Search in sources :

Example 36 with TestingTaskExecutorGatewayBuilder

use of org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder in project flink by apache.

the class JobMasterTest method testJobMasterDisconnectsOldTaskExecutorIfNewSessionIsSeen.

@Test
public void testJobMasterDisconnectsOldTaskExecutorIfNewSessionIsSeen() throws Exception {
    final JobMaster jobMaster = new JobMasterBuilder(jobGraph, rpcService).createJobMaster();
    final CompletableFuture<Void> firstTaskExecutorDisconnectedFuture = new CompletableFuture<>();
    final TestingTaskExecutorGateway firstTaskExecutorGateway = new TestingTaskExecutorGatewayBuilder().setAddress("firstTaskExecutor").setDisconnectJobManagerConsumer((jobID, throwable) -> firstTaskExecutorDisconnectedFuture.complete(null)).createTestingTaskExecutorGateway();
    final TestingTaskExecutorGateway secondTaskExecutorGateway = new TestingTaskExecutorGatewayBuilder().setAddress("secondTaskExecutor").createTestingTaskExecutorGateway();
    rpcService.registerGateway(firstTaskExecutorGateway.getAddress(), firstTaskExecutorGateway);
    rpcService.registerGateway(secondTaskExecutorGateway.getAddress(), secondTaskExecutorGateway);
    try {
        jobMaster.start();
        final LocalUnresolvedTaskManagerLocation taskManagerLocation = new LocalUnresolvedTaskManagerLocation();
        final UUID firstTaskManagerSessionId = UUID.randomUUID();
        final CompletableFuture<RegistrationResponse> firstRegistrationResponse = jobMaster.registerTaskManager(jobGraph.getJobID(), TaskManagerRegistrationInformation.create(firstTaskExecutorGateway.getAddress(), taskManagerLocation, firstTaskManagerSessionId), testingTimeout);
        assertThat(firstRegistrationResponse.get(), instanceOf(JMTMRegistrationSuccess.class));
        final UUID secondTaskManagerSessionId = UUID.randomUUID();
        final CompletableFuture<RegistrationResponse> secondRegistrationResponse = jobMaster.registerTaskManager(jobGraph.getJobID(), TaskManagerRegistrationInformation.create(secondTaskExecutorGateway.getAddress(), taskManagerLocation, secondTaskManagerSessionId), testingTimeout);
        assertThat(secondRegistrationResponse.get(), instanceOf(JMTMRegistrationSuccess.class));
        // the first TaskExecutor should be disconnected
        firstTaskExecutorDisconnectedFuture.get();
    } finally {
        RpcUtils.terminateRpcEndpoint(jobMaster, testingTimeout);
    }
}
Also used : TaskManagerGateway(org.apache.flink.runtime.jobmanager.slots.TaskManagerGateway) DefaultSchedulerFactory(org.apache.flink.runtime.scheduler.DefaultSchedulerFactory) TestingTaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGateway) Arrays(java.util.Arrays) Tuple3(org.apache.flink.api.java.tuple.Tuple3) SlotPoolService(org.apache.flink.runtime.jobmaster.slotpool.SlotPoolService) JobMasterBuilder(org.apache.flink.runtime.jobmaster.utils.JobMasterBuilder) RestartStrategyOptions(org.apache.flink.configuration.RestartStrategyOptions) PerJobCheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.PerJobCheckpointRecoveryFactory) ResultPartitionID(org.apache.flink.runtime.io.network.partition.ResultPartitionID) SettableLeaderRetrievalService(org.apache.flink.runtime.leaderretrieval.SettableLeaderRetrievalService) PhysicalSlot(org.apache.flink.runtime.jobmaster.slotpool.PhysicalSlot) TestingFatalErrorHandler(org.apache.flink.runtime.util.TestingFatalErrorHandler) Duration(java.time.Duration) Map(java.util.Map) Matchers.nullValue(org.hamcrest.Matchers.nullValue) CompletedCheckpoint(org.apache.flink.runtime.checkpoint.CompletedCheckpoint) ClassRule(org.junit.ClassRule) SimpleSlotContext(org.apache.flink.runtime.instance.SimpleSlotContext) SlotPoolServiceFactory(org.apache.flink.runtime.jobmaster.slotpool.SlotPoolServiceFactory) AfterClass(org.junit.AfterClass) BlockingQueue(java.util.concurrent.BlockingQueue) JobManagerOptions(org.apache.flink.configuration.JobManagerOptions) Category(org.junit.experimental.categories.Category) HeartbeatServices(org.apache.flink.runtime.heartbeat.HeartbeatServices) 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.is(org.hamcrest.Matchers.is) Time(org.apache.flink.api.common.time.Time) InputSplitSource(org.apache.flink.core.io.InputSplitSource) ResourceManagerGateway(org.apache.flink.runtime.resourcemanager.ResourceManagerGateway) FlinkException(org.apache.flink.util.FlinkException) ComponentMainThreadExecutor(org.apache.flink.runtime.concurrent.ComponentMainThreadExecutor) AccessExecution(org.apache.flink.runtime.executiongraph.AccessExecution) JobStatus(org.apache.flink.api.common.JobStatus) Supplier(java.util.function.Supplier) ArrayList(java.util.ArrayList) DefaultInputSplitAssigner(org.apache.flink.api.common.io.DefaultInputSplitAssigner) FutureUtils(org.apache.flink.util.concurrent.FutureUtils) PartitionProducerDisposedException(org.apache.flink.runtime.jobmanager.PartitionProducerDisposedException) BiConsumer(java.util.function.BiConsumer) Matchers.hasSize(org.hamcrest.Matchers.hasSize) MatcherAssert.assertThat(org.hamcrest.MatcherAssert.assertThat) DistributionPattern(org.apache.flink.runtime.jobgraph.DistributionPattern) Nullable(javax.annotation.Nullable) CheckpointProperties(org.apache.flink.runtime.checkpoint.CheckpointProperties) Before(org.junit.Before) InputSplitAssigner(org.apache.flink.core.io.InputSplitAssigner) Matchers.greaterThanOrEqualTo(org.hamcrest.Matchers.greaterThanOrEqualTo) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) FlinkRuntimeException(org.apache.flink.util.FlinkRuntimeException) InputSplit(org.apache.flink.core.io.InputSplit) ExecutionState(org.apache.flink.runtime.execution.ExecutionState) CheckpointsCleaner(org.apache.flink.runtime.checkpoint.CheckpointsCleaner) Test(org.junit.Test) IOException(java.io.IOException) StreamStateHandle(org.apache.flink.runtime.state.StreamStateHandle) File(java.io.File) ExecutionException(java.util.concurrent.ExecutionException) JobID(org.apache.flink.api.common.JobID) StandaloneCheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.StandaloneCheckpointRecoveryFactory) TestingTaskExecutorGatewayBuilder(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder) UnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.UnresolvedTaskManagerLocation) ArrayDeque(java.util.ArrayDeque) TestingHighAvailabilityServices(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices) SavepointRestoreSettings(org.apache.flink.runtime.jobgraph.SavepointRestoreSettings) CheckpointRetentionPolicy(org.apache.flink.runtime.checkpoint.CheckpointRetentionPolicy) Deadline(org.apache.flink.api.common.time.Deadline) TaskManagerLocation(org.apache.flink.runtime.taskmanager.TaskManagerLocation) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) TestingRpcService(org.apache.flink.runtime.rpc.TestingRpcService) BiFunction(java.util.function.BiFunction) JobGraph(org.apache.flink.runtime.jobgraph.JobGraph) TimeoutException(java.util.concurrent.TimeoutException) ExceptionUtils(org.apache.flink.util.ExceptionUtils) TaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TaskExecutorGateway) TaskExecutorToJobManagerHeartbeatPayload(org.apache.flink.runtime.taskexecutor.TaskExecutorToJobManagerHeartbeatPayload) AggregateFunction(org.apache.flink.api.common.functions.AggregateFunction) InstantiationUtil(org.apache.flink.util.InstantiationUtil) After(org.junit.After) TestLogger(org.apache.flink.util.TestLogger) TestingSchedulerNGFactory(org.apache.flink.runtime.scheduler.TestingSchedulerNGFactory) Assert.fail(org.junit.Assert.fail) BlobServerOptions(org.apache.flink.configuration.BlobServerOptions) CompletedCheckpointStorageLocation(org.apache.flink.runtime.state.CompletedCheckpointStorageLocation) Collection(java.util.Collection) AbstractInvokable(org.apache.flink.runtime.jobgraph.tasks.AbstractInvokable) ResourceManagerId(org.apache.flink.runtime.resourcemanager.ResourceManagerId) UUID(java.util.UUID) IntermediateDataSetID(org.apache.flink.runtime.jobgraph.IntermediateDataSetID) Collectors(java.util.stream.Collectors) SlotInfoWithUtilization(org.apache.flink.runtime.jobmaster.slotpool.SlotInfoWithUtilization) Acknowledge(org.apache.flink.runtime.messages.Acknowledge) ResourceProfile(org.apache.flink.runtime.clusterframework.types.ResourceProfile) Objects(java.util.Objects) TestingUtils(org.apache.flink.testutils.TestingUtils) List(java.util.List) Matchers.containsInAnyOrder(org.hamcrest.Matchers.containsInAnyOrder) ResultPartitionDeploymentDescriptor(org.apache.flink.runtime.deployment.ResultPartitionDeploymentDescriptor) Matchers.equalTo(org.hamcrest.Matchers.equalTo) ExecutionConfig(org.apache.flink.api.common.ExecutionConfig) Optional(java.util.Optional) Queue(java.util.Queue) Matchers.anyOf(org.hamcrest.Matchers.anyOf) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) IntStream(java.util.stream.IntStream) OneShotLatch(org.apache.flink.core.testutils.OneShotLatch) SavepointFormatType(org.apache.flink.core.execution.SavepointFormatType) JobVertex(org.apache.flink.runtime.jobgraph.JobVertex) BeforeClass(org.junit.BeforeClass) AccessExecutionVertex(org.apache.flink.runtime.executiongraph.AccessExecutionVertex) AtomicBoolean(java.util.concurrent.atomic.AtomicBoolean) ResultPartitionType(org.apache.flink.runtime.io.network.partition.ResultPartitionType) HashMap(java.util.HashMap) CompletableFuture(java.util.concurrent.CompletableFuture) RestartStrategies(org.apache.flink.api.common.restartstrategy.RestartStrategies) Function(java.util.function.Function) TaskDeploymentDescriptor(org.apache.flink.runtime.deployment.TaskDeploymentDescriptor) FailoverStrategyFactoryLoader(org.apache.flink.runtime.executiongraph.failover.flip1.FailoverStrategyFactoryLoader) JobVertexID(org.apache.flink.runtime.jobgraph.JobVertexID) TestingJobMasterPartitionTracker(org.apache.flink.runtime.io.network.partition.TestingJobMasterPartitionTracker) FailsWithAdaptiveScheduler(org.apache.flink.testutils.junit.FailsWithAdaptiveScheduler) JobGraphTestUtils(org.apache.flink.runtime.jobgraph.JobGraphTestUtils) TestingSlotPoolServiceBuilder(org.apache.flink.runtime.jobmaster.slotpool.TestingSlotPoolServiceBuilder) ResourceID(org.apache.flink.runtime.clusterframework.types.ResourceID) Nonnull(javax.annotation.Nonnull) StandaloneCompletedCheckpointStore(org.apache.flink.runtime.checkpoint.StandaloneCompletedCheckpointStore) ArchivedExecutionGraph(org.apache.flink.runtime.executiongraph.ArchivedExecutionGraph) SlotPool(org.apache.flink.runtime.jobmaster.slotpool.SlotPool) Matchers.empty(org.hamcrest.Matchers.empty) JobGraphBuilder(org.apache.flink.runtime.jobgraph.JobGraphBuilder) TestingSchedulerNG(org.apache.flink.runtime.scheduler.TestingSchedulerNG) Configuration(org.apache.flink.configuration.Configuration) TestingHeartbeatServices(org.apache.flink.runtime.heartbeat.TestingHeartbeatServices) Matchers(org.hamcrest.Matchers) RpcUtils(org.apache.flink.runtime.rpc.RpcUtils) ExecutionGraphInfo(org.apache.flink.runtime.scheduler.ExecutionGraphInfo) CheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.CheckpointRecoveryFactory) TimeUnit(java.util.concurrent.TimeUnit) ExecutionAttemptID(org.apache.flink.runtime.executiongraph.ExecutionAttemptID) TestingResourceManagerGateway(org.apache.flink.runtime.resourcemanager.utils.TestingResourceManagerGateway) ClosureCleaner(org.apache.flink.api.java.ClosureCleaner) TaskExecutionState(org.apache.flink.runtime.taskmanager.TaskExecutionState) 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) TestingTaskExecutorGatewayBuilder(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder) JobMasterBuilder(org.apache.flink.runtime.jobmaster.utils.JobMasterBuilder) CompletableFuture(java.util.concurrent.CompletableFuture) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) TestingTaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGateway) UUID(java.util.UUID) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) Test(org.junit.Test)

Example 37 with TestingTaskExecutorGatewayBuilder

use of org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder in project flink by apache.

the class JobMasterTest method runJobFailureWhenTaskExecutorTerminatesTest.

private void runJobFailureWhenTaskExecutorTerminatesTest(HeartbeatServices heartbeatServices, BiConsumer<LocalUnresolvedTaskManagerLocation, JobMasterGateway> jobReachedRunningState, BiFunction<JobMasterGateway, ResourceID, BiFunction<ResourceID, AllocatedSlotReport, CompletableFuture<Void>>> heartbeatConsumerFunction) throws Exception {
    final JobGraph jobGraph = JobGraphTestUtils.singleNoOpJobGraph();
    final JobMasterBuilder.TestingOnCompletionActions onCompletionActions = new JobMasterBuilder.TestingOnCompletionActions();
    final JobMaster jobMaster = new JobMasterBuilder(jobGraph, rpcService).withResourceId(jmResourceId).withHighAvailabilityServices(haServices).withHeartbeatServices(heartbeatServices).withOnCompletionActions(onCompletionActions).createJobMaster();
    try {
        jobMaster.start();
        final JobMasterGateway jobMasterGateway = jobMaster.getSelfGateway(JobMasterGateway.class);
        final LocalUnresolvedTaskManagerLocation taskManagerUnresolvedLocation = new LocalUnresolvedTaskManagerLocation();
        final CompletableFuture<ExecutionAttemptID> taskDeploymentFuture = new CompletableFuture<>();
        final TestingTaskExecutorGateway taskExecutorGateway = new TestingTaskExecutorGatewayBuilder().setSubmitTaskConsumer((taskDeploymentDescriptor, jobMasterId) -> {
            taskDeploymentFuture.complete(taskDeploymentDescriptor.getExecutionAttemptId());
            return CompletableFuture.completedFuture(Acknowledge.get());
        }).setHeartbeatJobManagerFunction(heartbeatConsumerFunction.apply(jobMasterGateway, taskManagerUnresolvedLocation.getResourceID())).createTestingTaskExecutorGateway();
        final Collection<SlotOffer> slotOffers = registerSlotsAtJobMaster(1, jobMasterGateway, jobGraph.getJobID(), taskExecutorGateway, taskManagerUnresolvedLocation);
        assertThat(slotOffers, hasSize(1));
        final ExecutionAttemptID executionAttemptId = taskDeploymentFuture.get();
        jobMasterGateway.updateTaskExecutionState(new TaskExecutionState(executionAttemptId, ExecutionState.INITIALIZING)).get();
        jobMasterGateway.updateTaskExecutionState(new TaskExecutionState(executionAttemptId, ExecutionState.RUNNING)).get();
        jobReachedRunningState.accept(taskManagerUnresolvedLocation, jobMasterGateway);
        final ArchivedExecutionGraph archivedExecutionGraph = onCompletionActions.getJobReachedGloballyTerminalStateFuture().get().getArchivedExecutionGraph();
        assertThat(archivedExecutionGraph.getState(), is(JobStatus.FAILED));
    } finally {
        RpcUtils.terminateRpcEndpoint(jobMaster, testingTimeout);
    }
}
Also used : ExecutionAttemptID(org.apache.flink.runtime.executiongraph.ExecutionAttemptID) SlotOffer(org.apache.flink.runtime.taskexecutor.slot.SlotOffer) ArchivedExecutionGraph(org.apache.flink.runtime.executiongraph.ArchivedExecutionGraph) TestingTaskExecutorGatewayBuilder(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder) JobMasterBuilder(org.apache.flink.runtime.jobmaster.utils.JobMasterBuilder) TaskExecutionState(org.apache.flink.runtime.taskmanager.TaskExecutionState) JobGraph(org.apache.flink.runtime.jobgraph.JobGraph) CompletableFuture(java.util.concurrent.CompletableFuture) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) TestingTaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGateway)

Example 38 with TestingTaskExecutorGatewayBuilder

use of org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder in project flink by apache.

the class JobMasterTest method testJobMasterAcknowledgesDuplicateTaskExecutorRegistrations.

@Test
public void testJobMasterAcknowledgesDuplicateTaskExecutorRegistrations() throws Exception {
    final JobMaster jobMaster = new JobMasterBuilder(jobGraph, rpcService).createJobMaster();
    final TestingTaskExecutorGateway testingTaskExecutorGateway = new TestingTaskExecutorGatewayBuilder().createTestingTaskExecutorGateway();
    rpcService.registerGateway(testingTaskExecutorGateway.getAddress(), testingTaskExecutorGateway);
    try {
        jobMaster.start();
        final TaskManagerRegistrationInformation taskManagerRegistrationInformation = TaskManagerRegistrationInformation.create(testingTaskExecutorGateway.getAddress(), new LocalUnresolvedTaskManagerLocation(), UUID.randomUUID());
        final CompletableFuture<RegistrationResponse> firstRegistrationResponse = jobMaster.registerTaskManager(jobGraph.getJobID(), taskManagerRegistrationInformation, testingTimeout);
        final CompletableFuture<RegistrationResponse> secondRegistrationResponse = jobMaster.registerTaskManager(jobGraph.getJobID(), taskManagerRegistrationInformation, testingTimeout);
        assertThat(firstRegistrationResponse.get(), instanceOf(JMTMRegistrationSuccess.class));
        assertThat(secondRegistrationResponse.get(), instanceOf(JMTMRegistrationSuccess.class));
    } finally {
        RpcUtils.terminateRpcEndpoint(jobMaster, testingTimeout);
    }
}
Also used : LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) TestingTaskExecutorGatewayBuilder(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder) TestingTaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGateway) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) JobMasterBuilder(org.apache.flink.runtime.jobmaster.utils.JobMasterBuilder) Test(org.junit.Test)

Example 39 with TestingTaskExecutorGatewayBuilder

use of org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder in project flink by apache.

the class JobMasterTest method testAllocatedSlotReportDoesNotContainStaleInformation.

/**
 * Tests that the {@link AllocatedSlotReport} contains up to date information and not stale
 * information about the allocated slots on the {@link JobMaster}.
 *
 * <p>This is a probabilistic test case which only fails if executed repeatedly without the fix
 * for FLINK-12863.
 */
@Test
public void testAllocatedSlotReportDoesNotContainStaleInformation() throws Exception {
    final CompletableFuture<Void> assertionFuture = new CompletableFuture<>();
    final UnresolvedTaskManagerLocation unresolvedTaskManagerLocation = new LocalUnresolvedTaskManagerLocation();
    final AtomicBoolean terminateHeartbeatVerification = new AtomicBoolean(false);
    final OneShotLatch hasReceivedSlotOffers = new OneShotLatch();
    final TestingTaskExecutorGateway taskExecutorGateway = new TestingTaskExecutorGatewayBuilder().setHeartbeatJobManagerFunction((taskManagerId, allocatedSlotReport) -> {
        try {
            if (hasReceivedSlotOffers.isTriggered()) {
                assertThat(allocatedSlotReport.getAllocatedSlotInfos(), hasSize(1));
            } else {
                assertThat(allocatedSlotReport.getAllocatedSlotInfos(), empty());
            }
        } catch (AssertionError e) {
            assertionFuture.completeExceptionally(e);
        }
        if (terminateHeartbeatVerification.get()) {
            assertionFuture.complete(null);
        }
        return FutureUtils.completedVoidFuture();
    }).createTestingTaskExecutorGateway();
    rpcService.registerGateway(taskExecutorGateway.getAddress(), taskExecutorGateway);
    final JobManagerSharedServices jobManagerSharedServices = new TestingJobManagerSharedServicesBuilder().build();
    final JobGraph jobGraph = JobGraphTestUtils.singleNoOpJobGraph();
    final JobMaster jobMaster = new JobMasterBuilder(jobGraph, rpcService).withHeartbeatServices(new HeartbeatServices(5L, 1000L)).withSlotPoolServiceSchedulerFactory(DefaultSlotPoolServiceSchedulerFactory.create(new TestingSlotPoolFactory(hasReceivedSlotOffers), new DefaultSchedulerFactory())).createJobMaster();
    jobMaster.start();
    try {
        final JobMasterGateway jobMasterGateway = jobMaster.getSelfGateway(JobMasterGateway.class);
        // register task manager will trigger monitor heartbeat target, schedule heartbeat
        // request at interval time
        CompletableFuture<RegistrationResponse> registrationResponse = jobMasterGateway.registerTaskManager(jobGraph.getJobID(), TaskManagerRegistrationInformation.create(taskExecutorGateway.getAddress(), unresolvedTaskManagerLocation, TestingUtils.zeroUUID()), testingTimeout);
        // wait for the completion of the registration
        registrationResponse.get();
        final SlotOffer slotOffer = new SlotOffer(new AllocationID(), 0, ResourceProfile.ANY);
        final CompletableFuture<Collection<SlotOffer>> slotOfferFuture = jobMasterGateway.offerSlots(unresolvedTaskManagerLocation.getResourceID(), Collections.singleton(slotOffer), testingTimeout);
        assertThat(slotOfferFuture.get(), containsInAnyOrder(slotOffer));
        terminateHeartbeatVerification.set(true);
        // make sure that no assertion has been violated
        assertionFuture.get();
    } finally {
        RpcUtils.terminateRpcEndpoint(jobMaster, testingTimeout);
        jobManagerSharedServices.shutdown();
    }
}
Also used : TaskManagerGateway(org.apache.flink.runtime.jobmanager.slots.TaskManagerGateway) DefaultSchedulerFactory(org.apache.flink.runtime.scheduler.DefaultSchedulerFactory) TestingTaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGateway) Arrays(java.util.Arrays) Tuple3(org.apache.flink.api.java.tuple.Tuple3) SlotPoolService(org.apache.flink.runtime.jobmaster.slotpool.SlotPoolService) JobMasterBuilder(org.apache.flink.runtime.jobmaster.utils.JobMasterBuilder) RestartStrategyOptions(org.apache.flink.configuration.RestartStrategyOptions) PerJobCheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.PerJobCheckpointRecoveryFactory) ResultPartitionID(org.apache.flink.runtime.io.network.partition.ResultPartitionID) SettableLeaderRetrievalService(org.apache.flink.runtime.leaderretrieval.SettableLeaderRetrievalService) PhysicalSlot(org.apache.flink.runtime.jobmaster.slotpool.PhysicalSlot) TestingFatalErrorHandler(org.apache.flink.runtime.util.TestingFatalErrorHandler) Duration(java.time.Duration) Map(java.util.Map) Matchers.nullValue(org.hamcrest.Matchers.nullValue) CompletedCheckpoint(org.apache.flink.runtime.checkpoint.CompletedCheckpoint) ClassRule(org.junit.ClassRule) SimpleSlotContext(org.apache.flink.runtime.instance.SimpleSlotContext) SlotPoolServiceFactory(org.apache.flink.runtime.jobmaster.slotpool.SlotPoolServiceFactory) AfterClass(org.junit.AfterClass) BlockingQueue(java.util.concurrent.BlockingQueue) JobManagerOptions(org.apache.flink.configuration.JobManagerOptions) Category(org.junit.experimental.categories.Category) HeartbeatServices(org.apache.flink.runtime.heartbeat.HeartbeatServices) 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.is(org.hamcrest.Matchers.is) Time(org.apache.flink.api.common.time.Time) InputSplitSource(org.apache.flink.core.io.InputSplitSource) ResourceManagerGateway(org.apache.flink.runtime.resourcemanager.ResourceManagerGateway) FlinkException(org.apache.flink.util.FlinkException) ComponentMainThreadExecutor(org.apache.flink.runtime.concurrent.ComponentMainThreadExecutor) AccessExecution(org.apache.flink.runtime.executiongraph.AccessExecution) JobStatus(org.apache.flink.api.common.JobStatus) Supplier(java.util.function.Supplier) ArrayList(java.util.ArrayList) DefaultInputSplitAssigner(org.apache.flink.api.common.io.DefaultInputSplitAssigner) FutureUtils(org.apache.flink.util.concurrent.FutureUtils) PartitionProducerDisposedException(org.apache.flink.runtime.jobmanager.PartitionProducerDisposedException) BiConsumer(java.util.function.BiConsumer) Matchers.hasSize(org.hamcrest.Matchers.hasSize) MatcherAssert.assertThat(org.hamcrest.MatcherAssert.assertThat) DistributionPattern(org.apache.flink.runtime.jobgraph.DistributionPattern) Nullable(javax.annotation.Nullable) CheckpointProperties(org.apache.flink.runtime.checkpoint.CheckpointProperties) Before(org.junit.Before) InputSplitAssigner(org.apache.flink.core.io.InputSplitAssigner) Matchers.greaterThanOrEqualTo(org.hamcrest.Matchers.greaterThanOrEqualTo) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) FlinkRuntimeException(org.apache.flink.util.FlinkRuntimeException) InputSplit(org.apache.flink.core.io.InputSplit) ExecutionState(org.apache.flink.runtime.execution.ExecutionState) CheckpointsCleaner(org.apache.flink.runtime.checkpoint.CheckpointsCleaner) Test(org.junit.Test) IOException(java.io.IOException) StreamStateHandle(org.apache.flink.runtime.state.StreamStateHandle) File(java.io.File) ExecutionException(java.util.concurrent.ExecutionException) JobID(org.apache.flink.api.common.JobID) StandaloneCheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.StandaloneCheckpointRecoveryFactory) TestingTaskExecutorGatewayBuilder(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder) UnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.UnresolvedTaskManagerLocation) ArrayDeque(java.util.ArrayDeque) TestingHighAvailabilityServices(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices) SavepointRestoreSettings(org.apache.flink.runtime.jobgraph.SavepointRestoreSettings) CheckpointRetentionPolicy(org.apache.flink.runtime.checkpoint.CheckpointRetentionPolicy) Deadline(org.apache.flink.api.common.time.Deadline) TaskManagerLocation(org.apache.flink.runtime.taskmanager.TaskManagerLocation) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) TestingRpcService(org.apache.flink.runtime.rpc.TestingRpcService) BiFunction(java.util.function.BiFunction) JobGraph(org.apache.flink.runtime.jobgraph.JobGraph) TimeoutException(java.util.concurrent.TimeoutException) ExceptionUtils(org.apache.flink.util.ExceptionUtils) TaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TaskExecutorGateway) TaskExecutorToJobManagerHeartbeatPayload(org.apache.flink.runtime.taskexecutor.TaskExecutorToJobManagerHeartbeatPayload) AggregateFunction(org.apache.flink.api.common.functions.AggregateFunction) InstantiationUtil(org.apache.flink.util.InstantiationUtil) After(org.junit.After) TestLogger(org.apache.flink.util.TestLogger) TestingSchedulerNGFactory(org.apache.flink.runtime.scheduler.TestingSchedulerNGFactory) Assert.fail(org.junit.Assert.fail) BlobServerOptions(org.apache.flink.configuration.BlobServerOptions) CompletedCheckpointStorageLocation(org.apache.flink.runtime.state.CompletedCheckpointStorageLocation) Collection(java.util.Collection) AbstractInvokable(org.apache.flink.runtime.jobgraph.tasks.AbstractInvokable) ResourceManagerId(org.apache.flink.runtime.resourcemanager.ResourceManagerId) UUID(java.util.UUID) IntermediateDataSetID(org.apache.flink.runtime.jobgraph.IntermediateDataSetID) Collectors(java.util.stream.Collectors) SlotInfoWithUtilization(org.apache.flink.runtime.jobmaster.slotpool.SlotInfoWithUtilization) Acknowledge(org.apache.flink.runtime.messages.Acknowledge) ResourceProfile(org.apache.flink.runtime.clusterframework.types.ResourceProfile) Objects(java.util.Objects) TestingUtils(org.apache.flink.testutils.TestingUtils) List(java.util.List) Matchers.containsInAnyOrder(org.hamcrest.Matchers.containsInAnyOrder) ResultPartitionDeploymentDescriptor(org.apache.flink.runtime.deployment.ResultPartitionDeploymentDescriptor) Matchers.equalTo(org.hamcrest.Matchers.equalTo) ExecutionConfig(org.apache.flink.api.common.ExecutionConfig) Optional(java.util.Optional) Queue(java.util.Queue) Matchers.anyOf(org.hamcrest.Matchers.anyOf) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) IntStream(java.util.stream.IntStream) OneShotLatch(org.apache.flink.core.testutils.OneShotLatch) SavepointFormatType(org.apache.flink.core.execution.SavepointFormatType) JobVertex(org.apache.flink.runtime.jobgraph.JobVertex) BeforeClass(org.junit.BeforeClass) AccessExecutionVertex(org.apache.flink.runtime.executiongraph.AccessExecutionVertex) AtomicBoolean(java.util.concurrent.atomic.AtomicBoolean) ResultPartitionType(org.apache.flink.runtime.io.network.partition.ResultPartitionType) HashMap(java.util.HashMap) CompletableFuture(java.util.concurrent.CompletableFuture) RestartStrategies(org.apache.flink.api.common.restartstrategy.RestartStrategies) Function(java.util.function.Function) TaskDeploymentDescriptor(org.apache.flink.runtime.deployment.TaskDeploymentDescriptor) FailoverStrategyFactoryLoader(org.apache.flink.runtime.executiongraph.failover.flip1.FailoverStrategyFactoryLoader) JobVertexID(org.apache.flink.runtime.jobgraph.JobVertexID) TestingJobMasterPartitionTracker(org.apache.flink.runtime.io.network.partition.TestingJobMasterPartitionTracker) FailsWithAdaptiveScheduler(org.apache.flink.testutils.junit.FailsWithAdaptiveScheduler) JobGraphTestUtils(org.apache.flink.runtime.jobgraph.JobGraphTestUtils) TestingSlotPoolServiceBuilder(org.apache.flink.runtime.jobmaster.slotpool.TestingSlotPoolServiceBuilder) ResourceID(org.apache.flink.runtime.clusterframework.types.ResourceID) Nonnull(javax.annotation.Nonnull) StandaloneCompletedCheckpointStore(org.apache.flink.runtime.checkpoint.StandaloneCompletedCheckpointStore) ArchivedExecutionGraph(org.apache.flink.runtime.executiongraph.ArchivedExecutionGraph) SlotPool(org.apache.flink.runtime.jobmaster.slotpool.SlotPool) Matchers.empty(org.hamcrest.Matchers.empty) JobGraphBuilder(org.apache.flink.runtime.jobgraph.JobGraphBuilder) TestingSchedulerNG(org.apache.flink.runtime.scheduler.TestingSchedulerNG) Configuration(org.apache.flink.configuration.Configuration) TestingHeartbeatServices(org.apache.flink.runtime.heartbeat.TestingHeartbeatServices) Matchers(org.hamcrest.Matchers) RpcUtils(org.apache.flink.runtime.rpc.RpcUtils) ExecutionGraphInfo(org.apache.flink.runtime.scheduler.ExecutionGraphInfo) CheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.CheckpointRecoveryFactory) TimeUnit(java.util.concurrent.TimeUnit) ExecutionAttemptID(org.apache.flink.runtime.executiongraph.ExecutionAttemptID) TestingResourceManagerGateway(org.apache.flink.runtime.resourcemanager.utils.TestingResourceManagerGateway) ClosureCleaner(org.apache.flink.api.java.ClosureCleaner) TaskExecutionState(org.apache.flink.runtime.taskmanager.TaskExecutionState) 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) DefaultSchedulerFactory(org.apache.flink.runtime.scheduler.DefaultSchedulerFactory) TestingTaskExecutorGatewayBuilder(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder) JobMasterBuilder(org.apache.flink.runtime.jobmaster.utils.JobMasterBuilder) CompletableFuture(java.util.concurrent.CompletableFuture) OneShotLatch(org.apache.flink.core.testutils.OneShotLatch) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) HeartbeatServices(org.apache.flink.runtime.heartbeat.HeartbeatServices) TestingHeartbeatServices(org.apache.flink.runtime.heartbeat.TestingHeartbeatServices) SlotOffer(org.apache.flink.runtime.taskexecutor.slot.SlotOffer) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) UnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.UnresolvedTaskManagerLocation) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) AtomicBoolean(java.util.concurrent.atomic.AtomicBoolean) JobGraph(org.apache.flink.runtime.jobgraph.JobGraph) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) Collection(java.util.Collection) TestingTaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGateway) Test(org.junit.Test)

Example 40 with TestingTaskExecutorGatewayBuilder

use of org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder in project flink by apache.

the class JobMasterTest method testRestoringFromSavepoint.

/**
 * Tests that a JobMaster will restore the given JobGraph from its savepoint upon initial
 * submission.
 */
@Test
public void testRestoringFromSavepoint() throws Exception {
    // create savepoint data
    final long savepointId = 42L;
    final File savepointFile = createSavepoint(savepointId);
    // set savepoint settings
    final SavepointRestoreSettings savepointRestoreSettings = SavepointRestoreSettings.forPath(savepointFile.getAbsolutePath(), true);
    final JobGraph jobGraph = createJobGraphWithCheckpointing(savepointRestoreSettings);
    final StandaloneCompletedCheckpointStore completedCheckpointStore = new StandaloneCompletedCheckpointStore(1);
    final CheckpointRecoveryFactory testingCheckpointRecoveryFactory = PerJobCheckpointRecoveryFactory.withoutCheckpointStoreRecovery(maxCheckpoints -> completedCheckpointStore);
    haServices.setCheckpointRecoveryFactory(testingCheckpointRecoveryFactory);
    final JobMaster jobMaster = new JobMasterBuilder(jobGraph, rpcService).withHighAvailabilityServices(haServices).createJobMaster();
    try {
        // we need to start and register the required slots to let the adaptive scheduler
        // restore from the savepoint
        jobMaster.start();
        final OneShotLatch taskSubmitLatch = new OneShotLatch();
        registerSlotsAtJobMaster(1, jobMaster.getSelfGateway(JobMasterGateway.class), jobGraph.getJobID(), new TestingTaskExecutorGatewayBuilder().setSubmitTaskConsumer((taskDeploymentDescriptor, jobMasterId) -> {
            taskSubmitLatch.trigger();
            return CompletableFuture.completedFuture(Acknowledge.get());
        }).createTestingTaskExecutorGateway(), new LocalUnresolvedTaskManagerLocation());
        // wait until a task has submitted because this guarantees that the ExecutionGraph has
        // been created
        taskSubmitLatch.await();
        final CompletedCheckpoint savepointCheckpoint = completedCheckpointStore.getLatestCheckpoint();
        assertThat(savepointCheckpoint, Matchers.notNullValue());
        assertThat(savepointCheckpoint.getCheckpointID(), is(savepointId));
    } finally {
        RpcUtils.terminateRpcEndpoint(jobMaster, testingTimeout);
    }
}
Also used : CompletedCheckpoint(org.apache.flink.runtime.checkpoint.CompletedCheckpoint) TestingTaskExecutorGatewayBuilder(org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder) PerJobCheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.PerJobCheckpointRecoveryFactory) StandaloneCheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.StandaloneCheckpointRecoveryFactory) CheckpointRecoveryFactory(org.apache.flink.runtime.checkpoint.CheckpointRecoveryFactory) JobMasterBuilder(org.apache.flink.runtime.jobmaster.utils.JobMasterBuilder) JobGraph(org.apache.flink.runtime.jobgraph.JobGraph) StandaloneCompletedCheckpointStore(org.apache.flink.runtime.checkpoint.StandaloneCompletedCheckpointStore) OneShotLatch(org.apache.flink.core.testutils.OneShotLatch) LocalUnresolvedTaskManagerLocation(org.apache.flink.runtime.taskmanager.LocalUnresolvedTaskManagerLocation) File(java.io.File) SavepointRestoreSettings(org.apache.flink.runtime.jobgraph.SavepointRestoreSettings) Test(org.junit.Test)

Aggregations

TestingTaskExecutorGatewayBuilder (org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGatewayBuilder)60 Test (org.junit.Test)56 CompletableFuture (java.util.concurrent.CompletableFuture)45 TestingTaskExecutorGateway (org.apache.flink.runtime.taskexecutor.TestingTaskExecutorGateway)45 ResourceID (org.apache.flink.runtime.clusterframework.types.ResourceID)43 JobID (org.apache.flink.api.common.JobID)39 AllocationID (org.apache.flink.runtime.clusterframework.types.AllocationID)37 TaskExecutorGateway (org.apache.flink.runtime.taskexecutor.TaskExecutorGateway)37 Acknowledge (org.apache.flink.runtime.messages.Acknowledge)36 ResourceProfile (org.apache.flink.runtime.clusterframework.types.ResourceProfile)35 Matchers.empty (org.hamcrest.Matchers.empty)30 Collections (java.util.Collections)29 TestLogger (org.apache.flink.util.TestLogger)29 Matchers.equalTo (org.hamcrest.Matchers.equalTo)29 Matchers.is (org.hamcrest.Matchers.is)29 ArrayList (java.util.ArrayList)28 ArrayBlockingQueue (java.util.concurrent.ArrayBlockingQueue)28 ResourceManagerId (org.apache.flink.runtime.resourcemanager.ResourceManagerId)28 TaskExecutorConnection (org.apache.flink.runtime.resourcemanager.registration.TaskExecutorConnection)28 SlotReport (org.apache.flink.runtime.taskexecutor.SlotReport)28