Merge pull request #26912 from taosdata/fix/create_tb

fix(stream): compare vg replica according to different db.
This commit is contained in:
Haojun Liao 2024-08-02 00:39:48 +08:00 committed by GitHub
commit 8be9337659
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
6 changed files with 101 additions and 56 deletions

View File

@ -88,17 +88,45 @@ void destroyStreamTaskIter(SStreamTaskIter* pIter) {
taosMemoryFree(pIter);
}
static bool checkStatusForEachReplica(SVgObj *pVgroup) {
for (int32_t i = 0; i < pVgroup->replica; ++i) {
if (!pVgroup->vnodeGid[i].syncRestore) {
mInfo("vgId:%d not restored, not ready for checkpoint or other operations", pVgroup->vgId);
return false;
}
ESyncState state = pVgroup->vnodeGid[i].syncState;
if (state == TAOS_SYNC_STATE_OFFLINE || state == TAOS_SYNC_STATE_ERROR || state == TAOS_SYNC_STATE_LEARNER ||
state == TAOS_SYNC_STATE_CANDIDATE) {
mInfo("vgId:%d state:%d , not ready for checkpoint or other operations, not check other vgroups", pVgroup->vgId,
state);
return false;
}
}
return true;
}
int32_t mndTakeVgroupSnapshot(SMnode *pMnode, bool *allReady, SArray **pList) {
SSdb *pSdb = pMnode->pSdb;
void *pIter = NULL;
SVgObj *pVgroup = NULL;
int32_t replica = -1; // do the replica check
int32_t code = 0;
SArray *pVgroupList = NULL;
SHashObj *pHash = NULL;
*allReady = true;
SArray *pVgroupList = taosArrayInit(4, sizeof(SNodeEntry));
pVgroupList = taosArrayInit(4, sizeof(SNodeEntry));
if (pVgroupList == NULL) {
return terrno;
mError("failed to prepare arraylist during take vgroup snapshot, code:%s", tstrerror(terrno));
code = terrno;
goto _err;
}
pHash = taosHashInit(10, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
if (pHash == NULL) {
mError("failed to prepare hashmap during take vgroup snapshot, code:%s", tstrerror(terrno));
code = terrno;
goto _err;
}
while (1) {
@ -110,36 +138,27 @@ int32_t mndTakeVgroupSnapshot(SMnode *pMnode, bool *allReady, SArray** pList) {
SNodeEntry entry = {.nodeId = pVgroup->vgId, .hbTimestamp = pVgroup->updateTime};
entry.epset = mndGetVgroupEpset(pMnode, pVgroup);
if (replica == -1) {
replica = pVgroup->replica;
} else {
if (replica != pVgroup->replica) {
mInfo("vgId:%d replica:%d inconsistent with other vgroups replica:%d, not ready for stream operations",
pVgroup->vgId, pVgroup->replica, replica);
*allReady = false;
int8_t *pReplica = taosHashGet(pHash, &pVgroup->dbUid, sizeof(pVgroup->dbUid));
if (pReplica == NULL) { // not exist, add it into hash map
code = taosHashPut(pHash, &pVgroup->dbUid, sizeof(pVgroup->dbUid), &pVgroup->replica, sizeof(pVgroup->replica));
if (code) {
mError("failed to put info into hashmap during task vgroup snapshot, code:%s", tstrerror(code));
sdbRelease(pSdb, pVgroup);
break;
sdbCancelFetch(pSdb, pIter);
goto _err; // take snapshot failed, and not all ready
}
} else {
if (*pReplica != pVgroup->replica) {
mInfo("vgId:%d replica:%d inconsistent with other vgroups replica:%d, not ready for stream operations",
pVgroup->vgId, pVgroup->replica, *pReplica);
*allReady = false; // task snap success, but not all ready
}
}
// if not all ready till now, no need to check the remaining vgroups.
// but still we need to put the info of the existed vgroups into the snapshot list
if (*allReady) {
for (int32_t i = 0; i < pVgroup->replica; ++i) {
if (!pVgroup->vnodeGid[i].syncRestore) {
mInfo("vgId:%d not restored, not ready for checkpoint or other operations", pVgroup->vgId);
*allReady = false;
break;
}
ESyncState state = pVgroup->vnodeGid[i].syncState;
if (state == TAOS_SYNC_STATE_OFFLINE || state == TAOS_SYNC_STATE_ERROR || state == TAOS_SYNC_STATE_LEARNER ||
state == TAOS_SYNC_STATE_CANDIDATE) {
mInfo("vgId:%d state:%d , not ready for checkpoint or other operations, not check other vgroups",
pVgroup->vgId, state);
*allReady = false;
break;
}
}
*allReady = checkStatusForEachReplica(pVgroup);
}
char buf[256] = {0};
@ -148,6 +167,10 @@ int32_t mndTakeVgroupSnapshot(SMnode *pMnode, bool *allReady, SArray** pList) {
void *p = taosArrayPush(pVgroupList, &entry);
if (p == NULL) {
mError("failed to put entry in vgroup list, nodeId:%d code:out of memory", entry.nodeId);
code = terrno;
sdbRelease(pSdb, pVgroup);
sdbCancelFetch(pSdb, pIter);
goto _err;
} else {
mDebug("take node snapshot, nodeId:%d %s", entry.nodeId, buf);
}
@ -166,7 +189,9 @@ int32_t mndTakeVgroupSnapshot(SMnode *pMnode, bool *allReady, SArray** pList) {
code = addEpIntoEpSet(&entry.epset, pObj->pDnode->fqdn, pObj->pDnode->port);
if (code) {
sdbRelease(pSdb, pObj);
continue;
sdbCancelFetch(pSdb, pIter);
mError("failed to extract epset for fqdn:%s during task vgroup snapshot", pObj->pDnode->fqdn);
goto _err;
}
char buf[256] = {0};
@ -174,7 +199,11 @@ int32_t mndTakeVgroupSnapshot(SMnode *pMnode, bool *allReady, SArray** pList) {
void *p = taosArrayPush(pVgroupList, &entry);
if (p == NULL) {
mError("failed to put entry in vgroup list, nodeId:%d code:out of memory", entry.nodeId);
code = terrno;
sdbRelease(pSdb, pObj);
sdbCancelFetch(pSdb, pIter);
mError("failed to put entry in vgroup list, nodeId:%d code:%s", entry.nodeId, tstrerror(code));
goto _err;
} else {
mDebug("take snode snapshot, nodeId:%d %s", entry.nodeId, buf);
}
@ -183,6 +212,14 @@ int32_t mndTakeVgroupSnapshot(SMnode *pMnode, bool *allReady, SArray** pList) {
}
*pList = pVgroupList;
taosHashCleanup(pHash);
return code;
_err:
*allReady = false;
taosArrayDestroy(pVgroupList);
taosHashCleanup(pHash);
return code;
}
@ -599,12 +636,9 @@ static int32_t doBuildStreamTaskUpdateMsg(void **pBuf, int32_t *pLen, SVgroupCha
static int32_t doSetUpdateTaskAction(SMnode *pMnode, STrans *pTrans, SStreamTask *pTask, SVgroupChangeInfo *pInfo) {
void *pBuf = NULL;
int32_t len = 0;
int32_t code = streamTaskUpdateEpsetInfo(pTask, pInfo->pUpdateNodeList);
if (code) {
return code;
}
(void)streamTaskUpdateEpsetInfo(pTask, pInfo->pUpdateNodeList);
code = doBuildStreamTaskUpdateMsg(&pBuf, &len, pInfo, pTask->info.nodeId, &pTask->id, pTrans->id);
int32_t code = doBuildStreamTaskUpdateMsg(&pBuf, &len, pInfo, pTask->info.nodeId, &pTask->id, pTrans->id);
if (code) {
return code;
}

View File

@ -253,6 +253,7 @@ int32_t streamProcessCheckpointTriggerBlock(SStreamTask* pTask, SStreamDataBlock
for (int32_t i = 0; i < taosArrayGetSize(pActiveInfo->pReadyMsgList); ++i) {
STaskCheckpointReadyInfo* p = taosArrayGet(pActiveInfo->pReadyMsgList, i);
if (p == NULL) {
streamMutexUnlock(&pTask->lock);
return TSDB_CODE_INVALID_PARA;
}
@ -390,6 +391,7 @@ int32_t streamProcessCheckpointReadyMsg(SStreamTask* pTask, int64_t checkpointId
for (int32_t i = 0; i < size; ++i) {
STaskDownstreamReadyInfo* p = taosArrayGet(pInfo->pCheckpointReadyRecvList, i);
if (p == NULL) {
streamMutexUnlock(&pInfo->lock);
return TSDB_CODE_INVALID_PARA;
}
@ -433,6 +435,7 @@ int32_t streamTaskProcessCheckpointReadyRsp(SStreamTask* pTask, int32_t upstream
for (int32_t i = 0; i < taosArrayGetSize(pInfo->pReadyMsgList); ++i) {
STaskCheckpointReadyInfo* pReadyInfo = taosArrayGet(pInfo->pReadyMsgList, i);
if (pReadyInfo == NULL) {
streamMutexUnlock(&pInfo->lock);
return TSDB_CODE_INVALID_PARA;
}
@ -447,6 +450,7 @@ int32_t streamTaskProcessCheckpointReadyRsp(SStreamTask* pTask, int32_t upstream
for (int32_t i = 0; i < taosArrayGetSize(pInfo->pReadyMsgList); ++i) {
STaskCheckpointReadyInfo* pReadyInfo = taosArrayGet(pInfo->pReadyMsgList, i);
if (pReadyInfo == NULL) {
streamMutexUnlock(&pInfo->lock);
return TSDB_CODE_INVALID_PARA;
}
@ -830,6 +834,7 @@ void checkpointTriggerMonitorFn(void* param, void* tmrId) {
if (pNotSendList == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
stDebug("s-task:%s start to triggerMonitor, reason:%s", id, tstrerror(terrno));
streamMutexUnlock(&pActiveInfo->lock);
return;
}
@ -938,13 +943,14 @@ bool streamTaskAlreadySendTrigger(SStreamTask* pTask, int32_t downstreamNodeId)
streamMutexLock(&pInfo->lock);
if (!pInfo->dispatchTrigger) {
streamMutexUnlock(&pTask->lock);
streamMutexUnlock(&pInfo->lock);
return false;
}
for (int32_t i = 0; i < taosArrayGetSize(pInfo->pDispatchTriggerList); ++i) {
STaskTriggerSendInfo* pSendInfo = taosArrayGet(pInfo->pDispatchTriggerList, i);
if (pSendInfo == NULL) {
streamMutexUnlock(&pInfo->lock);
return TSDB_CODE_INVALID_PARA;
}
@ -964,11 +970,11 @@ bool streamTaskAlreadySendTrigger(SStreamTask* pTask, int32_t downstreamNodeId)
id, pSendInfo->sendTs, before, pInfo->activeId, pInfo->transId);
}
streamMutexUnlock(&pTask->lock);
streamMutexUnlock(&pInfo->lock);
return true;
}
ASSERT(0);
streamMutexUnlock(&pInfo->lock);
return false;
}
@ -1028,6 +1034,7 @@ int32_t streamTaskGetNumOfConfirmed(SStreamTask* pTask) {
for (int32_t i = 0; i < taosArrayGetSize(pInfo->pDispatchTriggerList); ++i) {
STaskTriggerSendInfo* p = taosArrayGet(pInfo->pDispatchTriggerList, i);
if (p == NULL) {
streamMutexUnlock(&pInfo->lock);
return num;
}

View File

@ -907,8 +907,8 @@ int32_t streamTaskSendCheckpointReadyMsg(SStreamTask* pTask) {
STaskCheckpointReadyInfo* pInfo = taosArrayGet(pList, i);
SRpcMsg msg = {0};
int32_t code = initCheckpointReadyMsg(pTask, pInfo->upstreamNodeId, pInfo->upstreamTaskId, pInfo->childId, pInfo->checkpointId,
&msg);
int32_t code = initCheckpointReadyMsg(pTask, pInfo->upstreamNodeId, pInfo->upstreamTaskId, pInfo->childId,
pInfo->checkpointId, &msg);
if (code == TSDB_CODE_SUCCESS) {
code = tmsgSendReq(&pInfo->upstreamNodeEpset, &msg);
if (code == TSDB_CODE_SUCCESS) {

View File

@ -1199,14 +1199,17 @@ void streamMetaWUnLock(SStreamMeta* pMeta) {
}
int32_t streamMetaSendMsgBeforeCloseTasks(SStreamMeta* pMeta, SArray** pList) {
*pList = NULL;
QRY_OPTR_CHECK(pList);
int32_t code = 0;
SArray* pTaskList = taosArrayDup(pMeta->pTaskList, NULL);
if (pTaskList == NULL) {
stError("failed to generate the task list during send hbMsg to mnode, vgId:%d, code: out of memory", pMeta->vgId);
return TSDB_CODE_OUT_OF_MEMORY;
return terrno;
}
*pList = pTaskList;
bool sendMsg = pMeta->sendMsgBeforeClosing;
if (!sendMsg) {
stDebug("vgId:%d no need to send msg to mnode before closing tasks", pMeta->vgId);
@ -1239,9 +1242,9 @@ int32_t streamMetaSendMsgBeforeCloseTasks(SStreamMeta* pMeta, SArray** pList) {
streamMetaReleaseTask(pMeta, pTask);
}
code = streamMetaSendHbHelper(pMeta);
(void)streamMetaSendHbHelper(pMeta);
pMeta->sendMsgBeforeClosing = false;
return code;
return TSDB_CODE_SUCCESS; // always return true
}
void streamMetaUpdateStageRole(SStreamMeta* pMeta, int64_t stage, bool isLeader) {

View File

@ -602,9 +602,9 @@ int32_t streamTaskStop(SStreamTask* pTask) {
stError("failed to handle STOP event, s-task:%s", id);
}
if (pTask->info.taskLevel != TASK_LEVEL__SINK) {
if (pTask->info.taskLevel != TASK_LEVEL__SINK && pTask->exec.pExecutor != NULL) {
code = qKillTask(pTask->exec.pExecutor, TSDB_CODE_SUCCESS);
if (code) {
if (code != TSDB_CODE_SUCCESS) {
stError("s-task:%s failed to kill task related query handle", id);
}
}

View File

@ -41,8 +41,9 @@ class TDTestCase:
time.sleep(10)
tdSql.execute("use test", queryTimes=100)
tdSql.query("create stream if not exists s1 trigger at_once ignore expired 0 ignore update 0 fill_history 1 into st1 as select _wstart,sum(voltage),groupid from meters partition by groupid interval(2s)")
tdLog.debug("========create stream and insert data ok========")
time.sleep(5)
tdLog.debug("========create stream and insert data ok========")
tdSql.query("select _wstart,sum(voltage),groupid from meters partition by groupid interval(2s) order by groupid,_wstart")
rowCnt = tdSql.getRows()
results_meters = tdSql.queryResult