com.amazonaws.services.elasticmapreduce.model.JobFlowInstancesConfig Java Examples

@Test
public void test()
        throws Exception
{
    RunJobFlowRequest request = new RunJobFlowRequest()
            .withName("Digdag Test")
            .withReleaseLabel("emr-5.2.0")
            .withApplications(Stream.of("Hadoop", "Hive", "Spark", "Flink")
                    .map(s -> new Application().withName(s))
                    .collect(toList()))
            .withJobFlowRole("EMR_EC2_DefaultRole")
            .withServiceRole("EMR_DefaultRole")
            .withVisibleToAllUsers(true)
            .withLogUri(tmpS3FolderUri + "/logs/")
            .withInstances(new JobFlowInstancesConfig()
                    .withEc2KeyName("digdag-test")
                    .withInstanceCount(1)
                    .withKeepJobFlowAliveWhenNoSteps(true)
                    .withMasterInstanceType("m3.xlarge")
                    .withSlaveInstanceType("m3.xlarge"));

    RunJobFlowResult result = emr.runJobFlow(request);

    String clusterId = result.getJobFlowId();

    clusterIds.add(clusterId);

    Id attemptId = pushAndStart(server.endpoint(), projectDir, "emr", ImmutableMap.of(
            "test_s3_folder", tmpS3FolderUri.toString(),
            "test_cluster", clusterId,
            "outfile", outfile.toString()));
    expect(Duration.ofMinutes(30), attemptSuccess(server.endpoint(), attemptId));

    validateTdSparkQueryOutput();

    assertThat(Files.exists(outfile), is(true));
}
 

@Override
public String createCluster(String clusterName) {
  RunJobFlowRequest request = new RunJobFlowRequest()
      .withName(clusterName)
      .withReleaseLabel(EmrInfo.getVersion())
      .withServiceRole(emrClusterConfig.getServiceRole())
      .withJobFlowRole(emrClusterConfig.getJobFlowRole())
      .withVisibleToAllUsers(emrClusterConfig.isVisibleToAllUsers())
      .withInstances(new JobFlowInstancesConfig()
          .withEc2SubnetId(emrClusterConfig.getEc2SubnetId())
          .withEmrManagedMasterSecurityGroup(emrClusterConfig.getMasterSecurityGroup())
          .withEmrManagedSlaveSecurityGroup(emrClusterConfig.getSlaveSecurityGroup())
          .withInstanceCount(emrClusterConfig.getInstanceCount())
          .withKeepJobFlowAliveWhenNoSteps(true)
          .withMasterInstanceType(emrClusterConfig.getMasterInstanceType())
          .withSlaveInstanceType(emrClusterConfig.getSlaveInstanceType()));

  if (emrClusterConfig.isLoggingEnabled()) {
    request.withLogUri(emrClusterConfig.getS3LogUri());
    if (emrClusterConfig.isEnableEmrDebugging()) {
      String COMMAND_RUNNER = "command-runner.jar";
      String DEBUGGING_COMMAND = "state-pusher-script";
      String DEBUGGING_NAME = "Setup Hadoop Debugging";
      StepConfig enabledebugging = new StepConfig()
          .withName(DEBUGGING_NAME)
          .withActionOnFailure(ActionOnFailure.CONTINUE)
          .withHadoopJarStep(new HadoopJarStepConfig()
              .withJar(COMMAND_RUNNER)
              .withArgs(DEBUGGING_COMMAND));
      request.withSteps(enabledebugging);
    }
  }
  RunJobFlowResult result = getEmrClient(emrClusterConfig).runJobFlow(request);
  return result.getJobFlowId();
}
 

/**
 * Creates the job flow instance configuration containing specification of the number and type of Amazon EC2 instances.
 *
 * @param emrClusterDefinition the EMR cluster definition that contains all the EMR parameters
 *
 * @return the job flow instance configuration
 */
private JobFlowInstancesConfig getJobFlowInstancesConfig(EmrClusterDefinition emrClusterDefinition)
{
    // Create a new job flow instances configuration object.
    JobFlowInstancesConfig jobFlowInstancesConfig = new JobFlowInstancesConfig();

    // Set up master/slave security group
    jobFlowInstancesConfig.setEmrManagedMasterSecurityGroup(emrClusterDefinition.getMasterSecurityGroup());
    jobFlowInstancesConfig.setEmrManagedSlaveSecurityGroup(emrClusterDefinition.getSlaveSecurityGroup());

    // Set up service access security group
    jobFlowInstancesConfig.setServiceAccessSecurityGroup(emrClusterDefinition.getServiceAccessSecurityGroup());

    // Add additional security groups to master nodes.
    jobFlowInstancesConfig.setAdditionalMasterSecurityGroups(emrClusterDefinition.getAdditionalMasterSecurityGroups());

    // Add additional security groups to slave nodes.
    jobFlowInstancesConfig.setAdditionalSlaveSecurityGroups(emrClusterDefinition.getAdditionalSlaveSecurityGroups());

    // Fill-in the ssh key.
    if (StringUtils.isNotBlank(emrClusterDefinition.getSshKeyPairName()))
    {
        jobFlowInstancesConfig.setEc2KeyName(emrClusterDefinition.getSshKeyPairName());
    }

    // Fill in configuration for the instance groups in a cluster.
    jobFlowInstancesConfig.setInstanceGroups(getInstanceGroupConfigs(emrClusterDefinition.getInstanceDefinitions()));

    // Fill in instance fleet configuration.
    jobFlowInstancesConfig.setInstanceFleets(getInstanceFleets(emrClusterDefinition.getInstanceFleets()));

    // Fill-in subnet id.
    if (StringUtils.isNotBlank(emrClusterDefinition.getSubnetId()))
    {
        // Use collection of subnet IDs when instance fleet configuration is specified. Otherwise, we expect a single EC2 subnet ID to be passed here.
        if (CollectionUtils.isNotEmpty(jobFlowInstancesConfig.getInstanceFleets()))
        {
            jobFlowInstancesConfig.setEc2SubnetIds(herdStringHelper.splitAndTrim(emrClusterDefinition.getSubnetId(), ","));
        }
        else
        {
            jobFlowInstancesConfig.setEc2SubnetId(emrClusterDefinition.getSubnetId());
        }
    }

    // Fill in optional keep alive flag.
    if (emrClusterDefinition.isKeepAlive() != null)
    {
        jobFlowInstancesConfig.setKeepJobFlowAliveWhenNoSteps(emrClusterDefinition.isKeepAlive());
    }

    // Fill in optional termination protection flag.
    if (emrClusterDefinition.isTerminationProtection() != null)
    {
        jobFlowInstancesConfig.setTerminationProtected(emrClusterDefinition.isTerminationProtection());
    }

    // Fill in optional Hadoop version flag.
    if (StringUtils.isNotBlank(emrClusterDefinition.getHadoopVersion()))
    {
        jobFlowInstancesConfig.setHadoopVersion(emrClusterDefinition.getHadoopVersion());
    }

    // Return the object.
    return jobFlowInstancesConfig;
}
 

@Test
public void createEmrClusterAssertCallRunEmrJobFlowWithInstanceFleetAndMultipleSubnets()
{
    // Create objects required for testing.
    final String clusterName = "clusterName";
    final String clusterId = "clusterId";
    final String name = STRING_VALUE;
    final String instanceFleetType = STRING_VALUE_2;
    final Integer targetOnDemandCapacity = INTEGER_VALUE;
    final Integer targetSpotCapacity = INTEGER_VALUE_2;
    final List<EmrClusterDefinitionInstanceTypeConfig> emrClusterDefinitionInstanceTypeConfigs = null;
    final EmrClusterDefinitionLaunchSpecifications emrClusterDefinitionLaunchSpecifications = null;
    final EmrClusterDefinitionInstanceFleet emrClusterDefinitionInstanceFleet =
        new EmrClusterDefinitionInstanceFleet(name, instanceFleetType, targetOnDemandCapacity, targetSpotCapacity, emrClusterDefinitionInstanceTypeConfigs,
            emrClusterDefinitionLaunchSpecifications);

    // Create an EMR cluster definition with instance fleet configuration and multiple EC2 subnet IDs.
    EmrClusterDefinition emrClusterDefinition = new EmrClusterDefinition();
    emrClusterDefinition.setInstanceFleets(Lists.newArrayList(emrClusterDefinitionInstanceFleet));
    emrClusterDefinition.setSubnetId(String.format("%s , %s  ", EC2_SUBNET, EC2_SUBNET_2));
    emrClusterDefinition.setNodeTags(Lists.newArrayList(new NodeTag("tagName", "tagValue")));

    when(mockEmrOperations.runEmrJobFlow(any(), any())).then(new Answer<String>()
    {
        @Override
        public String answer(InvocationOnMock invocation)
        {
            // Assert that the given EMR cluster definition produced the correct RunJobFlowRequest.
            RunJobFlowRequest runJobFlowRequest = invocation.getArgument(1);
            JobFlowInstancesConfig jobFlowInstancesConfig = runJobFlowRequest.getInstances();
            assertEquals(0, CollectionUtils.size(jobFlowInstancesConfig.getInstanceGroups()));
            final List<InstanceTypeConfig> expectedInstanceTypeConfigs = null;
            assertEquals(Lists.newArrayList(
                new InstanceFleetConfig().withName(name).withInstanceFleetType(instanceFleetType).withTargetOnDemandCapacity(targetOnDemandCapacity)
                    .withTargetSpotCapacity(targetSpotCapacity).withInstanceTypeConfigs(expectedInstanceTypeConfigs).withLaunchSpecifications(null)),
                jobFlowInstancesConfig.getInstanceFleets());
            assertNull(jobFlowInstancesConfig.getEc2SubnetId());
            assertEquals(2, CollectionUtils.size(jobFlowInstancesConfig.getEc2SubnetIds()));
            assertTrue(jobFlowInstancesConfig.getEc2SubnetIds().contains(EC2_SUBNET));
            assertTrue(jobFlowInstancesConfig.getEc2SubnetIds().contains(EC2_SUBNET_2));
            assertEquals(herdStringHelper.getRequiredConfigurationValue(ConfigurationValue.EMR_DEFAULT_EC2_NODE_IAM_PROFILE_NAME),
                runJobFlowRequest.getJobFlowRole());
            assertEquals(herdStringHelper.getRequiredConfigurationValue(ConfigurationValue.EMR_DEFAULT_SERVICE_IAM_ROLE_NAME),
                runJobFlowRequest.getServiceRole());
            List<StepConfig> stepConfigs = runJobFlowRequest.getSteps();
            assertEquals(0, stepConfigs.size());
            List<Tag> tags = runJobFlowRequest.getTags();
            assertEquals(1, tags.size());
            {
                Tag tag = tags.get(0);
                assertEquals("tagName", tag.getKey());
                assertEquals("tagValue", tag.getValue());
            }

            return clusterId;
        }
    });

    assertEquals(clusterId, emrDao.createEmrCluster(clusterName, emrClusterDefinition, getAwsParamsDto()));
}
 

private NewCluster submitNewClusterRequest(AmazonElasticMapReduce emr, String tag, StepCompiler stepCompiler,
        Config cluster, Filer filer, ParameterCompiler parameterCompiler)
        throws IOException
{
    RemoteFile runner = prepareRunner(filer, tag);

    // Compile steps
    stepCompiler.compile(runner);

    List<StepConfig> stepConfigs = stepCompiler.stepConfigs();

    Config ec2 = cluster.getNested("ec2");
    Config master = ec2.getNestedOrGetEmpty("master");
    List<Config> core = ec2.getOptional("core", Config.class).transform(ImmutableList::of).or(ImmutableList.of());
    List<Config> task = ec2.getListOrEmpty("task", Config.class);

    List<String> applications = cluster.getListOrEmpty("applications", String.class);
    if (applications.isEmpty()) {
        applications = ImmutableList.of("Hadoop", "Hive", "Spark", "Flink");
    }

    // TODO: allow configuring additional application parameters
    List<Application> applicationConfigs = applications.stream()
            .map(application -> new Application().withName(application))
            .collect(toList());

    // TODO: merge configurations with the same classification?
    List<Configuration> configurations = cluster.getListOrEmpty("configurations", JsonNode.class).stream()
            .map(this::configurations)
            .flatMap(Collection::stream)
            .collect(toList());

    List<JsonNode> bootstrap = cluster.getListOrEmpty("bootstrap", JsonNode.class);
    List<BootstrapActionConfig> bootstrapActions = new ArrayList<>();
    for (int i = 0; i < bootstrap.size(); i++) {
        bootstrapActions.add(bootstrapAction(i + 1, bootstrap.get(i), tag, filer, runner, parameterCompiler));
    }

    // Stage files to S3
    filer.stageFiles();

    Optional<String> subnetId = ec2.getOptional("subnet_id", String.class);

    String defaultMasterInstanceType;
    String defaultCoreInstanceType;
    String defaultTaskInstanceType;

    if (subnetId.isPresent()) {
        // m4 requires VPC (subnet id)
        defaultMasterInstanceType = "m4.2xlarge";
        defaultCoreInstanceType = "m4.xlarge";
        defaultTaskInstanceType = "m4.xlarge";
    }
    else {
        defaultMasterInstanceType = "m3.2xlarge";
        defaultCoreInstanceType = "m3.xlarge";
        defaultTaskInstanceType = "m3.xlarge";
    }

    RunJobFlowRequest request = new RunJobFlowRequest()
            .withName(cluster.get("name", String.class, "Digdag") + " (" + tag + ")")
            .withReleaseLabel(cluster.get("release", String.class, "emr-5.2.0"))
            .withSteps(stepConfigs)
            .withBootstrapActions(bootstrapActions)
            .withApplications(applicationConfigs)
            .withLogUri(cluster.get("logs", String.class, null))
            .withJobFlowRole(cluster.get("cluster_role", String.class, "EMR_EC2_DefaultRole"))
            .withServiceRole(cluster.get("service_role", String.class, "EMR_DefaultRole"))
            .withTags(new Tag().withKey("DIGDAG_CLUSTER_ID").withValue(tag))
            .withVisibleToAllUsers(cluster.get("visible", boolean.class, true))
            .withConfigurations(configurations)
            .withInstances(new JobFlowInstancesConfig()
                    .withInstanceGroups(ImmutableList.<InstanceGroupConfig>builder()
                            // Master Node
                            .add(instanceGroupConfig("Master", master, "MASTER", defaultMasterInstanceType, 1))
                            // Core Group
                            .addAll(instanceGroupConfigs("Core", core, "CORE", defaultCoreInstanceType))
                            // Task Groups
                            .addAll(instanceGroupConfigs("Task %d", task, "TASK", defaultTaskInstanceType))
                            .build()
                    )
                    .withAdditionalMasterSecurityGroups(ec2.getListOrEmpty("additional_master_security_groups", String.class))
                    .withAdditionalSlaveSecurityGroups(ec2.getListOrEmpty("additional_slave_security_groups", String.class))
                    .withEmrManagedMasterSecurityGroup(ec2.get("emr_managed_master_security_group", String.class, null))
                    .withEmrManagedSlaveSecurityGroup(ec2.get("emr_managed_slave_security_group", String.class, null))
                    .withServiceAccessSecurityGroup(ec2.get("service_access_security_group", String.class, null))
                    .withTerminationProtected(cluster.get("termination_protected", boolean.class, false))
                    .withPlacement(cluster.getOptional("availability_zone", String.class)
                            .transform(zone -> new PlacementType().withAvailabilityZone(zone)).orNull())
                    .withEc2SubnetId(subnetId.orNull())
                    .withEc2KeyName(ec2.get("key", String.class))
                    .withKeepJobFlowAliveWhenNoSteps(!cluster.get("auto_terminate", boolean.class, true)));

    logger.info("Submitting EMR job with {} steps(s)", request.getSteps().size());
    RunJobFlowResult result = emr.runJobFlow(request);
    logger.info("Submitted EMR job with {} step(s): {}", request.getSteps().size(), result.getJobFlowId(), result);

    return NewCluster.of(result.getJobFlowId(), request.getSteps().size());
}
 

/**
 * This method uses method the AWS Java to launch an Apache HBase cluster on Amazon EMR. 
 * 
 * @param client - AmazonElasticMapReduce client that interfaces directly with the Amazon EMR Web Service
 * @param clusterIdentifier - identifier of an existing cluster
 * @param amiVersion - AMI to use for launching this cluster
 * @param keypair - A keypair for SSHing into the Amazon EMR master node
 * @param masterInstanceType - Master node Amazon EC2 instance type 
 * @param coreInstanceType - core nodes Amazon EC2 instance type 
 * @param logUri - An Amazon S3 bucket for your 
 * @param numberOfNodes - total number of nodes in this cluster including master node
 * @return
 */
public static String createCluster(AmazonElasticMapReduce client,
		String clusterIdentifier,
		String amiVersion,
		String keypair,
		String masterInstanceType,
		String coreInstanceType,
		String logUri,
		int numberOfNodes) {

	if (clusterExists(client, clusterIdentifier)) {
		LOG.info("Cluster " + clusterIdentifier + " is available");
		return clusterIdentifier;
	}
	
	//Error checking
	if (amiVersion == null || amiVersion.isEmpty()) throw new RuntimeException("ERROR: Please specify an AMI Version");
	if (keypair == null || keypair.isEmpty()) throw new RuntimeException("ERROR: Please specify a valid Amazon Key Pair");
	if (masterInstanceType == null || masterInstanceType.isEmpty()) throw new RuntimeException("ERROR: Please specify a Master Instance Type");
	if (logUri == null || logUri.isEmpty()) throw new RuntimeException("ERROR: Please specify a valid Amazon S3 bucket for your logs.");
	if (numberOfNodes < 0) throw new RuntimeException("ERROR: Please specify at least 1 node");
	  		
	  RunJobFlowRequest request = new RunJobFlowRequest()
	    .withAmiVersion(amiVersion)
		.withBootstrapActions(new BootstrapActionConfig()
		             .withName("Install HBase")
		             .withScriptBootstrapAction(new ScriptBootstrapActionConfig()
		             .withPath("s3://elasticmapreduce/bootstrap-actions/setup-hbase")))
		.withName("Job Flow With HBAse Actions")	 
		.withSteps(new StepConfig() //enable debugging step
					.withName("Enable debugging")
					.withActionOnFailure("TERMINATE_CLUSTER")
					.withHadoopJarStep(new StepFactory().newEnableDebuggingStep()), 
					//Start HBase step - after installing it with a bootstrap action
					createStepConfig("Start HBase","TERMINATE_CLUSTER", "/home/hadoop/lib/hbase.jar", getHBaseArgs()), 
					//add HBase backup step
					createStepConfig("Modify backup schedule","TERMINATE_JOB_FLOW", "/home/hadoop/lib/hbase.jar", getHBaseBackupArgs()))
		.withLogUri(logUri)
		.withInstances(new JobFlowInstancesConfig()
		.withEc2KeyName(keypair)
		.withInstanceCount(numberOfNodes)
		.withKeepJobFlowAliveWhenNoSteps(true)
		.withMasterInstanceType(masterInstanceType)
		.withSlaveInstanceType(coreInstanceType));

	RunJobFlowResult result = client.runJobFlow(request);
	
	String state = null;
	while (!(state = clusterState(client, result.getJobFlowId())).equalsIgnoreCase("waiting")) {
		try {
			Thread.sleep(10 * 1000);
			LOG.info(result.getJobFlowId() + " is " + state + ". Waiting for cluster to become available.");
		} catch (InterruptedException e) {

		}
		
		if (state.equalsIgnoreCase("TERMINATED_WITH_ERRORS")){
			LOG.error("Could not create EMR Cluster");
			System.exit(-1);	
		}
	}
	LOG.info("Created cluster " + result.getJobFlowId());
	LOG.info("Cluster " + clusterIdentifier + " is available");	
	return result.getJobFlowId();
}