#3562

Committed 20 Dec 2024 09:57AM UTC coverage: 26.655% (-32.2%) from 58.812%

Build # #3562

Build Type

push

travis-ci

Committed by

web-flow

Commit Message

Merge pull request #29229 from taosdata/enh/TS-5749-3.0

enh: seperate tsdb async tasks to different thread pools

Run Details

21498 of 109421 branches covered (19.65%)

Branch coverage included in aggregate %.

66 of 96 new or added lines in 7 files covered. (68.75%)

39441 existing lines in 157 files now uncovered.

35007 of 102566 relevant lines covered (34.13%)

53922.97 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

0.88

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

/*
 * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
 *
 * This program is free software: you can use, redistribute, and/or modify
 * it under the terms of the GNU Affero General Public License, version 3
 * or later ("AGPL"), as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.
 *
 * You should have received a copy of the GNU Affero General Public License
 * along with this program. If not, see <http://www.gnu.org/licenses/>.
 */

#include "tq.h"
#include "vnd.h"

#define MAX_REPEAT_SCAN_THRESHOLD 3
#define SCAN_WAL_IDLE_DURATION    500    // idle for 500ms to do next wal scan

typedef struct SBuildScanWalMsgParam {
  int64_t metaId;
  int32_t numOfTasks;
} SBuildScanWalMsgParam;

static int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta);
static int32_t setWalReaderStartOffset(SStreamTask* pTask, int32_t vgId);
static bool    handleFillhistoryScanComplete(SStreamTask* pTask, int64_t ver);
static bool    taskReadyForDataFromWal(SStreamTask* pTask);
static int32_t doPutDataIntoInputQ(SStreamTask* pTask, int64_t maxVer, int32_t* numOfItems, bool* pSucc);
static int32_t tqScanWalInFuture(STQ* pTq, int32_t numOfTasks, int32_t idleDuration);
static int32_t doScanWalAsync(STQ* pTq, bool ckPause);

// extract data blocks(submit/delete) from WAL, and add them into the input queue for all the sources tasks.
int32_t tqScanWal(STQ* pTq) {
  SStreamMeta* pMeta = pTq->pStreamMeta;
  int32_t      vgId = pMeta->vgId;
  int64_t      st = taosGetTimestampMs();
  int32_t      numOfTasks = 0;

  tqDebug("vgId:%d continue to check if data in wal are available, scanCounter:%d", vgId, pMeta->scanInfo.scanCounter);

  // check all tasks
  int32_t code = doScanWalForAllTasks(pMeta);
  if (code) {
    tqError("vgId:%d failed to start all tasks, try next time, code:%s", vgId, tstrerror(code));
    return code;
  }

  streamMetaWLock(pMeta);
  int32_t times = (--pMeta->scanInfo.scanCounter);
  if (times < 0) {
    tqError("vgId:%d invalid scan counter:%d, reset to 0", vgId, times);
    times = 0;
  }

  numOfTasks = taosArrayGetSize(pMeta->pTaskList);
  streamMetaWUnLock(pMeta);

  int64_t el = (taosGetTimestampMs() - st);
  tqDebug("vgId:%d scan wal for stream tasks completed, elapsed time:%" PRId64 " ms", vgId, el);

  if (times > 0) {
    tqDebug("vgId:%d scan wal for stream tasks for %d times in %dms", vgId, times, SCAN_WAL_IDLE_DURATION);
    code = tqScanWalInFuture(pTq, numOfTasks, SCAN_WAL_IDLE_DURATION);
    if (code) {
      tqError("vgId:%d sched scan wal in %dms failed, ignore this failure", vgId, SCAN_WAL_IDLE_DURATION);
    }
  }

  return code;
}

static void doStartScanWal(void* param, void* tmrId) {
  int32_t vgId = 0;
  STQ*    pTq = NULL;
  int32_t code = 0;

  SBuildScanWalMsgParam* pParam = (SBuildScanWalMsgParam*)param;

  SStreamMeta* pMeta = taosAcquireRef(streamMetaRefPool, pParam->metaId);
  if (pMeta == NULL) {
    tqError("metaRid:%" PRId64 " not valid now, stream meta has been freed", pParam->metaId);
    taosMemoryFree(pParam);
    return;
  }

  vgId = pMeta->vgId;
  pTq = pMeta->ahandle;

  tqDebug("vgId:%d create msg to start wal scan, numOfTasks:%d, vnd restored:%d", vgId, pParam->numOfTasks,
          pTq->pVnode->restored);

  code = streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);
  if (code) {
    tqError("vgId:%d failed sched task to scan wal, code:%s", vgId, tstrerror(code));
  }

  code = taosReleaseRef(streamMetaRefPool, pParam->metaId);
  if (code) {
    tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,
            tstrerror(code));
  }

  taosMemoryFree(pParam);
}

int32_t tqScanWalInFuture(STQ* pTq, int32_t numOfTasks, int32_t idleDuration) {
  SStreamMeta*           pMeta = pTq->pStreamMeta;
  int32_t                code = 0;
  int32_t                vgId = TD_VID(pTq->pVnode);
  tmr_h                  pTimer = NULL;
  SBuildScanWalMsgParam* pParam = NULL;

  pParam = taosMemoryMalloc(sizeof(SBuildScanWalMsgParam));
  if (pParam == NULL) {
    return terrno;
  }

  pParam->metaId = pMeta->rid;
  pParam->numOfTasks = numOfTasks;

  code = streamTimerGetInstance(&pTimer);
  if (code) {
    tqError("vgId:%d failed to get tmr ctrl during sched scan wal", vgId);
    taosMemoryFree(pParam);
  } else {
    streamTmrStart(doStartScanWal, idleDuration, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId, "scan-wal-fut");
  }

  return code;
}

int32_t tqScanWalAsync(STQ* pTq, bool ckPause) {
  SStreamMeta* pMeta = pTq->pStreamMeta;
  bool         alreadyRestored = pTq->pVnode->restored;
  int32_t      code = 0;

  // do not launch the stream tasks, if it is a follower or not restored vnode.
  if (!(vnodeIsRoleLeader(pTq->pVnode) && alreadyRestored)) {
    return TSDB_CODE_SUCCESS;
  }

  streamMetaWLock(pMeta);
  code = doScanWalAsync(pTq, ckPause);
  streamMetaWUnLock(pMeta);
  return code;
}

int32_t tqStopStreamTasksAsync(STQ* pTq) {
  SStreamMeta* pMeta = pTq->pStreamMeta;
  int32_t      vgId = pMeta->vgId;
  return streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_STOP_ALL_TASKS);
}

int32_t setWalReaderStartOffset(SStreamTask* pTask, int32_t vgId) {
  // seek the stored version and extract data from WAL
  int64_t firstVer = walReaderGetValidFirstVer(pTask->exec.pWalReader);
  if (pTask->chkInfo.nextProcessVer < firstVer) {
    tqWarn("vgId:%d s-task:%s ver:%" PRId64 " earlier than the first ver of wal range %" PRId64 ", forward to %" PRId64,
           vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer, firstVer, firstVer);

    pTask->chkInfo.nextProcessVer = firstVer;

    // todo need retry if failed
    int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);
    if (code != TSDB_CODE_SUCCESS) {
      return code;
    }

    // append the data for the stream
    tqDebug("vgId:%d s-task:%s wal reader seek to ver:%" PRId64, vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer);
  } else {
    int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);
    if (currentVer == -1) {  // we only seek the read for the first time
      int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);
      if (code != TSDB_CODE_SUCCESS) {  // no data in wal, quit
        return code;
      }

      // append the data for the stream
      tqDebug("vgId:%d s-task:%s wal reader initial seek to ver:%" PRId64, vgId, pTask->id.idStr,
              pTask->chkInfo.nextProcessVer);
    }
  }

  int64_t skipToVer = walReaderGetSkipToVersion(pTask->exec.pWalReader);
  if (skipToVer != 0 && skipToVer > pTask->chkInfo.nextProcessVer) {
    int32_t code = walReaderSeekVer(pTask->exec.pWalReader, skipToVer);
    if (code != TSDB_CODE_SUCCESS) {  // no data in wal, quit
      return code;
    }

    tqDebug("vgId:%d s-task:%s wal reader jump to ver:%" PRId64, vgId, pTask->id.idStr, skipToVer);
  }

  return TSDB_CODE_SUCCESS;
}

// todo handle memory error
bool handleFillhistoryScanComplete(SStreamTask* pTask, int64_t ver) {
  const char* id = pTask->id.idStr;
  int64_t     maxVer = pTask->step2Range.maxVer;

  if ((pTask->info.fillHistory == 1) && ver > maxVer) {
    if (!pTask->status.appendTranstateBlock) {
      qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the maximum ver:%" PRId64
            ", not scan wal anymore, add transfer-state block into inputQ",
            id, ver, maxVer);

      double el = (taosGetTimestampMs() - pTask->execInfo.step2Start) / 1000.0;
      qDebug("s-task:%s scan-history from WAL stage(step 2) ended, range:%" PRId64 "-%" PRId64 ", elapsed time:%.2fs",
             id, pTask->step2Range.minVer, maxVer, el);
      int32_t code = streamTaskPutTranstateIntoInputQ(pTask);
      if (code) {
        qError("s-task:%s failed to put trans-state into inputQ", id);
      }

      return true;
    } else {
      qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the ver range:%" PRId64 "-%" PRId64
            ", not scan wal",
            id, ver, pTask->step2Range.minVer, maxVer);
    }
  }

  return false;
}

bool taskReadyForDataFromWal(SStreamTask* pTask) {
  // non-source or fill-history tasks don't need to response the WAL scan action.
  SSTaskBasicInfo* pInfo = &pTask->info;
  if ((pInfo->taskLevel != TASK_LEVEL__SOURCE) || (pTask->status.downstreamReady == 0)) {
    return false;
  }

  if (pInfo->taskLevel == TASK_LEVEL__SOURCE && pInfo->trigger == STREAM_TRIGGER_FORCE_WINDOW_CLOSE) {
    return false;
  }

  // not in ready state, do not handle the data from wal
  SStreamTaskState pState = streamTaskGetStatus(pTask);
  if (pState.state != TASK_STATUS__READY) {
    tqTrace("s-task:%s not ready for submit block in wal, status:%s", pTask->id.idStr, pState.name);
    return false;
  }

  // fill-history task has entered into the last phase, no need to anything
  if ((pInfo->fillHistory == 1) && pTask->status.appendTranstateBlock) {
    // the maximum version of data in the WAL has reached already, the step2 is done
    tqDebug("s-task:%s fill-history reach the maximum ver:%" PRId64 ", not scan wal anymore", pTask->id.idStr,
            pTask->dataRange.range.maxVer);
    return false;
  }

  // check if input queue is full or not
  if (streamQueueIsFull(pTask->inputq.queue)) {
    tqTrace("s-task:%s input queue is full, do nothing", pTask->id.idStr);
    return false;
  }

  // the input queue of downstream task is full, so the output is blocked, stopped for a while
  if (pTask->inputq.status == TASK_INPUT_STATUS__BLOCKED) {
    tqDebug("s-task:%s inputQ is blocked, do nothing", pTask->id.idStr);
    return false;
  }

  return true;
}

int32_t doPutDataIntoInputQ(SStreamTask* pTask, int64_t maxVer, int32_t* numOfItems, bool* pSucc) {
  const char* id = pTask->id.idStr;
  int32_t     numOfNewItems = 0;
  int32_t     code = 0;
  *pSucc = false;

  while (1) {
    if ((pTask->info.fillHistory == 1) && pTask->status.appendTranstateBlock) {
      *numOfItems += numOfNewItems;
      return numOfNewItems > 0;
    }

    SStreamQueueItem* pItem = NULL;
    code = extractMsgFromWal(pTask->exec.pWalReader, (void**)&pItem, maxVer, id);
    if (code != TSDB_CODE_SUCCESS || pItem == NULL) {  // failed, continue
      int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);
      bool    itemInFillhistory = handleFillhistoryScanComplete(pTask, currentVer);
      if (itemInFillhistory) {
        numOfNewItems += 1;
      }
      break;
    }

    if (pItem != NULL) {
      code = streamTaskPutDataIntoInputQ(pTask, pItem);
      if (code == TSDB_CODE_SUCCESS) {
        numOfNewItems += 1;
        int64_t ver = walReaderGetCurrentVer(pTask->exec.pWalReader);
        pTask->chkInfo.nextProcessVer = ver;
        tqDebug("s-task:%s set ver:%" PRId64 " for reader after extract data from WAL", id, ver);

        bool itemInFillhistory = handleFillhistoryScanComplete(pTask, ver);
        if (itemInFillhistory) {
          break;
        }
      } else {
        if (code == TSDB_CODE_OUT_OF_MEMORY) {
          tqError("s-task:%s failed to put data into inputQ, since out of memory", id);
        } else {
          tqTrace("s-task:%s append input queue failed, code:inputQ is full, ver:%" PRId64, id,
                  pTask->chkInfo.nextProcessVer);
          code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);
          if (code) {
            tqError("s-task:%s failed to seek ver to:%" PRId64 " in wal", id, pTask->chkInfo.nextProcessVer);
          }

          code = 0;  // reset the error code
        }

        break;
      }
    }
  }

  *numOfItems += numOfNewItems;
  *pSucc = (numOfNewItems > 0);
  return code;
}

int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta) {
  int32_t vgId = pStreamMeta->vgId;
  int32_t numOfTasks = taosArrayGetSize(pStreamMeta->pTaskList);
  if (numOfTasks == 0) {
    return TSDB_CODE_SUCCESS;
  }

  // clone the task list, to avoid the task update during scan wal files
  SArray* pTaskList = NULL;
  streamMetaWLock(pStreamMeta);
  pTaskList = taosArrayDup(pStreamMeta->pTaskList, NULL);
  streamMetaWUnLock(pStreamMeta);
  if (pTaskList == NULL) {
    tqError("vgId:%d failed to create task list dup, code:%s", vgId, tstrerror(terrno));
    return terrno;
  }

  tqDebug("vgId:%d start to check wal to extract new submit block for %d tasks", vgId, numOfTasks);

  // update the new task number
  numOfTasks = taosArrayGetSize(pTaskList);

  for (int32_t i = 0; i < numOfTasks; ++i) {
    STaskId* pTaskId = taosArrayGet(pTaskList, i);
    if (pTaskId == NULL) {
      continue;
    }

    SStreamTask* pTask = NULL;
    int32_t      code = streamMetaAcquireTask(pStreamMeta, pTaskId->streamId, pTaskId->taskId, &pTask);
    if (pTask == NULL || code != 0) {
      continue;
    }

    if (!taskReadyForDataFromWal(pTask)) {
      streamMetaReleaseTask(pStreamMeta, pTask);
      continue;
    }

    // seek the stored version and extract data from WAL
    code = setWalReaderStartOffset(pTask, vgId);
    if (code != TSDB_CODE_SUCCESS) {
      streamMetaReleaseTask(pStreamMeta, pTask);
      continue;
    }

    int32_t numOfItems = streamQueueGetNumOfItems(pTask->inputq.queue);
    int64_t maxVer = (pTask->info.fillHistory == 1) ? pTask->step2Range.maxVer : INT64_MAX;

    streamMutexLock(&pTask->lock);

    SStreamTaskState state = streamTaskGetStatus(pTask);
    if (state.state != TASK_STATUS__READY) {
      tqDebug("s-task:%s not ready for submit block from wal, status:%s", pTask->id.idStr, state.name);
      streamMutexUnlock(&pTask->lock);
      streamMetaReleaseTask(pStreamMeta, pTask);
      continue;
    }

    bool hasNewData = false;
    code = doPutDataIntoInputQ(pTask, maxVer, &numOfItems, &hasNewData);
    streamMutexUnlock(&pTask->lock);

    if ((numOfItems > 0) || hasNewData) {
      code = streamTrySchedExec(pTask);
      if (code != TSDB_CODE_SUCCESS) {
        streamMetaReleaseTask(pStreamMeta, pTask);
        taosArrayDestroy(pTaskList);
        return code;
      }
    }

    streamMetaReleaseTask(pStreamMeta, pTask);
  }

  taosArrayDestroy(pTaskList);
  return TSDB_CODE_SUCCESS;
}

int32_t doScanWalAsync(STQ* pTq, bool ckPause) {
  SStreamMeta* pMeta = pTq->pStreamMeta;
  bool         alreadyRestored = pTq->pVnode->restored;
  int32_t      vgId = pMeta->vgId;
  int32_t      numOfTasks = taosArrayGetSize(pMeta->pTaskList);

  if (numOfTasks == 0) {
    tqDebug("vgId:%d no stream tasks existed to run", vgId);
    return 0;
  }

  if (pMeta->startInfo.startAllTasks) {
    tqTrace("vgId:%d in restart procedure, not scan wal", vgId);
    return 0;
  }

  pMeta->scanInfo.scanCounter += 1;
  if (pMeta->scanInfo.scanCounter > MAX_REPEAT_SCAN_THRESHOLD) {
    pMeta->scanInfo.scanCounter = MAX_REPEAT_SCAN_THRESHOLD;
  }

  if (pMeta->scanInfo.scanCounter > 1) {
    tqDebug("vgId:%d wal read task has been launched, remain scan times:%d", vgId, pMeta->scanInfo.scanCounter);
    return 0;
  }

  int32_t numOfPauseTasks = pMeta->numOfPausedTasks;
  if (ckPause && numOfTasks == numOfPauseTasks) {
    tqDebug("vgId:%d ignore all submit, all streams had been paused, reset the walScanCounter", vgId);

    // reset the counter value, since we do not launch the scan wal operation.
    pMeta->scanInfo.scanCounter = 0;
    return 0;
  }

  tqDebug("vgId:%d create msg to start wal scan to launch stream tasks, numOfTasks:%d, vnd restored:%d", vgId,
          numOfTasks, alreadyRestored);

  return streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);
}

1	/*
2	* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
3	*
4	* This program is free software: you can use, redistribute, and/or modify
5	* it under the terms of the GNU Affero General Public License, version 3
6	* or later ("AGPL"), as published by the Free Software Foundation.
7	*
8	* This program is distributed in the hope that it will be useful, but WITHOUT
9	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10	* FITNESS FOR A PARTICULAR PURPOSE.
11	*
12	* You should have received a copy of the GNU Affero General Public License
13	* along with this program. If not, see <http://www.gnu.org/licenses/>.
14	*/
15
16	#include "tq.h"
17	#include "vnd.h"
18
19	#define MAX_REPEAT_SCAN_THRESHOLD 3
20	#define SCAN_WAL_IDLE_DURATION 500 // idle for 500ms to do next wal scan
21
22	typedef struct SBuildScanWalMsgParam {
23	int64_t metaId;
24	int32_t numOfTasks;
25	} SBuildScanWalMsgParam;
26
27	static int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta);
28	static int32_t setWalReaderStartOffset(SStreamTask* pTask, int32_t vgId);
29	static bool handleFillhistoryScanComplete(SStreamTask* pTask, int64_t ver);
30	static bool taskReadyForDataFromWal(SStreamTask* pTask);
31	static int32_t doPutDataIntoInputQ(SStreamTask* pTask, int64_t maxVer, int32_t* numOfItems, bool* pSucc);
32	static int32_t tqScanWalInFuture(STQ* pTq, int32_t numOfTasks, int32_t idleDuration);
33	static int32_t doScanWalAsync(STQ* pTq, bool ckPause);
34
35	// extract data blocks(submit/delete) from WAL, and add them into the input queue for all the sources tasks.
UNCOV 36	int32_t tqScanWal(STQ* pTq) {	×
UNCOV 37	SStreamMeta* pMeta = pTq->pStreamMeta;	×
UNCOV 38	int32_t vgId = pMeta->vgId;	×
UNCOV 39	int64_t st = taosGetTimestampMs();	×
UNCOV 40	int32_t numOfTasks = 0;	×
41
UNCOV 42	tqDebug("vgId:%d continue to check if data in wal are available, scanCounter:%d", vgId, pMeta->scanInfo.scanCounter);	×
43
44	// check all tasks
UNCOV 45	int32_t code = doScanWalForAllTasks(pMeta);	×
UNCOV 46	if (code) {	×
47	tqError("vgId:%d failed to start all tasks, try next time, code:%s", vgId, tstrerror(code));	×
48	return code;	×
49	}
50
UNCOV 51	streamMetaWLock(pMeta);	×
UNCOV 52	int32_t times = (--pMeta->scanInfo.scanCounter);	×
UNCOV 53	if (times < 0) {	×
54	tqError("vgId:%d invalid scan counter:%d, reset to 0", vgId, times);	×
55	times = 0;	×
56	}
57
UNCOV 58	numOfTasks = taosArrayGetSize(pMeta->pTaskList);	×
UNCOV 59	streamMetaWUnLock(pMeta);	×
60
UNCOV 61	int64_t el = (taosGetTimestampMs() - st);	×
UNCOV 62	tqDebug("vgId:%d scan wal for stream tasks completed, elapsed time:%" PRId64 " ms", vgId, el);	×
63
UNCOV 64	if (times > 0) {	×
UNCOV 65	tqDebug("vgId:%d scan wal for stream tasks for %d times in %dms", vgId, times, SCAN_WAL_IDLE_DURATION);	×
UNCOV 66	code = tqScanWalInFuture(pTq, numOfTasks, SCAN_WAL_IDLE_DURATION);	×
UNCOV 67	if (code) {	×
68	tqError("vgId:%d sched scan wal in %dms failed, ignore this failure", vgId, SCAN_WAL_IDLE_DURATION);	×
69	}
70	}
71
UNCOV 72	return code;	×
73	}
74
UNCOV 75	static void doStartScanWal(void* param, void* tmrId) {	×
UNCOV 76	int32_t vgId = 0;	×
UNCOV 77	STQ* pTq = NULL;	×
UNCOV 78	int32_t code = 0;	×
79
UNCOV 80	SBuildScanWalMsgParam* pParam = (SBuildScanWalMsgParam*)param;	×
81
UNCOV 82	SStreamMeta* pMeta = taosAcquireRef(streamMetaRefPool, pParam->metaId);	×
UNCOV 83	if (pMeta == NULL) {	×
UNCOV 84	tqError("metaRid:%" PRId64 " not valid now, stream meta has been freed", pParam->metaId);	×
UNCOV 85	taosMemoryFree(pParam);	×
UNCOV 86	return;	×
87	}
88
UNCOV 89	vgId = pMeta->vgId;	×
UNCOV 90	pTq = pMeta->ahandle;	×
91
UNCOV 92	tqDebug("vgId:%d create msg to start wal scan, numOfTasks:%d, vnd restored:%d", vgId, pParam->numOfTasks,	×
93	pTq->pVnode->restored);
94
UNCOV 95	code = streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);	×
UNCOV 96	if (code) {	×
UNCOV 97	tqError("vgId:%d failed sched task to scan wal, code:%s", vgId, tstrerror(code));	×
98	}
99
UNCOV 100	code = taosReleaseRef(streamMetaRefPool, pParam->metaId);	×
UNCOV 101	if (code) {	×
102	tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,	×
103	tstrerror(code));
104	}
105
UNCOV 106	taosMemoryFree(pParam);	×
107	}
108
UNCOV 109	int32_t tqScanWalInFuture(STQ* pTq, int32_t numOfTasks, int32_t idleDuration) {	×
UNCOV 110	SStreamMeta* pMeta = pTq->pStreamMeta;	×
UNCOV 111	int32_t code = 0;	×
UNCOV 112	int32_t vgId = TD_VID(pTq->pVnode);	×
UNCOV 113	tmr_h pTimer = NULL;	×
UNCOV 114	SBuildScanWalMsgParam* pParam = NULL;	×
115
UNCOV 116	pParam = taosMemoryMalloc(sizeof(SBuildScanWalMsgParam));	×
UNCOV 117	if (pParam == NULL) {	×
118	return terrno;	×
119	}
120
UNCOV 121	pParam->metaId = pMeta->rid;	×
UNCOV 122	pParam->numOfTasks = numOfTasks;	×
123
UNCOV 124	code = streamTimerGetInstance(&pTimer);	×
UNCOV 125	if (code) {	×
126	tqError("vgId:%d failed to get tmr ctrl during sched scan wal", vgId);	×
127	taosMemoryFree(pParam);	×
128	} else {
UNCOV 129	streamTmrStart(doStartScanWal, idleDuration, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId, "scan-wal-fut");	×
130	}
131
UNCOV 132	return code;	×
133	}
134
UNCOV 135	int32_t tqScanWalAsync(STQ* pTq, bool ckPause) {	×
UNCOV 136	SStreamMeta* pMeta = pTq->pStreamMeta;	×
UNCOV 137	bool alreadyRestored = pTq->pVnode->restored;	×
UNCOV 138	int32_t code = 0;	×
139
140	// do not launch the stream tasks, if it is a follower or not restored vnode.
UNCOV 141	if (!(vnodeIsRoleLeader(pTq->pVnode) && alreadyRestored)) {	×
UNCOV 142	return TSDB_CODE_SUCCESS;	×
143	}
144
UNCOV 145	streamMetaWLock(pMeta);	×
UNCOV 146	code = doScanWalAsync(pTq, ckPause);	×
UNCOV 147	streamMetaWUnLock(pMeta);	×
UNCOV 148	return code;	×
149	}
150
151	int32_t tqStopStreamTasksAsync(STQ* pTq) {	824✔
152	SStreamMeta* pMeta = pTq->pStreamMeta;	824✔
153	int32_t vgId = pMeta->vgId;	824✔
154	return streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_STOP_ALL_TASKS);	824✔
155	}
156
UNCOV 157	int32_t setWalReaderStartOffset(SStreamTask* pTask, int32_t vgId) {	×
158	// seek the stored version and extract data from WAL
UNCOV 159	int64_t firstVer = walReaderGetValidFirstVer(pTask->exec.pWalReader);	×
UNCOV 160	if (pTask->chkInfo.nextProcessVer < firstVer) {	×
UNCOV 161	tqWarn("vgId:%d s-task:%s ver:%" PRId64 " earlier than the first ver of wal range %" PRId64 ", forward to %" PRId64,	×
162	vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer, firstVer, firstVer);
163
UNCOV 164	pTask->chkInfo.nextProcessVer = firstVer;	×
165
166	// todo need retry if failed
UNCOV 167	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	×
UNCOV 168	if (code != TSDB_CODE_SUCCESS) {	×
169	return code;	×
170	}
171
172	// append the data for the stream
UNCOV 173	tqDebug("vgId:%d s-task:%s wal reader seek to ver:%" PRId64, vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer);	×
174	} else {
UNCOV 175	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	×
UNCOV 176	if (currentVer == -1) { // we only seek the read for the first time	×
UNCOV 177	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	×
UNCOV 178	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	×
179	return code;	×
180	}
181
182	// append the data for the stream
UNCOV 183	tqDebug("vgId:%d s-task:%s wal reader initial seek to ver:%" PRId64, vgId, pTask->id.idStr,	×
184	pTask->chkInfo.nextProcessVer);
185	}
186	}
187
UNCOV 188	int64_t skipToVer = walReaderGetSkipToVersion(pTask->exec.pWalReader);	×
UNCOV 189	if (skipToVer != 0 && skipToVer > pTask->chkInfo.nextProcessVer) {	×
UNCOV 190	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, skipToVer);	×
UNCOV 191	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	×
192	return code;	×
193	}
194
UNCOV 195	tqDebug("vgId:%d s-task:%s wal reader jump to ver:%" PRId64, vgId, pTask->id.idStr, skipToVer);	×
196	}
197
UNCOV 198	return TSDB_CODE_SUCCESS;	×
199	}
200
201	// todo handle memory error
UNCOV 202	bool handleFillhistoryScanComplete(SStreamTask* pTask, int64_t ver) {	×
UNCOV 203	const char* id = pTask->id.idStr;	×
UNCOV 204	int64_t maxVer = pTask->step2Range.maxVer;	×
205
UNCOV 206	if ((pTask->info.fillHistory == 1) && ver > maxVer) {	×
UNCOV 207	if (!pTask->status.appendTranstateBlock) {	×
UNCOV 208	qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the maximum ver:%" PRId64	×
209	", not scan wal anymore, add transfer-state block into inputQ",
210	id, ver, maxVer);
211
UNCOV 212	double el = (taosGetTimestampMs() - pTask->execInfo.step2Start) / 1000.0;	×
UNCOV 213	qDebug("s-task:%s scan-history from WAL stage(step 2) ended, range:%" PRId64 "-%" PRId64 ", elapsed time:%.2fs",	×
214	id, pTask->step2Range.minVer, maxVer, el);
UNCOV 215	int32_t code = streamTaskPutTranstateIntoInputQ(pTask);	×
UNCOV 216	if (code) {	×
217	qError("s-task:%s failed to put trans-state into inputQ", id);	×
218	}
219
UNCOV 220	return true;	×
221	} else {
222	qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the ver range:%" PRId64 "-%" PRId64	×
223	", not scan wal",
224	id, ver, pTask->step2Range.minVer, maxVer);
225	}
226	}
227
UNCOV 228	return false;	×
229	}
230
UNCOV 231	bool taskReadyForDataFromWal(SStreamTask* pTask) {	×
232	// non-source or fill-history tasks don't need to response the WAL scan action.
UNCOV 233	SSTaskBasicInfo* pInfo = &pTask->info;	×
UNCOV 234	if ((pInfo->taskLevel != TASK_LEVEL__SOURCE) \|\| (pTask->status.downstreamReady == 0)) {	×
UNCOV 235	return false;	×
236	}
237
UNCOV 238	if (pInfo->taskLevel == TASK_LEVEL__SOURCE && pInfo->trigger == STREAM_TRIGGER_FORCE_WINDOW_CLOSE) {	×
UNCOV 239	return false;	×
240	}
241
242	// not in ready state, do not handle the data from wal
UNCOV 243	SStreamTaskState pState = streamTaskGetStatus(pTask);	×
UNCOV 244	if (pState.state != TASK_STATUS__READY) {	×
UNCOV 245	tqTrace("s-task:%s not ready for submit block in wal, status:%s", pTask->id.idStr, pState.name);	×
UNCOV 246	return false;	×
247	}
248
249	// fill-history task has entered into the last phase, no need to anything
UNCOV 250	if ((pInfo->fillHistory == 1) && pTask->status.appendTranstateBlock) {	×
251	// the maximum version of data in the WAL has reached already, the step2 is done
UNCOV 252	tqDebug("s-task:%s fill-history reach the maximum ver:%" PRId64 ", not scan wal anymore", pTask->id.idStr,	×
253	pTask->dataRange.range.maxVer);
UNCOV 254	return false;	×
255	}
256
257	// check if input queue is full or not
UNCOV 258	if (streamQueueIsFull(pTask->inputq.queue)) {	×
UNCOV 259	tqTrace("s-task:%s input queue is full, do nothing", pTask->id.idStr);	×
UNCOV 260	return false;	×
261	}
262
263	// the input queue of downstream task is full, so the output is blocked, stopped for a while
UNCOV 264	if (pTask->inputq.status == TASK_INPUT_STATUS__BLOCKED) {	×
265	tqDebug("s-task:%s inputQ is blocked, do nothing", pTask->id.idStr);	×
266	return false;	×
267	}
268
UNCOV 269	return true;	×
270	}
271
UNCOV 272	int32_t doPutDataIntoInputQ(SStreamTask* pTask, int64_t maxVer, int32_t* numOfItems, bool* pSucc) {	×
UNCOV 273	const char* id = pTask->id.idStr;	×
UNCOV 274	int32_t numOfNewItems = 0;	×
UNCOV 275	int32_t code = 0;	×
UNCOV 276	*pSucc = false;	×
277
UNCOV 278	while (1) {	×
UNCOV 279	if ((pTask->info.fillHistory == 1) && pTask->status.appendTranstateBlock) {	×
280	*numOfItems += numOfNewItems;	×
281	return numOfNewItems > 0;	×
282	}
283
UNCOV 284	SStreamQueueItem* pItem = NULL;	×
UNCOV 285	code = extractMsgFromWal(pTask->exec.pWalReader, (void**)&pItem, maxVer, id);	×
UNCOV 286	if (code != TSDB_CODE_SUCCESS \|\| pItem == NULL) { // failed, continue	×
UNCOV 287	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	×
UNCOV 288	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, currentVer);	×
UNCOV 289	if (itemInFillhistory) {	×
UNCOV 290	numOfNewItems += 1;	×
291	}
UNCOV 292	break;	×
293	}
294
UNCOV 295	if (pItem != NULL) {	×
UNCOV 296	code = streamTaskPutDataIntoInputQ(pTask, pItem);	×
UNCOV 297	if (code == TSDB_CODE_SUCCESS) {	×
UNCOV 298	numOfNewItems += 1;	×
UNCOV 299	int64_t ver = walReaderGetCurrentVer(pTask->exec.pWalReader);	×
UNCOV 300	pTask->chkInfo.nextProcessVer = ver;	×
UNCOV 301	tqDebug("s-task:%s set ver:%" PRId64 " for reader after extract data from WAL", id, ver);	×
302
UNCOV 303	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, ver);	×
UNCOV 304	if (itemInFillhistory) {	×
UNCOV 305	break;	×
306	}
307	} else {
UNCOV 308	if (code == TSDB_CODE_OUT_OF_MEMORY) {	×
309	tqError("s-task:%s failed to put data into inputQ, since out of memory", id);	×
310	} else {
UNCOV 311	tqTrace("s-task:%s append input queue failed, code:inputQ is full, ver:%" PRId64, id,	×
312	pTask->chkInfo.nextProcessVer);
UNCOV 313	code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	×
UNCOV 314	if (code) {	×
315	tqError("s-task:%s failed to seek ver to:%" PRId64 " in wal", id, pTask->chkInfo.nextProcessVer);	×
316	}
317
UNCOV 318	code = 0; // reset the error code	×
319	}
320
UNCOV 321	break;	×
322	}
323	}
324	}
325
UNCOV 326	*numOfItems += numOfNewItems;	×
UNCOV 327	*pSucc = (numOfNewItems > 0);	×
UNCOV 328	return code;	×
329	}
330
UNCOV 331	int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta) {	×
UNCOV 332	int32_t vgId = pStreamMeta->vgId;	×
UNCOV 333	int32_t numOfTasks = taosArrayGetSize(pStreamMeta->pTaskList);	×
UNCOV 334	if (numOfTasks == 0) {	×
UNCOV 335	return TSDB_CODE_SUCCESS;	×
336	}
337
338	// clone the task list, to avoid the task update during scan wal files
UNCOV 339	SArray* pTaskList = NULL;	×
UNCOV 340	streamMetaWLock(pStreamMeta);	×
UNCOV 341	pTaskList = taosArrayDup(pStreamMeta->pTaskList, NULL);	×
UNCOV 342	streamMetaWUnLock(pStreamMeta);	×
UNCOV 343	if (pTaskList == NULL) {	×
344	tqError("vgId:%d failed to create task list dup, code:%s", vgId, tstrerror(terrno));	×
345	return terrno;	×
346	}
347
UNCOV 348	tqDebug("vgId:%d start to check wal to extract new submit block for %d tasks", vgId, numOfTasks);	×
349
350	// update the new task number
UNCOV 351	numOfTasks = taosArrayGetSize(pTaskList);	×
352
UNCOV 353	for (int32_t i = 0; i < numOfTasks; ++i) {	×
UNCOV 354	STaskId* pTaskId = taosArrayGet(pTaskList, i);	×
UNCOV 355	if (pTaskId == NULL) {	×
UNCOV 356	continue;	×
357	}
358
UNCOV 359	SStreamTask* pTask = NULL;	×
UNCOV 360	int32_t code = streamMetaAcquireTask(pStreamMeta, pTaskId->streamId, pTaskId->taskId, &pTask);	×
UNCOV 361	if (pTask == NULL \|\| code != 0) {	×
UNCOV 362	continue;	×
363	}
364
UNCOV 365	if (!taskReadyForDataFromWal(pTask)) {	×
UNCOV 366	streamMetaReleaseTask(pStreamMeta, pTask);	×
UNCOV 367	continue;	×
368	}
369
370	// seek the stored version and extract data from WAL
UNCOV 371	code = setWalReaderStartOffset(pTask, vgId);	×
UNCOV 372	if (code != TSDB_CODE_SUCCESS) {	×
373	streamMetaReleaseTask(pStreamMeta, pTask);	×
374	continue;	×
375	}
376
UNCOV 377	int32_t numOfItems = streamQueueGetNumOfItems(pTask->inputq.queue);	×
UNCOV 378	int64_t maxVer = (pTask->info.fillHistory == 1) ? pTask->step2Range.maxVer : INT64_MAX;	×
379
UNCOV 380	streamMutexLock(&pTask->lock);	×
381
UNCOV 382	SStreamTaskState state = streamTaskGetStatus(pTask);	×
UNCOV 383	if (state.state != TASK_STATUS__READY) {	×
UNCOV 384	tqDebug("s-task:%s not ready for submit block from wal, status:%s", pTask->id.idStr, state.name);	×
UNCOV 385	streamMutexUnlock(&pTask->lock);	×
UNCOV 386	streamMetaReleaseTask(pStreamMeta, pTask);	×
UNCOV 387	continue;	×
388	}
389
UNCOV 390	bool hasNewData = false;	×
UNCOV 391	code = doPutDataIntoInputQ(pTask, maxVer, &numOfItems, &hasNewData);	×
UNCOV 392	streamMutexUnlock(&pTask->lock);	×
393
UNCOV 394	if ((numOfItems > 0) \|\| hasNewData) {	×
UNCOV 395	code = streamTrySchedExec(pTask);	×
UNCOV 396	if (code != TSDB_CODE_SUCCESS) {	×
397	streamMetaReleaseTask(pStreamMeta, pTask);	×
398	taosArrayDestroy(pTaskList);	×
399	return code;	×
400	}
401	}
402
UNCOV 403	streamMetaReleaseTask(pStreamMeta, pTask);	×
404	}
405
UNCOV 406	taosArrayDestroy(pTaskList);	×
UNCOV 407	return TSDB_CODE_SUCCESS;	×
408	}
409
UNCOV 410	int32_t doScanWalAsync(STQ* pTq, bool ckPause) {	×
UNCOV 411	SStreamMeta* pMeta = pTq->pStreamMeta;	×
UNCOV 412	bool alreadyRestored = pTq->pVnode->restored;	×
UNCOV 413	int32_t vgId = pMeta->vgId;	×
UNCOV 414	int32_t numOfTasks = taosArrayGetSize(pMeta->pTaskList);	×
415
UNCOV 416	if (numOfTasks == 0) {	×
UNCOV 417	tqDebug("vgId:%d no stream tasks existed to run", vgId);	×
UNCOV 418	return 0;	×
419	}
420
UNCOV 421	if (pMeta->startInfo.startAllTasks) {	×
UNCOV 422	tqTrace("vgId:%d in restart procedure, not scan wal", vgId);	×
UNCOV 423	return 0;	×
424	}
425
UNCOV 426	pMeta->scanInfo.scanCounter += 1;	×
UNCOV 427	if (pMeta->scanInfo.scanCounter > MAX_REPEAT_SCAN_THRESHOLD) {	×
UNCOV 428	pMeta->scanInfo.scanCounter = MAX_REPEAT_SCAN_THRESHOLD;	×
429	}
430
UNCOV 431	if (pMeta->scanInfo.scanCounter > 1) {	×
UNCOV 432	tqDebug("vgId:%d wal read task has been launched, remain scan times:%d", vgId, pMeta->scanInfo.scanCounter);	×
UNCOV 433	return 0;	×
434	}
435
UNCOV 436	int32_t numOfPauseTasks = pMeta->numOfPausedTasks;	×
UNCOV 437	if (ckPause && numOfTasks == numOfPauseTasks) {	×
UNCOV 438	tqDebug("vgId:%d ignore all submit, all streams had been paused, reset the walScanCounter", vgId);	×
439
440	// reset the counter value, since we do not launch the scan wal operation.
UNCOV 441	pMeta->scanInfo.scanCounter = 0;	×
UNCOV 442	return 0;	×
443	}
444
UNCOV 445	tqDebug("vgId:%d create msg to start wal scan to launch stream tasks, numOfTasks:%d, vnd restored:%d", vgId,	×
446	numOfTasks, alreadyRestored);
447
UNCOV 448	return streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);	×
449	}

taosdata / TDengine / #3562

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous