List of usage examples for org.apache.hadoop.mapred JobConf setNumMapTasks
public void setNumMapTasks(int n)
From source file:com.scaleoutsoftware.soss.hserver.JobScheduler.java
License:Apache License
/** * Runs the map-reduce job on ScaleOut hServer.* * * @param job the job to run/*from www . ja v a2 s . co m*/ * @param jobId the id of the job * @param sortEnabled if key sorting is enabled * @param jobParameter user defined parameter object for the job * @param grid the invocation grid to run the job * @throws IOException if errors occurred during the job * @throws InterruptedException if the processing thread is interrupted * @throws ClassNotFoundException if the invocation grid does not contain the dependency class */ @SuppressWarnings("unchecked") public void runOldApiJob(JobConf job, org.apache.hadoop.mapred.JobID jobId, boolean sortEnabled, Object jobParameter, InvocationGrid grid) throws IOException, InterruptedException, ClassNotFoundException { //Initialize user credential in advance int jobAppId = 0xFFFFFFF & BitConverter.hashStringOneInt(jobId.toString()); String hadoopVersion = VersionInfo.getVersion(); long time = System.currentTimeMillis(); CreateUserCredentials.run(grid); try { //Check output specs before running the job job.getOutputFormat().checkOutputSpecs(FileSystem.get(job), job); JobContext jContext = HadoopVersionSpecificCode.getInstance(hadoopVersion, job).createJobContext(job, jobId); org.apache.hadoop.mapred.OutputCommitter outputCommitter = job.getOutputCommitter(); outputCommitter.setupJob(jContext); //clear all temporary objects DataAccessor.clearObjects(jobAppId); //Calculating the partition layout com.scaleoutsoftware.soss.client.util.HostToPartitionsMapping hostNameToPartition = com.scaleoutsoftware.soss.client.util.HostToPartitionsMapping .getCurrent(); List<InetAddress> hostAddresses = new ArrayList<InetAddress>(hostNameToPartition.getHosts()); //Generating mapping of Hadoop partitions to SOSS partitions, so they are equally distributed across hosts int numHosts = hostAddresses.size(); int numberOfSlotsPerNode = Math .max(grid != null ? grid.getMaxNumberOfCores() : Runtime.getRuntime().availableProcessors(), 1); //Set the number of splits to the number of cores if (NamedMapInputFormatMapred.class.isAssignableFrom(job.getInputFormat().getClass())) { int numberOfSplits = HServerParameters.getSetting(MAP_SPLITS_PER_CORE, job) * numHosts * numberOfSlotsPerNode; job.setNumMapTasks(Math.min(numberOfSplits, HServerConstants.MAX_MAP_REDUCE_TASKS)); } //Generating split to hostname map org.apache.hadoop.mapred.InputFormat inputFormat = job.getInputFormat(); List<org.apache.hadoop.mapred.InputSplit> splitList = Arrays .asList(inputFormat.getSplits(job, job.getNumMapTasks())); Map<InetAddress, List<Integer>> splitToHostAddress = assignSplitsToHost(splitList, hostAddresses, null); //Choose the optimal number of reducers for GridOutputFormat if (job.getOutputFormat() instanceof NamedMapOutputFormatMapred) { job.setNumReduceTasks(numHosts * numberOfSlotsPerNode); sortEnabled = false; } int[] partitionMapping = hostNameToPartition.generateEvenItemDistribution(job.getNumReduceTasks()); //Generating invocation parameters Class<? extends org.apache.hadoop.mapred.InputSplit> splitType = splitList.size() > 0 ? splitList.get(0).getClass() : null; HadoopInvocationParameters hadoopParameters = new HadoopInvocationParameters(job, jobId, true); HServerInvocationParameters<org.apache.hadoop.mapred.InputSplit> parameters = new HServerInvocationParameters<org.apache.hadoop.mapred.InputSplit>( hadoopParameters, jobAppId, partitionMapping, hostNameToPartition, numberOfSlotsPerNode, splitType, splitList, splitToHostAddress, false, sortEnabled, hadoopVersion, jobParameter, SerializationMode.DEFAULT); StringBuilder stringBuilder = new StringBuilder(); stringBuilder.append("Splits created:\n"); for (InetAddress address : splitToHostAddress.keySet()) { stringBuilder.append("Host "); stringBuilder.append(address); stringBuilder.append(" has "); stringBuilder.append(splitToHostAddress.get(address).size()); stringBuilder.append(" splits.\n"); } System.out.println(stringBuilder.toString()); System.out.println("Job initialization completed in " + (System.currentTimeMillis() - time) + " ms."); time = System.currentTimeMillis(); InvokeResult<MapperResult> mapInvokeResult = MessagingHelper.invoke(grid, RunMapper.MapperInvokable.class, parameters, TimeSpan.INFINITE_TIMEOUT.getSeconds()); if (mapInvokeResult.getErrors() != null && mapInvokeResult.getErrors().size() > 0) { throw new IOException("Map invocation failed.", mapInvokeResult.getErrors().get(0)); } System.out.println("Map invocation done in " + (System.currentTimeMillis() - time) + " ms."); time = System.currentTimeMillis(); MapperResult resultObject = mapInvokeResult.getResult(); if (resultObject == null || mapInvokeResult.getNumFailed() != 0) { throw new IOException("Mapper invocation failed. Num failed = " + mapInvokeResult.getNumFailed()); } if (resultObject.getNumberOfSplitsProcessed() != splitList.size()) { throw new IOException("Number of splits does not match the number of invocations. Nsplits = " + splitList.size() + ", Ninvokes =" + resultObject.getNumberOfSplitsProcessed()); } if (partitionMapping.length > 0) { //Running the reduce step InvokeResult<Integer> reduceInvokeResult = MessagingHelper.invoke(grid, ReduceInvokable.class, jobAppId, TimeSpan.INFINITE_TIMEOUT.getSeconds()); System.out.println("Reduce invocation done in " + (System.currentTimeMillis() - time) + " ms."); DataAccessor.clearObjects(jobAppId); //clear all temporary objects if (reduceInvokeResult.getErrors() != null && reduceInvokeResult.getErrors().size() > 0) { throw new IOException("Reduce invocation failed.", reduceInvokeResult.getErrors().get(0)); } if (reduceInvokeResult.getNumFailed() != 0) { throw new IOException("Reduce invocation failed."); } if (reduceInvokeResult.getResult() != partitionMapping.length) { throw new IOException("Not all partitions were reduced. Expected = " + partitionMapping.length + " Actual = " + reduceInvokeResult.getResult()); } } outputCommitter.commitJob(jContext); } catch (StateServerException e) { throw new IOException("ScaleOut hServer access error.", e); } }
From source file:com.scaleunlimited.cascading.hadoop.HadoopUtils.java
License:Apache License
public static JobConf getDefaultJobConf() throws IOException, InterruptedException { JobConf conf = new JobConf(); // We explicitly set task counts to 1 for local so that code which depends on // things like the reducer count runs properly. if (isJobLocal(conf)) { conf.setNumMapTasks(1); conf.setNumReduceTasks(1);/*from w w w. j av a 2 s . co m*/ } else { conf.setNumReduceTasks(getNumReducers(conf)); // TODO - By default we want to use 0.95 * the number of reduce slots, as per // Hadoop wiki. But we want to round, versus truncate, to avoid setting it to // 0 if we have one reducer. This way it only impacts you if you have more // than 10 reducers. // conf.setNumReduceTasks((getNumReducers(conf) * 95) / 100); } conf.setMapSpeculativeExecution(false); conf.setReduceSpeculativeExecution(false); return conf; }
From source file:com.scaleunlimited.helpful.tools.AnalyzeEmail.java
License:Apache License
private static JobConf getDefaultJobConf() throws IOException { JobClient jobClient = new JobClient(new JobConf()); ClusterStatus status = jobClient.getClusterStatus(); int trackers = status.getTaskTrackers(); JobConf conf = new JobConf(); conf.setNumMapTasks(trackers * 10); conf.setNumReduceTasks((trackers * conf.getInt("mapred.tasktracker.reduce.tasks.maximum", 2))); conf.setMapSpeculativeExecution(false); conf.setReduceSpeculativeExecution(false); conf.set("mapred.child.java.opts", "-server -Xmx512m -Xss128k"); // Should match the value used for Xss above. Note no 'k' suffix for the ulimit command. // New support that one day will be in Hadoop. conf.set("mapred.child.ulimit.stack", "128"); return conf;/* w w w .j a v a2 s. c om*/ }
From source file:com.spotify.hdfs2cass.BulkLoader.java
License:Apache License
public int run(String[] args) throws Exception { CommandLine cmdLine = parseOptions(args); String[] inputPaths = cmdLine.getOptionValues('i'); String seedNodeHost = cmdLine.getOptionValue('h'); String seedNodePort = cmdLine.getOptionValue('p', "9160"); String keyspace = cmdLine.getOptionValue('k'); String colfamily = cmdLine.getOptionValue('c'); int mappers = Integer.parseInt(cmdLine.getOptionValue('m', "0")); Integer copiers = Integer.parseInt(cmdLine.getOptionValue('P', "0")); String poolName = cmdLine.getOptionValue("pool"); ClusterInfo clusterInfo = new ClusterInfo(seedNodeHost, seedNodePort); clusterInfo.init(keyspace);//w w w.j a va 2 s. co m final String partitionerClass = clusterInfo.getPartitionerClass(); final int reducers = adjustReducers(Integer.parseInt(cmdLine.getOptionValue('r', "0")), clusterInfo.getNumClusterNodes()); Configuration conf = new Configuration(); ConfigHelper.setOutputColumnFamily(conf, keyspace, colfamily); ConfigHelper.setOutputInitialAddress(conf, seedNodeHost); ConfigHelper.setOutputRpcPort(conf, seedNodePort); ConfigHelper.setOutputPartitioner(conf, partitionerClass); if (cmdLine.hasOption('s')) { conf.set("mapreduce.output.bulkoutputformat.buffersize", cmdLine.getOptionValue('s', "32")); } if (cmdLine.hasOption('M')) { conf.set("mapreduce.output.bulkoutputformat.streamthrottlembits", cmdLine.getOptionValue('M')); } if (cmdLine.hasOption('C')) { ConfigHelper.setOutputCompressionClass(conf, cmdLine.getOptionValue('C')); } if (cmdLine.hasOption('b')) { conf.setBoolean("com.spotify.hdfs2cass.base64", true); } JobConf job = new JobConf(conf); if (mappers > 0) job.setNumMapTasks(mappers); if (reducers > 0) job.setNumReduceTasks(reducers); if (copiers > 0) job.set("mapred.reduce.parallel.copies", copiers.toString()); if (poolName != null) job.set("mapred.fairscheduler.pool", poolName); // set the nodes as a param for the other hadoop nodes clusterInfo.setConf(job); String jobName = "bulkloader-hdfs-to-cassandra"; if (cmdLine.hasOption('n')) jobName += "-" + cmdLine.getOptionValue('n'); job.setJobName(jobName); job.setJarByClass(BulkLoader.class); job.setInputFormat(AvroAsTextInputFormat.class); for (String inputPath : inputPaths) { FileInputFormat.addInputPath(job, new Path(inputPath)); } //map just outputs text, reduce sends to cassandra job.setMapperClass(MapToText.class); job.setMapOutputKeyClass(Text.class); job.setMapOutputValueClass(Text.class); job.setPartitionerClass(CassandraPartitioner.class); job.setReducerClass(ReduceTextToCassandra.class); job.setOutputKeyClass(ByteBuffer.class); job.setOutputValueClass(List.class); if (cmdLine.hasOption('s')) job.setOutputFormat(BulkOutputFormat.class); else job.setOutputFormat(ColumnFamilyOutputFormat.class); JobClient.runJob(job); return 0; }
From source file:com.test.hadoop.JhhSort.java
License:Apache License
/** * The main driver for sort program. Invoke this method to submit the * map/reduce job./*ww w . ja v a2 s .c om*/ * * @throws IOException * When there is communication problems with the job tracker. */ @SuppressWarnings({ "rawtypes" }) public int run(String[] args) throws Exception { JobConf jobConf = new JobConf(getConf(), JhhSort.class); jobConf.setJobName("sorter"); jobConf.set("mapred.job.tracker", "192.168.12.200:9001"); jobConf.set("fs.default.name", "hdfs://192.168.12.200:9000"); jobConf.setMapperClass(IdentityMapper.class); jobConf.setReducerClass(IdentityReducer.class); JobClient client = new JobClient(jobConf); ClusterStatus cluster = client.getClusterStatus(); int num_reduces = (int) (cluster.getMaxReduceTasks() * 0.5); String sort_reduces = jobConf.get("test.sort.reduces_per_host"); if (sort_reduces != null) { num_reduces = cluster.getTaskTrackers() * Integer.parseInt(sort_reduces); } Class<? extends InputFormat> inputFormatClass = TextInputFormat.class; Class<? extends OutputFormat> outputFormatClass = TextOutputFormat.class; Class<? extends WritableComparable> outputKeyClass = LongWritable.class; Class<? extends Writable> outputValueClass = LongWritable.class; List<String> otherArgs = new ArrayList<String>(); InputSampler.Sampler<K, V> sampler = null; for (int i = 0; i < args.length; ++i) { try { if ("-m".equals(args[i])) { jobConf.setNumMapTasks(Integer.parseInt(args[++i])); } else if ("-r".equals(args[i])) { num_reduces = Integer.parseInt(args[++i]); } else if ("-inFormat".equals(args[i])) { inputFormatClass = Class.forName(args[++i]).asSubclass(InputFormat.class); } else if ("-outFormat".equals(args[i])) { outputFormatClass = Class.forName(args[++i]).asSubclass(OutputFormat.class); } else if ("-outKey".equals(args[i])) { outputKeyClass = Class.forName(args[++i]).asSubclass(WritableComparable.class); } else if ("-outValue".equals(args[i])) { outputValueClass = Class.forName(args[++i]).asSubclass(Writable.class); } else if ("-totalOrder".equals(args[i])) { double pcnt = Double.parseDouble(args[++i]); int numSamples = Integer.parseInt(args[++i]); int maxSplits = Integer.parseInt(args[++i]); if (0 >= maxSplits) maxSplits = Integer.MAX_VALUE; sampler = new InputSampler.RandomSampler<K, V>(pcnt, numSamples, maxSplits); } else { otherArgs.add(args[i]); } } catch (NumberFormatException except) { System.out.println("ERROR: Integer expected instead of " + args[i]); return printUsage(); } catch (ArrayIndexOutOfBoundsException except) { System.out.println("ERROR: Required parameter missing from " + args[i - 1]); return printUsage(); // exits } } // Set user-supplied (possibly default) job configs jobConf.setNumReduceTasks(num_reduces); jobConf.setInputFormat(inputFormatClass); jobConf.setOutputFormat(outputFormatClass); jobConf.setOutputKeyClass(outputKeyClass); jobConf.setOutputValueClass(outputValueClass); // Make sure there are exactly 2 parameters left. if (otherArgs.size() != 2) { System.out.println("ERROR: Wrong number of parameters: " + otherArgs.size() + " instead of 2."); return printUsage(); } FileInputFormat.setInputPaths(jobConf, otherArgs.get(0)); FileOutputFormat.setOutputPath(jobConf, new Path(otherArgs.get(1))); if (sampler != null) { System.out.println("Sampling input to effect total-order sort..."); jobConf.setPartitionerClass(TotalOrderPartitioner.class); Path inputDir = FileInputFormat.getInputPaths(jobConf)[0]; inputDir = inputDir.makeQualified(inputDir.getFileSystem(jobConf)); Path partitionFile = new Path(inputDir, "_sortPartitioning"); TotalOrderPartitioner.setPartitionFile(jobConf, partitionFile); InputSampler.<K, V>writePartitionFile(jobConf, sampler); URI partitionUri = new URI(partitionFile.toString() + "#" + "_sortPartitioning"); DistributedCache.addCacheFile(partitionUri, jobConf); DistributedCache.createSymlink(jobConf); } System.out.println("Running on " + cluster.getTaskTrackers() + " nodes to sort from " + FileInputFormat.getInputPaths(jobConf)[0] + " into " + FileOutputFormat.getOutputPath(jobConf) + " with " + num_reduces + " reduces."); Date startTime = new Date(); System.out.println("Job started: " + startTime); jobResult = JobClient.runJob(jobConf); Date end_time = new Date(); System.out.println("Job ended: " + end_time); System.out.println("The job took " + (end_time.getTime() - startTime.getTime()) / 1000 + " seconds."); return 0; }
From source file:com.test.PiEstimatorKrb.java
License:Apache License
/** * Run a map/reduce job for estimating Pi. * * @return the estimated value of Pi// w ww. j a v a2 s .com */ public static BigDecimal estimate(int numMaps, long numPoints, JobConf jobConf) throws IOException { //setup job conf jobConf.setJobName(PiEstimatorKrb.class.getSimpleName()); jobConf.setInputFormat(SequenceFileInputFormat.class); jobConf.setOutputKeyClass(BooleanWritable.class); jobConf.setOutputValueClass(LongWritable.class); jobConf.setOutputFormat(SequenceFileOutputFormat.class); jobConf.setMapperClass(PiMapper.class); jobConf.setNumMapTasks(numMaps); jobConf.setReducerClass(PiReducer.class); jobConf.setNumReduceTasks(1); // turn off speculative execution, because DFS doesn't handle // multiple writers to the same file. jobConf.setSpeculativeExecution(false); //setup input/output directories final Path inDir = new Path(TMP_DIR, "in"); final Path outDir = new Path(TMP_DIR, "out"); FileInputFormat.setInputPaths(jobConf, inDir); FileOutputFormat.setOutputPath(jobConf, outDir); final FileSystem fs = FileSystem.get(jobConf); if (fs.exists(TMP_DIR)) { throw new IOException( "Tmp directory " + fs.makeQualified(TMP_DIR) + " already exists. Please remove it first."); } if (!fs.mkdirs(inDir)) { throw new IOException("Cannot create input directory " + inDir); } try { //generate an input file for each map task for (int i = 0; i < numMaps; ++i) { final Path file = new Path(inDir, "part" + i); final LongWritable offset = new LongWritable(i * numPoints); final LongWritable size = new LongWritable(numPoints); final SequenceFile.Writer writer = SequenceFile.createWriter(fs, jobConf, file, LongWritable.class, LongWritable.class, CompressionType.NONE); try { writer.append(offset, size); } finally { writer.close(); } sLogger.info("Wrote input for Map #" + i); } //start a map/reduce job sLogger.info("Starting Job"); final long startTime = System.currentTimeMillis(); if (System.getenv("HADOOP_TOKEN_FILE_LOCATION") != null) { jobConf.set("mapreduce.job.credentials.binary", System.getenv("HADOOP_TOKEN_FILE_LOCATION")); } JobClient.runJob(jobConf); final double duration = (System.currentTimeMillis() - startTime) / 1000.0; sLogger.info("Job Finished in " + duration + " seconds"); //read outputs Path inFile = new Path(outDir, "reduce-out"); LongWritable numInside = new LongWritable(); LongWritable numOutside = new LongWritable(); SequenceFile.Reader reader = new SequenceFile.Reader(fs, inFile, jobConf); try { reader.next(numInside, numOutside); } finally { reader.close(); } //compute estimated value return BigDecimal.valueOf(4).setScale(20).multiply(BigDecimal.valueOf(numInside.get())) .divide(BigDecimal.valueOf(numMaps)).divide(BigDecimal.valueOf(numPoints)); } finally { fs.delete(TMP_DIR, true); } }
From source file:com.unstruct.demo.WordCount.java
License:Apache License
/** * The main driver for word count map/reduce program. * Invoke this method to submit the map/reduce job. * @throws IOException When there is communication problems with the * job tracker./*w w w . ja v a2 s. com*/ */ public int run(String[] args) throws Exception { JobConf conf = new JobConf(getConf(), WordCount.class); conf.setJobName("wordcount"); // the keys are words (strings) conf.setOutputKeyClass(Text.class); // the values are counts (ints) conf.setOutputValueClass(IntWritable.class); conf.setMapperClass(MapClass.class); conf.setCombinerClass(Reduce.class); conf.setReducerClass(Reduce.class); conf.setInputFormat(TextInputFormat.class); conf.setOutputFormat(TextOutputFormat.class); List<String> other_args = new ArrayList<String>(); for (int i = 0; i < args.length; ++i) { try { if ("-m".equals(args[i])) { conf.setNumMapTasks(Integer.parseInt(args[++i])); } else if ("-r".equals(args[i])) { conf.setNumReduceTasks(Integer.parseInt(args[++i])); } else { other_args.add(args[i]); } } catch (NumberFormatException except) { System.out.println("ERROR: Integer expected instead of " + args[i]); return printUsage(); } catch (ArrayIndexOutOfBoundsException except) { System.out.println("ERROR: Required parameter missing from " + args[i - 1]); return printUsage(); } } // Make sure there are exactly 2 parameters left. if (other_args.size() != 2) { System.out.println("ERROR: Wrong number of parameters: " + other_args.size() + " instead of 2."); return printUsage(); } FileInputFormat.setInputPaths(conf, other_args.get(0)); FileOutputFormat.setOutputPath(conf, new Path(other_args.get(1))); JobClient.runJob(conf); return 0; }
From source file:Corrector.Config.java
License:Apache License
public static void initializeConfiguration(JobConf conf) { validateConfiguration();/* w ww . j a v a 2 s .c o m*/ conf.setNumMapTasks(HADOOP_MAPPERS); conf.setNumReduceTasks(HADOOP_REDUCERS); conf.set("mapred.child.java.opts", HADOOP_JAVAOPTS); conf.set("mapred.task.timeout", Long.toString(HADOOP_TIMEOUT)); conf.setLong("LOCALNODES", HADOOP_LOCALNODES); conf.setLong("RANDOM_PASS", RANDOM_PASS); conf.setLong("UP_KMER", UP_KMER); conf.setLong("LOW_KMER", LOW_KMER); conf.setLong("K", K); conf.setLong("READLENGTH", READLEN); }
From source file:crunch.MaxTemperature.java
License:Apache License
@Override
public int run(String[] args) throws IOException {
JobConf conf = JobBuilder.parseInputAndOutput(this, getConf(), args);
if (conf == null) {
return -1;
}// ww w.jav a 2 s . co m
/*[*/conf.setInputFormat(TextInputFormat.class);
conf.setNumMapTasks(1);
conf.setMapperClass(IdentityMapper.class);
conf.setMapRunnerClass(MapRunner.class);
conf.setMapOutputKeyClass(LongWritable.class);
conf.setMapOutputValueClass(Text.class);
conf.setPartitionerClass(HashPartitioner.class);
conf.setNumReduceTasks(1);
conf.setReducerClass(IdentityReducer.class);
conf.setOutputKeyClass(LongWritable.class);
conf.setOutputValueClass(Text.class);
conf.setOutputFormat(TextOutputFormat.class);/*]*/
JobClient.runJob(conf);
return 0;
}
From source file:de.l3s.streamcorpus.mapreduce.TerrierIndexing.java
License:Mozilla Public License
/** Starts the MapReduce indexing. * @param args// ww w.j a v a2 s . c o m * @throws Exception */ public int run(String[] args) throws Exception { long time = System.currentTimeMillis(); // For the moment: Hard-code the terrier home to quick test System.setProperty("terrier.home", "/home/tuan.tran/executable/StreamCorpusIndexer"); boolean docPartitioned = false; int numberOfReducers = Integer .parseInt(ApplicationSetup.getProperty("terrier.hadoop.indexing.reducers", "26")); final HadoopPlugin.JobFactory jf = HadoopPlugin.getJobFactory("HOD-TerrierIndexing"); if (args.length == 2 && args[0].equals("-p")) { logger.debug("Document-partitioned Mode, " + numberOfReducers + " output indices."); numberOfReducers = Integer.parseInt(args[1]); docPartitioned = true; } else if (args.length == 1 && args[0].equals("--merge")) { if (numberOfReducers > 1) mergeLexiconInvertedFiles(ApplicationSetup.TERRIER_INDEX_PATH, numberOfReducers); else logger.error("No point merging 1 reduce task output"); return 0; } else if (args.length == 0) { logger.debug("Term-partitioned Mode, " + numberOfReducers + " reducers creating one inverted index."); docPartitioned = false; if (numberOfReducers > MAX_REDUCE) { logger.warn("Excessive reduce tasks (" + numberOfReducers + ") in use " + "- SplitEmittedTerm.SETPartitionerLowercaseAlphaTerm can use " + MAX_REDUCE + " at most"); } } /*else { logger.fatal(usage()); return 0; }*/ if (!(CompressionFactory.getCompressionConfiguration("inverted", new String[0], false) instanceof BitCompressionConfiguration)) { logger.error("Sorry, only default BitCompressionConfiguration is supported by HadoopIndexing" + " - you can recompress the inverted index later using IndexRecompressor"); return 0; } if (jf == null) throw new Exception("Could not get JobFactory from HadoopPlugin"); final JobConf conf = jf.newJob(); conf.setJarByClass(TerrierIndexing.class); conf.setJobName("StreamCorpusIndexer: Terrier Indexing"); if (Files.exists(ApplicationSetup.TERRIER_INDEX_PATH) && Index.existsIndex(ApplicationSetup.TERRIER_INDEX_PATH, ApplicationSetup.TERRIER_INDEX_PREFIX)) { logger.fatal("Cannot index while index exists at " + ApplicationSetup.TERRIER_INDEX_PATH + "," + ApplicationSetup.TERRIER_INDEX_PREFIX); return 0; } // boolean blockIndexing = ApplicationSetup.BLOCK_INDEXING; boolean blockIndexing = true; if (blockIndexing) { conf.setMapperClass(Hadoop_BlockSinglePassIndexer.class); conf.setReducerClass(Hadoop_BlockSinglePassIndexer.class); } else { conf.setMapperClass(Hadoop_BasicSinglePassIndexer.class); conf.setReducerClass(Hadoop_BasicSinglePassIndexer.class); } FileOutputFormat.setOutputPath(conf, new Path(ApplicationSetup.TERRIER_INDEX_PATH)); conf.set("indexing.hadoop.prefix", ApplicationSetup.TERRIER_INDEX_PREFIX); conf.setMapOutputKeyClass(SplitEmittedTerm.class); conf.setMapOutputValueClass(MapEmittedPostingList.class); conf.setBoolean("indexing.hadoop.multiple.indices", docPartitioned); if (!conf.get("mapred.job.tracker").equals("local")) { conf.setMapOutputCompressorClass(GzipCodec.class); conf.setCompressMapOutput(true); } else { conf.setCompressMapOutput(false); } conf.setInputFormat(MultiFileCollectionInputFormat.class); conf.setOutputFormat(NullOutputFormat.class); conf.setOutputKeyComparatorClass(SplitEmittedTerm.SETRawComparatorTermSplitFlush.class); conf.setOutputValueGroupingComparator(SplitEmittedTerm.SETRawComparatorTerm.class); conf.setReduceSpeculativeExecution(false); //parse the collection.spec BufferedReader specBR = Files.openFileReader(ApplicationSetup.COLLECTION_SPEC); String line = null; List<Path> paths = new ArrayList<Path>(); while ((line = specBR.readLine()) != null) { if (line.startsWith("#")) continue; paths.add(new Path(line)); } specBR.close(); FileInputFormat.setInputPaths(conf, paths.toArray(new Path[paths.size()])); // not sure if this is effective in YARN conf.setNumMapTasks(2000); // increase the heap usage conf.set("mapreduce.map.memory.mb", "6100"); conf.set("mapred.job.map.memory.mb", "6100"); conf.set("mapreduce.reduce.memory.mb", "6144"); conf.set("mapred.job.reduce.memory.mb", "6144"); conf.set("mapreduce.map.java.opts", "-Xmx6100m"); conf.set("mapred.map.child.java.opts", "-Xmx6100m"); conf.set("mapreduce.reduce.java.opts", "-Xmx6144m"); conf.set("mapred.reduce.child.opts", "-Xmx6144m"); //conf.setBoolean("mapred.used.genericoptionsparser", true) ; // This is the nasty thing in MapReduce v2 and YARN: They always prefer their ancient jars first. Set this on to say you don't like it conf.set("mapreduce.job.user.classpath.first", "true"); // increase the yarn memory to 10 GB conf.set("yarn.nodemanager.resource.memory-mb", "12288"); conf.set("yarn.nodemanager.resource.cpu-vcores", "16"); conf.set("yarn.scheduler.minimum-allocation-mb", "4096"); conf.setNumReduceTasks(numberOfReducers); if (numberOfReducers > 1) { if (docPartitioned) conf.setPartitionerClass(SplitEmittedTerm.SETPartitioner.class); else conf.setPartitionerClass(SplitEmittedTerm.SETPartitionerLowercaseAlphaTerm.class); } else { //for JUnit tests, we seem to need to restore the original partitioner class conf.setPartitionerClass(HashPartitioner.class); } /*JobID jobId = null; boolean ranOK = true; try{ RunningJob rj = JobClient.runJob(conf); jobId = rj.getID(); HadoopUtility.finishTerrierJob(conf); } catch (Exception e) { logger.error("Problem running job", e); e.printStackTrace(); ranOK = false; } if (jobId != null) { deleteTaskFiles(ApplicationSetup.TERRIER_INDEX_PATH, jobId); } */ //if (ranOK) //{ System.out.println("Merging indices"); if (!docPartitioned) { if (numberOfReducers > 1) mergeLexiconInvertedFiles(ApplicationSetup.TERRIER_INDEX_PATH, numberOfReducers); } Hadoop_BasicSinglePassIndexer.finish(ApplicationSetup.TERRIER_INDEX_PATH, docPartitioned ? numberOfReducers : 1, jf); //} System.out.println("Time Taken = " + ((System.currentTimeMillis() - time) / 1000) + " seconds"); jf.close(); return 0; }