Search in sources :

Example 56 with DrainDispatcher

use of org.apache.hadoop.yarn.event.DrainDispatcher in project hadoop by apache.

the class TestClientToAMTokens method testClientTokenRace.

@Test(timeout = 20000)
public void testClientTokenRace() throws Exception {
    conf.set(CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHENTICATION, "kerberos");
    UserGroupInformation.setConfiguration(conf);
    ContainerManagementProtocol containerManager = mock(ContainerManagementProtocol.class);
    StartContainersResponse mockResponse = mock(StartContainersResponse.class);
    when(containerManager.startContainers((StartContainersRequest) any())).thenReturn(mockResponse);
    final DrainDispatcher dispatcher = new DrainDispatcher();
    MockRM rm = new MockRMWithCustomAMLauncher(conf, containerManager) {

        protected ClientRMService createClientRMService() {
            return new ClientRMService(this.rmContext, scheduler, this.rmAppManager, this.applicationACLsManager, this.queueACLsManager, getRMContext().getRMDelegationTokenSecretManager());
        }

        ;

        @Override
        protected Dispatcher createDispatcher() {
            return dispatcher;
        }

        @Override
        protected void doSecureLogin() throws IOException {
        }
    };
    rm.start();
    // Submit an app
    RMApp app = rm.submitApp(1024);
    // Set up a node.
    MockNM nm1 = rm.registerNode("localhost:1234", 3072);
    nm1.nodeHeartbeat(true);
    dispatcher.await();
    nm1.nodeHeartbeat(true);
    dispatcher.await();
    ApplicationAttemptId appAttempt = app.getCurrentAppAttempt().getAppAttemptId();
    final MockAM mockAM = new MockAM(rm.getRMContext(), rm.getApplicationMasterService(), app.getCurrentAppAttempt().getAppAttemptId());
    UserGroupInformation appUgi = UserGroupInformation.createRemoteUser(appAttempt.toString());
    RegisterApplicationMasterResponse response = appUgi.doAs(new PrivilegedAction<RegisterApplicationMasterResponse>() {

        @Override
        public RegisterApplicationMasterResponse run() {
            RegisterApplicationMasterResponse response = null;
            try {
                response = mockAM.registerAppAttempt();
            } catch (Exception e) {
                Assert.fail("Exception was not expected");
            }
            return response;
        }
    });
    // Get the app-report.
    GetApplicationReportRequest request = Records.newRecord(GetApplicationReportRequest.class);
    request.setApplicationId(app.getApplicationId());
    GetApplicationReportResponse reportResponse = rm.getClientRMService().getApplicationReport(request);
    ApplicationReport appReport = reportResponse.getApplicationReport();
    org.apache.hadoop.yarn.api.records.Token originalClientToAMToken = appReport.getClientToAMToken();
    // ClientToAMToken master key should have been received on register
    // application master response.
    final ByteBuffer clientMasterKey = response.getClientToAMTokenMasterKey();
    Assert.assertNotNull(clientMasterKey);
    Assert.assertTrue(clientMasterKey.array().length > 0);
    // Start the AM with the correct shared-secret.
    ApplicationAttemptId appAttemptId = app.getAppAttempts().keySet().iterator().next();
    Assert.assertNotNull(appAttemptId);
    final CustomAM am = new CustomAM(appAttemptId, null);
    am.init(conf);
    am.start();
    // Now the real test!
    // Set up clients to be able to pick up correct tokens.
    SecurityUtil.setSecurityInfoProviders(new CustomSecurityInfo());
    Token<ClientToAMTokenIdentifier> token = ConverterUtils.convertFromYarn(originalClientToAMToken, am.address);
    // Schedule the key to be set after a significant delay
    Timer timer = new Timer();
    TimerTask timerTask = new TimerTask() {

        @Override
        public void run() {
            am.setClientSecretKey(clientMasterKey.array());
        }
    };
    timer.schedule(timerTask, 250);
    // connect should pause waiting for the master key to arrive
    verifyValidToken(conf, am, token);
    am.stop();
    rm.stop();
}
Also used : DrainDispatcher(org.apache.hadoop.yarn.event.DrainDispatcher) RMApp(org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp) MockRMWithCustomAMLauncher(org.apache.hadoop.yarn.server.resourcemanager.MockRMWithCustomAMLauncher) MockNM(org.apache.hadoop.yarn.server.resourcemanager.MockNM) MockRM(org.apache.hadoop.yarn.server.resourcemanager.MockRM) TimerTask(java.util.TimerTask) UserGroupInformation(org.apache.hadoop.security.UserGroupInformation) GetApplicationReportRequest(org.apache.hadoop.yarn.api.protocolrecords.GetApplicationReportRequest) StartContainersResponse(org.apache.hadoop.yarn.api.protocolrecords.StartContainersResponse) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) ByteBuffer(java.nio.ByteBuffer) ServiceException(com.google.protobuf.ServiceException) SaslException(javax.security.sasl.SaslException) IOException(java.io.IOException) RemoteException(org.apache.hadoop.ipc.RemoteException) YarnRuntimeException(org.apache.hadoop.yarn.exceptions.YarnRuntimeException) ClientRMService(org.apache.hadoop.yarn.server.resourcemanager.ClientRMService) ApplicationReport(org.apache.hadoop.yarn.api.records.ApplicationReport) ClientToAMTokenIdentifier(org.apache.hadoop.yarn.security.client.ClientToAMTokenIdentifier) ContainerManagementProtocol(org.apache.hadoop.yarn.api.ContainerManagementProtocol) Timer(java.util.Timer) RegisterApplicationMasterResponse(org.apache.hadoop.yarn.api.protocolrecords.RegisterApplicationMasterResponse) MockAM(org.apache.hadoop.yarn.server.resourcemanager.MockAM) GetApplicationReportResponse(org.apache.hadoop.yarn.api.protocolrecords.GetApplicationReportResponse) Test(org.junit.Test)

Example 57 with DrainDispatcher

use of org.apache.hadoop.yarn.event.DrainDispatcher in project hadoop by apache.

the class TestReservationSystemWithRMHA method addNodeCapacityToPlan.

private void addNodeCapacityToPlan(MockRM rm, int memory, int vCores) {
    try {
        rm.registerNode("127.0.0.1:1", memory, vCores);
        int attempts = 10;
        do {
            DrainDispatcher dispatcher = (DrainDispatcher) rm1.getRMContext().getDispatcher();
            dispatcher.await();
            rm.getRMContext().getReservationSystem().synchronizePlan(ReservationSystemTestUtil.reservationQ, false);
            if (rm.getRMContext().getReservationSystem().getPlan(ReservationSystemTestUtil.reservationQ).getTotalCapacity().getMemorySize() > 0) {
                break;
            }
            LOG.info("Waiting for node capacity to be added to plan");
            Thread.sleep(100);
        } while (attempts-- > 0);
        if (attempts <= 0) {
            Assert.fail("Exhausted attempts in checking if node capacity was " + "added to the plan");
        }
    } catch (Exception e) {
        Assert.fail(e.getMessage());
    }
}
Also used : DrainDispatcher(org.apache.hadoop.yarn.event.DrainDispatcher)

Example 58 with DrainDispatcher

use of org.apache.hadoop.yarn.event.DrainDispatcher in project hadoop by apache.

the class TestRMContainerImpl method testExpireWhileRunning.

@Test
public void testExpireWhileRunning() {
    DrainDispatcher drainDispatcher = new DrainDispatcher();
    EventHandler<RMAppAttemptEvent> appAttemptEventHandler = mock(EventHandler.class);
    EventHandler generic = mock(EventHandler.class);
    drainDispatcher.register(RMAppAttemptEventType.class, appAttemptEventHandler);
    drainDispatcher.register(RMNodeEventType.class, generic);
    drainDispatcher.init(new YarnConfiguration());
    drainDispatcher.start();
    NodeId nodeId = BuilderUtils.newNodeId("host", 3425);
    ApplicationId appId = BuilderUtils.newApplicationId(1, 1);
    ApplicationAttemptId appAttemptId = BuilderUtils.newApplicationAttemptId(appId, 1);
    ContainerId containerId = BuilderUtils.newContainerId(appAttemptId, 1);
    ContainerAllocationExpirer expirer = mock(ContainerAllocationExpirer.class);
    Resource resource = BuilderUtils.newResource(512, 1);
    Priority priority = BuilderUtils.newPriority(5);
    Container container = BuilderUtils.newContainer(containerId, nodeId, "host:3465", resource, priority, null);
    ConcurrentMap<ApplicationId, RMApp> appMap = new ConcurrentHashMap<>();
    RMApp rmApp = mock(RMApp.class);
    appMap.putIfAbsent(appId, rmApp);
    RMApplicationHistoryWriter writer = mock(RMApplicationHistoryWriter.class);
    SystemMetricsPublisher publisher = mock(SystemMetricsPublisher.class);
    RMContext rmContext = mock(RMContext.class);
    when(rmContext.getDispatcher()).thenReturn(drainDispatcher);
    when(rmContext.getContainerAllocationExpirer()).thenReturn(expirer);
    when(rmContext.getRMApplicationHistoryWriter()).thenReturn(writer);
    when(rmContext.getSystemMetricsPublisher()).thenReturn(publisher);
    YarnConfiguration conf = new YarnConfiguration();
    conf.setBoolean(YarnConfiguration.APPLICATION_HISTORY_SAVE_NON_AM_CONTAINER_META_INFO, true);
    when(rmContext.getYarnConfiguration()).thenReturn(conf);
    when(rmContext.getRMApps()).thenReturn(appMap);
    RMContainer rmContainer = new RMContainerImpl(container, SchedulerRequestKey.extractFrom(container), appAttemptId, nodeId, "user", rmContext);
    assertEquals(RMContainerState.NEW, rmContainer.getState());
    assertEquals(resource, rmContainer.getAllocatedResource());
    assertEquals(nodeId, rmContainer.getAllocatedNode());
    assertEquals(priority, rmContainer.getAllocatedSchedulerKey().getPriority());
    verify(writer).containerStarted(any(RMContainer.class));
    verify(publisher).containerCreated(any(RMContainer.class), anyLong());
    rmContainer.handle(new RMContainerEvent(containerId, RMContainerEventType.START));
    drainDispatcher.await();
    assertEquals(RMContainerState.ALLOCATED, rmContainer.getState());
    rmContainer.handle(new RMContainerEvent(containerId, RMContainerEventType.ACQUIRED));
    drainDispatcher.await();
    assertEquals(RMContainerState.ACQUIRED, rmContainer.getState());
    rmContainer.handle(new RMContainerEvent(containerId, RMContainerEventType.LAUNCHED));
    drainDispatcher.await();
    assertEquals(RMContainerState.RUNNING, rmContainer.getState());
    assertEquals("http://host:3465/node/containerlogs/container_1_0001_01_000001/user", rmContainer.getLogURL());
    // In RUNNING state. Verify EXPIRE and associated actions.
    reset(appAttemptEventHandler);
    ContainerStatus containerStatus = SchedulerUtils.createAbnormalContainerStatus(containerId, SchedulerUtils.EXPIRED_CONTAINER);
    rmContainer.handle(new RMContainerFinishedEvent(containerId, containerStatus, RMContainerEventType.EXPIRE));
    drainDispatcher.await();
    assertEquals(RMContainerState.RUNNING, rmContainer.getState());
    verify(writer, never()).containerFinished(any(RMContainer.class));
    verify(publisher, never()).containerFinished(any(RMContainer.class), anyLong());
}
Also used : DrainDispatcher(org.apache.hadoop.yarn.event.DrainDispatcher) RMApp(org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp) RMContext(org.apache.hadoop.yarn.server.resourcemanager.RMContext) Priority(org.apache.hadoop.yarn.api.records.Priority) Resource(org.apache.hadoop.yarn.api.records.Resource) EventHandler(org.apache.hadoop.yarn.event.EventHandler) RMApplicationHistoryWriter(org.apache.hadoop.yarn.server.resourcemanager.ahs.RMApplicationHistoryWriter) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) Container(org.apache.hadoop.yarn.api.records.Container) ContainerStatus(org.apache.hadoop.yarn.api.records.ContainerStatus) YarnConfiguration(org.apache.hadoop.yarn.conf.YarnConfiguration) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) SystemMetricsPublisher(org.apache.hadoop.yarn.server.resourcemanager.metrics.SystemMetricsPublisher) NodeId(org.apache.hadoop.yarn.api.records.NodeId) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) ConcurrentHashMap(java.util.concurrent.ConcurrentHashMap) RMAppAttemptEvent(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptEvent) Test(org.junit.Test)

Example 59 with DrainDispatcher

use of org.apache.hadoop.yarn.event.DrainDispatcher in project hadoop by apache.

the class TestRMAppTransitions method setUp.

@Before
public void setUp() throws Exception {
    conf = new YarnConfiguration();
    AuthenticationMethod authMethod = AuthenticationMethod.SIMPLE;
    if (isSecurityEnabled) {
        authMethod = AuthenticationMethod.KERBEROS;
    }
    SecurityUtil.setAuthenticationMethod(authMethod, conf);
    UserGroupInformation.setConfiguration(conf);
    rmDispatcher = new DrainDispatcher();
    ContainerAllocationExpirer containerAllocationExpirer = mock(ContainerAllocationExpirer.class);
    AMLivelinessMonitor amLivelinessMonitor = mock(AMLivelinessMonitor.class);
    AMLivelinessMonitor amFinishingMonitor = mock(AMLivelinessMonitor.class);
    store = mock(RMStateStore.class);
    writer = mock(RMApplicationHistoryWriter.class);
    DelegationTokenRenewer renewer = mock(DelegationTokenRenewer.class);
    RMContext realRMContext = new RMContextImpl(rmDispatcher, containerAllocationExpirer, amLivelinessMonitor, amFinishingMonitor, renewer, new AMRMTokenSecretManager(conf, this.rmContext), new RMContainerTokenSecretManager(conf), new NMTokenSecretManagerInRM(conf), new ClientToAMTokenSecretManagerInRM());
    ((RMContextImpl) realRMContext).setStateStore(store);
    publisher = mock(SystemMetricsPublisher.class);
    realRMContext.setSystemMetricsPublisher(publisher);
    realRMContext.setRMApplicationHistoryWriter(writer);
    this.rmContext = spy(realRMContext);
    ResourceScheduler resourceScheduler = mock(ResourceScheduler.class);
    doReturn(null).when(resourceScheduler).getAppResourceUsageReport((ApplicationAttemptId) Matchers.any());
    doReturn(resourceScheduler).when(rmContext).getScheduler();
    doReturn(mock(RMTimelineCollectorManager.class)).when(rmContext).getRMTimelineCollectorManager();
    rmDispatcher.register(RMAppAttemptEventType.class, new TestApplicationAttemptEventDispatcher(this.rmContext));
    rmDispatcher.register(RMAppEventType.class, new TestApplicationEventDispatcher(rmContext));
    rmDispatcher.register(RMAppManagerEventType.class, new TestApplicationManagerEventDispatcher());
    schedulerDispatcher = new TestSchedulerEventDispatcher();
    rmDispatcher.register(SchedulerEventType.class, schedulerDispatcher);
    rmDispatcher.init(conf);
    rmDispatcher.start();
}
Also used : DrainDispatcher(org.apache.hadoop.yarn.event.DrainDispatcher) RMContext(org.apache.hadoop.yarn.server.resourcemanager.RMContext) RMStateStore(org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore) ClientToAMTokenSecretManagerInRM(org.apache.hadoop.yarn.server.resourcemanager.security.ClientToAMTokenSecretManagerInRM) ContainerAllocationExpirer(org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.ContainerAllocationExpirer) RMApplicationHistoryWriter(org.apache.hadoop.yarn.server.resourcemanager.ahs.RMApplicationHistoryWriter) AuthenticationMethod(org.apache.hadoop.security.UserGroupInformation.AuthenticationMethod) AMRMTokenSecretManager(org.apache.hadoop.yarn.server.resourcemanager.security.AMRMTokenSecretManager) NMTokenSecretManagerInRM(org.apache.hadoop.yarn.server.resourcemanager.security.NMTokenSecretManagerInRM) DelegationTokenRenewer(org.apache.hadoop.yarn.server.resourcemanager.security.DelegationTokenRenewer) RMTimelineCollectorManager(org.apache.hadoop.yarn.server.resourcemanager.timelineservice.RMTimelineCollectorManager) YarnConfiguration(org.apache.hadoop.yarn.conf.YarnConfiguration) SystemMetricsPublisher(org.apache.hadoop.yarn.server.resourcemanager.metrics.SystemMetricsPublisher) AMLivelinessMonitor(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.AMLivelinessMonitor) ResourceScheduler(org.apache.hadoop.yarn.server.resourcemanager.scheduler.ResourceScheduler) RMContainerTokenSecretManager(org.apache.hadoop.yarn.server.resourcemanager.security.RMContainerTokenSecretManager) RMContextImpl(org.apache.hadoop.yarn.server.resourcemanager.RMContextImpl) Before(org.junit.Before)

Example 60 with DrainDispatcher

use of org.apache.hadoop.yarn.event.DrainDispatcher in project hadoop by apache.

the class TestAbstractYarnScheduler method testNodemanagerReconnect.

/**
   * Test the behavior of the scheduler when a node reconnects
   * with changed capabilities. This test is to catch any race conditions
   * that might occur due to the use of the RMNode object.
   * @throws Exception
   */
@Test(timeout = 60000)
public void testNodemanagerReconnect() throws Exception {
    configureScheduler();
    Configuration conf = getConf();
    MockRM rm = new MockRM(conf);
    try {
        rm.start();
        conf.setBoolean(Dispatcher.DISPATCHER_EXIT_ON_ERROR_KEY, false);
        DrainDispatcher privateDispatcher = new DrainDispatcher();
        SleepHandler sleepHandler = new SleepHandler();
        ResourceTrackerService privateResourceTrackerService = getPrivateResourceTrackerService(privateDispatcher, rm, sleepHandler);
        // Register node1
        String hostname1 = "localhost1";
        Resource capability = BuilderUtils.newResource(4096, 4);
        RecordFactory recordFactory = RecordFactoryProvider.getRecordFactory(null);
        RegisterNodeManagerRequest request1 = recordFactory.newRecordInstance(RegisterNodeManagerRequest.class);
        NodeId nodeId1 = NodeId.newInstance(hostname1, 0);
        request1.setNodeId(nodeId1);
        request1.setHttpPort(0);
        request1.setResource(capability);
        privateResourceTrackerService.registerNodeManager(request1);
        privateDispatcher.await();
        Resource clusterResource = rm.getResourceScheduler().getClusterResource();
        Assert.assertEquals("Initial cluster resources don't match", capability, clusterResource);
        Resource newCapability = BuilderUtils.newResource(1024, 1);
        RegisterNodeManagerRequest request2 = recordFactory.newRecordInstance(RegisterNodeManagerRequest.class);
        request2.setNodeId(nodeId1);
        request2.setHttpPort(0);
        request2.setResource(newCapability);
        // hold up the disaptcher and register the same node with lower capability
        sleepHandler.sleepFlag = true;
        privateResourceTrackerService.registerNodeManager(request2);
        privateDispatcher.await();
        Assert.assertEquals("Cluster resources don't match", newCapability, rm.getResourceScheduler().getClusterResource());
        privateResourceTrackerService.stop();
    } finally {
        rm.stop();
    }
}
Also used : DrainDispatcher(org.apache.hadoop.yarn.event.DrainDispatcher) RegisterNodeManagerRequest(org.apache.hadoop.yarn.server.api.protocolrecords.RegisterNodeManagerRequest) RecordFactory(org.apache.hadoop.yarn.factories.RecordFactory) Configuration(org.apache.hadoop.conf.Configuration) YarnConfiguration(org.apache.hadoop.yarn.conf.YarnConfiguration) Resource(org.apache.hadoop.yarn.api.records.Resource) NodeId(org.apache.hadoop.yarn.api.records.NodeId) MockRM(org.apache.hadoop.yarn.server.resourcemanager.MockRM) ResourceTrackerService(org.apache.hadoop.yarn.server.resourcemanager.ResourceTrackerService) Test(org.junit.Test)

Aggregations

DrainDispatcher (org.apache.hadoop.yarn.event.DrainDispatcher)84 Test (org.junit.Test)73 YarnConfiguration (org.apache.hadoop.yarn.conf.YarnConfiguration)54 Configuration (org.apache.hadoop.conf.Configuration)50 RMApp (org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp)41 ApplicationAttemptId (org.apache.hadoop.yarn.api.records.ApplicationAttemptId)35 MockNM (org.apache.hadoop.yarn.server.resourcemanager.MockNM)32 Path (org.apache.hadoop.fs.Path)24 ApplicationId (org.apache.hadoop.yarn.api.records.ApplicationId)24 Job (org.apache.hadoop.mapreduce.v2.app.job.Job)23 JobId (org.apache.hadoop.mapreduce.v2.api.records.JobId)21 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)20 IOException (java.io.IOException)16 ArrayList (java.util.ArrayList)16 DeletionService (org.apache.hadoop.yarn.server.nodemanager.DeletionService)16 LocalizerEvent (org.apache.hadoop.yarn.server.nodemanager.containermanager.localizer.event.LocalizerEvent)15 HashMap (java.util.HashMap)14 LocalDirsHandlerService (org.apache.hadoop.yarn.server.nodemanager.LocalDirsHandlerService)14 LocalResourceVisibility (org.apache.hadoop.yarn.api.records.LocalResourceVisibility)12 Container (org.apache.hadoop.yarn.server.nodemanager.containermanager.container.Container)12