Examples with JobHistory org.apache.hadoop.mapreduce.v2.hs.JobHistory used on opensource projects

Example 6 with JobHistory

use of org.apache.hadoop.mapreduce.v2.hs.JobHistory in project hadoop by apache.

the class TestRecovery method testCrashed.

/**
   * AM with 2 maps and 1 reduce. For 1st map, one attempt fails, one attempt
   * completely disappears because of failed launch, one attempt gets killed and
   * one attempt succeeds. AM crashes after the first tasks finishes and
   * recovers completely and succeeds in the second generation.
   * 
   * @throws Exception
   */
@Test
public void testCrashed() throws Exception {
    int runCount = 0;
    long am1StartTimeEst = System.currentTimeMillis();
    MRApp app = new MRAppWithHistory(2, 1, false, this.getClass().getName(), true, ++runCount);
    Configuration conf = new Configuration();
    conf.setBoolean("mapred.mapper.new-api", true);
    conf.setBoolean("mapred.reducer.new-api", true);
    conf.setBoolean(MRJobConfig.JOB_UBERTASK_ENABLE, false);
    conf.set(FileOutputFormat.OUTDIR, outputDir.toString());
    Job job = app.submit(conf);
    app.waitForState(job, JobState.RUNNING);
    long jobStartTime = job.getReport().getStartTime();
    //all maps would be running
    Assert.assertEquals("No of tasks not correct", 3, job.getTasks().size());
    Iterator<Task> it = job.getTasks().values().iterator();
    Task mapTask1 = it.next();
    Task mapTask2 = it.next();
    Task reduceTask = it.next();
    // all maps must be running
    app.waitForState(mapTask1, TaskState.RUNNING);
    app.waitForState(mapTask2, TaskState.RUNNING);
    TaskAttempt task1Attempt1 = mapTask1.getAttempts().values().iterator().next();
    TaskAttempt task2Attempt = mapTask2.getAttempts().values().iterator().next();
    //before sending the TA_DONE, event make sure attempt has come to 
    //RUNNING state
    app.waitForState(task1Attempt1, TaskAttemptState.RUNNING);
    app.waitForState(task2Attempt, TaskAttemptState.RUNNING);
    app.waitForState(reduceTask, TaskState.RUNNING);
    /////////// Play some games with the TaskAttempts of the first task //////
    //send the fail signal to the 1st map task attempt
    app.getContext().getEventHandler().handle(new TaskAttemptEvent(task1Attempt1.getID(), TaskAttemptEventType.TA_FAILMSG));
    app.waitForState(task1Attempt1, TaskAttemptState.FAILED);
    int timeOut = 0;
    while (mapTask1.getAttempts().size() != 2 && timeOut++ < 10) {
        Thread.sleep(2000);
        LOG.info("Waiting for next attempt to start");
    }
    Assert.assertEquals(2, mapTask1.getAttempts().size());
    Iterator<TaskAttempt> itr = mapTask1.getAttempts().values().iterator();
    itr.next();
    TaskAttempt task1Attempt2 = itr.next();
    // wait for the second task attempt to be assigned.
    waitForContainerAssignment(task1Attempt2);
    // This attempt will automatically fail because of the way ContainerLauncher
    // is setup
    // This attempt 'disappears' from JobHistory and so causes MAPREDUCE-3846
    app.getContext().getEventHandler().handle(new TaskAttemptEvent(task1Attempt2.getID(), TaskAttemptEventType.TA_CONTAINER_LAUNCH_FAILED));
    app.waitForState(task1Attempt2, TaskAttemptState.FAILED);
    timeOut = 0;
    while (mapTask1.getAttempts().size() != 3 && timeOut++ < 10) {
        Thread.sleep(2000);
        LOG.info("Waiting for next attempt to start");
    }
    Assert.assertEquals(3, mapTask1.getAttempts().size());
    itr = mapTask1.getAttempts().values().iterator();
    itr.next();
    itr.next();
    TaskAttempt task1Attempt3 = itr.next();
    app.waitForState(task1Attempt3, TaskAttemptState.RUNNING);
    //send the kill signal to the 1st map 3rd attempt
    app.getContext().getEventHandler().handle(new TaskAttemptEvent(task1Attempt3.getID(), TaskAttemptEventType.TA_KILL));
    app.waitForState(task1Attempt3, TaskAttemptState.KILLED);
    timeOut = 0;
    while (mapTask1.getAttempts().size() != 4 && timeOut++ < 10) {
        Thread.sleep(2000);
        LOG.info("Waiting for next attempt to start");
    }
    Assert.assertEquals(4, mapTask1.getAttempts().size());
    itr = mapTask1.getAttempts().values().iterator();
    itr.next();
    itr.next();
    itr.next();
    TaskAttempt task1Attempt4 = itr.next();
    app.waitForState(task1Attempt4, TaskAttemptState.RUNNING);
    //send the done signal to the 1st map 4th attempt
    app.getContext().getEventHandler().handle(new TaskAttemptEvent(task1Attempt4.getID(), TaskAttemptEventType.TA_DONE));
    /////////// End of games with the TaskAttempts of the first task //////
    //wait for first map task to complete
    app.waitForState(mapTask1, TaskState.SUCCEEDED);
    long task1StartTime = mapTask1.getReport().getStartTime();
    long task1FinishTime = mapTask1.getReport().getFinishTime();
    //stop the app
    app.stop();
    //rerun
    //in rerun the 1st map will be recovered from previous run
    long am2StartTimeEst = System.currentTimeMillis();
    app = new MRAppWithHistory(2, 1, false, this.getClass().getName(), false, ++runCount);
    conf = new Configuration();
    conf.setBoolean(MRJobConfig.MR_AM_JOB_RECOVERY_ENABLE, true);
    conf.setBoolean("mapred.mapper.new-api", true);
    conf.setBoolean("mapred.reducer.new-api", true);
    conf.set(FileOutputFormat.OUTDIR, outputDir.toString());
    conf.setBoolean(MRJobConfig.JOB_UBERTASK_ENABLE, false);
    job = app.submit(conf);
    app.waitForState(job, JobState.RUNNING);
    //all maps would be running
    Assert.assertEquals("No of tasks not correct", 3, job.getTasks().size());
    it = job.getTasks().values().iterator();
    mapTask1 = it.next();
    mapTask2 = it.next();
    reduceTask = it.next();
    // first map will be recovered, no need to send done
    app.waitForState(mapTask1, TaskState.SUCCEEDED);
    app.waitForState(mapTask2, TaskState.RUNNING);
    task2Attempt = mapTask2.getAttempts().values().iterator().next();
    //before sending the TA_DONE, event make sure attempt has come to 
    //RUNNING state
    app.waitForState(task2Attempt, TaskAttemptState.RUNNING);
    //send the done signal to the 2nd map task
    app.getContext().getEventHandler().handle(new TaskAttemptEvent(mapTask2.getAttempts().values().iterator().next().getID(), TaskAttemptEventType.TA_DONE));
    //wait to get it completed
    app.waitForState(mapTask2, TaskState.SUCCEEDED);
    //wait for reduce to be running before sending done
    app.waitForState(reduceTask, TaskState.RUNNING);
    //send the done signal to the reduce
    app.getContext().getEventHandler().handle(new TaskAttemptEvent(reduceTask.getAttempts().values().iterator().next().getID(), TaskAttemptEventType.TA_DONE));
    app.waitForState(job, JobState.SUCCEEDED);
    app.verifyCompleted();
    Assert.assertEquals("Job Start time not correct", jobStartTime, job.getReport().getStartTime());
    Assert.assertEquals("Task Start time not correct", task1StartTime, mapTask1.getReport().getStartTime());
    Assert.assertEquals("Task Finish time not correct", task1FinishTime, mapTask1.getReport().getFinishTime());
    Assert.assertEquals(2, job.getAMInfos().size());
    int attemptNum = 1;
    // Verify AMInfo
    for (AMInfo amInfo : job.getAMInfos()) {
        Assert.assertEquals(attemptNum++, amInfo.getAppAttemptId().getAttemptId());
        Assert.assertEquals(amInfo.getAppAttemptId(), amInfo.getContainerId().getApplicationAttemptId());
        Assert.assertEquals(MRApp.NM_HOST, amInfo.getNodeManagerHost());
        Assert.assertEquals(MRApp.NM_PORT, amInfo.getNodeManagerPort());
        Assert.assertEquals(MRApp.NM_HTTP_PORT, amInfo.getNodeManagerHttpPort());
    }
    long am1StartTimeReal = job.getAMInfos().get(0).getStartTime();
    long am2StartTimeReal = job.getAMInfos().get(1).getStartTime();
    Assert.assertTrue(am1StartTimeReal >= am1StartTimeEst && am1StartTimeReal <= am2StartTimeEst);
    Assert.assertTrue(am2StartTimeReal >= am2StartTimeEst && am2StartTimeReal <= System.currentTimeMillis());
// TODO Add verification of additional data from jobHistory - whatever was
// available in the failed attempt should be available here
}

Example 7 with JobHistory

use of org.apache.hadoop.mapreduce.v2.hs.JobHistory in project hadoop by apache.

the class ConfBlock method render.

/*
   * (non-Javadoc)
   * @see org.apache.hadoop.yarn.webapp.view.HtmlBlock#render(org.apache.hadoop.yarn.webapp.view.HtmlBlock.Block)
   */
@Override
protected void render(Block html) {
    String jid = $(JOB_ID);
    if (jid.isEmpty()) {
        html.p()._("Sorry, can't do anything without a JobID.")._();
        return;
    }
    JobId jobID = MRApps.toJobID(jid);
    Job job = appContext.getJob(jobID);
    if (job == null) {
        html.p()._("Sorry, ", jid, " not found.")._();
        return;
    }
    Path confPath = job.getConfFile();
    try {
        ConfInfo info = new ConfInfo(job);
        html.div().a("/jobhistory/downloadconf/" + jid, confPath.toString());
        TBODY<TABLE<Hamlet>> tbody = html.table("#conf").thead().tr().th(_TH, "key").th(_TH, "value").th(_TH, "source chain")._()._().tbody();
        for (ConfEntryInfo entry : info.getProperties()) {
            StringBuffer buffer = new StringBuffer();
            String[] sources = entry.getSource();
            //Skip the last entry, because it is always the same HDFS file, and
            // output them in reverse order so most recent is output first
            boolean first = true;
            for (int i = (sources.length - 2); i >= 0; i--) {
                if (!first) {
                    buffer.append(" <- ");
                }
                first = false;
                buffer.append(sources[i]);
            }
            tbody.tr().td(entry.getName()).td(entry.getValue()).td(buffer.toString())._();
        }
        tbody._().tfoot().tr().th().input("search_init").$type(InputType.text).$name("key").$value("key")._()._().th().input("search_init").$type(InputType.text).$name("value").$value("value")._()._().th().input("search_init").$type(InputType.text).$name("source chain").$value("source chain")._()._()._()._()._();
    } catch (IOException e) {
        LOG.error("Error while reading " + confPath, e);
        html.p()._("Sorry got an error while reading conf file. ", confPath);
    }
}

Example 8 with JobHistory

use of org.apache.hadoop.mapreduce.v2.hs.JobHistory in project hadoop by apache.

the class TestFileNameIndexUtils method testJobNameWithMultibyteChars.

/**
   * Verify the name of jobhistory file is not greater than 255 bytes
   * even if there are some multibyte characters in the job name.
   */
@Test
public void testJobNameWithMultibyteChars() throws IOException {
    JobIndexInfo info = new JobIndexInfo();
    JobID oldJobId = JobID.forName(JOB_ID);
    JobId jobId = TypeConverter.toYarn(oldJobId);
    info.setJobId(jobId);
    info.setSubmitTime(Long.parseLong(SUBMIT_TIME));
    info.setUser(USER_NAME);
    StringBuilder sb = new StringBuilder();
    info.setFinishTime(Long.parseLong(FINISH_TIME));
    info.setNumMaps(Integer.parseInt(NUM_MAPS));
    info.setNumReduces(Integer.parseInt(NUM_REDUCES));
    info.setJobStatus(JOB_STATUS);
    info.setQueueName(QUEUE_NAME);
    info.setJobStartTime(Long.parseLong(JOB_START_TIME));
    // which is encoded into 1 x 3 = 3 characters by URL encode.
    for (int i = 0; i < 100; i++) {
        sb.append('%');
    }
    String longJobName = sb.toString();
    info.setJobName(longJobName);
    String jobHistoryFile = FileNameIndexUtils.getDoneFileName(info, 50);
    Assert.assertTrue(jobHistoryFile.length() <= 255);
    String trimedJobName = jobHistoryFile.split(FileNameIndexUtils.DELIMITER)[// 3 is index of job name
    3];
    // 3 x 16 < 50 < 3 x 17 so the length of trimedJobName should be 48
    Assert.assertEquals(48, trimedJobName.getBytes(UTF_8).length);
    // validate whether trimmedJobName by testing reversibility
    byte[] trimedJobNameInByte = trimedJobName.getBytes(UTF_8);
    String reEncodedTrimedJobName = new String(trimedJobNameInByte, UTF_8);
    Assert.assertArrayEquals(trimedJobNameInByte, reEncodedTrimedJobName.getBytes(UTF_8));
    sb.setLength(0);
    // which is encoded into 2 x 3 = 6 characters by URL encode.
    for (int i = 0; i < 100; i++) {
        // large omega
        sb.append('Ω');
    }
    longJobName = sb.toString();
    info.setJobName(longJobName);
    jobHistoryFile = FileNameIndexUtils.getDoneFileName(info, 27);
    Assert.assertTrue(jobHistoryFile.length() <= 255);
    trimedJobName = jobHistoryFile.split(FileNameIndexUtils.DELIMITER)[// 3 is index of job name
    3];
    // 6 x 4 < 27 < 6 x 5 so the length of trimedJobName should be 24
    Assert.assertEquals(24, trimedJobName.getBytes(UTF_8).length);
    // validate whether trimmedJobName by testing reversibility
    trimedJobNameInByte = trimedJobName.getBytes(UTF_8);
    reEncodedTrimedJobName = new String(trimedJobNameInByte, UTF_8);
    Assert.assertArrayEquals(trimedJobNameInByte, reEncodedTrimedJobName.getBytes(UTF_8));
    sb.setLength(0);
    // which is encoded into 3 x 3 = 9 characters by URL encode.
    for (int i = 0; i < 100; i++) {
        // rightwards arrow
        sb.append('→');
    }
    longJobName = sb.toString();
    info.setJobName(longJobName);
    jobHistoryFile = FileNameIndexUtils.getDoneFileName(info, 40);
    Assert.assertTrue(jobHistoryFile.length() <= 255);
    trimedJobName = jobHistoryFile.split(FileNameIndexUtils.DELIMITER)[// 3 is index of job name
    3];
    // 9 x 4 < 40 < 9 x 5 so the length of trimedJobName should be 36
    Assert.assertEquals(36, trimedJobName.getBytes(UTF_8).length);
    // validate whether trimmedJobName by testing reversibility
    trimedJobNameInByte = trimedJobName.getBytes(UTF_8);
    reEncodedTrimedJobName = new String(trimedJobNameInByte, UTF_8);
    Assert.assertArrayEquals(trimedJobNameInByte, reEncodedTrimedJobName.getBytes(UTF_8));
    sb.setLength(0);
    // which is encoded into 4 x 3 = 12 characters by URL encode.
    for (int i = 0; i < 100; i++) {
        // Mugil cephalus in Kanji.
        sb.append("𩸽");
    }
    longJobName = sb.toString();
    info.setJobName(longJobName);
    jobHistoryFile = FileNameIndexUtils.getDoneFileName(info, 49);
    Assert.assertTrue(jobHistoryFile.length() <= 255);
    trimedJobName = jobHistoryFile.split(FileNameIndexUtils.DELIMITER)[// 3 is index of job name
    3];
    // 12 x 4 < 49 < 12 x 5 so the length of trimedJobName should be 48
    Assert.assertEquals(48, trimedJobName.getBytes(UTF_8).length);
    // validate whether trimmedJobName by testing reversibility
    trimedJobNameInByte = trimedJobName.getBytes(UTF_8);
    reEncodedTrimedJobName = new String(trimedJobNameInByte, UTF_8);
    Assert.assertArrayEquals(trimedJobNameInByte, reEncodedTrimedJobName.getBytes(UTF_8));
    sb.setLength(0);
    // Test for the combination of 1 to 4 bytes UTF-8 characters
    // cat in Kanji (encoded into 3 bytes x 3 characters)
    sb.append('猫').append(// (encoded into 1 byte x 3 characters)
    "[").append(// small lambda (encoded into 2 bytes x 3 characters)
    'λ').append(// (encoded into 1 byte x 3 characters)
    '/').append(// not url-encoded (1 byte x 1 character)
    'A').append(// flying fish in
    "𩹉").append(// dog in Kanji (encoded into 3 bytes x 3 characters)
    '犬');
    longJobName = sb.toString();
    info.setJobName(longJobName);
    jobHistoryFile = FileNameIndexUtils.getDoneFileName(info, 23);
    Assert.assertTrue(jobHistoryFile.length() <= 255);
    trimedJobName = jobHistoryFile.split(FileNameIndexUtils.DELIMITER)[// 3 is index of job name
    3];
    // total size of the first 5 characters = 22
    // 23 < total size of the first 6 characters
    Assert.assertEquals(22, trimedJobName.getBytes(UTF_8).length);
    // validate whether trimmedJobName by testing reversibility
    trimedJobNameInByte = trimedJobName.getBytes(UTF_8);
    reEncodedTrimedJobName = new String(trimedJobNameInByte, UTF_8);
    Assert.assertArrayEquals(trimedJobNameInByte, reEncodedTrimedJobName.getBytes(UTF_8));
}

Example 9 with JobHistory

use of org.apache.hadoop.mapreduce.v2.hs.JobHistory in project hadoop by apache.

the class HistoryClientService method initializeWebApp.

@VisibleForTesting
protected void initializeWebApp(Configuration conf) {
    webApp = new HsWebApp(history);
    InetSocketAddress bindAddress = MRWebAppUtil.getJHSWebBindAddress(conf);
    // NOTE: there should be a .at(InetSocketAddress)
    WebApps.$for("jobhistory", HistoryClientService.class, this, "ws").with(conf).withHttpSpnegoKeytabKey(JHAdminConfig.MR_WEBAPP_SPNEGO_KEYTAB_FILE_KEY).withHttpSpnegoPrincipalKey(JHAdminConfig.MR_WEBAPP_SPNEGO_USER_NAME_KEY).withCSRFProtection(JHAdminConfig.MR_HISTORY_CSRF_PREFIX).withXFSProtection(JHAdminConfig.MR_HISTORY_XFS_PREFIX).at(NetUtils.getHostPortString(bindAddress)).start(webApp);
    String connectHost = MRWebAppUtil.getJHSWebappURLWithoutScheme(conf).split(":")[0];
    MRWebAppUtil.setJHSWebappURLWithoutScheme(conf, connectHost + ":" + webApp.getListenerAddress().getPort());
}

Example 10 with JobHistory

use of org.apache.hadoop.mapreduce.v2.hs.JobHistory in project hadoop by apache.

the class JobHistoryServer method serviceInit.

@Override
protected void serviceInit(Configuration conf) throws Exception {
    Configuration config = new YarnConfiguration(conf);
    config.setBoolean(Dispatcher.DISPATCHER_EXIT_ON_ERROR_KEY, true);
    // This is required for WebApps to use https if enabled.
    MRWebAppUtil.initialize(getConfig());
    try {
        doSecureLogin(conf);
    } catch (IOException ie) {
        throw new YarnRuntimeException("History Server Failed to login", ie);
    }
    jobHistoryService = new JobHistory();
    historyContext = (HistoryContext) jobHistoryService;
    stateStore = createStateStore(conf);
    this.jhsDTSecretManager = createJHSSecretManager(conf, stateStore);
    clientService = createHistoryClientService();
    aggLogDelService = new AggregatedLogDeletionService();
    hsAdminServer = new HSAdminServer(aggLogDelService, jobHistoryService);
    addService(stateStore);
    addService(new HistoryServerSecretManagerService());
    addService(jobHistoryService);
    addService(clientService);
    addService(aggLogDelService);
    addService(hsAdminServer);
    DefaultMetricsSystem.initialize("JobHistoryServer");
    JvmMetrics jm = JvmMetrics.initSingleton("JobHistoryServer", null);
    pauseMonitor = new JvmPauseMonitor();
    addService(pauseMonitor);
    jm.setPauseMonitor(pauseMonitor);
    super.serviceInit(config);
}