Java Code Examples for org.apache.flink.runtime.concurrent.FutureUtils#retrySuccessfulWithDelay()

The following examples show how to use org.apache.flink.runtime.concurrent.FutureUtils#retrySuccessfulWithDelay() . You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may check out the related API usage on the sidebar.
Example 1
Source File: AbstractOperatorRestoreTestBase.java    From Flink-CEPplus with Apache License 2.0 6 votes vote down vote up 
private void restoreJob(ClassLoader classLoader, ClusterClient<?> clusterClient, Deadline deadline, String savepointPath) throws Exception {
	JobGraph jobToRestore = createJobGraph(ExecutionMode.RESTORE);
	jobToRestore.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointPath, allowNonRestoredState));

	assertNotNull("Job doesn't have a JobID.", jobToRestore.getJobID());

	clusterClient.submitJob(jobToRestore, classLoader);

	CompletableFuture<JobStatus> jobStatusFuture = FutureUtils.retrySuccessfulWithDelay(
		() -> clusterClient.getJobStatus(jobToRestore.getJobID()),
		Time.milliseconds(50),
		deadline,
		(jobStatus) -> jobStatus == JobStatus.FINISHED,
		TestingUtils.defaultScheduledExecutor());
	assertEquals(
		JobStatus.FINISHED,
		jobStatusFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));
}
Example 2
Source File: AbstractOperatorRestoreTestBase.java    From flink with Apache License 2.0 6 votes vote down vote up 
private void restoreJob(ClassLoader classLoader, ClusterClient<?> clusterClient, Deadline deadline, String savepointPath) throws Exception {
	JobGraph jobToRestore = createJobGraph(ExecutionMode.RESTORE);
	jobToRestore.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointPath, allowNonRestoredState));

	assertNotNull("Job doesn't have a JobID.", jobToRestore.getJobID());

	clusterClient.submitJob(jobToRestore, classLoader);

	CompletableFuture<JobStatus> jobStatusFuture = FutureUtils.retrySuccessfulWithDelay(
		() -> clusterClient.getJobStatus(jobToRestore.getJobID()),
		Time.milliseconds(50),
		deadline,
		(jobStatus) -> jobStatus == JobStatus.FINISHED,
		TestingUtils.defaultScheduledExecutor());
	assertEquals(
		JobStatus.FINISHED,
		jobStatusFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));
}
Example 3
Source File: AbstractOperatorRestoreTestBase.java    From flink with Apache License 2.0 6 votes vote down vote up 
private void restoreJob(ClusterClient<?> clusterClient, Deadline deadline, String savepointPath) throws Exception {
	JobGraph jobToRestore = createJobGraph(ExecutionMode.RESTORE);
	jobToRestore.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointPath, allowNonRestoredState));

	assertNotNull("Job doesn't have a JobID.", jobToRestore.getJobID());

	ClientUtils.submitJob(clusterClient, jobToRestore);

	CompletableFuture<JobStatus> jobStatusFuture = FutureUtils.retrySuccessfulWithDelay(
		() -> clusterClient.getJobStatus(jobToRestore.getJobID()),
		Time.milliseconds(50),
		deadline,
		(jobStatus) -> jobStatus == JobStatus.FINISHED,
		TestingUtils.defaultScheduledExecutor());
	assertEquals(
		JobStatus.FINISHED,
		jobStatusFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));
}
Example 4
Source File: SavepointITCase.java    From Flink-CEPplus with Apache License 2.0 5 votes vote down vote up 
private void restoreJobAndVerifyState(String savepointPath, MiniClusterResourceFactory clusterFactory, int parallelism) throws Exception {
	final JobGraph jobGraph = createJobGraph(parallelism, 0, 1000);
	jobGraph.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointPath));
	final JobID jobId = jobGraph.getJobID();
	StatefulCounter.resetForTest(parallelism);

	MiniClusterWithClientResource cluster = clusterFactory.get();
	cluster.before();
	ClusterClient<?> client = cluster.getClusterClient();

	try {
		client.setDetached(true);
		client.submitJob(jobGraph, SavepointITCase.class.getClassLoader());

		// Await state is restored
		StatefulCounter.getRestoreLatch().await();

		// Await some progress after restore
		StatefulCounter.getProgressLatch().await();

		client.cancel(jobId);

		FutureUtils.retrySuccessfulWithDelay(
			() -> client.getJobStatus(jobId),
			Time.milliseconds(50),
			Deadline.now().plus(Duration.ofSeconds(30)),
			status -> status == JobStatus.CANCELED,
			TestingUtils.defaultScheduledExecutor()
		);

		client.disposeSavepoint(savepointPath)
			.get();

		assertFalse("Savepoint not properly cleaned up.", new File(savepointPath).exists());
	} finally {
		cluster.after();
		StatefulCounter.resetForTest(parallelism);
	}
}
Example 5
Source File: MetricsAvailabilityITCase.java    From Flink-CEPplus with Apache License 2.0 5 votes vote down vote up 
private static <X> X fetchMetric(final SupplierWithException<CompletableFuture<X>, IOException> clientOperation, final Predicate<X> predicate) throws InterruptedException, ExecutionException, TimeoutException {
	final CompletableFuture<X> responseFuture = FutureUtils.retrySuccessfulWithDelay(() -> {
			try {
				return clientOperation.get();
			} catch (IOException e) {
				throw new RuntimeException(e);
			}
		},
		Time.seconds(1),
		Deadline.fromNow(Duration.ofSeconds(5)),
		predicate,
		new ScheduledExecutorServiceAdapter(scheduledExecutorService));

	return responseFuture.get(30, TimeUnit.SECONDS);
}
Example 6
Source File: AbstractQueryableStateTestBase.java    From Flink-CEPplus with Apache License 2.0 5 votes vote down vote up 
@Override
public void close() throws Exception {
	// Free cluster resources
	clusterClient.cancel(jobId);
	// cancel() is non-blocking so do this to make sure the job finished
	CompletableFuture<JobStatus> jobStatusFuture = FutureUtils.retrySuccessfulWithDelay(
		() -> clusterClient.getJobStatus(jobId),
		Time.milliseconds(50),
		deadline,
		(jobStatus) -> jobStatus.equals(JobStatus.CANCELED),
		TestingUtils.defaultScheduledExecutor());
	assertEquals(
		JobStatus.CANCELED,
		jobStatusFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));
}
Example 7
Source File: SavepointITCase.java    From flink with Apache License 2.0 5 votes vote down vote up 
private void restoreJobAndVerifyState(String savepointPath, MiniClusterResourceFactory clusterFactory, int parallelism) throws Exception {
	final JobGraph jobGraph = createJobGraph(parallelism, 0, 1000);
	jobGraph.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointPath));
	final JobID jobId = jobGraph.getJobID();
	StatefulCounter.resetForTest(parallelism);

	MiniClusterWithClientResource cluster = clusterFactory.get();
	cluster.before();
	ClusterClient<?> client = cluster.getClusterClient();

	try {
		client.setDetached(true);
		client.submitJob(jobGraph, SavepointITCase.class.getClassLoader());

		// Await state is restored
		StatefulCounter.getRestoreLatch().await();

		// Await some progress after restore
		StatefulCounter.getProgressLatch().await();

		client.cancel(jobId);

		FutureUtils.retrySuccessfulWithDelay(
			() -> client.getJobStatus(jobId),
			Time.milliseconds(50),
			Deadline.now().plus(Duration.ofSeconds(30)),
			status -> status == JobStatus.CANCELED,
			TestingUtils.defaultScheduledExecutor()
		);

		client.disposeSavepoint(savepointPath)
			.get();

		assertFalse("Savepoint not properly cleaned up.", new File(savepointPath).exists());
	} finally {
		cluster.after();
		StatefulCounter.resetForTest(parallelism);
	}
}
Example 8
Source File: MetricsAvailabilityITCase.java    From flink with Apache License 2.0 5 votes vote down vote up 
private static <X> X fetchMetric(final SupplierWithException<CompletableFuture<X>, IOException> clientOperation, final Predicate<X> predicate) throws InterruptedException, ExecutionException, TimeoutException {
	final CompletableFuture<X> responseFuture = FutureUtils.retrySuccessfulWithDelay(() -> {
			try {
				return clientOperation.get();
			} catch (IOException e) {
				throw new RuntimeException(e);
			}
		},
		Time.seconds(1),
		Deadline.fromNow(Duration.ofSeconds(5)),
		predicate,
		new ScheduledExecutorServiceAdapter(scheduledExecutorService));

	return responseFuture.get(30, TimeUnit.SECONDS);
}
Example 9
Source File: AbstractQueryableStateTestBase.java    From flink with Apache License 2.0 5 votes vote down vote up 
@Override
public void close() throws Exception {
	// Free cluster resources
	clusterClient.cancel(jobId);
	// cancel() is non-blocking so do this to make sure the job finished
	CompletableFuture<JobStatus> jobStatusFuture = FutureUtils.retrySuccessfulWithDelay(
		() -> clusterClient.getJobStatus(jobId),
		Time.milliseconds(50),
		deadline,
		(jobStatus) -> jobStatus.equals(JobStatus.CANCELED),
		TestingUtils.defaultScheduledExecutor());
	assertEquals(
		JobStatus.CANCELED,
		jobStatusFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));
}
Example 10
Source File: MetricsAvailabilityITCase.java    From flink with Apache License 2.0 5 votes vote down vote up 
private static <X> X fetchMetric(final SupplierWithException<CompletableFuture<X>, IOException> clientOperation, final Predicate<X> predicate) throws InterruptedException, ExecutionException, TimeoutException {
	final CompletableFuture<X> responseFuture = FutureUtils.retrySuccessfulWithDelay(() -> {
			try {
				return clientOperation.get();
			} catch (IOException e) {
				throw new RuntimeException(e);
			}
		},
		Time.seconds(1),
		Deadline.fromNow(Duration.ofSeconds(5)),
		predicate,
		new ScheduledExecutorServiceAdapter(scheduledExecutorService));

	return responseFuture.get(30, TimeUnit.SECONDS);
}
Example 11
Source File: AbstractQueryableStateTestBase.java    From flink with Apache License 2.0 5 votes vote down vote up 
@Override
public void close() throws Exception {
	// Free cluster resources
	clusterClient.cancel(jobId).get();
	// cancel() is non-blocking so do this to make sure the job finished
	CompletableFuture<JobStatus> jobStatusFuture = FutureUtils.retrySuccessfulWithDelay(
		() -> clusterClient.getJobStatus(jobId),
		Time.milliseconds(50),
		deadline,
		(jobStatus) -> jobStatus.equals(JobStatus.CANCELED),
		TestingUtils.defaultScheduledExecutor());
	assertEquals(
		JobStatus.CANCELED,
		jobStatusFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));
}
Example 12
Source File: AbstractOperatorRestoreTestBase.java    From Flink-CEPplus with Apache License 2.0 4 votes vote down vote up 
private String migrateJob(ClassLoader classLoader, ClusterClient<?> clusterClient, Deadline deadline) throws Throwable {

		URL savepointResource = AbstractOperatorRestoreTestBase.class.getClassLoader().getResource("operatorstate/" + getMigrationSavepointName());
		if (savepointResource == null) {
			throw new IllegalArgumentException("Savepoint file does not exist.");
		}
		JobGraph jobToMigrate = createJobGraph(ExecutionMode.MIGRATE);
		jobToMigrate.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointResource.getFile()));

		assertNotNull(jobToMigrate.getJobID());

		clusterClient.submitJob(jobToMigrate, classLoader);

		CompletableFuture<JobStatus> jobRunningFuture = FutureUtils.retrySuccessfulWithDelay(
			() -> clusterClient.getJobStatus(jobToMigrate.getJobID()),
			Time.milliseconds(50),
			deadline,
			(jobStatus) -> jobStatus == JobStatus.RUNNING,
			TestingUtils.defaultScheduledExecutor());
		assertEquals(
			JobStatus.RUNNING,
			jobRunningFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));

		// Trigger savepoint
		File targetDirectory = tmpFolder.newFolder();
		String savepointPath = null;

		// FLINK-6918: Retry cancel with savepoint message in case that StreamTasks were not running
		// TODO: The retry logic should be removed once the StreamTask lifecycle has been fixed (see FLINK-4714)
		while (deadline.hasTimeLeft() && savepointPath == null) {
			try {
				savepointPath = clusterClient.cancelWithSavepoint(
					jobToMigrate.getJobID(),
					targetDirectory.getAbsolutePath());
			} catch (Exception e) {
				String exceptionString = ExceptionUtils.stringifyException(e);
				if (!(exceptionString.matches("(.*\n)*.*savepoint for the job .* failed(.*\n)*") // legacy
						|| exceptionString.matches("(.*\n)*.*was not running(.*\n)*")
						|| exceptionString.matches("(.*\n)*.*Not all required tasks are currently running(.*\n)*") // new
						|| exceptionString.matches("(.*\n)*.*Checkpoint was declined \\(tasks not ready\\)(.*\n)*"))) { // new
					throw e;
				}
			}
		}

		assertNotNull("Could not take savepoint.", savepointPath);

		CompletableFuture<JobStatus> jobCanceledFuture = FutureUtils.retrySuccessfulWithDelay(
			() -> clusterClient.getJobStatus(jobToMigrate.getJobID()),
			Time.milliseconds(50),
			deadline,
			(jobStatus) -> jobStatus == JobStatus.CANCELED,
			TestingUtils.defaultScheduledExecutor());
		assertEquals(
			JobStatus.CANCELED,
			jobCanceledFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));

		return savepointPath;
	}
Example 13
Source File: AbstractOperatorRestoreTestBase.java    From flink with Apache License 2.0 4 votes vote down vote up 
private String migrateJob(ClassLoader classLoader, ClusterClient<?> clusterClient, Deadline deadline) throws Throwable {

		URL savepointResource = AbstractOperatorRestoreTestBase.class.getClassLoader().getResource("operatorstate/" + getMigrationSavepointName());
		if (savepointResource == null) {
			throw new IllegalArgumentException("Savepoint file does not exist.");
		}
		JobGraph jobToMigrate = createJobGraph(ExecutionMode.MIGRATE);
		jobToMigrate.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointResource.getFile()));

		assertNotNull(jobToMigrate.getJobID());

		clusterClient.submitJob(jobToMigrate, classLoader);

		CompletableFuture<JobStatus> jobRunningFuture = FutureUtils.retrySuccessfulWithDelay(
			() -> clusterClient.getJobStatus(jobToMigrate.getJobID()),
			Time.milliseconds(50),
			deadline,
			(jobStatus) -> jobStatus == JobStatus.RUNNING,
			TestingUtils.defaultScheduledExecutor());
		assertEquals(
			JobStatus.RUNNING,
			jobRunningFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));

		// Trigger savepoint
		File targetDirectory = tmpFolder.newFolder();
		String savepointPath = null;

		// FLINK-6918: Retry cancel with savepoint message in case that StreamTasks were not running
		// TODO: The retry logic should be removed once the StreamTask lifecycle has been fixed (see FLINK-4714)
		while (deadline.hasTimeLeft() && savepointPath == null) {
			try {
				savepointPath = clusterClient.cancelWithSavepoint(
					jobToMigrate.getJobID(),
					targetDirectory.getAbsolutePath());
			} catch (Exception e) {
				String exceptionString = ExceptionUtils.stringifyException(e);
				if (!PATTERN_CANCEL_WITH_SAVEPOINT_TOLERATED_EXCEPTIONS.matcher(exceptionString).find()) {
					throw e;
				}
			}
		}

		assertNotNull("Could not take savepoint.", savepointPath);

		CompletableFuture<JobStatus> jobCanceledFuture = FutureUtils.retrySuccessfulWithDelay(
			() -> clusterClient.getJobStatus(jobToMigrate.getJobID()),
			Time.milliseconds(50),
			deadline,
			(jobStatus) -> jobStatus == JobStatus.CANCELED,
			TestingUtils.defaultScheduledExecutor());
		assertEquals(
			JobStatus.CANCELED,
			jobCanceledFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));

		return savepointPath;
	}
Example 14
Source File: AbstractOperatorRestoreTestBase.java    From flink with Apache License 2.0 4 votes vote down vote up 
private String migrateJob(ClusterClient<?> clusterClient, Deadline deadline) throws Throwable {

		URL savepointResource = AbstractOperatorRestoreTestBase.class.getClassLoader().getResource("operatorstate/" + getMigrationSavepointName());
		if (savepointResource == null) {
			throw new IllegalArgumentException("Savepoint file does not exist.");
		}
		JobGraph jobToMigrate = createJobGraph(ExecutionMode.MIGRATE);
		jobToMigrate.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointResource.getFile()));

		assertNotNull(jobToMigrate.getJobID());

		ClientUtils.submitJob(clusterClient, jobToMigrate);

		CompletableFuture<JobStatus> jobRunningFuture = FutureUtils.retrySuccessfulWithDelay(
			() -> clusterClient.getJobStatus(jobToMigrate.getJobID()),
			Time.milliseconds(50),
			deadline,
			(jobStatus) -> jobStatus == JobStatus.RUNNING,
			TestingUtils.defaultScheduledExecutor());
		assertEquals(
			JobStatus.RUNNING,
			jobRunningFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));

		// Trigger savepoint
		File targetDirectory = tmpFolder.newFolder();
		String savepointPath = null;

		// FLINK-6918: Retry cancel with savepoint message in case that StreamTasks were not running
		// TODO: The retry logic should be removed once the StreamTask lifecycle has been fixed (see FLINK-4714)
		while (deadline.hasTimeLeft() && savepointPath == null) {
			try {
				savepointPath = clusterClient.cancelWithSavepoint(
					jobToMigrate.getJobID(),
					targetDirectory.getAbsolutePath()).get();
			} catch (Exception e) {
				String exceptionString = ExceptionUtils.stringifyException(e);
				if (!PATTERN_CANCEL_WITH_SAVEPOINT_TOLERATED_EXCEPTIONS.matcher(exceptionString).find()) {
					throw e;
				}
			}
		}

		assertNotNull("Could not take savepoint.", savepointPath);

		CompletableFuture<JobStatus> jobCanceledFuture = FutureUtils.retrySuccessfulWithDelay(
			() -> clusterClient.getJobStatus(jobToMigrate.getJobID()),
			Time.milliseconds(50),
			deadline,
			(jobStatus) -> jobStatus == JobStatus.CANCELED,
			TestingUtils.defaultScheduledExecutor());
		assertEquals(
			JobStatus.CANCELED,
			jobCanceledFuture.get(deadline.timeLeft().toMillis(), TimeUnit.MILLISECONDS));

		return savepointPath;
	}
Example 15
Source File: SavepointITCase.java    From flink with Apache License 2.0 4 votes vote down vote up 
private void restoreJobAndVerifyState(
	String savepointPath,
	MiniClusterResourceFactory clusterFactory,
	int parallelism) throws Exception {
	final JobGraph jobGraph = createJobGraph(parallelism, 0, 1000);
	jobGraph.setSavepointRestoreSettings(SavepointRestoreSettings.forPath(savepointPath, false));
	final JobID jobId = jobGraph.getJobID();
	StatefulCounter.resetForTest(parallelism);

	MiniClusterWithClientResource cluster = clusterFactory.get();
	cluster.before();
	ClusterClient<?> client = cluster.getClusterClient();

	try {
		ClientUtils.submitJob(client, jobGraph);

		// Await state is restored
		StatefulCounter.getRestoreLatch().await();

		// Await some progress after restore
		StatefulCounter.getProgressLatch().await();

		client.cancel(jobId).get();

		FutureUtils.retrySuccessfulWithDelay(
			() -> client.getJobStatus(jobId),
			Time.milliseconds(50),
			Deadline.now().plus(Duration.ofSeconds(30)),
			status -> status == JobStatus.CANCELED,
			TestingUtils.defaultScheduledExecutor()
		);

		client.disposeSavepoint(savepointPath)
			.get();

		assertFalse("Savepoint not properly cleaned up.", new File(savepointPath).exists());
	} finally {
		cluster.after();
		StatefulCounter.resetForTest(parallelism);
	}
}