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

Example 6 with JobInfo

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

the class JobHistoryFileReplayMapperV1 method map.

public void map(IntWritable key, IntWritable val, Context context) throws IOException {
    // collect the apps it needs to process
    TimelineClient tlc = new TimelineClientImpl();
    TimelineEntityConverterV1 converter = new TimelineEntityConverterV1();
    JobHistoryFileReplayHelper helper = new JobHistoryFileReplayHelper(context);
    int replayMode = helper.getReplayMode();
    Collection<JobFiles> jobs = helper.getJobFiles();
    JobHistoryFileParser parser = helper.getParser();
    if (jobs.isEmpty()) {
        LOG.info(context.getTaskAttemptID().getTaskID() + " will process no jobs");
    } else {
        LOG.info(context.getTaskAttemptID().getTaskID() + " will process " + jobs.size() + " jobs");
    }
    for (JobFiles job : jobs) {
        // process each job
        String jobIdStr = job.getJobId();
        LOG.info("processing " + jobIdStr + "...");
        JobId jobId = TypeConverter.toYarn(JobID.forName(jobIdStr));
        ApplicationId appId = jobId.getAppId();
        try {
            // parse the job info and configuration
            Path historyFilePath = job.getJobHistoryFilePath();
            Path confFilePath = job.getJobConfFilePath();
            if ((historyFilePath == null) || (confFilePath == null)) {
                continue;
            }
            JobInfo jobInfo = parser.parseHistoryFile(historyFilePath);
            Configuration jobConf = parser.parseConfiguration(confFilePath);
            LOG.info("parsed the job history file and the configuration file for job " + jobIdStr);
            // create entities from job history and write them
            long totalTime = 0;
            Set<TimelineEntity> entitySet = converter.createTimelineEntities(jobInfo, jobConf);
            LOG.info("converted them into timeline entities for job " + jobIdStr);
            // use the current user for this purpose
            UserGroupInformation ugi = UserGroupInformation.getCurrentUser();
            long startWrite = System.nanoTime();
            try {
                switch(replayMode) {
                    case JobHistoryFileReplayHelper.WRITE_ALL_AT_ONCE:
                        writeAllEntities(tlc, entitySet, ugi);
                        break;
                    case JobHistoryFileReplayHelper.WRITE_PER_ENTITY:
                        writePerEntity(tlc, entitySet, ugi);
                        break;
                    default:
                        break;
                }
            } catch (Exception e) {
                context.getCounter(PerfCounters.TIMELINE_SERVICE_WRITE_FAILURES).increment(1);
                LOG.error("writing to the timeline service failed", e);
            }
            long endWrite = System.nanoTime();
            totalTime += TimeUnit.NANOSECONDS.toMillis(endWrite - startWrite);
            int numEntities = entitySet.size();
            LOG.info("wrote " + numEntities + " entities in " + totalTime + " ms");
            context.getCounter(PerfCounters.TIMELINE_SERVICE_WRITE_TIME).increment(totalTime);
            context.getCounter(PerfCounters.TIMELINE_SERVICE_WRITE_COUNTER).increment(numEntities);
        } finally {
            // move it along
            context.progress();
        }
    }
}

Example 7 with JobInfo

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

the class MRAppMaster method parsePreviousJobHistory.

private void parsePreviousJobHistory() throws IOException {
    FSDataInputStream in = getPreviousJobHistoryStream(getConfig(), appAttemptID);
    JobHistoryParser parser = new JobHistoryParser(in);
    JobInfo jobInfo = parser.parse();
    Exception parseException = parser.getParseException();
    if (parseException != null) {
        LOG.info("Got an error parsing job-history file" + ", ignoring incomplete events.", parseException);
    }
    Map<org.apache.hadoop.mapreduce.TaskID, TaskInfo> taskInfos = jobInfo.getAllTasks();
    for (TaskInfo taskInfo : taskInfos.values()) {
        if (TaskState.SUCCEEDED.toString().equals(taskInfo.getTaskStatus())) {
            Iterator<Entry<TaskAttemptID, TaskAttemptInfo>> taskAttemptIterator = taskInfo.getAllTaskAttempts().entrySet().iterator();
            while (taskAttemptIterator.hasNext()) {
                Map.Entry<TaskAttemptID, TaskAttemptInfo> currentEntry = taskAttemptIterator.next();
                if (!jobInfo.getAllCompletedTaskAttempts().containsKey(currentEntry.getKey())) {
                    taskAttemptIterator.remove();
                }
            }
            completedTasksFromPreviousRun.put(TypeConverter.toYarn(taskInfo.getTaskId()), taskInfo);
            LOG.info("Read from history task " + TypeConverter.toYarn(taskInfo.getTaskId()));
        }
    }
    LOG.info("Read completed tasks from history " + completedTasksFromPreviousRun.size());
    recoveredJobStartTime = jobInfo.getLaunchTime();
    // recover AMInfos
    List<JobHistoryParser.AMInfo> jhAmInfoList = jobInfo.getAMInfos();
    if (jhAmInfoList != null) {
        for (JobHistoryParser.AMInfo jhAmInfo : jhAmInfoList) {
            AMInfo amInfo = MRBuilderUtils.newAMInfo(jhAmInfo.getAppAttemptId(), jhAmInfo.getStartTime(), jhAmInfo.getContainerId(), jhAmInfo.getNodeManagerHost(), jhAmInfo.getNodeManagerPort(), jhAmInfo.getNodeManagerHttpPort());
            amInfos.add(amInfo);
        }
    }
}

Example 8 with JobInfo

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

the class TestJobHistoryParsing method testDiagnosticsForKilledJob.

@Test(timeout = 60000)
public void testDiagnosticsForKilledJob() throws Exception {
    LOG.info("STARTING testDiagnosticsForKilledJob");
    try {
        final Configuration conf = new Configuration();
        conf.setClass(NET_TOPOLOGY_NODE_SWITCH_MAPPING_IMPL_KEY, MyResolver.class, DNSToSwitchMapping.class);
        RackResolver.init(conf);
        MRApp app = new MRAppWithHistoryWithJobKilled(2, 1, true, this.getClass().getName(), true);
        app.submit(conf);
        Job job = app.getContext().getAllJobs().values().iterator().next();
        JobId jobId = job.getID();
        app.waitForState(job, JobState.KILLED);
        // make sure all events are flushed
        app.waitForState(Service.STATE.STOPPED);
        JobHistory jobHistory = new JobHistory();
        jobHistory.init(conf);
        HistoryFileInfo fileInfo = jobHistory.getJobFileInfo(jobId);
        JobHistoryParser parser;
        JobInfo jobInfo;
        synchronized (fileInfo) {
            Path historyFilePath = fileInfo.getHistoryFile();
            FSDataInputStream in = null;
            FileContext fc = null;
            try {
                fc = FileContext.getFileContext(conf);
                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"));
            }
            parser = new JobHistoryParser(in);
            jobInfo = parser.parse();
        }
        Exception parseException = parser.getParseException();
        assertNull("Caught an expected exception " + parseException, parseException);
        final List<String> originalDiagnostics = job.getDiagnostics();
        final String historyError = jobInfo.getErrorInfo();
        assertTrue("No original diagnostics for a failed job", originalDiagnostics != null && !originalDiagnostics.isEmpty());
        assertNotNull("No history error info for a failed job ", historyError);
        for (String diagString : originalDiagnostics) {
            assertTrue(historyError.contains(diagString));
        }
        assertTrue("No killed message in diagnostics", historyError.contains(JobImpl.JOB_KILLED_DIAG));
    } finally {
        LOG.info("FINISHED testDiagnosticsForKilledJob");
    }
}

Example 9 with JobInfo

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

the class TestJobHistoryParsing method testCountersForFailedTask.

@Test(timeout = 60000)
public void testCountersForFailedTask() throws Exception {
    LOG.info("STARTING testCountersForFailedTask");
    try {
        Configuration conf = new Configuration();
        conf.setClass(NET_TOPOLOGY_NODE_SWITCH_MAPPING_IMPL_KEY, MyResolver.class, DNSToSwitchMapping.class);
        RackResolver.init(conf);
        MRApp app = new MRAppWithHistoryWithFailedTask(2, 1, true, this.getClass().getName(), true);
        app.submit(conf);
        Job job = app.getContext().getAllJobs().values().iterator().next();
        JobId jobId = job.getID();
        app.waitForState(job, JobState.FAILED);
        // make sure all events are flushed
        app.waitForState(Service.STATE.STOPPED);
        JobHistory jobHistory = new JobHistory();
        jobHistory.init(conf);
        HistoryFileInfo fileInfo = jobHistory.getJobFileInfo(jobId);
        JobHistoryParser parser;
        JobInfo jobInfo;
        synchronized (fileInfo) {
            Path historyFilePath = fileInfo.getHistoryFile();
            FSDataInputStream in = null;
            FileContext fc = null;
            try {
                fc = FileContext.getFileContext(conf);
                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"));
            }
            parser = new JobHistoryParser(in);
            jobInfo = parser.parse();
        }
        Exception parseException = parser.getParseException();
        Assert.assertNull("Caught an expected exception " + parseException, parseException);
        for (Map.Entry<TaskID, TaskInfo> entry : jobInfo.getAllTasks().entrySet()) {
            TaskId yarnTaskID = TypeConverter.toYarn(entry.getKey());
            CompletedTask ct = new CompletedTask(yarnTaskID, entry.getValue());
            Assert.assertNotNull("completed task report has null counters", ct.getReport().getCounters());
        }
        final List<String> originalDiagnostics = job.getDiagnostics();
        final String historyError = jobInfo.getErrorInfo();
        assertTrue("No original diagnostics for a failed job", originalDiagnostics != null && !originalDiagnostics.isEmpty());
        assertNotNull("No history error info for a failed job ", historyError);
        for (String diagString : originalDiagnostics) {
            assertTrue(historyError.contains(diagString));
        }
    } finally {
        LOG.info("FINISHED testCountersForFailedTask");
    }
}

Example 10 with JobInfo

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

the class TestJobHistoryParsing method testJobInfo.

@Test(timeout = 50000)
public void testJobInfo() throws Exception {
    JobInfo info = new JobInfo();
    Assert.assertEquals("NORMAL", info.getPriority());
    info.printAll();
}