org.apache.hadoop.yarn.api.records.ContainerStatus Java Examples

@Override
public void onContainersCompleted(final List<ContainerStatus> statuses) {
	runAsync(() -> {
			log.debug("YARN ResourceManager reported the following containers completed: {}.", statuses);
			for (final ContainerStatus containerStatus : statuses) {

				final ResourceID resourceId = new ResourceID(containerStatus.getContainerId().toString());
				final YarnWorkerNode yarnWorkerNode = workerNodeMap.remove(resourceId);

				if (yarnWorkerNode != null) {
					// Container completed unexpectedly ~> start a new one
					requestYarnContainerIfRequired();
				}
				// Eagerly close the connection with task manager.
				closeTaskManagerConnection(resourceId, new Exception(containerStatus.getDiagnostics()));
			}
		}
	);
}
 

@Override
public AllocateResponse allocate(AllocateRequest request)
    throws YarnException, IOException {
  Assert.assertEquals("response ID mismatch",
      responseId, request.getResponseId());
  ++responseId;
  org.apache.hadoop.yarn.api.records.Token yarnToken = null;
  if (amToken != null) {
    yarnToken = org.apache.hadoop.yarn.api.records.Token.newInstance(
        amToken.getIdentifier(), amToken.getKind().toString(),
        amToken.getPassword(), amToken.getService().toString());
  }
  return AllocateResponse.newInstance(responseId,
      Collections.<ContainerStatus>emptyList(),
      Collections.<Container>emptyList(),
      Collections.<NodeReport>emptyList(),
      Resources.none(), null, 1, null,
      Collections.<NMToken>emptyList(),
      yarnToken,
      Collections.<ContainerResourceIncrease>emptyList(),
      Collections.<ContainerResourceDecrease>emptyList());
}
 

@Override
public synchronized void containerCompleted(Object task, ContainerStatus containerStatus) {
  // Inform the Containers about completion.
  AMContainer amContainer = appContext.getAllContainers().get(containerStatus.getContainerId());
  if (amContainer != null) {
    String message = null;
    int exitStatus = containerStatus.getExitStatus();
    if (exitStatus == ContainerExitStatus.PREEMPTED) {
      message = "Container preempted externally. ";
    } else if (exitStatus == ContainerExitStatus.DISKS_FAILED) {
      message = "Container disk failed. ";
    } else {
      message = "Container failed. ";
    }
    if (containerStatus.getDiagnostics() != null) {
      message += containerStatus.getDiagnostics();
    }
    sendEvent(new AMContainerEventCompleted(amContainer.getContainerId(), exitStatus, message));
  }
}
 

/**
 * Get a list of container statuses running on this NodeManager
 */
@Override
public GetContainerStatusesResponse getContainerStatuses(
    GetContainerStatusesRequest request) throws YarnException, IOException {

  List<ContainerStatus> succeededRequests = new ArrayList<ContainerStatus>();
  Map<ContainerId, SerializedException> failedRequests =
      new HashMap<ContainerId, SerializedException>();
  UserGroupInformation remoteUgi = getRemoteUgi();
  NMTokenIdentifier identifier = selectNMTokenIdentifier(remoteUgi);
  for (ContainerId id : request.getContainerIds()) {
    try {
      ContainerStatus status = getContainerStatusInternal(id, identifier);
      succeededRequests.add(status);
    } catch (YarnException e) {
      failedRequests.put(id, SerializedException.newInstance(e));
    }
  }
  return GetContainerStatusesResponse.newInstance(succeededRequests,
    failedRequests);
}
 

public static AllocateResponse newAllocateResponse(int responseId,
    List<ContainerStatus> completedContainers,
    List<Container> allocatedContainers, List<NodeReport> updatedNodes,
    Resource availResources, AMCommand command, int numClusterNodes,
    PreemptionMessage preempt) {
  AllocateResponse response = recordFactory
      .newRecordInstance(AllocateResponse.class);
  response.setNumClusterNodes(numClusterNodes);
  response.setResponseId(responseId);
  response.setCompletedContainersStatuses(completedContainers);
  response.setAllocatedContainers(allocatedContainers);
  response.setUpdatedNodes(updatedNodes);
  response.setAvailableResources(availResources);
  response.setAMCommand(command);
  response.setPreemptionMessage(preempt);

  return response;
}
 

private static void publishContainerEndEvent(
    final TimelineClient timelineClient, ContainerStatus container,
    String domainId, UserGroupInformation ugi) {
  final TimelineEntity entity = new TimelineEntity();
  entity.setEntityId(container.getContainerId().toString());
  entity.setEntityType(DSEntity.DS_CONTAINER.toString());
  entity.setDomainId(domainId);
  entity.addPrimaryFilter("user", ugi.getShortUserName());
  TimelineEvent event = new TimelineEvent();
  event.setTimestamp(System.currentTimeMillis());
  event.setEventType(DSEvent.DS_CONTAINER_END.toString());
  event.addEventInfo("State", container.getState().name());
  event.addEventInfo("Exit Status", container.getExitStatus());
  entity.addEvent(event);
  try {
    timelineClient.putEntities(entity);
  } catch (YarnException | IOException e) {
    LOG.error("Container end event could not be published for "
        + container.getContainerId().toString(), e);
  }
}
 

private NodeStatus getNodeStatus(int responseId) throws IOException {

    NodeHealthStatus nodeHealthStatus = this.context.getNodeHealthStatus();
    nodeHealthStatus.setHealthReport(healthChecker.getHealthReport());
    nodeHealthStatus.setIsNodeHealthy(healthChecker.isHealthy());
    nodeHealthStatus.setLastHealthReportTime(healthChecker
      .getLastHealthReportTime());
    if (LOG.isDebugEnabled()) {
      LOG.debug("Node's health-status : " + nodeHealthStatus.getIsNodeHealthy()
          + ", " + nodeHealthStatus.getHealthReport());
    }
    List<ContainerStatus> containersStatuses = getContainerStatuses();
    NodeStatus nodeStatus =
        NodeStatus.newInstance(nodeId, responseId, containersStatuses,
          createKeepAliveApplicationList(), nodeHealthStatus);

    return nodeStatus;
  }
 

private String getAMContainerCrashedDiagnostics(
    RMAppAttemptContainerFinishedEvent finishEvent) {
  ContainerStatus status = finishEvent.getContainerStatus();
  StringBuilder diagnosticsBuilder = new StringBuilder();
  diagnosticsBuilder.append("AM Container for ").append(
    finishEvent.getApplicationAttemptId()).append(
    " exited with ").append(" exitCode: ").append(status.getExitStatus()).
    append("\n");
  if (this.getTrackingUrl() != null) {
    diagnosticsBuilder.append("For more detailed output,").append(
      " check application tracking page:").append(
      this.getTrackingUrl()).append(
      "Then, click on links to logs of each attempt.\n");
  }
  diagnosticsBuilder.append("Diagnostics: ").append(status.getDiagnostics())
      .append("Failing this attempt");
  return diagnosticsBuilder.toString();
}
 

public static void waitForContainerState(ContainerManagementProtocol containerManager,
      ContainerId containerID, ContainerState finalState, int timeOutMax)
      throws InterruptedException, YarnException, IOException {
List<ContainerId> list = new ArrayList<ContainerId>();
list.add(containerID);
GetContainerStatusesRequest request =
    GetContainerStatusesRequest.newInstance(list);
ContainerStatus containerStatus =
    containerManager.getContainerStatuses(request).getContainerStatuses()
      .get(0);
int timeoutSecs = 0;
  while (!containerStatus.getState().equals(finalState)
      && timeoutSecs++ < timeOutMax) {
      Thread.sleep(1000);
      LOG.info("Waiting for container to get into state " + finalState
          + ". Current state is " + containerStatus.getState());
      containerStatus = containerManager.getContainerStatuses(request).getContainerStatuses().get(0);
    }
    LOG.info("Container state is " + containerStatus.getState());
    Assert.assertEquals("ContainerState is not correct (timedout)",
        finalState, containerStatus.getState());
  }
 

@Override
public ContainerStatus getContainerStatus(ContainerId containerId,
    NodeId nodeId) throws YarnException, IOException {

  ContainerManagementProtocolProxyData proxy = null;
  List<ContainerId> containerIds = new ArrayList<ContainerId>();
  containerIds.add(containerId);
  try {
    proxy = cmProxy.getProxy(nodeId.toString(), containerId);
    GetContainerStatusesResponse response =
        proxy.getContainerManagementProtocol().getContainerStatuses(
            GetContainerStatusesRequest.newInstance(containerIds));
    if (response.getFailedRequests() != null
        && response.getFailedRequests().containsKey(containerId)) {
      Throwable t =
          response.getFailedRequests().get(containerId).deSerialize();
      parseAndThrowException(t);
    }
    ContainerStatus containerStatus = response.getContainerStatuses().get(0);
    return containerStatus;
  } finally {
    if (proxy != null) {
      cmProxy.mayBeCloseProxy(proxy);
    }
  }
}
 

private NodeStatus getNodeStatus(int responseId) throws IOException {

    NodeHealthStatus nodeHealthStatus = this.context.getNodeHealthStatus();
    nodeHealthStatus.setHealthReport(healthChecker.getHealthReport());
    nodeHealthStatus.setIsNodeHealthy(healthChecker.isHealthy());
    nodeHealthStatus.setLastHealthReportTime(healthChecker
      .getLastHealthReportTime());
    if (LOG.isDebugEnabled()) {
      LOG.debug("Node's health-status : " + nodeHealthStatus.getIsNodeHealthy()
          + ", " + nodeHealthStatus.getHealthReport());
    }
    List<ContainerStatus> containersStatuses = getContainerStatuses();
    NodeStatus nodeStatus =
        NodeStatus.newInstance(nodeId, responseId, containersStatuses,
          createKeepAliveApplicationList(), nodeHealthStatus);

    return nodeStatus;
  }
 

private ContainerStatus getContainerStatusInternal(ContainerId containerID,
    NMTokenIdentifier nmTokenIdentifier) throws YarnException {
  String containerIDStr = containerID.toString();
  Container container = this.context.getContainers().get(containerID);

  LOG.info("Getting container-status for " + containerIDStr);
  authorizeGetAndStopContainerRequest(containerID, container, false,
    nmTokenIdentifier);

  if (container == null) {
    if (nodeStatusUpdater.isContainerRecentlyStopped(containerID)) {
      throw RPCUtil.getRemoteException("Container " + containerIDStr
        + " was recently stopped on node manager.");
    } else {
      throw RPCUtil.getRemoteException("Container " + containerIDStr
        + " is not handled by this NodeManager");
    }
  }
  ContainerStatus containerStatus = container.cloneAndGetContainerStatus();
  LOG.info("Returning " + containerStatus);
  return containerStatus;
}
 

private String getAMContainerCrashedDiagnostics(
    RMAppAttemptContainerFinishedEvent finishEvent) {
  ContainerStatus status = finishEvent.getContainerStatus();
  StringBuilder diagnosticsBuilder = new StringBuilder();
  diagnosticsBuilder.append("AM Container for ").append(
    finishEvent.getApplicationAttemptId()).append(
    " exited with ").append(" exitCode: ").append(status.getExitStatus()).
    append("\n");
  if (this.getTrackingUrl() != null) {
    diagnosticsBuilder.append("For more detailed output,").append(
      " check application tracking page:").append(
      this.getTrackingUrl()).append(
      "Then, click on links to logs of each attempt.\n");
  }
  diagnosticsBuilder.append("Diagnostics: ").append(status.getDiagnostics())
      .append("Failing this attempt");
  return diagnosticsBuilder.toString();
}
 

@Override
public void onContainerStatusReceived(ContainerId containerId, ContainerStatus containerStatus)
{
  LOG.debug("Container Status: id={}, status={}", containerId, containerStatus);
  if (containerStatus.getState() != ContainerState.RUNNING) {
    recoverContainer(containerId);
  }
}
 

@Override
public void onContainersCompleted(List<ContainerStatus> statuses) {
  completedContainers = statuses;
  // wait for containers to be taken before returning
  synchronized (completedContainers) {
    while (completedContainers != null) {
      try {
        completedContainers.wait();
      } catch (InterruptedException ex) {
        LOG.error("Interrupted during wait", ex);
      }
    }
  }
}
 

@VisibleForTesting
protected Resource getResourcesUnderUse(RMNodeStatusEvent statusEvent) {
  Resource usedResources = Resource.newInstance(0, 0);
  for (ContainerStatus status : statusEvent.getContainers()) {
    if (containerInUse(status)) {
      RMContainer rmContainer = yarnScheduler.getRMContainer(status.getContainerId());
      // (sdaingade) This check is needed as RMContainer information may not be populated
      // immediately after a RM restart.
      if (rmContainer != null) {
        Resources.addTo(usedResources, rmContainer.getAllocatedResource());
      }
    }
  }
  return usedResources;
}
 

@Override
public RMContainerState transition(RMContainerImpl container,
    RMContainerEvent event) {
  NMContainerStatus report =
      ((RMContainerRecoverEvent) event).getContainerReport();
  if (report.getContainerState().equals(ContainerState.COMPLETE)) {
    ContainerStatus status =
        ContainerStatus.newInstance(report.getContainerId(),
          report.getContainerState(), report.getDiagnostics(),
          report.getContainerExitStatus());

    new FinishedTransition().transition(container,
      new RMContainerFinishedEvent(container.containerId, status,
        RMContainerEventType.FINISHED));
    return RMContainerState.COMPLETED;
  } else if (report.getContainerState().equals(ContainerState.RUNNING)) {
    // Tell the app
    container.eventHandler.handle(new RMAppRunningOnNodeEvent(container
        .getApplicationAttemptId().getApplicationId(), container.nodeId));
    return RMContainerState.RUNNING;
  } else {
    // This can never happen.
    LOG.warn("RMContainer received unexpected recover event with container"
        + " state " + report.getContainerState() + " while recovering.");
    return RMContainerState.RUNNING;
  }
}
 

public ContainerInfo(final Context nmContext, final Container container,
     String requestUri, String pathPrefix) {

  this.id = container.getContainerId().toString();
  this.nodeId = nmContext.getNodeId().toString();
  ContainerStatus containerData = container.cloneAndGetContainerStatus();
  this.exitCode = containerData.getExitStatus();
  this.exitStatus =
      (this.exitCode == ContainerExitStatus.INVALID) ?
          "N/A" : String.valueOf(exitCode);
  this.state = container.getContainerState().toString();
  this.diagnostics = containerData.getDiagnostics();
  if (this.diagnostics == null || this.diagnostics.isEmpty()) {
    this.diagnostics = "";
  }

  this.user = container.getUser();
  Resource res = container.getResource();
  if (res != null) {
    this.totalMemoryNeededMB = res.getMemory();
    this.totalVCoresNeeded = res.getVirtualCores();
  }
  this.containerLogsShortLink = ujoin("containerlogs", this.id,
      container.getUser());

  if (requestUri == null) {
    requestUri = "";
  }
  if (pathPrefix == null) {
    pathPrefix = "";
  }
  this.containerLogsLink = join(requestUri, pathPrefix,
      this.containerLogsShortLink);
}
 

@VisibleForTesting
public TaskAttemptEvent createContainerFinishedEvent(ContainerStatus cont,
    TaskAttemptId attemptID) {
  if (cont.getExitStatus() == ContainerExitStatus.ABORTED
      || cont.getExitStatus() == ContainerExitStatus.PREEMPTED) {
    // killed by framework
    return new TaskAttemptEvent(attemptID,
        TaskAttemptEventType.TA_KILL);
  } else {
    return new TaskAttemptEvent(attemptID,
        TaskAttemptEventType.TA_CONTAINER_COMPLETED);
  }
}
 

public RMNodeStatusEvent(NodeId nodeId, NodeHealthStatus nodeHealthStatus,
    List<ContainerStatus> collection, List<ApplicationId> keepAliveAppIds,
    NodeHeartbeatResponse latestResponse) {
  super(nodeId, RMNodeEventType.STATUS_UPDATE);
  this.nodeHealthStatus = nodeHealthStatus;
  this.containersCollection = collection;
  this.keepAliveAppIds = keepAliveAppIds;
  this.latestResponse = latestResponse;
}
 

public static NodeStatus newInstance(NodeId nodeId, int responseId,
    List<ContainerStatus> containerStatuses,
    List<ApplicationId> keepAliveApplications,
    NodeHealthStatus nodeHealthStatus) {
  NodeStatus nodeStatus = Records.newRecord(NodeStatus.class);
  nodeStatus.setResponseId(responseId);
  nodeStatus.setNodeId(nodeId);
  nodeStatus.setContainersStatuses(containerStatuses);
  nodeStatus.setKeepAliveApplications(keepAliveApplications);
  nodeStatus.setNodeHealthStatus(nodeHealthStatus);
  return nodeStatus;
}
 

public static Container getMockContainer(ContainerStatus containerStatus) {
  ContainerImpl container = mock(ContainerImpl.class);
  when(container.cloneAndGetContainerStatus()).thenReturn(containerStatus);
  when(container.getCurrentState()).thenReturn(containerStatus.getState());
  when(container.getContainerId()).thenReturn(
    containerStatus.getContainerId());
  if (containerStatus.getState().equals(ContainerState.COMPLETE)) {
    when(container.getContainerState())
      .thenReturn(org.apache.hadoop.yarn.server.nodemanager.containermanager.container.ContainerState.DONE);
  } else if (containerStatus.getState().equals(ContainerState.RUNNING)) {
    when(container.getContainerState())
    .thenReturn(org.apache.hadoop.yarn.server.nodemanager.containermanager.container.ContainerState.RUNNING);
  }
  return container;
}
 

private synchronized void initLocalFinishedContainerList() {
  if (this.completedContainersStatuses != null) {
    return;
  }
  AllocateResponseProtoOrBuilder p = viaProto ? proto : builder;
  List<ContainerStatusProto> list = p.getCompletedContainerStatusesList();
  completedContainersStatuses = new ArrayList<ContainerStatus>();

  for (ContainerStatusProto c : list) {
    completedContainersStatuses.add(convertFromProtoFormat(c));
  }
}
 

/**
 * Helper method to handle received ContainerStatus. If this corresponds to
 * the completion of a master-container of a managed AM,
 * we call the handler for RMAppAttemptContainerFinishedEvent.
 */
@SuppressWarnings("unchecked")
@VisibleForTesting
void handleNMContainerStatus(NMContainerStatus containerStatus, NodeId nodeId) {
  ApplicationAttemptId appAttemptId =
      containerStatus.getContainerId().getApplicationAttemptId();
  RMApp rmApp =
      rmContext.getRMApps().get(appAttemptId.getApplicationId());
  if (rmApp == null) {
    LOG.error("Received finished container : "
        + containerStatus.getContainerId()
        + " for unknown application " + appAttemptId.getApplicationId()
        + " Skipping.");
    return;
  }

  if (rmApp.getApplicationSubmissionContext().getUnmanagedAM()) {
    if (LOG.isDebugEnabled()) {
      LOG.debug("Ignoring container completion status for unmanaged AM "
          + rmApp.getApplicationId());
    }
    return;
  }

  RMAppAttempt rmAppAttempt = rmApp.getRMAppAttempt(appAttemptId);
  Container masterContainer = rmAppAttempt.getMasterContainer();
  if (masterContainer.getId().equals(containerStatus.getContainerId())
      && containerStatus.getContainerState() == ContainerState.COMPLETE) {
    ContainerStatus status =
        ContainerStatus.newInstance(containerStatus.getContainerId(),
          containerStatus.getContainerState(), containerStatus.getDiagnostics(),
          containerStatus.getContainerExitStatus());
    // sending master container finished event.
    RMAppAttemptContainerFinishedEvent evt =
        new RMAppAttemptContainerFinishedEvent(appAttemptId, status,
            nodeId);
    rmContext.getDispatcher().getEventHandler().handle(evt);
  }
}
 

private AllocateResponse createAllocateResponse(
    List<ContainerStatus> completed, List<Container> allocated,
    List<NMToken> nmTokens) {
  AllocateResponse response =
      AllocateResponse.newInstance(0, completed, allocated,
          new ArrayList<NodeReport>(), null, null, 1, null, nmTokens);
  return response;
}
 

private void testUnmanagedAMSuccess(String url) {
  unmanagedAM = true;
  when(submissionContext.getUnmanagedAM()).thenReturn(true);
  // submit AM and check it goes to LAUNCHED state
  scheduleApplicationAttempt();
  testAppAttemptLaunchedState(null);
  verify(amLivelinessMonitor, times(1)).register(
      applicationAttempt.getAppAttemptId());

  // launch AM
  runApplicationAttempt(null, "host", 8042, url, true);

  // complete a container
  Container container = mock(Container.class);
  when(container.getNodeId()).thenReturn(NodeId.newInstance("host", 1234));
  application.handle(new RMAppRunningOnNodeEvent(application.getApplicationId(),
      container.getNodeId()));
  applicationAttempt.handle(new RMAppAttemptContainerFinishedEvent(
      applicationAttempt.getAppAttemptId(), mock(ContainerStatus.class),
      container.getNodeId()));
  // complete AM
  String diagnostics = "Successful";
  FinalApplicationStatus finalStatus = FinalApplicationStatus.SUCCEEDED;
  applicationAttempt.handle(new RMAppAttemptUnregistrationEvent(
      applicationAttempt.getAppAttemptId(), url, finalStatus,
      diagnostics));
  testAppAttemptFinishedState(null, finalStatus, url, diagnostics, 1,
      true);
  assertFalse(transferStateFromPreviousAttempt);
}
 

@Override
public void
    transition(RMAppAttemptImpl appAttempt, RMAppAttemptEvent event) {
  RMAppAttemptContainerFinishedEvent containerFinishedEvent =
      (RMAppAttemptContainerFinishedEvent) event;
  ContainerStatus containerStatus =
      containerFinishedEvent.getContainerStatus();

  // If this is the AM container, it means the AM container is finished,
  // but we are not yet acknowledged that the final state has been saved.
  // Thus, we still return FINAL_SAVING state here.
  if (appAttempt.masterContainer.getId().equals(
    containerStatus.getContainerId())) {
    appAttempt.sendAMContainerToNM(appAttempt, containerFinishedEvent);

    if (appAttempt.targetedFinalState.equals(RMAppAttemptState.FAILED)
        || appAttempt.targetedFinalState.equals(RMAppAttemptState.KILLED)) {
      // ignore Container_Finished Event if we were supposed to reach
      // FAILED/KILLED state.
      return;
    }

    // pass in the earlier AMUnregistered Event also, as this is needed for
    // AMFinishedAfterFinalSavingTransition later on
    appAttempt.rememberTargetTransitions(event,
      new AMFinishedAfterFinalSavingTransition(
        appAttempt.eventCausingFinalSaving), RMAppAttemptState.FINISHED);
    return;
  }

  // Add all finished containers so that they can be acked to NM.
  addJustFinishedContainer(appAttempt, containerFinishedEvent);
}
 

@Private
@Unstable
public static GetContainerStatusesResponse newInstance(
    List<ContainerStatus> statuses,
    Map<ContainerId, SerializedException> failedRequests) {
  GetContainerStatusesResponse response =
      Records.newRecord(GetContainerStatusesResponse.class);
  response.setContainerStatuses(statuses);
  response.setFailedRequests(failedRequests);
  return response;
}
 

@Override
public void onContainerStatusReceived(ContainerId containerId, ContainerStatus containerStatus) {
  if (eventSubmitter.isPresent()) {
    eventSubmitter.get().submit(GobblinYarnEventConstants.EventNames.CONTAINER_STATUS_RECEIVED,
        buildContainerStatusEventMetadata(containerStatus).build());
  }

  LOGGER.info(String.format("Received container status for container %s: %s", containerId, containerStatus));
}
 

@Override
public void beforeCompletedContainer(RMContainer rmContainer, ContainerStatus containerStatus, RMContainerEventType event) {
  if (rmContainer != null && rmContainer.getContainer() != null) {
    NodeId nodeId = rmContainer.getContainer().getNodeId();
    for (YarnSchedulerInterceptor interceptor : interceptors.values()) {
      if (interceptor.getCallBackFilter().allowCallBacksForNode(nodeId)) {
        interceptor.beforeCompletedContainer(rmContainer, containerStatus, event);
      }
    }
  }
}