Search in sources :

Example 6 with VertexInitializedEvent

use of org.apache.tez.dag.history.events.VertexInitializedEvent in project tez by apache.

the class TestRecovery method testRecovery_OrderedWordCount.

@Test(timeout = 1800000)
public void testRecovery_OrderedWordCount() throws Exception {
    ApplicationId appId = ApplicationId.newInstance(System.currentTimeMillis(), 1);
    TezDAGID dagId = TezDAGID.getInstance(appId, 1);
    TezVertexID vertexId0 = TezVertexID.getInstance(dagId, 0);
    TezVertexID vertexId1 = TezVertexID.getInstance(dagId, 1);
    TezVertexID vertexId2 = TezVertexID.getInstance(dagId, 2);
    ContainerId containerId = ContainerId.newInstance(ApplicationAttemptId.newInstance(appId, 1), 1);
    NodeId nodeId = NodeId.newInstance("localhost", 10);
    List<TezEvent> initGeneratedEvents = Lists.newArrayList(new TezEvent(InputDataInformationEvent.createWithObjectPayload(0, new Object()), null));
    List<SimpleShutdownCondition> shutdownConditions = Lists.newArrayList(new SimpleShutdownCondition(TIMING.POST, new DAGInitializedEvent(dagId, 0L, "username", "dagName", null)), new SimpleShutdownCondition(TIMING.POST, new DAGStartedEvent(dagId, 0L, "username", "dagName")), new SimpleShutdownCondition(TIMING.POST, new DAGFinishedEvent(dagId, 0L, 0L, DAGState.SUCCEEDED, "", new TezCounters(), "username", "dagName", new HashMap<String, Integer>(), ApplicationAttemptId.newInstance(appId, 1), null)), new SimpleShutdownCondition(TIMING.POST, new VertexInitializedEvent(vertexId0, "Tokenizer", 0L, 0L, 0, "", null, initGeneratedEvents, null)), new SimpleShutdownCondition(TIMING.POST, new VertexInitializedEvent(vertexId1, "Summation", 0L, 0L, 0, "", null, null, null)), new SimpleShutdownCondition(TIMING.POST, new VertexInitializedEvent(vertexId2, "Sorter", 0L, 0L, 0, "", null, null, null)), new SimpleShutdownCondition(TIMING.POST, new VertexConfigurationDoneEvent(vertexId0, 0L, 2, null, null, null, true)), new SimpleShutdownCondition(TIMING.POST, new VertexConfigurationDoneEvent(vertexId1, 0L, 2, null, null, null, true)), new SimpleShutdownCondition(TIMING.POST, new VertexConfigurationDoneEvent(vertexId2, 0L, 2, null, null, null, true)), new SimpleShutdownCondition(TIMING.POST, new VertexStartedEvent(vertexId0, 0L, 0L)), new SimpleShutdownCondition(TIMING.POST, new VertexStartedEvent(vertexId1, 0L, 0L)), new SimpleShutdownCondition(TIMING.POST, new VertexStartedEvent(vertexId2, 0L, 0L)), new SimpleShutdownCondition(TIMING.POST, new VertexFinishedEvent(vertexId0, "vertexName", 1, 0L, 0L, 0L, 0L, 0L, VertexState.SUCCEEDED, "", new TezCounters(), new VertexStats(), new HashMap<String, Integer>(), null)), new SimpleShutdownCondition(TIMING.POST, new VertexFinishedEvent(vertexId1, "vertexName", 1, 0L, 0L, 0L, 0L, 0L, VertexState.SUCCEEDED, "", new TezCounters(), new VertexStats(), new HashMap<String, Integer>(), null)), new SimpleShutdownCondition(TIMING.POST, new VertexFinishedEvent(vertexId2, "vertexName", 1, 0L, 0L, 0L, 0L, 0L, VertexState.SUCCEEDED, "", new TezCounters(), new VertexStats(), new HashMap<String, Integer>(), null)), new SimpleShutdownCondition(TIMING.POST, new TaskStartedEvent(TezTaskID.getInstance(vertexId0, 0), "vertexName", 0L, 0L)), new SimpleShutdownCondition(TIMING.POST, new TaskStartedEvent(TezTaskID.getInstance(vertexId1, 0), "vertexName", 0L, 0L)), new SimpleShutdownCondition(TIMING.POST, new TaskStartedEvent(TezTaskID.getInstance(vertexId2, 0), "vertexName", 0L, 0L)), new SimpleShutdownCondition(TIMING.POST, new TaskFinishedEvent(TezTaskID.getInstance(vertexId0, 0), "vertexName", 0L, 0L, null, TaskState.SUCCEEDED, "", new TezCounters(), 0)), new SimpleShutdownCondition(TIMING.POST, new TaskFinishedEvent(TezTaskID.getInstance(vertexId1, 0), "vertexName", 0L, 0L, null, TaskState.SUCCEEDED, "", new TezCounters(), 0)), new SimpleShutdownCondition(TIMING.POST, new TaskFinishedEvent(TezTaskID.getInstance(vertexId2, 0), "vertexName", 0L, 0L, null, TaskState.SUCCEEDED, "", new TezCounters(), 0)), new SimpleShutdownCondition(TIMING.POST, new TaskAttemptStartedEvent(TezTaskAttemptID.getInstance(TezTaskID.getInstance(vertexId0, 0), 0), "vertexName", 0L, containerId, nodeId, "", "", "")), new SimpleShutdownCondition(TIMING.POST, new TaskAttemptStartedEvent(TezTaskAttemptID.getInstance(TezTaskID.getInstance(vertexId1, 0), 0), "vertexName", 0L, containerId, nodeId, "", "", "")), new SimpleShutdownCondition(TIMING.POST, new TaskAttemptStartedEvent(TezTaskAttemptID.getInstance(TezTaskID.getInstance(vertexId2, 0), 0), "vertexName", 0L, containerId, nodeId, "", "", "")));
    Random rand = new Random();
    for (int i = 0; i < shutdownConditions.size(); i++) {
        // timeout.
        if (rand.nextDouble() < 0.5) {
            // generate split in client side when HistoryEvent type is VERTEX_STARTED (TEZ-2976)
            testOrderedWordCount(shutdownConditions.get(i), true, shutdownConditions.get(i).getHistoryEvent().getEventType() == HistoryEventType.VERTEX_STARTED);
        }
    }
}
Also used : VertexInitializedEvent(org.apache.tez.dag.history.events.VertexInitializedEvent) VertexStats(org.apache.tez.dag.app.dag.impl.VertexStats) DAGInitializedEvent(org.apache.tez.dag.history.events.DAGInitializedEvent) Random(java.util.Random) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) TezDAGID(org.apache.tez.dag.records.TezDAGID) DAGStartedEvent(org.apache.tez.dag.history.events.DAGStartedEvent) VertexConfigurationDoneEvent(org.apache.tez.dag.history.events.VertexConfigurationDoneEvent) TezVertexID(org.apache.tez.dag.records.TezVertexID) VertexStartedEvent(org.apache.tez.dag.history.events.VertexStartedEvent) SimpleShutdownCondition(org.apache.tez.test.RecoveryServiceWithEventHandlingHook.SimpleShutdownCondition) TaskStartedEvent(org.apache.tez.dag.history.events.TaskStartedEvent) TezCounters(org.apache.tez.common.counters.TezCounters) TaskAttemptStartedEvent(org.apache.tez.dag.history.events.TaskAttemptStartedEvent) TaskFinishedEvent(org.apache.tez.dag.history.events.TaskFinishedEvent) NodeId(org.apache.hadoop.yarn.api.records.NodeId) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) DAGFinishedEvent(org.apache.tez.dag.history.events.DAGFinishedEvent) VertexFinishedEvent(org.apache.tez.dag.history.events.VertexFinishedEvent) Test(org.junit.Test)

Example 7 with VertexInitializedEvent

use of org.apache.tez.dag.history.events.VertexInitializedEvent in project tez by apache.

the class TestDAGRecovery method initMockDAGRecoveryDataForTaskAttempt.

// ///////////////////////////// TaskAttempt Recovery /////////////////////////////////////////////////////
private void initMockDAGRecoveryDataForTaskAttempt() {
    TaskStartedEvent t1StartedEvent = new TaskStartedEvent(t1v1Id, "vertex1", 0L, t1StartedTime);
    TaskRecoveryData taskRecoveryData = new TaskRecoveryData(t1StartedEvent, null, null);
    Map<TezTaskID, TaskRecoveryData> taskRecoveryDataMap = new HashMap<TezTaskID, TaskRecoveryData>();
    taskRecoveryDataMap.put(t1v1Id, taskRecoveryData);
    List<TezEvent> inputGeneratedTezEvents = new ArrayList<TezEvent>();
    VertexInitializedEvent v1InitedEvent = new VertexInitializedEvent(v1Id, "vertex1", 0L, v1InitedTime, v1NumTask, "", null, inputGeneratedTezEvents, null);
    Map<String, InputSpecUpdate> rootInputSpecs = new HashMap<String, InputSpecUpdate>();
    VertexConfigurationDoneEvent v1ReconfigureDoneEvent = new VertexConfigurationDoneEvent(v1Id, 0L, v1NumTask, null, null, rootInputSpecs, true);
    VertexStartedEvent v1StartedEvent = new VertexStartedEvent(v1Id, 0L, v1StartedTime);
    VertexRecoveryData v1RecoveryData = new VertexRecoveryData(v1InitedEvent, v1ReconfigureDoneEvent, v1StartedEvent, null, taskRecoveryDataMap, false);
    DAGInitializedEvent dagInitedEvent = new DAGInitializedEvent(dagId, dagInitedTime, "user", "dagName", null);
    DAGStartedEvent dagStartedEvent = new DAGStartedEvent(dagId, dagStartedTime, "user", "dagName");
    doReturn(v1RecoveryData).when(dagRecoveryData).getVertexRecoveryData(v1Id);
    doReturn(dagInitedEvent).when(dagRecoveryData).getDAGInitializedEvent();
    doReturn(dagStartedEvent).when(dagRecoveryData).getDAGStartedEvent();
}
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) InputSpecUpdate(org.apache.tez.runtime.api.InputSpecUpdate) TaskStartedEvent(org.apache.tez.dag.history.events.TaskStartedEvent) TezTaskID(org.apache.tez.dag.records.TezTaskID) DAGInitializedEvent(org.apache.tez.dag.history.events.DAGInitializedEvent) TaskRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskRecoveryData) DAGStartedEvent(org.apache.tez.dag.history.events.DAGStartedEvent) VertexConfigurationDoneEvent(org.apache.tez.dag.history.events.VertexConfigurationDoneEvent) VertexRecoveryData(org.apache.tez.dag.app.RecoveryParser.VertexRecoveryData) TezEvent(org.apache.tez.runtime.api.impl.TezEvent)

Example 8 with VertexInitializedEvent

use of org.apache.tez.dag.history.events.VertexInitializedEvent in project tez by apache.

the class TestDAGRecovery method testVertexRecoverFromInited.

/**
 * RecoveryEvents:
 *  DAG:  DAGInitedEvent -> DAGStartedEvent
 *  V1:   VertexInitializedEvent
 *
 * Reinitialize V1 again.
 */
@Test(timeout = 5000)
public void testVertexRecoverFromInited() {
    initMockDAGRecoveryDataForVertex();
    List<TezEvent> inputGeneratedTezEvents = new ArrayList<TezEvent>();
    VertexInitializedEvent v1InitedEvent = new VertexInitializedEvent(v1Id, "vertex1", 0L, v1InitedTime, v1NumTask, "", null, inputGeneratedTezEvents, null);
    VertexRecoveryData vertexRecoveryData = new VertexRecoveryData(v1InitedEvent, null, null, null, null, false);
    doReturn(vertexRecoveryData).when(dagRecoveryData).getVertexRecoveryData(v1Id);
    DAGEventRecoverEvent recoveryEvent = new DAGEventRecoverEvent(dagId, dagRecoveryData);
    dag.handle(recoveryEvent);
    dispatcher.await();
    assertEquals(DAGState.RUNNING, dag.getState());
    // reinitialize v1 again because its VertexManager is not completed
    VertexImpl v1 = (VertexImpl) dag.getVertex("vertex1");
    VertexImpl v2 = (VertexImpl) dag.getVertex("vertex2");
    VertexImpl v3 = (VertexImpl) dag.getVertex("vertex3");
    assertEquals(VertexState.INITIALIZING, v1.getState());
    assertEquals(VertexState.RUNNING, v2.getState());
    assertEquals(VertexState.INITED, v3.getState());
}
Also used : DAGEventRecoverEvent(org.apache.tez.dag.app.dag.event.DAGEventRecoverEvent) VertexInitializedEvent(org.apache.tez.dag.history.events.VertexInitializedEvent) ArrayList(java.util.ArrayList) VertexRecoveryData(org.apache.tez.dag.app.RecoveryParser.VertexRecoveryData) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) Test(org.junit.Test) StateChangeNotifierForTest(org.apache.tez.dag.app.dag.TestStateChangeNotifier.StateChangeNotifierForTest)

Example 9 with VertexInitializedEvent

use of org.apache.tez.dag.history.events.VertexInitializedEvent in project tez by apache.

the class TestDAGRecovery method testVertexRecoverFromInitedAndReconfigureDone.

/**
 * RecoveryEvents:
 *  DAG:  DAGInitedEvent -> DAGStartedEvent
 *  V1:   VertexReconfigrationDoneEvent -> VertexInitializedEvent
 *
 * V1 skip initialization.
 */
// (timeout=5000)
@Test
public void testVertexRecoverFromInitedAndReconfigureDone() {
    initMockDAGRecoveryDataForVertex();
    List<TezEvent> inputGeneratedTezEvents = new ArrayList<TezEvent>();
    VertexInitializedEvent v1InitedEvent = new VertexInitializedEvent(v1Id, "vertex1", 0L, v1InitedTime, v1NumTask, "", null, inputGeneratedTezEvents, null);
    VertexConfigurationDoneEvent v1ReconfigureDoneEvent = new VertexConfigurationDoneEvent(v1Id, 0L, v1NumTask, null, null, null, true);
    VertexRecoveryData vertexRecoveryData = new VertexRecoveryData(v1InitedEvent, v1ReconfigureDoneEvent, null, null, new HashMap<TezTaskID, TaskRecoveryData>(), false);
    doReturn(vertexRecoveryData).when(dagRecoveryData).getVertexRecoveryData(v1Id);
    DAGEventRecoverEvent recoveryEvent = new DAGEventRecoverEvent(dagId, dagRecoveryData);
    dag.handle(recoveryEvent);
    dispatcher.await();
    try {
        Thread.sleep(1000);
    } catch (InterruptedException e) {
        e.printStackTrace();
    }
    VertexImpl v1 = (VertexImpl) dag.getVertex("vertex1");
    VertexImpl v2 = (VertexImpl) dag.getVertex("vertex2");
    VertexImpl v3 = (VertexImpl) dag.getVertex("vertex3");
    assertEquals(DAGState.RUNNING, dag.getState());
    // v1 skip initialization
    assertEquals(VertexState.RUNNING, v1.getState());
    assertEquals(v1InitedTime, v1.initedTime);
    assertEquals(v1NumTask, v1.getTotalTasks());
    assertEquals(VertexState.RUNNING, v2.getState());
    assertEquals(VertexState.RUNNING, v3.getState());
}
Also used : DAGEventRecoverEvent(org.apache.tez.dag.app.dag.event.DAGEventRecoverEvent) VertexInitializedEvent(org.apache.tez.dag.history.events.VertexInitializedEvent) ArrayList(java.util.ArrayList) VertexConfigurationDoneEvent(org.apache.tez.dag.history.events.VertexConfigurationDoneEvent) VertexRecoveryData(org.apache.tez.dag.app.RecoveryParser.VertexRecoveryData) TezEvent(org.apache.tez.runtime.api.impl.TezEvent) TezTaskID(org.apache.tez.dag.records.TezTaskID) TaskRecoveryData(org.apache.tez.dag.app.RecoveryParser.TaskRecoveryData) Test(org.junit.Test) StateChangeNotifierForTest(org.apache.tez.dag.app.dag.TestStateChangeNotifier.StateChangeNotifierForTest)

Example 10 with VertexInitializedEvent

use of org.apache.tez.dag.history.events.VertexInitializedEvent 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

VertexInitializedEvent (org.apache.tez.dag.history.events.VertexInitializedEvent)16 VertexConfigurationDoneEvent (org.apache.tez.dag.history.events.VertexConfigurationDoneEvent)12 VertexStartedEvent (org.apache.tez.dag.history.events.VertexStartedEvent)12 TaskStartedEvent (org.apache.tez.dag.history.events.TaskStartedEvent)10 TezEvent (org.apache.tez.runtime.api.impl.TezEvent)10 Test (org.junit.Test)10 DAGInitializedEvent (org.apache.tez.dag.history.events.DAGInitializedEvent)9 DAGStartedEvent (org.apache.tez.dag.history.events.DAGStartedEvent)9 TaskAttemptStartedEvent (org.apache.tez.dag.history.events.TaskAttemptStartedEvent)9 TaskFinishedEvent (org.apache.tez.dag.history.events.TaskFinishedEvent)8 VertexFinishedEvent (org.apache.tez.dag.history.events.VertexFinishedEvent)8 VertexRecoveryData (org.apache.tez.dag.app.RecoveryParser.VertexRecoveryData)7 ArrayList (java.util.ArrayList)6 TaskRecoveryData (org.apache.tez.dag.app.RecoveryParser.TaskRecoveryData)6 DAGFinishedEvent (org.apache.tez.dag.history.events.DAGFinishedEvent)6 TaskAttemptFinishedEvent (org.apache.tez.dag.history.events.TaskAttemptFinishedEvent)6 TezTaskID (org.apache.tez.dag.records.TezTaskID)6 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)5 NodeId (org.apache.hadoop.yarn.api.records.NodeId)5 DAGSubmittedEvent (org.apache.tez.dag.history.events.DAGSubmittedEvent)5