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

Merge pull request #23946 from taosdata/fix/snodeBackendCrash 

Fix/snode backend crash
This commit is contained in:
Haojun Liao 2023-12-16 13:13:28 +08:00 committed by GitHub
parent 178b94a398 bb5f0d9112
commit 1a9701d401
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
10 changed files with 138 additions and 101 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
include/libs/executor/storageapi.h
View File

@ -102,11 +102,11 @@ typedef struct SMTbCursor {
} SMTbCursor;
typedef struct SMCtbCursor {
SMeta *pMeta;
void *pCur;
SMeta* pMeta;
void* pCur;
tb_uid_t suid;
void *pKey;
void *pVal;
void* pKey;
void* pVal;
int kLen;
int vLen;
int8_t paused;
@ -263,21 +263,22 @@ typedef struct SStoreMeta {
void* (*storeGetIndexInfo)();
void* (*getInvertIndex)(void* pVnode);
// support filter and non-filter cases. [vnodeGetCtbIdList & vnodeGetCtbIdListByFilter]
int32_t (*getChildTableList)( void* pVnode, int64_t suid, SArray* list);
int32_t (*getChildTableList)(void* pVnode, int64_t suid, SArray* list);
int32_t (*storeGetTableList)(void* pVnode, int8_t type, SArray* pList);
void* storeGetVersionRange;
void* storeGetLastTimestamp;
int32_t (*getTableSchema)(void* pVnode, int64_t uid, STSchema** pSchema, int64_t* suid); // tsdbGetTableSchema
int32_t (*getNumOfChildTables)( void* pVnode, int64_t uid, int64_t* numOfTables, int32_t* numOfCols);
void (*getBasicInfo)(void* pVnode, const char** dbname, int32_t* vgId, int64_t* numOfTables, int64_t* numOfNormalTables);
int32_t (*getNumOfChildTables)(void* pVnode, int64_t uid, int64_t* numOfTables, int32_t* numOfCols);
void (*getBasicInfo)(void* pVnode, const char** dbname, int32_t* vgId, int64_t* numOfTables,
int64_t* numOfNormalTables);
int64_t (*getNumOfRowsInMem)(void* pVnode);
SMCtbCursor* (*openCtbCursor)(void *pVnode, tb_uid_t uid, int lock);
SMCtbCursor* (*openCtbCursor)(void* pVnode, tb_uid_t uid, int lock);
int32_t (*resumeCtbCursor)(SMCtbCursor* pCtbCur, int8_t first);
void (*pauseCtbCursor)(SMCtbCursor* pCtbCur);
void (*closeCtbCursor)(SMCtbCursor *pCtbCur);
void (*closeCtbCursor)(SMCtbCursor* pCtbCur);
tb_uid_t (*ctbCursorNext)(SMCtbCursor* pCur);
} SStoreMeta;
@ -366,11 +367,11 @@ typedef struct SStateStore {
TSKEY (*updateInfoFillBlockData)(SUpdateInfo* pInfo, SSDataBlock* pBlock, int32_t primaryTsCol);
bool (*updateInfoIsUpdated)(SUpdateInfo* pInfo, uint64_t tableId, TSKEY ts);
bool (*updateInfoIsTableInserted)(SUpdateInfo* pInfo, int64_t tbUid);
bool (*isIncrementalTimeStamp)(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts);
bool (*isIncrementalTimeStamp)(SUpdateInfo* pInfo, uint64_t tableId, TSKEY ts);
void (*updateInfoDestroy)(SUpdateInfo* pInfo);
void (*windowSBfDelete)(SUpdateInfo *pInfo, uint64_t count);
void (*windowSBfAdd)(SUpdateInfo *pInfo, uint64_t count);
void (*windowSBfDelete)(SUpdateInfo* pInfo, uint64_t count);
void (*windowSBfAdd)(SUpdateInfo* pInfo, uint64_t count);
SUpdateInfo* (*updateInfoInitP)(SInterval* pInterval, int64_t watermark, bool igUp);
void (*updateInfoAddCloseWindowSBF)(SUpdateInfo* pInfo);
@ -397,6 +398,7 @@ typedef struct SStateStore {
void (*streamStateDestroy)(SStreamState* pState, bool remove);
int32_t (*streamStateDeleteCheckPoint)(SStreamState* pState, TSKEY mark);
void (*streamStateReloadInfo)(SStreamState* pState, TSKEY ts);
void (*streamStateCopyBackend)(SStreamState* src, SStreamState* dst);
} SStateStore;
typedef struct SStorageAPI {

1
include/libs/function/function.h
View File

@ -171,6 +171,7 @@ typedef struct {
int32_t taskId;
int64_t streamId;
int64_t streamBackendRid;
int8_t dump;
} SStreamState;
typedef struct SFunctionStateStore {

7
include/libs/stream/streamState.h
View File

@ -50,7 +50,7 @@ void streamStateSetNumber(SStreamState* pState, int32_t number);
int32_t streamStateSaveInfo(SStreamState* pState, void* pKey, int32_t keyLen, void* pVal, int32_t vLen);
int32_t streamStateGetInfo(SStreamState* pState, void* pKey, int32_t keyLen, void** pVal, int32_t* pLen);
//session window
// session window
int32_t streamStateSessionAddIfNotExist(SStreamState* pState, SSessionKey* key, TSKEY gap, void** pVal, int32_t* pVLen);
int32_t streamStateSessionPut(SStreamState* pState, const SSessionKey* key, void* value, int32_t vLen);
int32_t streamStateSessionGet(SStreamState* pState, SSessionKey* key, void** pVal, int32_t* pVLen);
@ -65,7 +65,7 @@ SStreamStateCur* streamStateSessionSeekKeyNext(SStreamState* pState, const SSess
SStreamStateCur* streamStateSessionSeekKeyCurrentPrev(SStreamState* pState, const SSessionKey* key);
SStreamStateCur* streamStateSessionSeekKeyCurrentNext(SStreamState* pState, const SSessionKey* key);
//state window
// state window
int32_t streamStateStateAddIfNotExist(SStreamState* pState, SSessionKey* key, char* pKeyData, int32_t keyDataLen,
state_key_cmpr_fn fn, void** pVal, int32_t* pVLen);
@ -96,6 +96,9 @@ int32_t streamStatePutParName(SStreamState* pState, int64_t groupId, const char*
int32_t streamStateGetParName(SStreamState* pState, int64_t groupId, void** pVal);
void streamStateReloadInfo(SStreamState* pState, TSKEY ts);
void streamStateCopyBackend(SStreamState* src, SStreamState* dst);
SStreamStateCur* createStreamStateCursor();
/***compare func **/

1
source/dnode/snode/src/snode.c
View File

@ -67,6 +67,7 @@ int32_t sndExpandTask(SSnode *pSnode, SStreamTask *pTask, int64_t nextProcessVer
sndDebug("s-task:%s state:%p", pTask->id.idStr, pTask->pState);
}
int32_t numOfVgroups = (int32_t)taosArrayGetSize(pTask->upstreamInfo.pList);
SReadHandle handle = {
.checkpointId = pTask->chkInfo.checkpointId,

5
source/dnode/snode/src/snodeInitApi.c
View File

@ -14,8 +14,8 @@
*/
#include "storageapi.h"
#include "tstreamUpdate.h"
#include "streamState.h"
#include "tstreamUpdate.h"
static void initStateStoreAPI(SStateStore* pStore);
static void initFunctionStateStore(SFunctionStateStore* pStore);
@ -100,9 +100,10 @@ void initStateStoreAPI(SStateStore* pStore) {
pStore->streamStateClose = streamStateClose;
pStore->streamStateBegin = streamStateBegin;
pStore->streamStateCommit = streamStateCommit;
pStore->streamStateDestroy= streamStateDestroy;
pStore->streamStateDestroy = streamStateDestroy;
pStore->streamStateDeleteCheckPoint = streamStateDeleteCheckPoint;
pStore->streamStateReloadInfo = streamStateReloadInfo;
pStore->streamStateCopyBackend = streamStateCopyBackend;
}
void initFunctionStateStore(SFunctionStateStore* pStore) {

35
source/dnode/vnode/src/tqCommon/tqCommon.c
View File

@ -13,9 +13,9 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "tstream.h"
#include "tmsgcb.h"
#include "tq.h"
#include "tstream.h"
typedef struct STaskUpdateEntry {
int64_t streamId;
@ -42,7 +42,7 @@ int32_t tqStreamTaskStartAsync(SStreamMeta* pMeta, SMsgCb* cb, bool restart) {
tqDebug("vgId:%d start all %d stream task(s) async", vgId, numOfTasks);
pRunReq->head.vgId = vgId;
pRunReq->streamId = 0;
pRunReq->taskId = restart? STREAM_EXEC_RESTART_ALL_TASKS_ID:STREAM_EXEC_START_ALL_TASKS_ID;
pRunReq->taskId = restart ? STREAM_EXEC_RESTART_ALL_TASKS_ID : STREAM_EXEC_START_ALL_TASKS_ID;
SRpcMsg msg = {.msgType = TDMT_STREAM_TASK_RUN, .pCont = pRunReq, .contLen = sizeof(SStreamTaskRunReq)};
tmsgPutToQueue(cb, STREAM_QUEUE, &msg);
@ -166,7 +166,8 @@ int32_t tqStreamTaskProcessUpdateReq(SStreamMeta* pMeta, SMsgCb* cb, SRpcMsg* pM
streamMetaWUnLock(pMeta);
} else {
if (!restored) {
tqDebug("vgId:%d vnode restore not completed, not restart the tasks, clear the start after nodeUpdate flag", vgId);
tqDebug("vgId:%d vnode restore not completed, not restart the tasks, clear the start after nodeUpdate flag",
vgId);
pMeta->startInfo.tasksWillRestart = 0;
streamMetaWUnLock(pMeta);
} else {
@ -251,7 +252,7 @@ int32_t tqStreamTaskProcessDispatchReq(SStreamMeta* pMeta, SRpcMsg* pMsg) {
SStreamTask* pTask = streamMetaAcquireTask(pMeta, req.streamId, req.taskId);
if (pTask) {
SRpcMsg rsp = {.info = pMsg->info, .code = 0};
if (streamProcessDispatchMsg(pTask, &req, &rsp) != 0){
if (streamProcessDispatchMsg(pTask, &req, &rsp) != 0) {
return -1;
}
tDeleteStreamDispatchReq(&req);
@ -355,8 +356,8 @@ int32_t tqStreamTaskProcessScanHistoryFinishReq(SStreamMeta* pMeta, SRpcMsg* pMs
SStreamTask* pTask = streamMetaAcquireTask(pMeta, req.streamId, req.downstreamTaskId);
if (pTask == NULL) {
tqError("vgId:%d process scan history finish msg, failed to find task:0x%x, it may be destroyed",
pMeta->vgId, req.downstreamTaskId);
tqError("vgId:%d process scan history finish msg, failed to find task:0x%x, it may be destroyed", pMeta->vgId,
req.downstreamTaskId);
return -1;
}
@ -381,8 +382,8 @@ int32_t tqStreamTaskProcessScanHistoryFinishRsp(SStreamMeta* pMeta, SRpcMsg* pMs
SStreamTask* pTask = streamMetaAcquireTask(pMeta, req.streamId, req.upstreamTaskId);
if (pTask == NULL) {
tqError("vgId:%d process scan history finish rsp, failed to find task:0x%x, it may be destroyed",
pMeta->vgId, req.upstreamTaskId);
tqError("vgId:%d process scan history finish rsp, failed to find task:0x%x, it may be destroyed", pMeta->vgId,
req.upstreamTaskId);
return -1;
}
@ -428,7 +429,8 @@ int32_t tqStreamTaskProcessCheckReq(SStreamMeta* pMeta, SRpcMsg* pMsg) {
// only the leader node handle the check request
if (pMeta->role == NODE_ROLE_FOLLOWER) {
tqError("s-task:0x%x invalid check msg from upstream:0x%x(vgId:%d), vgId:%d is follower, not handle check status msg",
tqError(
"s-task:0x%x invalid check msg from upstream:0x%x(vgId:%d), vgId:%d is follower, not handle check status msg",
taskId, req.upstreamTaskId, req.upstreamNodeId, pMeta->vgId);
rsp.status = TASK_DOWNSTREAM_NOT_LEADER;
} else {
@ -439,7 +441,8 @@ int32_t tqStreamTaskProcessCheckReq(SStreamMeta* pMeta, SRpcMsg* pMsg) {
char* p = NULL;
streamTaskGetStatus(pTask, &p);
tqDebug("s-task:%s status:%s, stage:%"PRId64" recv task check req(reqId:0x%" PRIx64 ") task:0x%x (vgId:%d), check_status:%d",
tqDebug("s-task:%s status:%s, stage:%" PRId64 " recv task check req(reqId:0x%" PRIx64
") task:0x%x (vgId:%d), check_status:%d",
pTask->id.idStr, p, rsp.oldStage, rsp.reqId, rsp.upstreamTaskId, rsp.upstreamNodeId, rsp.status);
} else {
rsp.status = TASK_DOWNSTREAM_NOT_READY;
@ -526,7 +529,8 @@ int32_t tqStreamTaskProcessCheckpointReadyMsg(SStreamMeta* pMeta, SRpcMsg* pMsg)
return code;
}
int32_t tqStreamTaskProcessDeployReq(SStreamMeta* pMeta, int64_t sversion, char* msg, int32_t msgLen, bool isLeader, bool restored) {
int32_t tqStreamTaskProcessDeployReq(SStreamMeta* pMeta, int64_t sversion, char* msg, int32_t msgLen, bool isLeader,
bool restored) {
int32_t code = 0;
int32_t vgId = pMeta->vgId;
@ -566,7 +570,8 @@ int32_t tqStreamTaskProcessDeployReq(SStreamMeta* pMeta, int64_t sversion, char*
streamMetaWUnLock(pMeta);
if (code < 0) {
tqError("failed to add s-task:0x%x into vgId:%d meta, total:%d, code:%s", vgId, taskId, numOfTasks, tstrerror(code));
tqError("failed to add s-task:0x%x into vgId:%d meta, total:%d, code:%s", vgId, taskId, numOfTasks,
tstrerror(code));
tFreeStreamTask(pTask);
return code;
}
@ -716,7 +721,7 @@ static int32_t restartStreamTasks(SStreamMeta* pMeta, bool isLeader) {
int32_t code = 0;
int64_t st = taosGetTimestampMs();
while(1) {
while (1) {
int32_t startVal = atomic_val_compare_exchange_32(&pMeta->startInfo.taskStarting, 0, 1);
if (startVal == 0) {
break;
@ -739,7 +744,7 @@ static int32_t restartStreamTasks(SStreamMeta* pMeta, bool isLeader) {
streamMetaClear(pMeta);
int64_t el = taosGetTimestampMs() - st;
tqInfo("vgId:%d close&reload state elapsed time:%.3fs", vgId, el/1000.);
tqInfo("vgId:%d close&reload state elapsed time:%.3fs", vgId, el / 1000.);
code = streamMetaLoadAllTasks(pMeta);
if (code != TSDB_CODE_SUCCESS) {
@ -800,5 +805,3 @@ int32_t tqStreamTaskProcessRunReq(SStreamMeta* pMeta, SRpcMsg* pMsg, bool isLead
return -1;
}
}

1
source/dnode/vnode/src/vnd/vnodeInitApi.c
View File

@ -215,6 +215,7 @@ void initStateStoreAPI(SStateStore* pStore) {
pStore->streamStateDestroy = streamStateDestroy;
pStore->streamStateDeleteCheckPoint = streamStateDeleteCheckPoint;
pStore->streamStateReloadInfo = streamStateReloadInfo;
pStore->streamStateCopyBackend = streamStateCopyBackend;
}
void initMetaReaderAPI(SStoreMetaReader* pMetaReader) {

14
source/libs/executor/src/streamtimewindowoperator.c
View File

@ -408,6 +408,11 @@ void destroyStreamFinalIntervalOperatorInfo(void* param) {
taosArrayDestroy(pInfo->pDelWins);
blockDataDestroy(pInfo->pDelRes);
pInfo->stateStore.streamFileStateDestroy(pInfo->pState->pFileState);
if (pInfo->pState->dump == 1) {
taosMemoryFreeClear(pInfo->pState->pTdbState->pOwner);
taosMemoryFreeClear(pInfo->pState->pTdbState);
}
taosMemoryFreeClear(pInfo->pState);
nodesDestroyNode((SNode*)pInfo->pPhyNode);
@ -1462,7 +1467,11 @@ SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
initBasicInfo(&pInfo->binfo, pResBlock);
pInfo->pState = taosMemoryCalloc(1, sizeof(SStreamState));
*(pInfo->pState) = *(pTaskInfo->streamInfo.pState);
qInfo("open state %p", pInfo->pState);
pAPI->stateStore.streamStateCopyBackend(pTaskInfo->streamInfo.pState, pInfo->pState);
//*(pInfo->pState) = *(pTaskInfo->streamInfo.pState);
qInfo("copy state %p to %p", pTaskInfo->streamInfo.pState, pInfo->pState);
pAPI->stateStore.streamStateSetNumber(pInfo->pState, -1);
int32_t code = initAggSup(&pOperator->exprSupp, &pInfo->aggSup, pExprInfo, numOfCols, keyBufSize, pTaskInfo->id.str,
@ -3338,7 +3347,8 @@ static void doStreamStateAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBl
SColumnInfoData* pKeyColInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->stateCol.slotId);
for (int32_t i = 0; i < rows; i += winRows) {
if (pInfo->ignoreExpiredData && checkExpiredData(&pInfo->streamAggSup.stateStore, pInfo->streamAggSup.pUpdateInfo,
&pInfo->twAggSup, pSDataBlock->info.id.uid, tsCols[i]) || colDataIsNull_s(pKeyColInfo, i)) {
&pInfo->twAggSup, pSDataBlock->info.id.uid, tsCols[i]) ||
colDataIsNull_s(pKeyColInfo, i)) {
i++;
continue;
}

19
source/libs/stream/src/streamState.c
View File

@ -91,8 +91,8 @@ int stateKeyCmpr(const void* pKey1, int kLen1, const void* pKey2, int kLen2) {
}
SStreamState* streamStateOpen(char* path, void* pTask, bool specPath, int32_t szPage, int32_t pages) {
stDebug("open stream state, %s", path);
SStreamState* pState = taosMemoryCalloc(1, sizeof(SStreamState));
stDebug("open stream state %p, %s", pState, path);
if (pState == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
@ -710,7 +710,8 @@ int32_t streamStateSessionPut(SStreamState* pState, const SSessionKey* key, void
#endif
}
int32_t streamStateSessionAllocWinBuffByNextPosition(SStreamState* pState, SStreamStateCur* pCur, const SSessionKey* pKey, void** pVal, int32_t* pVLen) {
int32_t streamStateSessionAllocWinBuffByNextPosition(SStreamState* pState, SStreamStateCur* pCur,
const SSessionKey* pKey, void** pVal, int32_t* pVLen) {
#ifdef USE_ROCKSDB
return allocSessioncWinBuffByNextPosition(pState->pFileState, pCur, pKey, pVal, pVLen);
#else
@ -1115,6 +1116,20 @@ int32_t streamStateDeleteCheckPoint(SStreamState* pState, TSKEY mark) {
void streamStateReloadInfo(SStreamState* pState, TSKEY ts) { streamFileStateReloadInfo(pState->pFileState, ts); }
void streamStateCopyBackend(SStreamState* src, SStreamState* dst) {
dst->pFileState = src->pFileState;
dst->parNameMap = src->parNameMap;
dst->number = src->number;
dst->taskId = src->taskId;
dst->streamId = src->streamId;
if (dst->pTdbState == NULL) {
dst->pTdbState = taosMemoryCalloc(1, sizeof(STdbState));
dst->pTdbState->pOwner = taosMemoryCalloc(1, sizeof(SStreamTask));
}
dst->dump = 1;
dst->pTdbState->pOwner->pBackend = src->pTdbState->pOwner->pBackend;
return;
}
SStreamStateCur* createStreamStateCursor() {
SStreamStateCur* pCur = taosMemoryCalloc(1, sizeof(SStreamStateCur));
pCur->buffIndex = -1;

2
tests/parallel_test/cases.task
View File

@ -26,7 +26,7 @@
,,y,system-test,./pytest.sh python3 ./test.py -f 8-stream/pause_resume_test.py
#,,n,system-test,python3 ./test.py -f 8-stream/vnode_restart.py -N 4
#,,n,system-test,python3 ./test.py -f 8-stream/snode_restart.py -N 4
#,,n,system-test,python3 ./test.py -f 8-stream/snode_restart_with_checkpoint.py -N 4
,,n,system-test,python3 ./test.py -f 8-stream/snode_restart_with_checkpoint.py -N 4
,,y,system-test,./pytest.sh python3 ./test.py -f 2-query/tbname_vgroup.py
,,y,system-test,./pytest.sh python3 ./test.py -f 2-query/stbJoin.py