#3583

Committed 17 Jan 2025 07:28AM UTC coverage: 63.876% (+0.07%) from 63.803%

Build # #3583

Build Type

push

travis-ci

Committed by

web-flow

Commit Message

Merge pull request #29594 from taosdata/fix/insert-when-2-replicas

fix/insert-when-2-replicas

Run Details

141608 of 284535 branches covered (49.77%)

Branch coverage included in aggregate %.

0 of 2 new or added lines in 1 file covered. (0.0%)

396 existing lines in 105 files now uncovered.

220075 of 281695 relevant lines covered (78.13%)

19864675.58 hits per line

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

78.95

/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.
36	int32_t tqScanWal(STQ* pTq) {	69,055✔
37	SStreamMeta* pMeta = pTq->pStreamMeta;	69,055✔
38	int32_t vgId = pMeta->vgId;	69,055✔
39	int64_t st = taosGetTimestampMs();	69,064✔
40	int32_t numOfTasks = 0;	69,064✔
41
42	tqDebug("vgId:%d continue to check if data in wal are available, scanCounter:%d", vgId, pMeta->scanInfo.scanCounter);	69,064✔
43
44	// check all tasks
45	int32_t code = doScanWalForAllTasks(pMeta);	69,065✔
46	if (code) {	69,036!
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);	69,036✔
52	int32_t times = (--pMeta->scanInfo.scanCounter);	69,053✔
53	if (times < 0) {	69,053!
54	tqError("vgId:%d invalid scan counter:%d, reset to 0", vgId, times);	×
55	times = 0;	×
56	}
57
58	numOfTasks = taosArrayGetSize(pMeta->pTaskList);	69,053✔
59	streamMetaWUnLock(pMeta);	69,037✔
60
61	int64_t el = (taosGetTimestampMs() - st);	69,049✔
62	tqDebug("vgId:%d scan wal for stream tasks completed, elapsed time:%" PRId64 " ms", vgId, el);	69,049✔
63
64	if (times > 0) {	69,049✔
65	tqDebug("vgId:%d scan wal for stream tasks for %d times in %dms", vgId, times, SCAN_WAL_IDLE_DURATION);	23,099✔
66	code = tqScanWalInFuture(pTq, numOfTasks, SCAN_WAL_IDLE_DURATION);	23,099✔
67	if (code) {	23,102!
68	tqError("vgId:%d sched scan wal in %dms failed, ignore this failure", vgId, SCAN_WAL_IDLE_DURATION);	×
69	}
70	}
71
72	return code;	69,019✔
73	}
74
75	static void doStartScanWal(void* param, void* tmrId) {	23,102✔
76	int32_t vgId = 0;	23,102✔
77	STQ* pTq = NULL;	23,102✔
78	int32_t code = 0;	23,102✔
79
80	SBuildScanWalMsgParam* pParam = (SBuildScanWalMsgParam*)param;	23,102✔
81
82	SStreamMeta* pMeta = taosAcquireRef(streamMetaRefPool, pParam->metaId);	23,102✔
83	if (pMeta == NULL) {	23,102✔
84	tqError("metaRid:%" PRId64 " not valid now, stream meta has been freed", pParam->metaId);	181!
85	taosMemoryFree(pParam);	181!
86	return;	181✔
87	}
88
89	vgId = pMeta->vgId;	22,921✔
90	pTq = pMeta->ahandle;	22,921✔
91
92	tqDebug("vgId:%d create msg to start wal scan, numOfTasks:%d, vnd restored:%d", vgId, pParam->numOfTasks,	22,921✔
93	pTq->pVnode->restored);
94
95	code = streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);	22,921✔
96	if (code) {	22,921✔
97	tqError("vgId:%d failed sched task to scan wal, code:%s", vgId, tstrerror(code));	565!
98	}
99
100	code = taosReleaseRef(streamMetaRefPool, pParam->metaId);	22,921✔
101	if (code) {	22,921!
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);	22,921!
107	}
108
109	int32_t tqScanWalInFuture(STQ* pTq, int32_t numOfTasks, int32_t idleDuration) {	23,096✔
110	SStreamMeta* pMeta = pTq->pStreamMeta;	23,096✔
111	int32_t code = 0;	23,096✔
112	int32_t vgId = TD_VID(pTq->pVnode);	23,096✔
113	tmr_h pTimer = NULL;	23,096✔
114	SBuildScanWalMsgParam* pParam = NULL;	23,096✔
115
116	pParam = taosMemoryMalloc(sizeof(SBuildScanWalMsgParam));	23,096!
117	if (pParam == NULL) {	23,101!
118	return terrno;	×
119	}
120
121	pParam->metaId = pMeta->rid;	23,101✔
122	pParam->numOfTasks = numOfTasks;	23,101✔
123
124	code = streamTimerGetInstance(&pTimer);	23,101✔
125	if (code) {	23,096!
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");	23,096✔
130	}
131
132	return code;	23,102✔
133	}
134
135	int32_t tqScanWalAsync(STQ* pTq, bool ckPause) {	528,189✔
136	SStreamMeta* pMeta = pTq->pStreamMeta;	528,189✔
137	bool alreadyRestored = pTq->pVnode->restored;	528,189✔
138	int32_t code = 0;	528,189✔
139
140	// do not launch the stream tasks, if it is a follower or not restored vnode.
141	if (!(vnodeIsRoleLeader(pTq->pVnode) && alreadyRestored)) {	528,189✔
142	return TSDB_CODE_SUCCESS;	40,089✔
143	}
144
145	streamMetaWLock(pMeta);	488,160✔
146	code = doScanWalAsync(pTq, ckPause);	488,138✔
147	streamMetaWUnLock(pMeta);	488,146✔
148	return code;	488,170✔
149	}
150
151	int32_t tqStopStreamTasksAsync(STQ* pTq) {	4,734✔
152	SStreamMeta* pMeta = pTq->pStreamMeta;	4,734✔
153	int32_t vgId = pMeta->vgId;	4,734✔
154	return streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_STOP_ALL_TASKS);	4,734✔
155	}
156
157	int32_t setWalReaderStartOffset(SStreamTask* pTask, int32_t vgId) {	74,726✔
158	// seek the stored version and extract data from WAL
159	int64_t firstVer = walReaderGetValidFirstVer(pTask->exec.pWalReader);	74,726✔
160	if (pTask->chkInfo.nextProcessVer < firstVer) {	74,726✔
161	tqWarn("vgId:%d s-task:%s ver:%" PRId64 " earlier than the first ver of wal range %" PRId64 ", forward to %" PRId64,	5!
162	vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer, firstVer, firstVer);
163
164	pTask->chkInfo.nextProcessVer = firstVer;	5✔
165
166	// todo need retry if failed
167	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	5✔
168	if (code != TSDB_CODE_SUCCESS) {	5!
169	return code;	×
170	}
171
172	// append the data for the stream
173	tqDebug("vgId:%d s-task:%s wal reader seek to ver:%" PRId64, vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer);	5!
174	} else {
175	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	74,721✔
176	if (currentVer == -1) { // we only seek the read for the first time	74,689✔
177	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	4,735✔
178	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	4,734!
179	return code;	×
180	}
181
182	// append the data for the stream
183	tqDebug("vgId:%d s-task:%s wal reader initial seek to ver:%" PRId64, vgId, pTask->id.idStr,	4,734✔
184	pTask->chkInfo.nextProcessVer);
185	}
186	}
187
188	int64_t skipToVer = walReaderGetSkipToVersion(pTask->exec.pWalReader);	74,693✔
189	if (skipToVer != 0 && skipToVer > pTask->chkInfo.nextProcessVer) {	74,694✔
190	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, skipToVer);	445✔
191	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	446!
192	return code;	×
193	}
194
195	tqDebug("vgId:%d s-task:%s wal reader jump to ver:%" PRId64, vgId, pTask->id.idStr, skipToVer);	446✔
196	}
197
198	return TSDB_CODE_SUCCESS;	74,682✔
199	}
200
201	// todo handle memory error
202	bool handleFillhistoryScanComplete(SStreamTask* pTask, int64_t ver) {	403,292✔
203	const char* id = pTask->id.idStr;	403,292✔
204	int64_t maxVer = pTask->step2Range.maxVer;	403,292✔
205
206	if ((pTask->info.fillHistory == 1) && ver > maxVer) {	403,292✔
207	if (!pTask->status.appendTranstateBlock) {	759!
208	qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the maximum ver:%" PRId64	759!
209	", not scan wal anymore, add transfer-state block into inputQ",
210	id, ver, maxVer);
211
212	double el = (taosGetTimestampMs() - pTask->execInfo.step2Start) / 1000.0;	759✔
213	qDebug("s-task:%s scan-history from WAL stage(step 2) ended, range:%" PRId64 "-%" PRId64 ", elapsed time:%.2fs",	759✔
214	id, pTask->step2Range.minVer, maxVer, el);
215	int32_t code = streamTaskPutTranstateIntoInputQ(pTask);	759✔
216	if (code) {	759!
217	qError("s-task:%s failed to put trans-state into inputQ", id);	×
218	}
219
220	return true;	759✔
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
228	return false;	402,542✔
229	}
230
231	bool taskReadyForDataFromWal(SStreamTask* pTask) {	263,998✔
232	// non-source or fill-history tasks don't need to response the WAL scan action.
233	SSTaskBasicInfo* pInfo = &pTask->info;	263,998✔
234	if ((pInfo->taskLevel != TASK_LEVEL__SOURCE) \|\| (pTask->status.downstreamReady == 0)) {	263,998✔
235	return false;	136,686✔
236	}
237
238	if (pInfo->taskLevel == TASK_LEVEL__SOURCE && pInfo->trigger == STREAM_TRIGGER_FORCE_WINDOW_CLOSE) {	127,312✔
239	return false;	12,581✔
240	}
241
242	// not in ready state, do not handle the data from wal
243	SStreamTaskState pState = streamTaskGetStatus(pTask);	114,731✔
244	if (pState.state != TASK_STATUS__READY) {	114,712✔
245	tqTrace("s-task:%s not ready for submit block in wal, status:%s", pTask->id.idStr, pState.name);	34,579✔
246	return false;	34,578✔
247	}
248
249	// fill-history task has entered into the last phase, no need to anything
250	if ((pInfo->fillHistory == 1) && pTask->status.appendTranstateBlock) {	80,133✔
251	// the maximum version of data in the WAL has reached already, the step2 is done
252	tqDebug("s-task:%s fill-history reach the maximum ver:%" PRId64 ", not scan wal anymore", pTask->id.idStr,	5,413✔
253	pTask->dataRange.range.maxVer);
254	return false;	5,413✔
255	}
256
257	// check if input queue is full or not
258	if (streamQueueIsFull(pTask->inputq.queue)) {	74,720✔
259	tqTrace("s-task:%s input queue is full, do nothing", pTask->id.idStr);	19!
UNCOV 260	return false;	×
261	}
262
263	// the input queue of downstream task is full, so the output is blocked, stopped for a while
264	if (pTask->inputq.status == TASK_INPUT_STATUS__BLOCKED) {	74,727!
265	tqDebug("s-task:%s inputQ is blocked, do nothing", pTask->id.idStr);	×
266	return false;	×
267	}
268
269	return true;	74,727✔
270	}
271
272	int32_t doPutDataIntoInputQ(SStreamTask* pTask, int64_t maxVer, int32_t* numOfItems, bool* pSucc) {	74,699✔
273	const char* id = pTask->id.idStr;	74,699✔
274	int32_t numOfNewItems = 0;	74,699✔
275	int32_t code = 0;	74,699✔
276	*pSucc = false;	74,699✔
277
278	while (1) {	328,649✔
279	if ((pTask->info.fillHistory == 1) && pTask->status.appendTranstateBlock) {	403,348!
280	*numOfItems += numOfNewItems;	×
281	return numOfNewItems > 0;	×
282	}
283
284	SStreamQueueItem* pItem = NULL;	403,348✔
285	code = extractMsgFromWal(pTask->exec.pWalReader, (void**)&pItem, maxVer, id);	403,348✔
286	if (code != TSDB_CODE_SUCCESS \|\| pItem == NULL) { // failed, continue	403,294✔
287	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	74,504✔
288	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, currentVer);	74,513✔
289	if (itemInFillhistory) {	74,503✔
290	numOfNewItems += 1;	620✔
291	}
292	break;	74,503✔
293	}
294
295	if (pItem != NULL) {	328,790!
296	code = streamTaskPutDataIntoInputQ(pTask, pItem);	328,791✔
297	if (code == TSDB_CODE_SUCCESS) {	328,795✔
298	numOfNewItems += 1;	328,794✔
299	int64_t ver = walReaderGetCurrentVer(pTask->exec.pWalReader);	328,794✔
300	pTask->chkInfo.nextProcessVer = ver;	328,793✔
301	tqDebug("s-task:%s set ver:%" PRId64 " for reader after extract data from WAL", id, ver);	328,793✔
302
303	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, ver);	328,793✔
304	if (itemInFillhistory) {	328,789✔
305	break;	139✔
306	}
307	} else {
308	if (code == TSDB_CODE_OUT_OF_MEMORY) {	1!
309	tqError("s-task:%s failed to put data into inputQ, since out of memory", id);	×
310	} else {
311	tqTrace("s-task:%s append input queue failed, code:inputQ is full, ver:%" PRId64, id,	1!
312	pTask->chkInfo.nextProcessVer);
313	code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	1✔
314	if (code) {	1!
315	tqError("s-task:%s failed to seek ver to:%" PRId64 " in wal", id, pTask->chkInfo.nextProcessVer);	×
316	}
317
318	code = 0; // reset the error code	1✔
319	}
320
321	break;	1✔
322	}
323	}
324	}
325
326	*numOfItems += numOfNewItems;	74,643✔
327	*pSucc = (numOfNewItems > 0);	74,643✔
328	return code;	74,643✔
329	}
330
331	int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta) {	69,049✔
332	int32_t vgId = pStreamMeta->vgId;	69,049✔
333	int32_t numOfTasks = taosArrayGetSize(pStreamMeta->pTaskList);	69,049✔
334	if (numOfTasks == 0) {	69,057✔
335	return TSDB_CODE_SUCCESS;	569✔
336	}
337
338	// clone the task list, to avoid the task update during scan wal files
339	SArray* pTaskList = NULL;	68,488✔
340	streamMetaWLock(pStreamMeta);	68,488✔
341	pTaskList = taosArrayDup(pStreamMeta->pTaskList, NULL);	68,497✔
342	streamMetaWUnLock(pStreamMeta);	68,492✔
343	if (pTaskList == NULL) {	68,498!
344	tqError("vgId:%d failed to create task list dup, code:%s", vgId, tstrerror(terrno));	×
345	return terrno;	×
346	}
347
348	tqDebug("vgId:%d start to check wal to extract new submit block for %d tasks", vgId, numOfTasks);	68,498✔
349
350	// update the new task number
351	numOfTasks = taosArrayGetSize(pTaskList);	68,498✔
352
353	for (int32_t i = 0; i < numOfTasks; ++i) {	333,124✔
354	STaskId* pTaskId = taosArrayGet(pTaskList, i);	264,587✔
355	if (pTaskId == NULL) {	264,561!
356	continue;	189,906✔
357	}
358
359	SStreamTask* pTask = NULL;	264,561✔
360	int32_t code = streamMetaAcquireTask(pStreamMeta, pTaskId->streamId, pTaskId->taskId, &pTask);	264,561✔
361	if (pTask == NULL \|\| code != 0) {	264,580!
362	continue;	540✔
363	}
364
365	if (!taskReadyForDataFromWal(pTask)) {	264,040✔
366	streamMetaReleaseTask(pStreamMeta, pTask);	189,249✔
367	continue;	189,366✔
368	}
369
370	// seek the stored version and extract data from WAL
371	code = setWalReaderStartOffset(pTask, vgId);	74,704✔
372	if (code != TSDB_CODE_SUCCESS) {	74,677!
373	streamMetaReleaseTask(pStreamMeta, pTask);	×
374	continue;	×
375	}
376
377	int32_t numOfItems = streamQueueGetNumOfItems(pTask->inputq.queue);	74,677✔
378	int64_t maxVer = (pTask->info.fillHistory == 1) ? pTask->step2Range.maxVer : INT64_MAX;	74,737✔
379
380	streamMutexLock(&pTask->lock);	74,737✔
381
382	SStreamTaskState state = streamTaskGetStatus(pTask);	74,735✔
383	if (state.state != TASK_STATUS__READY) {	74,727✔
384	tqDebug("s-task:%s not ready for submit block from wal, status:%s", pTask->id.idStr, state.name);	18✔
385	streamMutexUnlock(&pTask->lock);	18✔
386	streamMetaReleaseTask(pStreamMeta, pTask);	18✔
UNCOV 387	continue;	×
388	}
389
390	bool hasNewData = false;	74,709✔
391	code = doPutDataIntoInputQ(pTask, maxVer, &numOfItems, &hasNewData);	74,709✔
392	streamMutexUnlock(&pTask->lock);	74,622✔
393
394	if ((numOfItems > 0) \|\| hasNewData) {	74,717!
395	code = streamTrySchedExec(pTask);	25,613✔
396	if (code != TSDB_CODE_SUCCESS) {	25,616!
397	streamMetaReleaseTask(pStreamMeta, pTask);	×
398	taosArrayDestroy(pTaskList);	×
399	return code;	×
400	}
401	}
402
403	streamMetaReleaseTask(pStreamMeta, pTask);	74,720✔
404	}
405
406	taosArrayDestroy(pTaskList);	68,537✔
407	return TSDB_CODE_SUCCESS;	68,478✔
408	}
409
410	int32_t doScanWalAsync(STQ* pTq, bool ckPause) {	488,090✔
411	SStreamMeta* pMeta = pTq->pStreamMeta;	488,090✔
412	bool alreadyRestored = pTq->pVnode->restored;	488,090✔
413	int32_t vgId = pMeta->vgId;	488,090✔
414	int32_t numOfTasks = taosArrayGetSize(pMeta->pTaskList);	488,090✔
415
416	if (numOfTasks == 0) {	488,113✔
417	tqDebug("vgId:%d no stream tasks existed to run", vgId);	7!
418	return 0;	14✔
419	}
420
421	if (pMeta->startInfo.startAllTasks) {	488,106✔
422	tqTrace("vgId:%d in restart procedure, not scan wal", vgId);	26,567✔
423	return 0;	26,567✔
424	}
425
426	pMeta->scanInfo.scanCounter += 1;	461,539✔
427	if (pMeta->scanInfo.scanCounter > MAX_REPEAT_SCAN_THRESHOLD) {	461,539✔
428	pMeta->scanInfo.scanCounter = MAX_REPEAT_SCAN_THRESHOLD;	390,118✔
429	}
430
431	if (pMeta->scanInfo.scanCounter > 1) {	461,539✔
432	tqDebug("vgId:%d wal read task has been launched, remain scan times:%d", vgId, pMeta->scanInfo.scanCounter);	414,197✔
433	return 0;	414,205✔
434	}
435
436	int32_t numOfPauseTasks = pMeta->numOfPausedTasks;	47,342✔
437	if (ckPause && numOfTasks == numOfPauseTasks) {	47,342✔
438	tqDebug("vgId:%d ignore all submit, all streams had been paused, reset the walScanCounter", vgId);	138✔
439
440	// reset the counter value, since we do not launch the scan wal operation.
441	pMeta->scanInfo.scanCounter = 0;	138✔
442	return 0;	138✔
443	}
444
445	tqDebug("vgId:%d create msg to start wal scan to launch stream tasks, numOfTasks:%d, vnd restored:%d", vgId,	47,204✔
446	numOfTasks, alreadyRestored);
447
448	return streamTaskSchedTask(&pTq->pVnode->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA);	47,205✔
449	}

taosdata / TDengine / #3583

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