org.apache.flink.runtime.util.LeaderRetrievalUtils Java Examples

private static String determineTaskManagerBindAddressByConnectingToResourceManager(
		final Configuration configuration,
		final HighAvailabilityServices haServices) throws LeaderRetrievalException {

	final Time lookupTimeout = Time.milliseconds(AkkaUtils.getLookupTimeout(configuration).toMillis());

	final InetAddress taskManagerAddress = LeaderRetrievalUtils.findConnectingAddress(
		haServices.getResourceManagerLeaderRetriever(),
		lookupTimeout);

	LOG.info("TaskManager will use hostname/address '{}' ({}) for communication.",
		taskManagerAddress.getHostName(), taskManagerAddress.getHostAddress());

	HostBindPolicy bindPolicy = HostBindPolicy.fromString(configuration.getString(TaskManagerOptions.HOST_BIND_POLICY));
	return bindPolicy == HostBindPolicy.IP ? taskManagerAddress.getHostAddress() : taskManagerAddress.getHostName();
}
 

public static HighAvailabilityServices createAvailableOrEmbeddedServices(
	Configuration config,
	Executor executor) throws Exception {
	HighAvailabilityMode highAvailabilityMode = LeaderRetrievalUtils.getRecoveryMode(config);

	switch (highAvailabilityMode) {
		case NONE:
			return new EmbeddedHaServices(executor);

		case ZOOKEEPER:
			BlobStoreService blobStoreService = BlobUtils.createBlobStoreFromConfig(config);

			return new ZooKeeperHaServices(
				ZooKeeperUtils.startCuratorFramework(config),
				executor,
				config,
				blobStoreService);

		case FACTORY_CLASS:
			return createCustomHAServices(config, executor);

		default:
			throw new Exception("High availability mode " + highAvailabilityMode + " is not supported.");
	}
}
 

@Test
public void testTaskExecutorsReconnectToClusterWithLeadershipChange() throws Exception {
	final Deadline deadline = Deadline.fromNow(TESTING_TIMEOUT);
	waitUntilTaskExecutorsHaveConnected(NUM_TMS, deadline);
	highAvailabilityServices.revokeResourceManagerLeadership().get();
	highAvailabilityServices.grantResourceManagerLeadership();

	// wait for the ResourceManager to confirm the leadership
	assertThat(
		LeaderRetrievalUtils.retrieveLeaderConnectionInfo(
			highAvailabilityServices.getResourceManagerLeaderRetriever(),
			TESTING_TIMEOUT).getLeaderSessionId(),
		is(notNullValue()));

	waitUntilTaskExecutorsHaveConnected(NUM_TMS, deadline);
}
 

private static String determineTaskManagerBindAddressByConnectingToResourceManager(
		final Configuration configuration,
		final HighAvailabilityServices haServices) throws LeaderRetrievalException {

	final Duration lookupTimeout = AkkaUtils.getLookupTimeout(configuration);

	final InetAddress taskManagerAddress = LeaderRetrievalUtils.findConnectingAddress(
		haServices.getResourceManagerLeaderRetriever(),
		lookupTimeout);

	LOG.info("TaskManager will use hostname/address '{}' ({}) for communication.",
		taskManagerAddress.getHostName(), taskManagerAddress.getHostAddress());

	HostBindPolicy bindPolicy = HostBindPolicy.fromString(configuration.getString(TaskManagerOptions.HOST_BIND_POLICY));
	return bindPolicy == HostBindPolicy.IP ? taskManagerAddress.getHostAddress() : taskManagerAddress.getHostName();
}
 

private static String determineTaskManagerBindAddressByConnectingToResourceManager(
		final Configuration configuration,
		final HighAvailabilityServices haServices) throws LeaderRetrievalException {

	final Time lookupTimeout = Time.milliseconds(AkkaUtils.getLookupTimeout(configuration).toMillis());

	final InetAddress taskManagerAddress = LeaderRetrievalUtils.findConnectingAddress(
		haServices.getResourceManagerLeaderRetriever(),
		lookupTimeout);

	LOG.info("TaskManager will use hostname/address '{}' ({}) for communication.",
		taskManagerAddress.getHostName(), taskManagerAddress.getHostAddress());

	HostBindPolicy bindPolicy = HostBindPolicy.fromString(configuration.getString(TaskManagerOptions.HOST_BIND_POLICY));
	return bindPolicy == HostBindPolicy.IP ? taskManagerAddress.getHostAddress() : taskManagerAddress.getHostName();
}
 

/**
 * Helper method to wait until the {@link Dispatcher} has set its fencing token.
 *
 * @param rpcService to use to connect to the dispatcher
 * @param haServices high availability services to connect to the dispatcher
 * @return {@link DispatcherGateway}
 * @throws Exception if something goes wrong
 */
static DispatcherGateway retrieveDispatcherGateway(RpcService rpcService, HighAvailabilityServices haServices) throws Exception {
	final LeaderConnectionInfo leaderConnectionInfo = LeaderRetrievalUtils.retrieveLeaderConnectionInfo(haServices.getDispatcherLeaderRetriever(), Time.seconds(10L));

	return rpcService.connect(
		leaderConnectionInfo.getAddress(),
		DispatcherId.fromUuid(leaderConnectionInfo.getLeaderSessionID()),
		DispatcherGateway.class).get();
}
 

@Override
public void run() {
	try {
		result = LeaderRetrievalUtils.findConnectingAddress(
			leaderRetrievalService,
			timeout);
	} catch (Exception e) {
		exception = e;
	}
}
 

/**
 * Tests that the LeaderRetrievalUtils.findConnectingAddress stops trying to find the
 * connecting address if no leader address has been specified. The call should return
 * then InetAddress.getLocalHost().
 */
@Test
public void testTimeoutOfFindConnectingAddress() throws Exception {
	Duration timeout = Duration.ofSeconds(1L);

	LeaderRetrievalService leaderRetrievalService = highAvailabilityServices.getJobManagerLeaderRetriever(HighAvailabilityServices.DEFAULT_JOB_ID);
	InetAddress result = LeaderRetrievalUtils.findConnectingAddress(leaderRetrievalService, timeout);

	assertEquals(InetAddress.getLocalHost(), result);
}
 

/**
 * Helper method to wait until the {@link Dispatcher} has set its fencing token.
 *
 * @param rpcService to use to connect to the dispatcher
 * @param haServices high availability services to connect to the dispatcher
 * @return {@link DispatcherGateway}
 * @throws Exception if something goes wrong
 */
static DispatcherGateway retrieveDispatcherGateway(RpcService rpcService, HighAvailabilityServices haServices) throws Exception {
	final LeaderConnectionInfo leaderConnectionInfo = LeaderRetrievalUtils.retrieveLeaderConnectionInfo(
		haServices.getDispatcherLeaderRetriever(),
		Duration.ofSeconds(10L));

	return rpcService.connect(
		leaderConnectionInfo.getAddress(),
		DispatcherId.fromUuid(leaderConnectionInfo.getLeaderSessionId()),
		DispatcherGateway.class).get();
}
 

@Override
public void run() {
	try {
		result = LeaderRetrievalUtils.findConnectingAddress(
			leaderRetrievalService,
			timeout);
	} catch (Exception e) {
		exception = e;
	}
}
 

/**
 * Tests that the LeaderRetrievalUtils.findConnectingAddress stops trying to find the
 * connecting address if no leader address has been specified. The call should return
 * then InetAddress.getLocalHost().
 */
@Test
public void testTimeoutOfFindConnectingAddress() throws Exception {
	FiniteDuration timeout = new FiniteDuration(1L, TimeUnit.SECONDS);

	LeaderRetrievalService leaderRetrievalService = highAvailabilityServices.getJobManagerLeaderRetriever(HighAvailabilityServices.DEFAULT_JOB_ID);
	InetAddress result = LeaderRetrievalUtils.findConnectingAddress(leaderRetrievalService, timeout);

	assertEquals(InetAddress.getLocalHost(), result);
}
 

@Test
public void testTaskExecutorsReconnectToClusterWithLeadershipChange() throws Exception {
	final Deadline deadline = Deadline.fromNow(TESTING_TIMEOUT);
	waitUntilTaskExecutorsHaveConnected(NUM_TMS, deadline);
	highAvailabilityServices.revokeResourceManagerLeadership().get();
	highAvailabilityServices.grantResourceManagerLeadership();

	// wait for the ResourceManager to confirm the leadership
	assertThat(LeaderRetrievalUtils.retrieveLeaderConnectionInfo(highAvailabilityServices.getResourceManagerLeaderRetriever(), Time.minutes(TESTING_TIMEOUT.toMinutes())).getLeaderSessionID(), is(notNullValue()));

	waitUntilTaskExecutorsHaveConnected(NUM_TMS, deadline);
}
 

/**
 * Helper method to wait until the {@link Dispatcher} has set its fencing token.
 *
 * @param rpcService to use to connect to the dispatcher
 * @param haServices high availability services to connect to the dispatcher
 * @return {@link DispatcherGateway}
 * @throws Exception if something goes wrong
 */
static DispatcherGateway retrieveDispatcherGateway(RpcService rpcService, HighAvailabilityServices haServices) throws Exception {
	final LeaderConnectionInfo leaderConnectionInfo = LeaderRetrievalUtils.retrieveLeaderConnectionInfo(haServices.getDispatcherLeaderRetriever(), Time.seconds(10L));

	return rpcService.connect(
		leaderConnectionInfo.getAddress(),
		DispatcherId.fromUuid(leaderConnectionInfo.getLeaderSessionID()),
		DispatcherGateway.class).get();
}
 

/**
 * Returns the {@link ActorGateway} of the current job manager leader using
 * the {@link LeaderRetrievalService}.
 *
 * @return ActorGateway of the current job manager leader
 * @throws Exception
 */
public ActorGateway getJobManagerGateway() throws FlinkException {
	log.debug("Looking up JobManager");

	try {
		return LeaderRetrievalUtils.retrieveLeaderGateway(
			highAvailabilityServices.getJobManagerLeaderRetriever(HighAvailabilityServices.DEFAULT_JOB_ID),
			actorSystemLoader.get(),
			lookupTimeout);
	} catch (LeaderRetrievalException lre) {
		throw new FlinkException("Could not connect to the leading JobManager. Please check that the " +
			"JobManager is running.", lre);
	}
}
 

/**
 * Creates a new ActorSystem or returns an existing one.
 * @return ActorSystem
 * @throws Exception if the ActorSystem could not be created
 */
@Override
public ActorSystem get() throws FlinkException {

	if (!isLoaded()) {
		// start actor system
		log.info("Starting client actor system.");

		final InetAddress ownHostname;
		try {
			ownHostname = LeaderRetrievalUtils.findConnectingAddress(
				highAvailabilityServices.getJobManagerLeaderRetriever(HighAvailabilityServices.DEFAULT_JOB_ID),
				timeout);
		} catch (LeaderRetrievalException lre) {
			throw new FlinkException("Could not find out our own hostname by connecting to the " +
				"leading JobManager. Please make sure that the Flink cluster has been started.", lre);
		}

		try {
			actorSystem = BootstrapTools.startActorSystem(
				configuration,
				ownHostname.getCanonicalHostName(),
				0,
				log);
		} catch (Exception e) {
			throw new FlinkException("Could not start the ActorSystem lazily.", e);
		}
	}

	return actorSystem;
}
 

@Override
public void run() {
	try {
		result = LeaderRetrievalUtils.findConnectingAddress(
			leaderRetrievalService,
			timeout);
	} catch (Exception e) {
		exception = e;
	}
}
 

/**
 * Tests that the LeaderRetrievalUtils.findConnectingAddress stops trying to find the
 * connecting address if no leader address has been specified. The call should return
 * then InetAddress.getLocalHost().
 */
@Test
public void testTimeoutOfFindConnectingAddress() throws Exception {
	FiniteDuration timeout = new FiniteDuration(1L, TimeUnit.SECONDS);

	LeaderRetrievalService leaderRetrievalService = highAvailabilityServices.getJobManagerLeaderRetriever(HighAvailabilityServices.DEFAULT_JOB_ID);
	InetAddress result = LeaderRetrievalUtils.findConnectingAddress(leaderRetrievalService, timeout);

	assertEquals(InetAddress.getLocalHost(), result);
}
 

@Test
public void testTaskExecutorsReconnectToClusterWithLeadershipChange() throws Exception {
	final Deadline deadline = Deadline.fromNow(TESTING_TIMEOUT);
	waitUntilTaskExecutorsHaveConnected(NUM_TMS, deadline);
	highAvailabilityServices.revokeResourceManagerLeadership().get();
	highAvailabilityServices.grantResourceManagerLeadership();

	// wait for the ResourceManager to confirm the leadership
	assertThat(LeaderRetrievalUtils.retrieveLeaderConnectionInfo(highAvailabilityServices.getResourceManagerLeaderRetriever(), Time.minutes(TESTING_TIMEOUT.toMinutes())).getLeaderSessionID(), is(notNullValue()));

	waitUntilTaskExecutorsHaveConnected(NUM_TMS, deadline);
}
 

@Override
public LeaderConnectionInfo getClusterConnectionInfo() throws LeaderRetrievalException {
	return LeaderRetrievalUtils.retrieveLeaderConnectionInfo(
		highAvailabilityServices.getDispatcherLeaderRetriever(),
		timeout);
}
 

@Override
public LeaderConnectionInfo getClusterConnectionInfo() throws LeaderRetrievalException {
	return LeaderRetrievalUtils.retrieveLeaderConnectionInfo(
		highAvailabilityServices.getDispatcherLeaderRetriever(),
		timeout);
}
 

public static HighAvailabilityServices createHighAvailabilityServices(
	Configuration configuration,
	Executor executor,
	AddressResolution addressResolution) throws Exception {

	HighAvailabilityMode highAvailabilityMode = LeaderRetrievalUtils.getRecoveryMode(configuration);

	switch (highAvailabilityMode) {
		case NONE:
			final Tuple2<String, Integer> hostnamePort = getJobManagerAddress(configuration);

			final String jobManagerRpcUrl = AkkaRpcServiceUtils.getRpcUrl(
				hostnamePort.f0,
				hostnamePort.f1,
				JobMaster.JOB_MANAGER_NAME,
				addressResolution,
				configuration);
			final String resourceManagerRpcUrl = AkkaRpcServiceUtils.getRpcUrl(
				hostnamePort.f0,
				hostnamePort.f1,
				ResourceManager.RESOURCE_MANAGER_NAME,
				addressResolution,
				configuration);
			final String dispatcherRpcUrl = AkkaRpcServiceUtils.getRpcUrl(
				hostnamePort.f0,
				hostnamePort.f1,
				Dispatcher.DISPATCHER_NAME,
				addressResolution,
				configuration);

			final String address = checkNotNull(configuration.getString(RestOptions.ADDRESS),
				"%s must be set",
				RestOptions.ADDRESS.key());
			final int port = configuration.getInteger(RestOptions.PORT);
			final boolean enableSSL = SSLUtils.isRestSSLEnabled(configuration);
			final String protocol = enableSSL ? "https://" : "http://";

			return new StandaloneHaServices(
				resourceManagerRpcUrl,
				dispatcherRpcUrl,
				jobManagerRpcUrl,
				String.format("%s%s:%s", protocol, address, port));
		case ZOOKEEPER:
			BlobStoreService blobStoreService = BlobUtils.createBlobStoreFromConfig(configuration);

			return new ZooKeeperHaServices(
				ZooKeeperUtils.startCuratorFramework(configuration),
				executor,
				configuration,
				blobStoreService);

		case FACTORY_CLASS:
			return createCustomHAServices(configuration, executor);

		default:
			throw new Exception("Recovery mode " + highAvailabilityMode + " is not supported.");
	}
}
 

private void runLeaderRetrievalTest(LeaderElectionService leaderElectionService, LeaderRetrievalService leaderRetrievalService) throws Exception {
	LeaderRetrievalUtils.LeaderConnectionInfoListener leaderRetrievalListener = new LeaderRetrievalUtils.LeaderConnectionInfoListener();
	TestingLeaderContender leaderContender = new TestingLeaderContender();

	leaderRetrievalService.start(leaderRetrievalListener);
	leaderElectionService.start(leaderContender);

	final UUID leaderId = leaderContender.getLeaderSessionFuture().get();

	leaderElectionService.confirmLeadership(leaderId, ADDRESS);

	final LeaderConnectionInfo leaderConnectionInfo = leaderRetrievalListener.getLeaderConnectionInfoFuture().get();

	assertThat(leaderConnectionInfo.getAddress(), is(ADDRESS));
	assertThat(leaderConnectionInfo.getLeaderSessionId(), is(leaderId));
}
 

/**
 * Gets the current cluster connection info (may change in case of a HA setup).
 *
 * @return The the connection info to the leader component of the cluster
 * @throws LeaderRetrievalException if the leader could not be retrieved
 */
public LeaderConnectionInfo getClusterConnectionInfo() throws LeaderRetrievalException {
	return LeaderRetrievalUtils.retrieveLeaderConnectionInfo(
		highAvailabilityServices.getJobManagerLeaderRetriever(HighAvailabilityServices.DEFAULT_JOB_ID),
		timeout);
}
 

/**
 * Gets the current cluster connection info (may change in case of a HA setup).
 *
 * @return The the connection info to the leader component of the cluster
 * @throws LeaderRetrievalException if the leader could not be retrieved
 */
public LeaderConnectionInfo getClusterConnectionInfo() throws LeaderRetrievalException {
	return LeaderRetrievalUtils.retrieveLeaderConnectionInfo(
		highAvailabilityServices.getDispatcherLeaderRetriever(),
		timeout);
}