Merge pull request #12563 from taosdata/feature/stream

enh(tmq): cascade drop

include/common/tmsg.h

@@ -2097,6 +2097,18 @@ enum {
   TOPIC_SUB_TYPE__TABLE,
 };
 
+typedef struct {
+  SMsgHead head;
+  int64_t  leftForVer;
+  int32_t  vgId;
+  int64_t  consumerId;
+  char     subKey[TSDB_SUBSCRIBE_KEY_LEN];
+} SMqVDeleteReq;
+
+typedef struct {
+  int8_t reserved;
+} SMqVDeleteRsp;
+
 typedef struct {
   int64_t leftForVer;
   int32_t vgId;
@@ -2532,11 +2544,15 @@ static FORCE_INLINE void tDeleteSMqAskEpRsp(SMqAskEpRsp* pRsp) {
 }
 
 typedef struct {
-  void* data;
+  int64_t streamId;
+  int32_t taskId;
+  int32_t sourceVg;
+  int64_t sourceVer;
+  SArray* data;  // SArray<SSDataBlock>
 } SStreamDispatchReq;
 
 typedef struct {
-  int8_t status;
+  int8_t inputStatus;
 } SStreamDispatchRsp;
 
 #define TD_AUTO_CREATE_TABLE 0x1

include/common/tmsgdef.h

@@ -178,6 +178,7 @@ enum {
   TD_DEF_MSG_TYPE(TDMT_VND_MQ_CONNECT, "vnode-mq-connect", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_MQ_DISCONNECT, "vnode-mq-disconnect", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_MQ_VG_CHANGE, "vnode-mq-vg-change", SMqRebVgReq, SMqRebVgRsp)
+  TD_DEF_MSG_TYPE(TDMT_VND_MQ_VG_DELETE, "vnode-mq-vg-delete", SMqVDeleteReq, SMqVDeleteRsp)
   TD_DEF_MSG_TYPE(TDMT_VND_RES_READY, "vnode-res-ready", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_TASKS_STATUS, "vnode-tasks-status", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_CANCEL_TASK, "vnode-cancel-task", NULL, NULL)

include/util/taoserror.h

@@ -281,6 +281,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_MND_SUBSCRIBE_NOT_EXIST       TAOS_DEF_ERROR_CODE(0, 0x03E8)
 #define TSDB_CODE_MND_OFFSET_NOT_EXIST          TAOS_DEF_ERROR_CODE(0, 0x03E9)
 #define TSDB_CODE_MND_CONSUMER_NOT_READY        TAOS_DEF_ERROR_CODE(0, 0x03EA)
+#define TSDB_CODE_MND_TOPIC_SUBSCRIBED          TAOS_DEF_ERROR_CODE(0, 0x03EB)
 
 // mnode-stream
 #define TSDB_CODE_MND_STREAM_ALREADY_EXIST      TAOS_DEF_ERROR_CODE(0, 0x03F0)

source/dnode/mgmt/mgmt_mnode/src/mmHandle.c

@@ -216,6 +216,7 @@ SArray *mmGetMsgHandles() {
   if (dmSetMgmtHandle(pArray, TDMT_MND_MQ_COMMIT_OFFSET, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_MND_MQ_ASK_EP, mmPutNodeMsgToReadQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_MQ_VG_CHANGE_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_VND_MQ_VG_DELETE_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_MND_CREATE_STREAM, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_TASK_DEPLOY_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_MND_GET_DB_CFG, mmPutNodeMsgToReadQueue, 0) == NULL) goto _OVER;

source/dnode/mgmt/mgmt_vnode/src/vmHandle.c

@@ -308,6 +308,7 @@ SArray *vmGetMsgHandles() {
   if (dmSetMgmtHandle(pArray, TDMT_VND_DROP_SMA, vmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_SUBMIT_RSMA, vmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_MQ_VG_CHANGE, vmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_VND_MQ_VG_DELETE, vmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_CONSUME, vmPutNodeMsgToFetchQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_TASK_DEPLOY, vmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_QUERY_HEARTBEAT, vmPutNodeMsgToFetchQueue, 0) == NULL) goto _OVER;

source/dnode/mnode/impl/inc/mndDef.h

@@ -459,6 +459,7 @@ typedef struct {
   char*          ast;
   char*          physicalPlan;
   SSchemaWrapper schema;
+  int32_t        refConsumerCnt;
 } SMqTopicObj;
 
 typedef struct {

source/dnode/mnode/impl/inc/mndOffset.h

@@ -38,6 +38,9 @@ static FORCE_INLINE int32_t mndMakePartitionKey(char *key, const char *cgroup, c
 }
 
 int32_t mndDropOffsetByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb);
+int32_t mndDropOffsetByTopic(SMnode *pMnode, STrans *pTrans, const char *topic);
+
+bool mndOffsetFromTopic(SMqOffsetObj *pOffset, const char *topic);
 
 #ifdef __cplusplus
 }

source/dnode/mnode/impl/inc/mndSubscribe.h

@@ -32,6 +32,7 @@ void             mndReleaseSubscribe(SMnode *pMnode, SMqSubscribeObj *pSub);
 int32_t mndMakeSubscribeKey(char *key, const char *cgroup, const char *topicName);
 
 int32_t mndDropSubByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb);
+int32_t mndDropSubByTopic(SMnode *pMnode, STrans *pTrans, const char *topic);
 
 #ifdef __cplusplus
 }

source/dnode/mnode/impl/inc/mndTopic.h

@@ -35,6 +35,8 @@ int32_t mndDropTopicByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb);
 
 const char *mndTopicGetShowName(const char topic[TSDB_TOPIC_FNAME_LEN]);
 
+int32_t mndSetTopicRedoLogs(SMnode *pMnode, STrans *pTrans, SMqTopicObj *pTopic);
+
 #ifdef __cplusplus
 }
 #endif

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

@@ -399,6 +399,9 @@ static int32_t mndProcessSubscribeReq(SNodeMsg *pMsg) {
 
   int32_t newTopicNum = taosArrayGetSize(newSub);
   // check topic existance
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_TYPE_SUBSCRIBE, &pMsg->rpcMsg);
+  if (pTrans == NULL) goto SUBSCRIBE_OVER;
+
   for (int32_t i = 0; i < newTopicNum; i++) {
     char        *topic = taosArrayGetP(newSub, i);
     SMqTopicObj *pTopic = mndAcquireTopic(pMnode, topic);
@@ -406,7 +409,14 @@ static int32_t mndProcessSubscribeReq(SNodeMsg *pMsg) {
       terrno = TSDB_CODE_MND_TOPIC_NOT_EXIST;
       goto SUBSCRIBE_OVER;
     }
-    // TODO lock topic to prevent drop
+
+    // ref topic to prevent drop
+    // TODO make topic complete
+    SMqTopicObj topicObj = {0};
+    memcpy(&topicObj, pTopic, sizeof(SMqTopicObj));
+    topicObj.refConsumerCnt = pTopic->refConsumerCnt + 1;
+    if (mndSetTopicRedoLogs(pMnode, pTrans, &topicObj) != 0) goto SUBSCRIBE_OVER;
+
     mndReleaseTopic(pMnode, pTopic);
   }
 
@@ -422,8 +432,6 @@ static int32_t mndProcessSubscribeReq(SNodeMsg *pMsg) {
       taosArrayPush(pConsumerNew->assignedTopics, &newTopicCopy);
     }
 
-    STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_TYPE_SUBSCRIBE, &pMsg->rpcMsg);
-    if (pTrans == NULL) goto SUBSCRIBE_OVER;
     if (mndSetConsumerCommitLogs(pMnode, pTrans, pConsumerNew) != 0) goto SUBSCRIBE_OVER;
     if (mndTransPrepare(pMnode, pTrans) != 0) goto SUBSCRIBE_OVER;
 
@@ -494,8 +502,6 @@ static int32_t mndProcessSubscribeReq(SNodeMsg *pMsg) {
       goto SUBSCRIBE_OVER;
     }
 
-    STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_TYPE_SUBSCRIBE, &pMsg->rpcMsg);
-    if (pTrans == NULL) goto SUBSCRIBE_OVER;
     if (mndSetConsumerCommitLogs(pMnode, pTrans, pConsumerNew) != 0) goto SUBSCRIBE_OVER;
     if (mndTransPrepare(pMnode, pTrans) != 0) goto SUBSCRIBE_OVER;
   }
@@ -503,6 +509,8 @@ static int32_t mndProcessSubscribeReq(SNodeMsg *pMsg) {
   code = TSDB_CODE_MND_ACTION_IN_PROGRESS;
 
 SUBSCRIBE_OVER:
+  mndTransDrop(pTrans);
+
   if (pConsumerOld) {
     /*taosRUnLockLatch(&pConsumerOld->lock);*/
     mndReleaseConsumer(pMnode, pConsumerOld);

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

@@ -50,6 +50,14 @@ int32_t mndInitOffset(SMnode *pMnode) {
 
 void mndCleanupOffset(SMnode *pMnode) {}
 
+bool mndOffsetFromTopic(SMqOffsetObj *pOffset, const char *topic) {
+  int32_t i = 0;
+  while (pOffset->key[i] != ':') i++;
+  while (pOffset->key[i] != ':') i++;
+  if (strcmp(&pOffset->key[i + 1], topic) == 0) return true;
+  return false;
+}
+
 SSdbRaw *mndOffsetActionEncode(SMqOffsetObj *pOffset) {
   terrno = TSDB_CODE_OUT_OF_MEMORY;
   void   *buf = NULL;
@@ -134,10 +142,11 @@ int32_t mndCreateOffsets(STrans *pTrans, const char *cgroup, const char *topicNa
   int32_t sz = taosArrayGetSize(vgs);
   for (int32_t i = 0; i < sz; i++) {
     int32_t      vgId = *(int32_t *)taosArrayGet(vgs, i);
-    SMqOffsetObj offsetObj;
+    SMqOffsetObj offsetObj = {0};
     if (mndMakePartitionKey(offsetObj.key, cgroup, topicName, vgId) < 0) {
       return -1;
     }
+    // TODO assign db
     offsetObj.offset = -1;
     SSdbRaw *pOffsetRaw = mndOffsetActionEncode(&offsetObj);
     if (pOffsetRaw == NULL) {
@@ -240,6 +249,14 @@ static int32_t mndSetDropOffsetCommitLogs(SMnode *pMnode, STrans *pTrans, SMqOff
   return 0;
 }
 
+static int32_t mndSetDropOffsetRedoLogs(SMnode *pMnode, STrans *pTrans, SMqOffsetObj *pOffset) {
+  SSdbRaw *pRedoRaw = mndOffsetActionEncode(pOffset);
+  if (pRedoRaw == NULL) return -1;
+  if (mndTransAppendRedolog(pTrans, pRedoRaw) != 0) return -1;
+  if (sdbSetRawStatus(pRedoRaw, SDB_STATUS_DROPPED) != 0) return -1;
+  return 0;
+}
+
 int32_t mndDropOffsetByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb) {
   int32_t code = -1;
   SSdb   *pSdb = pMnode->pSdb;
@@ -247,7 +264,7 @@ int32_t mndDropOffsetByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb) {
   void         *pIter = NULL;
   SMqOffsetObj *pOffset = NULL;
   while (1) {
-    pIter = sdbFetch(pSdb, SDB_SUBSCRIBE, pIter, (void **)&pOffset);
+    pIter = sdbFetch(pSdb, SDB_OFFSET, pIter, (void **)&pOffset);
     if (pIter == NULL) break;
 
     if (pOffset->dbUid != pDb->uid) {
@@ -256,8 +273,39 @@ int32_t mndDropOffsetByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb) {
     }
 
     if (mndSetDropOffsetCommitLogs(pMnode, pTrans, pOffset) < 0) {
+      sdbRelease(pSdb, pOffset);
+      goto END;
+    }
+
+    sdbRelease(pSdb, pOffset);
+  }
+
+  code = 0;
+END:
+  return code;
+}
+
+int32_t mndDropOffsetByTopic(SMnode *pMnode, STrans *pTrans, const char *topic) {
+  int32_t code = -1;
+  SSdb   *pSdb = pMnode->pSdb;
+
+  void         *pIter = NULL;
+  SMqOffsetObj *pOffset = NULL;
+  while (1) {
+    pIter = sdbFetch(pSdb, SDB_OFFSET, pIter, (void **)&pOffset);
+    if (pIter == NULL) break;
+
+    if (!mndOffsetFromTopic(pOffset, topic)) {
+      sdbRelease(pSdb, pOffset);
+      continue;
+    }
+
+    if (mndSetDropOffsetRedoLogs(pMnode, pTrans, pOffset) < 0) {
+      sdbRelease(pSdb, pOffset);
       goto END;
     }
+
+    sdbRelease(pSdb, pOffset);
   }
 
   code = 0;

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

@@ -73,6 +73,7 @@ int32_t mndInitSubscribe(SMnode *pMnode) {
                      .deleteFp = (SdbDeleteFp)mndSubActionDelete};
 
   mndSetMsgHandle(pMnode, TDMT_VND_MQ_VG_CHANGE_RSP, mndProcessSubscribeInternalRsp);
+  mndSetMsgHandle(pMnode, TDMT_VND_MQ_VG_DELETE_RSP, mndProcessSubscribeInternalRsp);
   mndSetMsgHandle(pMnode, TDMT_MND_MQ_DO_REBALANCE, mndProcessRebalanceReq);
 
   mndAddShowRetrieveHandle(pMnode, TSDB_MGMT_TABLE_SUBSCRIPTIONS, mndRetrieveSubscribe);
@@ -389,7 +390,7 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
 }
 
 static int32_t mndPersistRebResult(SMnode *pMnode, SNodeMsg *pMsg, const SMqRebOutputObj *pOutput) {
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_TYPE_REBALANCE, &pMsg->rpcMsg);
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_TYPE_REBALANCE, &pMsg->rpcMsg);
   if (pTrans == NULL) {
     return -1;
   }
@@ -458,6 +459,20 @@ static int32_t mndPersistRebResult(SMnode *pMnode, SNodeMsg *pMsg, const SMqRebO
       goto REB_FAIL;
     }
   }
+  if (consumerNum) {
+    char topic[TSDB_TOPIC_FNAME_LEN];
+    char cgroup[TSDB_CGROUP_LEN];
+    mndSplitSubscribeKey(pOutput->pSub->key, topic, cgroup, true);
+    SMqTopicObj *pTopic = mndAcquireTopic(pMnode, topic);
+    if (pTopic) {
+      // TODO make topic complete
+      SMqTopicObj topicObj = {0};
+      memcpy(&topicObj, pTopic, sizeof(SMqTopicObj));
+      topicObj.refConsumerCnt = pTopic->refConsumerCnt - consumerNum;
+      if (mndSetTopicRedoLogs(pMnode, pTrans, &topicObj) != 0) goto REB_FAIL;
+    }
+  }
+
   // 4. TODO commit log: modification log
 
   // 5. set cb
@@ -688,6 +703,14 @@ static int32_t mndProcessSubscribeInternalRsp(SNodeMsg *pRsp) {
   return 0;
 }
 
+static int32_t mndSetDropSubRedoLogs(SMnode *pMnode, STrans *pTrans, SMqSubscribeObj *pSub) {
+  SSdbRaw *pRedoRaw = mndSubActionEncode(pSub);
+  if (pRedoRaw == NULL) return -1;
+  if (mndTransAppendRedolog(pTrans, pRedoRaw) != 0) return -1;
+  if (sdbSetRawStatus(pRedoRaw, SDB_STATUS_DROPPED) != 0) return -1;
+  return 0;
+}
+
 static int32_t mndSetDropSubCommitLogs(SMnode *pMnode, STrans *pTrans, SMqSubscribeObj *pSub) {
   SSdbRaw *pCommitRaw = mndSubActionEncode(pSub);
   if (pCommitRaw == NULL) return -1;
@@ -712,6 +735,57 @@ int32_t mndDropSubByDB(SMnode *pMnode, STrans *pTrans, SDbObj *pDb) {
     }
 
     if (mndSetDropSubCommitLogs(pMnode, pTrans, pSub) < 0) {
+      sdbRelease(pSdb, pSub);
+      goto END;
+    }
+  }
+
+  code = 0;
+END:
+  return code;
+}
+
+int32_t mndDropSubByTopic(SMnode *pMnode, STrans *pTrans, const char *topicName) {
+  int32_t code = -1;
+  SSdb   *pSdb = pMnode->pSdb;
+
+  void            *pIter = NULL;
+  SMqSubscribeObj *pSub = NULL;
+  while (1) {
+    pIter = sdbFetch(pSdb, SDB_SUBSCRIBE, pIter, (void **)&pSub);
+    if (pIter == NULL) break;
+
+    char topic[TSDB_TOPIC_FNAME_LEN];
+    char cgroup[TSDB_CGROUP_LEN];
+    mndSplitSubscribeKey(pSub->key, topic, cgroup, true);
+    if (strcmp(topic, topicName) != 0) {
+      sdbRelease(pSdb, pSub);
+      continue;
+    }
+
+    // iter all vnode to delete handle
+    ASSERT(taosHashGetSize(pSub->consumerHash) == 0);
+    int32_t sz = taosArrayGetSize(pSub->unassignedVgs);
+    for (int32_t i = 0; i < sz; i++) {
+      SMqVgEp       *pVgEp = taosArrayGetP(pSub->unassignedVgs, i);
+      SMqVDeleteReq *pReq = taosMemoryCalloc(1, sizeof(SMqVDeleteReq));
+      pReq->head.vgId = htonl(pVgEp->vgId);
+      pReq->vgId = pVgEp->vgId;
+      pReq->consumerId = -1;
+      memcpy(pReq->subKey, pSub->key, TSDB_SUBSCRIBE_KEY_LEN);
+      STransAction action = {0};
+      action.epSet = pVgEp->epSet;
+      action.pCont = pReq;
+      action.contLen = sizeof(SMqVDeleteReq);
+      action.msgType = TDMT_VND_MQ_VG_DELETE;
+      if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+        taosMemoryFree(pReq);
+        return -1;
+      }
+    }
+
+    if (mndSetDropSubRedoLogs(pMnode, pTrans, pSub) < 0) {
+      sdbRelease(pSdb, pSub);
       goto END;
     }
   }

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

@@ -18,8 +18,10 @@
 #include "mndDb.h"
 #include "mndDnode.h"
 #include "mndMnode.h"
+#include "mndOffset.h"
 #include "mndShow.h"
 #include "mndStb.h"
+#include "mndSubscribe.h"
 #include "mndTrans.h"
 #include "mndUser.h"
 #include "mndVgroup.h"
@@ -106,6 +108,7 @@ SSdbRaw *mndTopicActionEncode(SMqTopicObj *pTopic) {
   taosEncodeSSchemaWrapper(&aswBuf, &pTopic->schema);
   SDB_SET_INT32(pRaw, dataPos, schemaLen, TOPIC_ENCODE_OVER);
   SDB_SET_BINARY(pRaw, dataPos, swBuf, schemaLen, TOPIC_ENCODE_OVER);
+  SDB_SET_INT32(pRaw, dataPos, pTopic->refConsumerCnt, TOPIC_ENCODE_OVER);
 
   SDB_SET_RESERVE(pRaw, dataPos, MND_TOPIC_RESERVE_SIZE, TOPIC_ENCODE_OVER);
   SDB_SET_DATALEN(pRaw, dataPos, TOPIC_ENCODE_OVER);
@@ -190,6 +193,8 @@ SSdbRow *mndTopicActionDecode(SSdbRaw *pRaw) {
     goto TOPIC_DECODE_OVER;
   }
 
+  SDB_GET_INT32(pRaw, dataPos, &pTopic->refConsumerCnt, TOPIC_DECODE_OVER);
+
   SDB_GET_RESERVE(pRaw, dataPos, MND_TOPIC_RESERVE_SIZE, TOPIC_DECODE_OVER);
 
   terrno = TSDB_CODE_SUCCESS;
@@ -220,11 +225,13 @@ static int32_t mndTopicActionUpdate(SSdb *pSdb, SMqTopicObj *pOldTopic, SMqTopic
   atomic_exchange_64(&pOldTopic->updateTime, pNewTopic->updateTime);
   atomic_exchange_32(&pOldTopic->version, pNewTopic->version);
 
-  taosWLockLatch(&pOldTopic->lock);
+  atomic_store_32(&pOldTopic->refConsumerCnt, pNewTopic->refConsumerCnt);
+
+  /*taosWLockLatch(&pOldTopic->lock);*/
 
   // TODO handle update
 
-  taosWUnLockLatch(&pOldTopic->lock);
+  /*taosWUnLockLatch(&pOldTopic->lock);*/
   return 0;
 }
 
@@ -292,6 +299,7 @@ static int32_t mndCreateTopic(SMnode *pMnode, SNodeMsg *pReq, SCMCreateTopicReq
   topicObj.version = 1;
   topicObj.sql = strdup(pCreate->sql);
   topicObj.sqlLen = strlen(pCreate->sql) + 1;
+  topicObj.refConsumerCnt = 0;
 
   if (pCreate->ast && pCreate->ast[0]) {
     topicObj.ast = strdup(pCreate->ast);
@@ -436,15 +444,7 @@ CREATE_TOPIC_OVER:
   return code;
 }
 
-static int32_t mndDropTopic(SMnode *pMnode, SNodeMsg *pReq, SMqTopicObj *pTopic) {
-  // TODO: cannot drop when subscribed
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_TYPE_DROP_TOPIC, &pReq->rpcMsg);
-  if (pTrans == NULL) {
-    mError("topic:%s, failed to drop since %s", pTopic->name, terrstr());
-    return -1;
-  }
-  mDebug("trans:%d, used to drop topic:%s", pTrans->id, pTopic->name);
-
+static int32_t mndDropTopic(SMnode *pMnode, STrans *pTrans, SNodeMsg *pReq, SMqTopicObj *pTopic) {
   SSdbRaw *pRedoRaw = mndTopicActionEncode(pTopic);
   if (pRedoRaw == NULL || mndTransAppendRedolog(pTrans, pRedoRaw) != 0) {
     mError("trans:%d, failed to append redo log since %s", pTrans->id, terrstr());
@@ -465,6 +465,7 @@ static int32_t mndDropTopic(SMnode *pMnode, SNodeMsg *pReq, SMqTopicObj *pTopic)
 
 static int32_t mndProcessDropTopicReq(SNodeMsg *pReq) {
   SMnode        *pMnode = pReq->pNode;
+  SSdb          *pSdb = pMnode->pSdb;
   SMDropTopicReq dropReq = {0};
 
   if (tDeserializeSMDropTopicReq(pReq->rpcMsg.pCont, pReq->rpcMsg.contLen, &dropReq) != 0) {
@@ -485,10 +486,35 @@ static int32_t mndProcessDropTopicReq(SNodeMsg *pReq) {
       return -1;
     }
   }
-  // TODO: check ref
 
-  int32_t code = mndDropTopic(pMnode, pReq, pTopic);
-  // TODO: iterate and drop related subscriptions and offsets
+  // check ref
+  if (pTopic->refConsumerCnt != 0) {
+    terrno = TSDB_CODE_MND_TOPIC_SUBSCRIBED;
+    mError("topic:%s, failed to drop since %s", dropReq.name, terrstr());
+    return -1;
+  }
+
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_TYPE_DROP_TOPIC, &pReq->rpcMsg);
+  if (pTrans == NULL) {
+    mError("topic:%s, failed to drop since %s", pTopic->name, terrstr());
+    return -1;
+  }
+
+  mDebug("trans:%d, used to drop topic:%s", pTrans->id, pTopic->name);
+
+#if 1
+  if (mndDropOffsetByTopic(pMnode, pTrans, dropReq.name) < 0) {
+    ASSERT(0);
+    return -1;
+  }
+#endif
+
+  if (mndDropSubByTopic(pMnode, pTrans, dropReq.name) < 0) {
+    ASSERT(0);
+    return -1;
+  }
+
+  int32_t code = mndDropTopic(pMnode, pTrans, pReq, pTopic);
   mndReleaseTopic(pMnode, pTopic);
 
   if (code != 0) {
@@ -577,6 +603,15 @@ static int32_t mndRetrieveTopic(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock *pB
   return numOfRows;
 }
 
+int32_t mndSetTopicRedoLogs(SMnode *pMnode, STrans *pTrans, SMqTopicObj *pTopic) {
+  SSdbRaw *pRedoRaw = mndTopicActionEncode(pTopic);
+  if (pRedoRaw == NULL) return -1;
+  if (mndTransAppendCommitlog(pTrans, pRedoRaw) != 0) return -1;
+  if (sdbSetRawStatus(pRedoRaw, SDB_STATUS_READY) != 0) return -1;
+
+  return 0;
+}
+
 static int32_t mndSetDropTopicCommitLogs(SMnode *pMnode, STrans *pTrans, SMqTopicObj *pTopic) {
   SSdbRaw *pCommitRaw = mndTopicActionEncode(pTopic);
   if (pCommitRaw == NULL) return -1;

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

@@ -116,6 +116,7 @@ void    tqClose(STQ*);
 int     tqPushMsg(STQ*, void* msg, int32_t msgLen, tmsg_t msgType, int64_t ver);
 int     tqCommit(STQ*);
 int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen);
+int32_t tqProcessVgDeleteReq(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 workerId);

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

@@ -861,6 +861,14 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) {
 }
 #endif
 
+int32_t tqProcessVgDeleteReq(STQ* pTq, char* msg, int32_t msgLen) {
+  SMqVDeleteReq* pReq = (SMqVDeleteReq*)msg;
+
+  int32_t code = taosHashRemove(pTq->execs, pReq->subKey, strlen(pReq->subKey));
+  ASSERT(code == 0);
+  return 0;
+}
+
 // TODO: persist meta into tdb
 int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
   SMqRebVgReq req = {0};
@@ -1087,35 +1095,6 @@ int32_t tqProcessStreamTrigger2(STQ* pTq, SSubmitReq* pReq, int64_t ver) {
 }
 
 int32_t tqProcessTaskExec2(STQ* pTq, char* msg, int32_t msgLen) {
-  SStreamTaskExecReq req = {0};
-  tDecodeSStreamTaskExecReq(msg, &req);
-  int32_t taskId = req.taskId;
-
-  SStreamTask* pTask = taosHashGet(pTq->pStreamTasks, &taskId, sizeof(int32_t));
-  ASSERT(pTask);
-  ASSERT(pTask->inputType == TASK_INPUT_TYPE__DATA_BLOCK);
-
-  // enqueue
-  int32_t inputStatus = streamEnqueueDataBlk(pTask, (SStreamDataBlock*)req.data);
-  if (inputStatus == TASK_INPUT_STATUS__BLOCKED) {
-    // TODO rsp blocked
-    return 0;
-  }
-
-  // try exec
-  int8_t execStatus = atomic_val_compare_exchange_8(&pTask->status, TASK_STATUS__IDLE, TASK_STATUS__EXECUTING);
-  if (execStatus == TASK_STATUS__IDLE) {
-    if (streamTaskRun(pTask) < 0) {
-      atomic_store_8(&pTask->status, TASK_STATUS__CLOSING);
-
-      goto FAIL;
-    }
-  } else if (execStatus == TASK_STATUS__EXECUTING) {
-    return 0;
-  }
-
-  // TODO rsp success
+  //
   return 0;
-FAIL:
-  return -1;
 }

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

@@ -101,6 +101,11 @@ int vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRpcMsg
         // TODO: handle error
       }
       break;
+    case TDMT_VND_MQ_VG_DELETE:
+      if (tqProcessVgDeleteReq(pVnode->pTq, pMsg->pCont, pMsg->contLen) < 0) {
+        // TODO: handle error
+      }
+      break;
     case TDMT_VND_TASK_DEPLOY: {
       if (tqProcessTaskDeploy(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)),
                               pMsg->contLen - sizeof(SMsgHead)) < 0) {

source/libs/stream/src/tstream.c

@@ -35,7 +35,7 @@ void* streamDataBlockDecode(const void* buf, SStreamDataBlock* pInput) {
   return (void*)buf;
 }
 
-static int32_t streamBuildDispatchMsg(SStreamTask* pTask, SArray* data, SRpcMsg* pMsg, SEpSet** ppEpSet) {
+static int32_t streamBuildExecMsg(SStreamTask* pTask, SArray* data, SRpcMsg* pMsg, SEpSet** ppEpSet) {
   SStreamTaskExecReq req = {
       .streamId = pTask->streamId,
       .data = data,
@@ -107,7 +107,7 @@ static int32_t streamShuffleDispatch(SStreamTask* pTask, SMsgCb* pMsgCb, SHashOb
     SArray* pData = *(SArray**)pIter;
     SRpcMsg dispatchMsg = {0};
     SEpSet* pEpSet;
-    if (streamBuildDispatchMsg(pTask, pData, &dispatchMsg, &pEpSet) < 0) {
+    if (streamBuildExecMsg(pTask, pData, &dispatchMsg, &pEpSet) < 0) {
       ASSERT(0);
       return -1;
     }
@@ -133,7 +133,7 @@ int32_t streamEnqueueDataBlk(SStreamTask* pTask, SStreamDataBlock* input) {
   return inputStatus;
 }
 
-int32_t streamTaskExecImpl(SStreamTask* pTask, void* data, SArray* pRes) {
+static int32_t streamTaskExecImpl(SStreamTask* pTask, void* data, SArray* pRes) {
   void* exec = pTask->exec.runners[0].executor;
 
   // set input
@@ -265,87 +265,42 @@ FAIL:
   return -1;
 }
 
-int32_t streamTaskDispatchDown(SStreamTask* pTask, SMsgCb* pMsgCb) {
-  //
-  return 0;
-}
-
-int32_t streamTaskSink(SStreamTask* pTask) {
-  //
-  return 0;
-}
-
-int32_t streamTaskProcessInputReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDataBlock* pBlock, SRpcMsg* pRsp) {
-  // 1. handle input
-  // 1.1 enqueue
-  taosWriteQitem(pTask->inputQ, pBlock);
-  // 1.2 calc back pressure
-  // 1.3 rsp by input status
-  int8_t              inputStatus = atomic_load_8(&pTask->inputStatus);
-  SStreamDispatchRsp* pCont = rpcMallocCont(sizeof(SStreamDispatchRsp));
-  pCont->status = inputStatus;
-  pRsp->pCont = pCont;
-  pRsp->contLen = sizeof(SStreamDispatchRsp);
-  tmsgSendRsp(pRsp);
-  // 2. try exec
-  // 2.1. idle: exec
-  // 2.2. executing: return
-  // 2.3. closing: keep trying
+int32_t streamTaskSink(SStreamTask* pTask, SMsgCb* pMsgCb) {
+  bool firstRun = 1;
   while (1) {
-    int8_t execStatus = atomic_val_compare_exchange_8(&pTask->status, TASK_STATUS__IDLE, TASK_STATUS__EXECUTING);
-    if (execStatus == TASK_STATUS__IDLE) {
-      void*         exec = pTask->exec.runners[0].executor;
-      SArray*       pRes = taosArrayInit(0, sizeof(void*));
-      const SArray* blocks = pBlock->blocks;
-      qSetMultiStreamInput(exec, blocks->pData, blocks->size, STREAM_DATA_TYPE_SSDATA_BLOCK);
-      while (1) {
-        SSDataBlock* output;
-        uint64_t     ts = 0;
-        if (qExecTask(exec, &output, &ts) < 0) {
-          ASSERT(false);
-        }
-        if (output == NULL) break;
-        taosArrayPush(pRes, &output);
-      }
-      // TODO: wrap destroy block
-      taosArrayDestroyP(pBlock->blocks, (FDelete)blockDataDestroy);
-
-      if (taosArrayGetSize(pRes) != 0) {
-        SArray** resQ = taosAllocateQitem(sizeof(void**), DEF_QITEM);
-        *resQ = pRes;
-        taosWriteQitem(pTask->outputQ, resQ);
+    SStreamDataBlock* pBlock = NULL;
+    if (!firstRun) {
+      taosReadAllQitems(pTask->outputQ, pTask->outputQAll);
+    }
+    taosGetQitem(pTask->outputQAll, (void**)&pBlock);
+    if (pBlock == NULL) {
+      if (firstRun) {
+        firstRun = 0;
+        continue;
+      } else {
+        break;
       }
-
-    } else if (execStatus == TASK_STATUS__CLOSING) {
-      continue;
-    } else if (execStatus == TASK_STATUS__EXECUTING)
-      break;
-    else {
-      ASSERT(0);
     }
-  }
-  // 3. handle output
-  // 3.1 check and set status
-  // 3.2 dispatch / sink
-  STaosQall* qall = taosAllocateQall();
-  taosReadAllQitems(pTask->outputQ, qall);
-  SArray** ppRes = NULL;
-  while (1) {
-    taosGetQitem(qall, (void**)&ppRes);
-    if (ppRes == NULL) break;
 
-    SArray* pRes = *ppRes;
+    SArray* pRes = pBlock->blocks;
+
+    // sink
     if (pTask->sinkType == TASK_SINK__TABLE) {
-      pTask->tbSink.tbSinkFunc(pTask, pTask->tbSink.vnode, pBlock->sourceVer, pRes);
+      // blockDebugShowData(pRes);
+      pTask->tbSink.tbSinkFunc(pTask, pTask->tbSink.vnode, 0, pRes);
     } else if (pTask->sinkType == TASK_SINK__SMA) {
       pTask->smaSink.smaSink(pTask->ahandle, pTask->smaSink.smaId, pRes);
+      //
+    } else if (pTask->sinkType == TASK_SINK__FETCH) {
+      //
     } else {
+      ASSERT(pTask->sinkType == TASK_SINK__NONE);
     }
 
     // dispatch
     if (pTask->dispatchType == TASK_DISPATCH__INPLACE) {
       SRpcMsg dispatchMsg = {0};
-      if (streamBuildDispatchMsg(pTask, pRes, &dispatchMsg, NULL) < 0) {
+      if (streamBuildExecMsg(pTask, pRes, &dispatchMsg, NULL) < 0) {
         ASSERT(0);
         return -1;
       }
@@ -366,7 +321,7 @@ int32_t streamTaskProcessInputReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDat
     } else if (pTask->dispatchType == TASK_DISPATCH__FIXED) {
       SRpcMsg dispatchMsg = {0};
       SEpSet* pEpSet = NULL;
-      if (streamBuildDispatchMsg(pTask, pRes, &dispatchMsg, &pEpSet) < 0) {
+      if (streamBuildExecMsg(pTask, pRes, &dispatchMsg, &pEpSet) < 0) {
         ASSERT(0);
         return -1;
       }
@@ -401,75 +356,58 @@ int32_t streamTaskProcessInputReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDat
       ASSERT(pTask->dispatchType == TASK_DISPATCH__NONE);
     }
   }
-  //
   return 0;
 }
 
-int32_t streamTaskProcessDispatchRsp(SStreamTask* pTask, char* msg, int32_t msgLen) {
-  //
-  return 0;
-}
+int32_t streamTaskEnqueue(SStreamTask* pTask, SStreamDispatchReq* pReq, SRpcMsg* pRsp) {
+  SStreamDataBlock* pBlock = taosAllocateQitem(sizeof(SStreamDataBlock), DEF_QITEM);
+  int8_t            status;
+
+  // 1.1 update status
+  // TODO cal backpressure
+  if (pBlock == NULL) {
+    atomic_store_8(&pTask->inputStatus, TASK_INPUT_STATUS__FAILED);
+    status = TASK_INPUT_STATUS__FAILED;
+  } else {
+    status = atomic_load_8(&pTask->inputStatus);
+  }
+
+  // 1.2 enqueue
+  pBlock->type = STREAM_DATA_TYPE_SSDATA_BLOCK;
+  pBlock->sourceVg = pReq->sourceVg;
+  pBlock->sourceVer = pReq->sourceVer;
+  taosWriteQitem(pTask->inputQ, pBlock);
+
+  // 1.3 rsp by input status
+  SStreamDispatchRsp* pCont = rpcMallocCont(sizeof(SStreamDispatchRsp));
+  pCont->inputStatus = status;
+  pRsp->pCont = pCont;
+  pRsp->contLen = sizeof(SStreamDispatchRsp);
+  tmsgSendRsp(pRsp);
 
-int32_t streamTaskProcessRecoverReq(SStreamTask* pTask, char* msg) {
-  //
   return 0;
 }
 
-int32_t streamTaskRun(SStreamTask* pTask) {
-  SArray* pRes = NULL;
-  if (pTask->execType == TASK_EXEC__PIPE || pTask->execType == TASK_EXEC__MERGE) {
-    // TODO remove multi runner
-    void* exec = pTask->exec.runners[0].executor;
-
-    int8_t status = atomic_val_compare_exchange_8(&pTask->status, TASK_STATUS__IDLE, TASK_STATUS__EXECUTING);
-    if (status == TASK_STATUS__IDLE) {
-      pRes = taosArrayInit(0, sizeof(void*));
-      if (pRes == NULL) {
-        return -1;
-      }
+int32_t streamTaskProcessDispatchReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDispatchReq* pReq, SRpcMsg* pRsp) {
+  // 1. handle input
+  streamTaskEnqueue(pTask, pReq, pRsp);
 
-      void* input = NULL;
-      taosWriteQitem(pTask->inputQ, &input);
-      if (input == NULL) return 0;
-
-      // TODO: fix type
-      if (pTask->sourceType == TASK_SOURCE__SCAN) {
-        SStreamDataSubmit* pSubmit = (SStreamDataSubmit*)input;
-        qSetStreamInput(exec, pSubmit->data, STREAM_DATA_TYPE_SUBMIT_BLOCK);
-        while (1) {
-          SSDataBlock* output;
-          uint64_t     ts = 0;
-          if (qExecTask(exec, &output, &ts) < 0) {
-            ASSERT(false);
-          }
-          if (output == NULL) break;
-          taosArrayPush(pRes, &output);
-        }
-        streamDataSubmitRefDec(pSubmit);
-      } else {
-        SStreamDataBlock* pStreamBlock = (SStreamDataBlock*)input;
-        const SArray*     blocks = pStreamBlock->blocks;
-        qSetMultiStreamInput(exec, blocks->pData, blocks->size, STREAM_DATA_TYPE_SSDATA_BLOCK);
-        while (1) {
-          SSDataBlock* output;
-          uint64_t     ts = 0;
-          if (qExecTask(exec, &output, &ts) < 0) {
-            ASSERT(false);
-          }
-          if (output == NULL) break;
-          taosArrayPush(pRes, &output);
-        }
-        // TODO: wrap destroy block
-        taosArrayDestroyP(pStreamBlock->blocks, (FDelete)blockDataDestroy);
-      }
+  // 2. try exec
+  // 2.1. idle: exec
+  // 2.2. executing: return
+  // 2.3. closing: keep trying
+  streamTaskExec2(pTask, pMsgCb);
 
-      if (taosArrayGetSize(pRes) != 0) {
-        SArray** resQ = taosAllocateQitem(sizeof(void**), DEF_QITEM);
-        *resQ = pRes;
-        taosWriteQitem(pTask->outputQ, resQ);
-      }
-    }
-  }
+  // 3. handle output
+  // 3.1 check and set status
+  // 3.2 dispatch / sink
+  streamTaskSink(pTask, pMsgCb);
+
+  return 0;
+}
+
+int32_t streamTaskProcessRecoverReq(SStreamTask* pTask, char* msg) {
+  //
   return 0;
 }
 
@@ -545,7 +483,7 @@ int32_t streamExecTask(SStreamTask* pTask, SMsgCb* pMsgCb, const void* input, in
 
   if (pTask->dispatchType == TASK_DISPATCH__INPLACE) {
     SRpcMsg dispatchMsg = {0};
-    if (streamBuildDispatchMsg(pTask, pRes, &dispatchMsg, NULL) < 0) {
+    if (streamBuildExecMsg(pTask, pRes, &dispatchMsg, NULL) < 0) {
       ASSERT(0);
       return -1;
     }
@@ -566,7 +504,7 @@ int32_t streamExecTask(SStreamTask* pTask, SMsgCb* pMsgCb, const void* input, in
   } else if (pTask->dispatchType == TASK_DISPATCH__FIXED) {
     SRpcMsg dispatchMsg = {0};
     SEpSet* pEpSet = NULL;
-    if (streamBuildDispatchMsg(pTask, pRes, &dispatchMsg, &pEpSet) < 0) {
+    if (streamBuildExecMsg(pTask, pRes, &dispatchMsg, &pEpSet) < 0) {
       ASSERT(0);
       return -1;
     }

source/util/src/terror.c

@@ -285,6 +285,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_MND_INVALID_TOPIC_QUERY,      "Topic with invalid qu
 TAOS_DEFINE_ERROR(TSDB_CODE_MND_INVALID_TOPIC_OPTION,     "Topic with invalid option")
 TAOS_DEFINE_ERROR(TSDB_CODE_MND_CONSUMER_NOT_EXIST,       "Consumer not exist")
 TAOS_DEFINE_ERROR(TSDB_CODE_MND_CONSUMER_NOT_READY,       "Consumer waiting for rebalance")
+TAOS_DEFINE_ERROR(TSDB_CODE_MND_TOPIC_SUBSCRIBED,         "Topic subscribed cannot be dropped")
 
 // mnode-sma
 TAOS_DEFINE_ERROR(TSDB_CODE_MND_SMA_ALREADY_EXIST,        "SMA already exists")