           *pKey;
+  void           *pVal;
+  int             kLen;
+  int             vLen;
+  int             ret;
+  SSchemaDbKey    schemaDbKey;
+  SSchemaWrapper *pSchemaWrapper;
+  void           *pBuf;
+
+  // fetch
+  schemaDbKey.uid = uid;
+  schemaDbKey.sver = sver;
+  pKey = &schemaDbKey;
+  kLen = sizeof(schemaDbKey);
+  pVal = NULL;
+  ret = tdbDbGet(pMeta->pDB->pSchemaDB, pKey, kLen, &pVal, &vLen);
+  if (ret < 0) {
     return NULL;
   }
 
+  // decode
+  pBuf = pVal;
+  pSchemaWrapper = taosMemoryMalloc(sizeof(*pSchemaWrapper));
+  metaDecodeSchema(pBuf, pSchemaWrapper);
+
+  TDB_FREE(pVal);
+
+  return pSchemaWrapper;
+}
+
 STSchema *metaGetTbTSchema(SMeta *pMeta, tb_uid_t uid, int32_t sver) {
-  // TODO
+  tb_uid_t        quid;
+  SSchemaWrapper *pSW;
+  STSchemaBuilder sb;
+  SSchema        *pSchema;
+  STSchema       *pTSchema;
+  STbCfg         *pTbCfg;
+
+  pTbCfg = metaGetTbInfoByUid(pMeta, uid);
+  if (pTbCfg->type == META_CHILD_TABLE) {
+    quid = pTbCfg->ctbCfg.suid;
+  } else {
+    quid = uid;
+  }
+
+  pSW = metaGetTableSchema(pMeta, quid, sver, true);
+  if (pSW == NULL) {
     return NULL;
   }
 
+  tdInitTSchemaBuilder(&sb, 0);
+  for (int i = 0; i < pSW->nCols; i++) {
+    pSchema = pSW->pSchema + i;
+    tdAddColToSchema(&sb, pSchema->type, pSchema->colId, pSchema->bytes);
+  }
+  pTSchema = tdGetSchemaFromBuilder(&sb);
+  tdDestroyTSchemaBuilder(&sb);
+
+  return pTSchema;
+}
+
+struct SMTbCursor {
+  TDBC *pDbc;
+};
+
 SMTbCursor *metaOpenTbCursor(SMeta *pMeta) {
-  // TODO
+  SMTbCursor *pTbCur = NULL;
+  SMetaDB    *pDB = pMeta->pDB;
+
+  pTbCur = (SMTbCursor *)taosMemoryCalloc(1, sizeof(*pTbCur));
+  if (pTbCur == NULL) {
     return NULL;
   }
 
+  tdbDbcOpen(pDB->pTbDB, &pTbCur->pDbc);
+
+  return pTbCur;
+}
+
 void metaCloseTbCursor(SMTbCursor *pTbCur) {
-  // TODO
+  if (pTbCur) {
+    if (pTbCur->pDbc) {
+      tdbDbcClose(pTbCur->pDbc);
+    }
+    taosMemoryFree(pTbCur);
+  }
 }
 
 char *metaTbCursorNext(SMTbCursor *pTbCur) {
-  // TODO
+  void  *pKey = NULL;
+  void  *pVal = NULL;
+  int    kLen;
+  int    vLen;
+  int    ret;
+  void  *pBuf;
+  STbCfg tbCfg;
+
+  for (;;) {
+    ret = tdbDbNext(pTbCur->pDbc, &pKey, &kLen, &pVal, &vLen);
+    if (ret < 0) break;
+    pBuf = pVal;
+    metaDecodeTbInfo(pBuf, &tbCfg);
+    if (tbCfg.type == META_SUPER_TABLE) {
+      taosMemoryFree(tbCfg.name);
+      taosMemoryFree(tbCfg.stbCfg.pTagSchema);
+      continue;
+      ;
+    } else if (tbCfg.type == META_CHILD_TABLE) {
+      kvRowFree(tbCfg.ctbCfg.pTag);
+    }
+
+    return tbCfg.name;
+  }
+
   return NULL;
 }
 
+struct SMCtbCursor {
+  TDBC    *pCur;
+  tb_uid_t suid;
+  void    *pKey;
+  void    *pVal;
+  int      kLen;
+  int      vLen;
+};
+
 SMCtbCursor *metaOpenCtbCursor(SMeta *pMeta, tb_uid_t uid) {
-  // TODO
+  SMCtbCursor *pCtbCur = NULL;
+  SMetaDB     *pDB = pMeta->pDB;
+  int          ret;
+
+  pCtbCur = (SMCtbCursor *)taosMemoryCalloc(1, sizeof(*pCtbCur));
+  if (pCtbCur == NULL) {
     return NULL;
   }
 
+  pCtbCur->suid = uid;
+  ret = tdbDbcOpen(pDB->pCtbIdx, &pCtbCur->pCur);
+  if (ret < 0) {
+    taosMemoryFree(pCtbCur);
+    return NULL;
+  }
+
+  // TODO: move the cursor to the suid there
+
+  return pCtbCur;
+}
+
 void metaCloseCtbCurosr(SMCtbCursor *pCtbCur) {
-  // TODO
+  if (pCtbCur) {
+    if (pCtbCur->pCur) {
+      tdbDbcClose(pCtbCur->pCur);
+
+      TDB_FREE(pCtbCur->pKey);
+      TDB_FREE(pCtbCur->pVal);
+    }
+
+    taosMemoryFree(pCtbCur);
+  }
 }
 
 tb_uid_t metaCtbCursorNext(SMCtbCursor *pCtbCur) {
-  // TODO
+  int         ret;
+  SCtbIdxKey *pCtbIdxKey;
+
+  ret = tdbDbNext(pCtbCur->pCur, &pCtbCur->pKey, &pCtbCur->kLen, &pCtbCur->pVal, &pCtbCur->vLen);
+  if (ret < 0) {
     return 0;
   }
+
+  pCtbIdxKey = pCtbCur->pVal;
+
+  return pCtbIdxKey->uid;
+}
+
+int metaGetTbNum(SMeta *pMeta) {
+  // TODO
+  // ASSERT(0);
+  return 0;
+}
+
+STSmaWrapper *metaGetSmaInfoByTable(SMeta *pMeta, tb_uid_t uid) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+int metaRemoveSmaFromDb(SMeta *pMeta, int64_t indexUid) {
+  // TODO
+  ASSERT(0);
+  return 0;
+}
+
+int metaSaveSmaToDB(SMeta *pMeta, STSma *pSmaCfg) {
+  // TODO
+  ASSERT(0);
+  return 0;
+}
+
+STSma *metaGetSmaInfoByIndex(SMeta *pMeta, int64_t indexUid) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+const char *metaSmaCursorNext(SMSmaCursor *pCur) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+void metaCloseSmaCurosr(SMSmaCursor *pCur) {
+  // TODO
+  ASSERT(0);
+}
+
+SArray *metaGetSmaTbUids(SMeta *pMeta, bool isDup) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+SMSmaCursor *metaOpenSmaCursor(SMeta *pMeta, tb_uid_t uid) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+static int metaEncodeSchema(void **buf, SSchemaWrapper *pSW) {
+  int      tlen = 0;
+  SSchema *pSchema;
+
+  tlen += taosEncodeFixedU32(buf, pSW->nCols);
+  for (int i = 0; i < pSW->nCols; i++) {
+    pSchema = pSW->pSchema + i;
+    tlen += taosEncodeFixedI8(buf, pSchema->type);
+    tlen += taosEncodeFixedI16(buf, pSchema->colId);
+    tlen += taosEncodeFixedI32(buf, pSchema->bytes);
+    tlen += taosEncodeString(buf, pSchema->name);
+  }
+
+  return tlen;
+}
+
+static void *metaDecodeSchema(void *buf, SSchemaWrapper *pSW) {
+  SSchema *pSchema;
+
+  buf = taosDecodeFixedU32(buf, &pSW->nCols);
+  pSW->pSchema = (SSchema *)taosMemoryMalloc(sizeof(SSchema) * pSW->nCols);
+  for (int i = 0; i < pSW->nCols; i++) {
+    pSchema = pSW->pSchema + i;
+    buf = taosDecodeFixedI8(buf, &pSchema->type);
+    buf = taosDecodeFixedI16(buf, &pSchema->colId);
+    buf = taosDecodeFixedI32(buf, &pSchema->bytes);
+    buf = taosDecodeStringTo(buf, pSchema->name);
+  }
+
+  return buf;
+}
+
+static int metaEncodeTbInfo(void **buf, STbCfg *pTbCfg) {
+  int tsize = 0;
+
+  tsize += taosEncodeString(buf, pTbCfg->name);
+  tsize += taosEncodeFixedU32(buf, pTbCfg->ttl);
+  tsize += taosEncodeFixedU32(buf, pTbCfg->keep);
+  tsize += taosEncodeFixedU8(buf, pTbCfg->info);
+
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    SSchemaWrapper sw = {.nCols = pTbCfg->stbCfg.nTagCols, .pSchema = pTbCfg->stbCfg.pTagSchema};
+    tsize += metaEncodeSchema(buf, &sw);
+  } else if (pTbCfg->type == META_CHILD_TABLE) {
+    tsize += taosEncodeFixedU64(buf, pTbCfg->ctbCfg.suid);
+    tsize += tdEncodeKVRow(buf, pTbCfg->ctbCfg.pTag);
+  } else if (pTbCfg->type == META_NORMAL_TABLE) {
+    // TODO
+  } else {
+    ASSERT(0);
+  }
+
+  return tsize;
+}
+
+static void *metaDecodeTbInfo(void *buf, STbCfg *pTbCfg) {
+  buf = taosDecodeString(buf, &(pTbCfg->name));
+  buf = taosDecodeFixedU32(buf, &(pTbCfg->ttl));
+  buf = taosDecodeFixedU32(buf, &(pTbCfg->keep));
+  buf = taosDecodeFixedU8(buf, &(pTbCfg->info));
+
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    SSchemaWrapper sw;
+    buf = metaDecodeSchema(buf, &sw);
+    pTbCfg->stbCfg.nTagCols = sw.nCols;
+    pTbCfg->stbCfg.pTagSchema = sw.pSchema;
+  } else if (pTbCfg->type == META_CHILD_TABLE) {
+    buf = taosDecodeFixedU64(buf, &(pTbCfg->ctbCfg.suid));
+    buf = tdDecodeKVRow(buf, &(pTbCfg->ctbCfg.pTag));
+  } else if (pTbCfg->type == META_NORMAL_TABLE) {
+    // TODO
+  } else {
+    ASSERT(0);
+  }
+  return buf;
+}

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

@@ -275,7 +275,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
     pMsg->pCont = NULL;
     pMsg->contLen = 0;
     pMsg->code = -1;
-    rpcSendResponse(pMsg);
+    tmsgSendRsp(pMsg);
     return 0;
   }
 
@@ -340,7 +340,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
         pMsg->pCont = buf;
         pMsg->contLen = tlen;
         pMsg->code = 0;
-        rpcSendResponse(pMsg);
+        tmsgSendRsp(pMsg);
         return 0;
       }
     } else {
@@ -367,7 +367,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
   pMsg->pCont = buf;
   pMsg->contLen = tlen;
   pMsg->code = 0;
-  rpcSendResponse(pMsg);
+  tmsgSendRsp(pMsg);
   /*}*/
 
   return 0;
@@ -489,7 +489,7 @@ int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
   return 0;
 }
 
-int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen) {
+int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen, int32_t workerId) {
   void* pIter = NULL;
 
   while (1) {
@@ -497,14 +497,14 @@ int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen) {
     if (pIter == NULL) break;
     SStreamTask* pTask = (SStreamTask*)pIter;
 
-    if (streamExecTask(pTask, &pTq->pVnode->msgCb, data, STREAM_DATA_TYPE_SUBMIT_BLOCK, 0) < 0) {
+    if (streamExecTask(pTask, &pTq->pVnode->msgCb, data, STREAM_DATA_TYPE_SUBMIT_BLOCK, workerId) < 0) {
       // TODO
     }
   }
   return 0;
 }
 
-int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen) {
+int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen, int32_t workerId) {
   SStreamTaskExecReq req;
   tDecodeSStreamTaskExecReq(msg, &req);
 
@@ -514,7 +514,7 @@ int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen) {
   SStreamTask* pTask = taosHashGet(pTq->pStreamTasks, &taskId, sizeof(int32_t));
   ASSERT(pTask);
 
-  if (streamExecTask(pTask, &pTq->pVnode->msgCb, req.data, STREAM_DATA_TYPE_SSDATA_BLOCK, 0) < 0) {
+  if (streamExecTask(pTask, &pTq->pVnode->msgCb, req.data, STREAM_DATA_TYPE_SSDATA_BLOCK, workerId) < 0) {
     // TODO
   }
   return 0;

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

@@ -114,10 +114,9 @@ SArray* tqRetrieveDataBlock(STqReadHandle* pHandle) {
   SSchemaWrapper* pSchemaWrapper = pHandle->pSchemaWrapper;
 
   int32_t numOfRows = pHandle->pBlock->numOfRows;
-  int32_t numOfCols = pHandle->pSchema->numOfCols;
+  /*int32_t numOfCols = pHandle->pSchema->numOfCols;*/
   int32_t colNumNeed = taosArrayGetSize(pHandle->pColIdList);
 
-  // TODO: stable case
   if (colNumNeed > pSchemaWrapper->nCols) {
     colNumNeed = pSchemaWrapper->nCols;
   }
@@ -138,58 +137,24 @@ SArray* tqRetrieveDataBlock(STqReadHandle* pHandle) {
       colNeed++;
     } else {
       SColumnInfoData colInfo = {0};
-      int             sz = numOfRows * pColSchema->bytes;
+      /*int             sz = numOfRows * pColSchema->bytes;*/
       colInfo.info.bytes = pColSchema->bytes;
       colInfo.info.colId = pColSchema->colId;
       colInfo.info.type = pColSchema->type;
 
-      colInfo.pData = taosMemoryCalloc(1, sz);
+      if (blockDataEnsureColumnCapacity(&colInfo, numOfRows) < 0) {
-      if (colInfo.pData == NULL) {
+        taosArrayDestroyEx(pArray, (void (*)(void*))tDeleteSSDataBlock);
-        // TODO free
-        taosArrayDestroy(pArray);
         return NULL;
       }
-
-      blockDataEnsureColumnCapacity(&colInfo, numOfRows);
       taosArrayPush(pArray, &colInfo);
       colMeta++;
       colNeed++;
     }
   }
 
-#if 0
-  int j = 0;
-  for (int32_t i = 0; i < colNumNeed; i++) {
-    col_id_t colId = *(col_id_t*)taosArrayGet(pHandle->pColIdList, i);
-    while (j < pSchemaWrapper->nCols && pSchemaWrapper->pSchema[j].colId < colId) {
-      j++;
-    }
-    if (j >= pSchemaWrapper->nCols) {
-      continue;
-    }
-    SSchema*        pColSchema = &pSchemaWrapper->pSchema[j];
-    SColumnInfoData colInfo = {0};
-    int             sz = numOfRows * pColSchema->bytes;
-    colInfo.info.bytes = pColSchema->bytes;
-    colInfo.info.colId = colId;
-    colInfo.info.type = pColSchema->type;
-
-    colInfo.pData = taosMemoryCalloc(1, sz);
-    if (colInfo.pData == NULL) {
-      // TODO free
-      taosArrayDestroy(pArray);
-      return NULL;
-    }
-
-    blockDataEnsureColumnCapacity(&colInfo, numOfRows);
-    taosArrayPush(pArray, &colInfo);
-  }
-#endif
-
   STSRowIter iter = {0};
   tdSTSRowIterInit(&iter, pTschema);
   STSRow* row;
-  // int32_t kvIdx = 0;
   int32_t curRow = 0;
   tInitSubmitBlkIter(pHandle->pBlock, &pHandle->blkIter);
   while ((row = tGetSubmitBlkNext(&pHandle->blkIter)) != NULL) {
@@ -202,25 +167,11 @@ SArray* tqRetrieveDataBlock(STqReadHandle* pHandle) {
       if (!tdSTSRowIterNext(&iter, pColData->info.colId, pColData->info.type, &sVal)) {
         break;
       }
-      memcpy(POINTER_SHIFT(pColData->pData, curRow * pColData->info.bytes), sVal.val, pColData->info.bytes);
+      if (colDataAppend(pColData, curRow, sVal.val, sVal.valType == TD_VTYPE_NULL) < 0) {
+        taosArrayDestroyEx(pArray, (void (*)(void*))tDeleteSSDataBlock);
+        return NULL;
       }
-#if 0
-    for (int32_t i = 0; i < colNumNeed; i++) {
-      SColumnInfoData* pColData = taosArrayGet(pArray, i);
-      STColumn*        pCol = schemaColAt(pTschema, i);
-      // TODO
-      if(pCol->colId != pColData->info.colId) {
-        continue;
     }
-      // void* val = tdGetMemRowDataOfColEx(row, pCol->colId, pCol->type, TD_DATA_ROW_HEAD_SIZE + pCol->offset, &kvIdx);
-      SCellVal sVal = {0};
-      if (!tdSTSRowIterNext(&iter, pCol->colId, pCol->type, &sVal)) {
-        // TODO: reach end
-        break;
-      }
-      memcpy(POINTER_SHIFT(pColData->pData, curRow * pCol->bytes), sVal.val, pCol->bytes);
-    }
-#endif
     curRow++;
   }
   return pArray;

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

@@ -24,7 +24,7 @@ static const char *TSDB_SMA_DNAME[] = {
 #define SMA_STORAGE_TSDB_DAYS   30
 #define SMA_STORAGE_TSDB_TIMES  10
 #define SMA_STORAGE_SPLIT_HOURS 24
-#define SMA_KEY_LEN             18  // tableUid_colId_TSKEY 8+2+8
+#define SMA_KEY_LEN             16  // TSKEY+groupId 8+8
 #define SMA_DROP_EXPIRED_TIME   10  // default is 10 seconds
 
 #define SMA_STATE_HASH_SLOT      4
@@ -40,7 +40,7 @@ typedef enum {
 typedef struct {
   STsdb        *pTsdb;
   SDBFile       dFile;
-  SSDataBlock *pData;     // sma data
+  const SArray *pDataBlocks;  // sma data
   int32_t       interval;     // interval with the precision of DB
 } STSmaWriteH;
 
@@ -94,26 +94,24 @@ static int32_t  tsdbUnRefSmaStat(STsdb *pTsdb, SSmaStat *pStat);
 
 // read data
 // TODO: This is the basic params, and should wrap the params to a queryHandle.
-static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval,
+static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, char *pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult);
-                                   int8_t intervalUnit, tb_uid_t tableUid, col_id_t colId, TSKEY querySKey,
-                                   int32_t nMaxResult);
 
 // insert data
-static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, SSDataBlock *pData, int64_t interval,
+static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, const SArray *pDataBlocks, int64_t interval,
                                   int8_t intervalUnit);
 static void    tsdbDestroyTSmaWriteH(STSmaWriteH *pSmaH);
 static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, int64_t interval, int8_t intervalUnit);
 static int32_t tsdbGetSmaStorageLevel(int64_t interval, int8_t intervalUnit);
-static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *pData);
+static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, int32_t fid);
 static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t keyLen, void *pData, uint32_t dataLen);
-static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision);
+static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision, bool adjusted);
 static int32_t tsdbGetTSmaDays(STsdb *pTsdb, int64_t interval, int32_t storageLevel);
-static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int64_t indexUid, int32_t fid);
+static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, int64_t indexUid, int32_t fid);
 static int32_t tsdbInitTSmaFile(STSmaReadH *pSmaH, int64_t indexUid, TSKEY skey);
 static bool    tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, TSKEY *queryKey);
 static void    tsdbGetSmaDir(int32_t vgId, ETsdbSmaType smaType, char dirName[]);
-static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg);
+static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, int64_t indexUid, const char *msg);
-static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg);
+static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, const char *msg);
 
 // mgmt interface
 static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid);
@@ -387,7 +385,6 @@ static int32_t tsdbCheckAndInitSmaEnv(STsdb *pTsdb, int8_t smaType) {
   return TSDB_CODE_SUCCESS;
 };
 
-
 static int32_t tsdbSetExpiredWindow(STsdb *pTsdb, SHashObj *pItemsHash, int64_t indexUid, int64_t winSKey) {
   SSmaStatItem *pItem = taosHashGet(pItemsHash, &indexUid, sizeof(indexUid));
   if (pItem == NULL) {
@@ -480,18 +477,15 @@ int32_t tsdbUpdateExpiredWindowImpl(STsdb *pTsdb, const char *msg) {
   TSKEY expiredWindows[SMA_TEST_EXPIRED_WINDOW_SIZE];
 #endif
 
-
   // Firstly, assume that tSma can only be created on super table/normal table.
   // getActiveTimeWindow
 
-
   SSmaEnv  *pEnv = REPO_SMA_ENV(pTsdb, TSDB_SMA_TYPE_TIME_RANGE);
   SSmaStat *pStat = SMA_ENV_STAT(pEnv);
   SHashObj *pItemsHash = SMA_ENV_STAT_ITEMS(pEnv);
 
   TASSERT(pEnv != NULL && pStat != NULL && pItemsHash != NULL);
 
-
   // basic procedure
   // TODO: optimization
   tsdbRefSmaStat(pTsdb, pStat);
@@ -657,9 +651,8 @@ static int32_t tsdbGetSmaStorageLevel(int64_t interval, int8_t intervalUnit) {
  */
 static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t keyLen, void *pData, uint32_t dataLen) {
   SDBFile *pDBFile = &pSmaH->dFile;
-  tsdbDebug("vgId:%d insert sma data blocks into %s: smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64 ", dataLen %d",
+  printf("\nvgId:%d insert sma data blocks into %s: smaKey %" PRIx64 "-%" PRIx64 ", dataLen %" PRIu32 "\n",
-            REPO_ID(pSmaH->pTsdb), pDBFile->path, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
+         REPO_ID(pSmaH->pTsdb), pDBFile->path, *(int64_t *)smaKey, *(int64_t *)POINTER_SHIFT(smaKey, 8), dataLen);
-            *(int64_t *)POINTER_SHIFT(smaKey, 10), dataLen);
 
   // TODO: insert sma data blocks into B+Tree(TDB)
   if (tsdbSaveSmaToDB(pDBFile, smaKey, keyLen, pData, dataLen) != 0) {
@@ -683,9 +676,15 @@ static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t k
  * @param interval
  * @param intervalUnit
  * @param precision
+ * @param adjusted Interval already adjusted according to DB precision
  * @return int64_t
  */
-static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision) {
+static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision, bool adjusted) {
+  
+  if (adjusted) {
+    return interval;
+  }
+
   switch (intervalUnit) {
     case TIME_UNIT_YEAR:  // approximate value
       interval *= 365 * 86400 * 1e3;
@@ -753,59 +752,12 @@ static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit
   return interval;
 }
 
-/**
+static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, const SArray *pDataBlocks, int64_t interval,
- * @brief Split the TSma data blocks into expected size and insert into B+Tree.
+                                  int8_t intervalUnit) {
- *
- * @param pSmaH
- * @param pData
- * @param nOffset The nOffset of blocks since fid changes.
- * @param nBlocks The nBlocks with the same fid since nOffset.
- * @return int32_t
- */
-static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *pData) {
-  STsdb *pTsdb = pSmaH->pTsdb;
-
-  tsdbDebug("tsdbInsertTSmaDataSection: index %" PRIi64 ", skey %" PRIi64, pData->indexUid, pData->skey);
-
-  // TODO: check the data integrity
-
-  int32_t len = 0;
-  while (true) {
-    if (len >= pData->dataLen) {
-      break;
-    }
-    assert(pData->dataLen > 0);
-    STSmaTbData *pTbData = (STSmaTbData *)POINTER_SHIFT(pData->data, len);
-
-    int32_t tbLen = 0;
-    while (true) {
-      if (tbLen >= pTbData->dataLen) {
-        break;
-      }
-      assert(pTbData->dataLen > 0);
-      STSmaColData *pColData = (STSmaColData *)POINTER_SHIFT(pTbData->data, tbLen);
-      char          smaKey[SMA_KEY_LEN] = {0};
-      void         *pSmaKey = &smaKey;
-#if 0
-      printf("tsdbInsertTSmaDataSection: index %" PRIi64 ", skey %" PRIi64 " table[%" PRIi64 "]col[%" PRIu16 "]\n",
-             pData->indexUid, pData->skey, pTbData->tableUid, pColData->colId);
-#endif
-      tsdbEncodeTSmaKey(pTbData->tableUid, pColData->colId, pData->skey, (void **)&pSmaKey);
-      if (tsdbInsertTSmaBlocks(pSmaH, smaKey, SMA_KEY_LEN, pColData->data, pColData->blockSize) < 0) {
-        tsdbWarn("vgId:%d insert tSma blocks failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
-      }
-      tbLen += (sizeof(STSmaColData) + pColData->blockSize);
-    }
-    len += (sizeof(STSmaTbData) + pTbData->dataLen);
-  }
-
-  return TSDB_CODE_SUCCESS;
-}
-
-static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, SSDataBlock *pData, int64_t interval, int8_t intervalUnit) {
   pSmaH->pTsdb = pTsdb;
-  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision);
+  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision, true);
-  pSmaH->pData = pData;
+  pSmaH->pDataBlocks = pDataBlocks;
+  pSmaH->dFile.fid = TSDB_IVLD_FID;
   return TSDB_CODE_SUCCESS;
 }
 
@@ -815,7 +767,7 @@ static void tsdbDestroyTSmaWriteH(STSmaWriteH *pSmaH) {
   }
 }
 
-static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int64_t indexUid, int32_t fid) {
+static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, int64_t indexUid, int32_t fid) {
   STsdb *pTsdb = pSmaH->pTsdb;
   ASSERT(pSmaH->dFile.path == NULL && pSmaH->dFile.pDB == NULL);
 
@@ -859,11 +811,10 @@ static int32_t tsdbGetTSmaDays(STsdb *pTsdb, int64_t interval, int32_t storageLe
  * @param msg
  * @return int32_t
  */
-static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
+static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, int64_t indexUid, const char *msg) {
   STsdbCfg     *pCfg = REPO_CFG(pTsdb);
-  SSDataBlock *pData = (SSDataBlock *)msg;
+  const SArray *pDataBlocks = (const SArray *)msg;
   SSmaEnv      *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
-  int64_t      indexUid = SMA_TEST_INDEX_UID;
 
   if (pEnv == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
@@ -871,15 +822,15 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
     return terrno;
   }
 
-  if (pData == NULL) {
+  if (pDataBlocks == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
-    tsdbWarn("vgId:%d insert tSma data failed since pData is NULL", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert tSma data failed since pDataBlocks is NULL", REPO_ID(pTsdb));
     return terrno;
   }
 
-  if (taosArrayGetSize(pData->pDataBlock) <= 0) {
+  if (taosArrayGetSize(pDataBlocks) <= 0) {
     terrno = TSDB_CODE_INVALID_PARA;
-    tsdbWarn("vgId:%d insert tSma data failed since pDataBlock is empty", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert tSma data failed since pDataBlocks is empty", REPO_ID(pTsdb));
     return TSDB_CODE_FAILED;
   }
 
@@ -899,10 +850,9 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
   }
 
   STSma *pSma = pItem->pSma;
-
   STSmaWriteH tSmaH = {0};
 
-  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData, pSma->interval, pSma->intervalUnit) != 0) {
+  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pDataBlocks, pSma->interval, pSma->intervalUnit) != 0) {
     return TSDB_CODE_FAILED;
   }
 
@@ -921,14 +871,101 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
   int32_t storageLevel = tsdbGetSmaStorageLevel(pSma->interval, pSma->intervalUnit);
   int32_t daysPerFile = tsdbGetTSmaDays(pTsdb, tSmaH.interval, storageLevel);
 
+  // key: skey + groupId
+  char    smaKey[SMA_KEY_LEN] = {0};
+  char    dataBuf[512] = {0};
+  void   *pDataBuf = &dataBuf;
+  int32_t sz = taosArrayGetSize(pDataBlocks);
+  for (int32_t i = 0; i < sz; ++i) {
+    SSDataBlock *pDataBlock = *(SSDataBlock **)taosArrayGet(pDataBlocks, i);
+    int32_t      colNum = pDataBlock->info.numOfCols;
+    int32_t      rows = pDataBlock->info.rows;
+    int32_t      rowSize = pDataBlock->info.rowSize;
+    int64_t      groupId = pDataBlock->info.groupId;
+    for (int32_t j = 0; j < rows; ++j) {
+      printf("|");
+      TSKEY   skey = TSKEY_INITIAL_VAL;  // the start key of TS window by interval
+      void   *pSmaKey = &smaKey;
+      int32_t tlen = 0;
+      for (int32_t k = 0; k < colNum; ++k) {
+        SColumnInfoData *pColInfoData = *(SColumnInfoData **)taosArrayGet(pDataBlock->pDataBlock, k);
+        void            *var = POINTER_SHIFT(pColInfoData->pData, j * pColInfoData->info.bytes);
+        switch (pColInfoData->info.type) {
+          case TSDB_DATA_TYPE_TIMESTAMP:
+            skey = *(TSKEY *)var;
+            printf("==> skey = %" PRIi64 " groupId = %" PRId64 "|", skey, groupId);
+            tsdbEncodeTSmaKey(groupId, skey, &pSmaKey);
+            break;
+          case TSDB_DATA_TYPE_BOOL:
+          case TSDB_DATA_TYPE_UTINYINT:
+            printf(" %15d |", *(uint8_t *)var);
+            tlen += taosEncodeFixedU8(&pDataBuf, *(uint8_t *)var);
+            break;
+          case TSDB_DATA_TYPE_TINYINT:
+            printf(" %15d |", *(int8_t *)var);
+            tlen += taosEncodeFixedI8(&pDataBuf, *(int8_t *)var);
+            break;
+          case TSDB_DATA_TYPE_SMALLINT:
+            printf(" %15d |", *(int16_t *)var);
+            tlen += taosEncodeFixedI16(&pDataBuf, *(int16_t *)var);
+            break;
+          case TSDB_DATA_TYPE_USMALLINT:
+            printf(" %15d |", *(uint16_t *)var);
+            tlen += taosEncodeFixedU16(&pDataBuf, *(uint16_t *)var);
+            break;
+          case TSDB_DATA_TYPE_INT:
+            printf(" %15d |", *(int32_t *)var);
+            tlen += taosEncodeFixedI32(&pDataBuf, *(int32_t *)var);
+            break;
+          case TSDB_DATA_TYPE_FLOAT:
+          case TSDB_DATA_TYPE_UINT:
+            printf(" %15u |", *(uint32_t *)var);
+            tlen += taosEncodeFixedU32(&pDataBuf, *(uint32_t *)var);
+            break;
+          case TSDB_DATA_TYPE_BIGINT:
+            printf(" %15ld |", *(int64_t *)var);
+            tlen += taosEncodeFixedI64(&pDataBuf, *(int64_t *)var);
+            break;
+          case TSDB_DATA_TYPE_DOUBLE:
+          case TSDB_DATA_TYPE_UBIGINT:
+            printf(" %15lu |", *(uint64_t *)var);
+            tlen += taosEncodeFixedU64(&pDataBuf, *(uint64_t *)var);
+            break;
+          case TSDB_DATA_TYPE_NCHAR: {
+            char tmpChar[100] = {0};
+            strncpy(tmpChar, varDataVal(var), varDataLen(var));
+            printf(" %s |", tmpChar);
+            tlen += taosEncodeBinary(&pDataBuf, varDataVal(var), varDataLen(var));
+            break;
+          }
+          case TSDB_DATA_TYPE_VARCHAR: { // TSDB_DATA_TYPE_BINARY
+                      char tmpChar[100] = {0};
+            strncpy(tmpChar, varDataVal(var), varDataLen(var));
+            printf(" %s |", tmpChar);
+            tlen += taosEncodeBinary(&pDataBuf, varDataVal(var), varDataLen(var));
+            break;
+          }
+          case TSDB_DATA_TYPE_VARBINARY:
+            // TODO: add binary/varbinary
+            TASSERT(0);
+          default:
+            printf("the column type %" PRIi16 " is undefined\n", pColInfoData->info.type);
+            TASSERT(0);
+            break;
+        }
+      }
+      if ((tlen > 0) && (skey != TSKEY_INITIAL_VAL)) {
+        int32_t fid = (int32_t)(TSDB_KEY_FID(skey, daysPerFile, pCfg->precision));
 
-#if 0  
+        // Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index
-  int32_t fid = (int32_t)(TSDB_KEY_FID(pData->skey, daysPerFile, pCfg->precision));
+        // file
-
-  // Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index file
         //         - Set and open the DFile or the B+Tree file
         // TODO: tsdbStartTSmaCommit();
-  tsdbSetTSmaDataFile(&tSmaH, pData, indexUid, fid);
+        if (fid != tSmaH.dFile.fid) {
+          if (tSmaH.dFile.fid != TSDB_IVLD_FID) {
+            tsdbCloseDBF(&tSmaH.dFile);
+          }
+          tsdbSetTSmaDataFile(&tSmaH, indexUid, fid);
           if (tsdbOpenDBF(pTsdb->pTSmaEnv->dbEnv, &tSmaH.dFile) != 0) {
             tsdbWarn("vgId:%d open DB file %s failed since %s", REPO_ID(pTsdb),
                      tSmaH.dFile.path ? tSmaH.dFile.path : "path is NULL", tstrerror(terrno));
@@ -936,18 +973,32 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
             tsdbUnRefSmaStat(pTsdb, pStat);
             return TSDB_CODE_FAILED;
           }
+        }
 
-  if (tsdbInsertTSmaDataSection(&tSmaH, pData) != 0) {
+        if (tsdbInsertTSmaBlocks(&tSmaH, &smaKey, SMA_KEY_LEN, pDataBuf, tlen) != 0) {
-    tsdbWarn("vgId:%d insert tSma data section failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
+          tsdbWarn("vgId:%d insert tSma data blocks failed for index %" PRIi64 ", skey %" PRIi64 ", groupId %" PRIi64
+                   " since %s",
+                   REPO_ID(pTsdb), indexUid, skey, groupId, tstrerror(terrno));
           tsdbDestroyTSmaWriteH(&tSmaH);
           tsdbUnRefSmaStat(pTsdb, pStat);
           return TSDB_CODE_FAILED;
+        } else {
+          tsdbWarn("vgId:%d insert tSma data blocks success for index %" PRIi64 ", skey %" PRIi64 ", groupId %" PRIi64,
+                   REPO_ID(pTsdb), indexUid, skey, groupId);
         }
         // TODO:tsdbEndTSmaCommit();
 
         // Step 3: reset the SSmaStat
-  tsdbResetExpiredWindow(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv), pData->indexUid, pData->skey);
+        tsdbResetExpiredWindow(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv), indexUid, skey);
-#endif
+      } else {
+        tsdbWarn("vgId:%d invalid data skey:%" PRIi64 ", tlen %" PRIi32 " during insert tSma data for %" PRIi64,
+                 REPO_ID(pTsdb), skey, tlen, indexUid);
+      }
+
+      printf("\n");
+    }
+  }
+
   tsdbDestroyTSmaWriteH(&tSmaH);
   tsdbUnRefSmaStat(pTsdb, pStat);
   return TSDB_CODE_SUCCESS;
@@ -1002,7 +1053,7 @@ static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid) {
   // TODO:
 }
 
-static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int32_t fid) {
+static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, int32_t fid) {
   STsdb *pTsdb = pSmaH->pTsdb;
 
   char tSmaFile[TSDB_FILENAME_LEN] = {0};
@@ -1012,9 +1063,9 @@ static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData,
   return TSDB_CODE_SUCCESS;
 }
 
-static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
+static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, const char *msg) {
   STsdbCfg     *pCfg = REPO_CFG(pTsdb);
-  SSDataBlock *pData = (SSDataBlock *)msg;
+  const SArray *pDataBlocks = (const SArray *)msg;
   SSmaEnv      *pEnv = atomic_load_ptr(&pTsdb->pRSmaEnv);
   int64_t       indexUid = SMA_TEST_INDEX_UID;
 
@@ -1030,15 +1081,15 @@ static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
     return terrno;
   }
 
-  if (pData == NULL) {
+  if (pDataBlocks == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
-    tsdbWarn("vgId:%d insert rSma data failed since pData is NULL", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert rSma data failed since pDataBlocks is NULL", REPO_ID(pTsdb));
     return terrno;
   }
 
-  if (taosArrayGetSize(pData->pDataBlock) <= 0) {
+  if (taosArrayGetSize(pDataBlocks) <= 0) {
     terrno = TSDB_CODE_INVALID_PARA;
-    tsdbWarn("vgId:%d insert rSma data failed since pDataBlock is empty", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert rSma data failed since pDataBlocks is empty", REPO_ID(pTsdb));
     return TSDB_CODE_FAILED;
   }
 
@@ -1061,7 +1112,7 @@ static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
 
   STSmaWriteH tSmaH = {0};
 
-  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData, pSma->interval, pSma->intervalUnit) != 0) {
+  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pDataBlocks, pSma->interval, pSma->intervalUnit) != 0) {
     return TSDB_CODE_FAILED;
   }
 
@@ -1119,7 +1170,7 @@ static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
  */
 static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, int64_t interval, int8_t intervalUnit) {
   pSmaH->pTsdb = pTsdb;
-  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision);
+  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision, true);
   pSmaH->storageLevel = tsdbGetSmaStorageLevel(interval, intervalUnit);
   pSmaH->days = tsdbGetTSmaDays(pTsdb, pSmaH->interval, pSmaH->storageLevel);
 }
@@ -1185,17 +1236,11 @@ static bool tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, TSKEY *queryKey) {
  * @param pTsdb Return the data between queryWin and fill the pData.
  * @param pData
  * @param indexUid
- * @param interval
- * @param intervalUnit
- * @param tableUid
- * @param colId
  * @param pQuerySKey
  * @param nMaxResult The query invoker should control the nMaxResult need to return to avoid OOM.
  * @return int32_t
  */
-static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval,
+static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, char *pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult) {
-                                   int8_t intervalUnit, tb_uid_t tableUid, col_id_t colId, TSKEY querySKey,
-                                   int32_t nMaxResult) {
   SSmaEnv *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
 
   if (!pEnv) {
@@ -1243,14 +1288,19 @@ static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_
     tsdbDebug("vgId:%d skey %" PRIi64 " of window not in expired window for index %" PRIi64, REPO_ID(pTsdb), querySKey,
               indexUid);
   }
-  tsdbUnRefSmaStat(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv));
+
+  STSma *pTSma = pItem->pSma;
+
+  
 
 #endif
 
   STSmaReadH tReadH = {0};
-  tsdbInitTSmaReadH(&tReadH, pTsdb, interval, intervalUnit);
+  tsdbInitTSmaReadH(&tReadH, pTsdb, pTSma->interval, pTSma->intervalUnit);
   tsdbCloseDBF(&tReadH.dFile);
   
+  tsdbUnRefSmaStat(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv));
+
   tsdbInitTSmaFile(&tReadH, indexUid, querySKey);
   if (tsdbOpenDBF(SMA_ENV_ENV(pTsdb->pTSmaEnv), &tReadH.dFile) != 0) {
     tsdbWarn("vgId:%d open DBF %s failed since %s", REPO_ID(pTsdb), tReadH.dFile.path, tstrerror(terrno));
@@ -1259,19 +1309,17 @@ static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_
 
   char    smaKey[SMA_KEY_LEN] = {0};
   void   *pSmaKey = &smaKey;
-  tsdbEncodeTSmaKey(tableUid, colId, querySKey, (void **)&pSmaKey);
+  int64_t queryGroupId = 1;
+  tsdbEncodeTSmaKey(queryGroupId, querySKey, (void **)&pSmaKey);
 
-  tsdbDebug("vgId:%d get sma data from %s: smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64 ", keyLen %d", REPO_ID(pTsdb),
+  tsdbDebug("vgId:%d get sma data from %s: smaKey %" PRIx64 "-%" PRIx64 ", keyLen %d", REPO_ID(pTsdb),
-            tReadH.dFile.path, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
+            tReadH.dFile.path, *(int64_t *)smaKey, *(int64_t *)POINTER_SHIFT(smaKey, 8), SMA_KEY_LEN);
-            *(int64_t *)POINTER_SHIFT(smaKey, 10), SMA_KEY_LEN);
 
   void    *result = NULL;
   uint32_t valueSize = 0;
   if ((result = tsdbGetSmaDataByKey(&tReadH.dFile, smaKey, SMA_KEY_LEN, &valueSize)) == NULL) {
-    tsdbWarn("vgId:%d get sma data failed from smaIndex %" PRIi64 ", smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64
+    tsdbWarn("vgId:%d get sma data failed from smaIndex %" PRIi64 ", smaKey %" PRIx64 "-%" PRIx64 " since %s",
-             " since %s",
+             REPO_ID(pTsdb), indexUid, *(int64_t *)smaKey, *(int64_t *)POINTER_SHIFT(smaKey, 8), tstrerror(terrno));
-             REPO_ID(pTsdb), indexUid, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
-             *(int64_t *)POINTER_SHIFT(smaKey, 10), tstrerror(terrno));
     tsdbCloseDBF(&tReadH.dFile);
     return TSDB_CODE_FAILED;
   }
@@ -1347,11 +1395,10 @@ int32_t tsdbRemoveTSmaData(STsdb *pTsdb, void *smaIndex, STimeWindow *pWin) {
 }
 #endif
 
-
 // TODO: Who is responsible for resource allocate and release?
-int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg) {
+int32_t tsdbInsertTSmaData(STsdb *pTsdb, int64_t indexUid, const char *msg) {
   int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbInsertTSmaDataImpl(pTsdb, msg)) < 0) {
+  if ((code = tsdbInsertTSmaDataImpl(pTsdb, indexUid, msg)) < 0) {
     tsdbWarn("vgId:%d insert tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
   }
   return code;
@@ -1373,18 +1420,14 @@ int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
   return code;
 }
 
-
+int32_t tsdbGetTSmaData(STsdb *pTsdb, char*pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult) {
-int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
-                        tb_uid_t tableUid, col_id_t colId, TSKEY querySKey, int32_t nMaxResult) {
   int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbGetTSmaDataImpl(pTsdb, pData, indexUid, interval, intervalUnit, tableUid, colId, querySKey,
+  if ((code = tsdbGetTSmaDataImpl(pTsdb, pData, indexUid, querySKey, nMaxResult)) < 0) {
-                                  nMaxResult)) < 0) {
     tsdbWarn("vgId:%d get tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
   }
   return code;
 }
 
-
 int32_t tsdbDropTSmaData(STsdb *pTsdb, int64_t indexUid) {
   int32_t code = TSDB_CODE_SUCCESS;
   if ((code = tsdbDropTSmaDataImpl(pTsdb, indexUid)) < 0) {

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

@@ -41,7 +41,7 @@ int vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg) {
   }
 }
 
-int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg) {
+int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo) {
   vTrace("message in fetch queue is processing");
   char   *msgstr = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
   int32_t msgLen = pMsg->contLen - sizeof(SMsgHead);
@@ -69,9 +69,9 @@ int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg) {
       return tqProcessPollReq(pVnode->pTq, pMsg);
     case TDMT_VND_TASK_PIPE_EXEC:
     case TDMT_VND_TASK_MERGE_EXEC:
-      return tqProcessTaskExec(pVnode->pTq, msgstr, msgLen);
+      return tqProcessTaskExec(pVnode->pTq, msgstr, msgLen, pInfo->workerId);
     case TDMT_VND_STREAM_TRIGGER:
-      return tqProcessStreamTrigger(pVnode->pTq, pMsg->pCont, pMsg->contLen);
+      return tqProcessStreamTrigger(pVnode->pTq, pMsg->pCont, pMsg->contLen, pInfo->workerId);
     case TDMT_VND_QUERY_HEARTBEAT:
       return qWorkerProcessHbMsg(pVnode, pVnode->pQuery, pMsg);
     default:
@@ -205,8 +205,7 @@ _exit:
   rpcMsg.contLen = rspLen;
   rpcMsg.code = code;
 
-  rpcSendResponse(&rpcMsg);
+  tmsgSendRsp(&rpcMsg);
-
   return TSDB_CODE_SUCCESS;
 }
 
@@ -276,8 +275,7 @@ static int32_t vnodeGetTableList(SVnode *pVnode, SRpcMsg *pMsg) {
       .code = 0,
   };
 
-  rpcSendResponse(&rpcMsg);
+  tmsgSendRsp(&rpcMsg);
-
   taosArrayDestroyEx(pArray, freeItemHelper);
   return 0;
 }

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

@@ -17,7 +17,9 @@
 
 void smaHandleRes(void *pVnode, int64_t smaId, const SArray *data) {
   // TODO
+
   blockDebugShowData(data);
+  tsdbInsertTSmaData(((SVnode *)pVnode)->pTsdb, smaId, (const char *)data);
 }
 
 void vnodeProcessWMsgs(SVnode *pVnode, SArray *pMsgs) {
@@ -184,8 +186,8 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       }
     } break;
     case TDMT_VND_TASK_WRITE_EXEC: {
-      if (tqProcessTaskExec(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)),
+      if (tqProcessTaskExec(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), pMsg->contLen - sizeof(SMsgHead),
-                            pMsg->contLen - sizeof(SMsgHead)) < 0) {
+                            0) < 0) {
       }
     } break;
     case TDMT_VND_CREATE_SMA: {  // timeRangeSMA
@@ -194,14 +196,15 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       SSmaCfg vCreateSmaReq = {0};
       if (tDeserializeSVCreateTSmaReq(POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), &vCreateSmaReq) == NULL) {
         terrno = TSDB_CODE_OUT_OF_MEMORY;
-        vWarn("vgId%d: TDMT_VND_CREATE_SMA received but deserialize failed since %s", pVnode->config.vgId, terrstr(terrno));
+        vWarn("vgId%d: TDMT_VND_CREATE_SMA received but deserialize failed since %s", pVnode->config.vgId,
+              terrstr(terrno));
         return -1;
       }
       vWarn("vgId%d: TDMT_VND_CREATE_SMA received for %s:%" PRIi64, pVnode->config.vgId, vCreateSmaReq.tSma.indexName,
             vCreateSmaReq.tSma.indexUid);
 
       // record current timezone of server side
-      tstrncpy(vCreateSmaReq.tSma.timezone, tsTimezoneStr, TD_TIMEZONE_LEN);
+      vCreateSmaReq.tSma.timezoneInt = tsTimezone;
 
       if (metaCreateTSma(pVnode->pMeta, &vCreateSmaReq) < 0) {
         // TODO: handle error

source/dnode/vnode/test/tsdbSmaTest.cpp

@@ -15,6 +15,7 @@
 
 #include <gtest/gtest.h>
 #include <tsdbDef.h>
+
 #include <taoserror.h>
 #include <tglobal.h>
 #include <iostream>
@@ -58,20 +59,21 @@ TEST(testCase, unionEncodeDecodeTest) {
 
   void   *buf = taosMemoryMalloc(1024);
   void   *pBuf = buf;
+  void   *qBuf = buf;
   int32_t tlen = 0;
-  tlen += taosEncodeFixedU8(&buf, sut.info);
+  tlen += taosEncodeFixedU8(&pBuf, sut.info);
-  tlen += taosEncodeFixedI16(&buf, sut.nBSmaCols);
+  tlen += taosEncodeFixedI16(&pBuf, sut.nBSmaCols);
   for (col_id_t i = 0; i < sut.nBSmaCols; ++i) {
-    tlen += taosEncodeFixedI16(&buf, sut.pBSmaCols[i]);
+    tlen += taosEncodeFixedI16(&pBuf, sut.pBSmaCols[i]);
   }
 
   SUnionTest dut = {0};
-  pBuf = taosDecodeFixedU8(pBuf, &dut.info);
+  qBuf = taosDecodeFixedU8(qBuf, &dut.info);
-  pBuf = taosDecodeFixedI16(pBuf, &dut.nBSmaCols);
+  qBuf = taosDecodeFixedI16(qBuf, &dut.nBSmaCols);
   if (dut.nBSmaCols > 0) {
     dut.pBSmaCols = (col_id_t *)taosMemoryMalloc(dut.nBSmaCols * sizeof(col_id_t));
     for (col_id_t i = 0; i < dut.nBSmaCols; ++i) {
-      pBuf = taosDecodeFixedI16(pBuf, dut.pBSmaCols + i);
+      qBuf = taosDecodeFixedI16(qBuf, dut.pBSmaCols + i);
     }
   } else {
     dut.pBSmaCols = NULL;
@@ -80,13 +82,17 @@ TEST(testCase, unionEncodeDecodeTest) {
   printf("sut.rollup=%" PRIu8 ", type=%" PRIu8 ", info=%" PRIu8 "\n", sut.rollup, sut.type, sut.info);
   printf("dut.rollup=%" PRIu8 ", type=%" PRIu8 ", info=%" PRIu8 "\n", dut.rollup, dut.type, dut.info);
 
-  ASSERT_EQ(sut.rollup, dut.rollup);
+  EXPECT_EQ(sut.rollup, dut.rollup);
-  ASSERT_EQ(sut.type, dut.type);
+  EXPECT_EQ(sut.type, dut.type);
-  ASSERT_EQ(sut.nBSmaCols, dut.nBSmaCols);
+  EXPECT_EQ(sut.nBSmaCols, dut.nBSmaCols);
   for (col_id_t i = 0; i < sut.nBSmaCols; ++i) {
-    ASSERT_EQ(*(col_id_t *)(sut.pBSmaCols + i), sut.pBSmaCols[i]);
+    EXPECT_EQ(*(col_id_t *)(sut.pBSmaCols + i), sut.pBSmaCols[i]);
-    ASSERT_EQ(*(col_id_t *)(sut.pBSmaCols + i), dut.pBSmaCols[i]);
+    EXPECT_EQ(*(col_id_t *)(sut.pBSmaCols + i), dut.pBSmaCols[i]);
   }
+
+  taosMemoryFreeClear(buf);
+  taosMemoryFreeClear(dut.pBSmaCols);
+  taosMemoryFreeClear(sut.pBSmaCols);
 }
 #if 1
 TEST(testCase, tSma_Meta_Encode_Decode_Test) {
@@ -106,37 +112,37 @@ TEST(testCase, tSma_Meta_Encode_Decode_Test) {
   uint32_t     bufLen = tEncodeTSmaWrapper(NULL, &tSmaWrapper);
 
   void *buf = taosMemoryCalloc(1, bufLen);
-  ASSERT_NE(buf, nullptr);
+  EXPECT_NE(buf, nullptr);
 
   STSmaWrapper *pSW = (STSmaWrapper *)buf;
   uint32_t      len = tEncodeTSmaWrapper(&buf, &tSmaWrapper);
 
-  ASSERT_EQ(len, bufLen);
+  EXPECT_EQ(len, bufLen);
 
   // decode
   STSmaWrapper dstTSmaWrapper = {0};
   void        *result = tDecodeTSmaWrapper(pSW, &dstTSmaWrapper);
-  ASSERT_NE(result, nullptr);
+  EXPECT_NE(result, nullptr);
 
-  ASSERT_EQ(tSmaWrapper.number, dstTSmaWrapper.number);
+  EXPECT_EQ(tSmaWrapper.number, dstTSmaWrapper.number);
 
   for (int i = 0; i < tSmaWrapper.number; ++i) {
     STSma *pSma = tSmaWrapper.tSma + i;
     STSma *qSma = dstTSmaWrapper.tSma + i;
 
-    ASSERT_EQ(pSma->version, qSma->version);
+    EXPECT_EQ(pSma->version, qSma->version);
-    ASSERT_EQ(pSma->intervalUnit, qSma->intervalUnit);
+    EXPECT_EQ(pSma->intervalUnit, qSma->intervalUnit);
-    ASSERT_EQ(pSma->slidingUnit, qSma->slidingUnit);
+    EXPECT_EQ(pSma->slidingUnit, qSma->slidingUnit);
-    ASSERT_STRCASEEQ(pSma->indexName, qSma->indexName);
+    EXPECT_STRCASEEQ(pSma->indexName, qSma->indexName);
-    ASSERT_EQ(pSma->timezoneInt, qSma->timezoneInt);
+    EXPECT_EQ(pSma->timezoneInt, qSma->timezoneInt);
-    ASSERT_EQ(pSma->indexUid, qSma->indexUid);
+    EXPECT_EQ(pSma->indexUid, qSma->indexUid);
-    ASSERT_EQ(pSma->tableUid, qSma->tableUid);
+    EXPECT_EQ(pSma->tableUid, qSma->tableUid);
-    ASSERT_EQ(pSma->interval, qSma->interval);
+    EXPECT_EQ(pSma->interval, qSma->interval);
-    ASSERT_EQ(pSma->sliding, qSma->sliding);
+    EXPECT_EQ(pSma->sliding, qSma->sliding);
-    ASSERT_EQ(pSma->exprLen, qSma->exprLen);
+    EXPECT_EQ(pSma->exprLen, qSma->exprLen);
-    ASSERT_STRCASEEQ(pSma->expr, qSma->expr);
+    EXPECT_STRCASEEQ(pSma->expr, qSma->expr);
-    ASSERT_EQ(pSma->tagsFilterLen, qSma->tagsFilterLen);
+    EXPECT_EQ(pSma->tagsFilterLen, qSma->tagsFilterLen);
-    ASSERT_STRCASEEQ(pSma->tagsFilter, qSma->tagsFilter);
+    EXPECT_STRCASEEQ(pSma->tagsFilter, qSma->tagsFilter);
   }
 
   // resource release
@@ -172,12 +178,12 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
 
   tSma.exprLen = strlen(expr);
   tSma.expr = (char *)taosMemoryCalloc(1, tSma.exprLen + 1);
-  ASSERT_NE(tSma.expr, nullptr);
+  EXPECT_NE(tSma.expr, nullptr);
   tstrncpy(tSma.expr, expr, tSma.exprLen + 1);
 
   tSma.tagsFilterLen = strlen(tagsFilter);
   tSma.tagsFilter = (char *)taosMemoryCalloc(tSma.tagsFilterLen + 1, 1);
-  ASSERT_NE(tSma.tagsFilter, nullptr);
+  EXPECT_NE(tSma.tagsFilter, nullptr);
   tstrncpy(tSma.tagsFilter, tagsFilter, tSma.tagsFilterLen + 1);
 
   SMeta          *pMeta = NULL;
@@ -189,7 +195,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   pMeta = metaOpen(smaTestDir, pMetaCfg, NULL);
   assert(pMeta != NULL);
   // save index 1
-  ASSERT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
+  EXPECT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
 
   pSmaCfg->indexUid = indexUid2;
   tstrncpy(pSmaCfg->indexName, smaIndexName2, TSDB_INDEX_NAME_LEN);
@@ -200,7 +206,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   pSmaCfg->sliding = 5;
 
   // save index 2
-  ASSERT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
+  EXPECT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
 
   // get value by indexName
   STSma *qSmaCfg = NULL;
@@ -210,8 +216,8 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   printf("timezone1 = %" PRIi8 "\n", qSmaCfg->timezoneInt);
   printf("expr1 = %s\n", qSmaCfg->expr != NULL ? qSmaCfg->expr : "");
   printf("tagsFilter1 = %s\n", qSmaCfg->tagsFilter != NULL ? qSmaCfg->tagsFilter : "");
-  ASSERT_STRCASEEQ(qSmaCfg->indexName, smaIndexName1);
+  EXPECT_STRCASEEQ(qSmaCfg->indexName, smaIndexName1);
-  ASSERT_EQ(qSmaCfg->tableUid, tSma.tableUid);
+  EXPECT_EQ(qSmaCfg->tableUid, tSma.tableUid);
   tdDestroyTSma(qSmaCfg);
   taosMemoryFreeClear(qSmaCfg);
 
@@ -221,8 +227,8 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   printf("timezone2 = %" PRIi8 "\n", qSmaCfg->timezoneInt);
   printf("expr2 = %s\n", qSmaCfg->expr != NULL ? qSmaCfg->expr : "");
   printf("tagsFilter2 = %s\n", qSmaCfg->tagsFilter != NULL ? qSmaCfg->tagsFilter : "");
-  ASSERT_STRCASEEQ(qSmaCfg->indexName, smaIndexName2);
+  EXPECT_STRCASEEQ(qSmaCfg->indexName, smaIndexName2);
-  ASSERT_EQ(qSmaCfg->interval, tSma.interval);
+  EXPECT_EQ(qSmaCfg->interval, tSma.interval);
   tdDestroyTSma(qSmaCfg);
   taosMemoryFreeClear(qSmaCfg);
 
@@ -238,25 +244,25 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
     printf("indexName = %s\n", indexName);
     ++indexCnt;
   }
-  ASSERT_EQ(indexCnt, nCntTSma);
+  EXPECT_EQ(indexCnt, nCntTSma);
   metaCloseSmaCurosr(pSmaCur);
 
   // get wrapper by table uid
   STSmaWrapper *pSW = metaGetSmaInfoByTable(pMeta, tbUid);
   assert(pSW != NULL);
-  ASSERT_EQ(pSW->number, nCntTSma);
+  EXPECT_EQ(pSW->number, nCntTSma);
-  ASSERT_STRCASEEQ(pSW->tSma->indexName, smaIndexName1);
+  EXPECT_STRCASEEQ(pSW->tSma->indexName, smaIndexName1);
-  ASSERT_EQ(pSW->tSma->timezoneInt, timezone);
+  EXPECT_EQ(pSW->tSma->timezoneInt, timezone);
-  ASSERT_STRCASEEQ(pSW->tSma->expr, expr);
+  EXPECT_STRCASEEQ(pSW->tSma->expr, expr);
-  ASSERT_STRCASEEQ(pSW->tSma->tagsFilter, tagsFilter);
+  EXPECT_STRCASEEQ(pSW->tSma->tagsFilter, tagsFilter);
-  ASSERT_EQ(pSW->tSma->indexUid, indexUid1);
+  EXPECT_EQ(pSW->tSma->indexUid, indexUid1);
-  ASSERT_EQ(pSW->tSma->tableUid, tbUid);
+  EXPECT_EQ(pSW->tSma->tableUid, tbUid);
-  ASSERT_STRCASEEQ((pSW->tSma + 1)->indexName, smaIndexName2);
+  EXPECT_STRCASEEQ((pSW->tSma + 1)->indexName, smaIndexName2);
-  ASSERT_EQ((pSW->tSma + 1)->timezoneInt, timezone);
+  EXPECT_EQ((pSW->tSma + 1)->timezoneInt, timezone);
-  ASSERT_STRCASEEQ((pSW->tSma + 1)->expr, expr);
+  EXPECT_STRCASEEQ((pSW->tSma + 1)->expr, expr);
-  ASSERT_STRCASEEQ((pSW->tSma + 1)->tagsFilter, tagsFilter);
+  EXPECT_STRCASEEQ((pSW->tSma + 1)->tagsFilter, tagsFilter);
-  ASSERT_EQ((pSW->tSma + 1)->indexUid, indexUid2);
+  EXPECT_EQ((pSW->tSma + 1)->indexUid, indexUid2);
-  ASSERT_EQ((pSW->tSma + 1)->tableUid, tbUid);
+  EXPECT_EQ((pSW->tSma + 1)->tableUid, tbUid);
 
   tdDestroyTSmaWrapper(pSW);
   taosMemoryFreeClear(pSW);
@@ -268,7 +274,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
     printf("metaGetSmaTbUids: uid[%" PRIu32 "] = %" PRIi64 "\n", i, *(tb_uid_t *)taosArrayGet(pUids, i));
     // printf("metaGetSmaTbUids: index[%" PRIu32 "] = %s", i, (char *)taosArrayGet(pUids, i));
   }
-  ASSERT_EQ(taosArrayGetSize(pUids), 1);
+  EXPECT_EQ(taosArrayGetSize(pUids), 1);
   taosArrayDestroy(pUids);
 
   // resource release
@@ -280,7 +286,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
 }
 #endif
 
-#if 0
+#if 1
 TEST(testCase, tSma_Data_Insert_Query_Test) {
   // step 1: prepare meta
   const char    *smaIndexName1 = "sma_index_test_1";
@@ -299,9 +305,9 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   // encode
   STSma tSma = {0};
   tSma.version = 0;
-  tSma.intervalUnit = TIME_UNIT_DAY;
+  tSma.intervalUnit = TIME_UNIT_MINUTE;
   tSma.interval = 1;
-  tSma.slidingUnit = TIME_UNIT_HOUR;
+  tSma.slidingUnit = TIME_UNIT_MINUTE;
   tSma.sliding = 1;  // sliding = interval when it's convert window
   tSma.indexUid = indexUid1;
   tstrncpy(tSma.indexName, smaIndexName1, TSDB_INDEX_NAME_LEN);
@@ -310,12 +316,12 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
 
   tSma.exprLen = strlen(expr);
   tSma.expr = (char *)taosMemoryCalloc(1, tSma.exprLen + 1);
-  ASSERT_NE(tSma.expr, nullptr);
+  EXPECT_NE(tSma.expr, nullptr);
   tstrncpy(tSma.expr, expr, tSma.exprLen + 1);
 
   tSma.tagsFilterLen = strlen(tagsFilter);
   tSma.tagsFilter = (char *)taosMemoryCalloc(1, tSma.tagsFilterLen + 1);
-  ASSERT_NE(tSma.tagsFilter, nullptr);
+  EXPECT_NE(tSma.tagsFilter, nullptr);
   tstrncpy(tSma.tagsFilter, tagsFilter, tSma.tagsFilterLen + 1);
 
   SMeta          *pMeta = NULL;
@@ -327,7 +333,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   pMeta = metaOpen(smaTestDir, pMetaCfg, NULL);
   assert(pMeta != NULL);
   // save index 1
-  ASSERT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
+  EXPECT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
 
   // step 2: insert data
   STsdb    *pTsdb = (STsdb *)taosMemoryCalloc(1, sizeof(STsdb));
@@ -364,7 +370,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   strncpy(pDisks.dir, "/var/lib/taos", TSDB_FILENAME_LEN);
   int32_t numOfDisks = 1;
   pTsdb->pTfs = tfsOpen(&pDisks, numOfDisks);
-  ASSERT_NE(pTsdb->pTfs, nullptr);
+  EXPECT_NE(pTsdb->pTfs, nullptr);
 
   // generate SSubmitReq msg and update expired window
   int16_t  schemaVer = 0;
@@ -374,7 +380,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   uint32_t msgLen = sizeof(SSubmitReq) + mockBlkNum * sizeof(SSubmitBlk) + mockBlkNum * mockRowNum * mockRowLen;
 
   SSubmitReq *pMsg = (SSubmitReq *)taosMemoryCalloc(1, msgLen);
-  ASSERT_NE(pMsg, nullptr);
+  EXPECT_NE(pMsg, nullptr);
   pMsg->version = htobe64(schemaVer);
   pMsg->numOfBlocks = htonl(mockBlkNum);
   pMsg->length = htonl(msgLen);
@@ -400,20 +406,99 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
     }
   }
 
-  ASSERT_EQ(tdScanAndConvertSubmitMsg(pMsg), TSDB_CODE_SUCCESS);
+  EXPECT_EQ(tdScanAndConvertSubmitMsg(pMsg), TSDB_CODE_SUCCESS);
 
-  ASSERT_EQ(tsdbUpdateSmaWindow(pTsdb, (const char *)pMsg), 0);
+  EXPECT_EQ(tsdbUpdateSmaWindow(pTsdb, (const char *)pMsg), 0);
 
   // init
-  int32_t allocCnt = 0;
+  const int32_t  tSmaGroupSize = 4;
-  int32_t allocStep = 16384;
+  const int32_t  tSmaNumOfTags = 2;
-  int32_t buffer = 1024;
+  const int64_t  tSmaGroupId = 12345670;
-  void   *buf = NULL;
+  const col_id_t tSmaNumOfCols = 9;  // binary/nchar/varbinary/varchar are only used for tags for group by conditions.
-  ASSERT_EQ(tsdbMakeRoom(&buf, allocStep), 0);
+  const int32_t  tSmaNumOfRows = 2;
-  int32_t  bufSize = taosTSizeof(buf);
+
-  int32_t  numOfTables = 10;
+  SArray *pDataBlocks = taosArrayInit(tSmaGroupSize, sizeof(SSDataBlock *));
-  col_id_t numOfCols = 4096;
+  EXPECT_NE(pDataBlocks, nullptr);
-  ASSERT_GT(numOfCols, 0);
+  int32_t tSmaTypeArray[tSmaNumOfCols] = {TSDB_DATA_TYPE_TIMESTAMP, TSDB_DATA_TYPE_BOOL,     TSDB_DATA_TYPE_INT,
+                                          TSDB_DATA_TYPE_UBIGINT,   TSDB_DATA_TYPE_SMALLINT, TSDB_DATA_TYPE_FLOAT,
+                                          TSDB_DATA_TYPE_DOUBLE,    TSDB_DATA_TYPE_VARCHAR,  TSDB_DATA_TYPE_NCHAR};
+  // last 2 columns for group by tags
+  // int32_t tSmaTypeArray[tSmaNumOfCols] = {TSDB_DATA_TYPE_TIMESTAMP, TSDB_DATA_TYPE_BOOL};
+  const char *tSmaGroupbyTags[tSmaGroupSize * tSmaNumOfTags] = {"BeiJing",  "HaiDian", "BeiJing",   "ChaoYang",
+                                                                "ShangHai", "PuDong",  "ShangHai", "MinHang"};
+  TSKEY       tSmaSKeyMs = (int64_t)1648535332 * 1000;
+  int64_t     tSmaIntervalMs = tSma.interval * 60 * 1000;
+  int64_t     tSmaInitVal = 0;
+
+  for (int32_t g = 0; g < tSmaGroupSize; ++g) {
+    SSDataBlock *pDataBlock = (SSDataBlock *)taosMemoryCalloc(1, sizeof(SSDataBlock));
+    EXPECT_NE(pDataBlock, nullptr);
+    pDataBlock->pBlockAgg = NULL;
+    pDataBlock->info.numOfCols = tSmaNumOfCols;
+    pDataBlock->info.rows = tSmaNumOfRows;
+    pDataBlock->info.groupId = tSmaGroupId + g;
+
+    pDataBlock->pDataBlock = taosArrayInit(tSmaNumOfCols, sizeof(SColumnInfoData *));
+    EXPECT_NE(pDataBlock->pDataBlock, nullptr);
+    for (int32_t c = 0; c < tSmaNumOfCols; ++c) {
+      
+      SColumnInfoData *pColInfoData = (SColumnInfoData *)taosMemoryCalloc(1, sizeof(SColumnInfoData));
+      EXPECT_NE(pColInfoData, nullptr);
+
+      pColInfoData->info.type = tSmaTypeArray[c];
+      if (IS_VAR_DATA_TYPE(pColInfoData->info.type)) {
+        pColInfoData->info.bytes = 100;  // update accordingly
+      } else {
+        pColInfoData->info.bytes = TYPE_BYTES[pColInfoData->info.type];
+      }
+      pColInfoData->pData = (char *)taosMemoryCalloc(1, tSmaNumOfRows * pColInfoData->info.bytes);
+
+      for (int32_t r = 0; r < tSmaNumOfRows; ++r) {
+        void *pCellData = pColInfoData->pData + r * pColInfoData->info.bytes;
+        switch (pColInfoData->info.type) {
+          case TSDB_DATA_TYPE_TIMESTAMP:
+            *(TSKEY *)pCellData = tSmaSKeyMs + tSmaIntervalMs * r;
+            break;
+          case TSDB_DATA_TYPE_BOOL:
+            *(bool *)pCellData = (bool)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_INT:
+            *(int *)pCellData = (int)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_UBIGINT:
+            *(uint64_t *)pCellData = (uint64_t)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_SMALLINT:
+            *(int16_t *)pCellData = (int16_t)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_FLOAT:
+            *(float *)pCellData = (float)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_DOUBLE:
+            *(double *)pCellData = (double)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_VARCHAR:  // city
+            varDataSetLen(pCellData, strlen(tSmaGroupbyTags[g * 2]));
+            memcpy(varDataVal(pCellData), tSmaGroupbyTags[g * 2], varDataLen(pCellData));
+            break;
+          case TSDB_DATA_TYPE_NCHAR:  // district
+            varDataSetLen(pCellData, strlen(tSmaGroupbyTags[g * 2 + 1]));
+            memcpy(varDataVal(pCellData), tSmaGroupbyTags[g * 2 + 1], varDataLen(pCellData));
+            break;
+          default:
+            EXPECT_EQ(0, 1);  // add definition
+            break;
+        }
+      }
+      // push SColumnInfoData
+      taosArrayPush(pDataBlock->pDataBlock, &pColInfoData);
+    }
+    // push SSDataBlock
+    taosArrayPush(pDataBlocks, &pDataBlock);
+  }
+
+  // execute
+  EXPECT_EQ(tsdbInsertTSmaData(pTsdb, tSma.indexUid, (const char *)pDataBlocks), TSDB_CODE_SUCCESS);
 
 #if 0
   STSmaDataWrapper *pSmaData = NULL;
@@ -432,7 +517,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
     int32_t tableDataLen = sizeof(STSmaTbData);
     for (col_id_t c = 0; c < numOfCols; ++c) {
       if (bufSize - len - tableDataLen < buffer) {
-        ASSERT_EQ(tsdbMakeRoom(&buf, bufSize + allocStep), 0);
+        EXPECT_EQ(tsdbMakeRoom(&buf, bufSize + allocStep), 0);
         pSmaData = (STSmaDataWrapper *)buf;
         pTbData = (STSmaTbData *)POINTER_SHIFT(pSmaData, len);
         bufSize = taosTSizeof(buf);
@@ -459,31 +544,36 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   }
   pSmaData->dataLen = (len - sizeof(STSmaDataWrapper));
 
-  ASSERT_GE(bufSize, pSmaData->dataLen);
+  EXPECT_GE(bufSize, pSmaData->dataLen);
   // execute
-  ASSERT_EQ(tsdbInsertTSmaData(pTsdb, (char *)pSmaData), TSDB_CODE_SUCCESS);
+  EXPECT_EQ(tsdbInsertTSmaData(pTsdb, (char *)pSmaData), TSDB_CODE_SUCCESS);
 #endif
 
-  SSDataBlock *pSmaData = (SSDataBlock *)taosMemoryCalloc(1, sizeof(SSDataBlock));
-  
-
-
   // step 3: query
   uint32_t checkDataCnt = 0;
-  for (int32_t t = 0; t < numOfTables; ++t) {
+  EXPECT_EQ(tsdbGetTSmaData(pTsdb, NULL, indexUid1, skey1, 1), TSDB_CODE_SUCCESS);
-    for (col_id_t c = 0; c < numOfCols; ++c) {
-      ASSERT_EQ(tsdbGetTSmaData(pTsdb, NULL, indexUid1, interval1, intervalUnit1, tbUid + t,
-                                c + PRIMARYKEY_TIMESTAMP_COL_ID, skey1, 1),
-                TSDB_CODE_SUCCESS);
   ++checkDataCnt;
-    }
-  }
 
   printf("%s:%d The sma data check count for insert and query is %" PRIu32 "\n", __FILE__, __LINE__, checkDataCnt);
 
   // release data
   taosMemoryFreeClear(pMsg);
-  taosTZfree(buf);
+
+  for (int32_t i = 0; i < taosArrayGetSize(pDataBlocks); ++i) {
+    SSDataBlock *pDataBlock = *(SSDataBlock **)taosArrayGet(pDataBlocks, i);
+    int32_t      numOfOutput = taosArrayGetSize(pDataBlock->pDataBlock);
+    for (int32_t j = 0; j < numOfOutput; ++j) {
+      SColumnInfoData *pColInfoData = *(SColumnInfoData **)taosArrayGet(pDataBlock->pDataBlock, j);
+      colDataDestroy(pColInfoData);
+      taosMemoryFreeClear(pColInfoData);
+    }
+
+    taosArrayDestroy(pDataBlock->pDataBlock);
+    taosMemoryFreeClear(pDataBlock->pBlockAgg);
+    taosMemoryFreeClear(pDataBlock);
+  }
+  taosArrayDestroy(pDataBlocks);
+
   // release meta
   tdDestroyTSma(&tSma);
   tfsClose(pTsdb->pTfs);

source/libs/executor/inc/executorimpl.h

@@ -161,20 +161,8 @@ typedef struct STaskCostInfo {
 typedef struct SOperatorCostInfo {
   uint64_t openCost;
   uint64_t execCost;
-//  uint64_t totalRows;
-//  uint64_t totalBytes;
 } SOperatorCostInfo;
 
-typedef struct {
-  int64_t vgroupLimit;
-  int64_t ts;
-} SOrderedPrjQueryInfo;
-
-typedef struct {
-  char*   tags;
-  SArray* pResult;  // SArray<SStddevInterResult>
-} SInterResult;
-
 // The basic query information extracted from the SQueryInfo tree to support the
 // execution of query in a data node.
 typedef struct STaskAttr {
@@ -230,7 +218,6 @@ typedef struct STaskAttr {
   SColumnInfo*         tagColList;
   int32_t              numOfFilterCols;
   int64_t*             fillVal;
-  SOrderedPrjQueryInfo prjInfo;  // limit value for each vgroup, only available in global order projection query.
 
   SSingleColumnFilterInfo* pFilterInfo;
   //  SFilterInfo     *pFilters;
@@ -245,8 +232,9 @@ struct SOperatorInfo;
 
 typedef void (*__optr_encode_fn_t)(struct SOperatorInfo* pOperator, char **result, int32_t *length);
 typedef bool (*__optr_decode_fn_t)(struct SOperatorInfo* pOperator, char *result, int32_t length);
-typedef int32_t (*__optr_open_fn_t)(struct SOperatorInfo* param);
+
-typedef SSDataBlock* (*__optr_fn_t)(struct SOperatorInfo* param, bool* newgroup);
+typedef int32_t (*__optr_open_fn_t)(struct SOperatorInfo* pOptr);
+typedef SSDataBlock* (*__optr_fn_t)(struct SOperatorInfo* pOptr, bool* newgroup);
 typedef void (*__optr_close_fn_t)(void* param, int32_t num);
 
 typedef struct STaskIdInfo {
@@ -267,7 +255,8 @@ typedef struct SExecTaskInfo {
   uint64_t        totalRows;            // total number of rows
   STableGroupInfo tableqinfoGroupInfo;  // this is a group array list, including SArray<STableQueryInfo*> structure
   char*           sql;                  // query sql string
-  jmp_buf         env;                  //
+  jmp_buf         env;                  // jump to this position when error happens.
+  EOPTR_EXEC_MODEL execModel;            // operator execution model [batch model|stream model]
   struct SOperatorInfo* pRoot;
 } SExecTaskInfo;
 
@@ -330,11 +319,12 @@ typedef struct SOperatorInfo {
   SResultInfo            resultInfo;
   struct SOperatorInfo** pDownstream;      // downstram pointer list
   int32_t                numOfDownstream;  // number of downstream. The value is always ONE expect for join operator
-  __optr_fn_t            getNextFn;
-  __optr_fn_t            cleanupFn;
-  __optr_close_fn_t      closeFn;
   __optr_open_fn_t       _openFn;          // DO NOT invoke this function directly
-  __optr_encode_fn_t     encodeResultRow;           //
+  __optr_fn_t            getNextFn;
+  __optr_fn_t            getStreamResFn;   // execute the aggregate in the stream model.
+  __optr_fn_t            cleanupFn;        // call this function to release the allocated resources ASAP
+  __optr_close_fn_t      closeFn;
+  __optr_encode_fn_t     encodeResultRow;
   __optr_decode_fn_t     decodeResultRow;
 } SOperatorInfo;
 
@@ -363,18 +353,18 @@ typedef struct SQInfo {
   STaskCostInfo   summary;
 } SQInfo;
 
-enum {
+typedef enum {
-  DATA_NOT_READY = 0x1,
+  EX_SOURCE_DATA_NOT_READY = 0x1,
-  DATA_READY     = 0x2,
+  EX_SOURCE_DATA_READY     = 0x2,
-  DATA_EXHAUSTED = 0x3,
+  EX_SOURCE_DATA_EXHAUSTED = 0x3,
-};
+} EX_SOURCE_STATUS;
 
 typedef struct SSourceDataInfo {
   struct SExchangeInfo *pEx;
   int32_t               index;
   SRetrieveTableRsp    *pRsp;
   uint64_t              totalRows;
-  int32_t            status;
+  EX_SOURCE_STATUS      status;
 } SSourceDataInfo;
 
 typedef struct SLoadRemoteDataInfo {
@@ -383,12 +373,6 @@ typedef struct SLoadRemoteDataInfo {
   uint64_t           totalElapsed;  // total elapsed time
 } SLoadRemoteDataInfo;
 
-enum {
-  EX_SOURCE_DATA_NOT_READY = 0x1,
-  EX_SOURCE_DATA_READY     = 0x2,
-  EX_SOURCE_DATA_EXHAUSTED = 0x3,
-};
-
 typedef struct SExchangeInfo {
   SArray*            pSources;
   SArray*            pSourceDataInfo;
@@ -484,16 +468,18 @@ typedef struct SAggSupporter {
 } SAggSupporter;
 
 typedef struct STableIntervalOperatorInfo {
-  SOptrBasicInfo     binfo;
+  SOptrBasicInfo     binfo;                // basic info
-  SGroupResInfo      groupResInfo;
+  SGroupResInfo      groupResInfo;         // multiple results build supporter
-  SInterval          interval;
+  SInterval          interval;             // interval info
-  STimeWindow        win;
+  STimeWindow        win;                  // query time range
-  int32_t            precision;
+  bool               timeWindowInterpo;    // interpolation needed or not
-  bool               timeWindowInterpo;
+  char             **pRow;                 // previous row/tuple of already processed datablock
-  char             **pRow;
+  SAggSupporter      aggSup;               // aggregate supporter
-  SAggSupporter      aggSup;
+  STableQueryInfo   *pCurrent;             // current tableQueryInfo struct
-  STableQueryInfo   *pCurrent;
+  int32_t            order;                // current SSDataBlock scan order
-  int32_t            order;
+  EOPTR_EXEC_MODEL    execModel;            // operator execution model [batch model|stream model]
+  SArray            *pUpdatedWindow;       // updated time window due to the input data block from the downstream operator.
+  SColumnInfoData    timeWindowData;       // query time window info for scalar function execution.
 } STableIntervalOperatorInfo;
 
 typedef struct SAggOperatorInfo {
@@ -695,12 +681,6 @@ SOperatorInfo* createSLimitOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorI
 SOperatorInfo* createJoinOperatorInfo(SOperatorInfo** pdownstream, int32_t numOfDownstream, SSchema* pSchema,
                                       int32_t numOfOutput);
 
-void doSetFilterColumnInfo(SSingleColumnFilterInfo* pFilterInfo, int32_t numOfFilterCols, SSDataBlock* pBlock);
-bool doFilterDataBlock(SSingleColumnFilterInfo* pFilterInfo, int32_t numOfFilterCols, int32_t numOfRows, int8_t* p);
-void doCompactSDataBlock(SSDataBlock* pBlock, int32_t numOfRows, int8_t* p);
-
-SSDataBlock* createOutputBuf(SExprInfo* pExpr, int32_t numOfOutput, int32_t numOfRows);
-
 void* doDestroyFilterInfo(SSingleColumnFilterInfo* pFilterInfo, int32_t numOfFilterCols);
 
 void setInputDataBlock(SOperatorInfo* pOperator, SqlFunctionCtx* pCtx, SSDataBlock* pBlock, int32_t order);
@@ -734,7 +714,7 @@ int32_t getMaximumIdleDurationSec();
 
 void    doInvokeUdf(struct SUdfInfo* pUdfInfo, SqlFunctionCtx* pCtx, int32_t idx, int32_t type);
 void    setTaskStatus(SExecTaskInfo* pTaskInfo, int8_t status);
-int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId);
+int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId, EOPTR_EXEC_MODEL model);
 
 #ifdef __cplusplus
 }