Search in sources :

Example 61 with Task

use of org.apache.hadoop.mapreduce.v2.app.job.Task in project hadoop by apache.

the class TestTaskAttempt method testFetchFailureAttemptFinishTime.

@Test
public void testFetchFailureAttemptFinishTime() throws Exception {
    ApplicationId appId = ApplicationId.newInstance(1, 2);
    ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0);
    JobId jobId = MRBuilderUtils.newJobId(appId, 1);
    TaskId taskId = MRBuilderUtils.newTaskId(jobId, 1, TaskType.MAP);
    TaskAttemptId attemptId = MRBuilderUtils.newTaskAttemptId(taskId, 0);
    TaskId reducetaskId = MRBuilderUtils.newTaskId(jobId, 1, TaskType.REDUCE);
    TaskAttemptId reduceTAId = MRBuilderUtils.newTaskAttemptId(reducetaskId, 0);
    Path jobFile = mock(Path.class);
    MockEventHandler eventHandler = new MockEventHandler();
    TaskAttemptListener taListener = mock(TaskAttemptListener.class);
    when(taListener.getAddress()).thenReturn(new InetSocketAddress("localhost", 0));
    JobConf jobConf = new JobConf();
    jobConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class);
    jobConf.setBoolean("fs.file.impl.disable.cache", true);
    jobConf.set(JobConf.MAPRED_MAP_TASK_ENV, "");
    jobConf.set(MRJobConfig.APPLICATION_ATTEMPT_ID, "10");
    TaskSplitMetaInfo splits = mock(TaskSplitMetaInfo.class);
    when(splits.getLocations()).thenReturn(new String[] { "127.0.0.1" });
    AppContext appCtx = mock(AppContext.class);
    ClusterInfo clusterInfo = mock(ClusterInfo.class);
    when(appCtx.getClusterInfo()).thenReturn(clusterInfo);
    setupTaskAttemptFinishingMonitor(eventHandler, jobConf, appCtx);
    TaskAttemptImpl taImpl = new MapTaskAttemptImpl(taskId, 1, eventHandler, jobFile, 1, splits, jobConf, taListener, mock(Token.class), new Credentials(), SystemClock.getInstance(), appCtx);
    NodeId nid = NodeId.newInstance("127.0.0.1", 0);
    ContainerId contId = ContainerId.newContainerId(appAttemptId, 3);
    Container container = mock(Container.class);
    when(container.getId()).thenReturn(contId);
    when(container.getNodeId()).thenReturn(nid);
    when(container.getNodeHttpAddress()).thenReturn("localhost:0");
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_SCHEDULE));
    taImpl.handle(new TaskAttemptContainerAssignedEvent(attemptId, container, mock(Map.class)));
    taImpl.handle(new TaskAttemptContainerLaunchedEvent(attemptId, 0));
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_DONE));
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_CONTAINER_COMPLETED));
    assertEquals("Task attempt is not in succeeded state", taImpl.getState(), TaskAttemptState.SUCCEEDED);
    assertTrue("Task Attempt finish time is not greater than 0", taImpl.getFinishTime() > 0);
    Long finishTime = taImpl.getFinishTime();
    Thread.sleep(5);
    taImpl.handle(new TaskAttemptTooManyFetchFailureEvent(attemptId, reduceTAId, "Host"));
    assertEquals("Task attempt is not in Too Many Fetch Failure state", taImpl.getState(), TaskAttemptState.FAILED);
    assertEquals("After TA_TOO_MANY_FETCH_FAILURE," + " Task attempt finish time is not the same ", finishTime, Long.valueOf(taImpl.getFinishTime()));
}
Also used : TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) TaskAttemptListener(org.apache.hadoop.mapreduce.v2.app.TaskAttemptListener) InetSocketAddress(java.net.InetSocketAddress) Token(org.apache.hadoop.security.token.Token) TaskAttemptContainerAssignedEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptContainerAssignedEvent) Container(org.apache.hadoop.yarn.api.records.Container) TaskAttemptTooManyFetchFailureEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptTooManyFetchFailureEvent) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) JobConf(org.apache.hadoop.mapred.JobConf) JobId(org.apache.hadoop.mapreduce.v2.api.records.JobId) Path(org.apache.hadoop.fs.Path) TaskAttemptId(org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId) AppContext(org.apache.hadoop.mapreduce.v2.app.AppContext) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) TaskAttemptEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEvent) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) TaskAttemptContainerLaunchedEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptContainerLaunchedEvent) ClusterInfo(org.apache.hadoop.mapreduce.v2.app.ClusterInfo) NodeId(org.apache.hadoop.yarn.api.records.NodeId) TaskSplitMetaInfo(org.apache.hadoop.mapreduce.split.JobSplit.TaskSplitMetaInfo) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) Credentials(org.apache.hadoop.security.Credentials) Test(org.junit.Test)

Example 62 with Task

use of org.apache.hadoop.mapreduce.v2.app.job.Task in project hadoop by apache.

the class TestTaskAttempt method containerKillBeforeAssignment.

private void containerKillBeforeAssignment(boolean scheduleAttempt) throws Exception {
    MockEventHandler eventHandler = new MockEventHandler();
    ApplicationId appId = ApplicationId.newInstance(1, 2);
    JobId jobId = MRBuilderUtils.newJobId(appId, 1);
    TaskId taskId = MRBuilderUtils.newTaskId(jobId, 1, TaskType.MAP);
    TaskAttemptImpl taImpl = new MapTaskAttemptImpl(taskId, 1, eventHandler, mock(Path.class), 1, mock(TaskSplitMetaInfo.class), new JobConf(), mock(TaskAttemptListener.class), mock(Token.class), new Credentials(), SystemClock.getInstance(), mock(AppContext.class));
    if (scheduleAttempt) {
        taImpl.handle(new TaskAttemptEvent(taImpl.getID(), TaskAttemptEventType.TA_SCHEDULE));
    }
    taImpl.handle(new TaskAttemptKillEvent(taImpl.getID(), "", true));
    assertEquals("Task attempt is not in KILLED state", taImpl.getState(), TaskAttemptState.KILLED);
    assertEquals("Task attempt's internal state is not KILLED", taImpl.getInternalState(), TaskAttemptStateInternal.KILLED);
    assertFalse("InternalError occurred", eventHandler.internalError);
    TaskEvent event = eventHandler.lastTaskEvent;
    assertEquals(TaskEventType.T_ATTEMPT_KILLED, event.getType());
    // In NEW state, new map attempt should not be rescheduled.
    assertFalse(((TaskTAttemptKilledEvent) event).getRescheduleAttempt());
}
Also used : Path(org.apache.hadoop.fs.Path) TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) TaskAttemptListener(org.apache.hadoop.mapreduce.v2.app.TaskAttemptListener) AppContext(org.apache.hadoop.mapreduce.v2.app.AppContext) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) Token(org.apache.hadoop.security.token.Token) TaskAttemptEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEvent) TaskEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskEvent) TaskSplitMetaInfo(org.apache.hadoop.mapreduce.split.JobSplit.TaskSplitMetaInfo) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) TaskAttemptKillEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptKillEvent) JobConf(org.apache.hadoop.mapred.JobConf) JobId(org.apache.hadoop.mapreduce.v2.api.records.JobId) Credentials(org.apache.hadoop.security.Credentials)

Example 63 with Task

use of org.apache.hadoop.mapreduce.v2.app.job.Task in project hadoop by apache.

the class TestTaskAttempt method testKillMapTaskWhileSuccessFinishing.

@Test
public void testKillMapTaskWhileSuccessFinishing() throws Exception {
    MockEventHandler eventHandler = new MockEventHandler();
    TaskAttemptImpl taImpl = createTaskAttemptImpl(eventHandler);
    taImpl.handle(new TaskAttemptEvent(taImpl.getID(), TaskAttemptEventType.TA_DONE));
    assertEquals("Task attempt is not in SUCCEEDED state", taImpl.getState(), TaskAttemptState.SUCCEEDED);
    assertEquals("Task attempt's internal state is not " + "SUCCESS_FINISHING_CONTAINER", taImpl.getInternalState(), TaskAttemptStateInternal.SUCCESS_FINISHING_CONTAINER);
    // If the map task is killed when it is in SUCCESS_FINISHING_CONTAINER
    // state, the state will move to KILL_CONTAINER_CLEANUP
    taImpl.handle(new TaskAttemptEvent(taImpl.getID(), TaskAttemptEventType.TA_KILL));
    assertEquals("Task attempt is not in KILLED state", taImpl.getState(), TaskAttemptState.KILLED);
    assertEquals("Task attempt's internal state is not KILL_CONTAINER_CLEANUP", taImpl.getInternalState(), TaskAttemptStateInternal.KILL_CONTAINER_CLEANUP);
    taImpl.handle(new TaskAttemptEvent(taImpl.getID(), TaskAttemptEventType.TA_CONTAINER_CLEANED));
    assertEquals("Task attempt's internal state is not KILL_TASK_CLEANUP", taImpl.getInternalState(), TaskAttemptStateInternal.KILL_TASK_CLEANUP);
    taImpl.handle(new TaskAttemptEvent(taImpl.getID(), TaskAttemptEventType.TA_CLEANUP_DONE));
    assertEquals("Task attempt is not in KILLED state", taImpl.getState(), TaskAttemptState.KILLED);
    assertFalse("InternalError occurred", eventHandler.internalError);
}
Also used : MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) TaskAttemptEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEvent) Test(org.junit.Test)

Example 64 with Task

use of org.apache.hadoop.mapreduce.v2.app.job.Task in project hadoop by apache.

the class TestTaskHeartbeatHandler method testTimeout.

@SuppressWarnings({ "rawtypes", "unchecked" })
@Test
public void testTimeout() throws InterruptedException {
    EventHandler mockHandler = mock(EventHandler.class);
    Clock clock = SystemClock.getInstance();
    TaskHeartbeatHandler hb = new TaskHeartbeatHandler(mockHandler, clock, 1);
    Configuration conf = new Configuration();
    //10 ms
    conf.setInt(MRJobConfig.TASK_TIMEOUT, 10);
    // set TASK_PROGRESS_REPORT_INTERVAL to a value smaller than TASK_TIMEOUT
    // so that TASK_TIMEOUT is not overridden
    conf.setLong(MRJobConfig.TASK_PROGRESS_REPORT_INTERVAL, 5);
    //10 ms
    conf.setInt(MRJobConfig.TASK_TIMEOUT_CHECK_INTERVAL_MS, 10);
    hb.init(conf);
    hb.start();
    try {
        ApplicationId appId = ApplicationId.newInstance(0l, 5);
        JobId jobId = MRBuilderUtils.newJobId(appId, 4);
        TaskId tid = MRBuilderUtils.newTaskId(jobId, 3, TaskType.MAP);
        TaskAttemptId taid = MRBuilderUtils.newTaskAttemptId(tid, 2);
        hb.register(taid);
        Thread.sleep(100);
        //Events only happen when the task is canceled
        verify(mockHandler, times(2)).handle(any(Event.class));
    } finally {
        hb.stop();
    }
}
Also used : TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) Configuration(org.apache.hadoop.conf.Configuration) TaskAttemptId(org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId) EventHandler(org.apache.hadoop.yarn.event.EventHandler) Event(org.apache.hadoop.yarn.event.Event) Clock(org.apache.hadoop.yarn.util.Clock) SystemClock(org.apache.hadoop.yarn.util.SystemClock) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) JobId(org.apache.hadoop.mapreduce.v2.api.records.JobId) Test(org.junit.Test)

Example 65 with Task

use of org.apache.hadoop.mapreduce.v2.app.job.Task in project hadoop by apache.

the class TestTaskAttempt method testContainerKillWhileCommitPending.

@Test
public void testContainerKillWhileCommitPending() throws Exception {
    ApplicationId appId = ApplicationId.newInstance(1, 2);
    ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0);
    JobId jobId = MRBuilderUtils.newJobId(appId, 1);
    TaskId taskId = MRBuilderUtils.newTaskId(jobId, 1, TaskType.MAP);
    TaskAttemptId attemptId = MRBuilderUtils.newTaskAttemptId(taskId, 0);
    Path jobFile = mock(Path.class);
    MockEventHandler eventHandler = new MockEventHandler();
    TaskAttemptListener taListener = mock(TaskAttemptListener.class);
    when(taListener.getAddress()).thenReturn(new InetSocketAddress("localhost", 0));
    JobConf jobConf = new JobConf();
    jobConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class);
    jobConf.setBoolean("fs.file.impl.disable.cache", true);
    jobConf.set(JobConf.MAPRED_MAP_TASK_ENV, "");
    jobConf.set(MRJobConfig.APPLICATION_ATTEMPT_ID, "10");
    TaskSplitMetaInfo splits = mock(TaskSplitMetaInfo.class);
    when(splits.getLocations()).thenReturn(new String[] { "127.0.0.1" });
    AppContext appCtx = mock(AppContext.class);
    ClusterInfo clusterInfo = mock(ClusterInfo.class);
    Resource resource = mock(Resource.class);
    when(appCtx.getClusterInfo()).thenReturn(clusterInfo);
    when(resource.getMemorySize()).thenReturn(1024L);
    TaskAttemptImpl taImpl = new MapTaskAttemptImpl(taskId, 1, eventHandler, jobFile, 1, splits, jobConf, taListener, new Token(), new Credentials(), SystemClock.getInstance(), appCtx);
    NodeId nid = NodeId.newInstance("127.0.0.2", 0);
    ContainerId contId = ContainerId.newContainerId(appAttemptId, 3);
    Container container = mock(Container.class);
    when(container.getId()).thenReturn(contId);
    when(container.getNodeId()).thenReturn(nid);
    when(container.getNodeHttpAddress()).thenReturn("localhost:0");
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_SCHEDULE));
    taImpl.handle(new TaskAttemptContainerAssignedEvent(attemptId, container, mock(Map.class)));
    taImpl.handle(new TaskAttemptContainerLaunchedEvent(attemptId, 0));
    assertEquals("Task attempt is not in running state", taImpl.getState(), TaskAttemptState.RUNNING);
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_COMMIT_PENDING));
    assertEquals("Task should be in COMMIT_PENDING state", TaskAttemptStateInternal.COMMIT_PENDING, taImpl.getInternalState());
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_KILL));
    assertFalse("InternalError occurred trying to handle TA_KILL", eventHandler.internalError);
    assertEquals("Task should be in KILL_CONTAINER_CLEANUP state", TaskAttemptStateInternal.KILL_CONTAINER_CLEANUP, taImpl.getInternalState());
}
Also used : Path(org.apache.hadoop.fs.Path) TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) TaskAttemptId(org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId) TaskAttemptListener(org.apache.hadoop.mapreduce.v2.app.TaskAttemptListener) InetSocketAddress(java.net.InetSocketAddress) AppContext(org.apache.hadoop.mapreduce.v2.app.AppContext) Resource(org.apache.hadoop.yarn.api.records.Resource) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) Token(org.apache.hadoop.security.token.Token) TaskAttemptEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEvent) ApplicationAttemptId(org.apache.hadoop.yarn.api.records.ApplicationAttemptId) TaskAttemptContainerAssignedEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptContainerAssignedEvent) TaskAttemptContainerLaunchedEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptContainerLaunchedEvent) ClusterInfo(org.apache.hadoop.mapreduce.v2.app.ClusterInfo) Container(org.apache.hadoop.yarn.api.records.Container) ContainerId(org.apache.hadoop.yarn.api.records.ContainerId) NodeId(org.apache.hadoop.yarn.api.records.NodeId) TaskSplitMetaInfo(org.apache.hadoop.mapreduce.split.JobSplit.TaskSplitMetaInfo) ApplicationId(org.apache.hadoop.yarn.api.records.ApplicationId) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) JobConf(org.apache.hadoop.mapred.JobConf) JobId(org.apache.hadoop.mapreduce.v2.api.records.JobId) Credentials(org.apache.hadoop.security.Credentials) Test(org.junit.Test)

Aggregations

Task (org.apache.hadoop.mapreduce.v2.app.job.Task)157 Test (org.junit.Test)153 Job (org.apache.hadoop.mapreduce.v2.app.job.Job)150 JobId (org.apache.hadoop.mapreduce.v2.api.records.JobId)107 TaskAttempt (org.apache.hadoop.mapreduce.v2.app.job.TaskAttempt)94 TaskId (org.apache.hadoop.mapreduce.v2.api.records.TaskId)79 TaskAttemptId (org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId)73 Configuration (org.apache.hadoop.conf.Configuration)68 ClientResponse (com.sun.jersey.api.client.ClientResponse)56 WebResource (com.sun.jersey.api.client.WebResource)56 TaskAttemptEvent (org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEvent)52 JSONObject (org.codehaus.jettison.json.JSONObject)46 AppContext (org.apache.hadoop.mapreduce.v2.app.AppContext)25 ApplicationId (org.apache.hadoop.yarn.api.records.ApplicationId)23 Path (org.apache.hadoop.fs.Path)22 MapTaskAttemptImpl (org.apache.hadoop.mapred.MapTaskAttemptImpl)20 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)20 HashMap (java.util.HashMap)19 ApplicationAttemptId (org.apache.hadoop.yarn.api.records.ApplicationAttemptId)19 JobConf (org.apache.hadoop.mapred.JobConf)16