|
|
@@ -65,6 +65,7 @@ import org.apache.hadoop.util.concurrent.HadoopExecutors;
|
|
|
import org.apache.hadoop.yarn.api.protocolrecords.SignalContainerRequest;
|
|
|
import org.apache.hadoop.yarn.api.records.ApplicationAttemptId;
|
|
|
import org.apache.hadoop.yarn.api.records.ApplicationId;
|
|
|
+import org.apache.hadoop.yarn.api.records.ContainerExitStatus;
|
|
|
import org.apache.hadoop.yarn.api.records.ContainerId;
|
|
|
import org.apache.hadoop.yarn.api.records.ContainerLaunchContext;
|
|
|
import org.apache.hadoop.yarn.api.records.ContainerState;
|
|
|
@@ -1077,6 +1078,126 @@ public class TestNodeStatusUpdater {
|
|
|
Assert.assertTrue(containerIdSet.contains(runningContainerId));
|
|
|
}
|
|
|
|
|
|
+ @Test(timeout = 90000)
|
|
|
+ public void testKilledQueuedContainers() throws Exception {
|
|
|
+ NodeManager nm = new NodeManager();
|
|
|
+ YarnConfiguration conf = new YarnConfiguration();
|
|
|
+ conf.set(
|
|
|
+ NodeStatusUpdaterImpl
|
|
|
+ .YARN_NODEMANAGER_DURATION_TO_TRACK_STOPPED_CONTAINERS,
|
|
|
+ "10000");
|
|
|
+ nm.init(conf);
|
|
|
+ NodeStatusUpdaterImpl nodeStatusUpdater =
|
|
|
+ (NodeStatusUpdaterImpl) nm.getNodeStatusUpdater();
|
|
|
+ ApplicationId appId = ApplicationId.newInstance(0, 0);
|
|
|
+ ApplicationAttemptId appAttemptId =
|
|
|
+ ApplicationAttemptId.newInstance(appId, 0);
|
|
|
+
|
|
|
+ // Add application to context.
|
|
|
+ nm.getNMContext().getApplications().putIfAbsent(appId,
|
|
|
+ mock(Application.class));
|
|
|
+
|
|
|
+ // Create a running container and add it to the context.
|
|
|
+ ContainerId runningContainerId =
|
|
|
+ ContainerId.newContainerId(appAttemptId, 1);
|
|
|
+ Token runningContainerToken =
|
|
|
+ BuilderUtils.newContainerToken(runningContainerId, "anyHost",
|
|
|
+ 1234, "anyUser", BuilderUtils.newResource(1024, 1), 0, 123,
|
|
|
+ "password".getBytes(), 0);
|
|
|
+ Container runningContainer =
|
|
|
+ new ContainerImpl(conf, null, null, null, null,
|
|
|
+ BuilderUtils.newContainerTokenIdentifier(runningContainerToken),
|
|
|
+ nm.getNMContext()) {
|
|
|
+ @Override
|
|
|
+ public ContainerState getCurrentState() {
|
|
|
+ return ContainerState.RUNNING;
|
|
|
+ }
|
|
|
+
|
|
|
+ @Override
|
|
|
+ public org.apache.hadoop.yarn.server.nodemanager.containermanager.
|
|
|
+ container.ContainerState getContainerState() {
|
|
|
+ return org.apache.hadoop.yarn.server.nodemanager.containermanager.
|
|
|
+ container.ContainerState.RUNNING;
|
|
|
+ }
|
|
|
+ };
|
|
|
+
|
|
|
+ nm.getNMContext().getContainers()
|
|
|
+ .put(runningContainerId, runningContainer);
|
|
|
+
|
|
|
+ // Create two killed queued containers and add them to the queuing context.
|
|
|
+ ContainerId killedQueuedContainerId1 = ContainerId.newContainerId(
|
|
|
+ appAttemptId, 2);
|
|
|
+ ContainerTokenIdentifier killedQueuedContainerTokenId1 = BuilderUtils
|
|
|
+ .newContainerTokenIdentifier(BuilderUtils.newContainerToken(
|
|
|
+ killedQueuedContainerId1, "anyHost", 1234, "anyUser", BuilderUtils
|
|
|
+ .newResource(1024, 1), 0, 123, "password".getBytes(), 0));
|
|
|
+ ContainerId killedQueuedContainerId2 = ContainerId.newContainerId(
|
|
|
+ appAttemptId, 3);
|
|
|
+ ContainerTokenIdentifier killedQueuedContainerTokenId2 = BuilderUtils
|
|
|
+ .newContainerTokenIdentifier(BuilderUtils.newContainerToken(
|
|
|
+ killedQueuedContainerId2, "anyHost", 1234, "anyUser", BuilderUtils
|
|
|
+ .newResource(1024, 1), 0, 123, "password".getBytes(), 0));
|
|
|
+
|
|
|
+ nm.getNMContext().getQueuingContext().getKilledQueuedContainers().put(
|
|
|
+ killedQueuedContainerTokenId1, "Queued container killed.");
|
|
|
+ nm.getNMContext().getQueuingContext().getKilledQueuedContainers().put(
|
|
|
+ killedQueuedContainerTokenId2, "Queued container killed.");
|
|
|
+
|
|
|
+ List<ContainerStatus> containerStatuses = nodeStatusUpdater
|
|
|
+ .getContainerStatuses();
|
|
|
+
|
|
|
+ Assert.assertEquals(3, containerStatuses.size());
|
|
|
+
|
|
|
+ ContainerStatus runningContainerStatus = null;
|
|
|
+ ContainerStatus killedQueuedContainerStatus1 = null;
|
|
|
+ ContainerStatus killedQueuedContainerStatus2 = null;
|
|
|
+ for (ContainerStatus cStatus : containerStatuses) {
|
|
|
+ if (ContainerState.RUNNING == cStatus.getState()) {
|
|
|
+ runningContainerStatus = cStatus;
|
|
|
+ }
|
|
|
+ if (ContainerState.COMPLETE == cStatus.getState()) {
|
|
|
+ if (killedQueuedContainerId1.equals(cStatus.getContainerId())) {
|
|
|
+ killedQueuedContainerStatus1 = cStatus;
|
|
|
+ } else {
|
|
|
+ killedQueuedContainerStatus2 = cStatus;
|
|
|
+ }
|
|
|
+ }
|
|
|
+ }
|
|
|
+
|
|
|
+ // Check container IDs and Container Status.
|
|
|
+ Assert.assertNotNull(runningContainerId);
|
|
|
+ Assert.assertNotNull(killedQueuedContainerId1);
|
|
|
+ Assert.assertNotNull(killedQueuedContainerId2);
|
|
|
+
|
|
|
+ // Killed queued container should have ABORTED exit status.
|
|
|
+ Assert.assertEquals(ContainerExitStatus.ABORTED,
|
|
|
+ killedQueuedContainerStatus1.getExitStatus());
|
|
|
+ Assert.assertEquals(ContainerExitStatus.ABORTED,
|
|
|
+ killedQueuedContainerStatus2.getExitStatus());
|
|
|
+
|
|
|
+ // Killed queued container should appear in the recentlyStoppedContainers.
|
|
|
+ Assert.assertTrue(nodeStatusUpdater.isContainerRecentlyStopped(
|
|
|
+ killedQueuedContainerId1));
|
|
|
+ Assert.assertTrue(nodeStatusUpdater.isContainerRecentlyStopped(
|
|
|
+ killedQueuedContainerId2));
|
|
|
+
|
|
|
+ // Check if killed queued containers are successfully removed from the
|
|
|
+ // queuing context.
|
|
|
+ List<ContainerId> ackedContainers = new ArrayList<ContainerId>();
|
|
|
+ ackedContainers.add(killedQueuedContainerId1);
|
|
|
+ ackedContainers.add(killedQueuedContainerId2);
|
|
|
+
|
|
|
+ nodeStatusUpdater.removeOrTrackCompletedContainersFromContext(
|
|
|
+ ackedContainers);
|
|
|
+
|
|
|
+ containerStatuses = nodeStatusUpdater.getContainerStatuses();
|
|
|
+
|
|
|
+ // Only the running container should be in the container statuses now.
|
|
|
+ Assert.assertEquals(1, containerStatuses.size());
|
|
|
+ Assert.assertEquals(ContainerState.RUNNING,
|
|
|
+ containerStatuses.get(0).getState());
|
|
|
+ }
|
|
|
+
|
|
|
@Test(timeout = 10000)
|
|
|
public void testCompletedContainersIsRecentlyStopped() throws Exception {
|
|
|
NodeManager nm = new NodeManager();
|
Arun Suresh