#3519

Committed 05 Nov 2024 11:19AM UTC coverage: 57.706% (+8.4%) from 49.32%

Build # #3519

Build Type

push

travis-ci

Committed by

web-flow

Commit Message

Merge pull request #28652 from taosdata/fix/3_liaohj

refactor: always successfully put the retrieve msg

Run Details

109445 of 245179 branches covered (44.64%)

Branch coverage included in aggregate %.

187435 of 269288 relevant lines covered (69.6%)

12869818.21 hits per line

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

76.7

/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    100

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

static int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta, bool* pScanIdle);
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);

// 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;
  bool         shouldIdle = true;

  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, &shouldIdle);
  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) {
  int32_t      vgId = TD_VID(pTq->pVnode);
  SStreamMeta* pMeta = pTq->pStreamMeta;
  bool         alreadyRestored = pTq->pVnode->restored;
  int32_t      numOfTasks = 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);

  numOfTasks = taosArrayGetSize(pMeta->pTaskList);
  if (numOfTasks == 0) {
    tqDebug("vgId:%d no stream tasks existed to run", vgId);
    streamMetaWUnLock(pMeta);
    return 0;
  }

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

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

  int32_t code = streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);
  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.
  if ((pTask->info.taskLevel != TASK_LEVEL__SOURCE) || (pTask->status.downstreamReady == 0)) {
    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 ((pTask->info.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, bool* pScanIdle) {
  *pScanIdle = true;
  bool    noDataInWal = true;
  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;
    }

    *pScanIdle = false;

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

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

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

    streamMetaReleaseTask(pStreamMeta, pTask);
  }

  // all wal are checked, and no new data available in wal.
  if (noDataInWal) {
    *pScanIdle = true;
  }

  taosArrayDestroy(pTaskList);
  return TSDB_CODE_SUCCESS;
}

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 100
21
22	typedef struct SBuildScanWalMsgParam {
23	int64_t metaId;
24	int32_t numOfTasks;
25	} SBuildScanWalMsgParam;
26
27	static int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta, bool* pScanIdle);
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
34	// extract data blocks(submit/delete) from WAL, and add them into the input queue for all the sources tasks.
35	int32_t tqScanWal(STQ* pTq) {	50,797✔
36	SStreamMeta* pMeta = pTq->pStreamMeta;	50,797✔
37	int32_t vgId = pMeta->vgId;	50,797✔
38	int64_t st = taosGetTimestampMs();	50,798✔
39	int32_t numOfTasks = 0;	50,798✔
40	bool shouldIdle = true;	50,798✔
41
42	tqDebug("vgId:%d continue to check if data in wal are available, scanCounter:%d", vgId, pMeta->scanInfo.scanCounter);	50,798✔
43
44	// check all tasks
45	int32_t code = doScanWalForAllTasks(pMeta, &shouldIdle);	50,799✔
46	if (code) {	50,776!
47	tqError("vgId:%d failed to start all tasks, try next time, code:%s", vgId, tstrerror(code));	×
48	return code;	×
49	}
50
51	streamMetaWLock(pMeta);	50,776✔
52	int32_t times = (--pMeta->scanInfo.scanCounter);	50,785✔
53	if (times < 0) {	50,785!
54	tqError("vgId:%d invalid scan counter:%d, reset to 0", vgId, times);	×
55	times = 0;	×
56	}
57
58	numOfTasks = taosArrayGetSize(pMeta->pTaskList);	50,785✔
59	streamMetaWUnLock(pMeta);	50,780✔
60
61	int64_t el = (taosGetTimestampMs() - st);	50,783✔
62	tqDebug("vgId:%d scan wal for stream tasks completed, elapsed time:%" PRId64 " ms", vgId, el);	50,783✔
63
64	if (times > 0) {	50,783✔
65	tqDebug("vgId:%d scan wal for stream tasks for %d times in %dms", vgId, times, SCAN_WAL_IDLE_DURATION);	15,714✔
66	code = tqScanWalInFuture(pTq, numOfTasks, SCAN_WAL_IDLE_DURATION);	15,714✔
67	if (code) {	15,726!
68	tqError("vgId:%d sched scan wal in %dms failed, ignore this failure", vgId, SCAN_WAL_IDLE_DURATION);	×
69	}
70	}
71
72	return code;	50,788✔
73	}
74
75	static void doStartScanWal(void* param, void* tmrId) {	15,726✔
76	int32_t vgId = 0;	15,726✔
77	STQ* pTq = NULL;	15,726✔
78	int32_t code = 0;	15,726✔
79
80	SBuildScanWalMsgParam* pParam = (SBuildScanWalMsgParam*)param;	15,726✔
81
82	SStreamMeta* pMeta = taosAcquireRef(streamMetaRefPool, pParam->metaId);	15,726✔
83	if (pMeta == NULL) {	15,726!
84	tqError("metaRid:%" PRId64 " not valid now, stream meta has been freed", pParam->metaId);	×
85	taosMemoryFree(pParam);	×
86	return;	×
87	}
88
89	vgId = pMeta->vgId;	15,726✔
90	pTq = pMeta->ahandle;	15,726✔
91
92	tqDebug("vgId:%d create msg to start wal scan, numOfTasks:%d, vnd restored:%d", vgId, pParam->numOfTasks,	15,726✔
93	pTq->pVnode->restored);
94
95	code = streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);	15,726✔
96	if (code) {	15,726✔
97	tqError("vgId:%d failed sched task to scan wal, code:%s", vgId, tstrerror(code));	102!
98	}
99
100	code = taosReleaseRef(streamMetaRefPool, pParam->metaId);	15,726✔
101	if (code) {	15,726!
102	tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,	×
103	tstrerror(code));
104	}
105
106	taosMemoryFree(pParam);	15,726✔
107	}
108
109	int32_t tqScanWalInFuture(STQ* pTq, int32_t numOfTasks, int32_t idleDuration) {	15,711✔
110	SStreamMeta* pMeta = pTq->pStreamMeta;	15,711✔
111	int32_t code = 0;	15,711✔
112	int32_t vgId = TD_VID(pTq->pVnode);	15,711✔
113	tmr_h pTimer = NULL;	15,711✔
114	SBuildScanWalMsgParam* pParam = NULL;	15,711✔
115
116	pParam = taosMemoryMalloc(sizeof(SBuildScanWalMsgParam));	15,711✔
117	if (pParam == NULL) {	15,717!
118	return terrno;	×
119	}
120
121	pParam->metaId = pMeta->rid;	15,717✔
122	pParam->numOfTasks = numOfTasks;	15,717✔
123
124	code = streamTimerGetInstance(&pTimer);	15,717✔
125	if (code) {	15,710!
126	tqError("vgId:%d failed to get tmr ctrl during sched scan wal", vgId);	×
127	taosMemoryFree(pParam);	×
128	} else {
129	streamTmrStart(doStartScanWal, idleDuration, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId, "scan-wal-fut");	15,710✔
130	}
131
132	return code;	15,726✔
133	}
134
135	int32_t tqScanWalAsync(STQ* pTq, bool ckPause) {	444,687✔
136	int32_t vgId = TD_VID(pTq->pVnode);	444,687✔
137	SStreamMeta* pMeta = pTq->pStreamMeta;	444,687✔
138	bool alreadyRestored = pTq->pVnode->restored;	444,687✔
139	int32_t numOfTasks = 0;	444,687✔
140
141	// do not launch the stream tasks, if it is a follower or not restored vnode.
142	if (!(vnodeIsRoleLeader(pTq->pVnode) && alreadyRestored)) {	444,687✔
143	return TSDB_CODE_SUCCESS;	42,469✔
144	}
145
146	streamMetaWLock(pMeta);	402,299✔
147
148	numOfTasks = taosArrayGetSize(pMeta->pTaskList);	402,288✔
149	if (numOfTasks == 0) {	402,263✔
150	tqDebug("vgId:%d no stream tasks existed to run", vgId);	2!
151	streamMetaWUnLock(pMeta);	2✔
152	return 0;	2✔
153	}
154
155	if (pMeta->startInfo.startAllTasks) {	402,261✔
156	tqTrace("vgId:%d in restart procedure, not scan wal", vgId);	13,491!
157	streamMetaWUnLock(pMeta);	13,491✔
158	return 0;	13,493✔
159	}
160
161	pMeta->scanInfo.scanCounter += 1;	388,770✔
162	if (pMeta->scanInfo.scanCounter > MAX_REPEAT_SCAN_THRESHOLD) {	388,770✔
163	pMeta->scanInfo.scanCounter = MAX_REPEAT_SCAN_THRESHOLD;	337,130✔
164	}
165
166	if (pMeta->scanInfo.scanCounter > 1) {	388,770✔
167	tqDebug("vgId:%d wal read task has been launched, remain scan times:%d", vgId, pMeta->scanInfo.scanCounter);	353,031✔
168	streamMetaWUnLock(pMeta);	353,033✔
169	return 0;	353,016✔
170	}
171
172	int32_t numOfPauseTasks = pMeta->numOfPausedTasks;	35,739✔
173	if (ckPause && numOfTasks == numOfPauseTasks) {	35,739✔
174	tqDebug("vgId:%d ignore all submit, all streams had been paused, reset the walScanCounter", vgId);	127✔
175
176	// reset the counter value, since we do not launch the scan wal operation.
177	pMeta->scanInfo.scanCounter = 0;	127✔
178	streamMetaWUnLock(pMeta);	127✔
179	return 0;	127✔
180	}
181
182	tqDebug("vgId:%d create msg to start wal scan to launch stream tasks, numOfTasks:%d, vnd restored:%d", vgId,	35,612✔
183	numOfTasks, alreadyRestored);
184
185	int32_t code = streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);	35,612✔
186	streamMetaWUnLock(pMeta);	35,646✔
187
188	return code;	35,648✔
189	}
190
191	int32_t tqStopStreamTasksAsync(STQ* pTq) {	2,380✔
192	SStreamMeta* pMeta = pTq->pStreamMeta;	2,380✔
193	int32_t vgId = pMeta->vgId;	2,380✔
194	return streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_STOP_ALL_TASKS);	2,380✔
195	}
196
197	int32_t setWalReaderStartOffset(SStreamTask* pTask, int32_t vgId) {	62,095✔
198	// seek the stored version and extract data from WAL
199	int64_t firstVer = walReaderGetValidFirstVer(pTask->exec.pWalReader);	62,095✔
200	if (pTask->chkInfo.nextProcessVer < firstVer) {	62,106✔
201	tqWarn("vgId:%d s-task:%s ver:%" PRId64 " earlier than the first ver of wal range %" PRId64 ", forward to %" PRId64,	21!
202	vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer, firstVer, firstVer);
203
204	pTask->chkInfo.nextProcessVer = firstVer;	21✔
205
206	// todo need retry if failed
207	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	21✔
208	if (code != TSDB_CODE_SUCCESS) {	21!
209	return code;	×
210	}
211
212	// append the data for the stream
213	tqDebug("vgId:%d s-task:%s wal reader seek to ver:%" PRId64, vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer);	21✔
214	} else {
215	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	62,085✔
216	if (currentVer == -1) { // we only seek the read for the first time	62,071✔
217	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	2,964✔
218	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	2,963!
219	return code;	×
220	}
221
222	// append the data for the stream
223	tqDebug("vgId:%d s-task:%s wal reader initial seek to ver:%" PRId64, vgId, pTask->id.idStr,	2,963✔
224	pTask->chkInfo.nextProcessVer);
225	}
226	}
227
228	int64_t skipToVer = walReaderGetSkipToVersion(pTask->exec.pWalReader);	62,091✔
229	if (skipToVer != 0 && skipToVer > pTask->chkInfo.nextProcessVer) {	62,080✔
230	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, skipToVer);	126✔
231	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	127!
232	return code;	×
233	}
234
235	tqDebug("vgId:%d s-task:%s wal reader jump to ver:%" PRId64, vgId, pTask->id.idStr, skipToVer);	127✔
236	}
237
238	return TSDB_CODE_SUCCESS;	62,073✔
239	}
240
241	// todo handle memory error
242	bool handleFillhistoryScanComplete(SStreamTask* pTask, int64_t ver) {	492,003✔
243	const char* id = pTask->id.idStr;	492,003✔
244	int64_t maxVer = pTask->step2Range.maxVer;	492,003✔
245
246	if ((pTask->info.fillHistory == 1) && ver > maxVer) {	492,003✔
247	if (!pTask->status.appendTranstateBlock) {	624!
248	qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the maximum ver:%" PRId64	624!
249	", not scan wal anymore, add transfer-state block into inputQ",
250	id, ver, maxVer);
251
252	double el = (taosGetTimestampMs() - pTask->execInfo.step2Start) / 1000.0;	624✔
253	qDebug("s-task:%s scan-history from WAL stage(step 2) ended, range:%" PRId64 "-%" PRId64 ", elapsed time:%.2fs",	624✔
254	id, pTask->step2Range.minVer, maxVer, el);
255	int32_t code = streamTaskPutTranstateIntoInputQ(pTask);	624✔
256	if (code) {	624!
257	qError("s-task:%s failed to put trans-state into inputQ", id);	×
258	}
259
260	return true;	624✔
261	} else {
262	qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the ver range:%" PRId64 "-%" PRId64	×
263	", not scan wal",
264	id, ver, pTask->step2Range.minVer, maxVer);
265	}
266	}
267
268	return false;	491,361✔
269	}
270
271	bool taskReadyForDataFromWal(SStreamTask* pTask) {	198,243✔
272	// non-source or fill-history tasks don't need to response the WAL scan action.
273	if ((pTask->info.taskLevel != TASK_LEVEL__SOURCE) \|\| (pTask->status.downstreamReady == 0)) {	198,243✔
274	return false;	104,610✔
275	}
276
277	// not in ready state, do not handle the data from wal
278	SStreamTaskState pState = streamTaskGetStatus(pTask);	93,633✔
279	if (pState.state != TASK_STATUS__READY) {	93,607✔
280	tqTrace("s-task:%s not ready for submit block in wal, status:%s", pTask->id.idStr, pState.name);	24,689!
281	return false;	24,690✔
282	}
283
284	// fill-history task has entered into the last phase, no need to anything
285	if ((pTask->info.fillHistory == 1) && pTask->status.appendTranstateBlock) {	68,918✔
286	// the maximum version of data in the WAL has reached already, the step2 is done
287	tqDebug("s-task:%s fill-history reach the maximum ver:%" PRId64 ", not scan wal anymore", pTask->id.idStr,	6,825✔
288	pTask->dataRange.range.maxVer);
289	return false;	6,825✔
290	}
291
292	// check if input queue is full or not
293	if (streamQueueIsFull(pTask->inputq.queue)) {	62,093✔
294	tqTrace("s-task:%s input queue is full, do nothing", pTask->id.idStr);	3!
295	return false;	×
296	}
297
298	// the input queue of downstream task is full, so the output is blocked, stopped for a while
299	if (pTask->inputq.status == TASK_INPUT_STATUS__BLOCKED) {	62,142!
300	tqDebug("s-task:%s inputQ is blocked, do nothing", pTask->id.idStr);	×
301	return false;	×
302	}
303
304	return true;	62,142✔
305	}
306
307	int32_t doPutDataIntoInputQ(SStreamTask* pTask, int64_t maxVer, int32_t* numOfItems, bool* pSucc) {	62,099✔
308	const char* id = pTask->id.idStr;	62,099✔
309	int32_t numOfNewItems = 0;	62,099✔
310	int32_t code = 0;	62,099✔
311	*pSucc = false;	62,099✔
312
313	while (1) {	429,951✔
314	if ((pTask->info.fillHistory == 1) && pTask->status.appendTranstateBlock) {	492,050!
315	*numOfItems += numOfNewItems;	×
316	return numOfNewItems > 0;	×
317	}
318
319	SStreamQueueItem* pItem = NULL;	492,050✔
320	code = extractMsgFromWal(pTask->exec.pWalReader, (void**)&pItem, maxVer, id);	492,050✔
321	if (code != TSDB_CODE_SUCCESS \|\| pItem == NULL) { // failed, continue	492,051✔
322	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	61,979✔
323	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, currentVer);	61,986✔
324	if (itemInFillhistory) {	61,969✔
325	numOfNewItems += 1;	550✔
326	}
327	break;	61,969✔
328	}
329
330	if (pItem != NULL) {	430,072!
331	code = streamTaskPutDataIntoInputQ(pTask, pItem);	430,076✔
332	if (code == TSDB_CODE_SUCCESS) {	430,081!
333	numOfNewItems += 1;	430,081✔
334	int64_t ver = walReaderGetCurrentVer(pTask->exec.pWalReader);	430,081✔
335	pTask->chkInfo.nextProcessVer = ver;	430,047✔
336	tqDebug("s-task:%s set ver:%" PRId64 " for reader after extract data from WAL", id, ver);	430,047✔
337
338	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, ver);	430,047✔
339	if (itemInFillhistory) {	430,029✔
340	break;	74✔
341	}
342	} else {
343	if (code == TSDB_CODE_OUT_OF_MEMORY) {	×
344	tqError("s-task:%s failed to put data into inputQ, since out of memory", id);	×
345	} else {
346	tqTrace("s-task:%s append input queue failed, code:inputQ is full, ver:%" PRId64, id,	×
347	pTask->chkInfo.nextProcessVer);
348	code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	×
349	if (code) {	×
350	tqError("s-task:%s failed to seek ver to:%" PRId64 " in wal", id, pTask->chkInfo.nextProcessVer);	×
351	}
352
353	code = 0; // reset the error code	×
354	}
355
356	break;	×
357	}
358	}
359	}
360
361	*numOfItems += numOfNewItems;	62,043✔
362	*pSucc = (numOfNewItems > 0);	62,043✔
363	return code;	62,043✔
364	}
365
366	int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta, bool* pScanIdle) {	50,798✔
367	*pScanIdle = true;	50,798✔
368	bool noDataInWal = true;	50,798✔
369	int32_t vgId = pStreamMeta->vgId;	50,798✔
370
371	int32_t numOfTasks = taosArrayGetSize(pStreamMeta->pTaskList);	50,798✔
372	if (numOfTasks == 0) {	50,798✔
373	return TSDB_CODE_SUCCESS;	633✔
374	}
375
376	// clone the task list, to avoid the task update during scan wal files
377	SArray* pTaskList = NULL;	50,165✔
378	streamMetaWLock(pStreamMeta);	50,165✔
379	pTaskList = taosArrayDup(pStreamMeta->pTaskList, NULL);	50,166✔
380	streamMetaWUnLock(pStreamMeta);	50,164✔
381	if (pTaskList == NULL) {	50,156!
382	tqError("vgId:%d failed to create task list dup, code:%s", vgId, tstrerror(terrno));	×
383	return terrno;	×
384	}
385
386	tqDebug("vgId:%d start to check wal to extract new submit block for %d tasks", vgId, numOfTasks);	50,156✔
387
388	// update the new task number
389	numOfTasks = taosArrayGetSize(pTaskList);	50,156✔
390
391	for (int32_t i = 0; i < numOfTasks; ++i) {	248,812✔
392	STaskId* pTaskId = taosArrayGet(pTaskList, i);	198,608✔
393	if (pTaskId == NULL) {	198,543!
394	continue;	136,526✔
395	}
396
397	SStreamTask* pTask = NULL;	198,543✔
398	int32_t code = streamMetaAcquireTask(pStreamMeta, pTaskId->streamId, pTaskId->taskId, &pTask);	198,543✔
399	if (pTask == NULL \|\| code != 0) {	198,550!
400	continue;	256✔
401	}
402
403	if (!taskReadyForDataFromWal(pTask)) {	198,294✔
404	streamMetaReleaseTask(pStreamMeta, pTask);	136,121✔
405	continue;	136,278✔
406	}
407
408	*pScanIdle = false;	62,094✔
409
410	// seek the stored version and extract data from WAL
411	code = setWalReaderStartOffset(pTask, vgId);	62,094✔
412	if (code != TSDB_CODE_SUCCESS) {	62,072!
413	streamMetaReleaseTask(pStreamMeta, pTask);	×
414	continue;	×
415	}
416
417	int32_t numOfItems = streamQueueGetNumOfItems(pTask->inputq.queue);	62,072✔
418	int64_t maxVer = (pTask->info.fillHistory == 1) ? pTask->step2Range.maxVer : INT64_MAX;	62,134✔
419
420	streamMutexLock(&pTask->lock);	62,134✔
421
422	SStreamTaskState pState = streamTaskGetStatus(pTask);	62,125✔
423	if (pState.state != TASK_STATUS__READY) {	62,117✔
424	tqDebug("s-task:%s not ready for submit block from wal, status:%s", pTask->id.idStr, pState.name);	15!
425	streamMutexUnlock(&pTask->lock);	15✔
426	streamMetaReleaseTask(pStreamMeta, pTask);	15✔
427	continue;	×
428	}
429
430	bool hasNewData = false;	62,102✔
431	code = doPutDataIntoInputQ(pTask, maxVer, &numOfItems, &hasNewData);	62,102✔
432	streamMutexUnlock(&pTask->lock);	62,044✔
433
434	if ((numOfItems > 0) \|\| hasNewData) {	62,128!
435	noDataInWal = false;	24,161✔
436	code = streamTrySchedExec(pTask);	24,161✔
437	if (code != TSDB_CODE_SUCCESS) {	24,158!
438	streamMetaReleaseTask(pStreamMeta, pTask);	×
439	taosArrayDestroy(pTaskList);	×
440	return code;	×
441	}
442	}
443
444	streamMetaReleaseTask(pStreamMeta, pTask);	62,125✔
445	}
446
447	// all wal are checked, and no new data available in wal.
448	if (noDataInWal) {	50,204✔
449	*pScanIdle = true;	34,784✔
450	}
451
452	taosArrayDestroy(pTaskList);	50,204✔
453	return TSDB_CODE_SUCCESS;	50,157✔
454	}

taosdata / TDengine / #3519

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