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

Merge branch 'fix/snodeBackendCrash' of https://github.com/taosdata/TDengine into fix/snodeBackendCrash

This commit is contained in:
yihaoDeng 2023-12-15 18:16:38 +08:00
parent 7ab7a64eb3 5f2d93d62c
commit bb5f0d9112
1 changed files with 47 additions and 44 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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