#3824

Committed 01 Apr 2025 12:26PM UTC coverage: 34.064% (-0.001%) from 34.065%

Build # #3824

Build Type

push

travis-ci

Committed by happyguoxy

Commit Message

test:alter gcda dir

Run Details

148483 of 599532 branches covered (24.77%)

Branch coverage included in aggregate %.

222466 of 489445 relevant lines covered (45.45%)

762427.9 hits per line

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

51.47

/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 SCAN_WAL_IDLE_DURATION    250  // idle for 500ms to do next wal scan
#define SCAN_WAL_WAIT_COUNT       2

typedef struct SBuildScanWalMsgParam {
  int64_t metaId;
  SMsgCb  msgCb;
} SBuildScanWalMsgParam;

static int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta, int32_t* pNumOfTasks);
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);

// 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;
  int64_t      el = 0;
  int32_t      code = 0;

  int32_t old = atomic_val_compare_exchange_32(&pMeta->scanInfo.scanSentinel, 0, 1);
  if (old == 0) {
    tqDebug("vgId:%d try to scan wal to extract data", vgId);
  } else {
    tqDebug("vgId:%d already in wal scan, abort", vgId);
    return code;
  }

  // the scan wal interval less than 200, not scan, actually.
  if ((pMeta->scanInfo.lastScanTs > 0) && (st - pMeta->scanInfo.lastScanTs < 200)) {
    tqDebug("vgId:%d scan wal less than 200ms, do nothing", vgId);
    atomic_store_32(&pMeta->scanInfo.scanSentinel, 0);
    return code;
  }

  // check all tasks
  code = doScanWalForAllTasks(pMeta, &numOfTasks);

  pMeta->scanInfo.lastScanTs = taosGetTimestampMs();
  el = (pMeta->scanInfo.lastScanTs - st);

  if (code) {
    tqError("vgId:%d failed to scan wal for all tasks, try next time, elapsed time:%" PRId64 "ms code:%s", vgId, el,
            tstrerror(code));
  } else {
    tqDebug("vgId:%d scan wal for stream tasks completed, elapsed time:%" PRId64 "ms", vgId, el);
  }

  atomic_store_32(&pMeta->scanInfo.scanSentinel, 0);
  return code;
}

static bool waitEnoughDuration(SStreamMeta* pMeta) {
  if ((++pMeta->scanInfo.tickCounter) >= SCAN_WAL_WAIT_COUNT) {
    pMeta->scanInfo.tickCounter = 0;
    return true;
  }

  return false;
}

static void doStartScanWal(void* param, void* tmrId) {
  int32_t                vgId = 0;
  int32_t                code = 0;
  int32_t                numOfTasks = 0;
  tmr_h                  pTimer = NULL;
  SBuildScanWalMsgParam* pParam = (SBuildScanWalMsgParam*)param;

  tqTrace("start to do scan wal in tmr, metaRid:%" PRId64, pParam->metaId);

  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;
  code = streamTimerGetInstance(&pTimer);
  if (code) {
    tqFatal("vgId:%d failed to get tmr ctrl during sched scan wal, not scan wal, code:%s", vgId, tstrerror(code));
    taosMemoryFree(pParam);
    return;
  }

  if (pMeta->closeFlag) {
    code = taosReleaseRef(streamMetaRefPool, pParam->metaId);
    if (code == TSDB_CODE_SUCCESS) {
      tqInfo("vgId:%d jump out of scan wal timer since closed", vgId);
    } else {
      tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,
              tstrerror(code));
    }

    taosMemoryFree(pParam);
    return;
  }

  if (pMeta->role != NODE_ROLE_LEADER) {
    tqDebug("vgId:%d not leader, role:%d not scan wal anymore", vgId, pMeta->role);

    code = taosReleaseRef(streamMetaRefPool, pParam->metaId);
    if (code == TSDB_CODE_SUCCESS) {
      tqDebug("vgId:%d jump out of scan wal timer since not leader", vgId);
    } else {
      tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,
              tstrerror(code));
    }

    taosMemFree(pParam);
    return;
  }

  if (pMeta->startInfo.startAllTasks) {
    tqDebug("vgId:%d in restart procedure, not ready to scan wal", vgId);
    goto _end;
  }

  if (!waitEnoughDuration(pMeta)) {
    streamTmrStart(doStartScanWal, SCAN_WAL_IDLE_DURATION, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId,
                   "scan-wal");
    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));
    }
    return;
  }

  // failed to lock, try 500ms later
  code = streamMetaTryRlock(pMeta);
  if (code == 0) {
    numOfTasks = taosArrayGetSize(pMeta->pTaskList);
    streamMetaRUnLock(pMeta);
  } else {
    numOfTasks = 0;
  }

  if (numOfTasks > 0) {
    tqDebug("vgId:%d create msg to start wal scan, numOfTasks:%d", vgId, numOfTasks);

#if 0
  //  wait for the vnode is freed, and invalid read may occur.
  taosMsleep(10000);
#endif

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

_end:
  streamTmrStart(doStartScanWal, SCAN_WAL_IDLE_DURATION, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId, "scan-wal");
  tqTrace("vgId:%d try scan-wal will start in %dms", vgId, SCAN_WAL_IDLE_DURATION*SCAN_WAL_WAIT_COUNT);

  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));
  }
}

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

  // 1. the vnode should be the leader.
  // 2. the stream isn't disabled
  if ((pMeta->role != NODE_ROLE_LEADER) || tsDisableStream) {
    tqInfo("vgId:%d follower node or stream disabled, not scan wal", vgId);
    return;
  }

  pParam = taosMemoryMalloc(sizeof(SBuildScanWalMsgParam));
  if (pParam == NULL) {
    tqError("vgId:%d failed to start scan wal, stream not executes, code:%s", vgId, tstrerror(code));
    return;
  }

  pParam->metaId = pMeta->rid;
  pParam->msgCb = pTq->pVnode->msgCb;

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

int32_t tqStopStreamAllTasksAsync(SStreamMeta* pMeta, SMsgCb* pMsgCb) {
  return streamTaskSchedTask(pMsgCb, pMeta->vgId, 0, 0, STREAM_EXEC_T_STOP_ALL_TASKS, false);
}

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->trigger == STREAM_TRIGGER_FORCE_WINDOW_CLOSE) {
    return false;
  }

  if (pInfo->fillHistory == STREAM_RECALCUL_TASK) {
    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 do anything
  if ((pInfo->fillHistory == STREAM_HISTORY_TASK) && 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 whether input queue is full or not
  if (streamQueueIsFull(pTask->inputq.queue)) {
    tqTrace("s-task:%s input queue is full, launch task without scanning wal", pTask->id.idStr);
    int32_t code = streamTrySchedExec(pTask, false);
    if (code) {
      tqError("s-task:%s failed to start task while inputQ is full", 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* pNumOfTasks) {
  int32_t vgId = pStreamMeta->vgId;
  SArray* pTaskList = NULL;
  int32_t numOfTasks = 0;

  // clone the task list, to avoid the task update during scan wal files
  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;
  }

  // update the new task number
  numOfTasks = taosArrayGetSize(pTaskList);
  if (pNumOfTasks != NULL) {
    *pNumOfTasks = numOfTasks;
  }

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

  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, false);
      if (code != TSDB_CODE_SUCCESS) {
        streamMetaReleaseTask(pStreamMeta, pTask);
        taosArrayDestroy(pTaskList);
        return code;
      }
    }

    streamMetaReleaseTask(pStreamMeta, pTask);
  }

  taosArrayDestroy(pTaskList);
  return TSDB_CODE_SUCCESS;
}

void streamMetaFreeTQDuringScanWalError(STQ* pTq) {
  SBuildScanWalMsgParam* p = taosMemoryCalloc(1, sizeof(SBuildScanWalMsgParam));
  p->metaId = pTq->pStreamMeta->rid;

  doStartScanWal(p, 0);
}

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 SCAN_WAL_IDLE_DURATION 250 // idle for 500ms to do next wal scan
20	#define SCAN_WAL_WAIT_COUNT 2
21
22	typedef struct SBuildScanWalMsgParam {
23	int64_t metaId;
24	SMsgCb msgCb;
25	} SBuildScanWalMsgParam;
26
27	static int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta, int32_t* pNumOfTasks);
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
33	// extract data blocks(submit/delete) from WAL, and add them into the input queue for all the sources tasks.
34	int32_t tqScanWal(STQ* pTq) {	600✔
35	SStreamMeta* pMeta = pTq->pStreamMeta;	600✔
36	int32_t vgId = pMeta->vgId;	600✔
37	int64_t st = taosGetTimestampMs();	600✔
38	int32_t numOfTasks = 0;	600✔
39	int64_t el = 0;	600✔
40	int32_t code = 0;	600✔
41
42	int32_t old = atomic_val_compare_exchange_32(&pMeta->scanInfo.scanSentinel, 0, 1);	600✔
43	if (old == 0) {	600!
44	tqDebug("vgId:%d try to scan wal to extract data", vgId);	600!
45	} else {
46	tqDebug("vgId:%d already in wal scan, abort", vgId);	×
47	return code;	×
48	}
49
50	// the scan wal interval less than 200, not scan, actually.
51	if ((pMeta->scanInfo.lastScanTs > 0) && (st - pMeta->scanInfo.lastScanTs < 200)) {	600!
52	tqDebug("vgId:%d scan wal less than 200ms, do nothing", vgId);	×
53	atomic_store_32(&pMeta->scanInfo.scanSentinel, 0);	×
54	return code;	×
55	}
56
57	// check all tasks
58	code = doScanWalForAllTasks(pMeta, &numOfTasks);	600✔
59
60	pMeta->scanInfo.lastScanTs = taosGetTimestampMs();	600✔
61	el = (pMeta->scanInfo.lastScanTs - st);	600✔
62
63	if (code) {	600!
64	tqError("vgId:%d failed to scan wal for all tasks, try next time, elapsed time:%" PRId64 "ms code:%s", vgId, el,	×
65	tstrerror(code));
66	} else {
67	tqDebug("vgId:%d scan wal for stream tasks completed, elapsed time:%" PRId64 "ms", vgId, el);	600!
68	}
69
70	atomic_store_32(&pMeta->scanInfo.scanSentinel, 0);	600✔
71	return code;	600✔
72	}
73
74	static bool waitEnoughDuration(SStreamMeta* pMeta) {	23,260✔
75	if ((++pMeta->scanInfo.tickCounter) >= SCAN_WAL_WAIT_COUNT) {	23,260✔
76	pMeta->scanInfo.tickCounter = 0;	11,576✔
77	return true;	11,576✔
78	}
79
80	return false;	11,684✔
81	}
82
83	static void doStartScanWal(void* param, void* tmrId) {	23,572✔
84	int32_t vgId = 0;	23,572✔
85	int32_t code = 0;	23,572✔
86	int32_t numOfTasks = 0;	23,572✔
87	tmr_h pTimer = NULL;	23,572✔
88	SBuildScanWalMsgParam* pParam = (SBuildScanWalMsgParam*)param;	23,572✔
89
90	tqTrace("start to do scan wal in tmr, metaRid:%" PRId64, pParam->metaId);	23,572✔
91
92	SStreamMeta* pMeta = taosAcquireRef(streamMetaRefPool, pParam->metaId);	23,572✔
93	if (pMeta == NULL) {	23,572!
94	tqError("metaRid:%" PRId64 " not valid now, stream meta has been freed", pParam->metaId);	×
95	taosMemoryFree(pParam);	×
96	return;	11,913✔
97	}
98
99	vgId = pMeta->vgId;	23,572✔
100	code = streamTimerGetInstance(&pTimer);	23,572✔
101	if (code) {	23,572!
102	tqFatal("vgId:%d failed to get tmr ctrl during sched scan wal, not scan wal, code:%s", vgId, tstrerror(code));	×
103	taosMemoryFree(pParam);	×
104	return;	×
105	}
106
107	if (pMeta->closeFlag) {	23,572✔
108	code = taosReleaseRef(streamMetaRefPool, pParam->metaId);	229✔
109	if (code == TSDB_CODE_SUCCESS) {	229!
110	tqInfo("vgId:%d jump out of scan wal timer since closed", vgId);	229!
111	} else {
112	tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,	×
113	tstrerror(code));
114	}
115
116	taosMemoryFree(pParam);	229!
117	return;	229✔
118	}
119
120	if (pMeta->role != NODE_ROLE_LEADER) {	23,343!
121	tqDebug("vgId:%d not leader, role:%d not scan wal anymore", vgId, pMeta->role);	×
122
123	code = taosReleaseRef(streamMetaRefPool, pParam->metaId);	×
124	if (code == TSDB_CODE_SUCCESS) {	×
125	tqDebug("vgId:%d jump out of scan wal timer since not leader", vgId);	×
126	} else {
127	tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,	×
128	tstrerror(code));
129	}
130
131	taosMemFree(pParam);	×
132	return;	×
133	}
134
135	if (pMeta->startInfo.startAllTasks) {	23,343✔
136	tqDebug("vgId:%d in restart procedure, not ready to scan wal", vgId);	83✔
137	goto _end;	83✔
138	}
139
140	if (!waitEnoughDuration(pMeta)) {	23,260✔
141	streamTmrStart(doStartScanWal, SCAN_WAL_IDLE_DURATION, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId,	11,684✔
142	"scan-wal");
143	code = taosReleaseRef(streamMetaRefPool, pParam->metaId);	11,684✔
144	if (code) {	11,684!
145	tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,	×
146	tstrerror(code));
147	}
148	return;	11,684✔
149	}
150
151	// failed to lock, try 500ms later
152	code = streamMetaTryRlock(pMeta);	11,576✔
153	if (code == 0) {	11,576✔
154	numOfTasks = taosArrayGetSize(pMeta->pTaskList);	11,575✔
155	streamMetaRUnLock(pMeta);	11,575✔
156	} else {
157	numOfTasks = 0;	1✔
158	}
159
160	if (numOfTasks > 0) {	11,576✔
161	tqDebug("vgId:%d create msg to start wal scan, numOfTasks:%d", vgId, numOfTasks);	600!
162
163	#if 0
164	// wait for the vnode is freed, and invalid read may occur.
165	taosMsleep(10000);
166	#endif
167
168	code = streamTaskSchedTask(&pParam->msgCb, vgId, 0, 0, STREAM_EXEC_T_EXTRACT_WAL_DATA, false);	600✔
169	if (code) {	600!
170	tqError("vgId:%d failed sched task to scan wal, code:%s", vgId, tstrerror(code));	×
171	}
172	}
173
174	_end:	11,576✔
175	streamTmrStart(doStartScanWal, SCAN_WAL_IDLE_DURATION, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId, "scan-wal");	11,659✔
176	tqTrace("vgId:%d try scan-wal will start in %dms", vgId, SCAN_WAL_IDLE_DURATION*SCAN_WAL_WAIT_COUNT);	11,659✔
177
178	code = taosReleaseRef(streamMetaRefPool, pParam->metaId);	11,659✔
179	if (code) {	11,659!
180	tqError("vgId:%d failed to release ref for streamMeta, rid:%" PRId64 " code:%s", vgId, pParam->metaId,	×
181	tstrerror(code));
182	}
183	}
184
185	void tqScanWalAsync(STQ* pTq) {	229✔
186	SStreamMeta* pMeta = pTq->pStreamMeta;	229✔
187	int32_t code = 0;	229✔
188	int32_t vgId = TD_VID(pTq->pVnode);	229✔
189	tmr_h pTimer = NULL;	229✔
190	SBuildScanWalMsgParam* pParam = NULL;	229✔
191
192	// 1. the vnode should be the leader.
193	// 2. the stream isn't disabled
194	if ((pMeta->role != NODE_ROLE_LEADER) \|\| tsDisableStream) {	229!
195	tqInfo("vgId:%d follower node or stream disabled, not scan wal", vgId);	×
196	return;	×
197	}
198
199	pParam = taosMemoryMalloc(sizeof(SBuildScanWalMsgParam));	229!
200	if (pParam == NULL) {	229!
201	tqError("vgId:%d failed to start scan wal, stream not executes, code:%s", vgId, tstrerror(code));	×
202	return;	×
203	}
204
205	pParam->metaId = pMeta->rid;	229✔
206	pParam->msgCb = pTq->pVnode->msgCb;	229✔
207
208	code = streamTimerGetInstance(&pTimer);	229✔
209	if (code) {	229!
210	tqFatal("vgId:%d failed to get tmr ctrl during sched scan wal", vgId);	×
211	taosMemoryFree(pParam);	×
212	} else {
213	streamTmrStart(doStartScanWal, SCAN_WAL_IDLE_DURATION, pParam, pTimer, &pMeta->scanInfo.scanTimer, vgId,	229✔
214	"scan-wal");
215	}
216	}
217
218	int32_t tqStopStreamAllTasksAsync(SStreamMeta* pMeta, SMsgCb* pMsgCb) {	118✔
219	return streamTaskSchedTask(pMsgCb, pMeta->vgId, 0, 0, STREAM_EXEC_T_STOP_ALL_TASKS, false);	118✔
220	}
221
222	int32_t setWalReaderStartOffset(SStreamTask* pTask, int32_t vgId) {	630✔
223	// seek the stored version and extract data from WAL
224	int64_t firstVer = walReaderGetValidFirstVer(pTask->exec.pWalReader);	630✔
225	if (pTask->chkInfo.nextProcessVer < firstVer) {	630!
226	tqWarn("vgId:%d s-task:%s ver:%" PRId64 " earlier than the first ver of wal range %" PRId64 ", forward to %" PRId64,	×
227	vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer, firstVer, firstVer);
228
229	pTask->chkInfo.nextProcessVer = firstVer;	×
230
231	// todo need retry if failed
232	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	×
233	if (code != TSDB_CODE_SUCCESS) {	×
234	return code;	×
235	}
236
237	// append the data for the stream
238	tqDebug("vgId:%d s-task:%s wal reader seek to ver:%" PRId64, vgId, pTask->id.idStr, pTask->chkInfo.nextProcessVer);	×
239	} else {
240	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	630✔
241	if (currentVer == -1) { // we only seek the read for the first time	630✔
242	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	11✔
243	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	11!
244	return code;	×
245	}
246
247	// append the data for the stream
248	tqDebug("vgId:%d s-task:%s wal reader initial seek to ver:%" PRId64, vgId, pTask->id.idStr,	11!
249	pTask->chkInfo.nextProcessVer);
250	}
251	}
252
253	int64_t skipToVer = walReaderGetSkipToVersion(pTask->exec.pWalReader);	630✔
254	if (skipToVer != 0 && skipToVer > pTask->chkInfo.nextProcessVer) {	630!
255	int32_t code = walReaderSeekVer(pTask->exec.pWalReader, skipToVer);	×
256	if (code != TSDB_CODE_SUCCESS) { // no data in wal, quit	×
257	return code;	×
258	}
259
260	tqDebug("vgId:%d s-task:%s wal reader jump to ver:%" PRId64, vgId, pTask->id.idStr, skipToVer);	×
261	}
262
263	return TSDB_CODE_SUCCESS;	630✔
264	}
265
266	// todo handle memory error
267	bool handleFillhistoryScanComplete(SStreamTask* pTask, int64_t ver) {	1,835✔
268	const char* id = pTask->id.idStr;	1,835✔
269	int64_t maxVer = pTask->step2Range.maxVer;	1,835✔
270
271	if ((pTask->info.fillHistory == 1) && ver > maxVer) {	1,835!
272	if (!pTask->status.appendTranstateBlock) {	×
273	qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the maximum ver:%" PRId64	×
274	", not scan wal anymore, add transfer-state block into inputQ",
275	id, ver, maxVer);
276
277	double el = (taosGetTimestampMs() - pTask->execInfo.step2Start) / 1000.0;	×
278	qDebug("s-task:%s scan-history from WAL stage(step 2) ended, range:%" PRId64 "-%" PRId64 ", elapsed time:%.2fs",	×
279	id, pTask->step2Range.minVer, maxVer, el);
280	int32_t code = streamTaskPutTranstateIntoInputQ(pTask);	×
281	if (code) {	×
282	qError("s-task:%s failed to put trans-state into inputQ", id);	×
283	}
284
285	return true;	×
286	} else {
287	qWarn("s-task:%s fill-history scan WAL, nextProcessVer:%" PRId64 " out of the ver range:%" PRId64 "-%" PRId64	×
288	", not scan wal",
289	id, ver, pTask->step2Range.minVer, maxVer);
290	}
291	}
292
293	return false;	1,835✔
294	}
295
296	bool taskReadyForDataFromWal(SStreamTask* pTask) {	1,315✔
297	// non-source or fill-history tasks don't need to response the WAL scan action.
298	SSTaskBasicInfo* pInfo = &pTask->info;	1,315✔
299	if ((pInfo->taskLevel != TASK_LEVEL__SOURCE) \|\| (pTask->status.downstreamReady == 0)) {	1,315✔
300	return false;	605✔
301	}
302
303	if (pInfo->trigger == STREAM_TRIGGER_FORCE_WINDOW_CLOSE) {	710!
304	return false;	×
305	}
306
307	if (pInfo->fillHistory == STREAM_RECALCUL_TASK) {	710!
308	return false;	×
309	}
310
311	// not in ready state, do not handle the data from wal
312	SStreamTaskState pState = streamTaskGetStatus(pTask);	710✔
313	if (pState.state != TASK_STATUS__READY) {	710✔
314	tqTrace("s-task:%s not ready for submit block in wal, status:%s", pTask->id.idStr, pState.name);	80!
315	return false;	80✔
316	}
317
318	// fill-history task has entered into the last phase, no need to do anything
319	if ((pInfo->fillHistory == STREAM_HISTORY_TASK) && pTask->status.appendTranstateBlock) {	630!
320	// the maximum version of data in the WAL has reached already, the step2 is done
321	tqDebug("s-task:%s fill-history reach the maximum ver:%" PRId64 ", not scan wal anymore", pTask->id.idStr,	×
322	pTask->dataRange.range.maxVer);
323	return false;	×
324	}
325
326	// check whether input queue is full or not
327	if (streamQueueIsFull(pTask->inputq.queue)) {	630!
328	tqTrace("s-task:%s input queue is full, launch task without scanning wal", pTask->id.idStr);	×
329	int32_t code = streamTrySchedExec(pTask, false);	×
330	if (code) {	×
331	tqError("s-task:%s failed to start task while inputQ is full", pTask->id.idStr);	×
332	}
333	return false;	×
334	}
335
336	// the input queue of downstream task is full, so the output is blocked, stopped for a while
337	if (pTask->inputq.status == TASK_INPUT_STATUS__BLOCKED) {	630!
338	tqDebug("s-task:%s inputQ is blocked, do nothing", pTask->id.idStr);	×
339	return false;	×
340	}
341
342	return true;	630✔
343	}
344
345	int32_t doPutDataIntoInputQ(SStreamTask* pTask, int64_t maxVer, int32_t* numOfItems, bool* pSucc) {	630✔
346	const char* id = pTask->id.idStr;	630✔
347	int32_t numOfNewItems = 0;	630✔
348	int32_t code = 0;	630✔
349	*pSucc = false;	630✔
350
351	while (1) {	1,205✔
352	if ((pTask->info.fillHistory == 1) && pTask->status.appendTranstateBlock) {	1,835!
353	*numOfItems += numOfNewItems;	×
354	return numOfNewItems > 0;	×
355	}
356
357	SStreamQueueItem* pItem = NULL;	1,835✔
358	code = extractMsgFromWal(pTask->exec.pWalReader, (void**)&pItem, maxVer, id);	1,835✔
359	if (code != TSDB_CODE_SUCCESS \|\| pItem == NULL) { // failed, continue	1,835!
360	int64_t currentVer = walReaderGetCurrentVer(pTask->exec.pWalReader);	630✔
361	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, currentVer);	630✔
362	if (itemInFillhistory) {	630!
363	numOfNewItems += 1;	×
364	}
365	break;	630✔
366	}
367
368	if (pItem != NULL) {	1,205!
369	code = streamTaskPutDataIntoInputQ(pTask, pItem);	1,205✔
370	if (code == TSDB_CODE_SUCCESS) {	1,205!
371	numOfNewItems += 1;	1,205✔
372	int64_t ver = walReaderGetCurrentVer(pTask->exec.pWalReader);	1,205✔
373	pTask->chkInfo.nextProcessVer = ver;	1,205✔
374	tqDebug("s-task:%s set ver:%" PRId64 " for reader after extract data from WAL", id, ver);	1,205!
375
376	bool itemInFillhistory = handleFillhistoryScanComplete(pTask, ver);	1,205✔
377	if (itemInFillhistory) {	1,205!
378	break;	×
379	}
380	} else {
381	if (code == TSDB_CODE_OUT_OF_MEMORY) {	×
382	tqError("s-task:%s failed to put data into inputQ, since out of memory", id);	×
383	} else {
384	tqTrace("s-task:%s append input queue failed, code:inputQ is full, ver:%" PRId64, id,	×
385	pTask->chkInfo.nextProcessVer);
386	code = walReaderSeekVer(pTask->exec.pWalReader, pTask->chkInfo.nextProcessVer);	×
387	if (code) {	×
388	tqError("s-task:%s failed to seek ver to:%" PRId64 " in wal", id, pTask->chkInfo.nextProcessVer);	×
389	}
390
391	code = 0; // reset the error code	×
392	}
393
394	break;	×
395	}
396	}
397	}
398
399	*numOfItems += numOfNewItems;	630✔
400	*pSucc = (numOfNewItems > 0);	630✔
401	return code;	630✔
402	}
403
404	int32_t doScanWalForAllTasks(SStreamMeta* pStreamMeta, int32_t* pNumOfTasks) {	600✔
405	int32_t vgId = pStreamMeta->vgId;	600✔
406	SArray* pTaskList = NULL;	600✔
407	int32_t numOfTasks = 0;	600✔
408
409	// clone the task list, to avoid the task update during scan wal files
410	streamMetaWLock(pStreamMeta);	600✔
411	pTaskList = taosArrayDup(pStreamMeta->pTaskList, NULL);	600✔
412	streamMetaWUnLock(pStreamMeta);	600✔
413	if (pTaskList == NULL) {	600!
414	tqError("vgId:%d failed to create task list dup, code:%s", vgId, tstrerror(terrno));	×
415	return terrno;	×
416	}
417
418	// update the new task number
419	numOfTasks = taosArrayGetSize(pTaskList);	600✔
420	if (pNumOfTasks != NULL) {	600!
421	*pNumOfTasks = numOfTasks;	600✔
422	}
423
424	tqDebug("vgId:%d start to check wal to extract new submit block for %d tasks", vgId, numOfTasks);	600!
425
426	for (int32_t i = 0; i < numOfTasks; ++i) {	1,915✔
427	STaskId* pTaskId = taosArrayGet(pTaskList, i);	1,315✔
428	if (pTaskId == NULL) {	1,315!
429	continue;	685✔
430	}
431
432	SStreamTask* pTask = NULL;	1,315✔
433	int32_t code = streamMetaAcquireTask(pStreamMeta, pTaskId->streamId, pTaskId->taskId, &pTask);	1,315✔
434	if (pTask == NULL \|\| code != 0) {	1,315!
435	continue;	×
436	}
437
438	if (!taskReadyForDataFromWal(pTask)) {	1,315✔
439	streamMetaReleaseTask(pStreamMeta, pTask);	685✔
440	continue;	685✔
441	}
442
443	// seek the stored version and extract data from WAL
444	code = setWalReaderStartOffset(pTask, vgId);	630✔
445	if (code != TSDB_CODE_SUCCESS) {	630!
446	streamMetaReleaseTask(pStreamMeta, pTask);	×
447	continue;	×
448	}
449
450	int32_t numOfItems = streamQueueGetNumOfItems(pTask->inputq.queue);	630✔
451	int64_t maxVer = (pTask->info.fillHistory == 1) ? pTask->step2Range.maxVer : INT64_MAX;	630!
452
453	streamMutexLock(&pTask->lock);	630✔
454
455	SStreamTaskState state = streamTaskGetStatus(pTask);	630✔
456	if (state.state != TASK_STATUS__READY) {	630!
457	tqDebug("s-task:%s not ready for submit block from wal, status:%s", pTask->id.idStr, state.name);	×
458	streamMutexUnlock(&pTask->lock);	×
459	streamMetaReleaseTask(pStreamMeta, pTask);	×
460	continue;	×
461	}
462
463	bool hasNewData = false;	630✔
464	code = doPutDataIntoInputQ(pTask, maxVer, &numOfItems, &hasNewData);	630✔
465	streamMutexUnlock(&pTask->lock);	630✔
466
467	if ((numOfItems > 0) \|\| hasNewData) {	630!
468	code = streamTrySchedExec(pTask, false);	264✔
469	if (code != TSDB_CODE_SUCCESS) {	264!
470	streamMetaReleaseTask(pStreamMeta, pTask);	×
471	taosArrayDestroy(pTaskList);	×
472	return code;	×
473	}
474	}
475
476	streamMetaReleaseTask(pStreamMeta, pTask);	630✔
477	}
478
479	taosArrayDestroy(pTaskList);	600✔
480	return TSDB_CODE_SUCCESS;	600✔
481	}
482
483	void streamMetaFreeTQDuringScanWalError(STQ* pTq) {	×
484	SBuildScanWalMsgParam* p = taosMemoryCalloc(1, sizeof(SBuildScanWalMsgParam));	×
485	p->metaId = pTq->pStreamMeta->rid;	×
486
487	doStartScanWal(p, 0);	×
488	}	×

taosdata / TDengine / #3824

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