Examples with TaskAttemptInfo org.apache.hadoop.mapreduce.jobhistory.JobHistoryParser.TaskAttemptInfo used on opensource projects

Example 1 with TaskAttemptInfo

use of org.apache.hadoop.mapreduce.jobhistory.JobHistoryParser.TaskAttemptInfo in project hadoop by apache.

the class TimelineEntityConverterV2 method createTaskAttemptEntities.

private Set<TimelineEntity> createTaskAttemptEntities(TaskInfo taskInfo) {
    Set<TimelineEntity> taskAttempts = new HashSet<TimelineEntity>();
    Map<TaskAttemptID, TaskAttemptInfo> taskAttemptInfoMap = taskInfo.getAllTaskAttempts();
    LOG.info("task " + taskInfo.getTaskId() + " has " + taskAttemptInfoMap.size() + " task attempts");
    for (TaskAttemptInfo taskAttemptInfo : taskAttemptInfoMap.values()) {
        TimelineEntity taskAttempt = createTaskAttemptEntity(taskAttemptInfo);
        taskAttempts.add(taskAttempt);
    }
    return taskAttempts;
}

Example 2 with TaskAttemptInfo

use of org.apache.hadoop.mapreduce.jobhistory.JobHistoryParser.TaskAttemptInfo in project hadoop by apache.

the class CompletedTask method loadAllTaskAttempts.

private void loadAllTaskAttempts() {
    if (taskAttemptsLoaded.get()) {
        return;
    }
    taskAttemptsLock.lock();
    try {
        if (taskAttemptsLoaded.get()) {
            return;
        }
        for (TaskAttemptInfo attemptHistory : taskInfo.getAllTaskAttempts().values()) {
            CompletedTaskAttempt attempt = new CompletedTaskAttempt(taskId, attemptHistory);
            reportDiagnostics.addAll(attempt.getDiagnostics());
            attempts.put(attempt.getID(), attempt);
            if (successfulAttempt == null && attemptHistory.getTaskStatus() != null && attemptHistory.getTaskStatus().equals(TaskState.SUCCEEDED.toString())) {
                successfulAttempt = TypeConverter.toYarn(attemptHistory.getAttemptId());
            }
        }
        taskAttemptsLoaded.set(true);
    } finally {
        taskAttemptsLock.unlock();
    }
}

Example 3 with TaskAttemptInfo

use of org.apache.hadoop.mapreduce.jobhistory.JobHistoryParser.TaskAttemptInfo in project hadoop by apache.

the class TestJobHistoryParsing method checkHistoryParsing.

private void checkHistoryParsing(final int numMaps, final int numReduces, final int numSuccessfulMaps) throws Exception {
    Configuration conf = new Configuration();
    conf.set(MRJobConfig.USER_NAME, System.getProperty("user.name"));
    long amStartTimeEst = System.currentTimeMillis();
    conf.setClass(NET_TOPOLOGY_NODE_SWITCH_MAPPING_IMPL_KEY, MyResolver.class, DNSToSwitchMapping.class);
    RackResolver.init(conf);
    MRApp app = new MRAppWithHistory(numMaps, numReduces, true, this.getClass().getName(), true);
    app.submit(conf);
    Job job = app.getContext().getAllJobs().values().iterator().next();
    JobId jobId = job.getID();
    LOG.info("JOBID is " + TypeConverter.fromYarn(jobId).toString());
    app.waitForState(job, JobState.SUCCEEDED);
    // make sure all events are flushed
    app.waitForState(Service.STATE.STOPPED);
    String jobhistoryDir = JobHistoryUtils.getHistoryIntermediateDoneDirForUser(conf);
    FileContext fc = null;
    try {
        fc = FileContext.getFileContext(conf);
    } catch (IOException ioe) {
        LOG.info("Can not get FileContext", ioe);
        throw (new Exception("Can not get File Context"));
    }
    if (numMaps == numSuccessfulMaps) {
        String summaryFileName = JobHistoryUtils.getIntermediateSummaryFileName(jobId);
        Path summaryFile = new Path(jobhistoryDir, summaryFileName);
        String jobSummaryString = getJobSummary(fc, summaryFile);
        Assert.assertNotNull(jobSummaryString);
        Assert.assertTrue(jobSummaryString.contains("resourcesPerMap=100"));
        Assert.assertTrue(jobSummaryString.contains("resourcesPerReduce=100"));
        Map<String, String> jobSummaryElements = new HashMap<String, String>();
        StringTokenizer strToken = new StringTokenizer(jobSummaryString, ",");
        while (strToken.hasMoreTokens()) {
            String keypair = strToken.nextToken();
            jobSummaryElements.put(keypair.split("=")[0], keypair.split("=")[1]);
        }
        Assert.assertEquals("JobId does not match", jobId.toString(), jobSummaryElements.get("jobId"));
        Assert.assertEquals("JobName does not match", "test", jobSummaryElements.get("jobName"));
        Assert.assertTrue("submitTime should not be 0", Long.parseLong(jobSummaryElements.get("submitTime")) != 0);
        Assert.assertTrue("launchTime should not be 0", Long.parseLong(jobSummaryElements.get("launchTime")) != 0);
        Assert.assertTrue("firstMapTaskLaunchTime should not be 0", Long.parseLong(jobSummaryElements.get("firstMapTaskLaunchTime")) != 0);
        Assert.assertTrue("firstReduceTaskLaunchTime should not be 0", Long.parseLong(jobSummaryElements.get("firstReduceTaskLaunchTime")) != 0);
        Assert.assertTrue("finishTime should not be 0", Long.parseLong(jobSummaryElements.get("finishTime")) != 0);
        Assert.assertEquals("Mismatch in num map slots", numSuccessfulMaps, Integer.parseInt(jobSummaryElements.get("numMaps")));
        Assert.assertEquals("Mismatch in num reduce slots", numReduces, Integer.parseInt(jobSummaryElements.get("numReduces")));
        Assert.assertEquals("User does not match", System.getProperty("user.name"), jobSummaryElements.get("user"));
        Assert.assertEquals("Queue does not match", "default", jobSummaryElements.get("queue"));
        Assert.assertEquals("Status does not match", "SUCCEEDED", jobSummaryElements.get("status"));
    }
    JobHistory jobHistory = new JobHistory();
    jobHistory.init(conf);
    HistoryFileInfo fileInfo = jobHistory.getJobFileInfo(jobId);
    JobInfo jobInfo;
    long numFinishedMaps;
    synchronized (fileInfo) {
        Path historyFilePath = fileInfo.getHistoryFile();
        FSDataInputStream in = null;
        LOG.info("JobHistoryFile is: " + historyFilePath);
        try {
            in = fc.open(fc.makeQualified(historyFilePath));
        } catch (IOException ioe) {
            LOG.info("Can not open history file: " + historyFilePath, ioe);
            throw (new Exception("Can not open History File"));
        }
        JobHistoryParser parser = new JobHistoryParser(in);
        final EventReader realReader = new EventReader(in);
        EventReader reader = Mockito.mock(EventReader.class);
        if (numMaps == numSuccessfulMaps) {
            reader = realReader;
        } else {
            // Hack!
            final AtomicInteger numFinishedEvents = new AtomicInteger(0);
            Mockito.when(reader.getNextEvent()).thenAnswer(new Answer<HistoryEvent>() {

                public HistoryEvent answer(InvocationOnMock invocation) throws IOException {
                    HistoryEvent event = realReader.getNextEvent();
                    if (event instanceof TaskFinishedEvent) {
                        numFinishedEvents.incrementAndGet();
                    }
                    if (numFinishedEvents.get() <= numSuccessfulMaps) {
                        return event;
                    } else {
                        throw new IOException("test");
                    }
                }
            });
        }
        jobInfo = parser.parse(reader);
        numFinishedMaps = computeFinishedMaps(jobInfo, numMaps, numSuccessfulMaps);
        if (numFinishedMaps != numMaps) {
            Exception parseException = parser.getParseException();
            Assert.assertNotNull("Didn't get expected parse exception", parseException);
        }
    }
    Assert.assertEquals("Incorrect username ", System.getProperty("user.name"), jobInfo.getUsername());
    Assert.assertEquals("Incorrect jobName ", "test", jobInfo.getJobname());
    Assert.assertEquals("Incorrect queuename ", "default", jobInfo.getJobQueueName());
    Assert.assertEquals("incorrect conf path", "test", jobInfo.getJobConfPath());
    Assert.assertEquals("incorrect finishedMap ", numSuccessfulMaps, numFinishedMaps);
    Assert.assertEquals("incorrect finishedReduces ", numReduces, jobInfo.getFinishedReduces());
    Assert.assertEquals("incorrect uberized ", job.isUber(), jobInfo.getUberized());
    Map<TaskID, TaskInfo> allTasks = jobInfo.getAllTasks();
    int totalTasks = allTasks.size();
    Assert.assertEquals("total number of tasks is incorrect  ", (numMaps + numReduces), totalTasks);
    // Verify aminfo
    Assert.assertEquals(1, jobInfo.getAMInfos().size());
    Assert.assertEquals(MRApp.NM_HOST, jobInfo.getAMInfos().get(0).getNodeManagerHost());
    AMInfo amInfo = jobInfo.getAMInfos().get(0);
    Assert.assertEquals(MRApp.NM_PORT, amInfo.getNodeManagerPort());
    Assert.assertEquals(MRApp.NM_HTTP_PORT, amInfo.getNodeManagerHttpPort());
    Assert.assertEquals(1, amInfo.getAppAttemptId().getAttemptId());
    Assert.assertEquals(amInfo.getAppAttemptId(), amInfo.getContainerId().getApplicationAttemptId());
    Assert.assertTrue(amInfo.getStartTime() <= System.currentTimeMillis() && amInfo.getStartTime() >= amStartTimeEst);
    ContainerId fakeCid = MRApp.newContainerId(-1, -1, -1, -1);
    // Assert at taskAttempt level
    for (TaskInfo taskInfo : allTasks.values()) {
        int taskAttemptCount = taskInfo.getAllTaskAttempts().size();
        Assert.assertEquals("total number of task attempts ", 1, taskAttemptCount);
        TaskAttemptInfo taInfo = taskInfo.getAllTaskAttempts().values().iterator().next();
        Assert.assertNotNull(taInfo.getContainerId());
        // Verify the wrong ctor is not being used. Remove after mrv1 is removed.
        Assert.assertFalse(taInfo.getContainerId().equals(fakeCid));
    }
    // Deep compare Job and JobInfo
    for (Task task : job.getTasks().values()) {
        TaskInfo taskInfo = allTasks.get(TypeConverter.fromYarn(task.getID()));
        Assert.assertNotNull("TaskInfo not found", taskInfo);
        for (TaskAttempt taskAttempt : task.getAttempts().values()) {
            TaskAttemptInfo taskAttemptInfo = taskInfo.getAllTaskAttempts().get(TypeConverter.fromYarn((taskAttempt.getID())));
            Assert.assertNotNull("TaskAttemptInfo not found", taskAttemptInfo);
            Assert.assertEquals("Incorrect shuffle port for task attempt", taskAttempt.getShufflePort(), taskAttemptInfo.getShufflePort());
            if (numMaps == numSuccessfulMaps) {
                Assert.assertEquals(MRApp.NM_HOST, taskAttemptInfo.getHostname());
                Assert.assertEquals(MRApp.NM_PORT, taskAttemptInfo.getPort());
                // Verify rack-name
                Assert.assertEquals("rack-name is incorrect", taskAttemptInfo.getRackname(), RACK_NAME);
            }
        }
    }
    // test output for HistoryViewer
    PrintStream stdps = System.out;
    try {
        System.setOut(new PrintStream(outContent));
        HistoryViewer viewer;
        synchronized (fileInfo) {
            viewer = new HistoryViewer(fc.makeQualified(fileInfo.getHistoryFile()).toString(), conf, true);
        }
        viewer.print();
        for (TaskInfo taskInfo : allTasks.values()) {
            String test = (taskInfo.getTaskStatus() == null ? "" : taskInfo.getTaskStatus()) + " " + taskInfo.getTaskType() + " task list for " + taskInfo.getTaskId().getJobID();
            Assert.assertTrue(outContent.toString().indexOf(test) > 0);
            Assert.assertTrue(outContent.toString().indexOf(taskInfo.getTaskId().toString()) > 0);
        }
    } finally {
        System.setOut(stdps);
    }
}

Example 4 with TaskAttemptInfo

use of org.apache.hadoop.mapreduce.jobhistory.JobHistoryParser.TaskAttemptInfo in project hadoop by apache.

the class TestRecovery method testRecoveryAllAttemptsKilled.

@Test
public void testRecoveryAllAttemptsKilled() {
    LOG.info("--- START:  testRecoveryAllAttemptsKilled ---");
    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.KILLED);
    mockTaskAttempts.put(taId1, mockTAinfo1);
    TaskAttemptID taId2 = new TaskAttemptID(taskID, 1);
    TaskAttemptInfo mockTAinfo2 = getMockTaskAttemptInfo(taId2, TaskAttemptState.KILLED);
    mockTaskAttempts.put(taId2, mockTAinfo2);
    OutputCommitter mockCommitter = mock(OutputCommitter.class);
    TaskInfo mockTaskInfo = mock(TaskInfo.class);
    when(mockTaskInfo.getTaskStatus()).thenReturn("KILLED");
    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.KILLED);
    finalAttemptStates.put(taId2, TaskAttemptState.KILLED);
    List<EventType> jobHistoryEvents = new ArrayList<EventType>();
    jobHistoryEvents.add(EventType.TASK_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_KILLED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_STARTED);
    jobHistoryEvents.add(EventType.MAP_ATTEMPT_KILLED);
    jobHistoryEvents.add(EventType.TASK_FAILED);
    recoveryChecker(recoverMapTask, TaskState.KILLED, finalAttemptStates, arg, jobHistoryEvents, 2L, 0L);
}

Example 5 with TaskAttemptInfo

use of org.apache.hadoop.mapreduce.jobhistory.JobHistoryParser.TaskAttemptInfo 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);
}