Examples with OutputCommitter org.apache.hadoop.mapreduce.OutputCommitter used on opensource projects

Example 1 with OutputCommitter

use of org.apache.hadoop.mapreduce.OutputCommitter in project flink by apache.

the class HadoopOutputFormatTest method testCloseWithNeedsTaskCommitFalse.

@Test
public void testCloseWithNeedsTaskCommitFalse() throws Exception {
    RecordWriter<String, Long> recordWriter = Mockito.mock(DummyRecordWriter.class);
    OutputCommitter outputCommitter = setupOutputCommitter(false);
    HadoopOutputFormat<String, Long> hadoopOutputFormat = setupHadoopOutputFormat(new DummyOutputFormat(), Job.getInstance(), recordWriter, outputCommitter, new Configuration());
    hadoopOutputFormat.close();
    verify(outputCommitter, times(0)).commitTask(any(TaskAttemptContext.class));
    verify(recordWriter, times(1)).close(any(TaskAttemptContext.class));
}

Example 2 with OutputCommitter

use of org.apache.hadoop.mapreduce.OutputCommitter in project flink by apache.

the class HadoopOutputFormatTest method setupOutputCommitter.

private OutputCommitter setupOutputCommitter(boolean needsTaskCommit) throws IOException {
    OutputCommitter outputCommitter = Mockito.mock(OutputCommitter.class);
    when(outputCommitter.needsTaskCommit(any(TaskAttemptContext.class))).thenReturn(needsTaskCommit);
    doNothing().when(outputCommitter).commitTask(any(TaskAttemptContext.class));
    return outputCommitter;
}

Example 3 with OutputCommitter

use of org.apache.hadoop.mapreduce.OutputCommitter in project hadoop by apache.

the class TestRecovery method testRecoverySuccessAttempt.

@Test
public void testRecoverySuccessAttempt() {
    LOG.info("--- START: testRecoverySuccessAttempt ---");
    long clusterTimestamp = System.currentTimeMillis();
    EventHandler mockEventHandler = mock(EventHandler.class);
    MapTaskImpl recoverMapTask = getMockMapTask(clusterTimestamp, mockEventHandler);
    TaskId taskId = recoverMapTask.getID();
    JobID jobID = new JobID(Long.toString(clusterTimestamp), 1);
    TaskID taskID = new TaskID(jobID, org.apache.hadoop.mapreduce.TaskType.MAP, taskId.getId());
    //Mock up the TaskAttempts
    Map<TaskAttemptID, TaskAttemptInfo> mockTaskAttempts = new HashMap<TaskAttemptID, TaskAttemptInfo>();
    TaskAttemptID taId1 = new TaskAttemptID(taskID, 2);
    TaskAttemptInfo mockTAinfo1 = getMockTaskAttemptInfo(taId1, TaskAttemptState.SUCCEEDED);
    mockTaskAttempts.put(taId1, mockTAinfo1);
    TaskAttemptID taId2 = new TaskAttemptID(taskID, 1);
    TaskAttemptInfo mockTAinfo2 = getMockTaskAttemptInfo(taId2, TaskAttemptState.FAILED);
    mockTaskAttempts.put(taId2, mockTAinfo2);
    OutputCommitter mockCommitter = mock(OutputCommitter.class);
    TaskInfo mockTaskInfo = mock(TaskInfo.class);
    when(mockTaskInfo.getTaskStatus()).thenReturn("SUCCEEDED");
    when(mockTaskInfo.getTaskId()).thenReturn(taskID);
    when(mockTaskInfo.getAllTaskAttempts()).thenReturn(mockTaskAttempts);
    recoverMapTask.handle(new TaskRecoverEvent(taskId, mockTaskInfo, mockCommitter, true));
    ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class);
    verify(mockEventHandler, atLeast(1)).handle((org.apache.hadoop.yarn.event.Event) arg.capture());
    Map<TaskAttemptID, TaskAttemptState> finalAttemptStates = new HashMap<TaskAttemptID, TaskAttemptState>();
    finalAttemptStates.put(taId1, TaskAttemptState.SUCCEEDED);
    finalAttemptStates.put(taId2, TaskAttemptState.FAILED);
    List<EventType> jobHistoryEvents = new ArrayList<EventType>();
    jobHistoryEvents.add(EventType.TASK_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_FINISHED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_FAILED);
    jobHistoryEvents.add(EventType.TASK_FINISHED);
    recoveryChecker(recoverMapTask, TaskState.SUCCEEDED, finalAttemptStates, arg, jobHistoryEvents, 2L, 1L);
}

Example 4 with OutputCommitter

use of org.apache.hadoop.mapreduce.OutputCommitter in project hadoop by apache.

the class TestRecovery method testRecoveryTaskSuccessAllAttemptsFail.

@Test
public void testRecoveryTaskSuccessAllAttemptsFail() {
    LOG.info("--- START:  testRecoveryTaskSuccessAllAttemptsFail ---");
    long clusterTimestamp = System.currentTimeMillis();
    EventHandler mockEventHandler = mock(EventHandler.class);
    MapTaskImpl recoverMapTask = getMockMapTask(clusterTimestamp, mockEventHandler);
    TaskId taskId = recoverMapTask.getID();
    JobID jobID = new JobID(Long.toString(clusterTimestamp), 1);
    TaskID taskID = new TaskID(jobID, org.apache.hadoop.mapreduce.TaskType.MAP, taskId.getId());
    //Mock up the TaskAttempts
    Map<TaskAttemptID, TaskAttemptInfo> mockTaskAttempts = new HashMap<TaskAttemptID, TaskAttemptInfo>();
    TaskAttemptID taId1 = new TaskAttemptID(taskID, 2);
    TaskAttemptInfo mockTAinfo1 = getMockTaskAttemptInfo(taId1, TaskAttemptState.FAILED);
    mockTaskAttempts.put(taId1, mockTAinfo1);
    TaskAttemptID taId2 = new TaskAttemptID(taskID, 1);
    TaskAttemptInfo mockTAinfo2 = getMockTaskAttemptInfo(taId2, TaskAttemptState.FAILED);
    mockTaskAttempts.put(taId2, mockTAinfo2);
    OutputCommitter mockCommitter = mock(OutputCommitter.class);
    TaskInfo mockTaskInfo = mock(TaskInfo.class);
    when(mockTaskInfo.getTaskStatus()).thenReturn("SUCCEEDED");
    when(mockTaskInfo.getTaskId()).thenReturn(taskID);
    when(mockTaskInfo.getAllTaskAttempts()).thenReturn(mockTaskAttempts);
    recoverMapTask.handle(new TaskRecoverEvent(taskId, mockTaskInfo, mockCommitter, true));
    ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class);
    verify(mockEventHandler, atLeast(1)).handle((org.apache.hadoop.yarn.event.Event) arg.capture());
    Map<TaskAttemptID, TaskAttemptState> finalAttemptStates = new HashMap<TaskAttemptID, TaskAttemptState>();
    finalAttemptStates.put(taId1, TaskAttemptState.FAILED);
    finalAttemptStates.put(taId2, TaskAttemptState.FAILED);
    // check for one new attempt launched since successful attempt not found
    TaskAttemptID taId3 = new TaskAttemptID(taskID, 2000);
    finalAttemptStates.put(taId3, TaskAttemptState.NEW);
    List<EventType> jobHistoryEvents = new ArrayList<EventType>();
    jobHistoryEvents.add(EventType.TASK_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_FAILED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_FAILED);
    recoveryChecker(recoverMapTask, TaskState.RUNNING, finalAttemptStates, arg, jobHistoryEvents, 2L, 2L);
}

Example 5 with OutputCommitter

use of org.apache.hadoop.mapreduce.OutputCommitter in project hadoop by apache.

the class TestRecovery method writeOutput.

private void writeOutput(TaskAttempt attempt, Configuration conf) throws Exception {
    TaskAttemptContext tContext = new TaskAttemptContextImpl(conf, TypeConverter.fromYarn(attempt.getID()));
    TextOutputFormat<?, ?> theOutputFormat = new TextOutputFormat();
    RecordWriter theRecordWriter = theOutputFormat.getRecordWriter(tContext);
    NullWritable nullWritable = NullWritable.get();
    try {
        theRecordWriter.write(key1, val1);
        theRecordWriter.write(null, nullWritable);
        theRecordWriter.write(null, val1);
        theRecordWriter.write(nullWritable, val2);
        theRecordWriter.write(key2, nullWritable);
        theRecordWriter.write(key1, null);
        theRecordWriter.write(null, null);
        theRecordWriter.write(key2, val2);
    } finally {
        theRecordWriter.close(tContext);
    }
    OutputFormat outputFormat = ReflectionUtils.newInstance(tContext.getOutputFormatClass(), conf);
    OutputCommitter committer = outputFormat.getOutputCommitter(tContext);
    committer.commitTask(tContext);
}