Examples with NotifyWhenJobStatus org.apache.flink.runtime.testingUtils.TestingJobManagerMessages.NotifyWhenJobStatus used on opensource projects

Example 1 with NotifyWhenJobStatus

use of org.apache.flink.runtime.testingUtils.TestingJobManagerMessages.NotifyWhenJobStatus in project flink by apache.

the class JobManagerTest method testKvStateMessages.

/**
	 * Tests that the JobManager handles {@link org.apache.flink.runtime.query.KvStateMessage}
	 * instances as expected.
	 */
@Test
public void testKvStateMessages() throws Exception {
    Deadline deadline = new FiniteDuration(100, TimeUnit.SECONDS).fromNow();
    Configuration config = new Configuration();
    config.setString(ConfigConstants.AKKA_ASK_TIMEOUT, "100ms");
    UUID leaderSessionId = null;
    ActorGateway jobManager = new AkkaActorGateway(JobManager.startJobManagerActors(config, system, TestingUtils.defaultExecutor(), TestingUtils.defaultExecutor(), TestingJobManager.class, MemoryArchivist.class)._1(), leaderSessionId);
    LeaderRetrievalService leaderRetrievalService = new StandaloneLeaderRetrievalService(AkkaUtils.getAkkaURL(system, jobManager.actor()));
    Configuration tmConfig = new Configuration();
    tmConfig.setInteger(ConfigConstants.TASK_MANAGER_MEMORY_SIZE_KEY, 4);
    tmConfig.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, 8);
    ActorRef taskManager = TaskManager.startTaskManagerComponentsAndActor(tmConfig, ResourceID.generate(), system, "localhost", scala.Option.<String>empty(), scala.Option.apply(leaderRetrievalService), true, TestingTaskManager.class);
    Future<Object> registrationFuture = jobManager.ask(new NotifyWhenAtLeastNumTaskManagerAreRegistered(1), deadline.timeLeft());
    Await.ready(registrationFuture, deadline.timeLeft());
    //
    // Location lookup
    //
    LookupKvStateLocation lookupNonExistingJob = new LookupKvStateLocation(new JobID(), "any-name");
    Future<KvStateLocation> lookupFuture = jobManager.ask(lookupNonExistingJob, deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<KvStateLocation>apply(KvStateLocation.class));
    try {
        Await.result(lookupFuture, deadline.timeLeft());
        fail("Did not throw expected Exception");
    } catch (IllegalStateException ignored) {
    // Expected
    }
    JobGraph jobGraph = new JobGraph("croissant");
    JobVertex jobVertex1 = new JobVertex("cappuccino");
    jobVertex1.setParallelism(4);
    jobVertex1.setMaxParallelism(16);
    jobVertex1.setInvokableClass(BlockingNoOpInvokable.class);
    JobVertex jobVertex2 = new JobVertex("americano");
    jobVertex2.setParallelism(4);
    jobVertex2.setMaxParallelism(16);
    jobVertex2.setInvokableClass(BlockingNoOpInvokable.class);
    jobGraph.addVertex(jobVertex1);
    jobGraph.addVertex(jobVertex2);
    Future<JobSubmitSuccess> submitFuture = jobManager.ask(new SubmitJob(jobGraph, ListeningBehaviour.DETACHED), deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<JobSubmitSuccess>apply(JobSubmitSuccess.class));
    Await.result(submitFuture, deadline.timeLeft());
    Object lookupUnknownRegistrationName = new LookupKvStateLocation(jobGraph.getJobID(), "unknown");
    lookupFuture = jobManager.ask(lookupUnknownRegistrationName, deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<KvStateLocation>apply(KvStateLocation.class));
    try {
        Await.result(lookupFuture, deadline.timeLeft());
        fail("Did not throw expected Exception");
    } catch (UnknownKvStateLocation ignored) {
    // Expected
    }
    //
    // Registration
    //
    NotifyKvStateRegistered registerNonExistingJob = new NotifyKvStateRegistered(new JobID(), new JobVertexID(), new KeyGroupRange(0, 0), "any-name", new KvStateID(), new KvStateServerAddress(InetAddress.getLocalHost(), 1233));
    jobManager.tell(registerNonExistingJob);
    LookupKvStateLocation lookupAfterRegistration = new LookupKvStateLocation(registerNonExistingJob.getJobId(), registerNonExistingJob.getRegistrationName());
    lookupFuture = jobManager.ask(lookupAfterRegistration, deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<KvStateLocation>apply(KvStateLocation.class));
    try {
        Await.result(lookupFuture, deadline.timeLeft());
        fail("Did not throw expected Exception");
    } catch (IllegalStateException ignored) {
    // Expected
    }
    NotifyKvStateRegistered registerForExistingJob = new NotifyKvStateRegistered(jobGraph.getJobID(), jobVertex1.getID(), new KeyGroupRange(0, 0), "register-me", new KvStateID(), new KvStateServerAddress(InetAddress.getLocalHost(), 1293));
    jobManager.tell(registerForExistingJob);
    lookupAfterRegistration = new LookupKvStateLocation(registerForExistingJob.getJobId(), registerForExistingJob.getRegistrationName());
    lookupFuture = jobManager.ask(lookupAfterRegistration, deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<KvStateLocation>apply(KvStateLocation.class));
    KvStateLocation location = Await.result(lookupFuture, deadline.timeLeft());
    assertNotNull(location);
    assertEquals(jobGraph.getJobID(), location.getJobId());
    assertEquals(jobVertex1.getID(), location.getJobVertexId());
    assertEquals(jobVertex1.getMaxParallelism(), location.getNumKeyGroups());
    assertEquals(1, location.getNumRegisteredKeyGroups());
    KeyGroupRange keyGroupRange = registerForExistingJob.getKeyGroupRange();
    assertEquals(1, keyGroupRange.getNumberOfKeyGroups());
    assertEquals(registerForExistingJob.getKvStateId(), location.getKvStateID(keyGroupRange.getStartKeyGroup()));
    assertEquals(registerForExistingJob.getKvStateServerAddress(), location.getKvStateServerAddress(keyGroupRange.getStartKeyGroup()));
    //
    // Unregistration
    //
    NotifyKvStateUnregistered unregister = new NotifyKvStateUnregistered(registerForExistingJob.getJobId(), registerForExistingJob.getJobVertexId(), registerForExistingJob.getKeyGroupRange(), registerForExistingJob.getRegistrationName());
    jobManager.tell(unregister);
    lookupFuture = jobManager.ask(lookupAfterRegistration, deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<KvStateLocation>apply(KvStateLocation.class));
    try {
        Await.result(lookupFuture, deadline.timeLeft());
        fail("Did not throw expected Exception");
    } catch (UnknownKvStateLocation ignored) {
    // Expected
    }
    //
    // Duplicate registration fails task
    //
    NotifyKvStateRegistered register = new NotifyKvStateRegistered(jobGraph.getJobID(), jobVertex1.getID(), new KeyGroupRange(0, 0), "duplicate-me", new KvStateID(), new KvStateServerAddress(InetAddress.getLocalHost(), 1293));
    NotifyKvStateRegistered duplicate = new NotifyKvStateRegistered(jobGraph.getJobID(), // <--- different operator, but...
    jobVertex2.getID(), new KeyGroupRange(0, 0), // ...same name
    "duplicate-me", new KvStateID(), new KvStateServerAddress(InetAddress.getLocalHost(), 1293));
    Future<TestingJobManagerMessages.JobStatusIs> failedFuture = jobManager.ask(new NotifyWhenJobStatus(jobGraph.getJobID(), JobStatus.FAILED), deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<JobStatusIs>apply(JobStatusIs.class));
    jobManager.tell(register);
    jobManager.tell(duplicate);
    // Wait for failure
    JobStatusIs jobStatus = Await.result(failedFuture, deadline.timeLeft());
    assertEquals(JobStatus.FAILED, jobStatus.state());
}

Example 2 with NotifyWhenJobStatus

use of org.apache.flink.runtime.testingUtils.TestingJobManagerMessages.NotifyWhenJobStatus in project flink by apache.

the class TaskCancelAsyncProducerConsumerITCase method testCancelAsyncProducerAndConsumer.

/**
	 * Tests that a task waiting on an async producer/consumer that is stuck
	 * in a blocking buffer request can be properly cancelled.
	 *
	 * <p>This is currently required for the Flink Kafka sources, which spawn
	 * a separate Thread consuming from Kafka and producing the intermediate
	 * streams in the spawned Thread instead of the main task Thread.
	 */
@Test
public void testCancelAsyncProducerAndConsumer() throws Exception {
    Deadline deadline = new FiniteDuration(2, TimeUnit.MINUTES).fromNow();
    TestingCluster flink = null;
    try {
        // Cluster
        Configuration config = new Configuration();
        config.setInteger(ConfigConstants.LOCAL_NUMBER_TASK_MANAGER, 1);
        config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, 1);
        config.setInteger(ConfigConstants.TASK_MANAGER_MEMORY_SEGMENT_SIZE_KEY, 4096);
        config.setInteger(ConfigConstants.TASK_MANAGER_NETWORK_NUM_BUFFERS_KEY, 8);
        flink = new TestingCluster(config, true);
        flink.start();
        // Job with async producer and consumer
        JobVertex producer = new JobVertex("AsyncProducer");
        producer.setParallelism(1);
        producer.setInvokableClass(AsyncProducer.class);
        JobVertex consumer = new JobVertex("AsyncConsumer");
        consumer.setParallelism(1);
        consumer.setInvokableClass(AsyncConsumer.class);
        consumer.connectNewDataSetAsInput(producer, DistributionPattern.POINTWISE, ResultPartitionType.PIPELINED);
        SlotSharingGroup slot = new SlotSharingGroup(producer.getID(), consumer.getID());
        producer.setSlotSharingGroup(slot);
        consumer.setSlotSharingGroup(slot);
        JobGraph jobGraph = new JobGraph(producer, consumer);
        // Submit job and wait until running
        ActorGateway jobManager = flink.getLeaderGateway(deadline.timeLeft());
        flink.submitJobDetached(jobGraph);
        Object msg = new WaitForAllVerticesToBeRunning(jobGraph.getJobID());
        Future<?> runningFuture = jobManager.ask(msg, deadline.timeLeft());
        Await.ready(runningFuture, deadline.timeLeft());
        // Wait for blocking requests, cancel and wait for cancellation
        msg = new NotifyWhenJobStatus(jobGraph.getJobID(), JobStatus.CANCELED);
        Future<?> cancelledFuture = jobManager.ask(msg, deadline.timeLeft());
        boolean producerBlocked = false;
        for (int i = 0; i < 50; i++) {
            Thread thread = ASYNC_PRODUCER_THREAD;
            if (thread != null && thread.isAlive()) {
                StackTraceElement[] stackTrace = thread.getStackTrace();
                producerBlocked = isInBlockingBufferRequest(stackTrace);
            }
            if (producerBlocked) {
                break;
            } else {
                // Retry
                Thread.sleep(500);
            }
        }
        // Verify that async producer is in blocking request
        assertTrue("Producer thread is not blocked: " + Arrays.toString(ASYNC_CONSUMER_THREAD.getStackTrace()), producerBlocked);
        boolean consumerWaiting = false;
        for (int i = 0; i < 50; i++) {
            Thread thread = ASYNC_CONSUMER_THREAD;
            if (thread != null && thread.isAlive()) {
                consumerWaiting = thread.getState() == Thread.State.WAITING;
            }
            if (consumerWaiting) {
                break;
            } else {
                // Retry
                Thread.sleep(500);
            }
        }
        // Verify that async consumer is in blocking request
        assertTrue("Consumer thread is not blocked.", consumerWaiting);
        msg = new CancelJob(jobGraph.getJobID());
        Future<?> cancelFuture = jobManager.ask(msg, deadline.timeLeft());
        Await.ready(cancelFuture, deadline.timeLeft());
        Await.ready(cancelledFuture, deadline.timeLeft());
        // Verify the expected Exceptions
        assertNotNull(ASYNC_PRODUCER_EXCEPTION);
        assertEquals(IllegalStateException.class, ASYNC_PRODUCER_EXCEPTION.getClass());
        assertNotNull(ASYNC_CONSUMER_EXCEPTION);
        assertEquals(IllegalStateException.class, ASYNC_CONSUMER_EXCEPTION.getClass());
    } finally {
        if (flink != null) {
            flink.shutdown();
        }
    }
}

Example 3 with NotifyWhenJobStatus

use of org.apache.flink.runtime.testingUtils.TestingJobManagerMessages.NotifyWhenJobStatus in project flink by apache.

the class AbstractQueryableStateITCase method testDuplicateRegistrationFailsJob.

/**
	 * Tests that duplicate query registrations fail the job at the JobManager.
	 */
@Test
public void testDuplicateRegistrationFailsJob() throws Exception {
    // Config
    final Deadline deadline = TEST_TIMEOUT.fromNow();
    final int numKeys = 256;
    JobID jobId = null;
    try {
        //
        // Test program
        //
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setStateBackend(stateBackend);
        env.setParallelism(NUM_SLOTS);
        // Very important, because cluster is shared between tests and we
        // don't explicitly check that all slots are available before
        // submitting.
        env.setRestartStrategy(RestartStrategies.fixedDelayRestart(Integer.MAX_VALUE, 1000));
        DataStream<Tuple2<Integer, Long>> source = env.addSource(new TestKeyRangeSource(numKeys));
        // Reducing state
        ReducingStateDescriptor<Tuple2<Integer, Long>> reducingState = new ReducingStateDescriptor<>("any-name", new SumReduce(), source.getType());
        final String queryName = "duplicate-me";
        final QueryableStateStream<Integer, Tuple2<Integer, Long>> queryableState = source.keyBy(new KeySelector<Tuple2<Integer, Long>, Integer>() {

            @Override
            public Integer getKey(Tuple2<Integer, Long> value) throws Exception {
                return value.f0;
            }
        }).asQueryableState(queryName, reducingState);
        final QueryableStateStream<Integer, Tuple2<Integer, Long>> duplicate = source.keyBy(new KeySelector<Tuple2<Integer, Long>, Integer>() {

            @Override
            public Integer getKey(Tuple2<Integer, Long> value) throws Exception {
                return value.f0;
            }
        }).asQueryableState(queryName);
        // Submit the job graph
        JobGraph jobGraph = env.getStreamGraph().getJobGraph();
        jobId = jobGraph.getJobID();
        Future<JobStatusIs> failedFuture = cluster.getLeaderGateway(deadline.timeLeft()).ask(new NotifyWhenJobStatus(jobId, JobStatus.FAILED), deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<JobStatusIs>apply(JobStatusIs.class));
        cluster.submitJobDetached(jobGraph);
        JobStatusIs jobStatus = Await.result(failedFuture, deadline.timeLeft());
        assertEquals(JobStatus.FAILED, jobStatus.state());
        // Get the job and check the cause
        JobFound jobFound = Await.result(cluster.getLeaderGateway(deadline.timeLeft()).ask(new JobManagerMessages.RequestJob(jobId), deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<JobFound>apply(JobFound.class)), deadline.timeLeft());
        String failureCause = jobFound.executionGraph().getFailureCauseAsString();
        assertTrue("Not instance of SuppressRestartsException", failureCause.startsWith("org.apache.flink.runtime.execution.SuppressRestartsException"));
        int causedByIndex = failureCause.indexOf("Caused by: ");
        String subFailureCause = failureCause.substring(causedByIndex + "Caused by: ".length());
        assertTrue("Not caused by IllegalStateException", subFailureCause.startsWith("java.lang.IllegalStateException"));
        assertTrue("Exception does not contain registration name", subFailureCause.contains(queryName));
    } finally {
        // Free cluster resources
        if (jobId != null) {
            Future<CancellationSuccess> cancellation = cluster.getLeaderGateway(deadline.timeLeft()).ask(new JobManagerMessages.CancelJob(jobId), deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<CancellationSuccess>apply(CancellationSuccess.class));
            Await.ready(cancellation, deadline.timeLeft());
        }
    }
}

Example 4 with NotifyWhenJobStatus

use of org.apache.flink.runtime.testingUtils.TestingJobManagerMessages.NotifyWhenJobStatus in project flink by apache.

the class JobManagerTest method testCancelWithSavepoint.

@Test
public void testCancelWithSavepoint() throws Exception {
    File defaultSavepointDir = tmpFolder.newFolder();
    FiniteDuration timeout = new FiniteDuration(30, TimeUnit.SECONDS);
    Configuration config = new Configuration();
    config.setString(ConfigConstants.SAVEPOINT_DIRECTORY_KEY, defaultSavepointDir.getAbsolutePath());
    ActorSystem actorSystem = null;
    ActorGateway jobManager = null;
    ActorGateway archiver = null;
    ActorGateway taskManager = null;
    try {
        actorSystem = AkkaUtils.createLocalActorSystem(new Configuration());
        Tuple2<ActorRef, ActorRef> master = JobManager.startJobManagerActors(config, actorSystem, TestingUtils.defaultExecutor(), TestingUtils.defaultExecutor(), Option.apply("jm"), Option.apply("arch"), TestingJobManager.class, TestingMemoryArchivist.class);
        jobManager = new AkkaActorGateway(master._1(), null);
        archiver = new AkkaActorGateway(master._2(), null);
        ActorRef taskManagerRef = TaskManager.startTaskManagerComponentsAndActor(config, ResourceID.generate(), actorSystem, "localhost", Option.apply("tm"), Option.<LeaderRetrievalService>apply(new StandaloneLeaderRetrievalService(jobManager.path())), true, TestingTaskManager.class);
        taskManager = new AkkaActorGateway(taskManagerRef, null);
        // Wait until connected
        Object msg = new TestingTaskManagerMessages.NotifyWhenRegisteredAtJobManager(jobManager.actor());
        Await.ready(taskManager.ask(msg, timeout), timeout);
        // Create job graph
        JobVertex sourceVertex = new JobVertex("Source");
        sourceVertex.setInvokableClass(BlockingStatefulInvokable.class);
        sourceVertex.setParallelism(1);
        JobGraph jobGraph = new JobGraph("TestingJob", sourceVertex);
        JobSnapshottingSettings snapshottingSettings = new JobSnapshottingSettings(Collections.singletonList(sourceVertex.getID()), Collections.singletonList(sourceVertex.getID()), Collections.singletonList(sourceVertex.getID()), 3600000, 3600000, 0, Integer.MAX_VALUE, ExternalizedCheckpointSettings.none(), null, true);
        jobGraph.setSnapshotSettings(snapshottingSettings);
        // Submit job graph
        msg = new JobManagerMessages.SubmitJob(jobGraph, ListeningBehaviour.DETACHED);
        Await.result(jobManager.ask(msg, timeout), timeout);
        // Wait for all tasks to be running
        msg = new TestingJobManagerMessages.WaitForAllVerticesToBeRunning(jobGraph.getJobID());
        Await.result(jobManager.ask(msg, timeout), timeout);
        // Notify when canelled
        msg = new NotifyWhenJobStatus(jobGraph.getJobID(), JobStatus.CANCELED);
        Future<Object> cancelled = jobManager.ask(msg, timeout);
        // Cancel with savepoint
        String savepointPath = null;
        for (int i = 0; i < 10; i++) {
            msg = new JobManagerMessages.CancelJobWithSavepoint(jobGraph.getJobID(), null);
            CancellationResponse cancelResp = (CancellationResponse) Await.result(jobManager.ask(msg, timeout), timeout);
            if (cancelResp instanceof CancellationFailure) {
                CancellationFailure failure = (CancellationFailure) cancelResp;
                if (failure.cause().getMessage().contains(CheckpointDeclineReason.NOT_ALL_REQUIRED_TASKS_RUNNING.message())) {
                    // wait and retry
                    Thread.sleep(200);
                } else {
                    failure.cause().printStackTrace();
                    fail("Failed to cancel job: " + failure.cause().getMessage());
                }
            } else {
                savepointPath = ((CancellationSuccess) cancelResp).savepointPath();
                break;
            }
        }
        // Verify savepoint path
        assertNotEquals("Savepoint not triggered", null, savepointPath);
        // Wait for job status change
        Await.ready(cancelled, timeout);
        File savepointFile = new File(savepointPath);
        assertEquals(true, savepointFile.exists());
    } finally {
        if (actorSystem != null) {
            actorSystem.shutdown();
        }
        if (archiver != null) {
            archiver.actor().tell(PoisonPill.getInstance(), ActorRef.noSender());
        }
        if (jobManager != null) {
            jobManager.actor().tell(PoisonPill.getInstance(), ActorRef.noSender());
        }
        if (taskManager != null) {
            taskManager.actor().tell(PoisonPill.getInstance(), ActorRef.noSender());
        }
    }
}