public static void main(String[] args) throws IOException {
    Path baseDir = null;
    String localPath = null;
    String preservePath = null;
    String sIgnoreTablesFilename = null;
    String sNoPreserveFilename = null;
    String sDateString = null;
    long size = 0;

    // UNIX dates for right now
    long now = new java.util.Date().getTime() / 1000;
    long maxDate = now;

    for (int i = 0; i < args.length; i++) {
      if (args[i].equals("--hdfs-path")) {
        baseDir = new Path(args[++i]);
        continue;
      }
      if (args[i].equals("--local-path")) {
        localPath = args[++i];
        continue;
      }
      if (args[i].equals("--preserve-path")) {
        preservePath = args[++i];
        continue;
      }
      if (args[i].equals("--no-preserve")) {
        sNoPreserveFilename = args[++i];
        continue;
      }
      if (args[i].equals("--ignore-tables")) {
        sIgnoreTablesFilename = args[++i];
        continue;
      }
      if (args[i].equals("--sleep")) {
        try {
          m_nSleepSeconds = Integer.parseInt(args[++i]);
        } catch (Exception e) {
          System.err.println("ERROR: " + e.toString() + "\n");
          usage();
        }
        continue;
      }
      if (args[i].equals("--dry-run")) {
        m_bDryRun = true;
        continue;
      }
      if (args[i].equals("--date")) {
        sDateString = args[++i];
        continue;
      }
      if (args[i].equals("--max-date")) {
        maxDate = Long.parseLong(args[++i]);
        continue;
      }
      if (args[i].equals("--max-bytes")) {
        size = Long.parseLong(args[++i]);
        continue;
      }

      System.err.println("ERROR: unknown arg " + args[i]);
      usage();
    }

    if (baseDir == null || localPath == null || preservePath == null || sDateString == null) {
      usage();
    }

    long minDate;

    if ("yesterday".equals(sDateString)) {
      // figure out yesterday's dates
      Calendar cal = Calendar.getInstance();
      cal.roll(Calendar.DAY_OF_YEAR, -1);

      // yesterday midnight
      cal.set(Calendar.HOUR_OF_DAY, 0);
      cal.set(Calendar.MINUTE, 0);
      cal.set(Calendar.SECOND, 0);
      cal.set(Calendar.MILLISECOND, 0);

      minDate = cal.getTimeInMillis() / 1000;

      // yesterday end of day
      cal.set(Calendar.HOUR_OF_DAY, 23);
      cal.set(Calendar.MINUTE, 59);
      cal.set(Calendar.SECOND, 59);
      cal.set(Calendar.MILLISECOND, 999);

      maxDate = cal.getTimeInMillis() / 1000;
    } else if ("last-week".equals(sDateString)) {
      minDate = maxDate - (7 * 24 * 60 * 60);
    } else if ("last-day".equals(sDateString)) {
      minDate = maxDate - (24 * 60 * 60);
    } else {
      // UNIX date since epoch of last backup
      minDate = Long.parseLong(sDateString);
    }

    long tmpDate = 0;
    BackupHdfs bak = new BackupHdfs();

    // initialize the list of tables to ignore
    if (sIgnoreTablesFilename != null) {
      bak.initializeTablesToIgnore(sIgnoreTablesFilename);
    }

    // initialize list of files to not preserve
    if (sNoPreserveFilename != null) {
      bak.initializeNoPreserve(sNoPreserveFilename);
    }

    ArrayList<Path> pathList = new ArrayList<Path>(2000);
    HashMap<Path, Long> hmTimestamps = new HashMap<Path, Long>();

    Configuration conf = new Configuration();
    FileSystem fs = FileSystem.get(conf);

    // If the HDFS path is a dir continue
    if (fs.getFileStatus(baseDir).isDir()) {
      Calendar cal = Calendar.getInstance();

      System.err.println("");
      cal.setTimeInMillis(minDate * 1000);
      System.err.println("min date = " + cal.getTime().toString());

      cal.setTimeInMillis(maxDate * 1000);
      System.err.println("max date = " + cal.getTime().toString());

      System.err.println("");
      System.err.println("Searching filesystem: " + baseDir.toUri().getPath());

      bak.checkDir(fs, minDate, maxDate, baseDir, pathList, hmTimestamps);

      System.err.println("");
      System.err.println("Skipped " + m_nIgnoredTables + " files due to ignored tables");

      System.err.println("");
      System.err.println("Number of files to backup = " + pathList.size());

      System.err.println("Total bytes to backup = " + prettyPrintBytes(m_nTotalBytes));

      System.err.println("");
      System.err.println("sorting list of files...");
      Collections.sort(pathList, new DateComparator(hmTimestamps));
      System.err.println("done");

      System.err.println("");
      System.err.println("starting backup...");
      tmpDate = bak.backupFiles(localPath, preservePath, fs, pathList, size);

      bak.closeFiles();

      System.err.println("");
      System.err.println("backup completed...");
    }

    if (tmpDate == 0) {
      // If not size limit reached print out date for right now
      System.out.println(maxDate);
    } else {
      // Print out date for last file backed up
      System.err.println("Size limit reached.");
      System.out.println(tmpDate);
    }

    System.exit(0);
  }
예제 #2
0
 /**
  * Get the archive entries in classpath as an array of Path
  *
  * @param conf Configuration that contains the classpath setting
  */
 public static Path[] getArchiveClassPaths(Configuration conf) {
   String classpath = conf.get("mapred.job.classpath.archives");
   if (classpath == null) return null;
   ArrayList list =
       Collections.list(new StringTokenizer(classpath, System.getProperty("path.separator")));
   Path[] paths = new Path[list.size()];
   for (int i = 0; i < list.size(); i++) {
     paths[i] = new Path((String) list.get(i));
   }
   return paths;
 }
예제 #3
0
    /**
     * Produce splits such that each is no greater than the quotient of the total size and the
     * number of splits requested.
     *
     * @param job The handle to the JobConf object
     * @param numSplits Number of splits requested
     */
    public InputSplit[] getSplits(JobConf job, int numSplits) throws IOException {
      int cnfiles = job.getInt(SRC_COUNT_LABEL, -1);
      long cbsize = job.getLong(TOTAL_SIZE_LABEL, -1);
      String srcfilelist = job.get(SRC_LIST_LABEL, "");
      if (cnfiles < 0 || cbsize < 0 || "".equals(srcfilelist)) {
        throw new RuntimeException(
            "Invalid metadata: #files("
                + cnfiles
                + ") total_size("
                + cbsize
                + ") listuri("
                + srcfilelist
                + ")");
      }
      Path src = new Path(srcfilelist);
      FileSystem fs = src.getFileSystem(job);
      FileStatus srcst = fs.getFileStatus(src);

      ArrayList<FileSplit> splits = new ArrayList<FileSplit>(numSplits);
      LongWritable key = new LongWritable();
      FilePair value = new FilePair();
      final long targetsize = cbsize / numSplits;
      long pos = 0L;
      long last = 0L;
      long acc = 0L;
      long cbrem = srcst.getLen();
      SequenceFile.Reader sl = null;
      try {
        sl = new SequenceFile.Reader(fs, src, job);
        for (; sl.next(key, value); last = sl.getPosition()) {
          // if adding this split would put this split past the target size,
          // cut the last split and put this next file in the next split.
          if (acc + key.get() > targetsize && acc != 0) {
            long splitsize = last - pos;
            splits.add(new FileSplit(src, pos, splitsize, (String[]) null));
            cbrem -= splitsize;
            pos = last;
            acc = 0L;
          }
          acc += key.get();
        }
      } finally {
        checkAndClose(sl);
      }
      if (cbrem != 0) {
        splits.add(new FileSplit(src, pos, cbrem, (String[]) null));
      }

      return splits.toArray(new FileSplit[splits.size()]);
    }
예제 #4
0
  /**
   * A TaskTracker wants to know the physical locations of completed, but not yet closed, tasks.
   * This exists so the reduce task thread can locate map task outputs.
   */
  public synchronized MapOutputLocation[] locateMapOutputs(
      String taskId, String[][] mapTasksNeeded) {
    ArrayList v = new ArrayList();
    for (int i = 0; i < mapTasksNeeded.length; i++) {
      for (int j = 0; j < mapTasksNeeded[i].length; j++) {
        TaskInProgress tip = (TaskInProgress) taskidToTIPMap.get(mapTasksNeeded[i][j]);
        if (tip != null && tip.isComplete(mapTasksNeeded[i][j])) {
          String trackerId = (String) taskidToTrackerMap.get(mapTasksNeeded[i][j]);
          TaskTrackerStatus tracker;
          synchronized (taskTrackers) {
            tracker = (TaskTrackerStatus) taskTrackers.get(trackerId);
          }
          v.add(new MapOutputLocation(mapTasksNeeded[i][j], tracker.getHost(), tracker.getPort()));
          break;
        }
      }
    }
    // randomly shuffle results to load-balance map output requests
    Collections.shuffle(v);

    return (MapOutputLocation[]) v.toArray(new MapOutputLocation[v.size()]);
  }
예제 #5
0
  /**
   * Launch application for the dag represented by this client.
   *
   * @throws YarnException
   * @throws IOException
   */
  public void startApplication() throws YarnException, IOException {
    Class<?>[] defaultClasses;

    if (applicationType.equals(YARN_APPLICATION_TYPE)) {
      // TODO restrict the security check to only check if security is enabled for webservices.
      if (UserGroupInformation.isSecurityEnabled()) {
        defaultClasses = DATATORRENT_SECURITY_CLASSES;
      } else {
        defaultClasses = DATATORRENT_CLASSES;
      }
    } else {
      throw new IllegalStateException(applicationType + " is not a valid application type.");
    }

    LinkedHashSet<String> localJarFiles = findJars(dag, defaultClasses);

    if (resources != null) {
      localJarFiles.addAll(resources);
    }

    YarnClusterMetrics clusterMetrics = yarnClient.getYarnClusterMetrics();
    LOG.info(
        "Got Cluster metric info from ASM"
            + ", numNodeManagers="
            + clusterMetrics.getNumNodeManagers());

    // GetClusterNodesRequest clusterNodesReq = Records.newRecord(GetClusterNodesRequest.class);
    // GetClusterNodesResponse clusterNodesResp =
    // rmClient.clientRM.getClusterNodes(clusterNodesReq);
    // LOG.info("Got Cluster node info from ASM");
    // for (NodeReport node : clusterNodesResp.getNodeReports()) {
    //  LOG.info("Got node report from ASM for"
    //           + ", nodeId=" + node.getNodeId()
    //           + ", nodeAddress" + node.getHttpAddress()
    //           + ", nodeRackName" + node.getRackName()
    //           + ", nodeNumContainers" + node.getNumContainers()
    //           + ", nodeHealthStatus" + node.getHealthReport());
    // }
    List<QueueUserACLInfo> listAclInfo = yarnClient.getQueueAclsInfo();
    for (QueueUserACLInfo aclInfo : listAclInfo) {
      for (QueueACL userAcl : aclInfo.getUserAcls()) {
        LOG.info(
            "User ACL Info for Queue"
                + ", queueName="
                + aclInfo.getQueueName()
                + ", userAcl="
                + userAcl.name());
      }
    }

    // Get a new application id
    YarnClientApplication newApp = yarnClient.createApplication();
    appId = newApp.getNewApplicationResponse().getApplicationId();

    // Dump out information about cluster capability as seen by the resource manager
    int maxMem = newApp.getNewApplicationResponse().getMaximumResourceCapability().getMemory();
    LOG.info("Max mem capabililty of resources in this cluster " + maxMem);
    int amMemory = dag.getMasterMemoryMB();
    if (amMemory > maxMem) {
      LOG.info(
          "AM memory specified above max threshold of cluster. Using max value."
              + ", specified="
              + amMemory
              + ", max="
              + maxMem);
      amMemory = maxMem;
    }

    if (dag.getAttributes().get(LogicalPlan.APPLICATION_ID) == null) {
      dag.setAttribute(LogicalPlan.APPLICATION_ID, appId.toString());
    }

    // Create launch context for app master
    LOG.info("Setting up application submission context for ASM");
    ApplicationSubmissionContext appContext = Records.newRecord(ApplicationSubmissionContext.class);

    // set the application id
    appContext.setApplicationId(appId);
    // set the application name
    appContext.setApplicationName(dag.getValue(LogicalPlan.APPLICATION_NAME));
    appContext.setApplicationType(this.applicationType);
    if (YARN_APPLICATION_TYPE.equals(this.applicationType)) {
      // appContext.setMaxAppAttempts(1); // no retries until Stram is HA
    }

    // Set up the container launch context for the application master
    ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class);

    // Setup security tokens
    // If security is enabled get ResourceManager and NameNode delegation tokens.
    // Set these tokens on the container so that they are sent as part of application submission.
    // This also sets them up for renewal by ResourceManager. The NameNode delegation rmToken
    // is also used by ResourceManager to fetch the jars from HDFS and set them up for the
    // application master launch.
    if (UserGroupInformation.isSecurityEnabled()) {
      Credentials credentials = new Credentials();
      String tokenRenewer = conf.get(YarnConfiguration.RM_PRINCIPAL);
      if (tokenRenewer == null || tokenRenewer.length() == 0) {
        throw new IOException("Can't get Master Kerberos principal for the RM to use as renewer");
      }

      // For now, only getting tokens for the default file-system.
      FileSystem fs = StramClientUtils.newFileSystemInstance(conf);
      try {
        final Token<?> tokens[] = fs.addDelegationTokens(tokenRenewer, credentials);
        if (tokens != null) {
          for (Token<?> token : tokens) {
            LOG.info("Got dt for " + fs.getUri() + "; " + token);
          }
        }
      } finally {
        fs.close();
      }

      addRMDelegationToken(tokenRenewer, credentials);

      DataOutputBuffer dob = new DataOutputBuffer();
      credentials.writeTokenStorageToStream(dob);
      ByteBuffer fsTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
      amContainer.setTokens(fsTokens);
    }

    // set local resources for the application master
    // local files or archives as needed
    // In this scenario, the jar file for the application master is part of the local resources
    Map<String, LocalResource> localResources = new HashMap<String, LocalResource>();

    // copy required jar files to dfs, to be localized for containers
    FileSystem fs = StramClientUtils.newFileSystemInstance(conf);
    try {
      Path appsBasePath =
          new Path(StramClientUtils.getDTDFSRootDir(fs, conf), StramClientUtils.SUBDIR_APPS);
      Path appPath = new Path(appsBasePath, appId.toString());

      String libJarsCsv = copyFromLocal(fs, appPath, localJarFiles.toArray(new String[] {}));

      LOG.info("libjars: {}", libJarsCsv);
      dag.getAttributes().put(LogicalPlan.LIBRARY_JARS, libJarsCsv);
      LaunchContainerRunnable.addFilesToLocalResources(
          LocalResourceType.FILE, libJarsCsv, localResources, fs);

      if (archives != null) {
        String[] localFiles = archives.split(",");
        String archivesCsv = copyFromLocal(fs, appPath, localFiles);
        LOG.info("archives: {}", archivesCsv);
        dag.getAttributes().put(LogicalPlan.ARCHIVES, archivesCsv);
        LaunchContainerRunnable.addFilesToLocalResources(
            LocalResourceType.ARCHIVE, archivesCsv, localResources, fs);
      }

      if (files != null) {
        String[] localFiles = files.split(",");
        String filesCsv = copyFromLocal(fs, appPath, localFiles);
        LOG.info("files: {}", filesCsv);
        dag.getAttributes().put(LogicalPlan.FILES, filesCsv);
        LaunchContainerRunnable.addFilesToLocalResources(
            LocalResourceType.FILE, filesCsv, localResources, fs);
      }

      dag.getAttributes().put(LogicalPlan.APPLICATION_PATH, appPath.toString());
      if (dag.getAttributes().get(OperatorContext.STORAGE_AGENT) == null) {
          /* which would be the most likely case */
        Path checkpointPath = new Path(appPath, LogicalPlan.SUBDIR_CHECKPOINTS);
        // use conf client side to pickup any proxy settings from dt-site.xml
        dag.setAttribute(
            OperatorContext.STORAGE_AGENT, new FSStorageAgent(checkpointPath.toString(), conf));
      }
      if (dag.getAttributes().get(LogicalPlan.CONTAINER_OPTS_CONFIGURATOR) == null) {
        dag.setAttribute(
            LogicalPlan.CONTAINER_OPTS_CONFIGURATOR, new BasicContainerOptConfigurator());
      }

      // Set the log4j properties if needed
      if (!log4jPropFile.isEmpty()) {
        Path log4jSrc = new Path(log4jPropFile);
        Path log4jDst = new Path(appPath, "log4j.props");
        fs.copyFromLocalFile(false, true, log4jSrc, log4jDst);
        FileStatus log4jFileStatus = fs.getFileStatus(log4jDst);
        LocalResource log4jRsrc = Records.newRecord(LocalResource.class);
        log4jRsrc.setType(LocalResourceType.FILE);
        log4jRsrc.setVisibility(LocalResourceVisibility.APPLICATION);
        log4jRsrc.setResource(ConverterUtils.getYarnUrlFromURI(log4jDst.toUri()));
        log4jRsrc.setTimestamp(log4jFileStatus.getModificationTime());
        log4jRsrc.setSize(log4jFileStatus.getLen());
        localResources.put("log4j.properties", log4jRsrc);
      }

      if (originalAppId != null) {
        Path origAppPath = new Path(appsBasePath, this.originalAppId);
        LOG.info("Restart from {}", origAppPath);
        copyInitialState(origAppPath);
      }

      // push logical plan to DFS location
      Path cfgDst = new Path(appPath, LogicalPlan.SER_FILE_NAME);
      FSDataOutputStream outStream = fs.create(cfgDst, true);
      LogicalPlan.write(this.dag, outStream);
      outStream.close();

      Path launchConfigDst = new Path(appPath, LogicalPlan.LAUNCH_CONFIG_FILE_NAME);
      outStream = fs.create(launchConfigDst, true);
      conf.writeXml(outStream);
      outStream.close();

      FileStatus topologyFileStatus = fs.getFileStatus(cfgDst);
      LocalResource topologyRsrc = Records.newRecord(LocalResource.class);
      topologyRsrc.setType(LocalResourceType.FILE);
      topologyRsrc.setVisibility(LocalResourceVisibility.APPLICATION);
      topologyRsrc.setResource(ConverterUtils.getYarnUrlFromURI(cfgDst.toUri()));
      topologyRsrc.setTimestamp(topologyFileStatus.getModificationTime());
      topologyRsrc.setSize(topologyFileStatus.getLen());
      localResources.put(LogicalPlan.SER_FILE_NAME, topologyRsrc);

      // Set local resource info into app master container launch context
      amContainer.setLocalResources(localResources);

      // Set the necessary security tokens as needed
      // amContainer.setContainerTokens(containerToken);
      // Set the env variables to be setup in the env where the application master will be run
      LOG.info("Set the environment for the application master");
      Map<String, String> env = new HashMap<String, String>();

      // Add application jar(s) location to classpath
      // At some point we should not be required to add
      // the hadoop specific classpaths to the env.
      // It should be provided out of the box.
      // For now setting all required classpaths including
      // the classpath to "." for the application jar(s)
      // including ${CLASSPATH} will duplicate the class path in app master, removing it for now
      // StringBuilder classPathEnv = new StringBuilder("${CLASSPATH}:./*");
      StringBuilder classPathEnv = new StringBuilder("./*");
      String classpath = conf.get(YarnConfiguration.YARN_APPLICATION_CLASSPATH);
      for (String c :
          StringUtils.isBlank(classpath)
              ? YarnConfiguration.DEFAULT_YARN_APPLICATION_CLASSPATH
              : classpath.split(",")) {
        if (c.equals("$HADOOP_CLIENT_CONF_DIR")) {
          // SPOI-2501
          continue;
        }
        classPathEnv.append(':');
        classPathEnv.append(c.trim());
      }
      env.put("CLASSPATH", classPathEnv.toString());
      // propagate to replace node managers user name (effective in non-secure mode)
      env.put("HADOOP_USER_NAME", UserGroupInformation.getLoginUser().getUserName());

      amContainer.setEnvironment(env);

      // Set the necessary command to execute the application master
      ArrayList<CharSequence> vargs = new ArrayList<CharSequence>(30);

      // Set java executable command
      LOG.info("Setting up app master command");
      vargs.add(javaCmd);
      if (dag.isDebug()) {
        vargs.add("-agentlib:jdwp=transport=dt_socket,server=y,suspend=n");
      }
      // Set Xmx based on am memory size
      // default heap size 75% of total memory
      vargs.add("-Xmx" + (amMemory * 3 / 4) + "m");
      vargs.add("-XX:+HeapDumpOnOutOfMemoryError");
      vargs.add("-XX:HeapDumpPath=/tmp/dt-heap-" + appId.getId() + ".bin");
      vargs.add("-Dhadoop.root.logger=" + (dag.isDebug() ? "DEBUG" : "INFO") + ",RFA");
      vargs.add("-Dhadoop.log.dir=" + ApplicationConstants.LOG_DIR_EXPANSION_VAR);
      vargs.add(String.format("-D%s=%s", StreamingContainer.PROP_APP_PATH, dag.assertAppPath()));
      if (dag.isDebug()) {
        vargs.add("-Dlog4j.debug=true");
      }

      String loggersLevel = conf.get(DTLoggerFactory.DT_LOGGERS_LEVEL);
      if (loggersLevel != null) {
        vargs.add(String.format("-D%s=%s", DTLoggerFactory.DT_LOGGERS_LEVEL, loggersLevel));
      }
      vargs.add(StreamingAppMaster.class.getName());
      vargs.add("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stdout");
      vargs.add("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stderr");

      // Get final command
      StringBuilder command = new StringBuilder(9 * vargs.size());
      for (CharSequence str : vargs) {
        command.append(str).append(" ");
      }

      LOG.info("Completed setting up app master command " + command.toString());
      List<String> commands = new ArrayList<String>();
      commands.add(command.toString());
      amContainer.setCommands(commands);

      // Set up resource type requirements
      // For now, only memory is supported so we set memory requirements
      Resource capability = Records.newRecord(Resource.class);
      capability.setMemory(amMemory);
      appContext.setResource(capability);

      // Service data is a binary blob that can be passed to the application
      // Not needed in this scenario
      // amContainer.setServiceData(serviceData);
      appContext.setAMContainerSpec(amContainer);

      // Set the priority for the application master
      Priority pri = Records.newRecord(Priority.class);
      pri.setPriority(amPriority);
      appContext.setPriority(pri);
      // Set the queue to which this application is to be submitted in the RM
      appContext.setQueue(queueName);

      // Submit the application to the applications manager
      // SubmitApplicationResponse submitResp = rmClient.submitApplication(appRequest);
      // Ignore the response as either a valid response object is returned on success
      // or an exception thrown to denote some form of a failure
      String specStr =
          Objects.toStringHelper("Submitting application: ")
              .add("name", appContext.getApplicationName())
              .add("queue", appContext.getQueue())
              .add("user", UserGroupInformation.getLoginUser())
              .add("resource", appContext.getResource())
              .toString();
      LOG.info(specStr);
      if (dag.isDebug()) {
        // LOG.info("Full submission context: " + appContext);
      }
      yarnClient.submitApplication(appContext);
    } finally {
      fs.close();
    }
  }