Examples with ArchivedExecutionVertex org.apache.flink.runtime.executiongraph.ArchivedExecutionVertex used on opensource projects

Example 21 with ArchivedExecutionVertex

use of org.apache.flink.runtime.executiongraph.ArchivedExecutionVertex in project flink by apache.

the class DefaultSchedulerTest method vertexIsNotAffectedByOutdatedDeployment.

@Test
public void vertexIsNotAffectedByOutdatedDeployment() {
    final JobGraph jobGraph = singleJobVertexJobGraph(2);
    testExecutionSlotAllocator.disableAutoCompletePendingRequests();
    final DefaultScheduler scheduler = createSchedulerAndStartScheduling(jobGraph);
    final Iterator<ArchivedExecutionVertex> vertexIterator = scheduler.requestJob().getArchivedExecutionGraph().getAllExecutionVertices().iterator();
    final ArchivedExecutionVertex v1 = vertexIterator.next();
    final ArchivedExecutionVertex v2 = vertexIterator.next();
    final SchedulingExecutionVertex sv1 = scheduler.getSchedulingTopology().getVertices().iterator().next();
    // fail v1 and let it recover to SCHEDULED
    // the initial deployment of v1 will be outdated
    scheduler.updateTaskExecutionState(createFailedTaskExecutionState(v1.getCurrentExecutionAttempt().getAttemptId()));
    taskRestartExecutor.triggerScheduledTasks();
    // fail v2 to get all pending slot requests in the initial deployments to be done
    // this triggers the outdated deployment of v1
    scheduler.updateTaskExecutionState(createFailedTaskExecutionState(v2.getCurrentExecutionAttempt().getAttemptId()));
    // v1 should not be affected
    assertThat(sv1.getState(), is(equalTo(ExecutionState.SCHEDULED)));
}

Example 22 with ArchivedExecutionVertex

use of org.apache.flink.runtime.executiongraph.ArchivedExecutionVertex in project flink by apache.

the class DefaultSchedulerTest method failGlobalWhenRestoringStateFails.

@Test
public void failGlobalWhenRestoringStateFails() throws Exception {
    final JobGraph jobGraph = singleNonParallelJobVertexJobGraph();
    final JobVertex onlyJobVertex = getOnlyJobVertex(jobGraph);
    enableCheckpointing(jobGraph);
    final CountDownLatch checkpointTriggeredLatch = getCheckpointTriggeredLatch();
    final DefaultScheduler scheduler = createSchedulerAndStartScheduling(jobGraph);
    final ArchivedExecutionVertex onlyExecutionVertex = Iterables.getOnlyElement(scheduler.requestJob().getArchivedExecutionGraph().getAllExecutionVertices());
    final ExecutionAttemptID attemptId = onlyExecutionVertex.getCurrentExecutionAttempt().getAttemptId();
    transitionToRunning(scheduler, attemptId);
    final CheckpointCoordinator checkpointCoordinator = getCheckpointCoordinator(scheduler);
    // register a master hook to fail state restore
    final TestMasterHook masterHook = TestMasterHook.fromId("testHook");
    masterHook.enableFailOnRestore();
    checkpointCoordinator.addMasterHook(masterHook);
    // complete one checkpoint for state restore
    checkpointCoordinator.triggerCheckpoint(false);
    checkpointTriggeredLatch.await();
    final long checkpointId = checkpointCoordinator.getPendingCheckpoints().keySet().iterator().next();
    acknowledgePendingCheckpoint(scheduler, checkpointId);
    scheduler.updateTaskExecutionState(createFailedTaskExecutionState(attemptId));
    taskRestartExecutor.triggerScheduledTasks();
    final List<ExecutionVertexID> deployedExecutionVertices = testExecutionVertexOperations.getDeployedVertices();
    // the first task failover should be skipped on state restore failure
    final ExecutionVertexID executionVertexId = new ExecutionVertexID(onlyJobVertex.getID(), 0);
    assertThat(deployedExecutionVertices, contains(executionVertexId));
    // a global failure should be triggered on state restore failure
    masterHook.disableFailOnRestore();
    taskRestartExecutor.triggerScheduledTasks();
    assertThat(deployedExecutionVertices, contains(executionVertexId, executionVertexId));
}

Example 23 with ArchivedExecutionVertex

use of org.apache.flink.runtime.executiongraph.ArchivedExecutionVertex in project flink by apache.

the class DefaultSchedulerTest method testExceptionHistoryWithPreDeployFailure.

@Test
public void testExceptionHistoryWithPreDeployFailure() {
    // disable auto-completing slot requests to simulate timeout
    executionSlotAllocatorFactory.getTestExecutionSlotAllocator().disableAutoCompletePendingRequests();
    final DefaultScheduler scheduler = createSchedulerAndStartScheduling(singleNonParallelJobVertexJobGraph());
    executionSlotAllocatorFactory.getTestExecutionSlotAllocator().timeoutPendingRequests();
    final ArchivedExecutionVertex taskFailureExecutionVertex = Iterables.getOnlyElement(scheduler.requestJob().getArchivedExecutionGraph().getAllExecutionVertices());
    // pending slot request timeout triggers a task failure that needs to be processed
    taskRestartExecutor.triggerNonPeriodicScheduledTask();
    // sanity check that the TaskManagerLocation of the failed task is indeed null, as expected
    assertThat(taskFailureExecutionVertex.getCurrentAssignedResourceLocation(), is(nullValue()));
    final ErrorInfo failureInfo = taskFailureExecutionVertex.getFailureInfo().orElseThrow(() -> new AssertionError("A failureInfo should be set."));
    final Iterable<RootExceptionHistoryEntry> actualExceptionHistory = scheduler.getExceptionHistory();
    assertThat(actualExceptionHistory, IsIterableContainingInOrder.contains(ExceptionHistoryEntryMatcher.matchesFailure(failureInfo.getException(), failureInfo.getTimestamp(), taskFailureExecutionVertex.getTaskNameWithSubtaskIndex(), taskFailureExecutionVertex.getCurrentAssignedResourceLocation())));
}

Example 24 with ArchivedExecutionVertex

use of org.apache.flink.runtime.executiongraph.ArchivedExecutionVertex in project flink by apache.

the class AdaptiveSchedulerTest method runExceptionHistoryTests.

private Iterable<RootExceptionHistoryEntry> runExceptionHistoryTests(BiConsumer<AdaptiveScheduler, List<ExecutionAttemptID>> testLogic, Consumer<AdaptiveSchedulerBuilder> setupScheduler, Consumer<JobGraph> setupJobGraph) throws Exception {
    final int numAvailableSlots = 4;
    final JobGraph jobGraph = createJobGraph();
    setupJobGraph.accept(jobGraph);
    RunFailedJobListener listener = new RunFailedJobListener();
    List<ExecutionAttemptID> cancelledTasks = new ArrayList<>();
    final CompletedCheckpointStore completedCheckpointStore = new StandaloneCompletedCheckpointStore(1);
    final CheckpointIDCounter checkpointIDCounter = new StandaloneCheckpointIDCounter();
    final CheckpointsCleaner checkpointCleaner = new CheckpointsCleaner();
    TestingCheckpointRecoveryFactory checkpointRecoveryFactory = new TestingCheckpointRecoveryFactory(completedCheckpointStore, checkpointIDCounter);
    final DefaultDeclarativeSlotPool declarativeSlotPool = createDeclarativeSlotPool(jobGraph.getJobID());
    final Configuration configuration = new Configuration();
    configuration.set(JobManagerOptions.RESOURCE_WAIT_TIMEOUT, Duration.ofMillis(1L));
    AdaptiveSchedulerBuilder builder = new AdaptiveSchedulerBuilder(jobGraph, singleThreadMainThreadExecutor).setJobMasterConfiguration(configuration).setDeclarativeSlotPool(declarativeSlotPool).setCheckpointRecoveryFactory(checkpointRecoveryFactory).setCheckpointCleaner(checkpointCleaner).setJobStatusListener(listener);
    setupScheduler.accept(builder);
    final AdaptiveScheduler scheduler = builder.build();
    final SubmissionBufferingTaskManagerGateway taskManagerGateway = new SubmissionBufferingTaskManagerGateway(numAvailableSlots);
    taskManagerGateway.setCancelConsumer(cancelledTasks::add);
    singleThreadMainThreadExecutor.execute(() -> {
        scheduler.startScheduling();
        offerSlots(declarativeSlotPool, createSlotOffersForResourceRequirements(ResourceCounter.withResource(ResourceProfile.UNKNOWN, numAvailableSlots)), taskManagerGateway);
    });
    listener.waitForRunning();
    CompletableFuture<Iterable<ArchivedExecutionVertex>> vertexFuture = new CompletableFuture<>();
    singleThreadMainThreadExecutor.execute(() -> vertexFuture.complete(scheduler.requestJob().getArchivedExecutionGraph().getAllExecutionVertices()));
    final Iterable<ArchivedExecutionVertex> executionVertices = vertexFuture.get();
    final List<ExecutionAttemptID> attemptIds = IterableUtils.toStream(executionVertices).map(ArchivedExecutionVertex::getCurrentExecutionAttempt).map(ArchivedExecution::getAttemptId).collect(Collectors.toList());
    CompletableFuture<Void> runTestLogicFuture = CompletableFuture.runAsync(() -> testLogic.accept(scheduler, attemptIds), singleThreadMainThreadExecutor);
    runTestLogicFuture.get();
    Consumer<ExecutionAttemptID> canceller = attemptId -> scheduler.updateTaskExecutionState(new TaskExecutionStateTransition(new TaskExecutionState(attemptId, ExecutionState.CANCELED, null)));
    CompletableFuture<Void> cancelFuture = CompletableFuture.runAsync(() -> cancelledTasks.forEach(canceller), singleThreadMainThreadExecutor);
    cancelFuture.get();
    listener.waitForTerminal();
    return scheduler.requestJob().getExceptionHistory();
}