org.apache.hadoop.mapred.TaskLog Java Examples

public MRAppMaster(ApplicationAttemptId applicationAttemptId,
    ContainerId containerId, String nmHost, int nmPort, int nmHttpPort,
    Clock clock, long appSubmitTime) {
  super(MRAppMaster.class.getName());
  this.clock = clock;
  this.startTime = clock.getTime();
  this.appSubmitTime = appSubmitTime;
  this.appAttemptID = applicationAttemptId;
  this.containerID = containerId;
  this.nmHost = nmHost;
  this.nmPort = nmPort;
  this.nmHttpPort = nmHttpPort;
  this.metrics = MRAppMetrics.create();
  logSyncer = TaskLog.createLogSyncer();
  LOG.info("Created MRAppMaster for application " + applicationAttemptId);
}
 

public MRAppMaster(ApplicationAttemptId applicationAttemptId,
    ContainerId containerId, String nmHost, int nmPort, int nmHttpPort,
    Clock clock, long appSubmitTime) {
  super(MRAppMaster.class.getName());
  this.clock = clock;
  this.startTime = clock.getTime();
  this.appSubmitTime = appSubmitTime;
  this.appAttemptID = applicationAttemptId;
  this.containerID = containerId;
  this.nmHost = nmHost;
  this.nmPort = nmPort;
  this.nmHttpPort = nmHttpPort;
  this.metrics = MRAppMetrics.create();
  logSyncer = TaskLog.createLogSyncer();
  LOG.info("Created MRAppMaster for application " + applicationAttemptId);
}
 

/**
 * @param lInfo
 * @param taskLogFileDetails
 * @param updatedTaskLogFileDetails
 * @param logName
 */
private void revertIndexFileInfo(PerJVMInfo lInfo,
    Map<Task, Map<LogName, LogFileDetail>> taskLogFileDetails,
    Map<Task, Map<LogName, LogFileDetail>> updatedTaskLogFileDetails,
    LogName logName) {
  if (TaskLog.LOGS_TRACKED_BY_INDEX_FILES.contains(logName)) {
    for (Task task : lInfo.allAttempts) {
      if (!updatedTaskLogFileDetails.containsKey(task)) {
        updatedTaskLogFileDetails.put(task,
            new HashMap<LogName, LogFileDetail>());
      }
      updatedTaskLogFileDetails.get(task).put(logName,
          taskLogFileDetails.get(task).get(logName));
    }
  }
}
 

/**
 * clean previous std error and outs
 */

private void initStdOut(JobConf configuration) {
  TaskAttemptID taskId = TaskAttemptID.forName(configuration
          .get(MRJobConfig.TASK_ATTEMPT_ID));
  File stdOut = TaskLog.getTaskLogFile(taskId, false, TaskLog.LogName.STDOUT);
  File stdErr = TaskLog.getTaskLogFile(taskId, false, TaskLog.LogName.STDERR);
  // prepare folder
  if (!stdOut.getParentFile().exists()) {
    stdOut.getParentFile().mkdirs();
  } else { // clean logs
    stdOut.deleteOnExit();
    stdErr.deleteOnExit();
  }
}
 

private String readStdOut(JobConf conf) throws Exception {
  TaskAttemptID taskId = TaskAttemptID.forName(conf
          .get(MRJobConfig.TASK_ATTEMPT_ID));
  File stdOut = TaskLog.getTaskLogFile(taskId, false, TaskLog.LogName.STDOUT);

  return readFile(stdOut);

}
 

/**
 * Reads tasklog and returns it as string after trimming it.
 * 
 * @param filter
 *          Task log filter; can be STDOUT, STDERR, SYSLOG, DEBUGOUT, PROFILE
 * @param taskId
 *          The task id for which the log has to collected
 * @param isCleanup
 *          whether the task is a cleanup attempt or not.
 * @return task log as string
 * @throws IOException
 */
public static String readTaskLog(TaskLog.LogName filter,
    org.apache.hadoop.mapred.TaskAttemptID taskId, boolean isCleanup)
    throws IOException {
  // string buffer to store task log
  StringBuffer result = new StringBuffer();
  int res;

  // reads the whole tasklog into inputstream
  InputStream taskLogReader = new TaskLog.Reader(taskId, filter, 0, -1,
      isCleanup);
  // construct string log from inputstream.
  byte[] b = new byte[65536];
  while (true) {
    res = taskLogReader.read(b);
    if (res > 0) {
      result.append(new String(b));
    } else {
      break;
    }
  }
  taskLogReader.close();

  // trim the string and return it
  String str = result.toString();
  str = str.trim();
  return str;
}
 

/**
 * Truncation of logs is done. Now sync the index files to reflect the
 * truncated sizes.
 * 
 * @param firstAttempt
 * @param updatedTaskLogFileDetails
 */
private void updateIndicesAfterLogTruncation(TaskAttemptID firstAttempt,
    Map<Task, Map<LogName, LogFileDetail>> updatedTaskLogFileDetails) {
  for (Entry<Task, Map<LogName, LogFileDetail>> entry : 
                              updatedTaskLogFileDetails.entrySet()) {
    Task task = entry.getKey();
    Map<LogName, LogFileDetail> logFileDetails = entry.getValue();
    Map<LogName, Long[]> logLengths = new HashMap<LogName, Long[]>();
    // set current and previous lengths
    for (LogName logName : TaskLog.LOGS_TRACKED_BY_INDEX_FILES) {
      logLengths.put(logName, new Long[] { Long.valueOf(0L),
          Long.valueOf(0L) });
      LogFileDetail lfd = logFileDetails.get(logName);
      if (lfd != null) {
        // Set previous lengths
        logLengths.get(logName)[0] = Long.valueOf(lfd.start);
        // Set current lengths
        logLengths.get(logName)[1] = Long.valueOf(lfd.start + lfd.length);
      }
    }
    try {
      TaskLog.writeToIndexFile(firstAttempt, task.getTaskID(),
          task.isTaskCleanupTask(), logLengths);
    } catch (IOException ioe) {
      LOG.warn("Exception in updateIndicesAfterLogTruncation : "
          + StringUtils.stringifyException(ioe));
      LOG.warn("Exception encountered while updating index file of task "
          + task.getTaskID()
          + ". Ignoring and continuing with other tasks.");
    }
  }
}
 

void validateTaskStderr(StreamJob job, TaskType type)
    throws IOException {
  TaskAttemptID attemptId =
      new TaskAttemptID(new TaskID(job.jobId_, type, 0), 0);

  String log = MapReduceTestUtil.readTaskLog(TaskLog.LogName.STDERR,
      attemptId, false);

  // trim() is called on expectedStderr here because the method
  // MapReduceTestUtil.readTaskLog() returns trimmed String.
  assertTrue(log.equals(expectedStderr.trim()));
}
 

/**
 * Get the logFileDetails of all the list of attempts passed.
 * 
 * @param lInfo
 * @return a map of task to the log-file detail
 * @throws IOException
 */
private Map<Task, Map<LogName, LogFileDetail>> getAllLogsFileDetails(
    final List<Task> allAttempts) throws IOException {
  Map<Task, Map<LogName, LogFileDetail>> taskLogFileDetails =
      new HashMap<Task, Map<LogName, LogFileDetail>>();
  for (Task task : allAttempts) {
    Map<LogName, LogFileDetail> allLogsFileDetails;
    allLogsFileDetails =
        TaskLog.getAllLogsFileDetails(task.getTaskID(),
            task.isTaskCleanupTask());
    taskLogFileDetails.put(task, allLogsFileDetails);
  }
  return taskLogFileDetails;
}
 

/**
 * clean previous std error and outs
 */

private void initStdOut(JobConf configuration) {
  TaskAttemptID taskId = TaskAttemptID.forName(configuration
          .get(MRJobConfig.TASK_ATTEMPT_ID));
  File stdOut = TaskLog.getTaskLogFile(taskId, false, TaskLog.LogName.STDOUT);
  File stdErr = TaskLog.getTaskLogFile(taskId, false, TaskLog.LogName.STDERR);
  // prepare folder
  if (!stdOut.getParentFile().exists()) {
    stdOut.getParentFile().mkdirs();
  } else { // clean logs
    stdOut.deleteOnExit();
    stdErr.deleteOnExit();
  }
}
 

private String readStdOut(JobConf conf) throws Exception {
  TaskAttemptID taskId = TaskAttemptID.forName(conf
          .get(MRJobConfig.TASK_ATTEMPT_ID));
  File stdOut = TaskLog.getTaskLogFile(taskId, false, TaskLog.LogName.STDOUT);

  return readFile(stdOut);

}
 

/**
 * Reads tasklog and returns it as string after trimming it.
 * 
 * @param filter
 *          Task log filter; can be STDOUT, STDERR, SYSLOG, DEBUGOUT, PROFILE
 * @param taskId
 *          The task id for which the log has to collected
 * @param isCleanup
 *          whether the task is a cleanup attempt or not.
 * @return task log as string
 * @throws IOException
 */
public static String readTaskLog(TaskLog.LogName filter,
    org.apache.hadoop.mapred.TaskAttemptID taskId, boolean isCleanup)
    throws IOException {
  // string buffer to store task log
  StringBuffer result = new StringBuffer();
  int res;

  // reads the whole tasklog into inputstream
  InputStream taskLogReader = new TaskLog.Reader(taskId, filter, 0, -1,
      isCleanup);
  // construct string log from inputstream.
  byte[] b = new byte[65536];
  while (true) {
    res = taskLogReader.read(b);
    if (res > 0) {
      result.append(new String(b));
    } else {
      break;
    }
  }
  taskLogReader.close();

  // trim the string and return it
  String str = result.toString();
  str = str.trim();
  return str;
}
 

void validateTaskStderr(StreamJob job, TaskType type)
    throws IOException {
  TaskAttemptID attemptId =
      new TaskAttemptID(new TaskID(job.jobId_, type, 0), 0);

  String log = MapReduceTestUtil.readTaskLog(TaskLog.LogName.STDERR,
      attemptId, false);

  // trim() is called on expectedStderr here because the method
  // MapReduceTestUtil.readTaskLog() returns trimmed String.
  assertTrue(log.equals(expectedStderr.trim()));
}
 

/**
 * Start the child process to handle the task for us. 
 * @param conf the task's configuration
 * @param recordReader the fake record reader to update progress with
 * @param output the collector to send output to
 * @param reporter the reporter for the task
 * @param outputKeyClass the class of the output keys
 * @param outputValueClass the class of the output values
 * @param runOnGPU 
 * @throws IOException
 * @throws InterruptedException
 */
Application(JobConf conf, 
            RecordReader<FloatWritable, NullWritable> recordReader, 
            OutputCollector<K2,V2> output, Reporter reporter,
            Class<? extends K2> outputKeyClass,
            Class<? extends V2> outputValueClass,
            boolean runOnGPU
            ) throws IOException, InterruptedException {
  serverSocket = new ServerSocket(0);
  Map<String, String> env = new HashMap<String,String>();
  // add TMPDIR environment variable with the value of java.io.tmpdir
  env.put("TMPDIR", System.getProperty("java.io.tmpdir"));
  env.put("hadoop.pipes.command.port", 
          Integer.toString(serverSocket.getLocalPort()));
  List<String> cmd = new ArrayList<String>();
  String interpretor = conf.get("hadoop.pipes.executable.interpretor");
  if (interpretor != null) {
    cmd.add(interpretor);
  }

  // Check whether the applicaiton will run on GPU
  int i = runOnGPU ? 1 : 0;
  String executable = DistributedCache.getLocalCacheFiles(conf)[i].toString();
  FileUtil.chmod(executable, "a+x");
  cmd.add(executable);
  // wrap the command in a stdout/stderr capture
  TaskAttemptID taskid = TaskAttemptID.forName(conf.get("mapred.task.id"));
  File stdout = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDOUT);
  File stderr = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDERR);
  long logLength = TaskLog.getTaskLogLength(conf);
  cmd = TaskLog.captureOutAndError(cmd, stdout, stderr, logLength);

  process = runClient(cmd, env);
  clientSocket = serverSocket.accept();
  handler = new OutputHandler<K2, V2>(output, reporter, recordReader);
  K2 outputKey = (K2)
    ReflectionUtils.newInstance(outputKeyClass, conf);
  V2 outputValue = (V2) 
    ReflectionUtils.newInstance(outputValueClass, conf);
  downlink = new BinaryProtocol<K1, V1, K2, V2>(clientSocket, handler, 
                                outputKey, outputValue, conf);
  downlink.start();
  downlink.setJobConf(conf);
}
 

/**
 * Start the child process to handle the task for us.
 * @param conf the task's configuration
 * @param recordReader the fake record reader to update progress with
 * @param output the collector to send output to
 * @param reporter the reporter for the task
 * @param outputKeyClass the class of the output keys
 * @param outputValueClass the class of the output values
 * @throws IOException
 * @throws InterruptedException
 */
Application(JobConf conf, 
            RecordReader<FloatWritable, NullWritable> recordReader, 
            OutputCollector<K2,V2> output, Reporter reporter,
            Class<? extends K2> outputKeyClass,
            Class<? extends V2> outputValueClass
            ) throws IOException, InterruptedException {
  serverSocket = new ServerSocket(0);
  Map<String, String> env = new HashMap<String,String>();
  // add TMPDIR environment variable with the value of java.io.tmpdir
  env.put("TMPDIR", System.getProperty("java.io.tmpdir"));
  env.put("hadoop.pipes.command.port", 
          Integer.toString(serverSocket.getLocalPort()));
  List<String> cmd = new ArrayList<String>();
  String interpretor = conf.get("hadoop.pipes.executable.interpretor");
  if (interpretor != null) {
    cmd.add(interpretor);
  }

  String executable = DistributedCache.getLocalCacheFiles(conf)[0].toString();
  FileUtil.chmod(executable, "a+x");
  cmd.add(executable);
  // wrap the command in a stdout/stderr capture
  TaskAttemptID taskid = TaskAttemptID.forName(conf.get("mapred.task.id"));
  File stdout = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDOUT);
  File stderr = TaskLog.getTaskLogFile(taskid, TaskLog.LogName.STDERR);
  long logLength = TaskLog.getTaskLogLength(conf);
  cmd = TaskLog.captureOutAndError(null, cmd, stdout, stderr, logLength,
      false);

  process = runClient(cmd, env);
  clientSocket = serverSocket.accept();
  handler = new OutputHandler<K2, V2>(output, reporter, recordReader);
  K2 outputKey = (K2)
    ReflectionUtils.newInstance(outputKeyClass, conf);
  V2 outputValue = (V2) 
    ReflectionUtils.newInstance(outputValueClass, conf);
  downlink = new BinaryProtocol<K1, V1, K2, V2>(clientSocket, handler, 
                                outputKey, outputValue, conf);
  downlink.start();
  downlink.setJobConf(conf);
}
 

/**
 * Start the child process to handle the task for us.
 * @param conf the task's configuration
 * @param recordReader the fake record reader to update progress with
 * @param output the collector to send output to
 * @param reporter the reporter for the task
 * @param outputKeyClass the class of the output keys
 * @param outputValueClass the class of the output values
 * @throws IOException
 * @throws InterruptedException
 */
Application(JobConf conf, 
            RecordReader<FloatWritable, NullWritable> recordReader, 
            OutputCollector<K2,V2> output, Reporter reporter,
            Class<? extends K2> outputKeyClass,
            Class<? extends V2> outputValueClass
            ) throws IOException, InterruptedException {
  serverSocket = new ServerSocket(0);
  Map<String, String> env = new HashMap<String,String>();
  // add TMPDIR environment variable with the value of java.io.tmpdir
  env.put("TMPDIR", System.getProperty("java.io.tmpdir"));
  env.put(Submitter.PORT, 
          Integer.toString(serverSocket.getLocalPort()));
  
  //Add token to the environment if security is enabled
  Token<JobTokenIdentifier> jobToken = TokenCache.getJobToken(conf
      .getCredentials());
  // This password is used as shared secret key between this application and
  // child pipes process
  byte[]  password = jobToken.getPassword();
  String localPasswordFile = new File(".") + Path.SEPARATOR
      + "jobTokenPassword";
  writePasswordToLocalFile(localPasswordFile, password, conf);
  env.put("hadoop.pipes.shared.secret.location", localPasswordFile);
 
  List<String> cmd = new ArrayList<String>();
  String interpretor = conf.get(Submitter.INTERPRETOR);
  if (interpretor != null) {
    cmd.add(interpretor);
  }
  String executable = DistributedCache.getLocalCacheFiles(conf)[0].toString();
  if (!FileUtil.canExecute(new File(executable))) {
    // LinuxTaskController sets +x permissions on all distcache files already.
    // In case of DefaultTaskController, set permissions here.
    FileUtil.chmod(executable, "u+x");
  }
  cmd.add(executable);
  // wrap the command in a stdout/stderr capture
  // we are starting map/reduce task of the pipes job. this is not a cleanup
  // attempt. 
  TaskAttemptID taskid = 
    TaskAttemptID.forName(conf.get(MRJobConfig.TASK_ATTEMPT_ID));
  File stdout = TaskLog.getTaskLogFile(taskid, false, TaskLog.LogName.STDOUT);
  File stderr = TaskLog.getTaskLogFile(taskid, false, TaskLog.LogName.STDERR);
  long logLength = TaskLog.getTaskLogLength(conf);
  cmd = TaskLog.captureOutAndError(null, cmd, stdout, stderr, logLength,
                                   false);
  
  process = runClient(cmd, env);
  clientSocket = serverSocket.accept();
  
  String challenge = getSecurityChallenge();
  String digestToSend = createDigest(password, challenge);
  String digestExpected = createDigest(password, digestToSend);
  
  handler = new OutputHandler<K2, V2>(output, reporter, recordReader, 
      digestExpected);
  K2 outputKey = (K2)
    ReflectionUtils.newInstance(outputKeyClass, conf);
  V2 outputValue = (V2) 
    ReflectionUtils.newInstance(outputValueClass, conf);
  downlink = new BinaryProtocol<K1, V1, K2, V2>(clientSocket, handler, 
                                outputKey, outputValue, conf);
  
  downlink.authenticate(digestToSend, challenge);
  waitForAuthentication();
  LOG.debug("Authentication succeeded");
  downlink.start();
  downlink.setJobConf(conf);
}
 

@Override
public void stop() {
  super.stop();
  TaskLog.syncLogsShutdown(logSyncer);
}
 

/**
 * Start the child process to handle the task for us.
 * @param conf the task's configuration
 * @param recordReader the fake record reader to update progress with
 * @param output the collector to send output to
 * @param reporter the reporter for the task
 * @param outputKeyClass the class of the output keys
 * @param outputValueClass the class of the output values
 * @throws IOException
 * @throws InterruptedException
 */
Application(JobConf conf, 
            RecordReader<FloatWritable, NullWritable> recordReader, 
            OutputCollector<K2,V2> output, Reporter reporter,
            Class<? extends K2> outputKeyClass,
            Class<? extends V2> outputValueClass
            ) throws IOException, InterruptedException {
  serverSocket = new ServerSocket(0);
  Map<String, String> env = new HashMap<String,String>();
  // add TMPDIR environment variable with the value of java.io.tmpdir
  env.put("TMPDIR", System.getProperty("java.io.tmpdir"));
  env.put(Submitter.PORT, 
          Integer.toString(serverSocket.getLocalPort()));
  
  //Add token to the environment if security is enabled
  Token<JobTokenIdentifier> jobToken = TokenCache.getJobToken(conf
      .getCredentials());
  // This password is used as shared secret key between this application and
  // child pipes process
  byte[]  password = jobToken.getPassword();
  String localPasswordFile = new File(".") + Path.SEPARATOR
      + "jobTokenPassword";
  writePasswordToLocalFile(localPasswordFile, password, conf);
  env.put("hadoop.pipes.shared.secret.location", localPasswordFile);
 
  List<String> cmd = new ArrayList<String>();
  String interpretor = conf.get(Submitter.INTERPRETOR);
  if (interpretor != null) {
    cmd.add(interpretor);
  }
  String executable = DistributedCache.getLocalCacheFiles(conf)[0].toString();
  if (!FileUtil.canExecute(new File(executable))) {
    // LinuxTaskController sets +x permissions on all distcache files already.
    // In case of DefaultTaskController, set permissions here.
    FileUtil.chmod(executable, "u+x");
  }
  cmd.add(executable);
  // wrap the command in a stdout/stderr capture
  // we are starting map/reduce task of the pipes job. this is not a cleanup
  // attempt. 
  TaskAttemptID taskid = 
    TaskAttemptID.forName(conf.get(MRJobConfig.TASK_ATTEMPT_ID));
  File stdout = TaskLog.getTaskLogFile(taskid, false, TaskLog.LogName.STDOUT);
  File stderr = TaskLog.getTaskLogFile(taskid, false, TaskLog.LogName.STDERR);
  long logLength = TaskLog.getTaskLogLength(conf);
  cmd = TaskLog.captureOutAndError(null, cmd, stdout, stderr, logLength,
                                   false);
  
  process = runClient(cmd, env);
  clientSocket = serverSocket.accept();
  
  String challenge = getSecurityChallenge();
  String digestToSend = createDigest(password, challenge);
  String digestExpected = createDigest(password, digestToSend);
  
  handler = new OutputHandler<K2, V2>(output, reporter, recordReader, 
      digestExpected);
  K2 outputKey = (K2)
    ReflectionUtils.newInstance(outputKeyClass, conf);
  V2 outputValue = (V2) 
    ReflectionUtils.newInstance(outputValueClass, conf);
  downlink = new BinaryProtocol<K1, V1, K2, V2>(clientSocket, handler, 
                                outputKey, outputValue, conf);
  
  downlink.authenticate(digestToSend, challenge);
  waitForAuthentication();
  LOG.debug("Authentication succeeded");
  downlink.start();
  downlink.setJobConf(conf);
}
 

@Override
public void stop() {
  super.stop();
  TaskLog.syncLogsShutdown(logSyncer);
}