List of usage examples for org.apache.hadoop.mapreduce MRConfig LOCAL_DIR
String LOCAL_DIR
To view the source code for org.apache.hadoop.mapreduce MRConfig LOCAL_DIR.
Click Source Link
From source file:com.scaleoutsoftware.soss.hserver.hadoop.DistributedCacheManager.java
License:Apache License
/** * Set up the distributed cache by localizing the resources, and updating * the configuration with references to the localized resources. * @param conf job configuration//from w w w. j av a2s . c om * @throws IOException */ public void setup(Configuration conf) throws IOException { //If we are not 0th worker, wait for 0th worker to set up the cache if (InvocationWorker.getIgWorkerIndex() > 0 && InvocationWorker.getNumberOfWorkers() > 1) { try { InvocationWorker.getSynchronizationBarrier().waitForComplete(ACTION_NAME, SYNCHRONIZATION_WAIT_MS, WAIT_GRANULARITY_MS); } catch (Exception e) { throw new RuntimeException(e); } return; } File workDir = new File(System.getProperty("user.dir")); // Generate YARN local resources objects corresponding to the distributed // cache configuration Map<String, LocalResource> localResources = new LinkedHashMap<String, LocalResource>(); MRApps.setupDistributedCache(conf, localResources); //CODE CHANGE FROM ORIGINAL FILE: //We need to clear the resources from jar files, since they are distributed through the IG. // Iterator<Map.Entry<String, LocalResource>> iterator = localResources.entrySet().iterator(); while (iterator.hasNext()) { Entry<String, LocalResource> entry = iterator.next(); if (entry.getKey().endsWith(".jar")) { iterator.remove(); } } // Generating unique numbers for FSDownload. AtomicLong uniqueNumberGenerator = new AtomicLong(System.currentTimeMillis()); // Find which resources are to be put on the local classpath Map<String, Path> classpaths = new HashMap<String, Path>(); Path[] archiveClassPaths = DistributedCache.getArchiveClassPaths(conf); if (archiveClassPaths != null) { for (Path p : archiveClassPaths) { FileSystem remoteFS = p.getFileSystem(conf); p = remoteFS.resolvePath(p.makeQualified(remoteFS.getUri(), remoteFS.getWorkingDirectory())); classpaths.put(p.toUri().getPath().toString(), p); } } Path[] fileClassPaths = DistributedCache.getFileClassPaths(conf); if (fileClassPaths != null) { for (Path p : fileClassPaths) { FileSystem remoteFS = p.getFileSystem(conf); p = remoteFS.resolvePath(p.makeQualified(remoteFS.getUri(), remoteFS.getWorkingDirectory())); classpaths.put(p.toUri().getPath().toString(), p); } } // Localize the resources LocalDirAllocator localDirAllocator = new LocalDirAllocator(MRConfig.LOCAL_DIR); FileContext localFSFileContext = FileContext.getLocalFSFileContext(); UserGroupInformation ugi = UserGroupInformation.getCurrentUser(); ExecutorService exec = null; try { ThreadFactory tf = new ThreadFactoryBuilder() .setNameFormat("LocalDistributedCacheManager Downloader #%d").build(); exec = Executors.newCachedThreadPool(tf); Path destPath = localDirAllocator.getLocalPathForWrite(".", conf); Map<LocalResource, Future<Path>> resourcesToPaths = Maps.newHashMap(); for (LocalResource resource : localResources.values()) { Callable<Path> download = new FSDownload(localFSFileContext, ugi, conf, new Path(destPath, Long.toString(uniqueNumberGenerator.incrementAndGet())), resource); Future<Path> future = exec.submit(download); resourcesToPaths.put(resource, future); } for (Entry<String, LocalResource> entry : localResources.entrySet()) { LocalResource resource = entry.getValue(); Path path; try { path = resourcesToPaths.get(resource).get(); } catch (InterruptedException e) { throw new IOException(e); } catch (ExecutionException e) { throw new IOException(e); } String pathString = path.toUri().toString(); String link = entry.getKey(); String target = new File(path.toUri()).getPath(); symlink(workDir, target, link); if (resource.getType() == LocalResourceType.ARCHIVE) { localArchives.add(pathString); } else if (resource.getType() == LocalResourceType.FILE) { localFiles.add(pathString); } else if (resource.getType() == LocalResourceType.PATTERN) { //PATTERN is not currently used in local mode throw new IllegalArgumentException( "Resource type PATTERN is not " + "implemented yet. " + resource.getResource()); } Path resourcePath; try { resourcePath = ConverterUtils.getPathFromYarnURL(resource.getResource()); } catch (URISyntaxException e) { throw new IOException(e); } LOG.info(String.format("Localized %s as %s", resourcePath, path)); String cp = resourcePath.toUri().getPath(); if (classpaths.keySet().contains(cp)) { localClasspaths.add(path.toUri().getPath().toString()); } } } finally { if (exec != null) { exec.shutdown(); } } // Update the configuration object with localized data. if (!localArchives.isEmpty()) { conf.set(MRJobConfig.CACHE_LOCALARCHIVES, StringUtils.arrayToString(localArchives.toArray(new String[localArchives.size()]))); } if (!localFiles.isEmpty()) { conf.set(MRJobConfig.CACHE_LOCALFILES, StringUtils.arrayToString(localFiles.toArray(new String[localArchives.size()]))); } setupCalled = true; //If we are 0th worker, signal action complete if (InvocationWorker.getIgWorkerIndex() == 0 && InvocationWorker.getNumberOfWorkers() > 1) { try { InvocationWorker.getSynchronizationBarrier().signalComplete(ACTION_NAME); } catch (Exception e) { throw new RuntimeException(e); } } }
From source file:org.apache.tez.mapreduce.processor.map.TestMapProcessor.java
License:Apache License
public void setUpJobConf(JobConf job) { job.set(TezRuntimeFrameworkConfigs.LOCAL_DIRS, workDir.toString()); job.set(MRConfig.LOCAL_DIR, workDir.toString()); job.setClass(Constants.TEZ_RUNTIME_TASK_OUTPUT_MANAGER, TezTaskOutputFiles.class, TezTaskOutput.class); job.set(TezRuntimeConfiguration.TEZ_RUNTIME_PARTITIONER_CLASS, MRPartitioner.class.getName()); job.setNumReduceTasks(1);/* w w w. ja v a 2 s .c o m*/ }
From source file:org.apache.tez.mapreduce.task.MRRuntimeTask.java
License:Apache License
private static void configureMRTask(JobConf job, MRTask task) throws IOException, InterruptedException { Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); job.setCredentials(credentials);/*from ww w. ja va 2s . c o m*/ // TODO Can this be avoided all together. Have the MRTezOutputCommitter use // the Tez parameter. // TODO This could be fetched from the env if YARN is setting it for all // Containers. // Set it in conf, so as to be able to be used the the OutputCommitter. job.setInt(MRJobConfig.APPLICATION_ATTEMPT_ID, job.getInt(TezJobConfig.APPLICATION_ATTEMPT_ID, -1)); job.setClass(MRConfig.TASK_LOCAL_OUTPUT_CLASS, YarnOutputFiles.class, MapOutputFile.class); // MR Token<JobTokenIdentifier> jobToken = TokenCache.getJobToken(credentials); if (jobToken != null) { // Will MR ever run without a job token. SecretKey sk = JobTokenSecretManager.createSecretKey(jobToken.getPassword()); task.setJobTokenSecret(sk); } else { LOG.warn("No job token set"); } job.set(MRJobConfig.JOB_LOCAL_DIR, job.get(TezJobConfig.JOB_LOCAL_DIR)); job.set(MRConfig.LOCAL_DIR, job.get(TezJobConfig.LOCAL_DIRS)); if (job.get(TezJobConfig.DAG_CREDENTIALS_BINARY) != null) { job.set(MRJobConfig.MAPREDUCE_JOB_CREDENTIALS_BINARY, job.get(TezJobConfig.DAG_CREDENTIALS_BINARY)); } // setup the child's attempt directories // Do the task-type specific localization task.localizeConfiguration(job); // Set up the DistributedCache related configs setupDistributedCacheConfig(job); task.setConf(job); }