Merge pull request #9449 from taosdata/feature/qnode

Feature/qnode

include/common/tmsg.h

@@ -1018,7 +1018,7 @@ typedef struct {
 } SUpdateTagValRsp;
 
 typedef struct SSubQueryMsg {
-  uint64_t schedulerId;
+  uint64_t sId;
   uint64_t queryId;
   uint64_t taskId;
   uint32_t contentLen;
@@ -1026,7 +1026,7 @@ typedef struct SSubQueryMsg {
 } SSubQueryMsg;
 
 typedef struct SResReadyMsg {
-  uint64_t schedulerId;
+  uint64_t sId;
   uint64_t queryId;
   uint64_t taskId;
 } SResReadyMsg;
@@ -1036,13 +1036,13 @@ typedef struct SResReadyRsp {
 } SResReadyRsp;
 
 typedef struct SResFetchMsg {
-  uint64_t schedulerId;
+  uint64_t sId;
   uint64_t queryId;
   uint64_t taskId;
 } SResFetchMsg;
 
 typedef struct SSchTasksStatusMsg {
-  uint64_t schedulerId;
+  uint64_t sId;
 } SSchTasksStatusMsg;
 
 typedef struct STaskStatus {
@@ -1057,7 +1057,7 @@ typedef struct SSchedulerStatusRsp {
 } SSchedulerStatusRsp;
 
 typedef struct STaskCancelMsg {
-  uint64_t schedulerId;
+  uint64_t sId;
   uint64_t queryId;
   uint64_t taskId;
 } STaskCancelMsg;
@@ -1067,7 +1067,7 @@ typedef struct STaskCancelRsp {
 } STaskCancelRsp;
 
 typedef struct STaskDropMsg {
-  uint64_t schedulerId;
+  uint64_t sId;
   uint64_t queryId;
   uint64_t taskId;
 } STaskDropMsg;

source/client/src/clientEnv.c

@@ -227,7 +227,7 @@ void taos_init_imp(void) {
 
   rpcInit();
 
-  SCatalogCfg cfg = {.enableVgroupCache = true, .maxDBCacheNum = 100, .maxTblCacheNum = 100};
+  SCatalogCfg cfg = {.maxDBCacheNum = 100, .maxTblCacheNum = 100};
   catalogInit(&cfg);
 
   tscDebug("starting to initialize TAOS driver, local ep: %s", tsLocalEp);

source/dnode/vnode/impl/inc/vnodeDef.h

@@ -73,7 +73,7 @@ struct SVnode {
   SVnodeSync* pSync;
   SVnodeFS*   pFs;
   tsem_t      canCommit;
-  void*       pQuery;
+  SQHandle*   pQuery;
 };
 
 int vnodeScheduleTask(SVnodeTask* task);

source/dnode/vnode/impl/inc/vnodeQuery.h

@@ -22,6 +22,9 @@ extern "C" {
 #include "vnodeInt.h"
 #include "qworker.h"
 
+typedef struct SQWorkerMgmt SQHandle;
+
+
 int vnodeQueryOpen(SVnode *pVnode);
 
 #ifdef __cplusplus

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

@@ -22,13 +22,27 @@ int vnodeQueryOpen(SVnode *pVnode) {
 
 int vnodeProcessQueryReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
   vInfo("query message is processed");
-  qWorkerProcessQueryMsg(pVnode, pVnode->pQuery, pMsg);
-  return 0;
+  return qWorkerProcessQueryMsg(pVnode, pVnode->pQuery, pMsg);
 }
 
 int vnodeProcessFetchReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
   vInfo("fetch message is processed");
-  qWorkerProcessFetchMsg(pVnode, pVnode->pQuery, pMsg);
+  switch (pMsg->msgType) {
+    case TDMT_VND_FETCH:
+      return qWorkerProcessFetchMsg(pVnode, pVnode->pQuery, pMsg);
+    case TDMT_VND_RES_READY:
+      return qWorkerProcessReadyMsg(pVnode, pVnode->pQuery, pMsg);
+    case TDMT_VND_TASKS_STATUS:
+      return qWorkerProcessStatusMsg(pVnode, pVnode->pQuery, pMsg);
+    case TDMT_VND_CANCEL_TASK:
+      return qWorkerProcessCancelMsg(pVnode, pVnode->pQuery, pMsg);
+    case TDMT_VND_DROP_TASK:
+      return qWorkerProcessDropMsg(pVnode, pVnode->pQuery, pMsg);
+    default:
+      vError("unknown msg type:%d in fetch queue", pMsg->msgType);
+      return TSDB_CODE_VND_APP_ERROR;
+      break;
+  }
   return 0;
 }
 

source/libs/catalog/inc/catalogInt.h

@@ -46,7 +46,6 @@ typedef struct STableMetaCache {
 } STableMetaCache;
 
 typedef struct SCatalog {
-  SVgroupListCache vgroupCache;
   SDBVgroupCache   dbCache;
   STableMetaCache  tableCache;
 } SCatalog;
@@ -67,6 +66,7 @@ typedef uint32_t (*tableNameHashFp)(const char *, uint32_t);
 #define ctgTrace(...)  do { if (ctgDebugFlag & DEBUG_TRACE) { taosPrintLog("CTG ", ctgDebugFlag, __VA_ARGS__); }} while(0)
 #define ctgDebugL(...) do { if (ctgDebugFlag & DEBUG_DEBUG) { taosPrintLongString("CTG ", ctgDebugFlag, __VA_ARGS__); }} while(0)
 
+#define CTG_CACHE_ENABLED() (ctgMgmt.cfg.maxDBCacheNum > 0 || ctgMgmt.cfg.maxTblCacheNum > 0)
 
 #define CTG_ERR_RET(c) do { int32_t _code = c; if (_code != TSDB_CODE_SUCCESS) { terrno = _code; return _code; } } while (0)
 #define CTG_RET(c) do { int32_t _code = c; if (_code != TSDB_CODE_SUCCESS) { terrno = _code; } return _code; } while (0)

source/libs/catalog/src/catalog.c

@@ -146,8 +146,44 @@ void ctgGenEpSet(SEpSet *epSet, SVgroupInfo *vgroupInfo) {
   }
 }
 
+int32_t ctgGetTableMetaFromMnode(struct SCatalog* pCatalog, void *pRpc, const SEpSet* pMgmtEps, const char *pDBName, const char* pTableName, STableMetaOutput* output) {
+  if (NULL == pCatalog || NULL == pRpc || NULL == pMgmtEps || NULL == pDBName || NULL == pTableName || NULL == output) {
+    CTG_ERR_RET(TSDB_CODE_CTG_INVALID_INPUT);
+  }
+
+  char tbFullName[TSDB_TABLE_FNAME_LEN];
+
+  snprintf(tbFullName, sizeof(tbFullName), "%s.%s", pDBName, pTableName);
+
+  SBuildTableMetaInput bInput = {.vgId = 0, .tableFullName = tbFullName};
+  char *msg = NULL;
+  SEpSet *pVnodeEpSet = NULL;
+  int32_t msgLen = 0;
+
+  CTG_ERR_RET(queryBuildMsg[TDMT_MND_STB_META](&bInput, &msg, 0, &msgLen));
+
+  SRpcMsg rpcMsg = {
+      .msgType = TDMT_MND_STB_META,
+      .pCont   = msg,
+      .contLen = msgLen,
+  };
+
+  SRpcMsg rpcRsp = {0};
 
-int32_t ctgGetTableMetaFromMnode(struct SCatalog* pCatalog, void *pRpc, const SEpSet* pMgmtEps, const char *pDBName, const char* pTableName, SVgroupInfo *vgroupInfo, STableMetaOutput* output) {
+  rpcSendRecv(pRpc, (SEpSet*)pMgmtEps, &rpcMsg, &rpcRsp);
+  
+  if (TSDB_CODE_SUCCESS != rpcRsp.code) {
+    ctgError("error rsp for table meta, code:%x", rpcRsp.code);
+    CTG_ERR_RET(rpcRsp.code);
+  }
+
+  CTG_ERR_RET(queryProcessMsgRsp[TDMT_MND_STB_META](output, rpcRsp.pCont, rpcRsp.contLen));
+
+  return TSDB_CODE_SUCCESS;
+}
+
+
+int32_t ctgGetTableMetaFromVnode(struct SCatalog* pCatalog, void *pRpc, const SEpSet* pMgmtEps, const char *pDBName, const char* pTableName, SVgroupInfo *vgroupInfo, STableMetaOutput* output) {
   if (NULL == pCatalog || NULL == pRpc || NULL == pMgmtEps || NULL == pDBName || NULL == pTableName || NULL == vgroupInfo || NULL == output) {
     CTG_ERR_RET(TSDB_CODE_CTG_INVALID_INPUT);
   }
@@ -307,7 +343,9 @@ int32_t ctgUpdateTableMetaCache(struct SCatalog *pCatalog, STableMetaOutput *out
       ctgError("init hash[%d] for tablemeta cache failed", ctgMgmt.cfg.maxTblCacheNum);
       CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR);
     }
+  }
 
+  if (NULL == pCatalog->tableCache.stableCache) {
     pCatalog->tableCache.stableCache = taosHashInit(ctgMgmt.cfg.maxTblCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), true, HASH_ENTRY_LOCK);
     if (NULL == pCatalog->tableCache.stableCache) {
       ctgError("init hash[%d] for stablemeta cache failed", ctgMgmt.cfg.maxTblCacheNum);
@@ -318,55 +356,51 @@ int32_t ctgUpdateTableMetaCache(struct SCatalog *pCatalog, STableMetaOutput *out
   if (output->metaNum == 2) {
     if (taosHashPut(pCatalog->tableCache.cache, output->ctbFname, strlen(output->ctbFname), &output->ctbMeta, sizeof(output->ctbMeta)) != 0) {
       ctgError("push ctable[%s] to table cache failed", output->ctbFname);
-      goto error_exit;
+      CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR);
     }
 
     if (TSDB_SUPER_TABLE != output->tbMeta->tableType) {
       ctgError("table type[%d] error, expected:%d", output->tbMeta->tableType, TSDB_SUPER_TABLE);
-      goto error_exit;
+      CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR);
     }    
   }
 
   int32_t tbSize = sizeof(*output->tbMeta) + sizeof(SSchema) * (output->tbMeta->tableInfo.numOfColumns + output->tbMeta->tableInfo.numOfTags);
   if (taosHashPut(pCatalog->tableCache.cache, output->tbFname, strlen(output->tbFname), output->tbMeta, tbSize) != 0) {
     ctgError("push table[%s] to table cache failed", output->tbFname);
-    goto error_exit;
+    CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR);
   }
 
   if (TSDB_SUPER_TABLE == output->tbMeta->tableType) {
     if (taosHashPut(pCatalog->tableCache.stableCache, &output->tbMeta->suid, sizeof(output->tbMeta->suid), &output->tbMeta, POINTER_BYTES) != 0) {
       ctgError("push suid[%"PRIu64"] to stable cache failed", output->tbMeta->suid);
-      goto error_exit;
+      CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR);
     }
   }
   
   return TSDB_CODE_SUCCESS;
-
-error_exit:
-  if (pCatalog->vgroupCache.cache) {
-    taosHashCleanup(pCatalog->vgroupCache.cache);
-    pCatalog->vgroupCache.cache = NULL;
-  }
-
-  pCatalog->vgroupCache.vgroupVersion = CTG_DEFAULT_INVALID_VERSION;
-
-  CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR);
 }
 
 int32_t catalogInit(SCatalogCfg *cfg) {
-  ctgMgmt.pCluster = taosHashInit(CTG_DEFAULT_CACHE_CLUSTER_NUMBER, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK);
-  if (NULL == ctgMgmt.pCluster) {
-    CTG_ERR_LRET(TSDB_CODE_CTG_INTERNAL_ERROR, "init %d cluster cache failed", CTG_DEFAULT_CACHE_CLUSTER_NUMBER);
+  if (ctgMgmt.pCluster) {
+    ctgError("catalog already init");
+    CTG_ERR_RET(TSDB_CODE_CTG_INVALID_INPUT);
   }
 
   if (cfg) {
     memcpy(&ctgMgmt.cfg, cfg, sizeof(*cfg));
   } else {
-    ctgMgmt.cfg.enableVgroupCache = true;
     ctgMgmt.cfg.maxDBCacheNum = CTG_DEFAULT_CACHE_DB_NUMBER;
     ctgMgmt.cfg.maxTblCacheNum = CTG_DEFAULT_CACHE_TABLEMETA_NUMBER;
   }
 
+  if (CTG_CACHE_ENABLED()) {
+    ctgMgmt.pCluster = taosHashInit(CTG_DEFAULT_CACHE_CLUSTER_NUMBER, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK);
+    if (NULL == ctgMgmt.pCluster) {
+      CTG_ERR_LRET(TSDB_CODE_CTG_INTERNAL_ERROR, "init %d cluster cache failed", CTG_DEFAULT_CACHE_CLUSTER_NUMBER);
+    }
+  }
+
   return TSDB_CODE_SUCCESS;
 }
 
@@ -381,21 +415,19 @@ int32_t catalogGetHandle(const char* clusterId , struct SCatalog** catalogHandle
   }
 
   size_t clen = strlen(clusterId);
-  SCatalog *clusterCtg = (SCatalog *)taosHashGet(ctgMgmt.pCluster, clusterId, clen);
+  SCatalog **ctg = (SCatalog **)taosHashGet(ctgMgmt.pCluster, clusterId, clen);
 
-  if (clusterCtg) {
-    *catalogHandle = clusterCtg;
+  if (ctg && (*ctg)) {
+    *catalogHandle = *ctg;
     return TSDB_CODE_SUCCESS;
   }
 
-  clusterCtg = calloc(1, sizeof(*clusterCtg));
+  SCatalog *clusterCtg = calloc(1, sizeof(SCatalog));
   if (NULL == clusterCtg) {
-    ctgError("calloc %d failed", (int32_t)sizeof(*clusterCtg));
+    ctgError("calloc %d failed", (int32_t)sizeof(SCatalog));
     CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR);
   }
 
-  clusterCtg->vgroupCache.vgroupVersion = CTG_DEFAULT_INVALID_VERSION;
-
   if (taosHashPut(ctgMgmt.pCluster, clusterId, clen, &clusterCtg, POINTER_BYTES)) {
     ctgError("put cluster %s cache to hash failed", clusterId);
     tfree(clusterCtg);
@@ -443,7 +475,7 @@ int32_t catalogUpdateDBVgroupCache(struct SCatalog* pCatalog, const char* dbName
   }
 
   if (NULL == pCatalog->dbCache.cache) {
-    pCatalog->dbCache.cache = taosHashInit(CTG_DEFAULT_CACHE_DB_NUMBER, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK);
+    pCatalog->dbCache.cache = taosHashInit(ctgMgmt.cfg.maxDBCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK);
     if (NULL == pCatalog->dbCache.cache) {
       ctgError("init hash[%d] for db cache failed", CTG_DEFAULT_CACHE_DB_NUMBER);
       CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR);
@@ -515,7 +547,9 @@ int32_t catalogRenewTableMeta(struct SCatalog* pCatalog, void *pRpc, const SEpSe
 
   STableMetaOutput output = {0};
   
-  CTG_ERR_RET(ctgGetTableMetaFromMnode(pCatalog, pRpc, pMgmtEps, pDBName, pTableName, &vgroupInfo, &output));
+  //CTG_ERR_RET(ctgGetTableMetaFromVnode(pCatalog, pRpc, pMgmtEps, pDBName, pTableName, &vgroupInfo, &output));
+
+  CTG_ERR_RET(ctgGetTableMetaFromMnode(pCatalog, pRpc, pMgmtEps, pDBName, pTableName, &output));
 
   CTG_ERR_RET(ctgUpdateTableMetaCache(pCatalog, &output));
 

source/libs/parser/src/dCDAstProcess.c

@@ -313,7 +313,7 @@ int32_t doCheckForCreateCTable(SSqlInfo* pInfo, SParseBasicCtx *pCtx, SMsgBuf* p
       return code;
     }
 
-    code = tNameGetTableName(&name, pCreateTableInfo->tagdata.name);
+    const char* pSTableName = tNameGetTableName(&name);
 
     SArray* pValList = pCreateTableInfo->pTagVals;
     if (code != TSDB_CODE_SUCCESS) {
@@ -326,7 +326,7 @@ int32_t doCheckForCreateCTable(SSqlInfo* pInfo, SParseBasicCtx *pCtx, SMsgBuf* p
     char dbName[TSDB_DB_FNAME_LEN] = {0};
     tNameGetFullDbName(&name, dbName);
 
-    catalogGetTableMeta(pCtx->pCatalog, pCtx->pTransporter, &pCtx->mgmtEpSet, dbName, pCreateTableInfo->tagdata.name, &pSuperTableMeta);
+    catalogGetTableMeta(pCtx->pCatalog, pCtx->pTransporter, &pCtx->mgmtEpSet, dbName, pSTableName, &pSuperTableMeta);
 
     // too long tag values will return invalid sql, not be truncated automatically
     SSchema  *pTagSchema = getTableTagSchema(pSuperTableMeta);

source/libs/qcom/src/querymsg.c

@@ -266,9 +266,11 @@ int32_t queryProcessTableMetaRsp(void* output, char *msg, int32_t msgSize) {
 
 void initQueryModuleMsgHandle() {
   queryBuildMsg[TDMT_VND_TABLE_META] = queryBuildTableMetaReqMsg;
+  queryBuildMsg[TDMT_MND_STB_META] = queryBuildTableMetaReqMsg;
   queryBuildMsg[TDMT_MND_USE_DB] = queryBuildUseDbMsg;
 
   queryProcessMsgRsp[TDMT_VND_TABLE_META] = queryProcessTableMetaRsp;
+  queryProcessMsgRsp[TDMT_MND_STB_META] = queryProcessTableMetaRsp;
   queryProcessMsgRsp[TDMT_MND_USE_DB] = queryProcessUseDBRsp;
 }
 

source/libs/qworker/inc/qworkerInt.h

@@ -42,25 +42,41 @@ enum {
   QW_WRITE,
 };
 
-typedef struct SQWorkerTaskStatus {  
+enum {
+  QW_EXIST_ACQUIRE = 1,
+  QW_EXIST_RET_ERR,
+};
+
+enum {
+  QW_NOT_EXIST_RET_ERR = 1,
+  QW_NOT_EXIST_ADD,
+};
+
+enum {
+  QW_ADD_RET_ERR = 1,
+  QW_ADD_ACQUIRE,
+};
+
+
+typedef struct SQWTaskStatus {  
   SRWLatch lock;
   int32_t  code;
   int8_t   status;
   int8_t   ready; 
   bool     cancel;
   bool     drop;
-} SQWorkerTaskStatus;
+} SQWTaskStatus;
 
 typedef struct SQWorkerResCache {
   SRWLatch lock;
   void *data;
 } SQWorkerResCache;
 
-typedef struct SQWorkerSchStatus {
+typedef struct SQWSchStatus {
   int32_t   lastAccessTs; // timestamp in second
   SRWLatch  tasksLock;
   SHashObj *tasksHash;   // key:queryId+taskId, value: SQWorkerTaskStatus
-} SQWorkerSchStatus;
+} SQWSchStatus;
 
 // Qnode/Vnode level task management
 typedef struct SQWorkerMgmt {
@@ -71,7 +87,7 @@ typedef struct SQWorkerMgmt {
   SHashObj *resHash;       //key: queryId+taskId, value: SQWorkerResCache
 } SQWorkerMgmt;
 
-#define QW_GOT_RES_DATA(data) (false)
+#define QW_GOT_RES_DATA(data) (true)
 #define QW_LOW_RES_DATA(data) (false)
 
 #define QW_TASK_NOT_EXIST(code) (TSDB_CODE_QRY_SCH_NOT_EXIST == (code) || TSDB_CODE_QRY_TASK_NOT_EXIST == (code))
@@ -86,8 +102,31 @@ typedef struct SQWorkerMgmt {
 #define QW_ERR_LRET(c,...) do { int32_t _code = c; if (_code != TSDB_CODE_SUCCESS) { qError(__VA_ARGS__); terrno = _code; return _code; } } while (0)
 #define QW_ERR_JRET(c) do { code = c; if (code != TSDB_CODE_SUCCESS) { terrno = code; goto _return; } } while (0)
 
-#define QW_LOCK(type, _lock) (QW_READ == (type) ? taosRLockLatch(_lock) : taosWLockLatch(_lock))
-#define QW_UNLOCK(type, _lock) (QW_READ == (type) ? taosRUnLockLatch(_lock) : taosWUnLockLatch(_lock))
+#define QW_LOCK(type, _lock) do {   \
+  if (QW_READ == (type)) {          \
+    if ((*(_lock)) < 0) assert(0);    \
+    taosRLockLatch(_lock);          \
+    qDebug("RLOCK%p, %s:%d", (_lock), __FILE__, __LINE__); \
+  } else {                                                \
+    if ((*(_lock)) < 0) assert(0);                          \
+    taosWLockLatch(_lock);                                \
+    qDebug("WLOCK%p, %s:%d", (_lock), __FILE__, __LINE__);  \
+  }                                                       \
+} while (0)
+
+#define QW_UNLOCK(type, _lock) do {                       \
+  if (QW_READ == (type)) {                                \
+    if ((*(_lock)) <= 0) assert(0);                         \
+    taosRUnLockLatch(_lock);                              \
+    qDebug("RULOCK%p, %s:%d", (_lock), __FILE__, __LINE__); \
+  } else {                                                \
+    if ((*(_lock)) <= 0) assert(0);                         \
+    taosWUnLockLatch(_lock);                              \
+    qDebug("WULOCK%p, %s:%d", (_lock), __FILE__, __LINE__); \
+  }                                                       \
+} while (0)
+
+static int32_t qwAcquireScheduler(int32_t rwType, SQWorkerMgmt *mgmt, uint64_t sId, SQWSchStatus **sch, int32_t nOpt);
 
 
 #ifdef __cplusplus

source/libs/qworker/test/qworkerTests.cpp

@@ -36,10 +36,25 @@
 
 namespace {
 
+bool testStop = false;
+
 int32_t qwtStringToPlan(const char* str, SSubplan** subplan) {
   return 0;
 }
 
+void qwtRpcSendResponse(const SRpcMsg *pRsp) {
+  if (TDMT_VND_TASKS_STATUS_RSP == pRsp->msgType) {
+    SSchedulerStatusRsp *rsp = (SSchedulerStatusRsp *)pRsp->pCont;
+    printf("task num:%d\n", rsp->num);
+    for (int32_t i = 0; i < rsp->num; ++i) {
+      STaskStatus *task = &rsp->status[i];
+      printf("qId:%"PRIx64",tId:%"PRIx64",status:%d\n", task->queryId, task->taskId, task->status);
+    }
+  }
+  return;
+}
+
+
 
 void stubSetStringToPlan() {
   static Stub stub;
@@ -54,11 +69,148 @@ void stubSetStringToPlan() {
   }
 }
 
+void stubSetRpcSendResponse() {
+  static Stub stub;
+  stub.set(rpcSendResponse, qwtRpcSendResponse);
+  {
+    AddrAny any("libplanner.so");
+    std::map<std::string,void*> result;
+    any.get_global_func_addr_dynsym("^rpcSendResponse$", result);
+    for (const auto& f : result) {
+      stub.set(f.second, qwtRpcSendResponse);
+    }
+  }
+}
+
+void *queryThread(void *param) {
+  SRpcMsg queryRpc = {0};
+  int32_t code = 0;
+  uint32_t n = 0;
+  void *mockPointer = (void *)0x1;    
+  void *mgmt = param;
+  SSubQueryMsg *queryMsg = (SSubQueryMsg *)calloc(1, sizeof(SSubQueryMsg) + 100);
+  queryMsg->queryId = htobe64(1);
+  queryMsg->sId = htobe64(1);
+  queryMsg->taskId = htobe64(1);
+  queryMsg->contentLen = htonl(100);
+  queryRpc.pCont = queryMsg;
+  queryRpc.contLen = sizeof(SSubQueryMsg) + 100;
+
+  while (!testStop) {
+    qWorkerProcessQueryMsg(mockPointer, mgmt, &queryRpc);
+    usleep(rand()%5);
+    if (++n % 50000 == 0) {
+      printf("query:%d\n", n);
+    }
+  }
+
+  return NULL;
+}
+
+void *readyThread(void *param) {
+  SRpcMsg readyRpc = {0};
+  int32_t code = 0;
+  uint32_t n = 0;  
+  void *mockPointer = (void *)0x1;    
+  void *mgmt = param;
+  SResReadyMsg readyMsg = {0};
+  readyMsg.sId = htobe64(1);
+  readyMsg.queryId = htobe64(1);
+  readyMsg.taskId = htobe64(1);
+  readyRpc.pCont = &readyMsg;
+  readyRpc.contLen = sizeof(SResReadyMsg);
+
+  while (!testStop) {
+    code = qWorkerProcessReadyMsg(mockPointer, mgmt, &readyRpc);
+    usleep(rand()%5);
+    if (++n % 50000 == 0) {
+      printf("ready:%d\n", n);
+    }    
+  }
+
+  return NULL;
+}
+
+void *fetchThread(void *param) {
+  SRpcMsg fetchRpc = {0};
+  int32_t code = 0;
+  uint32_t n = 0;  
+  void *mockPointer = (void *)0x1;    
+  void *mgmt = param;
+  SResFetchMsg fetchMsg = {0};
+  fetchMsg.sId = htobe64(1);
+  fetchMsg.queryId = htobe64(1);
+  fetchMsg.taskId = htobe64(1);
+  fetchRpc.pCont = &fetchMsg;
+  fetchRpc.contLen = sizeof(SResFetchMsg);
+
+  while (!testStop) {
+    code = qWorkerProcessFetchMsg(mockPointer, mgmt, &fetchRpc);
+    usleep(rand()%5);
+    if (++n % 50000 == 0) {
+      printf("fetch:%d\n", n);
+    }    
+  }
+
+  return NULL;
+}
+
+void *dropThread(void *param) {
+  SRpcMsg dropRpc = {0};
+  int32_t code = 0;
+  uint32_t n = 0;  
+  void *mockPointer = (void *)0x1;    
+  void *mgmt = param;
+  STaskDropMsg dropMsg = {0};  
+  dropMsg.sId = htobe64(1);
+  dropMsg.queryId = htobe64(1);
+  dropMsg.taskId = htobe64(1);
+  dropRpc.pCont = &dropMsg;
+  dropRpc.contLen = sizeof(STaskDropMsg);
+
+  while (!testStop) {
+    code = qWorkerProcessDropMsg(mockPointer, mgmt, &dropRpc);
+    usleep(rand()%5);
+    if (++n % 50000 == 0) {
+      printf("drop:%d\n", n);
+    }    
+  }
+
+  return NULL;
+}
+
+void *statusThread(void *param) {
+  SRpcMsg statusRpc = {0};
+  int32_t code = 0;
+  uint32_t n = 0;  
+  void *mockPointer = (void *)0x1;    
+  void *mgmt = param;
+  SSchTasksStatusMsg statusMsg = {0};
+  statusMsg.sId = htobe64(1);
+  statusRpc.pCont = &statusMsg;
+  statusRpc.contLen = sizeof(SSchTasksStatusMsg);
+  statusRpc.msgType = TDMT_VND_TASKS_STATUS;
+
+  while (!testStop) {
+    statusMsg.sId = htobe64(1);
+    code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+    usleep(rand()%5);
+    if (++n % 50000 == 0) {
+      printf("status:%d\n", n);
+    }    
+  }
+
+  return NULL;
+}
+
+
+
+
 
 }
 
 
-TEST(testCase, normalCase) {
+TEST(seqTest, normalCase) {
   void *mgmt = NULL;
   int32_t code = 0;
   void *mockPointer = (void *)0x1;
@@ -66,48 +218,254 @@ TEST(testCase, normalCase) {
   SRpcMsg readyRpc = {0};
   SRpcMsg fetchRpc = {0};
   SRpcMsg dropRpc = {0};
+  SRpcMsg statusRpc = {0};
+  
   SSubQueryMsg *queryMsg = (SSubQueryMsg *)calloc(1, sizeof(SSubQueryMsg) + 100);
   queryMsg->queryId = htobe64(1);
-  queryMsg->schedulerId = htobe64(1);
+  queryMsg->sId = htobe64(1);
   queryMsg->taskId = htobe64(1);
   queryMsg->contentLen = htonl(100);
   queryRpc.pCont = queryMsg;
+  queryRpc.contLen = sizeof(SSubQueryMsg) + 100;
 
   SResReadyMsg readyMsg = {0};
-  readyMsg.schedulerId = htobe64(1);
+  readyMsg.sId = htobe64(1);
   readyMsg.queryId = htobe64(1);
   readyMsg.taskId = htobe64(1);
   readyRpc.pCont = &readyMsg;
+  readyRpc.contLen = sizeof(SResReadyMsg);
 
   SResFetchMsg fetchMsg = {0};
-  fetchMsg.schedulerId = htobe64(1);
+  fetchMsg.sId = htobe64(1);
   fetchMsg.queryId = htobe64(1);
   fetchMsg.taskId = htobe64(1);
   fetchRpc.pCont = &fetchMsg;
+  fetchRpc.contLen = sizeof(SResFetchMsg);
 
   STaskDropMsg dropMsg = {0};  
-  dropMsg.schedulerId = htobe64(1);
+  dropMsg.sId = htobe64(1);
   dropMsg.queryId = htobe64(1);
   dropMsg.taskId = htobe64(1);
   dropRpc.pCont = &dropMsg;
+  dropRpc.contLen = sizeof(STaskDropMsg);
 
+  SSchTasksStatusMsg statusMsg = {0};
+  statusMsg.sId = htobe64(1);
+  statusRpc.pCont = &statusMsg;
+  statusRpc.contLen = sizeof(SSchTasksStatusMsg);
+  statusRpc.msgType = TDMT_VND_TASKS_STATUS;
+  
   stubSetStringToPlan();
+  stubSetRpcSendResponse();
   
   code = qWorkerInit(NULL, &mgmt);
   ASSERT_EQ(code, 0);
 
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
   code = qWorkerProcessQueryMsg(mockPointer, mgmt, &queryRpc);
   ASSERT_EQ(code, 0);
 
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
   code = qWorkerProcessReadyMsg(mockPointer, mgmt, &readyRpc);
   ASSERT_EQ(code, 0);
 
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
   code = qWorkerProcessFetchMsg(mockPointer, mgmt, &fetchRpc);
   ASSERT_EQ(code, 0);
 
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
   code = qWorkerProcessDropMsg(mockPointer, mgmt, &dropRpc);
   ASSERT_EQ(code, 0);
+
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
+  qWorkerDestroy(&mgmt);
+}
+
+TEST(seqTest, cancelFirst) {
+  void *mgmt = NULL;
+  int32_t code = 0;
+  void *mockPointer = (void *)0x1;
+  SRpcMsg queryRpc = {0};
+  SRpcMsg dropRpc = {0};
+  SRpcMsg statusRpc = {0};
+  
+  SSubQueryMsg *queryMsg = (SSubQueryMsg *)calloc(1, sizeof(SSubQueryMsg) + 100);
+  queryMsg->queryId = htobe64(1);
+  queryMsg->sId = htobe64(1);
+  queryMsg->taskId = htobe64(1);
+  queryMsg->contentLen = htonl(100);
+  queryRpc.pCont = queryMsg;
+  queryRpc.contLen = sizeof(SSubQueryMsg) + 100;
+
+  STaskDropMsg dropMsg = {0};  
+  dropMsg.sId = htobe64(1);
+  dropMsg.queryId = htobe64(1);
+  dropMsg.taskId = htobe64(1);
+  dropRpc.pCont = &dropMsg;
+  dropRpc.contLen = sizeof(STaskDropMsg);
+
+  SSchTasksStatusMsg statusMsg = {0};
+  statusMsg.sId = htobe64(1);
+  statusRpc.pCont = &statusMsg;
+  statusRpc.contLen = sizeof(SSchTasksStatusMsg);
+  statusRpc.msgType = TDMT_VND_TASKS_STATUS;
+
+  stubSetStringToPlan();
+  stubSetRpcSendResponse();
+  
+  code = qWorkerInit(NULL, &mgmt);
+  ASSERT_EQ(code, 0);
+
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
+  code = qWorkerProcessDropMsg(mockPointer, mgmt, &dropRpc);
+  ASSERT_EQ(code, 0);
+
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
+  code = qWorkerProcessQueryMsg(mockPointer, mgmt, &queryRpc);
+  ASSERT_EQ(code, 0);
+
+  statusMsg.sId = htobe64(1);
+  code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+  ASSERT_EQ(code, 0);
+
+  qWorkerDestroy(&mgmt);
+}
+
+TEST(seqTest, randCase) {
+  void *mgmt = NULL;
+  int32_t code = 0;
+  void *mockPointer = (void *)0x1;
+  SRpcMsg queryRpc = {0};
+  SRpcMsg readyRpc = {0};
+  SRpcMsg fetchRpc = {0};
+  SRpcMsg dropRpc = {0};
+  SRpcMsg statusRpc = {0};
+  
+  SSubQueryMsg *queryMsg = (SSubQueryMsg *)calloc(1, sizeof(SSubQueryMsg) + 100);
+  queryMsg->queryId = htobe64(1);
+  queryMsg->sId = htobe64(1);
+  queryMsg->taskId = htobe64(1);
+  queryMsg->contentLen = htonl(100);
+  queryRpc.pCont = queryMsg;
+  queryRpc.contLen = sizeof(SSubQueryMsg) + 100;
+
+  SResReadyMsg readyMsg = {0};
+  readyMsg.sId = htobe64(1);
+  readyMsg.queryId = htobe64(1);
+  readyMsg.taskId = htobe64(1);
+  readyRpc.pCont = &readyMsg;
+  readyRpc.contLen = sizeof(SResReadyMsg);
+
+  SResFetchMsg fetchMsg = {0};
+  fetchMsg.sId = htobe64(1);
+  fetchMsg.queryId = htobe64(1);
+  fetchMsg.taskId = htobe64(1);
+  fetchRpc.pCont = &fetchMsg;
+  fetchRpc.contLen = sizeof(SResFetchMsg);
+
+  STaskDropMsg dropMsg = {0};  
+  dropMsg.sId = htobe64(1);
+  dropMsg.queryId = htobe64(1);
+  dropMsg.taskId = htobe64(1);
+  dropRpc.pCont = &dropMsg;
+  dropRpc.contLen = sizeof(STaskDropMsg);
+
+  SSchTasksStatusMsg statusMsg = {0};
+  statusMsg.sId = htobe64(1);
+  statusRpc.pCont = &statusMsg;
+  statusRpc.contLen = sizeof(SSchTasksStatusMsg);
+  statusRpc.msgType = TDMT_VND_TASKS_STATUS;
+  
+  stubSetStringToPlan();
+  stubSetRpcSendResponse();
+
+  srand(time(NULL));
+  
+  code = qWorkerInit(NULL, &mgmt);
+  ASSERT_EQ(code, 0);
+
+  int32_t t = 0;
+  int32_t maxr = 10001;
+  while (true) {
+    int32_t r = rand() % maxr;
+    
+    if (r >= 0 && r < maxr/5) {
+      printf("Query,%d\n", t++);
+      code = qWorkerProcessQueryMsg(mockPointer, mgmt, &queryRpc);
+    } else if (r >= maxr/5 && r < maxr * 2/5) {
+      printf("Ready,%d\n", t++);
+      code = qWorkerProcessReadyMsg(mockPointer, mgmt, &readyRpc);
+    } else if (r >= maxr * 2/5 && r < maxr* 3/5) {
+      printf("Fetch,%d\n", t++);
+      code = qWorkerProcessFetchMsg(mockPointer, mgmt, &fetchRpc);
+    } else if (r >= maxr * 3/5 && r < maxr * 4/5) {
+      printf("Drop,%d\n", t++);
+      code = qWorkerProcessDropMsg(mockPointer, mgmt, &dropRpc);
+    } else if (r >= maxr * 4/5 && r < maxr-1) {
+      printf("Status,%d\n", t++);
+      statusMsg.sId = htobe64(1);
+      code = qWorkerProcessStatusMsg(mockPointer, mgmt, &statusRpc);
+      ASSERT_EQ(code, 0);
+    } else {
+      printf("QUIT RAND NOW");
+      break;
+    }
+  }
+
+  qWorkerDestroy(&mgmt);
+}
+
+TEST(seqTest, multithreadRand) {
+  void *mgmt = NULL;
+  int32_t code = 0;
+  void *mockPointer = (void *)0x1;
+  
+  stubSetStringToPlan();
+  stubSetRpcSendResponse();
+
+  srand(time(NULL));
+  
+  code = qWorkerInit(NULL, &mgmt);
+  ASSERT_EQ(code, 0);
+
+  pthread_attr_t thattr;
+  pthread_attr_init(&thattr);
+
+  pthread_t t1,t2,t3,t4,t5;
+  pthread_create(&(t1), &thattr, queryThread, mgmt);
+  pthread_create(&(t2), &thattr, readyThread, NULL);
+  pthread_create(&(t3), &thattr, fetchThread, NULL);
+  pthread_create(&(t4), &thattr, dropThread, NULL);
+  pthread_create(&(t5), &thattr, statusThread, NULL);
+
+  int32_t t = 0;
+  int32_t maxr = 10001;
+  sleep(300);
+  testStop = true;
+  sleep(1);
   
+  qWorkerDestroy(&mgmt);
 }
 
 

source/libs/scheduler/inc/schedulerInt.h

@@ -38,11 +38,16 @@ enum {
 
 typedef struct SSchedulerMgmt {
   uint64_t  taskId; 
-  uint64_t  schedulerId;
+  uint64_t  sId;
   SSchedulerCfg cfg;
   SHashObj *jobs;  // key: queryId, value: SQueryJob*
 } SSchedulerMgmt;
 
+typedef struct SSchCallbackParam {
+  uint64_t queryId;
+  uint64_t taskId;
+} SSchCallbackParam;
+
 typedef struct SSchLevel {
   int32_t  level;
   int8_t   status;
@@ -120,6 +125,7 @@ typedef struct SSchJob {
 
 
 extern int32_t schLaunchTask(SSchJob *job, SSchTask *task);
+extern int32_t schBuildAndSendMsg(SSchJob *job, SSchTask *task, int32_t msgType);
 
 #ifdef __cplusplus
 }

source/libs/scheduler/src/scheduler.c

@@ -21,36 +21,6 @@
 SSchedulerMgmt schMgmt = {0};
 
 
-int32_t schBuildAndSendRequest(void *pRpc, const SEpSet* pMgmtEps, __taos_async_fn_t fp) {
-/*
-  SRequestObj *pRequest = createRequest(pTscObj, fp, param, TSDB_SQL_CONNECT);
-  if (pRequest == NULL) {
-    return TSDB_CODE_TSC_OUT_OF_MEMORY;
-  }
-
-  SRequestMsgBody body = {0};
-  buildConnectMsg(pRequest, &body);
-
-  int64_t transporterId = 0;
-  asyncSendMsgToServer(pTscObj->pTransporter, &pTscObj->pAppInfo->mgmtEp.epSet, &body, &transporterId);
-
-  tsem_wait(&pRequest->body.rspSem);
-  destroyConnectMsg(&body);
-
-  if (pRequest->code != TSDB_CODE_SUCCESS) {
-    const char *errorMsg = (pRequest->code == TSDB_CODE_RPC_FQDN_ERROR) ? taos_errstr(pRequest) : tstrerror(terrno);
-    printf("failed to connect to server, reason: %s\n\n", errorMsg);
-
-    destroyRequest(pRequest);
-    taos_close(pTscObj);
-    pTscObj = NULL;
-  } else {
-    tscDebug("0x%"PRIx64" connection is opening, connId:%d, dnodeConn:%p", pTscObj->id, pTscObj->connId, pTscObj->pTransporter);
-    destroyRequest(pRequest);
-  }
-*/  
-}
-
 int32_t schBuildTaskRalation(SSchJob *job, SHashObj *planToTask) {
   for (int32_t i = 0; i < job->levelNum; ++i) {
     SSchLevel *level = taosArrayGet(job->levels, i);
@@ -312,100 +282,6 @@ int32_t schMoveTaskToFailList(SSchJob *job, SSchTask *task, bool *moved) {
   return TSDB_CODE_SUCCESS;
 }
 
-
-int32_t schAsyncSendMsg(SSchJob *job, SSchTask *task, int32_t msgType) {
-  int32_t msgSize = 0;
-  void *msg = NULL;
-  
-  switch (msgType) {
-    case TDMT_VND_SUBMIT: {
-      if (NULL == task->msg || task->msgLen <= 0) {
-        qError("submit msg is NULL");
-        SCH_ERR_RET(TSDB_CODE_SCH_INTERNAL_ERROR);
-      }
-
-      msgSize = task->msgLen;
-      msg = task->msg;
-      break;
-    }
-    case TDMT_VND_QUERY: {
-      if (NULL == task->msg) {
-        qError("query msg is NULL");
-        SCH_ERR_RET(TSDB_CODE_SCH_INTERNAL_ERROR);
-      }
-
-      msgSize = sizeof(SSubQueryMsg) + task->msgLen;
-      msg = calloc(1, msgSize);
-      if (NULL == msg) {
-        qError("calloc %d failed", msgSize);
-        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
-      }
-
-      SSubQueryMsg *pMsg = msg;
-      
-      pMsg->schedulerId = htobe64(schMgmt.schedulerId);
-      pMsg->queryId = htobe64(job->queryId);
-      pMsg->taskId = htobe64(task->taskId);
-      pMsg->contentLen = htonl(task->msgLen);
-      memcpy(pMsg->msg, task->msg, task->msgLen);
-      break;
-    }    
-    case TDMT_VND_RES_READY: {
-      msgSize = sizeof(SResReadyMsg);
-      msg = calloc(1, msgSize);
-      if (NULL == msg) {
-        qError("calloc %d failed", msgSize);
-        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
-      }
-
-      SResReadyMsg *pMsg = msg;
-      pMsg->schedulerId = htobe64(schMgmt.schedulerId);      
-      pMsg->queryId = htobe64(job->queryId);
-      pMsg->taskId = htobe64(task->taskId);      
-      break;
-    }
-    case TDMT_VND_FETCH: {
-      if (NULL == task) {
-        SCH_ERR_RET(TSDB_CODE_QRY_APP_ERROR);
-      }
-      msgSize = sizeof(SResFetchMsg);
-      msg = calloc(1, msgSize);
-      if (NULL == msg) {
-        qError("calloc %d failed", msgSize);
-        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
-      }
-    
-      SResFetchMsg *pMsg = msg;
-      pMsg->schedulerId = htobe64(schMgmt.schedulerId);      
-      pMsg->queryId = htobe64(job->queryId);
-      pMsg->taskId = htobe64(task->taskId);      
-      break;
-    }
-    case TDMT_VND_DROP_TASK:{
-      msgSize = sizeof(STaskDropMsg);
-      msg = calloc(1, msgSize);
-      if (NULL == msg) {
-        qError("calloc %d failed", msgSize);
-        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
-      }
-    
-      STaskDropMsg *pMsg = msg;
-      pMsg->schedulerId = htobe64(schMgmt.schedulerId);      
-      pMsg->queryId = htobe64(job->queryId);
-      pMsg->taskId = htobe64(task->taskId);      
-      break;
-    }
-    default:
-      qError("unknown msg type:%d", msgType);
-      break;
-  }
-
-  //TODO SEND MSG
-  //taosAsyncExec(rpcSendRequest(void * shandle, const SEpSet * pEpSet, SRpcMsg * pMsg, int64_t * pRid), p, &code);
-
-  return TSDB_CODE_SUCCESS;
-}
-
 int32_t schTaskCheckAndSetRetry(SSchJob *job, SSchTask *task, int32_t errCode, bool *needRetry) {
   // TODO set retry or not based on task type/errCode/retry times/job status/available eps...
   // TODO if needRetry, set task retry info
@@ -424,7 +300,7 @@ int32_t schFetchFromRemote(SSchJob *job) {
     return TSDB_CODE_SUCCESS;
   }
 
-  SCH_ERR_JRET(schAsyncSendMsg(job, job->fetchTask, TDMT_VND_FETCH));
+  SCH_ERR_JRET(schBuildAndSendMsg(job, job->fetchTask, TDMT_VND_FETCH));
 
   return TSDB_CODE_SUCCESS;
   
@@ -577,11 +453,11 @@ int32_t schProcessOnTaskFailure(SSchJob *job, SSchTask *task, int32_t errCode) {
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t schHandleRspMsg(SSchJob *job, SSchTask *task, int32_t msgType, char *msg, int32_t msgSize, int32_t rspCode) {
+int32_t schProcessRspMsg(SSchJob *job, SSchTask *task, int32_t msgType, char *msg, int32_t msgSize, int32_t rspCode) {
   int32_t code = 0;
   
   switch (msgType) {
-    case TDMT_VND_SUBMIT: {
+    case TDMT_VND_SUBMIT_RSP: {
         SShellSubmitRspMsg *rsp = (SShellSubmitRspMsg *)msg;
         if (rsp->code != TSDB_CODE_SUCCESS) {
           SCH_ERR_JRET(schProcessOnTaskFailure(job, task, rsp->code));
@@ -595,20 +471,20 @@ int32_t schHandleRspMsg(SSchJob *job, SSchTask *task, int32_t msgType, char *msg
         }
         break;
       }
-    case TDMT_VND_QUERY: {
+    case TDMT_VND_QUERY_RSP: {
         SQueryTableRsp *rsp = (SQueryTableRsp *)msg;
         
         if (rsp->code != TSDB_CODE_SUCCESS) {
           SCH_ERR_JRET(schProcessOnTaskFailure(job, task, rsp->code));
         } else {
-          code = schAsyncSendMsg(job, task, TDMT_VND_RES_READY);
+          code = schBuildAndSendMsg(job, task, TDMT_VND_RES_READY);
           if (code) {
             goto _task_error;
           }
         }
         break;
       }
-    case TDMT_VND_RES_READY: {
+    case TDMT_VND_RES_READY_RSP: {
         SResReadyRsp *rsp = (SResReadyRsp *)msg;
         
         if (rsp->code != TSDB_CODE_SUCCESS) {
@@ -621,7 +497,7 @@ int32_t schHandleRspMsg(SSchJob *job, SSchTask *task, int32_t msgType, char *msg
         }
         break;
       }
-    case TDMT_VND_FETCH: {
+    case TDMT_VND_FETCH_RSP: {
         SCH_ERR_JRET(rspCode);
         SRetrieveTableRsp *rsp = (SRetrieveTableRsp *)msg;
 
@@ -631,6 +507,9 @@ int32_t schHandleRspMsg(SSchJob *job, SSchTask *task, int32_t msgType, char *msg
         SCH_ERR_JRET(schProcessOnDataFetched(job));
         break;
       }
+    case TDMT_VND_DROP_TASK: {
+
+      }
     default:
       qError("unknown msg type:%d received", msgType);
       return TSDB_CODE_QRY_INVALID_INPUT;
@@ -648,6 +527,211 @@ _return:
 }
 
 
+int32_t schHandleCallback(void* param, const SDataBuf* pMsg, int32_t msgType, int32_t rspCode) {
+  int32_t code = 0;
+  SSchCallbackParam *pParam = (SSchCallbackParam *)param;
+  
+  SSchJob **job = taosHashGet(schMgmt.jobs, &pParam->queryId, sizeof(pParam->queryId));
+  if (NULL == job || NULL == (*job)) {
+    qError("taosHashGet queryId:%"PRIx64" not exist", pParam->queryId);
+    SCH_ERR_JRET(TSDB_CODE_SCH_INTERNAL_ERROR);
+  }
+
+  SSchTask **task = taosHashGet((*job)->execTasks, &pParam->taskId, sizeof(pParam->taskId));
+  if (NULL == task || NULL == (*task)) {
+    qError("taosHashGet taskId:%"PRIx64" not exist", pParam->taskId);
+    SCH_ERR_JRET(TSDB_CODE_SCH_INTERNAL_ERROR);
+  }
+  
+  schProcessRspMsg(*job, *task, msgType, pMsg->pData, pMsg->len, rspCode);
+
+_return:  
+  tfree(param);
+
+  SCH_RET(code);
+}
+
+int32_t schHandleSubmitCallback(void* param, const SDataBuf* pMsg, int32_t code) {
+  return schHandleCallback(param, pMsg, TDMT_VND_SUBMIT_RSP, code);
+}
+int32_t schHandleQueryCallback(void* param, const SDataBuf* pMsg, int32_t code) {
+  return schHandleCallback(param, pMsg, TDMT_VND_QUERY_RSP, code);
+}
+int32_t schHandleFetchCallback(void* param, const SDataBuf* pMsg, int32_t code) {
+  return schHandleCallback(param, pMsg, TDMT_VND_FETCH_RSP, code);
+}
+int32_t schHandleReadyCallback(void* param, const SDataBuf* pMsg, int32_t code) {
+  return schHandleCallback(param, pMsg, TDMT_VND_RES_READY_RSP, code);
+}
+int32_t schHandleDropCallback(void* param, const SDataBuf* pMsg, int32_t code) {  
+  SSchCallbackParam *pParam = (SSchCallbackParam *)param;
+  qDebug("drop task rsp received, queryId:%"PRIx64 ",taksId:%"PRIx64 ",code:%d", pParam->queryId, pParam->taskId, code);
+}
+
+int32_t schGetCallbackFp(int32_t msgType, __async_send_cb_fn_t *fp) {
+  switch (msgType) {
+    case TDMT_VND_SUBMIT: 
+      *fp = schHandleSubmitCallback;
+      break;
+    case TDMT_VND_QUERY: 
+      *fp = schHandleQueryCallback;
+      break;
+    case TDMT_VND_RES_READY: 
+      *fp = schHandleReadyCallback;
+      break;
+    case TDMT_VND_FETCH: 
+      *fp = schHandleFetchCallback;
+      break;
+    case TDMT_VND_DROP_TASK:
+      *fp = schHandleDropCallback;
+      break;
+    default:
+      qError("unknown msg type:%d", msgType);
+      SCH_ERR_RET(TSDB_CODE_QRY_APP_ERROR);
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+
+
+int32_t schAsyncSendMsg(void *transport, SEpSet* epSet, uint64_t qId, uint64_t tId, int32_t msgType, void *msg, uint32_t msgSize) {
+  int32_t code = 0;
+  SMsgSendInfo* pMsgSendInfo = calloc(1, sizeof(SMsgSendInfo));
+  if (NULL == pMsgSendInfo) {
+    qError("calloc %d failed", (int32_t)sizeof(SMsgSendInfo));
+    SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
+  }
+
+  SSchCallbackParam *param = calloc(1, sizeof(SSchCallbackParam));
+  if (NULL == param) {
+    qError("calloc %d failed", (int32_t)sizeof(SSchCallbackParam));
+    SCH_ERR_JRET(TSDB_CODE_QRY_OUT_OF_MEMORY);
+  }
+
+  __async_send_cb_fn_t fp = NULL;
+  SCH_ERR_JRET(schGetCallbackFp(msgType, &fp));
+
+  param->queryId = qId;
+  param->taskId = tId;
+
+  pMsgSendInfo->param = param;
+  pMsgSendInfo->msgInfo.pData = msg;
+  pMsgSendInfo->msgInfo.len = msgSize;
+  pMsgSendInfo->msgType = msgType;
+
+  pMsgSendInfo->fp = fp;
+  
+  int64_t  transporterId = 0;
+  SCH_ERR_JRET(asyncSendMsgToServer(transport, epSet, &transporterId, pMsgSendInfo));
+  
+  return TSDB_CODE_SUCCESS;
+
+_return:
+  tfree(param);
+  tfree(pMsgSendInfo);
+
+  SCH_RET(code);
+}
+
+
+int32_t schBuildAndSendMsg(SSchJob *job, SSchTask *task, int32_t msgType) {
+  uint32_t msgSize = 0;
+  void *msg = NULL;
+  int32_t code = 0;
+  
+  switch (msgType) {
+    case TDMT_VND_SUBMIT: {
+      if (NULL == task->msg || task->msgLen <= 0) {
+        qError("submit msg is NULL");
+        SCH_ERR_RET(TSDB_CODE_SCH_INTERNAL_ERROR);
+      }
+
+      msgSize = task->msgLen;
+      msg = task->msg;
+      break;
+    }
+    case TDMT_VND_QUERY: {
+      if (NULL == task->msg) {
+        qError("query msg is NULL");
+        SCH_ERR_RET(TSDB_CODE_SCH_INTERNAL_ERROR);
+      }
+
+      msgSize = sizeof(SSubQueryMsg) + task->msgLen;
+      msg = calloc(1, msgSize);
+      if (NULL == msg) {
+        qError("calloc %d failed", msgSize);
+        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
+      }
+
+      SSubQueryMsg *pMsg = msg;
+      
+      pMsg->sId = htobe64(schMgmt.sId);
+      pMsg->queryId = htobe64(job->queryId);
+      pMsg->taskId = htobe64(task->taskId);
+      pMsg->contentLen = htonl(task->msgLen);
+      memcpy(pMsg->msg, task->msg, task->msgLen);
+      break;
+    }    
+    case TDMT_VND_RES_READY: {
+      msgSize = sizeof(SResReadyMsg);
+      msg = calloc(1, msgSize);
+      if (NULL == msg) {
+        qError("calloc %d failed", msgSize);
+        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
+      }
+
+      SResReadyMsg *pMsg = msg;
+      pMsg->sId = htobe64(schMgmt.sId);      
+      pMsg->queryId = htobe64(job->queryId);
+      pMsg->taskId = htobe64(task->taskId);      
+      break;
+    }
+    case TDMT_VND_FETCH: {
+      if (NULL == task) {
+        SCH_ERR_RET(TSDB_CODE_QRY_APP_ERROR);
+      }
+      msgSize = sizeof(SResFetchMsg);
+      msg = calloc(1, msgSize);
+      if (NULL == msg) {
+        qError("calloc %d failed", msgSize);
+        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
+      }
+    
+      SResFetchMsg *pMsg = msg;
+      pMsg->sId = htobe64(schMgmt.sId);      
+      pMsg->queryId = htobe64(job->queryId);
+      pMsg->taskId = htobe64(task->taskId);      
+      break;
+    }
+    case TDMT_VND_DROP_TASK:{
+      msgSize = sizeof(STaskDropMsg);
+      msg = calloc(1, msgSize);
+      if (NULL == msg) {
+        qError("calloc %d failed", msgSize);
+        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
+      }
+    
+      STaskDropMsg *pMsg = msg;
+      pMsg->sId = htobe64(schMgmt.sId);      
+      pMsg->queryId = htobe64(job->queryId);
+      pMsg->taskId = htobe64(task->taskId);      
+      break;
+    }
+    default:
+      qError("unknown msg type:%d", msgType);
+      SCH_ERR_RET(TSDB_CODE_SCH_INTERNAL_ERROR);
+      break;
+  }
+
+  SCH_ERR_JRET(schAsyncSendMsg(job->transport, &task->plan->execEpSet, job->queryId, task->taskId, msgType, msg, msgSize));
+
+  return TSDB_CODE_SUCCESS;
+
+_return:
+
+  tfree(msg);
+  SCH_RET(code);
+}
 
 
 int32_t schLaunchTask(SSchJob *job, SSchTask *task) {
@@ -664,7 +748,7 @@ int32_t schLaunchTask(SSchJob *job, SSchTask *task) {
 
   int32_t msgType = (plan->type == QUERY_TYPE_MODIFY) ? TDMT_VND_SUBMIT : TDMT_VND_QUERY;
   
-  SCH_ERR_RET(schAsyncSendMsg(job, task, msgType));
+  SCH_ERR_RET(schBuildAndSendMsg(job, task, msgType));
 
   SCH_ERR_RET(schPushTaskToExecList(job, task));
 
@@ -673,6 +757,8 @@ int32_t schLaunchTask(SSchJob *job, SSchTask *task) {
   return TSDB_CODE_SUCCESS;
 }
 
+
+
 int32_t schLaunchJob(SSchJob *job) {
   SSchLevel *level = taosArrayGet(job->levels, job->levelIdx);
   for (int32_t i = 0; i < level->taskNum; ++i) {
@@ -690,7 +776,7 @@ void schDropJobAllTasks(SSchJob *job) {
   while (pIter) {
     SSchTask *task = *(SSchTask **)pIter;
   
-    schAsyncSendMsg(job, task, TDMT_VND_DROP_TASK);
+    schBuildAndSendMsg(job, task, TDMT_VND_DROP_TASK);
     
     pIter = taosHashIterate(job->succTasks, pIter);
   }  
@@ -699,7 +785,7 @@ void schDropJobAllTasks(SSchJob *job) {
   while (pIter) {
     SSchTask *task = *(SSchTask **)pIter;
   
-    schAsyncSendMsg(job, task, TDMT_VND_DROP_TASK);
+    schBuildAndSendMsg(job, task, TDMT_VND_DROP_TASK);
     
     pIter = taosHashIterate(job->succTasks, pIter);
   }  
@@ -717,7 +803,7 @@ int32_t schedulerInit(SSchedulerCfg *cfg) {
     SCH_ERR_LRET(TSDB_CODE_QRY_OUT_OF_MEMORY, "init %d schduler jobs failed", schMgmt.cfg.maxJobNum);
   }
 
-  schMgmt.schedulerId = 1; //TODO GENERATE A UUID
+  schMgmt.sId = 1; //TODO GENERATE A UUID
   
   return TSDB_CODE_SUCCESS;
 }

source/libs/scheduler/test/schedulerTests.cpp

@@ -36,7 +36,7 @@
 
 namespace {
 
-extern "C" int32_t schHandleRspMsg(SSchJob *job, SSchTask *task, int32_t msgType, char *msg, int32_t msgSize, int32_t rspCode);
+extern "C" int32_t schProcessRspMsg(SSchJob *job, SSchTask *task, int32_t msgType, char *msg, int32_t msgSize, int32_t rspCode);
 
 void schtBuildQueryDag(SQueryDag *dag) {
   uint64_t qId = 0x0000000000000001;
@@ -182,7 +182,7 @@ void *schtSendRsp(void *param) {
 
     SShellSubmitRspMsg rsp = {0};
     rsp.affectedRows = 10;
-    schHandleRspMsg(job, task, TDMT_VND_SUBMIT, (char *)&rsp, sizeof(rsp), 0);
+    schProcessRspMsg(job, task, TDMT_VND_SUBMIT, (char *)&rsp, sizeof(rsp), 0);
     
     pIter = taosHashIterate(job->execTasks, pIter);
   }    
@@ -227,7 +227,7 @@ TEST(queryTest, normalCase) {
     SSchTask *task = *(SSchTask **)pIter;
 
     SQueryTableRsp rsp = {0};
-    code = schHandleRspMsg(job, task, TDMT_VND_QUERY, (char *)&rsp, sizeof(rsp), 0);
+    code = schProcessRspMsg(job, task, TDMT_VND_QUERY, (char *)&rsp, sizeof(rsp), 0);
     
     ASSERT_EQ(code, 0);
     pIter = taosHashIterate(job->execTasks, pIter);
@@ -238,7 +238,7 @@ TEST(queryTest, normalCase) {
     SSchTask *task = *(SSchTask **)pIter;
 
     SResReadyRsp rsp = {0};
-    code = schHandleRspMsg(job, task, TDMT_VND_RES_READY, (char *)&rsp, sizeof(rsp), 0);
+    code = schProcessRspMsg(job, task, TDMT_VND_RES_READY, (char *)&rsp, sizeof(rsp), 0);
     
     ASSERT_EQ(code, 0);
     pIter = taosHashIterate(job->execTasks, pIter);
@@ -249,7 +249,7 @@ TEST(queryTest, normalCase) {
     SSchTask *task = *(SSchTask **)pIter;
 
     SQueryTableRsp rsp = {0};
-    code = schHandleRspMsg(job, task, TDMT_VND_QUERY, (char *)&rsp, sizeof(rsp), 0);
+    code = schProcessRspMsg(job, task, TDMT_VND_QUERY, (char *)&rsp, sizeof(rsp), 0);
     
     ASSERT_EQ(code, 0);
     pIter = taosHashIterate(job->execTasks, pIter);
@@ -260,7 +260,7 @@ TEST(queryTest, normalCase) {
     SSchTask *task = *(SSchTask **)pIter;
 
     SResReadyRsp rsp = {0};
-    code = schHandleRspMsg(job, task, TDMT_VND_RES_READY, (char *)&rsp, sizeof(rsp), 0);
+    code = schProcessRspMsg(job, task, TDMT_VND_RES_READY, (char *)&rsp, sizeof(rsp), 0);
     ASSERT_EQ(code, 0);
     
     pIter = taosHashIterate(job->execTasks, pIter);
@@ -269,7 +269,7 @@ TEST(queryTest, normalCase) {
   SRetrieveTableRsp rsp = {0};
   rsp.completed = 1;
   rsp.numOfRows = 10;
-  code = schHandleRspMsg(job, NULL, TDMT_VND_FETCH, (char *)&rsp, sizeof(rsp), 0);
+  code = schProcessRspMsg(job, NULL, TDMT_VND_FETCH, (char *)&rsp, sizeof(rsp), 0);
     
   ASSERT_EQ(code, 0);