Search in sources :

Example 26 with TezTaskAttemptID

use of org.apache.tez.dag.records.TezTaskAttemptID in project tez by apache.

the class TestTaskAttempt method testContainerTerminatedAfterSuccess.

@Test(timeout = 5000)
public // the TaskAttempt
void testContainerTerminatedAfterSuccess() throws Exception {
    ApplicationId appId = ApplicationId.newInstance(1, 2);
    ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0);
    TezDAGID dagID = TezDAGID.getInstance(appId, 1);
    TezVertexID vertexID = TezVertexID.getInstance(dagID, 1);
    TezTaskID taskID = TezTaskID.getInstance(vertexID, 1);
    MockEventHandler eventHandler = spy(new MockEventHandler());
    TaskCommunicatorManagerInterface taListener = createMockTaskAttemptListener();
    Configuration taskConf = new Configuration();
    taskConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class);
    taskConf.setBoolean("fs.file.impl.disable.cache", true);
    locationHint = TaskLocationHint.createTaskLocationHint(new HashSet<String>(Arrays.asList(new String[] { "127.0.0.1" })), null);
    Resource resource = Resource.newInstance(1024, 1);
    NodeId nid = NodeId.newInstance("127.0.0.1", 0);
    @SuppressWarnings("deprecation") ContainerId contId = ContainerId.newInstance(appAttemptId, 3);
    Container container = mock(Container.class);
    when(container.getId()).thenReturn(contId);
    when(container.getNodeId()).thenReturn(nid);
    when(container.getNodeHttpAddress()).thenReturn("localhost:0");
    AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskCommunicatorManagerInterface.class), new ContainerContextMatcher(), appCtx);
    containers.addContainerIfNew(container, 0, 0, 0);
    doReturn(new ClusterInfo()).when(appCtx).getClusterInfo();
    doReturn(containers).when(appCtx).getAllContainers();
    TaskHeartbeatHandler mockHeartbeatHandler = mock(TaskHeartbeatHandler.class);
    TaskAttemptImpl taImpl = new MockTaskAttemptImpl(taskID, 1, eventHandler, taListener, taskConf, new SystemClock(), mockHeartbeatHandler, appCtx, false, resource, createFakeContainerContext(), false);
    TezTaskAttemptID taskAttemptID = taImpl.getID();
    ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class);
    taImpl.handle(new TaskAttemptEventSchedule(taskAttemptID, 0, 0));
    taImpl.handle(new TaskAttemptEventSubmitted(taskAttemptID, contId));
    taImpl.handle(new TaskAttemptEventStartedRemotely(taskAttemptID));
    assertEquals("Task attempt is not in the RUNNING state", taImpl.getState(), TaskAttemptState.RUNNING);
    verify(mockHeartbeatHandler).register(taskAttemptID);
    int expectedEventsAtRunning = 5;
    verify(eventHandler, times(expectedEventsAtRunning)).handle(arg.capture());
    taImpl.handle(new TaskAttemptEvent(taskAttemptID, TaskAttemptEventType.TA_DONE));
    assertEquals("Task attempt is not in the  SUCCEEDED state", taImpl.getState(), TaskAttemptState.SUCCEEDED);
    verify(mockHeartbeatHandler).unregister(taskAttemptID);
    assertEquals(0, taImpl.getDiagnostics().size());
    int expectedEvenstAfterTerminating = expectedEventsAtRunning + 3;
    arg = ArgumentCaptor.forClass(Event.class);
    verify(eventHandler, times(expectedEvenstAfterTerminating)).handle(arg.capture());
    verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), TaskEventTASucceeded.class, 1);
    verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), AMSchedulerEventTAEnded.class, 1);
    verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), DAGEventCounterUpdate.class, 1);
    taImpl.handle(new TaskAttemptEventContainerTerminated(contId, taskAttemptID, "Terminated", TaskAttemptTerminationCause.CONTAINER_EXITED));
    // verify unregister is not invoked again
    verify(mockHeartbeatHandler, times(1)).unregister(taskAttemptID);
    int expectedEventAfterTerminated = expectedEvenstAfterTerminating + 0;
    arg = ArgumentCaptor.forClass(Event.class);
    verify(eventHandler, times(expectedEventAfterTerminated)).handle(arg.capture());
    // Verify that the diagnostic message included in the Terminated event is not
    // captured - TA already succeeded. Error cause is the default value.
    assertEquals(0, taImpl.getDiagnostics().size());
    assertEquals(TaskAttemptTerminationCause.UNKNOWN_ERROR, taImpl.getTerminationCause());
}
Also used : Configuration(org.apache.hadoop.conf.Configuration) TezConfiguration(org.apache.tez.dag.api.TezConfiguration) AMContainerMap(org.apache.tez.dag.app.rm.container.AMContainerMap) Container(org.apache.hadoop.yarn.api.records.Container) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) TezDAGID(org.apache.tez.dag.records.TezDAGID) TaskHeartbeatHandler(org.apache.tez.dag.app.TaskHeartbeatHandler) TaskAttemptEventSchedule(org.apache.tez.dag.app.dag.event.TaskAttemptEventSchedule) TezVertexID(org.apache.tez.dag.records.TezVertexID) ContainerHeartbeatHandler(org.apache.tez.dag.app.ContainerHeartbeatHandler) HashSet(java.util.HashSet) TaskAttemptEventStartedRemotely(org.apache.tez.dag.app.dag.event.TaskAttemptEventStartedRemotely) SystemClock(org.apache.hadoop.yarn.util.SystemClock) Resource(org.apache.hadoop.yarn.api.records.Resource) LocalResource(org.apache.hadoop.yarn.api.records.LocalResource) TaskAttemptEvent(org.apache.tez.dag.app.dag.event.TaskAttemptEvent) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) TaskCommunicatorManagerInterface(org.apache.tez.dag.app.TaskCommunicatorManagerInterface) ContainerContextMatcher(org.apache.tez.dag.app.rm.container.ContainerContextMatcher) TaskAttemptEventSubmitted(org.apache.tez.dag.app.dag.event.TaskAttemptEventSubmitted) TezTaskID(org.apache.tez.dag.records.TezTaskID) TaskLocationHint(org.apache.tez.dag.api.TaskLocationHint) ClusterInfo(org.apache.tez.dag.app.ClusterInfo) NodeId(org.apache.hadoop.yarn.api.records.NodeId) TaskStatusUpdateEvent(org.apache.tez.runtime.api.events.TaskStatusUpdateEvent) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) TaskEvent(org.apache.tez.dag.app.dag.event.TaskEvent) DAGEvent(org.apache.tez.dag.app.dag.event.DAGEvent) InputReadErrorEvent(org.apache.tez.runtime.api.events.InputReadErrorEvent) TaskAttemptEvent(org.apache.tez.dag.app.dag.event.TaskAttemptEvent) Event(org.apache.hadoop.yarn.event.Event) DAGHistoryEvent(org.apache.tez.dag.history.DAGHistoryEvent) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) TaskAttemptEventContainerTerminated(org.apache.tez.dag.app.dag.event.TaskAttemptEventContainerTerminated) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) TezTaskAttemptID(org.apache.tez.dag.records.TezTaskAttemptID) Test(org.junit.Test)

Example 27 with TezTaskAttemptID

use of org.apache.tez.dag.records.TezTaskAttemptID in project tez by apache.

the class TestTaskAttempt method testCompletedAtSubmitted.

@Test(timeout = 5000)
public void testCompletedAtSubmitted() throws ServicePluginException {
    ApplicationId appId = ApplicationId.newInstance(1, 2);
    ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0);
    TezDAGID dagID = TezDAGID.getInstance(appId, 1);
    TezVertexID vertexID = TezVertexID.getInstance(dagID, 1);
    TezTaskID taskID = TezTaskID.getInstance(vertexID, 1);
    MockEventHandler eventHandler = spy(new MockEventHandler());
    TaskCommunicatorManagerInterface taListener = createMockTaskAttemptListener();
    Configuration taskConf = new Configuration();
    taskConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class);
    taskConf.setBoolean("fs.file.impl.disable.cache", true);
    locationHint = TaskLocationHint.createTaskLocationHint(new HashSet<String>(Arrays.asList(new String[] { "127.0.0.1" })), null);
    Resource resource = Resource.newInstance(1024, 1);
    NodeId nid = NodeId.newInstance("127.0.0.1", 0);
    @SuppressWarnings("deprecation") ContainerId contId = ContainerId.newInstance(appAttemptId, 3);
    Container container = mock(Container.class);
    when(container.getId()).thenReturn(contId);
    when(container.getNodeId()).thenReturn(nid);
    when(container.getNodeHttpAddress()).thenReturn("localhost:0");
    AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskCommunicatorManagerInterface.class), new ContainerContextMatcher(), appCtx);
    containers.addContainerIfNew(container, 0, 0, 0);
    doReturn(new ClusterInfo()).when(appCtx).getClusterInfo();
    doReturn(containers).when(appCtx).getAllContainers();
    TaskHeartbeatHandler mockHeartbeatHandler = mock(TaskHeartbeatHandler.class);
    TaskAttemptImpl taImpl = new MockTaskAttemptImpl(taskID, 1, eventHandler, taListener, taskConf, new SystemClock(), mockHeartbeatHandler, appCtx, false, resource, createFakeContainerContext(), false);
    TezTaskAttemptID taskAttemptID = taImpl.getID();
    ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class);
    taImpl.handle(new TaskAttemptEventSchedule(taskAttemptID, 0, 0));
    taImpl.handle(new TaskAttemptEventSubmitted(taskAttemptID, contId));
    assertEquals("Task attempt is not in the RUNNING state", taImpl.getState(), TaskAttemptState.STARTING);
    verify(mockHeartbeatHandler).register(taskAttemptID);
    int expectedEventsAtStarting = 4;
    verify(eventHandler, times(expectedEventsAtStarting)).handle(arg.capture());
    // Ensure status_updates are handled in the submitted state.
    taImpl.handle(new TaskAttemptEventStatusUpdate(taskAttemptID, new TaskStatusUpdateEvent(null, 0.1f, null, false)));
    taImpl.handle(new TaskAttemptEvent(taskAttemptID, TaskAttemptEventType.TA_DONE));
    assertEquals("Task attempt is not in the  SUCCEEDED state", taImpl.getState(), TaskAttemptState.SUCCEEDED);
    verify(mockHeartbeatHandler).unregister(taskAttemptID);
    assertEquals(0, taImpl.getDiagnostics().size());
    int expectedEvenstAfterTerminating = expectedEventsAtStarting + 3;
    arg = ArgumentCaptor.forClass(Event.class);
    verify(eventHandler, times(expectedEvenstAfterTerminating)).handle(arg.capture());
    Event e = verifyEventType(arg.getAllValues().subList(expectedEventsAtStarting, expectedEvenstAfterTerminating), TaskEventTASucceeded.class, 1);
    assertEquals(TaskEventType.T_ATTEMPT_SUCCEEDED, e.getType());
    verifyEventType(arg.getAllValues().subList(expectedEventsAtStarting, expectedEvenstAfterTerminating), AMSchedulerEventTAEnded.class, 1);
    verifyEventType(arg.getAllValues().subList(expectedEventsAtStarting, expectedEvenstAfterTerminating), DAGEventCounterUpdate.class, 1);
}
Also used : Configuration(org.apache.hadoop.conf.Configuration) TezConfiguration(org.apache.tez.dag.api.TezConfiguration) TaskAttemptEventStatusUpdate(org.apache.tez.dag.app.dag.event.TaskAttemptEventStatusUpdate) TaskStatusUpdateEvent(org.apache.tez.runtime.api.events.TaskStatusUpdateEvent) AMContainerMap(org.apache.tez.dag.app.rm.container.AMContainerMap) Container(org.apache.hadoop.yarn.api.records.Container) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) TezDAGID(org.apache.tez.dag.records.TezDAGID) TaskHeartbeatHandler(org.apache.tez.dag.app.TaskHeartbeatHandler) TaskAttemptEventSchedule(org.apache.tez.dag.app.dag.event.TaskAttemptEventSchedule) TezVertexID(org.apache.tez.dag.records.TezVertexID) ContainerHeartbeatHandler(org.apache.tez.dag.app.ContainerHeartbeatHandler) HashSet(java.util.HashSet) SystemClock(org.apache.hadoop.yarn.util.SystemClock) Resource(org.apache.hadoop.yarn.api.records.Resource) LocalResource(org.apache.hadoop.yarn.api.records.LocalResource) TaskAttemptEvent(org.apache.tez.dag.app.dag.event.TaskAttemptEvent) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) TaskCommunicatorManagerInterface(org.apache.tez.dag.app.TaskCommunicatorManagerInterface) ContainerContextMatcher(org.apache.tez.dag.app.rm.container.ContainerContextMatcher) TaskAttemptEventSubmitted(org.apache.tez.dag.app.dag.event.TaskAttemptEventSubmitted) TezTaskID(org.apache.tez.dag.records.TezTaskID) TaskLocationHint(org.apache.tez.dag.api.TaskLocationHint) ClusterInfo(org.apache.tez.dag.app.ClusterInfo) NodeId(org.apache.hadoop.yarn.api.records.NodeId) TaskStatusUpdateEvent(org.apache.tez.runtime.api.events.TaskStatusUpdateEvent) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) TaskEvent(org.apache.tez.dag.app.dag.event.TaskEvent) DAGEvent(org.apache.tez.dag.app.dag.event.DAGEvent) InputReadErrorEvent(org.apache.tez.runtime.api.events.InputReadErrorEvent) TaskAttemptEvent(org.apache.tez.dag.app.dag.event.TaskAttemptEvent) Event(org.apache.hadoop.yarn.event.Event) DAGHistoryEvent(org.apache.tez.dag.history.DAGHistoryEvent) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) TezTaskAttemptID(org.apache.tez.dag.records.TezTaskAttemptID) Test(org.junit.Test)

Example 28 with TezTaskAttemptID

use of org.apache.tez.dag.records.TezTaskAttemptID in project tez by apache.

the class TestTaskImpl method testTaskSucceedAndRetroActiveFailure.

@SuppressWarnings("rawtypes")
@Test(timeout = 5000)
public void testTaskSucceedAndRetroActiveFailure() {
    TezTaskID taskId = getNewTaskID();
    scheduleTaskAttempt(taskId);
    launchTaskAttempt(mockTask.getLastAttempt().getID());
    updateAttemptState(mockTask.getLastAttempt(), TaskAttemptState.RUNNING);
    mockTask.handle(createTaskTASucceededEvent(mockTask.getLastAttempt().getID()));
    // The task should now have succeeded
    assertTaskSucceededState();
    verify(mockTask.stateChangeNotifier).taskSucceeded(any(String.class), eq(taskId), eq(mockTask.getLastAttempt().getID().getId()));
    ArgumentCaptor<DAGHistoryEvent> argumentCaptor = ArgumentCaptor.forClass(DAGHistoryEvent.class);
    verify(mockHistoryHandler).handle(argumentCaptor.capture());
    DAGHistoryEvent dagHistoryEvent = argumentCaptor.getValue();
    HistoryEvent historyEvent = dagHistoryEvent.getHistoryEvent();
    assertTrue(historyEvent instanceof TaskFinishedEvent);
    TaskFinishedEvent taskFinishedEvent = (TaskFinishedEvent) historyEvent;
    assertEquals(taskFinishedEvent.getFinishTime(), mockTask.getFinishTime());
    eventHandler.events.clear();
    // Now fail the attempt after it has succeeded
    TezTaskAttemptID mockDestId = mock(TezTaskAttemptID.class);
    TezEvent mockTezEvent = mock(TezEvent.class);
    EventMetaData meta = new EventMetaData(EventProducerConsumerType.INPUT, "Vertex", "Edge", mockDestId);
    when(mockTezEvent.getSourceInfo()).thenReturn(meta);
    TaskAttemptEventOutputFailed outputFailedEvent = new TaskAttemptEventOutputFailed(mockDestId, mockTezEvent, 1);
    mockTask.handle(createTaskTAFailedEvent(mockTask.getLastAttempt().getID(), TaskFailureType.NON_FATAL, outputFailedEvent));
    // The task should still be in the scheduled state
    assertTaskScheduledState();
    Event event = eventHandler.events.get(0);
    Assert.assertEquals(AMNodeEventType.N_TA_ENDED, event.getType());
    event = eventHandler.events.get(eventHandler.events.size() - 1);
    Assert.assertEquals(VertexEventType.V_TASK_RESCHEDULED, event.getType());
    // report of output read error should be the causal TA
    List<MockTaskAttemptImpl> attempts = mockTask.getAttemptList();
    Assert.assertEquals(2, attempts.size());
    MockTaskAttemptImpl newAttempt = attempts.get(1);
    Assert.assertEquals(mockDestId, newAttempt.getSchedulingCausalTA());
}
Also used : DAGHistoryEvent(org.apache.tez.dag.history.DAGHistoryEvent) HistoryEvent(org.apache.tez.dag.history.HistoryEvent) DAGHistoryEvent(org.apache.tez.dag.history.DAGHistoryEvent) TezTaskID(org.apache.tez.dag.records.TezTaskID) TaskFinishedEvent(org.apache.tez.dag.history.events.TaskFinishedEvent) TaskAttemptEventOutputFailed(org.apache.tez.dag.app.dag.event.TaskAttemptEventOutputFailed) TaskFinishedEvent(org.apache.tez.dag.history.events.TaskFinishedEvent) TaskAttemptEvent(org.apache.tez.dag.app.dag.event.TaskAttemptEvent) Event(org.apache.hadoop.yarn.event.Event) HistoryEvent(org.apache.tez.dag.history.HistoryEvent) TezAbstractEvent(org.apache.tez.common.TezAbstractEvent) DataMovementEvent(org.apache.tez.runtime.api.events.DataMovementEvent) TaskEvent(org.apache.tez.dag.app.dag.event.TaskEvent) DAGHistoryEvent(org.apache.tez.dag.history.DAGHistoryEvent) InputReadErrorEvent(org.apache.tez.runtime.api.events.InputReadErrorEvent) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) TaskAttemptEventTerminationCauseEvent(org.apache.tez.dag.app.dag.event.TaskAttemptEventTerminationCauseEvent) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) EventMetaData(org.apache.tez.runtime.api.impl.EventMetaData) TezTaskAttemptID(org.apache.tez.dag.records.TezTaskAttemptID) Test(org.junit.Test)

Example 29 with TezTaskAttemptID

use of org.apache.tez.dag.records.TezTaskAttemptID in project tez by apache.

the class TestTaskImpl method testFetchedEventsModifyUnderlyingList.

@Test(timeout = 5000)
public void testFetchedEventsModifyUnderlyingList() {
    // Tests to ensure that adding an event to a task, does not affect the
    // result of past getTaskAttemptTezEvents calls.
    List<TezEvent> fetchedList;
    TezTaskID taskId = getNewTaskID();
    scheduleTaskAttempt(taskId);
    sendTezEventsToTask(taskId, 2);
    TezTaskAttemptID attemptID = mockTask.getAttemptList().iterator().next().getID();
    fetchedList = mockTask.getTaskAttemptTezEvents(attemptID, 0, 100);
    assertEquals(2, fetchedList.size());
    // Add events, make sure underlying list is the same, and no exceptions are
    // thrown while accessing the previous list
    sendTezEventsToTask(taskId, 4);
    assertEquals(2, fetchedList.size());
    fetchedList = mockTask.getTaskAttemptTezEvents(attemptID, 0, 100);
    assertEquals(6, fetchedList.size());
}
Also used : TezEvent(org.apache.tez.runtime.api.impl.TezEvent) TezTaskID(org.apache.tez.dag.records.TezTaskID) TezTaskAttemptID(org.apache.tez.dag.records.TezTaskAttemptID) Test(org.junit.Test)

Example 30 with TezTaskAttemptID

use of org.apache.tez.dag.records.TezTaskAttemptID in project tez by apache.

the class TestTaskImpl method testKillRunningTaskAttempt.

@Test(timeout = 5000)
public /**
 * Kill running attempt
 * {@link TaskState#RUNNING}->{@link TaskState#RUNNING}
 */
void testKillRunningTaskAttempt() {
    LOG.info("--- START: testKillRunningTaskAttempt ---");
    TezTaskID taskId = getNewTaskID();
    scheduleTaskAttempt(taskId);
    TezTaskAttemptID lastTAId = mockTask.getLastAttempt().getID();
    launchTaskAttempt(mockTask.getLastAttempt().getID());
    killRunningTaskAttempt(mockTask.getLastAttempt().getID());
    // last killed attempt should be causal TA of next attempt
    Assert.assertEquals(lastTAId, mockTask.getLastAttempt().getSchedulingCausalTA());
}
Also used : TezTaskID(org.apache.tez.dag.records.TezTaskID) TezTaskAttemptID(org.apache.tez.dag.records.TezTaskAttemptID) Test(org.junit.Test)

Aggregations

TezTaskAttemptID (org.apache.tez.dag.records.TezTaskAttemptID)167 Test (org.junit.Test)124 TezTaskID (org.apache.tez.dag.records.TezTaskID)61 TezVertexID (org.apache.tez.dag.records.TezVertexID)54 Container (org.apache.hadoop.yarn.api.records.Container)48 TezEvent (org.apache.tez.runtime.api.impl.TezEvent)46 TezDAGID (org.apache.tez.dag.records.TezDAGID)43 Configuration (org.apache.hadoop.conf.Configuration)42 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)41 Priority (org.apache.hadoop.yarn.api.records.Priority)41 TezConfiguration (org.apache.tez.dag.api.TezConfiguration)41 LocalResource (org.apache.hadoop.yarn.api.records.LocalResource)33 Resource (org.apache.hadoop.yarn.api.records.Resource)30 TaskCommunicatorManagerInterface (org.apache.tez.dag.app.TaskCommunicatorManagerInterface)28 EventMetaData (org.apache.tez.runtime.api.impl.EventMetaData)28 ClusterInfo (org.apache.tez.dag.app.ClusterInfo)27 ContainerHeartbeatHandler (org.apache.tez.dag.app.ContainerHeartbeatHandler)27 AMContainerMap (org.apache.tez.dag.app.rm.container.AMContainerMap)27 ContainerContextMatcher (org.apache.tez.dag.app.rm.container.ContainerContextMatcher)27 ApplicationId (org.apache.hadoop.yarn.api.records.ApplicationId)25