org.apache.flink.runtime.highavailability.RunningJobsRegistry Java Examples

private CompletableFuture<Boolean> tryRunRecoveredJobGraph(JobGraph jobGraph, DispatcherId dispatcherId) throws Exception {
	if (leaderElectionService.hasLeadership(dispatcherId.toUUID())) {
		final JobID jobId = jobGraph.getJobID();
		if (jobManagerRunnerFutures.containsKey(jobId)) {
			// we must not release the job graph lock since it can only be locked once and
			// is currently being executed. Once we support multiple locks, we must release
			// the JobGraph here
			log.debug("Ignore added JobGraph because the job {} is already running.", jobId);
			return CompletableFuture.completedFuture(true);
		} else if (runningJobsRegistry.getJobSchedulingStatus(jobId) != RunningJobsRegistry.JobSchedulingStatus.DONE) {
			return waitForTerminatingJobManager(jobId, jobGraph, this::runJob).thenApply(ignored -> true);
		} else {
			log.debug("Ignore added JobGraph because the job {} has already been completed.", jobId);
		}
	}

	return CompletableFuture.completedFuture(false);
}
 

private CompletableFuture<Boolean> tryRunRecoveredJobGraph(JobGraph jobGraph, DispatcherId dispatcherId) throws Exception {
	if (leaderElectionService.hasLeadership(dispatcherId.toUUID())) {
		final JobID jobId = jobGraph.getJobID();
		if (jobManagerRunnerFutures.containsKey(jobId)) {
			// we must not release the job graph lock since it can only be locked once and
			// is currently being executed. Once we support multiple locks, we must release
			// the JobGraph here
			log.debug("Ignore added JobGraph because the job {} is already running.", jobId);
			return CompletableFuture.completedFuture(true);
		} else if (runningJobsRegistry.getJobSchedulingStatus(jobId) != RunningJobsRegistry.JobSchedulingStatus.DONE) {
			return waitForTerminatingJobManager(jobId, jobGraph, this::runJob).thenApply(ignored -> true);
		} else {
			log.debug("Ignore added JobGraph because the job {} has already been completed.", jobId);
		}
	}

	return CompletableFuture.completedFuture(false);
}
 

private void runCleanupTest(
		Configuration configuration,
		TestingBlobStoreService blobStoreService,
		ThrowingConsumer<ZooKeeperHaServices, Exception> zooKeeperHaServicesConsumer) throws Exception {
	try (ZooKeeperHaServices zooKeeperHaServices = new ZooKeeperHaServices(
		ZooKeeperUtils.startCuratorFramework(configuration),
		Executors.directExecutor(),
		configuration,
		blobStoreService)) {

		// create some Zk services to trigger the generation of paths
		final LeaderRetrievalService resourceManagerLeaderRetriever = zooKeeperHaServices.getResourceManagerLeaderRetriever();
		final LeaderElectionService resourceManagerLeaderElectionService = zooKeeperHaServices.getResourceManagerLeaderElectionService();
		final RunningJobsRegistry runningJobsRegistry = zooKeeperHaServices.getRunningJobsRegistry();

		final TestingListener listener = new TestingListener();
		resourceManagerLeaderRetriever.start(listener);
		resourceManagerLeaderElectionService.start(new TestingContender("foobar", resourceManagerLeaderElectionService));
		final JobID jobId = new JobID();
		runningJobsRegistry.setJobRunning(jobId);

		listener.waitForNewLeader(2000L);

		resourceManagerLeaderRetriever.stop();
		resourceManagerLeaderElectionService.stop();
		runningJobsRegistry.clearJob(jobId);

		zooKeeperHaServicesConsumer.accept(zooKeeperHaServices);
	}
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() throws IOException {
	enter();
	try {
		// IMPORTANT: The registry must NOT place its data in a directory that is
		// cleaned up by these services.
		return new FsNegativeRunningJobsRegistry(flinkFileSystem, workingDirectory);
	}
	finally {
		exit();
	}
}
 

/**
 * Tests that the function of ZookeeperRegistry, setJobRunning(), setJobFinished(), isJobRunning()
 */
@Test
public void testZooKeeperRegistry() throws Exception {
	Configuration configuration = new Configuration();
	configuration.setString(HighAvailabilityOptions.HA_ZOOKEEPER_QUORUM, testingServer.getConnectString());
	configuration.setString(HighAvailabilityOptions.HA_MODE, "zookeeper");

	final HighAvailabilityServices zkHaService = new ZooKeeperHaServices(
			ZooKeeperUtils.startCuratorFramework(configuration),
		Executors.directExecutor(),
		configuration,
		new VoidBlobStore());

	final RunningJobsRegistry zkRegistry = zkHaService.getRunningJobsRegistry();

	try {
		JobID jobID = JobID.generate();
		assertEquals(JobSchedulingStatus.PENDING, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.setJobRunning(jobID);
		assertEquals(JobSchedulingStatus.RUNNING, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.setJobFinished(jobID);
		assertEquals(JobSchedulingStatus.DONE, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.clearJob(jobID);
		assertEquals(JobSchedulingStatus.PENDING, zkRegistry.getJobSchedulingStatus(jobID));
	} finally {
		zkHaService.close();
	}
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() throws Exception {
	synchronized (lock) {
		checkNotShutdown();

		return runningJobsRegistry;
	}
}
 

/**
 * Checks whether the given job has already been submitted or executed.
 *
 * @param jobId identifying the submitted job
 * @return true if the job has already been submitted (is running) or has been executed
 * @throws FlinkException if the job scheduling status cannot be retrieved
 */
private boolean isDuplicateJob(JobID jobId) throws FlinkException {
	final RunningJobsRegistry.JobSchedulingStatus jobSchedulingStatus;

	try {
		jobSchedulingStatus = runningJobsRegistry.getJobSchedulingStatus(jobId);
	} catch (IOException e) {
		throw new FlinkException(String.format("Failed to retrieve job scheduling status for job %s.", jobId), e);
	}

	return jobSchedulingStatus == RunningJobsRegistry.JobSchedulingStatus.DONE || jobManagerRunnerFutures.containsKey(jobId);
}
 

private void runCleanupTest(
		Configuration configuration,
		TestingBlobStoreService blobStoreService,
		ThrowingConsumer<ZooKeeperHaServices, Exception> zooKeeperHaServicesConsumer) throws Exception {
	try (ZooKeeperHaServices zooKeeperHaServices = new ZooKeeperHaServices(
		ZooKeeperUtils.startCuratorFramework(configuration),
		Executors.directExecutor(),
		configuration,
		blobStoreService)) {

		// create some Zk services to trigger the generation of paths
		final LeaderRetrievalService resourceManagerLeaderRetriever = zooKeeperHaServices.getResourceManagerLeaderRetriever();
		final LeaderElectionService resourceManagerLeaderElectionService = zooKeeperHaServices.getResourceManagerLeaderElectionService();
		final RunningJobsRegistry runningJobsRegistry = zooKeeperHaServices.getRunningJobsRegistry();

		final TestingListener listener = new TestingListener();
		resourceManagerLeaderRetriever.start(listener);
		resourceManagerLeaderElectionService.start(new TestingContender("foobar", resourceManagerLeaderElectionService));
		final JobID jobId = new JobID();
		runningJobsRegistry.setJobRunning(jobId);

		listener.waitForNewLeader(2000L);

		resourceManagerLeaderRetriever.stop();
		resourceManagerLeaderElectionService.stop();
		runningJobsRegistry.clearJob(jobId);

		zooKeeperHaServicesConsumer.accept(zooKeeperHaServices);
	}
}
 

/**
 * Tests that the function of ZookeeperRegistry, setJobRunning(), setJobFinished(), isJobRunning()
 */
@Test
public void testZooKeeperRegistry() throws Exception {
	Configuration configuration = new Configuration();
	configuration.setString(HighAvailabilityOptions.HA_ZOOKEEPER_QUORUM, testingServer.getConnectString());
	configuration.setString(HighAvailabilityOptions.HA_MODE, "zookeeper");

	final HighAvailabilityServices zkHaService = new ZooKeeperHaServices(
			ZooKeeperUtils.startCuratorFramework(configuration),
		Executors.directExecutor(),
		configuration,
		new VoidBlobStore());

	final RunningJobsRegistry zkRegistry = zkHaService.getRunningJobsRegistry();

	try {
		JobID jobID = JobID.generate();
		assertEquals(JobSchedulingStatus.PENDING, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.setJobRunning(jobID);
		assertEquals(JobSchedulingStatus.RUNNING, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.setJobFinished(jobID);
		assertEquals(JobSchedulingStatus.DONE, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.clearJob(jobID);
		assertEquals(JobSchedulingStatus.PENDING, zkRegistry.getJobSchedulingStatus(jobID));
	} finally {
		zkHaService.close();
	}
}
 

/**
 * Tests that a reelected Dispatcher can recover jobs.
 */
@Test
public void testJobRecovery() throws Exception {
	dispatcher = createAndStartDispatcher(heartbeatServices, haServices, new ExpectedJobIdJobManagerRunnerFactory(TEST_JOB_ID, createdJobManagerRunnerLatch));

	final DispatcherGateway dispatcherGateway = dispatcher.getSelfGateway(DispatcherGateway.class);

	// elect the initial dispatcher as the leader
	dispatcherLeaderElectionService.isLeader(UUID.randomUUID()).get();

	// submit the job to the current leader
	dispatcherGateway.submitJob(jobGraph, TIMEOUT).get();

	// check that the job has been persisted
	assertThat(submittedJobGraphStore.getJobIds(), contains(jobGraph.getJobID()));

	jobMasterLeaderElectionService.isLeader(UUID.randomUUID()).get();

	assertThat(runningJobsRegistry.getJobSchedulingStatus(jobGraph.getJobID()), is(RunningJobsRegistry.JobSchedulingStatus.RUNNING));

	// revoke the leadership which will stop all currently running jobs
	dispatcherLeaderElectionService.notLeader();

	// re-grant the leadership, this should trigger the job recovery
	dispatcherLeaderElectionService.isLeader(UUID.randomUUID()).get();

	// wait until we have recovered the job
	createdJobManagerRunnerLatch.await();

	// check whether the job has been recovered
	final Collection<JobID> jobIds = dispatcherGateway.listJobs(TIMEOUT).get();

	assertThat(jobIds, hasSize(1));
	assertThat(jobIds, contains(jobGraph.getJobID()));
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() throws Exception {
	synchronized (lock) {
		checkNotShutdown();

		return runningJobsRegistry;
	}
}
 

/**
 * Checks whether the given job has already been submitted or executed.
 *
 * @param jobId identifying the submitted job
 * @return true if the job has already been submitted (is running) or has been executed
 * @throws FlinkException if the job scheduling status cannot be retrieved
 */
private boolean isDuplicateJob(JobID jobId) throws FlinkException {
	final RunningJobsRegistry.JobSchedulingStatus jobSchedulingStatus;

	try {
		jobSchedulingStatus = runningJobsRegistry.getJobSchedulingStatus(jobId);
	} catch (IOException e) {
		throw new FlinkException(String.format("Failed to retrieve job scheduling status for job %s.", jobId), e);
	}

	return jobSchedulingStatus == RunningJobsRegistry.JobSchedulingStatus.DONE || jobManagerRunnerFutures.containsKey(jobId);
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() throws IOException {
	enter();
	try {
		// IMPORTANT: The registry must NOT place its data in a directory that is
		// cleaned up by these services.
		return new FsNegativeRunningJobsRegistry(flinkFileSystem, workingDirectory);
	}
	finally {
		exit();
	}
}
 

private void runCleanupTest(
		Configuration configuration,
		TestingBlobStoreService blobStoreService,
		ThrowingConsumer<ZooKeeperHaServices, Exception> zooKeeperHaServicesConsumer) throws Exception {
	try (ZooKeeperHaServices zooKeeperHaServices = new ZooKeeperHaServices(
		ZooKeeperUtils.startCuratorFramework(configuration),
		Executors.directExecutor(),
		configuration,
		blobStoreService)) {

		// create some Zk services to trigger the generation of paths
		final LeaderRetrievalService resourceManagerLeaderRetriever = zooKeeperHaServices.getResourceManagerLeaderRetriever();
		final LeaderElectionService resourceManagerLeaderElectionService = zooKeeperHaServices.getResourceManagerLeaderElectionService();
		final RunningJobsRegistry runningJobsRegistry = zooKeeperHaServices.getRunningJobsRegistry();

		final TestingListener listener = new TestingListener();
		resourceManagerLeaderRetriever.start(listener);
		resourceManagerLeaderElectionService.start(new TestingContender("foobar", resourceManagerLeaderElectionService));
		final JobID jobId = new JobID();
		runningJobsRegistry.setJobRunning(jobId);

		listener.waitForNewLeader(2000L);

		resourceManagerLeaderRetriever.stop();
		resourceManagerLeaderElectionService.stop();
		runningJobsRegistry.clearJob(jobId);

		zooKeeperHaServicesConsumer.accept(zooKeeperHaServices);
	}
}
 

/**
 * Tests that the function of ZookeeperRegistry, setJobRunning(), setJobFinished(), isJobRunning()
 */
@Test
public void testZooKeeperRegistry() throws Exception {
	Configuration configuration = new Configuration();
	configuration.setString(HighAvailabilityOptions.HA_ZOOKEEPER_QUORUM, testingServer.getConnectString());
	configuration.setString(HighAvailabilityOptions.HA_MODE, "zookeeper");

	final HighAvailabilityServices zkHaService = new ZooKeeperHaServices(
			ZooKeeperUtils.startCuratorFramework(configuration),
		Executors.directExecutor(),
		configuration,
		new VoidBlobStore());

	final RunningJobsRegistry zkRegistry = zkHaService.getRunningJobsRegistry();

	try {
		JobID jobID = JobID.generate();
		assertEquals(JobSchedulingStatus.PENDING, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.setJobRunning(jobID);
		assertEquals(JobSchedulingStatus.RUNNING, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.setJobFinished(jobID);
		assertEquals(JobSchedulingStatus.DONE, zkRegistry.getJobSchedulingStatus(jobID));

		zkRegistry.clearJob(jobID);
		assertEquals(JobSchedulingStatus.PENDING, zkRegistry.getJobSchedulingStatus(jobID));
	} finally {
		zkHaService.close();
	}
}
 

/**
 * Tests that a reelected Dispatcher can recover jobs.
 */
@Test
public void testJobRecovery() throws Exception {
	dispatcher = createAndStartDispatcher(heartbeatServices, haServices, new ExpectedJobIdJobManagerRunnerFactory(TEST_JOB_ID, createdJobManagerRunnerLatch));

	final DispatcherGateway dispatcherGateway = dispatcher.getSelfGateway(DispatcherGateway.class);

	// elect the initial dispatcher as the leader
	dispatcherLeaderElectionService.isLeader(UUID.randomUUID()).get();

	// submit the job to the current leader
	dispatcherGateway.submitJob(jobGraph, TIMEOUT).get();

	// check that the job has been persisted
	assertThat(submittedJobGraphStore.getJobIds(), contains(jobGraph.getJobID()));

	jobMasterLeaderElectionService.isLeader(UUID.randomUUID()).get();

	assertThat(runningJobsRegistry.getJobSchedulingStatus(jobGraph.getJobID()), is(RunningJobsRegistry.JobSchedulingStatus.RUNNING));

	// revoke the leadership which will stop all currently running jobs
	dispatcherLeaderElectionService.notLeader();

	// re-grant the leadership, this should trigger the job recovery
	dispatcherLeaderElectionService.isLeader(UUID.randomUUID()).get();

	// wait until we have recovered the job
	createdJobManagerRunnerLatch.await();

	// check whether the job has been recovered
	final Collection<JobID> jobIds = dispatcherGateway.listJobs(TIMEOUT).get();

	assertThat(jobIds, hasSize(1));
	assertThat(jobIds, contains(jobGraph.getJobID()));
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() throws Exception {
	synchronized (lock) {
		checkNotShutdown();

		return runningJobsRegistry;
	}
}
 

/**
 * Checks whether the given job has already been submitted or executed.
 *
 * @param jobId identifying the submitted job
 * @return true if the job has already been submitted (is running) or has been executed
 * @throws FlinkException if the job scheduling status cannot be retrieved
 */
private boolean isDuplicateJob(JobID jobId) throws FlinkException {
	final RunningJobsRegistry.JobSchedulingStatus jobSchedulingStatus;

	try {
		jobSchedulingStatus = runningJobsRegistry.getJobSchedulingStatus(jobId);
	} catch (IOException e) {
		throw new FlinkException(String.format("Failed to retrieve job scheduling status for job %s.", jobId), e);
	}

	return jobSchedulingStatus == RunningJobsRegistry.JobSchedulingStatus.DONE || jobManagerRunnerFutures.containsKey(jobId);
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() {
	return runningJobsRegistry;
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() {
	return runningJobsRegistry;
}
 

@Override
public RunningJobsRegistry getRunningJobsRegistry() {
	return runningJobsRegistry;
}