Search in sources :

Example 1 with TaskExecutorRegistration

use of org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration in project flink by apache.

the class SlotProtocolTest method testSlotAvailableRequest.

/**
	 * Tests whether
	 * 1) a SlotRequest is routed to the SlotManager
	 * 2) a SlotRequest is confirmed
	 * 3) a SlotRequest leads to an allocation of a registered slot
	 * 4) a SlotRequest is routed to the TaskExecutor
	 */
@Test
public void testSlotAvailableRequest() throws Exception {
    final String rmAddress = "/rm1";
    final String jmAddress = "/jm1";
    final String tmAddress = "/tm1";
    final JobID jobID = new JobID();
    testRpcService.registerGateway(jmAddress, mock(JobMasterGateway.class));
    final TestingHighAvailabilityServices testingHaServices = new TestingHighAvailabilityServices();
    final UUID rmLeaderID = UUID.randomUUID();
    final UUID jmLeaderID = UUID.randomUUID();
    TestingLeaderElectionService rmLeaderElectionService = configureHA(testingHaServices, jobID, rmAddress, rmLeaderID, jmAddress, jmLeaderID);
    TaskExecutorGateway taskExecutorGateway = mock(TaskExecutorGateway.class);
    Mockito.when(taskExecutorGateway.requestSlot(any(SlotID.class), any(JobID.class), any(AllocationID.class), any(String.class), any(UUID.class), any(Time.class))).thenReturn(new FlinkCompletableFuture<TMSlotRequestReply>());
    testRpcService.registerGateway(tmAddress, taskExecutorGateway);
    ResourceManagerConfiguration resourceManagerConfiguration = new ResourceManagerConfiguration(Time.seconds(5L), Time.seconds(5L), Time.minutes(5L));
    JobLeaderIdService jobLeaderIdService = new JobLeaderIdService(testingHaServices, testRpcService.getScheduledExecutor(), resourceManagerConfiguration.getJobTimeout());
    TestingSlotManagerFactory slotManagerFactory = new TestingSlotManagerFactory();
    ResourceManager<ResourceID> resourceManager = Mockito.spy(new StandaloneResourceManager(testRpcService, resourceManagerConfiguration, testingHaServices, slotManagerFactory, mock(MetricRegistry.class), jobLeaderIdService, mock(FatalErrorHandler.class)));
    resourceManager.start();
    rmLeaderElectionService.isLeader(rmLeaderID);
    Thread.sleep(1000);
    Future<RegistrationResponse> registrationFuture = resourceManager.registerJobManager(rmLeaderID, jmLeaderID, jmAddress, jobID);
    try {
        registrationFuture.get(5L, TimeUnit.SECONDS);
    } catch (Exception e) {
        Assert.fail("JobManager registration Future didn't become ready.");
    }
    final SlotManager slotManager = slotManagerFactory.slotManager;
    final ResourceID resourceID = ResourceID.generate();
    final AllocationID allocationID = new AllocationID();
    final ResourceProfile resourceProfile = new ResourceProfile(1.0, 100);
    final SlotID slotID = new SlotID(resourceID, 0);
    final SlotStatus slotStatus = new SlotStatus(slotID, resourceProfile);
    final SlotReport slotReport = new SlotReport(Collections.singletonList(slotStatus));
    // register slot at SlotManager
    slotManager.registerTaskExecutor(resourceID, new TaskExecutorRegistration(taskExecutorGateway), slotReport);
    SlotRequest slotRequest = new SlotRequest(jobID, allocationID, resourceProfile);
    RMSlotRequestReply slotRequestReply = resourceManager.requestSlot(jmLeaderID, rmLeaderID, slotRequest);
    // 1) a SlotRequest is routed to the SlotManager
    verify(slotManager).requestSlot(slotRequest);
    // 2) a SlotRequest is confirmed
    Assert.assertEquals(slotRequestReply.getAllocationID(), allocationID);
    // 3) a SlotRequest leads to an allocation of a registered slot
    Assert.assertTrue(slotManager.isAllocated(slotID));
    Assert.assertTrue(slotManager.isAllocated(allocationID));
    // 4) a SlotRequest is routed to the TaskExecutor
    verify(taskExecutorGateway, timeout(5000)).requestSlot(eq(slotID), eq(jobID), eq(allocationID), any(String.class), any(UUID.class), any(Time.class));
}
Also used : TMSlotRequestReply(org.apache.flink.runtime.resourcemanager.messages.taskexecutor.TMSlotRequestReply) TaskExecutorRegistration(org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration) JobLeaderIdService(org.apache.flink.runtime.resourcemanager.JobLeaderIdService) Time(org.apache.flink.api.common.time.Time) StandaloneResourceManager(org.apache.flink.runtime.resourcemanager.StandaloneResourceManager) JobMasterGateway(org.apache.flink.runtime.jobmaster.JobMasterGateway) SlotRequest(org.apache.flink.runtime.resourcemanager.SlotRequest) TestingHighAvailabilityServices(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices) ResourceID(org.apache.flink.runtime.clusterframework.types.ResourceID) UUID(java.util.UUID) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) ResourceProfile(org.apache.flink.runtime.clusterframework.types.ResourceProfile) TestingLeaderElectionService(org.apache.flink.runtime.leaderelection.TestingLeaderElectionService) SlotStatus(org.apache.flink.runtime.taskexecutor.SlotStatus) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) SlotReport(org.apache.flink.runtime.taskexecutor.SlotReport) RMSlotRequestReply(org.apache.flink.runtime.resourcemanager.messages.jobmanager.RMSlotRequestReply) TaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TaskExecutorGateway) ResourceManagerConfiguration(org.apache.flink.runtime.resourcemanager.ResourceManagerConfiguration) SlotID(org.apache.flink.runtime.clusterframework.types.SlotID) TestingSlotManager(org.apache.flink.runtime.resourcemanager.TestingSlotManager) JobID(org.apache.flink.api.common.JobID) Test(org.junit.Test)

Example 2 with TaskExecutorRegistration

use of org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration in project flink by apache.

the class SlotManagerTest method setUp.

@BeforeClass
public static void setUp() {
    taskExecutorRegistration = Mockito.mock(TaskExecutorRegistration.class);
    TaskExecutorGateway gateway = Mockito.mock(TaskExecutorGateway.class);
    Mockito.when(taskExecutorRegistration.getTaskExecutorGateway()).thenReturn(gateway);
    Mockito.when(gateway.requestSlot(any(SlotID.class), any(JobID.class), any(AllocationID.class), any(String.class), any(UUID.class), any(Time.class))).thenReturn(new FlinkCompletableFuture<TMSlotRequestReply>());
}
Also used : SlotID(org.apache.flink.runtime.clusterframework.types.SlotID) TMSlotRequestReply(org.apache.flink.runtime.resourcemanager.messages.taskexecutor.TMSlotRequestReply) TaskExecutorRegistration(org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) TaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TaskExecutorGateway) Time(org.apache.flink.api.common.time.Time) UUID(java.util.UUID) JobID(org.apache.flink.api.common.JobID) BeforeClass(org.junit.BeforeClass)

Example 3 with TaskExecutorRegistration

use of org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration in project flink by apache.

the class SlotManager method sendSlotRequest.

private void sendSlotRequest(final ResourceSlot freeSlot, final SlotRequest slotRequest) {
    final AllocationID allocationID = slotRequest.getAllocationId();
    final TaskExecutorRegistration registration = freeSlot.getTaskExecutorRegistration();
    final Future<TMSlotRequestReply> slotRequestReplyFuture = registration.getTaskExecutorGateway().requestSlot(freeSlot.getSlotId(), slotRequest.getJobId(), allocationID, // TODO: set proper JM address
    "foobar", rmServices.getLeaderID(), timeout);
    slotRequestReplyFuture.handleAsync(new BiFunction<TMSlotRequestReply, Throwable, Void>() {

        @Override
        public Void apply(TMSlotRequestReply slotRequestReply, Throwable throwable) {
            TaskExecutorRegistration current = taskManagers.get(slotRequestReply.getResourceID());
            if (current != null && current.getInstanceID().equals(slotRequestReply.getInstanceID())) {
                if (throwable != null || slotRequestReply instanceof TMSlotRequestRejected) {
                    handleSlotRequestFailedAtTaskManager(slotRequest, freeSlot.getSlotId());
                } else {
                    LOG.debug("Successfully registered slot {} ", freeSlot.getSlotId());
                }
            } else {
                LOG.debug("Discarding message from obsolete TaskExecutor with InstanceID {}", slotRequestReply.getInstanceID());
            }
            return null;
        }
    }, rmServices.getMainThreadExecutor());
}
Also used : TMSlotRequestRejected(org.apache.flink.runtime.resourcemanager.messages.taskexecutor.TMSlotRequestRejected) TMSlotRequestReply(org.apache.flink.runtime.resourcemanager.messages.taskexecutor.TMSlotRequestReply) TaskExecutorRegistration(org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID)

Example 4 with TaskExecutorRegistration

use of org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration in project flink by apache.

the class SlotManager method registerTaskExecutor.

/**
	 * Registers a TaskExecutor
	 * @param resourceID TaskExecutor's ResourceID
	 * @param registration TaskExecutor's registration
	 * @param slotReport TaskExecutor's free and allocated slots
	 */
public void registerTaskExecutor(ResourceID resourceID, TaskExecutorRegistration registration, SlotReport slotReport) {
    if (taskManagers.get(resourceID) != null) {
        notifyTaskManagerFailure(resourceID);
    }
    this.taskManagers.put(resourceID, registration);
    for (SlotStatus slotStatus : slotReport.getSlotsStatus()) {
        final SlotID slotId = slotStatus.getSlotID();
        final TaskExecutorRegistration taskExecutorRegistration = taskManagers.get(slotId.getResourceID());
        if (taskExecutorRegistration == null) {
            LOG.info("Received SlotStatus but ResourceID {} is unknown to the SlotManager", slotId.getResourceID());
            return;
        }
        final ResourceSlot slot = new ResourceSlot(slotId, slotStatus.getProfiler(), taskExecutorRegistration);
        registerNewSlot(slot);
        LOG.info("New slot appeared, SlotID:{}, AllocationID:{}", slotId, slotStatus.getAllocationID());
        if (slotStatus.getAllocationID() != null) {
            // slot in use, record this in bookkeeping
            allocationMap.addAllocation(slotId, slotStatus.getAllocationID());
        } else {
            handleFreeSlot(slot);
        }
    }
}
Also used : SlotID(org.apache.flink.runtime.clusterframework.types.SlotID) TaskExecutorRegistration(org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration) SlotStatus(org.apache.flink.runtime.taskexecutor.SlotStatus) ResourceSlot(org.apache.flink.runtime.clusterframework.types.ResourceSlot)

Example 5 with TaskExecutorRegistration

use of org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration in project flink by apache.

the class SlotProtocolTest method testSlotsUnavailableRequest.

/**
	 * Tests whether
	 * 1) SlotRequest is routed to the SlotManager
	 * 2) SlotRequest is confirmed
	 * 3) SlotRequest leads to a container allocation
	 * 4) Slot becomes available and TaskExecutor gets a SlotRequest
	 */
@Test
public void testSlotsUnavailableRequest() throws Exception {
    final String rmAddress = "/rm1";
    final String jmAddress = "/jm1";
    final JobID jobID = new JobID();
    testRpcService.registerGateway(jmAddress, mock(JobMasterGateway.class));
    final TestingHighAvailabilityServices testingHaServices = new TestingHighAvailabilityServices();
    final UUID rmLeaderID = UUID.randomUUID();
    final UUID jmLeaderID = UUID.randomUUID();
    TestingLeaderElectionService rmLeaderElectionService = configureHA(testingHaServices, jobID, rmAddress, rmLeaderID, jmAddress, jmLeaderID);
    ResourceManagerConfiguration resourceManagerConfiguration = new ResourceManagerConfiguration(Time.seconds(5L), Time.seconds(5L), Time.minutes(5L));
    JobLeaderIdService jobLeaderIdService = new JobLeaderIdService(testingHaServices, testRpcService.getScheduledExecutor(), resourceManagerConfiguration.getJobTimeout());
    final TestingSlotManagerFactory slotManagerFactory = new TestingSlotManagerFactory();
    SpiedResourceManager resourceManager = new SpiedResourceManager(testRpcService, resourceManagerConfiguration, testingHaServices, slotManagerFactory, mock(MetricRegistry.class), jobLeaderIdService, mock(FatalErrorHandler.class));
    resourceManager.start();
    rmLeaderElectionService.isLeader(rmLeaderID);
    Future<RegistrationResponse> registrationFuture = resourceManager.registerJobManager(rmLeaderID, jmLeaderID, jmAddress, jobID);
    try {
        registrationFuture.get(5, TimeUnit.SECONDS);
    } catch (Exception e) {
        Assert.fail("JobManager registration Future didn't become ready.");
    }
    final SlotManager slotManager = slotManagerFactory.slotManager;
    final AllocationID allocationID = new AllocationID();
    final ResourceProfile resourceProfile = new ResourceProfile(1.0, 100);
    SlotRequest slotRequest = new SlotRequest(jobID, allocationID, resourceProfile);
    RMSlotRequestReply slotRequestReply = resourceManager.requestSlot(jmLeaderID, rmLeaderID, slotRequest);
    // 1) SlotRequest is routed to the SlotManager
    verify(slotManager).requestSlot(slotRequest);
    // 2) SlotRequest is confirmed
    Assert.assertEquals(slotRequestReply.getAllocationID(), allocationID);
    // 3) SlotRequest leads to a container allocation
    Assert.assertEquals(1, resourceManager.startNewWorkerCalled);
    Assert.assertFalse(slotManager.isAllocated(allocationID));
    // slot becomes available
    final String tmAddress = "/tm1";
    TaskExecutorGateway taskExecutorGateway = mock(TaskExecutorGateway.class);
    Mockito.when(taskExecutorGateway.requestSlot(any(SlotID.class), any(JobID.class), any(AllocationID.class), any(String.class), any(UUID.class), any(Time.class))).thenReturn(new FlinkCompletableFuture<TMSlotRequestReply>());
    testRpcService.registerGateway(tmAddress, taskExecutorGateway);
    final ResourceID resourceID = ResourceID.generate();
    final SlotID slotID = new SlotID(resourceID, 0);
    final SlotStatus slotStatus = new SlotStatus(slotID, resourceProfile);
    final SlotReport slotReport = new SlotReport(Collections.singletonList(slotStatus));
    // register slot at SlotManager
    slotManager.registerTaskExecutor(resourceID, new TaskExecutorRegistration(taskExecutorGateway), slotReport);
    // 4) Slot becomes available and TaskExecutor gets a SlotRequest
    verify(taskExecutorGateway, timeout(5000)).requestSlot(eq(slotID), eq(jobID), eq(allocationID), any(String.class), any(UUID.class), any(Time.class));
}
Also used : TMSlotRequestReply(org.apache.flink.runtime.resourcemanager.messages.taskexecutor.TMSlotRequestReply) TaskExecutorRegistration(org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration) JobLeaderIdService(org.apache.flink.runtime.resourcemanager.JobLeaderIdService) Time(org.apache.flink.api.common.time.Time) JobMasterGateway(org.apache.flink.runtime.jobmaster.JobMasterGateway) SlotRequest(org.apache.flink.runtime.resourcemanager.SlotRequest) TestingHighAvailabilityServices(org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices) ResourceID(org.apache.flink.runtime.clusterframework.types.ResourceID) UUID(java.util.UUID) RegistrationResponse(org.apache.flink.runtime.registration.RegistrationResponse) ResourceProfile(org.apache.flink.runtime.clusterframework.types.ResourceProfile) TestingLeaderElectionService(org.apache.flink.runtime.leaderelection.TestingLeaderElectionService) SlotStatus(org.apache.flink.runtime.taskexecutor.SlotStatus) MetricRegistry(org.apache.flink.runtime.metrics.MetricRegistry) AllocationID(org.apache.flink.runtime.clusterframework.types.AllocationID) RMSlotRequestReply(org.apache.flink.runtime.resourcemanager.messages.jobmanager.RMSlotRequestReply) SlotReport(org.apache.flink.runtime.taskexecutor.SlotReport) ResourceManagerConfiguration(org.apache.flink.runtime.resourcemanager.ResourceManagerConfiguration) TaskExecutorGateway(org.apache.flink.runtime.taskexecutor.TaskExecutorGateway) FatalErrorHandler(org.apache.flink.runtime.rpc.FatalErrorHandler) SlotID(org.apache.flink.runtime.clusterframework.types.SlotID) TestingSlotManager(org.apache.flink.runtime.resourcemanager.TestingSlotManager) JobID(org.apache.flink.api.common.JobID) Test(org.junit.Test)

Aggregations

TaskExecutorRegistration (org.apache.flink.runtime.resourcemanager.registration.TaskExecutorRegistration)5 AllocationID (org.apache.flink.runtime.clusterframework.types.AllocationID)4 SlotID (org.apache.flink.runtime.clusterframework.types.SlotID)4 TMSlotRequestReply (org.apache.flink.runtime.resourcemanager.messages.taskexecutor.TMSlotRequestReply)4 UUID (java.util.UUID)3 JobID (org.apache.flink.api.common.JobID)3 Time (org.apache.flink.api.common.time.Time)3 SlotStatus (org.apache.flink.runtime.taskexecutor.SlotStatus)3 TaskExecutorGateway (org.apache.flink.runtime.taskexecutor.TaskExecutorGateway)3 ResourceID (org.apache.flink.runtime.clusterframework.types.ResourceID)2 ResourceProfile (org.apache.flink.runtime.clusterframework.types.ResourceProfile)2 TestingHighAvailabilityServices (org.apache.flink.runtime.highavailability.TestingHighAvailabilityServices)2 JobMasterGateway (org.apache.flink.runtime.jobmaster.JobMasterGateway)2 TestingLeaderElectionService (org.apache.flink.runtime.leaderelection.TestingLeaderElectionService)2 RegistrationResponse (org.apache.flink.runtime.registration.RegistrationResponse)2 JobLeaderIdService (org.apache.flink.runtime.resourcemanager.JobLeaderIdService)2 ResourceManagerConfiguration (org.apache.flink.runtime.resourcemanager.ResourceManagerConfiguration)2 SlotRequest (org.apache.flink.runtime.resourcemanager.SlotRequest)2 TestingSlotManager (org.apache.flink.runtime.resourcemanager.TestingSlotManager)2 RMSlotRequestReply (org.apache.flink.runtime.resourcemanager.messages.jobmanager.RMSlotRequestReply)2