Examples with RMContainer org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer used on opensource projects

Example 51 with RMContainer

use of org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer in project hadoop by apache.

the class TestRMAppAttemptTransitions method testLaunchedFailWhileAHSEnabled.

@SuppressWarnings("unchecked")
@Test(timeout = 10000)
public void testLaunchedFailWhileAHSEnabled() {
    Configuration myConf = new Configuration(conf);
    myConf.setBoolean(YarnConfiguration.APPLICATION_HISTORY_ENABLED, true);
    ApplicationId applicationId = MockApps.newAppID(appId);
    ApplicationAttemptId applicationAttemptId = ApplicationAttemptId.newInstance(applicationId, 2);
    RMAppAttempt myApplicationAttempt = new RMAppAttemptImpl(applicationAttempt.getAppAttemptId(), spyRMContext, scheduler, masterService, submissionContext, myConf, BuilderUtils.newResourceRequest(RMAppAttemptImpl.AM_CONTAINER_PRIORITY, ResourceRequest.ANY, submissionContext.getResource(), 1), application);
    //submit, schedule and allocate app attempt
    myApplicationAttempt.handle(new RMAppAttemptEvent(myApplicationAttempt.getAppAttemptId(), RMAppAttemptEventType.START));
    myApplicationAttempt.handle(new RMAppAttemptEvent(myApplicationAttempt.getAppAttemptId(), RMAppAttemptEventType.ATTEMPT_ADDED));
    Container amContainer = mock(Container.class);
    Resource resource = BuilderUtils.newResource(2048, 1);
    when(amContainer.getId()).thenReturn(BuilderUtils.newContainerId(myApplicationAttempt.getAppAttemptId(), 1));
    when(amContainer.getResource()).thenReturn(resource);
    Allocation allocation = mock(Allocation.class);
    when(allocation.getContainers()).thenReturn(Collections.singletonList(amContainer));
    when(scheduler.allocate(any(ApplicationAttemptId.class), any(List.class), any(List.class), any(List.class), any(List.class), any(ContainerUpdates.class))).thenReturn(allocation);
    RMContainer rmContainer = mock(RMContainerImpl.class);
    when(scheduler.getRMContainer(amContainer.getId())).thenReturn(rmContainer);
    myApplicationAttempt.handle(new RMAppAttemptEvent(myApplicationAttempt.getAppAttemptId(), RMAppAttemptEventType.CONTAINER_ALLOCATED));
    assertEquals(RMAppAttemptState.ALLOCATED_SAVING, myApplicationAttempt.getAppAttemptState());
    myApplicationAttempt.handle(new RMAppAttemptEvent(myApplicationAttempt.getAppAttemptId(), RMAppAttemptEventType.ATTEMPT_NEW_SAVED));
    // launch app attempt
    myApplicationAttempt.handle(new RMAppAttemptEvent(myApplicationAttempt.getAppAttemptId(), RMAppAttemptEventType.LAUNCHED));
    assertEquals(YarnApplicationAttemptState.LAUNCHED, myApplicationAttempt.createApplicationAttemptState());
    //fail container right after launched
    NodeId anyNodeId = NodeId.newInstance("host", 1234);
    myApplicationAttempt.handle(new RMAppAttemptContainerFinishedEvent(myApplicationAttempt.getAppAttemptId(), BuilderUtils.newContainerStatus(amContainer.getId(), ContainerState.COMPLETE, "", 0, amContainer.getResource()), anyNodeId));
    sendAttemptUpdateSavedEvent(myApplicationAttempt);
    assertEquals(RMAppAttemptState.FAILED, myApplicationAttempt.getAppAttemptState());
    String rmAppPageUrl = pjoin(AHS_WEBAPP_ADDR, "applicationhistory", "app", myApplicationAttempt.getAppAttemptId().getApplicationId());
    assertEquals(rmAppPageUrl, myApplicationAttempt.getOriginalTrackingUrl());
    assertEquals(rmAppPageUrl, myApplicationAttempt.getTrackingUrl());
}

Example 52 with RMContainer

use of org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer in project hadoop by apache.

the class TestCapacitySchedulerLazyPreemption method testSimplePreemption.

@Test(timeout = 60000)
public void testSimplePreemption() throws Exception {
    /**
     * Test case: Submit two application (app1/app2) to different queues, queue
     * structure:
     *
     * <pre>
     *             Root
     *            /  |  \
     *           a   b   c
     *          10   20  70
     * </pre>
     *
     * 1) Two nodes in the cluster, each of them has 4G.
     *
     * 2) app1 submit to queue-a first, it asked 7 * 1G containers, so there's no
     * more resource available.
     *
     * 3) app2 submit to queue-c, ask for one 1G container (for AM)
     *
     * Now the cluster is fulfilled.
     *
     * 4) app2 asks for another 1G container, system will preempt one container
     * from app1, and app2 will receive the preempted container
     */
    MockRM rm1 = new MockRM(conf);
    rm1.getRMContext().setNodeLabelManager(mgr);
    rm1.start();
    MockNM nm1 = rm1.registerNode("h1:1234", 4 * GB);
    MockNM nm2 = rm1.registerNode("h2:1234", 4 * GB);
    CapacityScheduler cs = (CapacityScheduler) rm1.getResourceScheduler();
    RMNode rmNode1 = rm1.getRMContext().getRMNodes().get(nm1.getNodeId());
    RMNode rmNode2 = rm1.getRMContext().getRMNodes().get(nm2.getNodeId());
    // launch an app to queue, AM container should be launched in nm1
    RMApp app1 = rm1.submitApp(1 * GB, "app", "user", null, "a");
    MockAM am1 = MockRM.launchAndRegisterAM(app1, rm1, nm1);
    am1.allocate("*", 1 * GB, 7, new ArrayList<ContainerId>());
    // Do allocation 3 times for node1/node2
    for (int i = 0; i < 3; i++) {
        cs.handle(new NodeUpdateSchedulerEvent(rmNode1));
        cs.handle(new NodeUpdateSchedulerEvent(rmNode2));
    }
    // App1 should have 7 containers now, and no available resource for cluster
    FiCaSchedulerApp schedulerApp1 = cs.getApplicationAttempt(am1.getApplicationAttemptId());
    Assert.assertEquals(7, schedulerApp1.getLiveContainers().size());
    // Submit app2 to queue-c and asks for a 1G container for AM
    RMApp app2 = rm1.submitApp(1 * GB, "app", "user", null, "c");
    MockAM am2 = MockRM.launchAndRegisterAM(app2, rm1, nm2);
    // NM1/NM2 has available resource = 0G
    Assert.assertEquals(0 * GB, cs.getNode(nm1.getNodeId()).getUnallocatedResource().getMemorySize());
    Assert.assertEquals(0 * GB, cs.getNode(nm2.getNodeId()).getUnallocatedResource().getMemorySize());
    // AM asks for a 1 * GB container
    am2.allocate(Arrays.asList(ResourceRequest.newInstance(Priority.newInstance(1), ResourceRequest.ANY, Resources.createResource(1 * GB), 1)), null);
    // Get edit policy and do one update
    SchedulingEditPolicy editPolicy = getSchedulingEditPolicy(rm1);
    // Call edit schedule twice, and check if one container from app1 marked
    // to be "killable"
    editPolicy.editSchedule();
    editPolicy.editSchedule();
    PreemptionManager pm = cs.getPreemptionManager();
    Map<ContainerId, RMContainer> killableContainers = waitKillableContainersSize(pm, "a", RMNodeLabelsManager.NO_LABEL, 1);
    Assert.assertEquals(1, killableContainers.size());
    Assert.assertEquals(killableContainers.entrySet().iterator().next().getKey().getApplicationAttemptId(), am1.getApplicationAttemptId());
    // Call CS.handle once to see if container preempted
    cs.handle(new NodeUpdateSchedulerEvent(rmNode2));
    FiCaSchedulerApp schedulerApp2 = cs.getApplicationAttempt(am2.getApplicationAttemptId());
    // App1 has 6 containers, and app2 has 2 containers
    Assert.assertEquals(6, schedulerApp1.getLiveContainers().size());
    Assert.assertEquals(2, schedulerApp2.getLiveContainers().size());
    rm1.close();
}

Example 53 with RMContainer

use of org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer in project hadoop by apache.

the class CapacitySchedulerPreemptionTestBase method checkNumberOfPreemptionCandidateFromApp.

public void checkNumberOfPreemptionCandidateFromApp(ProportionalCapacityPreemptionPolicy policy, int expected, ApplicationAttemptId attemptId) {
    int total = 0;
    for (RMContainer rmContainer : policy.getToPreemptContainers().keySet()) {
        if (rmContainer.getApplicationAttemptId().equals(attemptId)) {
            ++total;
        }
    }
    Assert.assertEquals(expected, total);
}

Example 54 with RMContainer

use of org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer in project hadoop by apache.

the class TestCapacityScheduler method sentRMContainerLaunched.

private void sentRMContainerLaunched(MockRM rm, ContainerId containerId) {
    CapacityScheduler cs = (CapacityScheduler) rm.getResourceScheduler();
    RMContainer rmContainer = cs.getRMContainer(containerId);
    if (rmContainer != null) {
        rmContainer.handle(new RMContainerEvent(containerId, RMContainerEventType.LAUNCHED));
    } else {
        Assert.fail("Cannot find RMContainer");
    }
}

Example 55 with RMContainer

use of org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer in project hadoop by apache.

the class TestCapacityScheduler method testRecoverRequestAfterPreemption.

@Test(timeout = 300000)
public void testRecoverRequestAfterPreemption() throws Exception {
    Configuration conf = new Configuration();
    conf.setClass(YarnConfiguration.RM_SCHEDULER, CapacityScheduler.class, ResourceScheduler.class);
    MockRM rm1 = new MockRM(conf);
    rm1.start();
    MockNM nm1 = rm1.registerNode("127.0.0.1:1234", 8000);
    RMApp app1 = rm1.submitApp(1024);
    MockAM am1 = MockRM.launchAndRegisterAM(app1, rm1, nm1);
    CapacityScheduler cs = (CapacityScheduler) rm1.getResourceScheduler();
    // request a container.
    am1.allocate("127.0.0.1", 1024, 1, new ArrayList<ContainerId>());
    ContainerId containerId1 = ContainerId.newContainerId(am1.getApplicationAttemptId(), 2);
    rm1.waitForState(nm1, containerId1, RMContainerState.ALLOCATED);
    RMContainer rmContainer = cs.getRMContainer(containerId1);
    List<ResourceRequest> requests = rmContainer.getResourceRequests();
    FiCaSchedulerApp app = cs.getApplicationAttempt(am1.getApplicationAttemptId());
    FiCaSchedulerNode node = cs.getNode(rmContainer.getAllocatedNode());
    for (ResourceRequest request : requests) {
        // Skip the OffRack and RackLocal resource requests.
        if (request.getResourceName().equals(node.getRackName()) || request.getResourceName().equals(ResourceRequest.ANY)) {
            continue;
        }
        // Already the node local resource request is cleared from RM after
        // allocation.
        Assert.assertEquals(0, app.getOutstandingAsksCount(SchedulerRequestKey.create(request), request.getResourceName()));
    }
    // Call killContainer to preempt the container
    cs.markContainerForKillable(rmContainer);
    Assert.assertEquals(3, requests.size());
    for (ResourceRequest request : requests) {
        // Resource request must have added back in RM after preempt event
        // handling.
        Assert.assertEquals(1, app.getOutstandingAsksCount(SchedulerRequestKey.create(request), request.getResourceName()));
    }
    // New container will be allocated and will move to ALLOCATED state
    ContainerId containerId2 = ContainerId.newContainerId(am1.getApplicationAttemptId(), 3);
    rm1.waitForState(nm1, containerId2, RMContainerState.ALLOCATED);
    // allocate container
    List<Container> containers = am1.allocate(new ArrayList<ResourceRequest>(), new ArrayList<ContainerId>()).getAllocatedContainers();
    // Now with updated ResourceRequest, a container is allocated for AM.
    Assert.assertTrue(containers.size() == 1);
}