Examples with SplitsInClientOptionParser org.apache.tez.mapreduce.examples.helpers.SplitsInClientOptionParser used on opensource projects

Example 1 with SplitsInClientOptionParser

use of org.apache.tez.mapreduce.examples.helpers.SplitsInClientOptionParser in project tez by apache.

the class TestOrderedWordCount method run.

@Override
public int run(String[] args) throws Exception {
    Configuration conf = getConf();
    String[] otherArgs = new GenericOptionsParser(conf, args).getRemainingArgs();
    boolean generateSplitsInClient;
    SplitsInClientOptionParser splitCmdLineParser = new SplitsInClientOptionParser();
    try {
        generateSplitsInClient = splitCmdLineParser.parse(otherArgs, false);
        otherArgs = splitCmdLineParser.getRemainingArgs();
    } catch (ParseException e1) {
        System.err.println("Invalid options");
        printUsage();
        return 2;
    }
    boolean useTezSession = conf.getBoolean("USE_TEZ_SESSION", true);
    long interJobSleepTimeout = conf.getInt("INTER_JOB_SLEEP_INTERVAL", 0) * 1000;
    boolean retainStagingDir = conf.getBoolean("RETAIN_STAGING_DIR", false);
    boolean useMRSettings = conf.getBoolean("USE_MR_CONFIGS", true);
    // TODO needs to use auto reduce parallelism
    int intermediateNumReduceTasks = conf.getInt("IREDUCE_NUM_TASKS", 2);
    int maxDataLengthThroughIPC = conf.getInt(MAX_IPC_DATA_LENGTH, -1);
    int exceedDataLimit = conf.getInt(EXCEED_IPC_DATA_LIMIT, 3);
    if (maxDataLengthThroughIPC > 0) {
        conf.setInt(CommonConfigurationKeys.IPC_MAXIMUM_DATA_LENGTH, maxDataLengthThroughIPC * 1024 * 1024);
    }
    if (((otherArgs.length % 2) != 0) || (!useTezSession && otherArgs.length != 2)) {
        printUsage();
        return 2;
    }
    List<String> inputPaths = new ArrayList<String>();
    List<String> outputPaths = new ArrayList<String>();
    TezConfiguration tezConf = new TezConfiguration(conf);
    for (int i = 0; i < otherArgs.length; i += 2) {
        FileSystem inputPathFs = new Path(otherArgs[i]).getFileSystem(tezConf);
        inputPaths.add(inputPathFs.makeQualified(new Path(otherArgs[i])).toString());
        FileSystem outputPathFs = new Path(otherArgs[i + 1]).getFileSystem(tezConf);
        outputPaths.add(outputPathFs.makeQualified(new Path(otherArgs[i + 1])).toString());
    }
    UserGroupInformation.setConfiguration(conf);
    HadoopShim hadoopShim = new HadoopShimsLoader(tezConf).getHadoopShim();
    TestOrderedWordCount instance = new TestOrderedWordCount();
    FileSystem fs = FileSystem.get(conf);
    String stagingDirStr = conf.get(TezConfiguration.TEZ_AM_STAGING_DIR, TezConfiguration.TEZ_AM_STAGING_DIR_DEFAULT) + Path.SEPARATOR + Long.toString(System.currentTimeMillis());
    Path stagingDir = new Path(stagingDirStr);
    FileSystem pathFs = stagingDir.getFileSystem(tezConf);
    pathFs.mkdirs(new Path(stagingDirStr));
    tezConf.set(TezConfiguration.TEZ_AM_STAGING_DIR, stagingDirStr);
    stagingDir = pathFs.makeQualified(new Path(stagingDirStr));
    TokenCache.obtainTokensForNamenodes(instance.credentials, new Path[] { stagingDir }, conf);
    TezClientUtils.ensureStagingDirExists(tezConf, stagingDir);
    if (useTezSession) {
        LOG.info("Creating Tez Session");
        tezConf.setBoolean(TezConfiguration.TEZ_AM_SESSION_MODE, true);
    } else {
        tezConf.setBoolean(TezConfiguration.TEZ_AM_SESSION_MODE, false);
    }
    TezClient tezSession = TezClient.create("OrderedWordCountSession", tezConf, null, instance.credentials);
    tezSession.start();
    if (tezSession.getAppMasterApplicationId() != null) {
        TezUtilsInternal.setHadoopCallerContext(hadoopShim, tezSession.getAppMasterApplicationId());
    }
    DAGStatus dagStatus = null;
    DAGClient dagClient = null;
    String[] vNames = { "initialmap", "intermediate_reducer", "finalreduce" };
    Set<StatusGetOpts> statusGetOpts = EnumSet.of(StatusGetOpts.GET_COUNTERS);
    try {
        for (int dagIndex = 1; dagIndex <= inputPaths.size(); ++dagIndex) {
            if (dagIndex != 1 && interJobSleepTimeout > 0) {
                try {
                    LOG.info("Sleeping between jobs, sleepInterval=" + (interJobSleepTimeout / 1000));
                    Thread.sleep(interJobSleepTimeout);
                } catch (InterruptedException e) {
                    LOG.info("Main thread interrupted. Breaking out of job loop");
                    break;
                }
            }
            String inputPath = inputPaths.get(dagIndex - 1);
            String outputPath = outputPaths.get(dagIndex - 1);
            if (fs.exists(new Path(outputPath))) {
                throw new FileAlreadyExistsException("Output directory " + outputPath + " already exists");
            }
            LOG.info("Running OrderedWordCount DAG" + ", dagIndex=" + dagIndex + ", inputPath=" + inputPath + ", outputPath=" + outputPath);
            Map<String, LocalResource> localResources = new TreeMap<String, LocalResource>();
            DAG dag = instance.createDAG(fs, tezConf, localResources, stagingDir, dagIndex, inputPath, outputPath, generateSplitsInClient, useMRSettings, intermediateNumReduceTasks, maxDataLengthThroughIPC, exceedDataLimit);
            String callerType = "TestOrderedWordCount";
            String callerId = tezSession.getAppMasterApplicationId() == null ? ("UnknownApp_" + System.currentTimeMillis() + dagIndex) : (tezSession.getAppMasterApplicationId().toString() + "_" + dagIndex);
            dag.setCallerContext(CallerContext.create("Tez", callerId, callerType, "TestOrderedWordCount Job"));
            boolean doPreWarm = dagIndex == 1 && useTezSession && conf.getBoolean("PRE_WARM_SESSION", true);
            int preWarmNumContainers = 0;
            if (doPreWarm) {
                preWarmNumContainers = conf.getInt("PRE_WARM_NUM_CONTAINERS", 0);
                if (preWarmNumContainers <= 0) {
                    doPreWarm = false;
                }
            }
            if (doPreWarm) {
                LOG.info("Pre-warming Session");
                PreWarmVertex preWarmVertex = PreWarmVertex.create("PreWarm", preWarmNumContainers, dag.getVertex("initialmap").getTaskResource());
                preWarmVertex.addTaskLocalFiles(dag.getVertex("initialmap").getTaskLocalFiles());
                preWarmVertex.setTaskEnvironment(dag.getVertex("initialmap").getTaskEnvironment());
                preWarmVertex.setTaskLaunchCmdOpts(dag.getVertex("initialmap").getTaskLaunchCmdOpts());
                tezSession.preWarm(preWarmVertex);
            }
            if (useTezSession) {
                LOG.info("Waiting for TezSession to get into ready state");
                waitForTezSessionReady(tezSession);
                LOG.info("Submitting DAG to Tez Session, dagIndex=" + dagIndex);
                dagClient = tezSession.submitDAG(dag);
                LOG.info("Submitted DAG to Tez Session, dagIndex=" + dagIndex);
            } else {
                LOG.info("Submitting DAG as a new Tez Application");
                dagClient = tezSession.submitDAG(dag);
            }
            while (true) {
                dagStatus = dagClient.getDAGStatus(statusGetOpts);
                if (dagStatus.getState() == DAGStatus.State.RUNNING || dagStatus.getState() == DAGStatus.State.SUCCEEDED || dagStatus.getState() == DAGStatus.State.FAILED || dagStatus.getState() == DAGStatus.State.KILLED || dagStatus.getState() == DAGStatus.State.ERROR) {
                    break;
                }
                try {
                    Thread.sleep(500);
                } catch (InterruptedException e) {
                // continue;
                }
            }
            while (dagStatus.getState() != DAGStatus.State.SUCCEEDED && dagStatus.getState() != DAGStatus.State.FAILED && dagStatus.getState() != DAGStatus.State.KILLED && dagStatus.getState() != DAGStatus.State.ERROR) {
                if (dagStatus.getState() == DAGStatus.State.RUNNING) {
                    ExampleDriver.printDAGStatus(dagClient, vNames);
                }
                try {
                    try {
                        Thread.sleep(1000);
                    } catch (InterruptedException e) {
                    // continue;
                    }
                    dagStatus = dagClient.getDAGStatus(statusGetOpts);
                } catch (TezException e) {
                    LOG.error("Failed to get application progress. Exiting");
                    return -1;
                }
            }
            ExampleDriver.printDAGStatus(dagClient, vNames, true, true);
            LOG.info("DAG " + dagIndex + " completed. " + "FinalState=" + dagStatus.getState());
            if (dagStatus.getState() != DAGStatus.State.SUCCEEDED) {
                LOG.info("DAG " + dagIndex + " diagnostics: " + dagStatus.getDiagnostics());
            }
        }
    } catch (Exception e) {
        LOG.error("Error occurred when submitting/running DAGs", e);
        throw e;
    } finally {
        if (!retainStagingDir) {
            pathFs.delete(stagingDir, true);
        }
        LOG.info("Shutting down session");
        tezSession.stop();
    }
    if (!useTezSession) {
        ExampleDriver.printDAGStatus(dagClient, vNames);
        LOG.info("Application completed. " + "FinalState=" + dagStatus.getState());
    }
    return dagStatus.getState() == DAGStatus.State.SUCCEEDED ? 0 : 1;
}

Example 2 with SplitsInClientOptionParser

use of org.apache.tez.mapreduce.examples.helpers.SplitsInClientOptionParser in project tez by apache.

the class FilterLinesByWordOneToOne method run.

@Override
public int run(String[] otherArgs) throws Exception {
    boolean generateSplitsInClient = false;
    SplitsInClientOptionParser splitCmdLineParser = new SplitsInClientOptionParser();
    try {
        generateSplitsInClient = splitCmdLineParser.parse(otherArgs, false);
        otherArgs = splitCmdLineParser.getRemainingArgs();
    } catch (ParseException e1) {
        System.err.println("Invalid options");
        printUsage();
        return 2;
    }
    if (otherArgs.length != 3) {
        printUsage();
        return 2;
    }
    String inputPath = otherArgs[0];
    String outputPath = otherArgs[1];
    String filterWord = otherArgs[2];
    Configuration conf = getConf();
    FileSystem fs = FileSystem.get(conf);
    if (fs.exists(new Path(outputPath))) {
        System.err.println("Output directory : " + outputPath + " already exists");
        return 2;
    }
    TezConfiguration tezConf = new TezConfiguration(conf);
    fs.getWorkingDirectory();
    Path stagingDir = new Path(fs.getWorkingDirectory(), UUID.randomUUID().toString());
    tezConf.set(TezConfiguration.TEZ_AM_STAGING_DIR, stagingDir.toString());
    TezClientUtils.ensureStagingDirExists(tezConf, stagingDir);
    String jarPath = ClassUtil.findContainingJar(FilterLinesByWordOneToOne.class);
    if (jarPath == null) {
        throw new TezUncheckedException("Could not find any jar containing" + FilterLinesByWordOneToOne.class.getName() + " in the classpath");
    }
    Path remoteJarPath = fs.makeQualified(new Path(stagingDir, "dag_job.jar"));
    fs.copyFromLocalFile(new Path(jarPath), remoteJarPath);
    FileStatus remoteJarStatus = fs.getFileStatus(remoteJarPath);
    Map<String, LocalResource> commonLocalResources = new TreeMap<String, LocalResource>();
    LocalResource dagJarLocalRsrc = LocalResource.newInstance(ConverterUtils.getYarnUrlFromPath(remoteJarPath), LocalResourceType.FILE, LocalResourceVisibility.APPLICATION, remoteJarStatus.getLen(), remoteJarStatus.getModificationTime());
    commonLocalResources.put("dag_job.jar", dagJarLocalRsrc);
    TezClient tezSession = TezClient.create("FilterLinesByWordSession", tezConf, commonLocalResources, null);
    // Why do I need to start the TezSession.
    tezSession.start();
    Configuration stage1Conf = new JobConf(conf);
    stage1Conf.set(FILTER_PARAM_NAME, filterWord);
    Configuration stage2Conf = new JobConf(conf);
    stage2Conf.set(FileOutputFormat.OUTDIR, outputPath);
    stage2Conf.setBoolean("mapred.mapper.new-api", false);
    UserPayload stage1Payload = TezUtils.createUserPayloadFromConf(stage1Conf);
    // Setup stage1 Vertex
    Vertex stage1Vertex = Vertex.create("stage1", ProcessorDescriptor.create(FilterByWordInputProcessor.class.getName()).setUserPayload(stage1Payload)).addTaskLocalFiles(commonLocalResources);
    DataSourceDescriptor dsd;
    if (generateSplitsInClient) {
        // TODO TEZ-1406. Dont' use MRInputLegacy
        stage1Conf.set(FileInputFormat.INPUT_DIR, inputPath);
        stage1Conf.setBoolean("mapred.mapper.new-api", false);
        dsd = MRInputHelpers.configureMRInputWithLegacySplitGeneration(stage1Conf, stagingDir, true);
    } else {
        dsd = MRInputLegacy.createConfigBuilder(stage1Conf, TextInputFormat.class, inputPath).groupSplits(false).build();
    }
    stage1Vertex.addDataSource("MRInput", dsd);
    // Setup stage2 Vertex
    Vertex stage2Vertex = Vertex.create("stage2", ProcessorDescriptor.create(FilterByWordOutputProcessor.class.getName()).setUserPayload(TezUtils.createUserPayloadFromConf(stage2Conf)), dsd.getNumberOfShards());
    stage2Vertex.addTaskLocalFiles(commonLocalResources);
    // Configure the Output for stage2
    stage2Vertex.addDataSink("MROutput", DataSinkDescriptor.create(OutputDescriptor.create(MROutput.class.getName()).setUserPayload(TezUtils.createUserPayloadFromConf(stage2Conf)), OutputCommitterDescriptor.create(MROutputCommitter.class.getName()), null));
    UnorderedKVEdgeConfig edgeConf = UnorderedKVEdgeConfig.newBuilder(Text.class.getName(), TextLongPair.class.getName()).setFromConfiguration(tezConf).build();
    DAG dag = DAG.create("FilterLinesByWord");
    Edge edge = Edge.create(stage1Vertex, stage2Vertex, edgeConf.createDefaultOneToOneEdgeProperty());
    dag.addVertex(stage1Vertex).addVertex(stage2Vertex).addEdge(edge);
    LOG.info("Submitting DAG to Tez Session");
    DAGClient dagClient = tezSession.submitDAG(dag);
    LOG.info("Submitted DAG to Tez Session");
    DAGStatus dagStatus = null;
    String[] vNames = { "stage1", "stage2" };
    try {
        while (true) {
            dagStatus = dagClient.getDAGStatus(null);
            if (dagStatus.getState() == DAGStatus.State.RUNNING || dagStatus.getState() == DAGStatus.State.SUCCEEDED || dagStatus.getState() == DAGStatus.State.FAILED || dagStatus.getState() == DAGStatus.State.KILLED || dagStatus.getState() == DAGStatus.State.ERROR) {
                break;
            }
            try {
                Thread.sleep(500);
            } catch (InterruptedException e) {
            // continue;
            }
        }
        while (dagStatus.getState() == DAGStatus.State.RUNNING) {
            try {
                ExampleDriver.printDAGStatus(dagClient, vNames);
                try {
                    Thread.sleep(1000);
                } catch (InterruptedException e) {
                // continue;
                }
                dagStatus = dagClient.getDAGStatus(null);
            } catch (TezException e) {
                LOG.error("Failed to get application progress. Exiting");
                return -1;
            }
        }
    } finally {
        fs.delete(stagingDir, true);
        tezSession.stop();
    }
    ExampleDriver.printDAGStatus(dagClient, vNames);
    LOG.info("Application completed. " + "FinalState=" + dagStatus.getState());
    return dagStatus.getState() == DAGStatus.State.SUCCEEDED ? 0 : 1;
}

Example 3 with SplitsInClientOptionParser

use of org.apache.tez.mapreduce.examples.helpers.SplitsInClientOptionParser in project tez by apache.

the class FilterLinesByWord method run.

@Override
public int run(String[] args) throws Exception {
    Configuration conf = getConf();
    String[] otherArgs = new GenericOptionsParser(conf, args).getRemainingArgs();
    Credentials credentials = new Credentials();
    boolean generateSplitsInClient = false;
    SplitsInClientOptionParser splitCmdLineParser = new SplitsInClientOptionParser();
    try {
        generateSplitsInClient = splitCmdLineParser.parse(otherArgs, false);
        otherArgs = splitCmdLineParser.getRemainingArgs();
    } catch (ParseException e1) {
        System.err.println("Invalid options");
        printUsage();
        return 2;
    }
    if (otherArgs.length != 3) {
        printUsage();
        return 2;
    }
    String inputPath = otherArgs[0];
    String outputPath = otherArgs[1];
    String filterWord = otherArgs[2];
    FileSystem fs = FileSystem.get(conf);
    if (fs.exists(new Path(outputPath))) {
        System.err.println("Output directory : " + outputPath + " already exists");
        return 2;
    }
    TezConfiguration tezConf = new TezConfiguration(conf);
    fs.getWorkingDirectory();
    Path stagingDir = new Path(fs.getWorkingDirectory(), UUID.randomUUID().toString());
    tezConf.set(TezConfiguration.TEZ_AM_STAGING_DIR, stagingDir.toString());
    TezClientUtils.ensureStagingDirExists(tezConf, stagingDir);
    String jarPath = ClassUtil.findContainingJar(FilterLinesByWord.class);
    if (jarPath == null) {
        throw new TezUncheckedException("Could not find any jar containing" + FilterLinesByWord.class.getName() + " in the classpath");
    }
    Path remoteJarPath = fs.makeQualified(new Path(stagingDir, "dag_job.jar"));
    fs.copyFromLocalFile(new Path(jarPath), remoteJarPath);
    FileStatus remoteJarStatus = fs.getFileStatus(remoteJarPath);
    TokenCache.obtainTokensForNamenodes(credentials, new Path[] { remoteJarPath }, conf);
    Map<String, LocalResource> commonLocalResources = new TreeMap<String, LocalResource>();
    LocalResource dagJarLocalRsrc = LocalResource.newInstance(ConverterUtils.getYarnUrlFromPath(remoteJarPath), LocalResourceType.FILE, LocalResourceVisibility.APPLICATION, remoteJarStatus.getLen(), remoteJarStatus.getModificationTime());
    commonLocalResources.put("dag_job.jar", dagJarLocalRsrc);
    TezClient tezSession = TezClient.create("FilterLinesByWordSession", tezConf, commonLocalResources, credentials);
    // Why do I need to start the TezSession.
    tezSession.start();
    Configuration stage1Conf = new JobConf(conf);
    stage1Conf.set(FILTER_PARAM_NAME, filterWord);
    Configuration stage2Conf = new JobConf(conf);
    stage2Conf.set(FileOutputFormat.OUTDIR, outputPath);
    stage2Conf.setBoolean("mapred.mapper.new-api", false);
    UserPayload stage1Payload = TezUtils.createUserPayloadFromConf(stage1Conf);
    // Setup stage1 Vertex
    Vertex stage1Vertex = Vertex.create("stage1", ProcessorDescriptor.create(FilterByWordInputProcessor.class.getName()).setUserPayload(stage1Payload)).addTaskLocalFiles(commonLocalResources);
    DataSourceDescriptor dsd;
    if (generateSplitsInClient) {
        // TODO TEZ-1406. Dont' use MRInputLegacy
        stage1Conf.set(FileInputFormat.INPUT_DIR, inputPath);
        stage1Conf.setBoolean("mapred.mapper.new-api", false);
        dsd = MRInputHelpers.configureMRInputWithLegacySplitGeneration(stage1Conf, stagingDir, true);
    } else {
        dsd = MRInputLegacy.createConfigBuilder(stage1Conf, TextInputFormat.class, inputPath).groupSplits(false).build();
    }
    stage1Vertex.addDataSource("MRInput", dsd);
    // Setup stage2 Vertex
    Vertex stage2Vertex = Vertex.create("stage2", ProcessorDescriptor.create(FilterByWordOutputProcessor.class.getName()).setUserPayload(TezUtils.createUserPayloadFromConf(stage2Conf)), 1);
    stage2Vertex.addTaskLocalFiles(commonLocalResources);
    // Configure the Output for stage2
    OutputDescriptor od = OutputDescriptor.create(MROutput.class.getName()).setUserPayload(TezUtils.createUserPayloadFromConf(stage2Conf));
    OutputCommitterDescriptor ocd = OutputCommitterDescriptor.create(MROutputCommitter.class.getName());
    stage2Vertex.addDataSink("MROutput", DataSinkDescriptor.create(od, ocd, null));
    UnorderedKVEdgeConfig edgeConf = UnorderedKVEdgeConfig.newBuilder(Text.class.getName(), TextLongPair.class.getName()).setFromConfiguration(tezConf).build();
    DAG dag = DAG.create("FilterLinesByWord");
    Edge edge = Edge.create(stage1Vertex, stage2Vertex, edgeConf.createDefaultBroadcastEdgeProperty());
    dag.addVertex(stage1Vertex).addVertex(stage2Vertex).addEdge(edge);
    LOG.info("Submitting DAG to Tez Session");
    DAGClient dagClient = tezSession.submitDAG(dag);
    LOG.info("Submitted DAG to Tez Session");
    DAGStatus dagStatus = null;
    String[] vNames = { "stage1", "stage2" };
    try {
        while (true) {
            dagStatus = dagClient.getDAGStatus(null);
            if (dagStatus.getState() == DAGStatus.State.RUNNING || dagStatus.getState() == DAGStatus.State.SUCCEEDED || dagStatus.getState() == DAGStatus.State.FAILED || dagStatus.getState() == DAGStatus.State.KILLED || dagStatus.getState() == DAGStatus.State.ERROR) {
                break;
            }
            try {
                Thread.sleep(500);
            } catch (InterruptedException e) {
            // continue;
            }
        }
        while (dagStatus.getState() == DAGStatus.State.RUNNING) {
            try {
                ExampleDriver.printDAGStatus(dagClient, vNames);
                try {
                    Thread.sleep(1000);
                } catch (InterruptedException e) {
                // continue;
                }
                dagStatus = dagClient.getDAGStatus(null);
            } catch (TezException e) {
                LOG.error("Failed to get application progress. Exiting");
                return -1;
            }
        }
        dagStatus = dagClient.getDAGStatus(Sets.newHashSet(StatusGetOpts.GET_COUNTERS));
    } finally {
        fs.delete(stagingDir, true);
        tezSession.stop();
    }
    ExampleDriver.printDAGStatus(dagClient, vNames, true, true);
    LOG.info("Application completed. " + "FinalState=" + dagStatus.getState());
    return dagStatus.getState() == DAGStatus.State.SUCCEEDED ? 0 : 1;
}