Examples with InstanceID org.apache.flink.runtime.instance.InstanceID used on opensource projects

Example 1 with InstanceID

use of org.apache.flink.runtime.instance.InstanceID in project flink by apache.

the class ResourceManager method notifySlotAvailable.

/**
	 * Notification from a TaskExecutor that a slot has become available
	 * @param resourceManagerLeaderId TaskExecutor's resource manager leader id
	 * @param instanceID TaskExecutor's instance id
	 * @param slotId The slot id of the available slot
	 * @return SlotAvailableReply
	 */
@RpcMethod
public void notifySlotAvailable(final UUID resourceManagerLeaderId, final InstanceID instanceID, final SlotID slotId) {
    if (resourceManagerLeaderId.equals(leaderSessionId)) {
        final ResourceID resourceId = slotId.getResourceID();
        WorkerRegistration<WorkerType> registration = taskExecutors.get(resourceId);
        if (registration != null) {
            InstanceID registrationId = registration.getInstanceID();
            if (registrationId.equals(instanceID)) {
                slotManager.notifySlotAvailable(resourceId, slotId);
            } else {
                log.debug("Invalid registration id for slot available message. This indicates an" + " outdated request.");
            }
        } else {
            log.debug("Could not find registration for resource id {}. Discarding the slot available" + "message {}.", resourceId, slotId);
        }
    } else {
        log.debug("Discarding notify slot available message for slot {}, because the " + "leader id {} did not match the expected leader id {}.", slotId, resourceManagerLeaderId, leaderSessionId);
    }
}

Example 2 with InstanceID

use of org.apache.flink.runtime.instance.InstanceID in project flink by apache.

the class TaskExecutorTest method testJobLeaderDetection.

/**
	 * Tests that a TaskManager detects a job leader for which has reserved slots. Upon detecting
	 * the job leader, it will offer all reserved slots to the JobManager.
	 */
@Test
public void testJobLeaderDetection() throws Exception {
    final JobID jobId = new JobID();
    final TestingSerialRpcService rpc = new TestingSerialRpcService();
    final Configuration configuration = new Configuration();
    final TaskManagerConfiguration taskManagerConfiguration = TaskManagerConfiguration.fromConfiguration(configuration);
    final ResourceID resourceId = new ResourceID("foobar");
    final TaskManagerLocation taskManagerLocation = new TaskManagerLocation(resourceId, InetAddress.getLoopbackAddress(), 1234);
    final TestingHighAvailabilityServices haServices = new TestingHighAvailabilityServices();
    final TimerService<AllocationID> timerService = mock(TimerService.class);
    final TaskSlotTable taskSlotTable = new TaskSlotTable(Arrays.asList(mock(ResourceProfile.class)), timerService);
    final JobManagerTable jobManagerTable = new JobManagerTable();
    final JobLeaderService jobLeaderService = new JobLeaderService(taskManagerLocation);
    final TestingFatalErrorHandler testingFatalErrorHandler = new TestingFatalErrorHandler();
    final TestingLeaderRetrievalService resourceManagerLeaderRetrievalService = new TestingLeaderRetrievalService();
    final TestingLeaderRetrievalService jobManagerLeaderRetrievalService = new TestingLeaderRetrievalService();
    haServices.setResourceManagerLeaderRetriever(resourceManagerLeaderRetrievalService);
    haServices.setJobMasterLeaderRetriever(jobId, jobManagerLeaderRetrievalService);
    final String resourceManagerAddress = "rm";
    final UUID resourceManagerLeaderId = UUID.randomUUID();
    final ResourceManagerGateway resourceManagerGateway = mock(ResourceManagerGateway.class);
    final InstanceID registrationId = new InstanceID();
    when(resourceManagerGateway.registerTaskExecutor(eq(resourceManagerLeaderId), any(String.class), eq(resourceId), any(SlotReport.class), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new TaskExecutorRegistrationSuccess(registrationId, 1000L)));
    final String jobManagerAddress = "jm";
    final UUID jobManagerLeaderId = UUID.randomUUID();
    final ResourceID jmResourceId = new ResourceID(jobManagerAddress);
    final int blobPort = 42;
    final JobMasterGateway jobMasterGateway = mock(JobMasterGateway.class);
    when(jobMasterGateway.registerTaskManager(any(String.class), eq(taskManagerLocation), eq(jobManagerLeaderId), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new JMTMRegistrationSuccess(jmResourceId, blobPort)));
    when(jobMasterGateway.getHostname()).thenReturn(jobManagerAddress);
    rpc.registerGateway(resourceManagerAddress, resourceManagerGateway);
    rpc.registerGateway(jobManagerAddress, jobMasterGateway);
    final AllocationID allocationId = new AllocationID();
    final SlotID slotId = new SlotID(resourceId, 0);
    final SlotOffer slotOffer = new SlotOffer(allocationId, 0, ResourceProfile.UNKNOWN);
    try {
        TaskExecutor taskManager = new TaskExecutor(taskManagerConfiguration, taskManagerLocation, rpc, mock(MemoryManager.class), mock(IOManager.class), mock(NetworkEnvironment.class), haServices, mock(HeartbeatServices.class, RETURNS_MOCKS), mock(MetricRegistry.class), mock(TaskManagerMetricGroup.class), mock(BroadcastVariableManager.class), mock(FileCache.class), taskSlotTable, jobManagerTable, jobLeaderService, testingFatalErrorHandler);
        taskManager.start();
        // tell the task manager about the rm leader
        resourceManagerLeaderRetrievalService.notifyListener(resourceManagerAddress, resourceManagerLeaderId);
        // request slots from the task manager under the given allocation id
        TMSlotRequestReply reply = taskManager.requestSlot(slotId, jobId, allocationId, jobManagerAddress, resourceManagerLeaderId);
        // this is hopefully successful :-)
        assertTrue(reply instanceof TMSlotRequestRegistered);
        // now inform the task manager about the new job leader
        jobManagerLeaderRetrievalService.notifyListener(jobManagerAddress, jobManagerLeaderId);
        // the job leader should get the allocation id offered
        verify(jobMasterGateway).offerSlots(any(ResourceID.class), (Iterable<SlotOffer>) Matchers.argThat(contains(slotOffer)), eq(jobManagerLeaderId), any(Time.class));
        // check if a concurrent error occurred
        testingFatalErrorHandler.rethrowError();
    } finally {
        rpc.stopService();
    }
}

Example 3 with InstanceID

use of org.apache.flink.runtime.instance.InstanceID in project flink by apache.

the class TaskExecutorTest method testSubmitTaskBeforeAcceptSlot.

/**
	 * This tests task executor receive SubmitTask before OfferSlot response.
	 */
@Test
public void testSubmitTaskBeforeAcceptSlot() throws Exception {
    final JobID jobId = new JobID();
    final TestingSerialRpcService rpc = new TestingSerialRpcService();
    final Configuration configuration = new Configuration();
    final TaskManagerConfiguration taskManagerConfiguration = TaskManagerConfiguration.fromConfiguration(configuration);
    final ResourceID resourceId = new ResourceID("foobar");
    final TaskManagerLocation taskManagerLocation = new TaskManagerLocation(resourceId, InetAddress.getLoopbackAddress(), 1234);
    final TestingHighAvailabilityServices haServices = new TestingHighAvailabilityServices();
    final TimerService<AllocationID> timerService = mock(TimerService.class);
    final TaskSlotTable taskSlotTable = new TaskSlotTable(Arrays.asList(mock(ResourceProfile.class), mock(ResourceProfile.class)), timerService);
    final JobManagerTable jobManagerTable = new JobManagerTable();
    final JobLeaderService jobLeaderService = new JobLeaderService(taskManagerLocation);
    final TestingFatalErrorHandler testingFatalErrorHandler = new TestingFatalErrorHandler();
    final String resourceManagerAddress = "rm";
    final UUID resourceManagerLeaderId = UUID.randomUUID();
    final String jobManagerAddress = "jm";
    final UUID jobManagerLeaderId = UUID.randomUUID();
    final LeaderRetrievalService resourceManagerLeaderRetrievalService = new TestingLeaderRetrievalService(resourceManagerAddress, resourceManagerLeaderId);
    final LeaderRetrievalService jobManagerLeaderRetrievalService = new TestingLeaderRetrievalService(jobManagerAddress, jobManagerLeaderId);
    haServices.setResourceManagerLeaderRetriever(resourceManagerLeaderRetrievalService);
    haServices.setJobMasterLeaderRetriever(jobId, jobManagerLeaderRetrievalService);
    final ResourceManagerGateway resourceManagerGateway = mock(ResourceManagerGateway.class);
    final InstanceID registrationId = new InstanceID();
    when(resourceManagerGateway.registerTaskExecutor(eq(resourceManagerLeaderId), any(String.class), eq(resourceId), any(SlotReport.class), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new TaskExecutorRegistrationSuccess(registrationId, 1000L)));
    final ResourceID jmResourceId = new ResourceID(jobManagerAddress);
    final int blobPort = 42;
    final AllocationID allocationId1 = new AllocationID();
    final AllocationID allocationId2 = new AllocationID();
    final SlotOffer offer1 = new SlotOffer(allocationId1, 0, ResourceProfile.UNKNOWN);
    final JobMasterGateway jobMasterGateway = mock(JobMasterGateway.class);
    when(jobMasterGateway.registerTaskManager(any(String.class), eq(taskManagerLocation), eq(jobManagerLeaderId), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new JMTMRegistrationSuccess(jmResourceId, blobPort)));
    when(jobMasterGateway.getHostname()).thenReturn(jobManagerAddress);
    rpc.registerGateway(resourceManagerAddress, resourceManagerGateway);
    rpc.registerGateway(jobManagerAddress, jobMasterGateway);
    final LibraryCacheManager libraryCacheManager = mock(LibraryCacheManager.class);
    when(libraryCacheManager.getClassLoader(eq(jobId))).thenReturn(getClass().getClassLoader());
    final JobManagerConnection jobManagerConnection = new JobManagerConnection(jobId, jmResourceId, jobMasterGateway, jobManagerLeaderId, mock(TaskManagerActions.class), mock(CheckpointResponder.class), libraryCacheManager, mock(ResultPartitionConsumableNotifier.class), mock(PartitionProducerStateChecker.class));
    jobManagerTable.put(jobId, jobManagerConnection);
    try {
        final TaskExecutor taskManager = new TaskExecutor(taskManagerConfiguration, taskManagerLocation, rpc, mock(MemoryManager.class), mock(IOManager.class), mock(NetworkEnvironment.class), haServices, mock(HeartbeatServices.class, RETURNS_MOCKS), mock(MetricRegistry.class), mock(TaskManagerMetricGroup.class), mock(BroadcastVariableManager.class), mock(FileCache.class), taskSlotTable, jobManagerTable, jobLeaderService, testingFatalErrorHandler);
        taskManager.start();
        taskSlotTable.allocateSlot(0, jobId, allocationId1, Time.milliseconds(10000L));
        taskSlotTable.allocateSlot(1, jobId, allocationId2, Time.milliseconds(10000L));
        final JobVertexID jobVertexId = new JobVertexID();
        JobInformation jobInformation = new JobInformation(jobId, name.getMethodName(), new SerializedValue<>(new ExecutionConfig()), new Configuration(), Collections.<BlobKey>emptyList(), Collections.<URL>emptyList());
        TaskInformation taskInformation = new TaskInformation(jobVertexId, "test task", 1, 1, TestInvokable.class.getName(), new Configuration());
        SerializedValue<JobInformation> serializedJobInformation = new SerializedValue<>(jobInformation);
        SerializedValue<TaskInformation> serializedJobVertexInformation = new SerializedValue<>(taskInformation);
        final TaskDeploymentDescriptor tdd = new TaskDeploymentDescriptor(serializedJobInformation, serializedJobVertexInformation, new ExecutionAttemptID(), allocationId1, 0, 0, 0, null, Collections.<ResultPartitionDeploymentDescriptor>emptyList(), Collections.<InputGateDeploymentDescriptor>emptyList());
        CompletableFuture<Iterable<SlotOffer>> offerResultFuture = new FlinkCompletableFuture<>();
        // submit task first and then return acceptance response
        when(jobMasterGateway.offerSlots(any(ResourceID.class), any(Iterable.class), eq(jobManagerLeaderId), any(Time.class))).thenReturn(offerResultFuture);
        // we have to add the job after the TaskExecutor, because otherwise the service has not
        // been properly started. This will also offer the slots to the job master
        jobLeaderService.addJob(jobId, jobManagerAddress);
        verify(jobMasterGateway).offerSlots(any(ResourceID.class), any(Iterable.class), eq(jobManagerLeaderId), any(Time.class));
        // submit the task without having acknowledge the offered slots
        taskManager.submitTask(tdd, jobManagerLeaderId);
        // acknowledge the offered slots
        offerResultFuture.complete(Collections.singleton(offer1));
        verify(resourceManagerGateway).notifySlotAvailable(eq(resourceManagerLeaderId), eq(registrationId), eq(new SlotID(resourceId, 1)));
        assertTrue(taskSlotTable.existsActiveSlot(jobId, allocationId1));
        assertFalse(taskSlotTable.existsActiveSlot(jobId, allocationId2));
        assertTrue(taskSlotTable.isSlotFree(1));
        // check if a concurrent error occurred
        testingFatalErrorHandler.rethrowError();
    } finally {
        rpc.stopService();
    }
}

Example 4 with InstanceID

use of org.apache.flink.runtime.instance.InstanceID in project flink by apache.

the class TaskManagerTest method testSubmitAndExecuteTask.

@Test
public void testSubmitAndExecuteTask() throws IOException {
    new JavaTestKit(system) {

        {
            ActorGateway taskManager = null;
            final ActorGateway jobManager = TestingUtils.createForwardingActor(system, getTestActor(), Option.<String>empty());
            try {
                taskManager = TestingUtils.createTaskManager(system, jobManager, new Configuration(), true, false);
                final ActorGateway tm = taskManager;
                // handle the registration
                new Within(d) {

                    @Override
                    protected void run() {
                        expectMsgClass(RegistrationMessages.RegisterTaskManager.class);
                        final InstanceID iid = new InstanceID();
                        assertEquals(tm.actor(), getLastSender());
                        tm.tell(new RegistrationMessages.AcknowledgeRegistration(iid, 12345), jobManager);
                    }
                };
                final JobID jid = new JobID();
                final JobVertexID vid = new JobVertexID();
                final ExecutionAttemptID eid = new ExecutionAttemptID();
                final SerializedValue<ExecutionConfig> executionConfig = new SerializedValue<>(new ExecutionConfig());
                final TaskDeploymentDescriptor tdd = createTaskDeploymentDescriptor(jid, "TestJob", vid, eid, executionConfig, "TestTask", 7, 2, 7, 0, new Configuration(), new Configuration(), TestInvokableCorrect.class.getName(), Collections.<ResultPartitionDeploymentDescriptor>emptyList(), Collections.<InputGateDeploymentDescriptor>emptyList(), new ArrayList<BlobKey>(), Collections.<URL>emptyList(), 0);
                new Within(d) {

                    @Override
                    protected void run() {
                        tm.tell(new SubmitTask(tdd), jobManager);
                        // TaskManager should acknowledge the submission
                        // heartbeats may be interleaved
                        long deadline = System.currentTimeMillis() + 10000;
                        do {
                            Object message = receiveOne(d);
                            if (message.equals(Acknowledge.get())) {
                                break;
                            }
                        } while (System.currentTimeMillis() < deadline);
                        // task should have switched to running
                        Object toRunning = new TaskMessages.UpdateTaskExecutionState(new TaskExecutionState(jid, eid, ExecutionState.RUNNING));
                        // task should have switched to finished
                        Object toFinished = new TaskMessages.UpdateTaskExecutionState(new TaskExecutionState(jid, eid, ExecutionState.FINISHED));
                        deadline = System.currentTimeMillis() + 10000;
                        do {
                            Object message = receiveOne(d);
                            if (message.equals(toRunning)) {
                                break;
                            } else if (!(message instanceof TaskManagerMessages.Heartbeat)) {
                                fail("Unexpected message: " + message);
                            }
                        } while (System.currentTimeMillis() < deadline);
                        deadline = System.currentTimeMillis() + 10000;
                        do {
                            Object message = receiveOne(d);
                            if (message.equals(toFinished)) {
                                break;
                            } else if (!(message instanceof TaskManagerMessages.Heartbeat)) {
                                fail("Unexpected message: " + message);
                            }
                        } while (System.currentTimeMillis() < deadline);
                    }
                };
            } finally {
                // shut down the actors
                TestingUtils.stopActor(taskManager);
                TestingUtils.stopActor(jobManager);
            }
        }
    };
}

Example 5 with InstanceID

use of org.apache.flink.runtime.instance.InstanceID in project flink by apache.

the class ExecutionGraphTestUtils method getInstance.

public static Instance getInstance(final TaskManagerGateway gateway, final int numberOfSlots) throws Exception {
    ResourceID resourceID = ResourceID.generate();
    HardwareDescription hardwareDescription = new HardwareDescription(4, 2L * 1024 * 1024 * 1024, 1024 * 1024 * 1024, 512 * 1024 * 1024);
    InetAddress address = InetAddress.getByName("127.0.0.1");
    TaskManagerLocation connection = new TaskManagerLocation(resourceID, address, 10001);
    return new Instance(gateway, connection, new InstanceID(), hardwareDescription, numberOfSlots);
}