xxq250
/
homework-jianmu
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

fix stream backend crash when transfer

This commit is contained in:
yihaoDeng 2023-12-18 15:48:01 +08:00
parent 0840e6a0d8
commit 1dea66d4f5
1 changed files with 46 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
source/libs/stream/src/streamStart.c
View File

@ -14,10 +14,10 @@
*/ */
#include "streamInt.h" #include "streamInt.h"
#include "streamsm.h"
#include "trpc.h" #include "trpc.h"
#include "ttimer.h" #include "ttimer.h"
#include "wal.h" #include "wal.h"
#include "streamsm.h"
#define SCANHISTORY_IDLE_TIME_SLICE 100 // 100ms #define SCANHISTORY_IDLE_TIME_SLICE 100 // 100ms
#define SCANHISTORY_MAX_IDLE_TIME 10 // 10 sec #define SCANHISTORY_MAX_IDLE_TIME 10 // 10 sec
@ -115,8 +115,7 @@ static void doReExecScanhistory(void* param, void* tmrId) {
// release the task. // release the task.
streamMetaReleaseTask(pTask->pMeta, pTask); streamMetaReleaseTask(pTask->pMeta, pTask);
} else { } else {
taosTmrReset(doReExecScanhistory, SCANHISTORY_IDLE_TIME_SLICE, pTask, streamTimer, taosTmrReset(doReExecScanhistory, SCANHISTORY_IDLE_TIME_SLICE, pTask, streamTimer, &pTask->schedHistoryInfo.pTimer);
&pTask->schedHistoryInfo.pTimer);
} }
} }
@ -290,7 +289,8 @@ static void recheckDownstreamTasks(void* param, void* tmrId) {
stDebug("s-task:%s complete send check in timer, ref:%d", pTask->id.idStr, ref); stDebug("s-task:%s complete send check in timer, ref:%d", pTask->id.idStr, ref);
} }
int32_t streamTaskCheckStatus(SStreamTask* pTask, int32_t upstreamTaskId, int32_t vgId, int64_t stage, int64_t* oldStage) { int32_t streamTaskCheckStatus(SStreamTask* pTask, int32_t upstreamTaskId, int32_t vgId, int64_t stage,
int64_t* oldStage) {
SStreamChildEpInfo* pInfo = streamTaskGetUpstreamTaskEpInfo(pTask, upstreamTaskId); SStreamChildEpInfo* pInfo = streamTaskGetUpstreamTaskEpInfo(pTask, upstreamTaskId);
ASSERT(pInfo != NULL); ASSERT(pInfo != NULL);
@ -418,7 +418,6 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs
if (pRsp->status == TASK_DOWNSTREAM_READY) { if (pRsp->status == TASK_DOWNSTREAM_READY) {
if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) {
bool found = false; bool found = false;
int32_t numOfReqs = taosArrayGetSize(pTask->checkReqIds); int32_t numOfReqs = taosArrayGetSize(pTask->checkReqIds);
for (int32_t i = 0; i < numOfReqs; i++) { for (int32_t i = 0; i < numOfReqs; i++) {
@ -457,8 +456,9 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs
} else { // not ready, wait for 100ms and retry } else { // not ready, wait for 100ms and retry
if (pRsp->status == TASK_UPSTREAM_NEW_STAGE || pRsp->status == TASK_DOWNSTREAM_NOT_LEADER) { if (pRsp->status == TASK_UPSTREAM_NEW_STAGE || pRsp->status == TASK_DOWNSTREAM_NOT_LEADER) {
if (pRsp->status == TASK_UPSTREAM_NEW_STAGE) { if (pRsp->status == TASK_UPSTREAM_NEW_STAGE) {
stError( stError("s-task:%s vgId:%d self vnode-transfer/leader-change/restart detected, old stage:%" PRId64
"s-task:%s vgId:%d self vnode-transfer/leader-change/restart detected, old stage:%"PRId64", current stage:%"PRId64", " ", current stage:%" PRId64
", "
"not check wait for downstream task nodeUpdate, and all tasks restart", "not check wait for downstream task nodeUpdate, and all tasks restart",
id, pRsp->upstreamNodeId, pRsp->oldStage, pTask->pMeta->stage); id, pRsp->upstreamNodeId, pRsp->oldStage, pTask->pMeta->stage);
addIntoNodeUpdateList(pTask, pRsp->upstreamNodeId); addIntoNodeUpdateList(pTask, pRsp->upstreamNodeId);
@ -476,11 +476,16 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs
// automatically set the related fill-history task to be failed. // automatically set the related fill-history task to be failed.
if (HAS_RELATED_FILLHISTORY_TASK(pTask)) { if (HAS_RELATED_FILLHISTORY_TASK(pTask)) {
STaskId* pId = &pTask->hTaskInfo.id; STaskId* pId = &pTask->hTaskInfo.id;
int64_t current = taosGetTimestampMs();
SStreamTask* pHTask = streamMetaAcquireTask(pTask->pMeta, pId->streamId, pId->taskId); SStreamTask* pHTask = streamMetaAcquireTask(pTask->pMeta, pId->streamId, pId->taskId);
streamMetaUpdateTaskDownstreamStatus(pHTask->pMeta, pId->streamId, pId->taskId, pHTask->execInfo.init, if (pHTask != NULL) {
taosGetTimestampMs(), false); streamMetaUpdateTaskDownstreamStatus(pTask->pMeta, pId->streamId, pId->taskId, pHTask->execInfo.init, current,
false);
streamMetaReleaseTask(pTask->pMeta, pHTask); streamMetaReleaseTask(pTask->pMeta, pHTask);
} else {
streamMetaUpdateTaskDownstreamStatus(pTask->pMeta, pId->streamId, pId->taskId, 0, current, false);
}
} }
} else { // TASK_DOWNSTREAM_NOT_READY, let's retry in 100ms } else { // TASK_DOWNSTREAM_NOT_READY, let's retry in 100ms
STaskRecheckInfo* pInfo = createRecheckInfo(pTask, pRsp); STaskRecheckInfo* pInfo = createRecheckInfo(pTask, pRsp);
@ -601,8 +606,8 @@ int32_t streamProcessScanHistoryFinishReq(SStreamTask* pTask, SStreamScanHistory
ETaskStatus status = streamTaskGetStatus(pTask, &p); ETaskStatus status = streamTaskGetStatus(pTask, &p);
if (status != TASK_STATUS__SCAN_HISTORY && status != TASK_STATUS__STREAM_SCAN_HISTORY) { if (status != TASK_STATUS__SCAN_HISTORY && status != TASK_STATUS__STREAM_SCAN_HISTORY) {
stError("s-task:%s not in scan-history status, status:%s return upstream:0x%x scan-history finish directly", stError("s-task:%s not in scan-history status, status:%s return upstream:0x%x scan-history finish directly", id, p,
id, p, pReq->upstreamTaskId); pReq->upstreamTaskId);
void* pBuf = NULL; void* pBuf = NULL;
int32_t len = 0; int32_t len = 0;
@ -612,8 +617,8 @@ int32_t streamProcessScanHistoryFinishReq(SStreamTask* pTask, SStreamScanHistory
initRpcMsg(&msg, 0, pBuf, sizeof(SMsgHead) + len); initRpcMsg(&msg, 0, pBuf, sizeof(SMsgHead) + len);
tmsgSendRsp(&msg); tmsgSendRsp(&msg);
stDebug("s-task:%s level:%d notify upstream:0x%x(vgId:%d) to continue process data in WAL", id, stDebug("s-task:%s level:%d notify upstream:0x%x(vgId:%d) to continue process data in WAL", id, taskLevel,
taskLevel, pReq->upstreamTaskId, pReq->upstreamNodeId); pReq->upstreamTaskId, pReq->upstreamNodeId);
return 0; return 0;
} }
@ -647,8 +652,8 @@ int32_t streamProcessScanHistoryFinishReq(SStreamTask* pTask, SStreamScanHistory
streamTaskHandleEvent(pTask->status.pSM, TASK_EVENT_SCANHIST_DONE); streamTaskHandleEvent(pTask->status.pSM, TASK_EVENT_SCANHIST_DONE);
} }
} else { } else {
stDebug("s-task:%s receive scan-history data finish msg from upstream:0x%x(index:%d), unfinished:%d", stDebug("s-task:%s receive scan-history data finish msg from upstream:0x%x(index:%d), unfinished:%d", id,
id, pReq->upstreamTaskId, pReq->childId, left); pReq->upstreamTaskId, pReq->childId, left);
} }
return 0; return 0;
@ -688,8 +693,8 @@ static void checkFillhistoryTaskStatus(SStreamTask* pTask, SStreamTask* pHTask)
if (pTask->info.taskLevel == TASK_LEVEL__SOURCE) { if (pTask->info.taskLevel == TASK_LEVEL__SOURCE) {
stDebug("s-task:%s set the launch condition for fill-history s-task:%s, window:%" PRId64 " - %" PRId64 stDebug("s-task:%s set the launch condition for fill-history s-task:%s, window:%" PRId64 " - %" PRId64
" verRange:%" PRId64 " - %" PRId64 ", init:%" PRId64, " verRange:%" PRId64 " - %" PRId64 ", init:%" PRId64,
pTask->id.idStr, pHTask->id.idStr, pRange->window.skey, pRange->window.ekey, pTask->id.idStr, pHTask->id.idStr, pRange->window.skey, pRange->window.ekey, pRange->range.minVer,
pRange->range.minVer, pRange->range.maxVer, pHTask->execInfo.init); pRange->range.maxVer, pHTask->execInfo.init);
} else { } else {
stDebug("s-task:%s no fill-history condition for non-source task:%s", pTask->id.idStr, pHTask->id.idStr); stDebug("s-task:%s no fill-history condition for non-source task:%s", pTask->id.idStr, pHTask->id.idStr);
} }
@ -724,7 +729,6 @@ static void tryLaunchHistoryTask(void* param, void* tmrId) {
SStreamTask* pTask = streamMetaAcquireTask(pMeta, pInfo->id.streamId, pInfo->id.taskId); SStreamTask* pTask = streamMetaAcquireTask(pMeta, pInfo->id.streamId, pInfo->id.taskId);
if (pTask != NULL) { if (pTask != NULL) {
SHistoryTaskInfo* pHTaskInfo = &pTask->hTaskInfo; SHistoryTaskInfo* pHTaskInfo = &pTask->hTaskInfo;
pHTaskInfo->tickCount -= 1; pHTaskInfo->tickCount -= 1;