Search in sources :

Example 61 with TaskId

use of org.apache.hadoop.mapreduce.v2.api.records.TaskId 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)

Example 62 with TaskId

use of org.apache.hadoop.mapreduce.v2.api.records.TaskId in project hadoop by apache.

the class TestTaskAttempt method testLaunchFailedWhileKilling.

@Test
public void testLaunchFailedWhileKilling() 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" });
    TaskAttemptImpl taImpl = new MapTaskAttemptImpl(taskId, 1, eventHandler, jobFile, 1, splits, jobConf, taListener, new Token(), new Credentials(), SystemClock.getInstance(), null);
    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);
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_SCHEDULE));
    taImpl.handle(new TaskAttemptContainerAssignedEvent(attemptId, container, mock(Map.class)));
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_KILL));
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_CONTAINER_CLEANED));
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_CONTAINER_LAUNCH_FAILED));
    assertFalse(eventHandler.internalError);
    assertEquals("Task attempt is not assigned on the local node", Locality.NODE_LOCAL, taImpl.getLocality());
}
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) 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) 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)

Example 63 with TaskId

use of org.apache.hadoop.mapreduce.v2.api.records.TaskId in project hadoop by apache.

the class TestTaskImpl method testScheduleTask.

@Test
public /**
   * {@link TaskState#NEW}->{@link TaskState#SCHEDULED}
   */
void testScheduleTask() {
    LOG.info("--- START: testScheduleTask ---");
    mockTask = createMockTask(TaskType.MAP);
    TaskId taskId = getNewTaskID();
    scheduleTaskAttempt(taskId);
}
Also used : TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) Test(org.junit.Test)

Example 64 with TaskId

use of org.apache.hadoop.mapreduce.v2.api.records.TaskId in project hadoop by apache.

the class TestTaskImpl method testFailedTransitions.

@Test
public void testFailedTransitions() {
    mockTask = new MockTaskImpl(jobId, partition, dispatcher.getEventHandler(), remoteJobConfFile, conf, taskAttemptListener, jobToken, credentials, clock, startCount, metrics, appContext, TaskType.MAP) {

        @Override
        protected int getMaxAttempts() {
            return 1;
        }
    };
    TaskId taskId = getNewTaskID();
    scheduleTaskAttempt(taskId);
    launchTaskAttempt(getLastAttempt().getAttemptId());
    // add three more speculative attempts
    mockTask.handle(new TaskTAttemptEvent(getLastAttempt().getAttemptId(), TaskEventType.T_ADD_SPEC_ATTEMPT));
    launchTaskAttempt(getLastAttempt().getAttemptId());
    mockTask.handle(new TaskTAttemptEvent(getLastAttempt().getAttemptId(), TaskEventType.T_ADD_SPEC_ATTEMPT));
    launchTaskAttempt(getLastAttempt().getAttemptId());
    mockTask.handle(new TaskTAttemptEvent(getLastAttempt().getAttemptId(), TaskEventType.T_ADD_SPEC_ATTEMPT));
    launchTaskAttempt(getLastAttempt().getAttemptId());
    assertEquals(4, taskAttempts.size());
    // have the first attempt fail, verify task failed due to no retries
    MockTaskAttemptImpl taskAttempt = taskAttempts.get(0);
    taskAttempt.setState(TaskAttemptState.FAILED);
    mockTask.handle(new TaskTAttemptEvent(taskAttempt.getAttemptId(), TaskEventType.T_ATTEMPT_FAILED));
    assertEquals(TaskState.FAILED, mockTask.getState());
    // verify task can no longer be killed
    mockTask.handle(new TaskEvent(taskId, TaskEventType.T_KILL));
    assertEquals(TaskState.FAILED, mockTask.getState());
    // verify speculative doesn't launch new tasks
    mockTask.handle(new TaskTAttemptEvent(getLastAttempt().getAttemptId(), TaskEventType.T_ADD_SPEC_ATTEMPT));
    mockTask.handle(new TaskTAttemptEvent(getLastAttempt().getAttemptId(), TaskEventType.T_ATTEMPT_LAUNCHED));
    assertEquals(TaskState.FAILED, mockTask.getState());
    assertEquals(4, taskAttempts.size());
    // verify attempt events from active tasks don't knock task out of FAILED
    taskAttempt = taskAttempts.get(1);
    taskAttempt.setState(TaskAttemptState.COMMIT_PENDING);
    mockTask.handle(new TaskTAttemptEvent(taskAttempt.getAttemptId(), TaskEventType.T_ATTEMPT_COMMIT_PENDING));
    assertEquals(TaskState.FAILED, mockTask.getState());
    taskAttempt.setState(TaskAttemptState.FAILED);
    mockTask.handle(new TaskTAttemptEvent(taskAttempt.getAttemptId(), TaskEventType.T_ATTEMPT_FAILED));
    assertEquals(TaskState.FAILED, mockTask.getState());
    taskAttempt = taskAttempts.get(2);
    taskAttempt.setState(TaskAttemptState.SUCCEEDED);
    mockTask.handle(new TaskTAttemptEvent(taskAttempt.getAttemptId(), TaskEventType.T_ATTEMPT_SUCCEEDED));
    assertEquals(TaskState.FAILED, mockTask.getState());
    taskAttempt = taskAttempts.get(3);
    taskAttempt.setState(TaskAttemptState.KILLED);
    mockTask.handle(new TaskTAttemptKilledEvent(taskAttempt.getAttemptId(), false));
    assertEquals(TaskState.FAILED, mockTask.getState());
}
Also used : TaskTAttemptKilledEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskTAttemptKilledEvent) TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) TaskEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskEvent) TaskTAttemptEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskTAttemptEvent) Test(org.junit.Test)

Example 65 with TaskId

use of org.apache.hadoop.mapreduce.v2.api.records.TaskId in project hadoop by apache.

the class TestTaskImpl method testFailedTransitionWithHangingSpeculativeMap.

@Test
public void testFailedTransitionWithHangingSpeculativeMap() {
    mockTask = new MockTaskImpl(jobId, partition, new PartialAttemptEventHandler(), remoteJobConfFile, conf, taskAttemptListener, jobToken, credentials, clock, startCount, metrics, appContext, TaskType.MAP) {

        @Override
        protected int getMaxAttempts() {
            return 4;
        }
    };
    // start a new task, schedule and launch a new attempt
    TaskId taskId = getNewTaskID();
    scheduleTaskAttempt(taskId);
    launchTaskAttempt(getLastAttempt().getAttemptId());
    // add a speculative attempt(#2), but not launch it
    mockTask.handle(new TaskTAttemptEvent(getLastAttempt().getAttemptId(), TaskEventType.T_ADD_SPEC_ATTEMPT));
    // have the first attempt(#1) fail, verify task still running since the
    // max attempts is 4
    MockTaskAttemptImpl taskAttempt = taskAttempts.get(0);
    taskAttempt.setState(TaskAttemptState.FAILED);
    mockTask.handle(new TaskTAttemptEvent(taskAttempt.getAttemptId(), TaskEventType.T_ATTEMPT_FAILED));
    assertEquals(TaskState.RUNNING, mockTask.getState());
    // verify a new attempt(#3) added because the speculative attempt(#2)
    // is hanging
    assertEquals(3, taskAttempts.size());
    // verify the speculative attempt(#2) is not a rescheduled attempt
    assertEquals(false, taskAttempts.get(1).getRescheduled());
    // verify the third attempt is a rescheduled attempt
    assertEquals(true, taskAttempts.get(2).getRescheduled());
    // now launch the latest attempt(#3) and set the internal state to running
    launchTaskAttempt(getLastAttempt().getAttemptId());
    // have the speculative attempt(#2) fail, verify task still since it
    // hasn't reach the max attempts which is 4
    MockTaskAttemptImpl taskAttempt1 = taskAttempts.get(1);
    taskAttempt1.setState(TaskAttemptState.FAILED);
    mockTask.handle(new TaskTAttemptEvent(taskAttempt1.getAttemptId(), TaskEventType.T_ATTEMPT_FAILED));
    assertEquals(TaskState.RUNNING, mockTask.getState());
    // verify there's no new attempt added because of the running attempt(#3)
    assertEquals(3, taskAttempts.size());
}
Also used : TaskId(org.apache.hadoop.mapreduce.v2.api.records.TaskId) TaskTAttemptEvent(org.apache.hadoop.mapreduce.v2.app.job.event.TaskTAttemptEvent) Test(org.junit.Test)

Aggregations

TaskId (org.apache.hadoop.mapreduce.v2.api.records.TaskId)113 Test (org.junit.Test)75 Task (org.apache.hadoop.mapreduce.v2.app.job.Task)69 Job (org.apache.hadoop.mapreduce.v2.app.job.Job)60 JobId (org.apache.hadoop.mapreduce.v2.api.records.JobId)58 TaskAttemptId (org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId)56 TaskAttempt (org.apache.hadoop.mapreduce.v2.app.job.TaskAttempt)42 Configuration (org.apache.hadoop.conf.Configuration)29 AppContext (org.apache.hadoop.mapreduce.v2.app.AppContext)24 TaskAttemptEvent (org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEvent)24 Path (org.apache.hadoop.fs.Path)23 ApplicationId (org.apache.hadoop.yarn.api.records.ApplicationId)22 HashMap (java.util.HashMap)20 JobConf (org.apache.hadoop.mapred.JobConf)17 ApplicationAttemptId (org.apache.hadoop.yarn.api.records.ApplicationAttemptId)17 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)17 MapTaskAttemptImpl (org.apache.hadoop.mapred.MapTaskAttemptImpl)16 TaskAttemptListener (org.apache.hadoop.mapreduce.v2.app.TaskAttemptListener)16 InetSocketAddress (java.net.InetSocketAddress)15 TaskSplitMetaInfo (org.apache.hadoop.mapreduce.split.JobSplit.TaskSplitMetaInfo)15