Search in sources :

Example 6 with MapTaskAttemptImpl

use of org.apache.hadoop.mapred.MapTaskAttemptImpl in project hadoop by apache.

the class TestTaskAttempt method createMapTaskAttemptImplForTest.

private TaskAttemptImpl createMapTaskAttemptImplForTest(EventHandler eventHandler, TaskSplitMetaInfo taskSplitMetaInfo, Clock clock) {
    ApplicationId appId = ApplicationId.newInstance(1, 1);
    JobId jobId = MRBuilderUtils.newJobId(appId, 1);
    TaskId taskId = MRBuilderUtils.newTaskId(jobId, 1, TaskType.MAP);
    TaskAttemptListener taListener = mock(TaskAttemptListener.class);
    Path jobFile = mock(Path.class);
    JobConf jobConf = new JobConf();
    TaskAttemptImpl taImpl = new MapTaskAttemptImpl(taskId, 1, eventHandler, jobFile, 1, taskSplitMetaInfo, jobConf, taListener, null, null, clock, null);
    return taImpl;
}
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) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) 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)

Example 7 with MapTaskAttemptImpl

use of org.apache.hadoop.mapred.MapTaskAttemptImpl 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 8 with MapTaskAttemptImpl

use of org.apache.hadoop.mapred.MapTaskAttemptImpl 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 9 with MapTaskAttemptImpl

use of org.apache.hadoop.mapred.MapTaskAttemptImpl in project hadoop by apache.

the class TestShuffleProvider method testShuffleProviders.

@Test
public void testShuffleProviders() throws Exception {
    ApplicationId appId = ApplicationId.newInstance(1, 1);
    JobId jobId = MRBuilderUtils.newJobId(appId, 1);
    TaskId taskId = MRBuilderUtils.newTaskId(jobId, 1, TaskType.MAP);
    Path jobFile = mock(Path.class);
    EventHandler eventHandler = mock(EventHandler.class);
    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(YarnConfiguration.NM_AUX_SERVICES, TestShuffleHandler1.MAPREDUCE_TEST_SHUFFLE_SERVICEID + "," + TestShuffleHandler2.MAPREDUCE_TEST_SHUFFLE_SERVICEID);
    String serviceName = TestShuffleHandler1.MAPREDUCE_TEST_SHUFFLE_SERVICEID;
    String serviceStr = String.format(YarnConfiguration.NM_AUX_SERVICE_FMT, serviceName);
    jobConf.set(serviceStr, TestShuffleHandler1.class.getName());
    serviceName = TestShuffleHandler2.MAPREDUCE_TEST_SHUFFLE_SERVICEID;
    serviceStr = String.format(YarnConfiguration.NM_AUX_SERVICE_FMT, serviceName);
    jobConf.set(serviceStr, TestShuffleHandler2.class.getName());
    jobConf.set(MRJobConfig.MAPREDUCE_JOB_SHUFFLE_PROVIDER_SERVICES, TestShuffleHandler1.MAPREDUCE_TEST_SHUFFLE_SERVICEID + "," + TestShuffleHandler2.MAPREDUCE_TEST_SHUFFLE_SERVICEID);
    Credentials credentials = new Credentials();
    Token<JobTokenIdentifier> jobToken = new Token<JobTokenIdentifier>(("tokenid").getBytes(), ("tokenpw").getBytes(), new Text("tokenkind"), new Text("tokenservice"));
    TaskAttemptImpl taImpl = new MapTaskAttemptImpl(taskId, 1, eventHandler, jobFile, 1, mock(TaskSplitMetaInfo.class), jobConf, taListener, jobToken, credentials, SystemClock.getInstance(), null);
    jobConf.set(MRJobConfig.APPLICATION_ATTEMPT_ID, taImpl.getID().toString());
    ContainerLaunchContext launchCtx = TaskAttemptImpl.createContainerLaunchContext(null, jobConf, jobToken, taImpl.createRemoteTask(), TypeConverter.fromYarn(jobId), mock(WrappedJvmID.class), taListener, credentials);
    Map<String, ByteBuffer> serviceDataMap = launchCtx.getServiceData();
    Assert.assertNotNull("TestShuffleHandler1 is missing", serviceDataMap.get(TestShuffleHandler1.MAPREDUCE_TEST_SHUFFLE_SERVICEID));
    Assert.assertNotNull("TestShuffleHandler2 is missing", serviceDataMap.get(TestShuffleHandler2.MAPREDUCE_TEST_SHUFFLE_SERVICEID));
    // 2 that we entered + 1 for the built-in shuffle-provider
    Assert.assertTrue("mismatch number of services in map", serviceDataMap.size() == 3);
}
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) InetSocketAddress(java.net.InetSocketAddress) EventHandler(org.apache.hadoop.yarn.event.EventHandler) JobTokenIdentifier(org.apache.hadoop.mapreduce.security.token.JobTokenIdentifier) MapTaskAttemptImpl(org.apache.hadoop.mapred.MapTaskAttemptImpl) Token(org.apache.hadoop.security.token.Token) Text(org.apache.hadoop.io.Text) ContainerLaunchContext(org.apache.hadoop.yarn.api.records.ContainerLaunchContext) ByteBuffer(java.nio.ByteBuffer) WrappedJvmID(org.apache.hadoop.mapred.WrappedJvmID) 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 10 with MapTaskAttemptImpl

use of org.apache.hadoop.mapred.MapTaskAttemptImpl in project hadoop by apache.

the class TestTaskAttempt method testTooManyFetchFailureAfterKill.

@Test
public void testTooManyFetchFailureAfterKill() 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);
    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);
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_KILL));
    assertEquals("Task attempt is not in KILLED state", taImpl.getState(), TaskAttemptState.KILLED);
    taImpl.handle(new TaskAttemptEvent(attemptId, TaskAttemptEventType.TA_TOO_MANY_FETCH_FAILURE));
    assertEquals("Task attempt is not in KILLED state, still", taImpl.getState(), TaskAttemptState.KILLED);
    assertFalse("InternalError occurred trying to handle TA_CONTAINER_CLEANED", eventHandler.internalError);
}
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

Path (org.apache.hadoop.fs.Path)16 JobConf (org.apache.hadoop.mapred.JobConf)16 MapTaskAttemptImpl (org.apache.hadoop.mapred.MapTaskAttemptImpl)16 JobId (org.apache.hadoop.mapreduce.v2.api.records.JobId)16 TaskId (org.apache.hadoop.mapreduce.v2.api.records.TaskId)16 TaskAttemptListener (org.apache.hadoop.mapreduce.v2.app.TaskAttemptListener)16 ApplicationId (org.apache.hadoop.yarn.api.records.ApplicationId)16 TaskSplitMetaInfo (org.apache.hadoop.mapreduce.split.JobSplit.TaskSplitMetaInfo)15 Credentials (org.apache.hadoop.security.Credentials)15 Token (org.apache.hadoop.security.token.Token)15 InetSocketAddress (java.net.InetSocketAddress)14 Test (org.junit.Test)13 TaskAttemptId (org.apache.hadoop.mapreduce.v2.api.records.TaskAttemptId)12 AppContext (org.apache.hadoop.mapreduce.v2.app.AppContext)12 TaskAttemptEvent (org.apache.hadoop.mapreduce.v2.app.job.event.TaskAttemptEvent)12 ApplicationAttemptId (org.apache.hadoop.yarn.api.records.ApplicationAttemptId)12 Container (org.apache.hadoop.yarn.api.records.Container)12 ContainerId (org.apache.hadoop.yarn.api.records.ContainerId)12 NodeId (org.apache.hadoop.yarn.api.records.NodeId)12 ClusterInfo (org.apache.hadoop.mapreduce.v2.app.ClusterInfo)11