Search in sources :

Example 6 with TaskAttemptRecoveryData

use of org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData in project tez by apache.

the class TestDAGRecovery method testTaskRecoverFromSucceeded.

/**
 * RecoveryEvent: TaskStartedEvent -> TaskFinishedEvent
 * Recover it to Scheduled
 */
@Test(timeout = 5000)
public void testTaskRecoverFromSucceeded() {
    initMockDAGRecoveryDataForTask();
    TaskStartedEvent taskStartedEvent = new TaskStartedEvent(t1v1Id, "v1", 0L, 0L);
    TaskFinishedEvent taskFinishedEvent = new TaskFinishedEvent(t1v1Id, "v1", 0L, 0L, null, TaskState.SUCCEEDED, "", null, 4);
    TaskAttemptStartedEvent taStartedEvent = new TaskAttemptStartedEvent(ta1t1v1Id, "v1", 0L, mock(ContainerId.class), mock(NodeId.class), "", "", "");
    List<TezEvent> taGeneratedEvents = new ArrayList<TezEvent>();
    EventMetaData metadata = new EventMetaData(EventProducerConsumerType.OUTPUT, "vertex1", "vertex3", ta1t1v2Id);
    taGeneratedEvents.add(new TezEvent(DataMovementEvent.create(ByteBuffer.wrap(new byte[0])), metadata));
    TaskAttemptFinishedEvent taFinishedEvent = new TaskAttemptFinishedEvent(ta1t1v1Id, "v1", 0L, 0L, TaskAttemptState.SUCCEEDED, null, null, "", null, null, taGeneratedEvents, 0L, null, 0L, null, null, null, null, null);
    TaskAttemptRecoveryData taRecoveryData = new TaskAttemptRecoveryData(taStartedEvent, taFinishedEvent);
    Map<TezTaskAttemptID, TaskAttemptRecoveryData> taRecoveryDataMap = new HashMap<TezTaskAttemptID, TaskAttemptRecoveryData>();
    taRecoveryDataMap.put(ta1t1v1Id, taRecoveryData);
    TaskRecoveryData taskRecoveryData = new TaskRecoveryData(taskStartedEvent, taskFinishedEvent, taRecoveryDataMap);
    doReturn(taskRecoveryData).when(dagRecoveryData).getTaskRecoveryData(t1v1Id);
    doReturn(taRecoveryData).when(dagRecoveryData).getTaskAttemptRecoveryData(ta1t1v1Id);
    dag.handle(new DAGEventRecoverEvent(dagId, dagRecoveryData));
    dispatcher.await();
    VertexImpl vertex1 = (VertexImpl) dag.getVertex(v1Id);
    TaskImpl task = (TaskImpl) vertex1.getTask(t1v1Id);
    TaskAttemptImpl taskAttempt = (TaskAttemptImpl) task.getAttempt(ta1t1v1Id);
    assertEquals(VertexState.RUNNING, vertex1.getState());
    assertEquals(1, vertex1.getCompletedTasks());
    assertEquals(TaskStateInternal.SUCCEEDED, task.getInternalState());
    assertEquals(TaskAttemptStateInternal.SUCCEEDED, taskAttempt.getInternalState());
}
Also used : HashMap(java.util.HashMap) ArrayList(java.util.ArrayList) TaskStartedEvent(org.apache.tez.dag.history.events.TaskStartedEvent) TaskAttemptStartedEvent(org.apache.tez.dag.history.events.TaskAttemptStartedEvent) TaskAttemptRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData) TaskRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskRecoveryData) DAGEventRecoverEvent(org.apache.tez.dag.app.dag.event.DAGEventRecoverEvent) TaskFinishedEvent(org.apache.tez.dag.history.events.TaskFinishedEvent) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) NodeId(org.apache.hadoop.yarn.api.records.NodeId) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) EventMetaData(org.apache.tez.runtime.api.impl.EventMetaData) TezTaskAttemptID(org.apache.tez.dag.records.TezTaskAttemptID) Test(org.junit.Test) StateChangeNotifierForTest(org.apache.tez.dag.app.dag.TestStateChangeNotifier.StateChangeNotifierForTest)

Example 7 with TaskAttemptRecoveryData

use of org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData in project tez by apache.

the class TestDAGRecovery method testTARecoverFromKilled.

/**
 * RecoveryEvents: TaskAttemptStartedEvent -> TaskAttemptFinishedEvent (KILLED)
 * Recover it to KILLED
 */
@Test(timeout = 5000)
public void testTARecoverFromKilled() {
    initMockDAGRecoveryDataForTaskAttempt();
    TaskAttemptStartedEvent taStartedEvent = new TaskAttemptStartedEvent(ta1t1v1Id, "v1", ta1LaunchTime, mock(ContainerId.class), mock(NodeId.class), "", "", "");
    TaskAttemptFinishedEvent taFinishedEvent = new TaskAttemptFinishedEvent(ta1t1v1Id, "v1", ta1FinishedTime, ta1FinishedTime, TaskAttemptState.KILLED, null, TaskAttemptTerminationCause.TERMINATED_BY_CLIENT, "", null, null, null, 0L, null, 0L, null, null, null, null, null);
    TaskAttemptRecoveryData taRecoveryData = new TaskAttemptRecoveryData(taStartedEvent, taFinishedEvent);
    doReturn(taRecoveryData).when(dagRecoveryData).getTaskAttemptRecoveryData(ta1t1v1Id);
    dag.handle(new DAGEventRecoverEvent(dagId, dagRecoveryData));
    dispatcher.await();
    TaskImpl task = (TaskImpl) dag.getVertex(v1Id).getTask(t1v1Id);
    TaskAttemptImpl taskAttempt = (TaskAttemptImpl) task.getAttempt(ta1t1v1Id);
    assertEquals(TaskAttemptStateInternal.KILLED, taskAttempt.getInternalState());
    assertEquals(TaskAttemptTerminationCause.TERMINATED_BY_CLIENT, taskAttempt.getTerminationCause());
    historyEventHandler.verifyHistoryEvent(0, HistoryEventType.TASK_ATTEMPT_FINISHED);
    assertEquals(ta1LaunchTime, taskAttempt.getLaunchTime());
    assertEquals(ta1FinishedTime, taskAttempt.getFinishTime());
}
Also used : DAGEventRecoverEvent(org.apache.tez.dag.app.dag.event.DAGEventRecoverEvent) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) NodeId(org.apache.hadoop.yarn.api.records.NodeId) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) TaskAttemptStartedEvent(org.apache.tez.dag.history.events.TaskAttemptStartedEvent) TaskAttemptRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData) Test(org.junit.Test) StateChangeNotifierForTest(org.apache.tez.dag.app.dag.TestStateChangeNotifier.StateChangeNotifierForTest)

Example 8 with TaskAttemptRecoveryData

use of org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData in project tez by apache.

the class TestDAGRecovery method testTARecoverFromNewToKilled.

/**
 * RecoveryEvents: TaskAttemptFinishedEvent (KILLED)
 * Recover it to KILLED
 */
@Test(timeout = 5000)
public void testTARecoverFromNewToKilled() {
    initMockDAGRecoveryDataForTaskAttempt();
    TaskAttemptFinishedEvent taFinishedEvent = new TaskAttemptFinishedEvent(ta1t1v1Id, "v1", ta1LaunchTime, ta1FinishedTime, TaskAttemptState.KILLED, null, TaskAttemptTerminationCause.TERMINATED_BY_CLIENT, "", null, null, null, 0L, null, 0L, null, null, null, null, null);
    TaskAttemptRecoveryData taRecoveryData = new TaskAttemptRecoveryData(null, taFinishedEvent);
    doReturn(taRecoveryData).when(dagRecoveryData).getTaskAttemptRecoveryData(ta1t1v1Id);
    dag.handle(new DAGEventRecoverEvent(dagId, dagRecoveryData));
    dispatcher.await();
    TaskImpl task = (TaskImpl) dag.getVertex(v1Id).getTask(t1v1Id);
    TaskAttemptImpl taskAttempt = (TaskAttemptImpl) task.getAttempt(ta1t1v1Id);
    assertEquals(TaskAttemptStateInternal.KILLED, taskAttempt.getInternalState());
    assertEquals(TaskAttemptTerminationCause.TERMINATED_BY_CLIENT, taskAttempt.getTerminationCause());
    historyEventHandler.verifyHistoryEvent(0, HistoryEventType.TASK_ATTEMPT_STARTED);
    historyEventHandler.verifyHistoryEvent(0, HistoryEventType.TASK_ATTEMPT_FINISHED);
    assertEquals(0, task.failedAttempts);
    assertEquals(ta1FinishedTime, taskAttempt.getFinishTime());
}
Also used : DAGEventRecoverEvent(org.apache.tez.dag.app.dag.event.DAGEventRecoverEvent) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) TaskAttemptRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData) Test(org.junit.Test) StateChangeNotifierForTest(org.apache.tez.dag.app.dag.TestStateChangeNotifier.StateChangeNotifierForTest)

Example 9 with TaskAttemptRecoveryData

use of org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData in project tez by apache.

the class TestDAGRecovery method testTARecoverFromSucceeded_OutputCommitterRecoveryNotSupported.

/**
 * RecoveryEvents: TaskAttemptStartedEvent -> TaskAttemptFinishedEvent (SUCCEEDED)
 * Recovered it SUCCEEDED, but task schedule new task attempt
 * V2's committer is not recovery supported
 */
// (timeout=5000)
@Test
public void testTARecoverFromSucceeded_OutputCommitterRecoveryNotSupported() throws Exception {
    initMockDAGRecoveryDataForTaskAttempt();
    // set up v2 recovery data
    // ta1t1v2: TaskAttemptStartedEvent -> TaskAttemptFinishedEvent(SUCCEEDED)
    // t1v2: TaskStartedEvent
    // v2: VertexInitializedEvent -> VertexConfigurationDoneEvent -> VertexStartedEvent
    TaskAttemptStartedEvent taStartedEvent = new TaskAttemptStartedEvent(ta1t1v2Id, "vertex2", ta1LaunchTime, mock(ContainerId.class), mock(NodeId.class), "", "", "");
    List<TezEvent> taGeneratedEvents = new ArrayList<TezEvent>();
    EventMetaData metadata = new EventMetaData(EventProducerConsumerType.OUTPUT, "vertex2", "vertex3", ta1t1v2Id);
    taGeneratedEvents.add(new TezEvent(DataMovementEvent.create(ByteBuffer.wrap(new byte[0])), metadata));
    TaskAttemptFinishedEvent taFinishedEvent = new TaskAttemptFinishedEvent(ta1t1v2Id, "vertex2", ta1LaunchTime, ta1FinishedTime, TaskAttemptState.SUCCEEDED, null, null, "", null, null, taGeneratedEvents, 0L, null, 0L, null, null, null, null, null);
    TaskAttemptRecoveryData taRecoveryData = new TaskAttemptRecoveryData(taStartedEvent, taFinishedEvent);
    doReturn(taRecoveryData).when(dagRecoveryData).getTaskAttemptRecoveryData(ta1t1v2Id);
    Map<TezTaskAttemptID, TaskAttemptRecoveryData> taRecoveryDataMap = new HashMap<TezTaskAttemptID, TaskAttemptRecoveryData>();
    taRecoveryDataMap.put(ta1t1v2Id, taRecoveryData);
    TaskStartedEvent t1StartedEvent = new TaskStartedEvent(t1v2Id, "vertex2", 0L, t1StartedTime);
    TaskRecoveryData taskRecoveryData = new TaskRecoveryData(t1StartedEvent, null, taRecoveryDataMap);
    Map<TezTaskID, TaskRecoveryData> taskRecoveryDataMap = new HashMap<TezTaskID, TaskRecoveryData>();
    taskRecoveryDataMap.put(t1v2Id, taskRecoveryData);
    doReturn(taskRecoveryData).when(dagRecoveryData).getTaskRecoveryData(t1v2Id);
    VertexInitializedEvent v2InitedEvent = new VertexInitializedEvent(v2Id, "vertex2", 0L, v1InitedTime, v1NumTask, "", null, null, null);
    VertexConfigurationDoneEvent v2ReconfigureDoneEvent = new VertexConfigurationDoneEvent(v2Id, 0L, v1NumTask, null, null, null, false);
    VertexStartedEvent v2StartedEvent = new VertexStartedEvent(v2Id, 0L, v1StartedTime);
    VertexRecoveryData v2RecoveryData = new VertexRecoveryData(v2InitedEvent, v2ReconfigureDoneEvent, v2StartedEvent, null, taskRecoveryDataMap, false);
    doReturn(v2RecoveryData).when(dagRecoveryData).getVertexRecoveryData(v2Id);
    dag.handle(new DAGEventRecoverEvent(dagId, dagRecoveryData));
    dispatcher.await();
    TaskImpl task = (TaskImpl) dag.getVertex(v2Id).getTask(t1v2Id);
    TaskAttemptImpl taskAttempt = (TaskAttemptImpl) task.getAttempt(ta1t1v2Id);
    assertEquals(TaskAttemptStateInternal.KILLED, taskAttempt.getInternalState());
    historyEventHandler.verifyHistoryEvent(1, HistoryEventType.TASK_ATTEMPT_FINISHED);
    assertEquals(TaskStateInternal.RUNNING, task.getInternalState());
    // new task attempt is scheduled
    assertEquals(2, task.getAttempts().size());
    assertEquals(ta1LaunchTime, taskAttempt.getLaunchTime());
    assertEquals(ta1FinishedTime, taskAttempt.getFinishTime());
}
Also used : VertexStartedEvent(org.apache.tez.dag.history.events.VertexStartedEvent) VertexInitializedEvent(org.apache.tez.dag.history.events.VertexInitializedEvent) HashMap(java.util.HashMap) ArrayList(java.util.ArrayList) TaskStartedEvent(org.apache.tez.dag.history.events.TaskStartedEvent) TaskAttemptStartedEvent(org.apache.tez.dag.history.events.TaskAttemptStartedEvent) TaskAttemptRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData) TezTaskID(org.apache.tez.dag.records.TezTaskID) TaskRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskRecoveryData) DAGEventRecoverEvent(org.apache.tez.dag.app.dag.event.DAGEventRecoverEvent) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) NodeId(org.apache.hadoop.yarn.api.records.NodeId) VertexConfigurationDoneEvent(org.apache.tez.dag.history.events.VertexConfigurationDoneEvent) VertexRecoveryData(org.apache.tez.dag.app.RecoveryParser.VertexRecoveryData) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) TaskAttemptFinishedEvent(org.apache.tez.dag.history.events.TaskAttemptFinishedEvent) EventMetaData(org.apache.tez.runtime.api.impl.EventMetaData) TezTaskAttemptID(org.apache.tez.dag.records.TezTaskAttemptID) Test(org.junit.Test) StateChangeNotifierForTest(org.apache.tez.dag.app.dag.TestStateChangeNotifier.StateChangeNotifierForTest)

Aggregations

TaskAttemptRecoveryData (org.apache.tez.dag.app.RecoveryParser.TaskAttemptRecoveryData)9 StateChangeNotifierForTest (org.apache.tez.dag.app.dag.TestStateChangeNotifier.StateChangeNotifierForTest)8 DAGEventRecoverEvent (org.apache.tez.dag.app.dag.event.DAGEventRecoverEvent)8 TaskAttemptFinishedEvent (org.apache.tez.dag.history.events.TaskAttemptFinishedEvent)8 Test (org.junit.Test)8 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)7 NodeId (org.apache.hadoop.yarn.api.records.NodeId)7 TaskAttemptStartedEvent (org.apache.tez.dag.history.events.TaskAttemptStartedEvent)7 TezEvent (org.apache.tez.runtime.api.impl.TezEvent)4 ArrayList (java.util.ArrayList)3 TaskRecoveryData (org.apache.tez.dag.app.RecoveryParser.TaskRecoveryData)3 TaskStartedEvent (org.apache.tez.dag.history.events.TaskStartedEvent)3 TezTaskAttemptID (org.apache.tez.dag.records.TezTaskAttemptID)3 EventMetaData (org.apache.tez.runtime.api.impl.EventMetaData)3 HashMap (java.util.HashMap)2 VertexRecoveryData (org.apache.tez.dag.app.RecoveryParser.VertexRecoveryData)2 TaskFinishedEvent (org.apache.tez.dag.history.events.TaskFinishedEvent)2 VertexInitializedEvent (org.apache.tez.dag.history.events.VertexInitializedEvent)2 VertexStartedEvent (org.apache.tez.dag.history.events.VertexStartedEvent)2 TezTaskID (org.apache.tez.dag.records.TezTaskID)2