Package org.apache.hadoop.tools.rumen

Examples of org.apache.hadoop.tools.rumen.JobStory


  private void testRandomLocation(int locations, int njobs, UserGroupInformation ugi) throws Exception {
    Configuration conf = new Configuration();
    conf.setInt(JobCreator.SLEEPJOB_RANDOM_LOCATIONS, locations);
    DebugJobProducer jobProducer = new DebugJobProducer(njobs, conf);
    JobConf jconf = GridmixTestUtils.mrCluster.createJobConf(new JobConf(conf));
    JobStory story;
    int seq=1;
    while ((story = jobProducer.getNextJob()) != null) {
      GridmixJob gridmixJob = JobCreator.SLEEPJOB.createGridmixJob(jconf, 0,
          story, new Path("ignored"), ugi, seq++);
      gridmixJob.buildSplits(null);
View Full Code Here


    Configuration conf = new Configuration();
    conf.setBoolean(SleepJob.SLEEPJOB_MAPTASK_ONLY, true);
    DebugJobProducer jobProducer = new DebugJobProducer(5, conf);
    JobConf jconf = GridmixTestUtils.mrCluster.createJobConf(new JobConf(conf));
    UserGroupInformation ugi = UserGroupInformation.getLoginUser();
    JobStory story;
    int seq = 1;
    while ((story = jobProducer.getNextJob()) != null) {
      GridmixJob gridmixJob = JobCreator.SLEEPJOB.createGridmixJob(jconf, 0,
          story, new Path("ignored"), ugi, seq++);
      gridmixJob.buildSplits(null);
View Full Code Here

                       job.getConfiguration().get("mapred.job.queue.name"),
                       sub.get(job.getConfiguration().get(GridmixJob.ORIGNAME))
                          .getQueueName());
        }

        final JobStory spec =
          sub.get(job.getConfiguration().get(GridmixJob.ORIGNAME));
        assertNotNull("No spec for " + job.getJobName(), spec);
        assertNotNull("No counters for " + job.getJobName(), job.getCounters());
        final String specname = spec.getName();
        final FileStatus stat =
          GridmixTestUtils.dfs.getFileStatus(
            new Path(GridmixTestUtils.DEST,
            "" + Integer.valueOf(specname.substring(specname.length() - 5))));
        assertEquals("Wrong owner for " + job.getJobName(), spec.getUser(),
                     stat.getOwner());

        final int nMaps = spec.getNumberMaps();
        final int nReds = spec.getNumberReduces();

        // TODO Blocked by MAPREDUCE-118
        if (true) return;
        // TODO
        System.out.println(jobname + ": " + nMaps + "/" + nReds);
View Full Code Here

   * @throws IOException
   */
  private int buildDistCacheFilesList(JobStoryProducer jsp) throws IOException {
    // Read all the jobs from the trace file and build the list of unique
    // distributed cache files.
    JobStory jobStory;
    while ((jobStory = jsp.getNextJob()) != null) {
      if (jobStory.getOutcome() == Pre21JobHistoryConstants.Values.SUCCESS &&
         jobStory.getSubmissionTime() >= 0) {
        updateHDFSDistCacheFilesList(jobStory);
      }
    }
    jsp.close();

View Full Code Here

              if (LOG.isDebugEnabled()) {
                LOG.debug("[STRESS] Cluster underloaded in run! Stressing...");
              }
              try {
                //TODO This in-line read can block submission for large jobs.
                final JobStory job = getNextJobFiltered();
                if (null == job) {
                  LOG.warn("[STRESS] Finished consuming the input trace. "
                           + "Exiting..");
                  return;
                }
                if (LOG.isDebugEnabled()) {
                  LOG.debug("Job Selected: " + job.getJobID());
                }
               
                UserGroupInformation ugi =
                  UserGroupInformation.createRemoteUser(job.getUser());
                UserGroupInformation tgtUgi = userResolver.getTargetUgi(ugi);
                GridmixJob tJob =
                  jobCreator.createGridmixJob(conf, 0L, job, scratch,
                               tgtUgi, sequence.getAndIncrement());
               
                // submit the job
                submitter.add(tJob);
               
                // TODO: We need to take care of scenario when one map/reduce
                // takes more than 1 slot.
               
                // Lock the loadjob as we are making updates
                int incompleteMapTasks = (int) calcEffectiveIncompleteMapTasks(
                                                 loadStatus.getMapCapacity(),
                                                 job.getNumberMaps(), 0.0f);
                loadStatus.decrementMapLoad(incompleteMapTasks);
               
                int incompleteReduceTasks =
                  (int) calcEffectiveIncompleteReduceTasks(
                          loadStatus.getReduceCapacity(),
                          job.getNumberReduces(), 0.0f);
                loadStatus.decrementReduceLoad(incompleteReduceTasks);
                 
                loadStatus.decrementJobLoad(1);
              } catch (IOException e) {
                LOG.error("[STRESS] Error while submitting the job ", e);
View Full Code Here

  protected abstract Thread createReaderThread() ;

  // gets the next job from the trace and does some bookkeeping for the same
  private JobStory getNextJobFromTrace() throws IOException {
    JobStory story = jobProducer.getNextJob();
    if (story != null) {
      ++numJobsInTrace;
    }
    return story;
  }
View Full Code Here

    }
    return story;
  }
 
  protected JobStory getNextJobFiltered() throws IOException {
    JobStory job = getNextJobFromTrace();
    // filter out the following jobs
    //    - unsuccessful jobs
    //    - jobs with missing submit-time
    //    - reduce only jobs
    // These jobs are not yet supported in Gridmix
    while (job != null &&
      (job.getOutcome() != Pre21JobHistoryConstants.Values.SUCCESS ||
        job.getSubmissionTime() < 0 || job.getNumberMaps() == 0)) {
      if (LOG.isDebugEnabled()) {
        List<String> reason = new ArrayList<String>();
        if (job.getOutcome() != Pre21JobHistoryConstants.Values.SUCCESS) {
          reason.add("STATE (" + job.getOutcome().name() + ")");
        }
        if (job.getSubmissionTime() < 0) {
          reason.add("SUBMISSION-TIME (" + job.getSubmissionTime() + ")");
        }
        if (job.getNumberMaps() == 0) {
          reason.add("ZERO-MAPS-JOB");
        }
       
        // TODO This should never happen. Probably we missed something!
        if (reason.size() == 0) {
          reason.add("N/A");
        }
       
        LOG.debug("Ignoring job " + job.getJobID() + " from the input trace."
                  + " Reason: " + StringUtils.join(reason, ","));
      }
      job = getNextJobFromTrace();
    }
    return null == job ? null : new FilterJobStory(job) {
View Full Code Here

  @Override
  void buildSplits(FilePool inputDir) throws IOException {
    long mapInputBytesTotal = 0L;
    long mapOutputBytesTotal = 0L;
    long mapOutputRecordsTotal = 0L;
    final JobStory jobdesc = getJobDesc();
    if (null == jobdesc) {
      return;
    }
    final int maps = jobdesc.getNumberMaps();
    final int reds = jobdesc.getNumberReduces();
    for (int i = 0; i < maps; ++i) {
      final TaskInfo info = jobdesc.getTaskInfo(TaskType.MAP, i);
      mapInputBytesTotal += info.getInputBytes();
      mapOutputBytesTotal += info.getOutputBytes();
      mapOutputRecordsTotal += info.getOutputRecords();
    }
    final double[] reduceRecordRatio = new double[reds];
    final double[] reduceByteRatio = new double[reds];
    for (int i = 0; i < reds; ++i) {
      final TaskInfo info = jobdesc.getTaskInfo(TaskType.REDUCE, i);
      reduceByteRatio[i] = info.getInputBytes() / (1.0 * mapOutputBytesTotal);
      reduceRecordRatio[i] =
        info.getInputRecords() / (1.0 * mapOutputRecordsTotal);
    }
    final InputStriper striper = new InputStriper(inputDir, mapInputBytesTotal);
    final List<InputSplit> splits = new ArrayList<InputSplit>();
    for (int i = 0; i < maps; ++i) {
      final int nSpec = reds / maps + ((reds % maps) > i ? 1 : 0);
      final long[] specBytes = new long[nSpec];
      final long[] specRecords = new long[nSpec];
      final ResourceUsageMetrics[] metrics = new ResourceUsageMetrics[nSpec];
      for (int j = 0; j < nSpec; ++j) {
        final TaskInfo info =
          jobdesc.getTaskInfo(TaskType.REDUCE, i + j * maps);
        specBytes[j] = info.getOutputBytes();
        specRecords[j] = info.getOutputRecords();
        metrics[j] = info.getResourceUsageMetrics();
        if (LOG.isDebugEnabled()) {
          LOG.debug(String.format("SPEC(%d) %d -> %d %d %d %d %d %d %d", id(), i,
                    i + j * maps, info.getOutputRecords(),
                    info.getOutputBytes(),
                    info.getResourceUsageMetrics().getCumulativeCpuUsage(),
                    info.getResourceUsageMetrics().getPhysicalMemoryUsage(),
                    info.getResourceUsageMetrics().getVirtualMemoryUsage(),
                    info.getResourceUsageMetrics().getHeapUsage()));
        }
      }
      final TaskInfo info = jobdesc.getTaskInfo(TaskType.MAP, i);
      long possiblyCompressedInputBytes = info.getInputBytes();
      Configuration conf = job.getConfiguration();
      long uncompressedInputBytes =
          CompressionEmulationUtil.getUncompressedInputBytes(
          possiblyCompressedInputBytes, conf);
View Full Code Here

          return;
        }
        LOG.info("START SERIAL @ " + System.currentTimeMillis());
        GridmixJob prevJob;
        while (!Thread.currentThread().isInterrupted()) {
          final JobStory job;
          try {
            job = getNextJobFiltered();
            if (null == job) {
              return;
            }
            if (LOG.isDebugEnabled()) {
              LOG.debug(
                "Serial mode submitting job " + job.getName());
            }
            prevJob = jobCreator.createGridmixJob(
              conf, 0L, job, scratch,
              userResolver.getTargetUgi(
                UserGroupInformation.createRemoteUser(job.getUser())),
              sequence.getAndIncrement());

            lock.lock();
            try {
              LOG.info(" Submitted the job " + prevJob);
              submitter.add(prevJob);
            } finally {
              lock.unlock();
            }
          } catch (IOException e) {
            error = e;
            //If submission of current job fails , try to submit the next job.
            return;
          }

          if (prevJob != null) {
            //Wait till previous job submitted is completed.
            lock.lock();
            try {
              while (true) {
                try {
                  jobCompleted.await();
                } catch (InterruptedException ie) {
                  LOG.error(
                    " Error in SerialJobFactory while waiting for job completion ",
                    ie);
                  return;
                }
                if (LOG.isDebugEnabled()) {
                  LOG.info(" job " + job.getName() + " completed ");
                }
                break;
              }
            } finally {
              lock.unlock();
View Full Code Here

    assertTrue("Incorrect job status", status == stats.getJobStatus());
  }
 
  private static JobStory getCustomJobStory(final int numMaps,
                                            final int numReds) {
    return new JobStory() {
      @Override
      public InputSplit[] getInputSplits() {
        return null;
      }
      @Override
View Full Code Here

TOP

Related Classes of org.apache.hadoop.tools.rumen.JobStory

Copyright © 2018 www.massapicom. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.