Search in sources :

Example 1 with JMTMRegistrationRejection

use of org.apache.flink.runtime.jobmaster.JMTMRegistrationRejection in project flink by apache.

the class TaskExecutorTest method testReleaseOfJobResourcesIfJobMasterIsNotCorrect.

/**
 * Tests that the TaskExecutor releases all of its job resources if the JobMaster is not running
 * the specified job. See FLINK-21606.
 */
@Test
public void testReleaseOfJobResourcesIfJobMasterIsNotCorrect() throws Exception {
    final TaskManagerServices taskManagerServices = new TaskManagerServicesBuilder().setTaskSlotTable(TaskSlotUtils.createTaskSlotTable(1)).build();
    final TestingTaskExecutorPartitionTracker taskExecutorPartitionTracker = new TestingTaskExecutorPartitionTracker();
    final CompletableFuture<JobID> jobPartitionsReleaseFuture = new CompletableFuture<>();
    // simulate that we have some partitions tracked
    taskExecutorPartitionTracker.setIsTrackingPartitionsForFunction(ignored -> true);
    taskExecutorPartitionTracker.setStopTrackingAndReleaseAllPartitionsConsumer(jobPartitionsReleaseFuture::complete);
    final TaskExecutor taskExecutor = createTaskExecutor(taskManagerServices, HEARTBEAT_SERVICES, taskExecutorPartitionTracker);
    final TestingJobMasterGateway jobMasterGateway = new TestingJobMasterGatewayBuilder().setRegisterTaskManagerFunction((ignoredJobId, ignoredTaskManagerRegistrationInformation) -> CompletableFuture.completedFuture(new JMTMRegistrationRejection("foobar"))).build();
    rpc.registerGateway(jobMasterGateway.getAddress(), jobMasterGateway);
    final InstanceID registrationId = new InstanceID();
    final OneShotLatch taskExecutorIsRegistered = new OneShotLatch();
    final CompletableFuture<Tuple3<InstanceID, SlotID, AllocationID>> availableSlotFuture = new CompletableFuture<>();
    final TestingResourceManagerGateway resourceManagerGateway = createRmWithTmRegisterAndNotifySlotHooks(registrationId, taskExecutorIsRegistered, availableSlotFuture);
    rpc.registerGateway(resourceManagerGateway.getAddress(), resourceManagerGateway);
    resourceManagerLeaderRetriever.notifyListener(resourceManagerGateway.getAddress(), resourceManagerGateway.getFencingToken().toUUID());
    try {
        taskExecutor.start();
        final TaskExecutorGateway taskExecutorGateway = taskExecutor.getSelfGateway(TaskExecutorGateway.class);
        taskExecutorIsRegistered.await();
        final AllocationID allocationId = new AllocationID();
        final SlotID slotId = new SlotID(taskExecutor.getResourceID(), 0);
        requestSlot(taskExecutorGateway, jobId, allocationId, slotId, ResourceProfile.UNKNOWN, jobMasterGateway.getAddress(), resourceManagerGateway.getFencingToken());
        // The JobManager should reject the registration which should release all job resources
        // on the TaskExecutor
        jobManagerLeaderRetriever.notifyListener(jobMasterGateway.getAddress(), jobMasterGateway.getFencingToken().toUUID());
        // the slot should be freed
        assertThat(availableSlotFuture.get().f1, is(slotId));
        assertThat(availableSlotFuture.get().f2, is(allocationId));
        // all job partitions should be released
        assertThat(jobPartitionsReleaseFuture.get(), is(jobId));
    } finally {
        RpcUtils.terminateRpcEndpoint(taskExecutor, timeout);
    }
}
Also used : Arrays(java.util.Arrays) Tuple3(org.apache.flink.api.java.tuple.Tuple3) TaskSlotUtils.createDefaultTimerService(org.apache.flink.runtime.taskexecutor.slot.TaskSlotUtils.createDefaultTimerService) MemorySize(org.apache.flink.configuration.MemorySize) NetUtils(org.apache.flink.util.NetUtils) InetAddress(java.net.InetAddress) IOManagerAsync(org.apache.flink.runtime.io.disk.iomanager.IOManagerAsync) ResultPartitionID(org.apache.flink.runtime.io.network.partition.ResultPartitionID) SettableLeaderRetrievalService(org.apache.flink.runtime.leaderretrieval.SettableLeaderRetrievalService) TestingFatalErrorHandler(org.apache.flink.runtime.util.TestingFatalErrorHandler) TaskExecutorPartitionTracker(org.apache.flink.runtime.io.network.partition.TaskExecutorPartitionTracker) FunctionUtils(org.apache.flink.util.function.FunctionUtils) Matchers.nullValue(org.hamcrest.Matchers.nullValue) SlotID(org.apache.flink.runtime.clusterframework.types.SlotID) TestingJobMasterGatewayBuilder(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGatewayBuilder) Builder(org.apache.flink.runtime.taskexecutor.TaskSubmissionTestEnvironment.Builder) Matchers.notNullValue(org.hamcrest.Matchers.notNullValue) Failure(org.apache.flink.runtime.registration.RegistrationResponse.Failure) NoOpTaskExecutorBlobService(org.apache.flink.runtime.blob.NoOpTaskExecutorBlobService) TestingTaskExecutorPartitionTracker(org.apache.flink.runtime.io.network.partition.TestingTaskExecutorPartitionTracker) NoOpMetricRegistry(org.apache.flink.runtime.metrics.NoOpMetricRegistry) BlockingQueue(java.util.concurrent.BlockingQueue) Matchers.startsWith(org.hamcrest.Matchers.startsWith) HeartbeatServices(org.apache.flink.runtime.heartbeat.HeartbeatServices) TaskExecutorStateChangelogStoragesManager(org.apache.flink.runtime.state.TaskExecutorStateChangelogStoragesManager) SlotOffer(org.apache.flink.runtime.taskexecutor.slot.SlotOffer) Matchers.instanceOf(org.hamcrest.Matchers.instanceOf) ArrayBlockingQueue(java.util.concurrent.ArrayBlockingQueue) CountDownLatch(java.util.concurrent.CountDownLatch) TimeUtils(org.apache.flink.util.TimeUtils) Matchers.contains(org.hamcrest.Matchers.contains) Matchers.is(org.hamcrest.Matchers.is) Matchers.containsString(org.hamcrest.Matchers.containsString) Time(org.apache.flink.api.common.time.Time) TaskExecutorRegistration(org.apache.flink.runtime.resourcemanager.TaskExecutorRegistration) TaskExecutorPartitionTrackerImpl(org.apache.flink.runtime.io.network.partition.TaskExecutorPartitionTrackerImpl) FlinkException(org.apache.flink.util.FlinkException) TaskSlotTableImpl(org.apache.flink.runtime.taskexecutor.slot.TaskSlotTableImpl) MemoryManager(org.apache.flink.runtime.memory.MemoryManager) TaskSubmissionException(org.apache.flink.runtime.taskexecutor.exceptions.TaskSubmissionException) Callable(java.util.concurrent.Callable) DEFAULT_RESOURCE_PROFILE(org.apache.flink.runtime.taskexecutor.slot.TaskSlotUtils.DEFAULT_RESOURCE_PROFILE) TestingJobMasterGateway(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGateway) TestingTaskSlotTable(org.apache.flink.runtime.taskexecutor.slot.TestingTaskSlotTable) ArrayList(java.util.ArrayList) TaskManagerOptions(org.apache.flink.configuration.TaskManagerOptions) TestingClassLoaderLease(org.apache.flink.runtime.execution.librarycache.TestingClassLoaderLease) FutureUtils(org.apache.flink.util.concurrent.FutureUtils) TestName(org.junit.rules.TestName) ScheduledExecutorService(java.util.concurrent.ScheduledExecutorService) Matchers.hasSize(org.hamcrest.Matchers.hasSize) TriConsumer(org.apache.flink.util.function.TriConsumer) TestFileUtils(org.apache.flink.testutils.TestFileUtils) Before(org.junit.Before) RetryingRegistrationConfiguration(org.apache.flink.runtime.registration.RetryingRegistrationConfiguration) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) SlotNotFoundException(org.apache.flink.runtime.taskexecutor.slot.SlotNotFoundException) TriConsumerWithException(org.apache.flink.util.function.TriConsumerWithException) Assert.assertTrue(org.junit.Assert.assertTrue) Test(org.junit.Test) IOException(java.io.IOException) InstanceID(org.apache.flink.runtime.instance.InstanceID) File(java.io.File) ExecutionException(java.util.concurrent.ExecutionException) Executors(org.apache.flink.util.concurrent.Executors) ThreadSafeTaskSlotTable(org.apache.flink.runtime.taskexecutor.slot.ThreadSafeTaskSlotTable) JobID(org.apache.flink.api.common.JobID) UnregisteredMetricGroups(org.apache.flink.runtime.metrics.groups.UnregisteredMetricGroups) Task(org.apache.flink.runtime.taskmanager.Task) Assert.assertNull(org.junit.Assert.assertNull) UnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.UnresolvedTaskManagerLocation) ArrayDeque(java.util.ArrayDeque) TestingHighAvailabilityServices(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices) Assert.assertEquals(org.junit.Assert.assertEquals) CPUResource(org.apache.flink.api.common.resources.CPUResource) MultiShotLatch(org.apache.flink.core.testutils.MultiShotLatch) Deadline(org.apache.flink.api.common.time.Deadline) ClusterPartitionReport(org.apache.flink.runtime.taskexecutor.partition.ClusterPartitionReport) IntStream.range(java.util.stream.IntStream.range) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) TestingRpcService(org.apache.flink.runtime.rpc.TestingRpcService) ShuffleEnvironment(org.apache.flink.runtime.shuffle.ShuffleEnvironment) IOManager(org.apache.flink.runtime.io.disk.iomanager.IOManager) BlockingNoOpInvokable(org.apache.flink.runtime.testtasks.BlockingNoOpInvokable) TimeoutException(java.util.concurrent.TimeoutException) ExceptionUtils(org.apache.flink.util.ExceptionUtils) Lists(org.apache.flink.shaded.guava30.com.google.common.collect.Lists) Assert.assertThat(org.junit.Assert.assertThat) After(org.junit.After) TestLogger(org.apache.flink.util.TestLogger) Assert.fail(org.junit.Assert.fail) TransientBlobKey(org.apache.flink.runtime.blob.TransientBlobKey) RegistrationTimeoutException(org.apache.flink.runtime.taskexecutor.exceptions.RegistrationTimeoutException) Collection(java.util.Collection) KvStateRegistry(org.apache.flink.runtime.query.KvStateRegistry) ResourceManagerId(org.apache.flink.runtime.resourcemanager.ResourceManagerId) CompletionException(java.util.concurrent.CompletionException) UUID(java.util.UUID) NettyShuffleEnvironment(org.apache.flink.runtime.io.network.NettyShuffleEnvironment) IntermediateDataSetID(org.apache.flink.runtime.jobgraph.IntermediateDataSetID) TaskManagerException(org.apache.flink.runtime.taskexecutor.exceptions.TaskManagerException) Collectors(java.util.stream.Collectors) Acknowledge(org.apache.flink.runtime.messages.Acknowledge) ResourceProfile(org.apache.flink.runtime.clusterframework.types.ResourceProfile) ExecutorUtils(org.apache.flink.util.ExecutorUtils) TaskSlotUtils.createTotalResourceProfile(org.apache.flink.runtime.taskexecutor.slot.TaskSlotUtils.createTotalResourceProfile) List(java.util.List) Matchers.containsInAnyOrder(org.hamcrest.Matchers.containsInAnyOrder) Matchers.equalTo(org.hamcrest.Matchers.equalTo) Queue(java.util.Queue) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) AllocatedSlotInfo(org.apache.flink.runtime.jobmaster.AllocatedSlotInfo) OneShotLatch(org.apache.flink.core.testutils.OneShotLatch) LeaderRetrievalListener(org.apache.flink.runtime.leaderretrieval.LeaderRetrievalListener) TaskInvokable(org.apache.flink.runtime.jobgraph.tasks.TaskInvokable) TaskExecutorLocalStateStoresManager(org.apache.flink.runtime.state.TaskExecutorLocalStateStoresManager) JMTMRegistrationSuccess(org.apache.flink.runtime.jobmaster.JMTMRegistrationSuccess) CompletableFuture(java.util.concurrent.CompletableFuture) TaskDeploymentDescriptorBuilder(org.apache.flink.runtime.deployment.TaskDeploymentDescriptorBuilder) TaskDeploymentDescriptor(org.apache.flink.runtime.deployment.TaskDeploymentDescriptor) NettyShuffleEnvironmentOptions(org.apache.flink.configuration.NettyShuffleEnvironmentOptions) TaskSlotTable(org.apache.flink.runtime.taskexecutor.slot.TaskSlotTable) ExternalResourceInfoProvider(org.apache.flink.runtime.externalresource.ExternalResourceInfoProvider) ConfigConstants(org.apache.flink.configuration.ConfigConstants) ClusterInformation(org.apache.flink.runtime.entrypoint.ClusterInformation) ResourceID(org.apache.flink.runtime.clusterframework.types.ResourceID) Nonnull(javax.annotation.Nonnull) TaskSlotUtils(org.apache.flink.runtime.taskexecutor.slot.TaskSlotUtils) JMTMRegistrationRejection(org.apache.flink.runtime.jobmaster.JMTMRegistrationRejection) Matchers.empty(org.hamcrest.Matchers.empty) NettyShuffleEnvironmentBuilder(org.apache.flink.runtime.io.network.NettyShuffleEnvironmentBuilder) Assert.assertNotNull(org.junit.Assert.assertNotNull) Configuration(org.apache.flink.configuration.Configuration) TestingHeartbeatServices(org.apache.flink.runtime.heartbeat.TestingHeartbeatServices) TaskManagerMetricGroup(org.apache.flink.runtime.metrics.groups.TaskManagerMetricGroup) Reference(org.apache.flink.util.Reference) RpcUtils(org.apache.flink.runtime.rpc.RpcUtils) AllocatedSlotReport(org.apache.flink.runtime.jobmaster.AllocatedSlotReport) TimeUnit(java.util.concurrent.TimeUnit) Consumer(java.util.function.Consumer) ExecutionAttemptID(org.apache.flink.runtime.executiongraph.ExecutionAttemptID) TestingResourceManagerGateway(org.apache.flink.runtime.resourcemanager.utils.TestingResourceManagerGateway) Rule(org.junit.Rule) UnregisteredMetricGroups.createUnregisteredTaskManagerMetricGroup(org.apache.flink.runtime.metrics.groups.UnregisteredMetricGroups.createUnregisteredTaskManagerMetricGroup) CommonTestUtils(org.apache.flink.runtime.testutils.CommonTestUtils) Collections(java.util.Collections) TemporaryFolder(org.junit.rules.TemporaryFolder) RecipientUnreachableException(org.apache.flink.runtime.rpc.exceptions.RecipientUnreachableException) NoOpInvokable(org.apache.flink.runtime.testtasks.NoOpInvokable) InstanceID(org.apache.flink.runtime.instance.InstanceID) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) JMTMRegistrationRejection(org.apache.flink.runtime.jobmaster.JMTMRegistrationRejection) SlotID(org.apache.flink.runtime.clusterframework.types.SlotID) CompletableFuture(java.util.concurrent.CompletableFuture) TestingJobMasterGateway(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGateway) TestingTaskExecutorPartitionTracker(org.apache.flink.runtime.io.network.partition.TestingTaskExecutorPartitionTracker) TestingJobMasterGatewayBuilder(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGatewayBuilder) Tuple3(org.apache.flink.api.java.tuple.Tuple3) TestingResourceManagerGateway(org.apache.flink.runtime.resourcemanager.utils.TestingResourceManagerGateway) OneShotLatch(org.apache.flink.core.testutils.OneShotLatch) JobID(org.apache.flink.api.common.JobID) Test(org.junit.Test)

Example 2 with JMTMRegistrationRejection

use of org.apache.flink.runtime.jobmaster.JMTMRegistrationRejection in project flink by apache.

the class DefaultJobLeaderServiceTest method rejectedJobManagerRegistrationCallsJobLeaderListener.

@Test
public void rejectedJobManagerRegistrationCallsJobLeaderListener() throws Exception {
    final SettableLeaderRetrievalService leaderRetrievalService = new SettableLeaderRetrievalService();
    final TestingHighAvailabilityServices haServices = new TestingHighAvailabilityServicesBuilder().setJobMasterLeaderRetrieverFunction(ignored -> leaderRetrievalService).build();
    final JobID jobId = new JobID();
    final CompletableFuture<JobID> rejectedRegistrationFuture = new CompletableFuture<>();
    final TestingJobLeaderListener testingJobLeaderListener = new TestingJobLeaderListener(ignored -> {
    }, rejectedRegistrationFuture::complete);
    final JobLeaderService jobLeaderService = createAndStartJobLeaderService(haServices, testingJobLeaderListener);
    final TestingJobMasterGateway jobMasterGateway = new TestingJobMasterGatewayBuilder().setRegisterTaskManagerFunction((jobID, taskManagerRegistrationInformation) -> CompletableFuture.completedFuture(new JMTMRegistrationRejection("foobar"))).build();
    rpcServiceResource.getTestingRpcService().registerGateway(jobMasterGateway.getAddress(), jobMasterGateway);
    try {
        jobLeaderService.addJob(jobId, "foobar");
        leaderRetrievalService.notifyListener(jobMasterGateway.getAddress(), jobMasterGateway.getFencingToken().toUUID());
        assertThat(rejectedRegistrationFuture.get(), is(jobId));
    } finally {
        jobLeaderService.stop();
    }
}
Also used : CoreMatchers.is(org.hamcrest.CoreMatchers.is) OneShotLatch(org.apache.flink.core.testutils.OneShotLatch) FlinkException(org.apache.flink.util.FlinkException) TimeoutException(java.util.concurrent.TimeoutException) JMTMRegistrationSuccess(org.apache.flink.runtime.jobmaster.JMTMRegistrationSuccess) CompletableFuture(java.util.concurrent.CompletableFuture) TestingJobMasterGateway(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGateway) JobMasterGateway(org.apache.flink.runtime.jobmaster.JobMasterGateway) Assert.assertThat(org.junit.Assert.assertThat) CheckedThread(org.apache.flink.core.testutils.CheckedThread) SettableLeaderRetrievalService(org.apache.flink.runtime.leaderretrieval.SettableLeaderRetrievalService) TestLogger(org.apache.flink.util.TestLogger) TestingJobMasterGatewayBuilder(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGatewayBuilder) Assert.fail(org.junit.Assert.fail) TestingHighAvailabilityServicesBuilder(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServicesBuilder) RetryingRegistrationConfiguration(org.apache.flink.runtime.registration.RetryingRegistrationConfiguration) JMTMRegistrationRejection(org.apache.flink.runtime.jobmaster.JMTMRegistrationRejection) HighAvailabilityServices(org.apache.flink.runtime.highavailability.HighAvailabilityServices) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) JobMasterId(org.apache.flink.runtime.jobmaster.JobMasterId) Test(org.junit.Test) BlockingQueue(java.util.concurrent.BlockingQueue) UUID(java.util.UUID) TimeUnit(java.util.concurrent.TimeUnit) Consumer(java.util.function.Consumer) ArrayBlockingQueue(java.util.concurrent.ArrayBlockingQueue) CountDownLatch(java.util.concurrent.CountDownLatch) JobID(org.apache.flink.api.common.JobID) Rule(org.junit.Rule) TestingRpcServiceResource(org.apache.flink.runtime.rpc.TestingRpcServiceResource) TestingHighAvailabilityServices(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices) JMTMRegistrationRejection(org.apache.flink.runtime.jobmaster.JMTMRegistrationRejection) TestingHighAvailabilityServicesBuilder(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServicesBuilder) TestingHighAvailabilityServices(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices) CompletableFuture(java.util.concurrent.CompletableFuture) TestingJobMasterGateway(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGateway) SettableLeaderRetrievalService(org.apache.flink.runtime.leaderretrieval.SettableLeaderRetrievalService) TestingJobMasterGatewayBuilder(org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGatewayBuilder) JobID(org.apache.flink.api.common.JobID) Test(org.junit.Test)

Aggregations

UUID (java.util.UUID)2 ArrayBlockingQueue (java.util.concurrent.ArrayBlockingQueue)2 BlockingQueue (java.util.concurrent.BlockingQueue)2 CompletableFuture (java.util.concurrent.CompletableFuture)2 CountDownLatch (java.util.concurrent.CountDownLatch)2 TimeUnit (java.util.concurrent.TimeUnit)2 TimeoutException (java.util.concurrent.TimeoutException)2 Consumer (java.util.function.Consumer)2 JobID (org.apache.flink.api.common.JobID)2 OneShotLatch (org.apache.flink.core.testutils.OneShotLatch)2 TestingHighAvailabilityServices (org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices)2 JMTMRegistrationRejection (org.apache.flink.runtime.jobmaster.JMTMRegistrationRejection)2 JMTMRegistrationSuccess (org.apache.flink.runtime.jobmaster.JMTMRegistrationSuccess)2 TestingJobMasterGateway (org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGateway)2 TestingJobMasterGatewayBuilder (org.apache.flink.runtime.jobmaster.utils.TestingJobMasterGatewayBuilder)2 SettableLeaderRetrievalService (org.apache.flink.runtime.leaderretrieval.SettableLeaderRetrievalService)2 RetryingRegistrationConfiguration (org.apache.flink.runtime.registration.RetryingRegistrationConfiguration)2 LocalUnresolvedTaskManagerLocation (org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation)2 FlinkException (org.apache.flink.util.FlinkException)2 TestLogger (org.apache.flink.util.TestLogger)2