Example usage for com.amazonaws.services.elasticmapreduce.util StepFactory StepFactory

Introduction

In this page you can find the example usage for com.amazonaws.services.elasticmapreduce.util StepFactory StepFactory.

Prototype

public StepFactory(String bucket) 

Source Link

Document

Creates a new StepFactory using the specified Amazon S3 bucket to load resources.

Usage

From source file:org.deeplearning4j.legacyExamples.EmrSparkExample.java

License:Apache License

public void entryPoint(String[] args) {
    JCommander jcmdr = new JCommander(this);
    try {//from w w w. j a v a2s.  co m
        jcmdr.parse(args);
    } catch (ParameterException e) {
        jcmdr.usage();
        try {
            Thread.sleep(500);
        } catch (Exception e2) {
        }
        throw e;
    }

    AmazonElasticMapReduceClientBuilder builder = AmazonElasticMapReduceClientBuilder.standard();
    builder.withRegion(region);
    builder.withCredentials(getCredentialsProvider());

    AmazonElasticMapReduce emr = builder.build();

    List<StepConfig> steps = new ArrayList<>();

    if (upload) {
        log.info("uploading uber jar");

        AmazonS3ClientBuilder s3builder = AmazonS3ClientBuilder.standard();
        s3builder.withRegion(region);
        s3builder.withCredentials(getCredentialsProvider());
        AmazonS3 s3Client = s3builder.build();

        if (!s3Client.doesBucketExist(bucketName)) {
            s3Client.createBucket(bucketName);
        }

        File uberJarFile = new File(uberJar);

        s3Client.putObject(new PutObjectRequest(bucketName, uberJarFile.getName(), uberJarFile));
    }

    if (debug) {
        log.info("enable debug");

        StepFactory stepFactory = new StepFactory(builder.getRegion() + ".elasticmapreduce");
        StepConfig enableDebugging = new StepConfig().withName("Enable Debugging")
                .withActionOnFailure(ActionOnFailure.TERMINATE_JOB_FLOW)
                .withHadoopJarStep(stepFactory.newEnableDebuggingStep());
        steps.add(enableDebugging);
    }

    if (execute) {
        log.info("execute spark step");

        HadoopJarStepConfig sparkStepConf = new HadoopJarStepConfig();
        sparkStepConf.withJar("command-runner.jar");
        sparkStepConf.withArgs("spark-submit", "--deploy-mode", "cluster", "--class", className,
                getS3UberJarUrl(), "-useSparkLocal", "false");

        ActionOnFailure action = ActionOnFailure.TERMINATE_JOB_FLOW;

        if (keepAlive) {
            action = ActionOnFailure.CONTINUE;
        }

        StepConfig sparkStep = new StepConfig().withName("Spark Step").withActionOnFailure(action)
                .withHadoopJarStep(sparkStepConf);
        steps.add(sparkStep);
    }

    log.info("create spark cluster");

    Application sparkApp = new Application().withName("Spark");

    // service and job flow role will be created automatically when
    // launching cluster in aws console, better do that first or create
    // manually

    RunJobFlowRequest request = new RunJobFlowRequest().withName("Spark Cluster").withSteps(steps)
            .withServiceRole("EMR_DefaultRole").withJobFlowRole("EMR_EC2_DefaultRole")
            .withApplications(sparkApp).withReleaseLabel(emrVersion).withLogUri(getS3BucketLogsUrl())
            .withInstances(new JobFlowInstancesConfig().withEc2KeyName("spark").withInstanceCount(instanceCount)
                    .withKeepJobFlowAliveWhenNoSteps(keepAlive).withMasterInstanceType(instanceType)
                    .withSlaveInstanceType(instanceType));

    RunJobFlowResult result = emr.runJobFlow(request);

    log.info(result.toString());

    log.info("done");
}

---