Search in sources :

Example 16 with RMAppAttempt

use of org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt in project hadoop by apache.

the class MockAM method waitForState.

/**
   * Wait until an attempt has reached a specified state.
   * The timeout is 40 seconds.
   * @param finalState the attempt state waited
   * @throws InterruptedException
   *         if interrupted while waiting for the state transition
   */
private void waitForState(RMAppAttemptState finalState) throws InterruptedException {
    RMApp app = context.getRMApps().get(attemptId.getApplicationId());
    RMAppAttempt attempt = app.getRMAppAttempt(attemptId);
    MockRM.waitForState(attempt, finalState);
}
Also used : RMApp(org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp) RMAppAttempt(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt)

Example 17 with RMAppAttempt

use of org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt in project hadoop by apache.

the class ApplicationMasterLauncher method handle.

@Override
public synchronized void handle(AMLauncherEvent appEvent) {
    AMLauncherEventType event = appEvent.getType();
    RMAppAttempt application = appEvent.getAppAttempt();
    switch(event) {
        case LAUNCH:
            launch(application);
            break;
        case CLEANUP:
            cleanup(application);
            break;
        default:
            break;
    }
}
Also used : RMAppAttempt(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt)

Example 18 with RMAppAttempt

use of org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt in project hadoop by apache.

the class TestClientRMService method getRMApp.

private RMAppImpl getRMApp(RMContext rmContext, YarnScheduler yarnScheduler, ApplicationId applicationId3, YarnConfiguration config, String queueName, final long memorySeconds, final long vcoreSeconds, String appNodeLabelExpression, String amNodeLabelExpression) {
    ApplicationSubmissionContext asContext = mock(ApplicationSubmissionContext.class);
    when(asContext.getMaxAppAttempts()).thenReturn(1);
    when(asContext.getNodeLabelExpression()).thenReturn(appNodeLabelExpression);
    when(asContext.getPriority()).thenReturn(Priority.newInstance(0));
    RMAppImpl app = spy(new RMAppImpl(applicationId3, rmContext, config, null, null, queueName, asContext, yarnScheduler, null, System.currentTimeMillis(), "YARN", null, BuilderUtils.newResourceRequest(RMAppAttemptImpl.AM_CONTAINER_PRIORITY, ResourceRequest.ANY, Resource.newInstance(1024, 1), 1)) {

        @Override
        public ApplicationReport createAndGetApplicationReport(String clientUserName, boolean allowAccess) {
            ApplicationReport report = super.createAndGetApplicationReport(clientUserName, allowAccess);
            ApplicationResourceUsageReport usageReport = report.getApplicationResourceUsageReport();
            usageReport.setMemorySeconds(memorySeconds);
            usageReport.setVcoreSeconds(vcoreSeconds);
            report.setApplicationResourceUsageReport(usageReport);
            return report;
        }
    });
    app.getAMResourceRequest().setNodeLabelExpression(amNodeLabelExpression);
    ApplicationAttemptId attemptId = ApplicationAttemptId.newInstance(ApplicationId.newInstance(123456, 1), 1);
    RMAppAttemptImpl rmAppAttemptImpl = spy(new RMAppAttemptImpl(attemptId, rmContext, yarnScheduler, null, asContext, config, null, app));
    Container container = Container.newInstance(ContainerId.newContainerId(attemptId, 1), null, "", null, null, null);
    RMContainerImpl containerimpl = spy(new RMContainerImpl(container, SchedulerRequestKey.extractFrom(container), attemptId, null, "", rmContext));
    Map<ApplicationAttemptId, RMAppAttempt> attempts = new HashMap<ApplicationAttemptId, RMAppAttempt>();
    attempts.put(attemptId, rmAppAttemptImpl);
    when(app.getCurrentAppAttempt()).thenReturn(rmAppAttemptImpl);
    when(app.getAppAttempts()).thenReturn(attempts);
    when(app.getApplicationPriority()).thenReturn(Priority.newInstance(0));
    when(rmAppAttemptImpl.getMasterContainer()).thenReturn(container);
    ResourceScheduler rs = mock(ResourceScheduler.class);
    when(rmContext.getScheduler()).thenReturn(rs);
    when(rmContext.getScheduler().getRMContainer(any(ContainerId.class))).thenReturn(containerimpl);
    SchedulerAppReport sAppReport = mock(SchedulerAppReport.class);
    when(rmContext.getScheduler().getSchedulerAppInfo(any(ApplicationAttemptId.class))).thenReturn(sAppReport);
    List<RMContainer> rmContainers = new ArrayList<RMContainer>();
    rmContainers.add(containerimpl);
    when(rmContext.getScheduler().getSchedulerAppInfo(attemptId).getLiveContainers()).thenReturn(rmContainers);
    ContainerStatus cs = mock(ContainerStatus.class);
    when(containerimpl.completed()).thenReturn(false);
    when(containerimpl.getDiagnosticsInfo()).thenReturn("N/A");
    when(containerimpl.getContainerExitStatus()).thenReturn(0);
    when(containerimpl.getContainerState()).thenReturn(ContainerState.COMPLETE);
    return app;
}
Also used : RMAppImpl(org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppImpl) RMAppAttempt(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt) ConcurrentHashMap(java.util.concurrent.ConcurrentHashMap) HashMap(java.util.HashMap) ArrayList(java.util.ArrayList) Matchers.anyString(org.mockito.Matchers.anyString) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) RMContainer(org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer) ApplicationReport(org.apache.hadoop.yarn.api.records.ApplicationReport) RMContainer(org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer) Container(org.apache.hadoop.yarn.api.records.Container) RMContainerImpl(org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainerImpl) ContainerStatus(org.apache.hadoop.yarn.api.records.ContainerStatus) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) ApplicationResourceUsageReport(org.apache.hadoop.yarn.api.records.ApplicationResourceUsageReport) RMAppAttemptImpl(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl) ApplicationSubmissionContext(org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext) ResourceScheduler(org.apache.hadoop.yarn.server.resourcemanager.scheduler.ResourceScheduler) SchedulerAppReport(org.apache.hadoop.yarn.server.resourcemanager.scheduler.SchedulerAppReport)

Example 19 with RMAppAttempt

use of org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt in project hadoop by apache.

the class TestNodeBlacklistingOnAMFailures method testNodeBlacklistingOnAMFailure.

@Test(timeout = 100000)
public void testNodeBlacklistingOnAMFailure() throws Exception {
    YarnConfiguration conf = new YarnConfiguration();
    conf.setClass(YarnConfiguration.RM_SCHEDULER, CapacityScheduler.class, ResourceScheduler.class);
    conf.setBoolean(YarnConfiguration.AM_SCHEDULING_NODE_BLACKLISTING_ENABLED, true);
    DrainDispatcher dispatcher = new DrainDispatcher();
    MockRM rm = startRM(conf, dispatcher);
    CapacityScheduler scheduler = (CapacityScheduler) rm.getResourceScheduler();
    // Register 5 nodes, so that we can blacklist atleast one if AM container
    // is failed. As per calculation it will be like, 5nodes * 0.2 (default)=1.
    // First register 2 nodes, and after AM lauched register 3 more nodes.
    MockNM nm1 = new MockNM("127.0.0.1:1234", 8000, rm.getResourceTrackerService());
    nm1.registerNode();
    MockNM nm2 = new MockNM("127.0.0.2:2345", 8000, rm.getResourceTrackerService());
    nm2.registerNode();
    RMApp app = rm.submitApp(200);
    MockAM am1 = MockRM.launchAndRegisterAM(app, rm, nm1);
    ContainerId amContainerId = ContainerId.newContainerId(am1.getApplicationAttemptId(), 1);
    RMContainer rmContainer = scheduler.getRMContainer(amContainerId);
    NodeId nodeWhereAMRan = rmContainer.getAllocatedNode();
    MockNM currentNode, otherNode;
    if (nodeWhereAMRan.equals(nm1.getNodeId())) {
        currentNode = nm1;
        otherNode = nm2;
    } else {
        currentNode = nm2;
        otherNode = nm1;
    }
    // register 3 nodes now
    MockNM nm3 = new MockNM("127.0.0.3:2345", 8000, rm.getResourceTrackerService());
    nm3.registerNode();
    MockNM nm4 = new MockNM("127.0.0.4:2345", 8000, rm.getResourceTrackerService());
    nm4.registerNode();
    MockNM nm5 = new MockNM("127.0.0.5:2345", 8000, rm.getResourceTrackerService());
    nm5.registerNode();
    // Set the exist status to INVALID so that we can verify that the system
    // automatically blacklisting the node
    makeAMContainerExit(rm, amContainerId, currentNode, ContainerExitStatus.INVALID);
    // restart the am
    RMAppAttempt attempt = MockRM.waitForAttemptScheduled(app, rm);
    System.out.println("New AppAttempt launched " + attempt.getAppAttemptId());
    // Try the current node a few times
    for (int i = 0; i <= 2; i++) {
        currentNode.nodeHeartbeat(true);
        dispatcher.await();
        Assert.assertEquals("AppAttemptState should still be SCHEDULED if currentNode is " + "blacklisted correctly", RMAppAttemptState.SCHEDULED, attempt.getAppAttemptState());
    }
    // Now try the other node
    otherNode.nodeHeartbeat(true);
    dispatcher.await();
    // Now the AM container should be allocated
    MockRM.waitForState(attempt, RMAppAttemptState.ALLOCATED, 20000);
    MockAM am2 = rm.sendAMLaunched(attempt.getAppAttemptId());
    rm.waitForState(attempt.getAppAttemptId(), RMAppAttemptState.LAUNCHED);
    amContainerId = ContainerId.newContainerId(am2.getApplicationAttemptId(), 1);
    rmContainer = scheduler.getRMContainer(amContainerId);
    nodeWhereAMRan = rmContainer.getAllocatedNode();
    // The other node should now receive the assignment
    Assert.assertEquals("After blacklisting, AM should have run on the other node", otherNode.getNodeId(), nodeWhereAMRan);
    am2.registerAppAttempt();
    rm.waitForState(app.getApplicationId(), RMAppState.RUNNING);
    List<Container> allocatedContainers = TestAMRestart.allocateContainers(currentNode, am2, 1);
    Assert.assertEquals("Even though AM is blacklisted from the node, application can " + "still allocate non-AM containers there", currentNode.getNodeId(), allocatedContainers.get(0).getNodeId());
}
Also used : DrainDispatcher(org.apache.hadoop.yarn.event.DrainDispatcher) RMApp(org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp) RMAppAttempt(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt) RMContainer(org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer) RMContainer(org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer) Container(org.apache.hadoop.yarn.api.records.Container) YarnConfiguration(org.apache.hadoop.yarn.conf.YarnConfiguration) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) NodeId(org.apache.hadoop.yarn.api.records.NodeId) CapacityScheduler(org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler) Test(org.junit.Test)

Example 20 with RMAppAttempt

use of org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt in project hadoop by apache.

the class TestApplicationMasterLauncher method testSetupTokens.

@Test
public void testSetupTokens() throws Exception {
    MockRM rm = new MockRM();
    rm.start();
    MockNM nm1 = rm.registerNode("h1:1234", 5000);
    RMApp app = rm.submitApp(2000);
    /// kick the scheduling
    nm1.nodeHeartbeat(true);
    RMAppAttempt attempt = app.getCurrentAppAttempt();
    MyAMLauncher launcher = new MyAMLauncher(rm.getRMContext(), attempt, AMLauncherEventType.LAUNCH, rm.getConfig());
    DataOutputBuffer dob = new DataOutputBuffer();
    Credentials ts = new Credentials();
    ts.writeTokenStorageToStream(dob);
    ByteBuffer securityTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
    ContainerLaunchContext amContainer = ContainerLaunchContext.newInstance(null, null, null, null, securityTokens, null);
    ContainerId containerId = ContainerId.newContainerId(attempt.getAppAttemptId(), 0L);
    try {
        launcher.setupTokens(amContainer, containerId);
    } catch (Exception e) {
    // ignore the first fake exception
    }
    try {
        launcher.setupTokens(amContainer, containerId);
    } catch (java.io.EOFException e) {
        Assert.fail("EOFException should not happen.");
    }
}
Also used : RMApp(org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp) RMAppAttempt(org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt) ContainerLaunchContext(org.apache.hadoop.yarn.api.records.ContainerLaunchContext) ByteBuffer(java.nio.ByteBuffer) ApplicationMasterNotRegisteredException(org.apache.hadoop.yarn.exceptions.ApplicationMasterNotRegisteredException) YarnException(org.apache.hadoop.yarn.exceptions.YarnException) SerializedException(org.apache.hadoop.yarn.api.records.SerializedException) IOException(java.io.IOException) ApplicationAttemptNotFoundException(org.apache.hadoop.yarn.exceptions.ApplicationAttemptNotFoundException) NMNotYetReadyException(org.apache.hadoop.yarn.exceptions.NMNotYetReadyException) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) DataOutputBuffer(org.apache.hadoop.io.DataOutputBuffer) Credentials(org.apache.hadoop.security.Credentials) Test(org.junit.Test)

Aggregations

RMAppAttempt (org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttempt)123 RMApp (org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp)91 Test (org.junit.Test)71 ApplicationAttemptId (org.apache.hadoop.yarn.api.records.ApplicationAttemptId)40 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)35 Container (org.apache.hadoop.yarn.api.records.Container)31 YarnConfiguration (org.apache.hadoop.yarn.conf.YarnConfiguration)30 ApplicationId (org.apache.hadoop.yarn.api.records.ApplicationId)28 MockNM (org.apache.hadoop.yarn.server.resourcemanager.MockNM)28 ArrayList (java.util.ArrayList)26 MockAM (org.apache.hadoop.yarn.server.resourcemanager.MockAM)22 MockRM (org.apache.hadoop.yarn.server.resourcemanager.MockRM)22 ResourceRequest (org.apache.hadoop.yarn.api.records.ResourceRequest)21 AllocateResponse (org.apache.hadoop.yarn.api.protocolrecords.AllocateResponse)19 RMContainer (org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer)18 MemoryRMStateStore (org.apache.hadoop.yarn.server.resourcemanager.recovery.MemoryRMStateStore)16 NMContainerStatus (org.apache.hadoop.yarn.server.api.protocolrecords.NMContainerStatus)14 HashMap (java.util.HashMap)13 ApplicationStateData (org.apache.hadoop.yarn.server.resourcemanager.recovery.records.ApplicationStateData)13 UserGroupInformation (org.apache.hadoop.security.UserGroupInformation)12