Examples with JobManagerTaskRestore org.apache.flink.runtime.checkpoint.JobManagerTaskRestore used on opensource projects

Example 1 with JobManagerTaskRestore

use of org.apache.flink.runtime.checkpoint.JobManagerTaskRestore in project flink by apache.

the class TaskExecutor method submitTask.

// ----------------------------------------------------------------------
// Task lifecycle RPCs
// ----------------------------------------------------------------------
@Override
public CompletableFuture<Acknowledge> submitTask(TaskDeploymentDescriptor tdd, JobMasterId jobMasterId, Time timeout) {
    try {
        final JobID jobId = tdd.getJobId();
        final ExecutionAttemptID executionAttemptID = tdd.getExecutionAttemptId();
        final JobTable.Connection jobManagerConnection = jobTable.getConnection(jobId).orElseThrow(() -> {
            final String message = "Could not submit task because there is no JobManager " + "associated for the job " + jobId + '.';
            log.debug(message);
            return new TaskSubmissionException(message);
        });
        if (!Objects.equals(jobManagerConnection.getJobMasterId(), jobMasterId)) {
            final String message = "Rejecting the task submission because the job manager leader id " + jobMasterId + " does not match the expected job manager leader id " + jobManagerConnection.getJobMasterId() + '.';
            log.debug(message);
            throw new TaskSubmissionException(message);
        }
        if (!taskSlotTable.tryMarkSlotActive(jobId, tdd.getAllocationId())) {
            final String message = "No task slot allocated for job ID " + jobId + " and allocation ID " + tdd.getAllocationId() + '.';
            log.debug(message);
            throw new TaskSubmissionException(message);
        }
        // re-integrate offloaded data:
        try {
            tdd.loadBigData(taskExecutorBlobService.getPermanentBlobService());
        } catch (IOException | ClassNotFoundException e) {
            throw new TaskSubmissionException("Could not re-integrate offloaded TaskDeploymentDescriptor data.", e);
        }
        // deserialize the pre-serialized information
        final JobInformation jobInformation;
        final TaskInformation taskInformation;
        try {
            jobInformation = tdd.getSerializedJobInformation().deserializeValue(getClass().getClassLoader());
            taskInformation = tdd.getSerializedTaskInformation().deserializeValue(getClass().getClassLoader());
        } catch (IOException | ClassNotFoundException e) {
            throw new TaskSubmissionException("Could not deserialize the job or task information.", e);
        }
        if (!jobId.equals(jobInformation.getJobId())) {
            throw new TaskSubmissionException("Inconsistent job ID information inside TaskDeploymentDescriptor (" + tdd.getJobId() + " vs. " + jobInformation.getJobId() + ")");
        }
        TaskManagerJobMetricGroup jobGroup = taskManagerMetricGroup.addJob(jobInformation.getJobId(), jobInformation.getJobName());
        // note that a pre-existing job group can NOT be closed concurrently - this is done by
        // the same TM thread in removeJobMetricsGroup
        TaskMetricGroup taskMetricGroup = jobGroup.addTask(taskInformation.getJobVertexId(), tdd.getExecutionAttemptId(), taskInformation.getTaskName(), tdd.getSubtaskIndex(), tdd.getAttemptNumber());
        InputSplitProvider inputSplitProvider = new RpcInputSplitProvider(jobManagerConnection.getJobManagerGateway(), taskInformation.getJobVertexId(), tdd.getExecutionAttemptId(), taskManagerConfiguration.getRpcTimeout());
        final TaskOperatorEventGateway taskOperatorEventGateway = new RpcTaskOperatorEventGateway(jobManagerConnection.getJobManagerGateway(), executionAttemptID, (t) -> runAsync(() -> failTask(executionAttemptID, t)));
        TaskManagerActions taskManagerActions = jobManagerConnection.getTaskManagerActions();
        CheckpointResponder checkpointResponder = jobManagerConnection.getCheckpointResponder();
        GlobalAggregateManager aggregateManager = jobManagerConnection.getGlobalAggregateManager();
        LibraryCacheManager.ClassLoaderHandle classLoaderHandle = jobManagerConnection.getClassLoaderHandle();
        ResultPartitionConsumableNotifier resultPartitionConsumableNotifier = jobManagerConnection.getResultPartitionConsumableNotifier();
        PartitionProducerStateChecker partitionStateChecker = jobManagerConnection.getPartitionStateChecker();
        final TaskLocalStateStore localStateStore = localStateStoresManager.localStateStoreForSubtask(jobId, tdd.getAllocationId(), taskInformation.getJobVertexId(), tdd.getSubtaskIndex());
        // TODO: Pass config value from user program and do overriding here.
        final StateChangelogStorage<?> changelogStorage;
        try {
            changelogStorage = changelogStoragesManager.stateChangelogStorageForJob(jobId, taskManagerConfiguration.getConfiguration(), jobGroup);
        } catch (IOException e) {
            throw new TaskSubmissionException(e);
        }
        final JobManagerTaskRestore taskRestore = tdd.getTaskRestore();
        final TaskStateManager taskStateManager = new TaskStateManagerImpl(jobId, tdd.getExecutionAttemptId(), localStateStore, changelogStorage, taskRestore, checkpointResponder);
        MemoryManager memoryManager;
        try {
            memoryManager = taskSlotTable.getTaskMemoryManager(tdd.getAllocationId());
        } catch (SlotNotFoundException e) {
            throw new TaskSubmissionException("Could not submit task.", e);
        }
        Task task = new Task(jobInformation, taskInformation, tdd.getExecutionAttemptId(), tdd.getAllocationId(), tdd.getSubtaskIndex(), tdd.getAttemptNumber(), tdd.getProducedPartitions(), tdd.getInputGates(), memoryManager, taskExecutorServices.getIOManager(), taskExecutorServices.getShuffleEnvironment(), taskExecutorServices.getKvStateService(), taskExecutorServices.getBroadcastVariableManager(), taskExecutorServices.getTaskEventDispatcher(), externalResourceInfoProvider, taskStateManager, taskManagerActions, inputSplitProvider, checkpointResponder, taskOperatorEventGateway, aggregateManager, classLoaderHandle, fileCache, taskManagerConfiguration, taskMetricGroup, resultPartitionConsumableNotifier, partitionStateChecker, getRpcService().getScheduledExecutor());
        taskMetricGroup.gauge(MetricNames.IS_BACK_PRESSURED, task::isBackPressured);
        log.info("Received task {} ({}), deploy into slot with allocation id {}.", task.getTaskInfo().getTaskNameWithSubtasks(), tdd.getExecutionAttemptId(), tdd.getAllocationId());
        boolean taskAdded;
        try {
            taskAdded = taskSlotTable.addTask(task);
        } catch (SlotNotFoundException | SlotNotActiveException e) {
            throw new TaskSubmissionException("Could not submit task.", e);
        }
        if (taskAdded) {
            task.startTaskThread();
            setupResultPartitionBookkeeping(tdd.getJobId(), tdd.getProducedPartitions(), task.getTerminationFuture());
            return CompletableFuture.completedFuture(Acknowledge.get());
        } else {
            final String message = "TaskManager already contains a task for id " + task.getExecutionId() + '.';
            log.debug(message);
            throw new TaskSubmissionException(message);
        }
    } catch (TaskSubmissionException e) {
        return FutureUtils.completedExceptionally(e);
    }
}

Example 2 with JobManagerTaskRestore

use of org.apache.flink.runtime.checkpoint.JobManagerTaskRestore in project flink by apache.

the class StateInitializationContextImplTest method setUp.

@Before
public void setUp() throws Exception {
    this.writtenKeyGroups = 0;
    this.writtenOperatorStates = new HashSet<>();
    this.closableRegistry = new CloseableRegistry();
    ByteArrayOutputStreamWithPos out = new ByteArrayOutputStreamWithPos(64);
    List<KeyedStateHandle> keyedStateHandles = new ArrayList<>(NUM_HANDLES);
    int prev = 0;
    for (int i = 0; i < NUM_HANDLES; ++i) {
        out.reset();
        int size = i % 4;
        int end = prev + size;
        DataOutputView dov = new DataOutputViewStreamWrapper(out);
        KeyGroupRangeOffsets offsets = new KeyGroupRangeOffsets(i == 9 ? KeyGroupRange.EMPTY_KEY_GROUP_RANGE : new KeyGroupRange(prev, end));
        prev = end + 1;
        for (int kg : offsets.getKeyGroupRange()) {
            offsets.setKeyGroupOffset(kg, out.getPosition());
            dov.writeInt(kg);
            ++writtenKeyGroups;
        }
        KeyedStateHandle handle = new KeyGroupsStateHandle(offsets, new ByteStateHandleCloseChecking("kg-" + i, out.toByteArray()));
        keyedStateHandles.add(handle);
    }
    List<OperatorStateHandle> operatorStateHandles = new ArrayList<>(NUM_HANDLES);
    for (int i = 0; i < NUM_HANDLES; ++i) {
        int size = i % 4;
        out.reset();
        DataOutputView dov = new DataOutputViewStreamWrapper(out);
        LongArrayList offsets = new LongArrayList(size);
        for (int s = 0; s < size; ++s) {
            offsets.add(out.getPosition());
            int val = i * NUM_HANDLES + s;
            dov.writeInt(val);
            writtenOperatorStates.add(val);
        }
        Map<String, OperatorStateHandle.StateMetaInfo> offsetsMap = new HashMap<>();
        offsetsMap.put(DefaultOperatorStateBackend.DEFAULT_OPERATOR_STATE_NAME, new OperatorStateHandle.StateMetaInfo(offsets.toArray(), OperatorStateHandle.Mode.SPLIT_DISTRIBUTE));
        OperatorStateHandle operatorStateHandle = new OperatorStreamStateHandle(offsetsMap, new ByteStateHandleCloseChecking("os-" + i, out.toByteArray()));
        operatorStateHandles.add(operatorStateHandle);
    }
    OperatorSubtaskState operatorSubtaskState = OperatorSubtaskState.builder().setRawOperatorState(new StateObjectCollection<>(operatorStateHandles)).setRawKeyedState(new StateObjectCollection<>(keyedStateHandles)).build();
    OperatorID operatorID = new OperatorID();
    TaskStateSnapshot taskStateSnapshot = new TaskStateSnapshot();
    taskStateSnapshot.putSubtaskStateByOperatorID(operatorID, operatorSubtaskState);
    JobManagerTaskRestore jobManagerTaskRestore = new JobManagerTaskRestore(0L, taskStateSnapshot);
    TaskStateManager manager = new TaskStateManagerImpl(new JobID(), new ExecutionAttemptID(), new TestTaskLocalStateStore(), new InMemoryStateChangelogStorage(), jobManagerTaskRestore, mock(CheckpointResponder.class));
    DummyEnvironment environment = new DummyEnvironment("test", 1, 0, prev);
    environment.setTaskStateManager(manager);
    StateBackend stateBackend = new MemoryStateBackend(1024);
    StreamTaskStateInitializer streamTaskStateManager = new StreamTaskStateInitializerImpl(environment, stateBackend, TtlTimeProvider.DEFAULT, new InternalTimeServiceManager.Provider() {

        @Override
        public <K> InternalTimeServiceManager<K> create(CheckpointableKeyedStateBackend<K> keyedStatedBackend, ClassLoader userClassloader, KeyContext keyContext, ProcessingTimeService processingTimeService, Iterable<KeyGroupStatePartitionStreamProvider> rawKeyedStates) throws Exception {
            // stream.
            return null;
        }
    });
    AbstractStreamOperator<?> mockOperator = mock(AbstractStreamOperator.class);
    when(mockOperator.getOperatorID()).thenReturn(operatorID);
    StreamOperatorStateContext stateContext = streamTaskStateManager.streamOperatorStateContext(operatorID, "TestOperatorClass", mock(ProcessingTimeService.class), mockOperator, // consumed by the timer service.
    IntSerializer.INSTANCE, closableRegistry, new UnregisteredMetricsGroup(), 1.0, false);
    OptionalLong restoredCheckpointId = stateContext.getRestoredCheckpointId();
    this.initializationContext = new StateInitializationContextImpl(restoredCheckpointId.isPresent() ? restoredCheckpointId.getAsLong() : null, stateContext.operatorStateBackend(), mock(KeyedStateStore.class), stateContext.rawKeyedStateInputs(), stateContext.rawOperatorStateInputs());
}

Example 3 with JobManagerTaskRestore

use of org.apache.flink.runtime.checkpoint.JobManagerTaskRestore in project flink by apache.

the class StatefulOperatorChainedTaskTest method createRunAndCheckpointOperatorChain.

private JobManagerTaskRestore createRunAndCheckpointOperatorChain(OperatorID headId, OneInputStreamOperator<String, String> headOperator, OperatorID tailId, OneInputStreamOperator<String, String> tailOperator, Optional<JobManagerTaskRestore> restore) throws Exception {
    File localRootDir = temporaryFolder.newFolder();
    final OneInputStreamTaskTestHarness<String, String> testHarness = new OneInputStreamTaskTestHarness<>(OneInputStreamTask::new, 1, 1, BasicTypeInfo.STRING_TYPE_INFO, BasicTypeInfo.STRING_TYPE_INFO, localRootDir);
    testHarness.setupOperatorChain(headId, headOperator).chain(tailId, tailOperator, StringSerializer.INSTANCE, true).finish();
    if (restore.isPresent()) {
        JobManagerTaskRestore taskRestore = restore.get();
        testHarness.setTaskStateSnapshot(taskRestore.getRestoreCheckpointId(), taskRestore.getTaskStateSnapshot());
    }
    StreamMockEnvironment environment = new StreamMockEnvironment(testHarness.jobConfig, testHarness.taskConfig, testHarness.getExecutionConfig(), testHarness.memorySize, new MockInputSplitProvider(), testHarness.bufferSize, testHarness.getTaskStateManager());
    Configuration configuration = new Configuration();
    configuration.setString(STATE_BACKEND.key(), "rocksdb");
    File file = temporaryFolder.newFolder();
    configuration.setString(CHECKPOINTS_DIRECTORY.key(), file.toURI().toString());
    configuration.setString(INCREMENTAL_CHECKPOINTS.key(), "true");
    environment.setTaskManagerInfo(new TestingTaskManagerRuntimeInfo(configuration, System.getProperty("java.io.tmpdir").split(",|" + File.pathSeparator)));
    testHarness.invoke(environment);
    testHarness.waitForTaskRunning();
    OneInputStreamTask<String, String> streamTask = testHarness.getTask();
    processRecords(testHarness);
    triggerCheckpoint(testHarness, streamTask);
    TestTaskStateManager taskStateManager = testHarness.getTaskStateManager();
    JobManagerTaskRestore jobManagerTaskRestore = new JobManagerTaskRestore(taskStateManager.getReportedCheckpointId(), taskStateManager.getLastJobManagerTaskStateSnapshot());
    testHarness.endInput();
    testHarness.waitForTaskCompletion();
    return jobManagerTaskRestore;
}

Example 4 with JobManagerTaskRestore

use of org.apache.flink.runtime.checkpoint.JobManagerTaskRestore in project flink by apache.

the class TaskStateManagerImplTest method testAcquringRestoreCheckpointId.

public void testAcquringRestoreCheckpointId() {
    TaskStateManagerImpl emptyStateManager = new TaskStateManagerImpl(new JobID(), new ExecutionAttemptID(), new TestTaskLocalStateStore(), null, null, new TestCheckpointResponder());
    Assert.assertFalse(emptyStateManager.getRestoreCheckpointId().isPresent());
    TaskStateManagerImpl nonEmptyStateManager = new TaskStateManagerImpl(new JobID(), new ExecutionAttemptID(), new TestTaskLocalStateStore(), null, new JobManagerTaskRestore(2, new TaskStateSnapshot()), new TestCheckpointResponder());
    Assert.assertEquals(2L, (long) nonEmptyStateManager.getRestoreCheckpointId().get());
}

Example 5 with JobManagerTaskRestore

use of org.apache.flink.runtime.checkpoint.JobManagerTaskRestore in project flink by apache.

the class InterruptSensitiveRestoreTest method createTask.

// ------------------------------------------------------------------------
// Utilities
// ------------------------------------------------------------------------
private static Task createTask(StreamConfig streamConfig, Configuration taskConfig, StreamStateHandle state, int mode) throws IOException {
    ShuffleEnvironment<?, ?> shuffleEnvironment = new NettyShuffleEnvironmentBuilder().build();
    Collection<KeyedStateHandle> keyedStateFromBackend = Collections.emptyList();
    Collection<KeyedStateHandle> keyedStateFromStream = Collections.emptyList();
    Collection<OperatorStateHandle> operatorStateBackend = Collections.emptyList();
    Collection<OperatorStateHandle> operatorStateStream = Collections.emptyList();
    Map<String, OperatorStateHandle.StateMetaInfo> operatorStateMetadata = new HashMap<>(1);
    OperatorStateHandle.StateMetaInfo metaInfo = new OperatorStateHandle.StateMetaInfo(new long[] { 0 }, OperatorStateHandle.Mode.SPLIT_DISTRIBUTE);
    operatorStateMetadata.put(DefaultOperatorStateBackend.DEFAULT_OPERATOR_STATE_NAME, metaInfo);
    KeyGroupRangeOffsets keyGroupRangeOffsets = new KeyGroupRangeOffsets(new KeyGroupRange(0, 0));
    Collection<OperatorStateHandle> operatorStateHandles = Collections.singletonList(new OperatorStreamStateHandle(operatorStateMetadata, state));
    List<KeyedStateHandle> keyedStateHandles = Collections.singletonList(new KeyGroupsStateHandle(keyGroupRangeOffsets, state));
    switch(mode) {
        case OPERATOR_MANAGED:
            operatorStateBackend = operatorStateHandles;
            break;
        case OPERATOR_RAW:
            operatorStateStream = operatorStateHandles;
            break;
        case KEYED_MANAGED:
            keyedStateFromBackend = keyedStateHandles;
            break;
        case KEYED_RAW:
            keyedStateFromStream = keyedStateHandles;
            break;
        default:
            throw new IllegalArgumentException();
    }
    OperatorSubtaskState operatorSubtaskState = OperatorSubtaskState.builder().setManagedOperatorState(new StateObjectCollection<>(operatorStateBackend)).setRawOperatorState(new StateObjectCollection<>(operatorStateStream)).setManagedKeyedState(new StateObjectCollection<>(keyedStateFromBackend)).setRawKeyedState(new StateObjectCollection<>(keyedStateFromStream)).build();
    JobVertexID jobVertexID = new JobVertexID();
    OperatorID operatorID = OperatorID.fromJobVertexID(jobVertexID);
    streamConfig.setOperatorID(operatorID);
    TaskStateSnapshot stateSnapshot = new TaskStateSnapshot();
    stateSnapshot.putSubtaskStateByOperatorID(operatorID, operatorSubtaskState);
    JobManagerTaskRestore taskRestore = new JobManagerTaskRestore(1L, stateSnapshot);
    JobInformation jobInformation = new JobInformation(new JobID(), "test job name", new SerializedValue<>(new ExecutionConfig()), new Configuration(), Collections.emptyList(), Collections.emptyList());
    TaskInformation taskInformation = new TaskInformation(jobVertexID, "test task name", 1, 1, SourceStreamTask.class.getName(), taskConfig);
    TestTaskStateManager taskStateManager = TestTaskStateManager.builder().setReportedCheckpointId(taskRestore.getRestoreCheckpointId()).setJobManagerTaskStateSnapshotsByCheckpointId(Collections.singletonMap(taskRestore.getRestoreCheckpointId(), taskRestore.getTaskStateSnapshot())).build();
    return new Task(jobInformation, taskInformation, new ExecutionAttemptID(), new AllocationID(), 0, 0, Collections.<ResultPartitionDeploymentDescriptor>emptyList(), Collections.<InputGateDeploymentDescriptor>emptyList(), mock(MemoryManager.class), mock(IOManager.class), shuffleEnvironment, new KvStateService(new KvStateRegistry(), null, null), mock(BroadcastVariableManager.class), new TaskEventDispatcher(), ExternalResourceInfoProvider.NO_EXTERNAL_RESOURCES, taskStateManager, mock(TaskManagerActions.class), mock(InputSplitProvider.class), mock(CheckpointResponder.class), new NoOpTaskOperatorEventGateway(), new TestGlobalAggregateManager(), TestingClassLoaderLease.newBuilder().build(), new FileCache(new String[] { EnvironmentInformation.getTemporaryFileDirectory() }, VoidPermanentBlobService.INSTANCE), new TestingTaskManagerRuntimeInfo(), UnregisteredMetricGroups.createUnregisteredTaskMetricGroup(), new NoOpResultPartitionConsumableNotifier(), mock(PartitionProducerStateChecker.class), mock(Executor.class));
}