# Using managed scaling in Amazon EMR **Important** We strongly recommend that you use the latest Amazon EMR release (Amazon EMR 7.12.0) for managed scaling. In some early releases, you might experience intermittent application failures or delays in scaling. Amazon EMR resolved this issue with 5.x releases 5.30.2, 5.31.1, 5.32.1, 5.33.1 and higher, and with 6.x releases 6.1.1, 6.2.1, 6.3.1 and higher. For more information Region and release availability, see [Managed scaling availability](#emr-managed-scaling-availability). ## Overview With Amazon EMR versions 5.30.0 and higher (except for Amazon EMR 6.0.0), you can enable Amazon EMR managed scaling. Managed scaling lets you automatically increase or decrease the number of instances or units in your cluster based on workload. Amazon EMR continuously evaluates cluster metrics to make scaling decisions that optimize your clusters for cost and speed. Managed scaling is available for clusters composed of either instance groups or instance fleets. ## Managed scaling availability + In the following AWS Regions, Amazon EMR managed scaling is available with Amazon EMR 6.14.0 and higher: + Asia Pacific (Taipei) (ap-east-2) + Asia Pacific (Melbourne) (ap-southeast-4) + Asia Pacific (Malaysia) (ap-southeast-5) + Asia Pacific (New Zealand) (ap-southeast-6) + Asia Pacific (Thailand) (ap-southeast-7) + Canada West (Calgary) (ca-west-1) + Europe (Spain) (eu-south-2) + Mexico (Central) (mx-central-1) + In the following AWS Regions, Amazon EMR managed scaling is available with Amazon EMR 5.30.0 and 6.1.0 and higher: + US East (N. Virginia) (us-east-1) + US East (Ohio) (us-east-2) + US West (Oregon) (us-west-2) + US West (N. California) (us-west-1) + Africa (Cape Town) (af-south-1) + Asia Pacific (Hong Kong) (ap-east-1) + Asia Pacific (Mumbai) (ap-south-1) + Asia Pacific (Hyderabad) (ap-south-2) + Asia Pacific (Seoul) (ap-northeast-2) + Asia Pacific (Singapore) (ap-southeast-1) + Asia Pacific (Sydney) (ap-southeast-2) + Asia Pacific (Jakarta) (ap-southeast-3) + Asia Pacific (Tokyo) (ap-northeast-1) + Asia Pacific (Osaka) (ap-northeast-3) + Canada (Central) (ca-central-1) + South America (São Paulo) (sa-east-1) + Europe (Frankfurt) (eu-central-1) + Europe (Zurich) (eu-central-2) + Europe (Ireland) (eu-west-1) + Europe (London) (eu-west-2) + Europe (Milan) (eu-south-1) + Europe (Paris) (eu-west-3) + Europe (Stockholm) (eu-north-1) + Israel (Tel Aviv) (il-central-1) + Middle East (UAE) (me-central-1) + China (Beijing) (cn-north-1) + China (Ningxia) (cn-northwest-1) + AWS GovCloud (US-East) (us-gov-east-1) + AWS GovCloud (US-West) (us-gov-west-1) + Amazon EMR managed scaling only works with YARN applications, such as Spark, Hadoop, Hive, and Flink. It doesn't support applications that are not based on YARN, such as Presto and HBase. ## Managed scaling parameters You must configure the following parameters for managed scaling. The limit only applies to the core and task nodes. You cannot scale the primary node after initial configuration. + **Minimum** (`MinimumCapacityUnits`) – The lower boundary of allowed EC2 capacity in a cluster. It is measured through virtual central processing unit (vCPU) cores or instances for instance groups. It is measured through units for instance fleets. + **Maximum** (`MaximumCapacityUnits`) – The upper boundary of allowed EC2 capacity in a cluster. It is measured through virtual central processing unit (vCPU) cores or instances for instance groups. It is measured through units for instance fleets. + **On-Demand limit** (`MaximumOnDemandCapacityUnits`) (Optional) – The upper boundary of allowed EC2 capacity for On-Demand market type in a cluster. If this parameter is not specified, it defaults to the value of `MaximumCapacityUnits`. + This parameter is used to split capacity allocation between On-Demand and Spot Instances. For example, if you set the minimum parameter as 2 instances, the maximum parameter as 100 instances, the On-Demand limit as 10 instances, then Amazon EMR managed scaling scales up to 10 On-Demand Instances and allocates the remaining capacity to Spot Instances. For more information, see [Node allocation scenarios](managed-scaling-allocation-strategy.md#node-allocation-scenarios). + **Maximum core nodes **(`MaximumCoreCapacityUnits`) (Optional) – The upper boundary of allowed EC2 capacity for core node type in a cluster. If this parameter is not specified, it defaults to the value of `MaximumCapacityUnits`. + This parameter is used to split capacity allocation between core and task nodes. For example, if you set the minimum parameter as 2 instances, the maximum as 100 instances, the maximum core node as 17 instances, then Amazon EMR managed scaling scales up to 17 core nodes and allocates the remaining 83 instances to task nodes. For more information, see [Node allocation scenarios](managed-scaling-allocation-strategy.md#node-allocation-scenarios). For more information about managed scaling parameters, see [https://docs.aws.amazon.com/emr/latest/APIReference/API_ComputeLimits.html](https://docs.aws.amazon.com/emr/latest/APIReference/API_ComputeLimits.html). ## Considerations for Amazon EMR managed scaling + Managed scaling is supported in limited AWS Regions and Amazon EMR releases. For more information, see [Managed scaling availability](#emr-managed-scaling-availability). + You must configure the required parameters for Amazon EMR managed scaling. For more information, see [Managed scaling parameters](#emr-managed-scaling-parameters). + To use managed scaling, the metrics-collector process must be able to connect to the public API endpoint for managed scaling in API Gateway. If you use a private DNS name with Amazon Virtual Private Cloud, managed scaling won't function properly. To ensure that managed scaling works, we recommend that you take one of the following actions: + Remove the API Gateway interface VPC endpoint from your Amazon VPC. + Follow the instructions in [Why do I get an HTTP 403 Forbidden error when connecting to my API Gateway APIs from a VPC?](https://aws.amazon.com/premiumsupport/knowledge-center/api-gateway-vpc-connections/) to disable the private DNS name setting. + Launch your cluster in a private subnet instead. For more information, see the topic on [Private subnets](emr-clusters-in-a-vpc.md#emr-vpc-private-subnet). + If your YARN jobs are intermittently slow during scale down, and YARN Resource Manager logs show that most of your nodes were deny-listed during that time, you can adjust the decommissioning timeout threshold. Reduce the `spark.blacklist.decommissioning.timeout` from one hour to one minute to make the node available for other pending containers to continue task processing. You should also set `YARN.resourcemanager.nodemanager-graceful-decommission-timeout-secs` to a larger value to ensure Amazon EMR doesn’t force terminate the node while the longest “Spark Task” is still running on the node. The current default is 60 minutes, which means YARN force-terminates the container after 60 minutes once the node enters the decomissioning state. The following example YARN Resource Manager Log line shows nodes added to the decomissioning state: ``` 2021-10-20 15:55:26,994 INFO org.apache.hadoop.YARN.server.resourcemanager.DefaultAMSProcessor (IPC Server handler 37 on default port 8030): blacklist are updated in Scheduler.blacklistAdditions: [ip-10-10-27-207.us-west-2.compute.internal, ip-10-10-29-216.us-west-2.compute.internal, ip-10-10-31-13.us-west-2.compute.internal, ... , ip-10-10-30-77.us-west-2.compute.internal], blacklistRemovals: [] ``` See more [details on how Amazon EMR integrates with YARN deny listing during decommissioning of nodes](https://aws.amazon.com/blogs/big-data/spark-enhancements-for-elasticity-and-resiliency-on-amazon-emr/), [cases when nodes in Amazon EMR can be deny listed](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-troubleshoot-error-resource-3.html), and [configuring Spark node-decommissioning behavior](https://docs.aws.amazon.com/emr/latest/ReleaseGuide/emr-spark-configure.html#spark-decommissioning). + For Spark workloads, disabling Spark Dynamic Resource Allocator (DRA) by changing the Spark property **spark.dynamicAllocation.enabled** to `FALSE` can cause Managed Scaling issues, where clusters can be scaled up more than required for your workloads (up to the maximum compute). When using Managed Scaling for these workloads, we recommend that you keep Spark DRA enabled, which is the default state of this property. + Over-utilization of EBS volumes can cause Managed Scaling issues. We recommend that you maintain EBS volume below 90% utilization. For more information, see [Instance storage options and behavior in Amazon EMR](emr-plan-storage.md). + Amazon CloudWatch metrics are critical for Amazon EMR managed scaling to operate. We recommend that you closely monitor Amazon CloudWatch metrics to make sure data is not missing. For more information about how you can configure CloudWatch alarms to detect missing metrics, see [Using Amazon CloudWatch alarms](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/AlarmThatSendsEmail.html). + Managed scaling operations on 5.30.0 and 5.30.1 clusters without Presto installed may cause application failures or cause a uniform instance group or instance fleet to stay in the `ARRESTED` state, particularly when a scale down operation is followed quickly by a scale up operation. As a workaround, choose Presto as an application to install when you create a cluster with Amazon EMR releases 5.30.0 and 5.30.1, even if your job does not require Presto. + When you set the maximum core node and the On-Demand limit for Amazon EMR managed scaling, consider the differences between instance groups and instance fleets. Each instance group consists of the same instance type and the same purchasing option for instances: On-Demand or Spot. For each instance fleet, you can specify up to five instance types, which can be provisioned as On-Demand and Spot Instances. For more information, see [Create a cluster with instance fleets or uniform instance groups](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-instance-group-configuration.html), [Instance fleet options](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-instance-fleet.html#emr-instance-fleet-options), and [Node allocation scenarios](managed-scaling-allocation-strategy.md#node-allocation-scenarios). + With Amazon EMR 5.30.0 and higher, if you remove the default **Allow All** outbound rule to 0.0.0.0/ for the master security group, you must add a rule that allows outbound TCP connectivity to your security group for service access on port 9443. Your security group for service access must also allow inbound TCP traffic on port 9443 from the master security group. For more information about configuring security groups, see [Amazon EMR-managed security group for the primary instance (private subnets)](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-man-sec-groups.html#emr-sg-elasticmapreduce-master-private). + You can use AWS CloudFormation to configure Amazon EMR managed scaling. For more information, see [AWS::EMR::Cluster](https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-resource-elasticmapreduce-cluster.html) in the *AWS CloudFormation User Guide*. + If you're using Spot nodes, consider using node labels to prevent Amazon EMR from removing application processes when Amazon EMR removes Spot nodes. For more information about node labels, see [Task nodes](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-master-core-task-nodes.html#emr-plan-task). + Node labeling is not supported by default in Amazon EMR releases 6.15 or lower. For more information, see [Understand node types: primary, core, and task nodes.](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-master-core-task-nodes.html) + If you're using Amazon EMR releases 6.15 or lower, you can only assign node labels by node type, such as core and task nodes. However, if you're using Amazon EMR release 7.0 or higher, you can configure node labels by node type and market type, such as On-Demand and Spot. + If application process demand increases and executor demand decreases when you restricted the application process to core nodes, you can add back core nodes and remove task nodes in the same resize operation. For more information, see [Understanding node allocation strategy and scenarios](https://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html). + Amazon EMR doesn't label task nodes, so you can't set the YARN properties to restrict application processes only for task nodes. However, if you want to use market types as node labels, you can use the `ON_DEMAND` or `SPOT` labels for application process placement. We don't recommend using Spot nodes for application primary processes. + When using node labels, the total running units in the cluster can temporarily exceed the max compute set in your managed scaling policy while Amazon EMR decommissions some of your instances. Total requested units will always stay at or below your policy’s max compute. + Managed scaling only supports the node labels `ON_DEMAND` and `SPOT` or `CORE` and `TASK`. Custom node labels aren't supported. + Amazon EMR creates node labels when creating the cluster and provisioning resources. Amazon EMR doesn't support adding node labels when you reconfigure the cluster. You also can't modify the node labels when configuring managed scaling after launching the cluster. + Managed scaling scales core and task nodes independently based on application process and executor demand. To prevent HDFS data loss issues during core scale down, follow standard practice for core nodes. To learn more about best practices about core nodes and HDFS replication, see [Considerations and best practices](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-plan-ha-considerations.html). + You can't place both the application process and executors on only the `core` or the `ON_DEMAND` node. If you want to add both the application process and executors on one of the nodes, don't use the `yarn.node-labels.am.default-node-label-expression` configuration. For example, to place both the application process and executors in `ON_DEMAND` nodes, set max compute to the same as the maximum in the `ON_DEMAND` node. Also remove the `yarn.node-labels.am.default-node-label-expression` configuration. To add both the application process and executors on `core` nodes, remove the `yarn.node-labels.am.default-node-label-expression` configuration. + When you use managed scaling with node labels, set the property `yarn.scheduler.capacity.maximum-am-resource-percent: 1` if you plan to run multiple applications in parallel. Doing so ensures that your application processes fully utilize the available `CORE` or `ON_DEMAND` nodes. + If you use managed scaling with node labels, set the property `yarn.resourcemanager.decommissioning.timeout` to a value that's longer than the longest running application on your cluster. Doing so reduces the chance that Amazon EMR managed scalling needs to reschedule your applications to recommission `CORE` or `ON_DEMAND` nodes. + To reduce the risk of application failures due to shuffle data loss, Amazon EMR collects metrics from the cluster to determine nodes that have existing transient shuffle data from the current and previous stage. In rare cases, metrics can continue to report stale data for applications that are already completed or terminated. This can impact timely scale down of instances in your cluster. For clusters that have large amount of shuffle data, consider using EMR versions 6.13 and later. ## Feature history This table lists updates to the Amazon EMR managed scaling capability. | Release date | Capability | Amazon EMR versions | | --- | --- | --- | | November 20, 2024 | Managed scaling is available in the il-central-1 Israel (Tel Aviv), me-central-1 Middle East (UAE), and ap-northeast-3 Asia Pacific (Osaka) regions. | 5.30.0 and 6.1.0 and higher | | November 15, 2024 | Managed scaling is available in the eu-central-2 Europe (Zurich) Region. | 5.30.0 and 6.1.0 and higher | | August 20, 2024 | Node labels are now available in managed scaling, so you can label your instances based on market type or node type to improve automatic scaling. | 7.2.0 and higher | | March 31, 2024 | Managed scaling is available in the ap-south-2 Asia Pacific (Hyderabad) Region. | 6.14.0 and higher | | February 13, 2024 | Managed scaling is available in the eu-south-2 Europe (Spain) Region. | 6.14.0 and higher | | October 10, 2023 | Managed scaling is available in the ap-southeast-3 Asia Pacific (Jakarta) Region. | 6.14.0 and higher | | July 28, 2023 | Enhanced managed scaling to switch to different task instance group on scale-up when Amazon EMR experiences a delay in scale-up with the current instance group. | 5.34.0 and higher, 6.4.0 and higher | | June 16, 2023 | Enhanced managed scaling to be aware of the nodes running application master so that those nodes are not scaled down. For more information, see [Understanding Amazon EMR node allocation strategy and scenarios](managed-scaling-allocation-strategy.md). | 5.34.0 and higher, 6.4.0 and higher | | March 21, 2022 | Added Spark shuffle data awareness used when scaling-down clusters. For Amazon EMR clusters with Apache Spark and the managed scaling feature enabled, Amazon EMR continuously monitors Spark executors and intermediate shuffle data locations. Using this information, Amazon EMR scales-down only under-utilized instances which don't contain actively used shuffle data. This prevents recomputation of lost shuffle data, helping to lower cost and improve job performance. For more information, see the [Spark Programming Guide](https://spark.apache.org/docs/latest/rdd-programming-guide.html#shuffle-operations). | 5.34.0 and higher, 6.4.0 and higher | # Configure managed scaling for Amazon EMR The following sections explain how to launch an EMR cluster that uses managed scaling with the AWS Management Console, the AWS SDK for Java, or the AWS Command Line Interface. **Topics** + [Use the AWS Management Console to configure managed scaling](#managed-scaling-console) + [Use the AWS CLI to configure managed scaling](#managed-scaling-cli) + [Use AWS SDK for Java to configure managed scaling](#managed-scaling-sdk) ## Use the AWS Management Console to configure managed scaling You can use the Amazon EMR console to configure managed scaling when you create a cluster or to change a managed scaling policy for a running cluster. ------ #### [ Console ] **To configure managed scaling when you create a cluster with the console** 1. Sign in to the AWS Management Console, and open the Amazon EMR console at [https://console.aws.amazon.com/emr](https://console.aws.amazon.com/emr). 1. Under **EMR on EC2** in the left navigation pane, choose **Clusters**, and then choose **Create cluster**. 1. Choose an Amazon EMR release **emr-5.30.0** or later, except version **emr-6.0.0**. 1. Under **Cluster scaling and provisioning option**, choose **Use EMR-managed scaling**. Specify the **Minimum** and **Maximum** number of instances, the **Maximum core node** instances, and the **Maximum On-Demand** instances. 1. Choose any other options that apply to your cluster. 1. To launch your cluster, choose **Create cluster**. **To configure managed scaling on an existing cluster with the console** 1. Sign in to the AWS Management Console, and open the Amazon EMR console at [https://console.aws.amazon.com/emr](https://console.aws.amazon.com/emr). 1. Under **EMR on EC2** in the left navigation pane, choose **Clusters**, and select the cluster that you want to update. 1. On the **Instances** tab of the cluster details page, find the **Instance group settings** section. Select **Edit cluster scaling** to specify new values for the **Minimum** and **Maximum** number of instances and the **On-Demand** limit. ------ ## Use the AWS CLI to configure managed scaling You can use AWS CLI commands for Amazon EMR to configure managed scaling when you create a cluster. You can use a shorthand syntax, specifying the JSON configuration inline within the relevant commands, or you can reference a file containing the configuration JSON. You can also apply a managed scaling policy to an existing cluster and remove a managed scaling policy that was previously applied. In addition, you can retrieve details of a scaling policy configuration from a running cluster. **Enabling Managed Scaling During Cluster Launch** You can enable managed scaling during cluster launch as the following example demonstrates. ``` aws emr create-cluster \ --service-role EMR_DefaultRole \ --release-label emr-7.12.0 \ --name EMR_Managed_Scaling_Enabled_Cluster \ --applications Name=Spark Name=Hbase \ --ec2-attributes KeyName=keyName,InstanceProfile=EMR_EC2_DefaultRole \ --instance-groups InstanceType=m4.xlarge,InstanceGroupType=MASTER,InstanceCount=1 InstanceType=m4.xlarge,InstanceGroupType=CORE,InstanceCount=2 \ --region us-east-1 \ --managed-scaling-policy ComputeLimits='{MinimumCapacityUnits=2,MaximumCapacityUnits=4,UnitType=Instances}' ``` You can also specify a managed policy configuration using the --managed-scaling-policy option when you use `create-cluster`. **Applying a Managed Scaling Policy to an Existing Cluster** You can apply a managed scaling policy to an existing cluster as the following example demonstrates. ``` aws emr put-managed-scaling-policy --cluster-id j-123456 --managed-scaling-policy ComputeLimits='{MinimumCapacityUnits=1, MaximumCapacityUnits=10, MaximumOnDemandCapacityUnits=10, UnitType=Instances}' ``` You can also apply a managed scaling policy to an existing cluster by using the `aws emr put-managed-scaling-policy` command. The following example uses a reference to a JSON file, `managedscaleconfig.json`, that specifies the managed scaling policy configuration. ``` aws emr put-managed-scaling-policy --cluster-id j-123456 --managed-scaling-policy file://./managedscaleconfig.json ``` The following example shows the contents of the `managedscaleconfig.json` file, which defines the managed scaling policy. ``` { "ComputeLimits": { "UnitType": "Instances", "MinimumCapacityUnits": 1, "MaximumCapacityUnits": 10, "MaximumOnDemandCapacityUnits": 10 } } ``` **Retrieving a Managed Scaling Policy Configuration** The `GetManagedScalingPolicy` command retrieves the policy configuration. For example, the following command retrieves the configuration for the cluster with a cluster ID of `j-123456`. ``` aws emr get-managed-scaling-policy --cluster-id j-123456 ``` The command produces the following example output. ``` 1. { 2. "ManagedScalingPolicy": { 3. "ComputeLimits": { 4. "MinimumCapacityUnits": 1, 5. "MaximumOnDemandCapacityUnits": 10, 6. "MaximumCapacityUnits": 10, 7. "UnitType": "Instances" 8. } 9. } 10. } ``` For more information about using Amazon EMR commands in the AWS CLI, see [https://docs.aws.amazon.com/cli/latest/reference/emr](https://docs.aws.amazon.com/cli/latest/reference/emr). **Removing Managed Scaling Policy** The `RemoveManagedScalingPolicy` command removes the policy configuration. For example, the following command removes the configuration for the cluster with a cluster ID of `j-123456`. ``` aws emr remove-managed-scaling-policy --cluster-id j-123456 ``` ## Use AWS SDK for Java to configure managed scaling The following program excerpt shows how to configure managed scaling using the AWS SDK for Java: ``` package com.amazonaws.emr.sample; import java.util.ArrayList; import java.util.List; import com.amazonaws.AmazonClientException; import com.amazonaws.auth.AWSCredentials; import com.amazonaws.auth.AWSStaticCredentialsProvider; import com.amazonaws.auth.profile.ProfileCredentialsProvider; import com.amazonaws.regions.Regions; import com.amazonaws.services.elasticmapreduce.AmazonElasticMapReduce; import com.amazonaws.services.elasticmapreduce.AmazonElasticMapReduceClientBuilder; import com.amazonaws.services.elasticmapreduce.model.Application; import com.amazonaws.services.elasticmapreduce.model.ComputeLimits; import com.amazonaws.services.elasticmapreduce.model.ComputeLimitsUnitType; import com.amazonaws.services.elasticmapreduce.model.InstanceGroupConfig; import com.amazonaws.services.elasticmapreduce.model.JobFlowInstancesConfig; import com.amazonaws.services.elasticmapreduce.model.ManagedScalingPolicy; import com.amazonaws.services.elasticmapreduce.model.RunJobFlowRequest; import com.amazonaws.services.elasticmapreduce.model.RunJobFlowResult; public class CreateClusterWithManagedScalingWithIG { public static void main(String[] args) { AWSCredentials credentialsFromProfile = getCreadentials("AWS-Profile-Name-Here"); /** * Create an Amazon EMR client with the credentials and region specified in order to create the cluster */ AmazonElasticMapReduce emr = AmazonElasticMapReduceClientBuilder.standard() .withCredentials(new AWSStaticCredentialsProvider(credentialsFromProfile)) .withRegion(Regions.US_EAST_1) .build(); /** * Create Instance Groups - Primary, Core, Task */ InstanceGroupConfig instanceGroupConfigMaster = new InstanceGroupConfig() .withInstanceCount(1) .withInstanceRole("MASTER") .withInstanceType("m4.large") .withMarket("ON_DEMAND"); InstanceGroupConfig instanceGroupConfigCore = new InstanceGroupConfig() .withInstanceCount(4) .withInstanceRole("CORE") .withInstanceType("m4.large") .withMarket("ON_DEMAND"); InstanceGroupConfig instanceGroupConfigTask = new InstanceGroupConfig() .withInstanceCount(5) .withInstanceRole("TASK") .withInstanceType("m4.large") .withMarket("ON_DEMAND"); List igConfigs = new ArrayList<>(); igConfigs.add(instanceGroupConfigMaster); igConfigs.add(instanceGroupConfigCore); igConfigs.add(instanceGroupConfigTask); /** * specify applications to be installed and configured when Amazon EMR creates the cluster */ Application hive = new Application().withName("Hive"); Application spark = new Application().withName("Spark"); Application ganglia = new Application().withName("Ganglia"); Application zeppelin = new Application().withName("Zeppelin"); /** * Managed Scaling Configuration - * Using UnitType=Instances for clusters composed of instance groups * * Other options are: * UnitType = VCPU ( for clusters composed of instance groups) * UnitType = InstanceFleetUnits ( for clusters composed of instance fleets) **/ ComputeLimits computeLimits = new ComputeLimits() .withMinimumCapacityUnits(1) .withMaximumCapacityUnits(20) .withUnitType(ComputeLimitsUnitType.Instances); ManagedScalingPolicy managedScalingPolicy = new ManagedScalingPolicy(); managedScalingPolicy.setComputeLimits(computeLimits); // create the cluster with a managed scaling policy RunJobFlowRequest request = new RunJobFlowRequest() .withName("EMR_Managed_Scaling_TestCluster") .withReleaseLabel("emr-7.12.0") // Specifies the version label for the Amazon EMR release; we recommend the latest release .withApplications(hive,spark,ganglia,zeppelin) .withLogUri("s3://path/to/my/emr/logs") // A URI in S3 for log files is required when debugging is enabled. .withServiceRole("EMR_DefaultRole") // If you use a custom IAM service role, replace the default role with the custom role. .withJobFlowRole("EMR_EC2_DefaultRole") // If you use a custom Amazon EMR role for EC2 instance profile, replace the default role with the custom Amazon EMR role. .withInstances(new JobFlowInstancesConfig().withInstanceGroups(igConfigs) .withEc2SubnetId("subnet-123456789012345") .withEc2KeyName("my-ec2-key-name") .withKeepJobFlowAliveWhenNoSteps(true)) .withManagedScalingPolicy(managedScalingPolicy); RunJobFlowResult result = emr.runJobFlow(request); System.out.println("The cluster ID is " + result.toString()); } public static AWSCredentials getCredentials(String profileName) { // specifies any named profile in .aws/credentials as the credentials provider try { return new ProfileCredentialsProvider("AWS-Profile-Name-Here") .getCredentials(); } catch (Exception e) { throw new AmazonClientException( "Cannot load credentials from .aws/credentials file. " + "Make sure that the credentials file exists and that the profile name is defined within it.", e); } } public CreateClusterWithManagedScalingWithIG() { } } ``` # Advanced Scaling for Amazon EMR Starting with Amazon EMR on EC2 version 7.0, you can leverage Advanced Scaling to control your cluster's resource utilization. Advanced Scaling introduces a utilization-performance scale for tuning your resource utilization and performance level according to your business needs. The value you set determines whether your cluster is weighted more to resource conservation or to scaling up to handle service-level-agreement (SLA) sensitive workloads, where quick completion is critical. When the scaling value is adjusted, managed scaling interprets your intent and intelligently scales to optimize resources. For more information about managed scaling, see [Configure managed scaling for Amazon EMR](https://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-configure.html). ## Advanced Scaling settings The value your set for Advanced Scaling optimizes your cluster to your requirements. Values range from **1**-**100**. Possible values are **1**, **25**, **50**, **75** and **100**. If you set the index to values other than these, it results in a validation error. Scaling values map to resource-utilization strategies. The following list defines several of these: + **Utilization optimized [1]** – This setting prevents resource over provisioning. Use a low value when you want to keep costs low and to prioritize efficient resource utilization. It causes the cluster to scale up less aggressively. This works well for the use case when there are regularly occurring workload spikes and you don't want resources to ramp up too quickly. + **Balanced [50]** – This balances resource utilization and job performance. This setting is suitable for steady workloads where most stages have a stable runtime. It's also suitable for workloads with a mix of short and long-running stages. We recommend starting with this setting if you aren't sure which to choose. + ** Performance optimized [100]** – This strategy prioritizes performance. The cluster scales up aggressively to ensure that jobs complete quickly and meet performance targets. Performance optimized is suitable for service-level-agreement (SLA) sensitive workloads where fast run time is critical. **Note** The intermediate values available provide a middle ground between strategies in order to fine tune your cluster's Advanced Scaling behavior. ## Benefits of Advanced Scaling As you have variability in your environment and requirements, such as changing data volumes, cost-target adjustments, and SLA implementations, cluster scaling can help you adjust your cluster configuration to achieve your objectives. Key benefits include: + **Enhanced granular control** – The introduction of the utilization-performance setting allows you to easily adjust your cluster's scaling behavior according to your requirements. You can scale up to meet demand for compute resources or scale down to save resources, based on your use patterns. + **Improved cost optimization** – You can choose a low utilization value as requirements dictate to more easily meet your cost objectives. ## Getting started with optimization **Setup and configuration** Use these steps to set the performance index and optimize your scaling strategy. 1. The following command updates an existing cluster with the utilization-optimized `[1]` scaling strategy: ``` aws emr put-managed-scaling-policy --cluster-id 'cluster-id' \ --managed-scaling-policy '{ "ComputeLimits": { "UnitType": "Instances", "MinimumCapacityUnits": 1, "MaximumCapacityUnits": 2, "MaximumOnDemandCapacityUnits": 2, "MaximumCoreCapacityUnits": 2 }, "ScalingStrategy": "ADVANCED", "UtilizationPerformanceIndex": "1" }' \ --region "region-name" ``` The attributes `ScalingStrategy` and `UtilizationPerformanceIndex` are new and relevant to scaling optimization. You can select different scaling strategies by setting corresponding values (1, 25, 50, 75, and 100) for the `UtilizationPerformanceIndex` attribute in the managed-scaling policy. 1. To revert to the default managed-scaling strategy, run the `put-managed-scaling-policy` command without including the `ScalingStrategy` and `UtilizationPerformanceIndex` attributes. (This is optional.) This sample shows how to do this: ``` aws emr put-managed-scaling-policy \ --cluster-id 'cluster-id' \ --managed-scaling-policy '{"ComputeLimits":{"UnitType":"Instances","MinimumCapacityUnits":1,"MaximumCapacityUnits":2,"MaximumOnDemandCapacityUnits":2,"MaximumCoreCapacityUnits":2}}' \ --region "region-name" ``` **Using monitoring metrics to track cluster utilization** Starting with EMR version 7.3.0, Amazon EMR publishes four new metrics related to memory and virtual CPU. You can use these to measure cluster utilization across scaling strategies. These metrics are available for any use case, but you can use the details provided here for monitoring Advanced Scaling. Helpful metrics available include the following: + **YarnContainersUsedMemoryGBSeconds** – Amount of memory consumed by applications managed by YARN. + **YarnContainersTotalMemoryGBSeconds** – Total memory capacity allocated to YARN within the cluster. + **YarnNodesUsedVCPUSeconds** – Total VCPU seconds for each application managed by YARN. + **YarnNodesTotalVCPUSeconds** – Aggregated total VCPU seconds for memory consumed, including the time window when yarn is not ready. You can analyze resource metrics using Amazon CloudWatch Logs Insights. Features include a purpose-built query language that helps you extract metrics specific to resource use and scaling. The following query, which you can run in the Amazon CloudWatch console, uses metric math to calculate the average memory utilization (e1) by dividing the running sum of consumed memory (e2) by the running sum of total memory (e3): ``` { "metrics": [ [ { "expression": "e2/e3", "label": "Average Mem Utilization", "id": "e1", "yAxis": "right" } ], [ { "expression": "RUNNING_SUM(m1)", "label": "RunningTotal-YarnContainersUsedMemoryGBSeconds", "id": "e2", "visible": false } ], [ { "expression": "RUNNING_SUM(m2)", "label": "RunningTotal-YarnContainersTotalMemoryGBSeconds", "id": "e3", "visible": false } ], [ "AWS_EMR_ManagedResize", "YarnContainersUsedMemoryGBSeconds", "ACCOUNT_ID", "793684541905", "COMPONENT", "ManagerService", "JOB_FLOW_ID", "cluster-id", { "id": "m1", "label": "YarnContainersUsedMemoryGBSeconds" } ], [ ".", "YarnContainersTotalMemoryGBSeconds", ".", ".", ".", ".", ".", ".", { "id": "m2", "label": "YarnContainersTotalMemoryGBSeconds" } ] ], "view": "timeSeries", "stacked": false, "region": "region", "period": 60, "stat": "Sum", "title": "Memory Utilization" } ``` To query logs, you can select CloudWatch in the AWS console. For more information about writing queries for CloudWatch, see [Analyzing log data with CloudWatch Logs Insights](https://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/AnalyzingLogData.html) in the Amazon CloudWatch Logs User Guide. The following image shows these metrics for a sample cluster: ![\[Graph that shows utilization statistics.\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/images/scaling_graph_EMR.png) ## Considerations and limitations + The effectiveness of scaling strategies might vary, depending on your unique workload characteristics and cluster configuration. We encourage you to experiment with the scaling setting to determine an optimal index value for your use case. + Amazon EMR Advanced Scaling is particularly well suited for batch workloads. For SQL/data-warehousing and streaming workloads, we recommend using the default managed-scaling strategy for optimal performance. + Amazon EMR Advanced Scaling is not supported when Node Label Configurations are enabled in the cluster. If both Advanced Scaling and Node Label Configurations are enabled together in a cluster, then the scaling behavior would be as if the default managed scaling setting was enabled. + The performance-optimized scaling strategy enables faster job execution by maintaining high compute resources for a longer period than the default managed-scaling strategy. This mode prioritizes quickly scaling up to meet resource demands, resulting in quicker job completion. This might result in higher costs when compared with the default strategy. + In cases where the cluster is already optimized and fully utilized, enabling Advanced Scaling might not provide additional benefits. In some situations, enabling Advanced Scaling might lead to increased costs as workloads may run longer. In these cases, we recommend using the default managed-scaling strategy to ensure optimal resource allocation and cost efficiency. + In the context of managed scaling, the emphasis shifts towards resource utilization over execution time as the setting is adjusted from performance-optimized [**100**] to utilization-optimized [**1**]. However, it is important to note that the outcomes might vary, based on the nature of the workload and the cluster's topology. To ensure optimal results for your use case, we strongly recommend testing the scaling strategies with your workloads to determine the most suitable setting. + The **PerformanceUtilizationIndex** accepts only the following values: + **1** + **25** + **50** + **75** + **100** Any other values submitted result in a validation error. # Understanding Amazon EMR node allocation strategy and scenarios This section gives an overview of node allocation strategy and common scaling scenarios that you can use with Amazon EMR managed scaling. ## Node allocation strategy Amazon EMR managed scaling allocates core and task nodes based on the following scale-up and scale-down strategies: **Scale-up strategy ** + For Amazon EMR releases 7.2 and higher, managed scaling first adds nodes based on node labels and the application process restriction YARN property. + For Amazon EMR releases 7.2 and higher, if you enabled node labels and restricted application processes to `CORE` nodes, Amazon EMR managed scaling scales up core nodes and task nodes if application process demand increases and executor demand increases. Similarly, if you enabled node labels and restricted application processes to `ON_DEMAND` nodes, managed scaling scales up on-demand nodes if application process demand increases and scales up spot nodes if executor demand increases. + If node labels aren't enabled, application process placement aren't restricted to any node or market type. + By using node labels, managed scaling can scale up and scale down different instance groups and instance fleets in the same resize operation. For example, in a scenario in which `instance_group1` has `ON_DEMAND` node and `instance_group2` has a `SPOT` node, and node labels are enabled and application processes are restricted to nodes with the `ON_DEMAND` label. Managed scaling will scale down `instance_group1` and scale up `instance_group2` if application process demand decreases and executor demand increases. + When Amazon EMR experiences a delay in scale-up with the current instance group, clusters that use managed scaling automatically switch to a different task instance group. + If the `MaximumCoreCapacityUnits` parameter is set, then Amazon EMR scales core nodes until the core units reach the maximum allowed limit. All the remaining capacity is added to task nodes. + If the `MaximumOnDemandCapacityUnits` parameter is set, then Amazon EMR scales the cluster by using the On-Demand Instances until the On-Demand units reach the maximum allowed limit. All the remaining capacity is added using Spot Instances. + If both the `MaximumCoreCapacityUnits` and `MaximumOnDemandCapacityUnits` parameters are set, Amazon EMR considers both limits during scaling. For example, if the `MaximumCoreCapacityUnits` is less than `MaximumOnDemandCapacityUnits`, Amazon EMR first scales core nodes until the core capacity limit is reached. For the remaining capacity, Amazon EMR first uses On-Demand Instances to scale task nodes until the On-Demand limit is reached, and then uses Spot Instances for task nodes. **Scale-down strategy** + Similar to the scale-up strategy, Amazon EMR removes nodes based on node labels. For more information about node labels, see [Understand node types: primary, core, and task nodes](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-master-core-task-nodes.html). + If you haven't enabled node labels, managed scaling removes task nodes and then removes core nodes until it achieves the desired scale-down target capacity. Managed scaling never scales down the cluster below the minimum constraints specified in the managed scaling policy. + Amazon EMR versions 5.34.0 and higher, and Amazon EMR versions 6.4.0 and higher, support Spark shuffle data awareness, which prevents an instance from scaling down while Managed Scaling is aware of existing shuffle data. For more information on shuffle operations, see the [Spark Programming Guide](https://spark.apache.org/docs/latest/rdd-programming-guide.html#shuffle-operations). Managed Scaling makes best effort to prevent scaling-down nodes with shuffle data from the current and previous stage of any active Spark application, up to a maximum of 30 minutes. This helps minimize unintended shuffle data loss, avoiding the need for job re-attempts and recomputation of intermediate data. However, prevention of shuffle data loss is not guaranteed. For improved Spark shuffle protection, we recommend shuffle awareness on clusters with release label 7.4 or higher. Add the following flags to the cluster configuration to enable improved Spark shuffle protection. + If either the `yarn.nodemanager.shuffledata-monitor.interval-ms` flag (default 30000 ms) or the `spark.dynamicAllocation.executorIdleTimeout` (default 60 sec) has been changed from the default values, ensure the condition `spark.dynamicAllocation.executorIdleTimeout > yarn.nodemanager.shuffledata-monitor.interval-ms` remains `true` by updating the necessary flag. ``` [ { "Classification": "yarn-site", "Properties": { "yarn.resourcemanager.decommissioning-nodes-watcher.wait-for-shuffle-data": "true" } }, { "Classification": "spark-defaults", "Properties": { "spark.dynamicAllocation.enabled": "true", "spark.shuffle.service.removeShuffle": "true" } } ] ``` + Managed scaling first removes task nodes and then removes core nodes until it achieves the desired scale-down target capacity. The cluster never scales below the minimum constraints specified in the managed scaling policy. + For clusters that are launched with Amazon EMR 5.x releases 5.34.0 and higher, and 6.x releases 6.4.0 and higher, Amazon EMR Managed Scaling doesn’t scale down nodes that have `ApplicationMaster` for Apache Spark, if there are active stages in the applications running on them. This minimizes job failures and retries, which helps to improve job performance and reduce costs. To confirm which nodes in your cluster are running `ApplicationMaster`, visit the Spark History Server and filter for the driver under the **Executors** tab of your Spark application ID. + While the intelligent scaling with EMR Managed Scaling minimizes shuffle data loss for Spark, there can be instances when transient shuffle data might be not be protected during a scale-down. To provide enhanced resiliency of shuffle data during scale-down, we recommend enabling **Graceful Decommissioning for Shuffle Data** in YARN. When **Graceful Decommissioning for Shuffle Data** is enabled in YARN, nodes selected for scale-down that have shuffle data will enter the **Decommissioning** state and continue to serve shuffle files. The YARN ResourceManager waits until nodes report no shuffle files present before removing the nodes from the cluster. + Amazon EMR version 6.11.0 and higher support Yarn-based graceful decommissioning for **Hive** shuffle data for both the Tez and MapReduce Shuffle Handlers. + Enable Graceful Decommissioning for Shuffle Data by setting `yarn.resourcemanager.decommissioning-nodes-watcher.wait-for-shuffle-data` to `true`. + Amazon EMR version 7.4.0 and higher support Yarn-based graceful decommissioning for Spark shuffle data when the external shuffle service is enabled (enabled by default in EMR on EC2). + The default behavior of the Spark external shuffle service, when running Spark on Yarn, is for the Yarn NodeManager to remove application shuffle files at time of application termination. This may have an impact on the speed of node decommissioning and compute utilization. For long running applications, consider setting `spark.shuffle.service.removeShuffle` to `true` to remove shuffle files no longer in use to enable faster decommissioning of nodes with no active shuffle data. + To minimize Spark shuffle data loss in Amazon EMR version 7.4.0 and higher, consider setting the following flags. + If either the `yarn.nodemanager.shuffledata-monitor.interval-ms` flag (default 30000 ms) or the `spark.dynamicAllocation.executorIdleTimeout` (default 60 sec) has been changed from the default values, ensure that the condition `spark.dynamicAllocation.executorIdleTimeout > yarn.nodemanager.shuffledata-monitor.interval-ms` remains `true` by updating the necessary flag. ``` [ { "Classification": "yarn-site", "Properties": { "yarn.resourcemanager.decommissioning-nodes-watcher.wait-for-shuffle-data": "true" } }, { "Classification": "spark-defaults", "Properties": { "spark.dynamicAllocation.enabled": "true", "spark.shuffle.service.removeShuffle": "true" } } ] ``` If the cluster does not have any load, then Amazon EMR cancels the addition of new instances from a previous evaluation and performs scale-down operations. If the cluster has a heavy load, Amazon EMR cancels the removal of instances and performs scale-up operations. ## Node allocation considerations We recommend that you use the On-Demand purchasing option for core nodes to avoid HDFS data loss in case of Spot reclamation. You can use the Spot purchasing option for task nodes to reduce costs and get faster job execution when more Spot Instances are added to task nodes. ## Node allocation scenarios You can create various scaling scenarios based on your needs by setting up the Maximum, Minimum, On-Demand limit, and Maximum core node parameters in different combinations. **Scenario 1: Scale Core Nodes Only** To scale core nodes only, the managed scaling parameters must meet the following requirements: + The On-Demand limit is equal to the maximum boundary. + The maximum core node is equal to the maximum boundary. When the On-Demand limit and the maximum core node parameters are not specified, both parameters default to the maximum boundary. This scenario isn't applicable if you use managed scaling with node labels and restrict your application processes to only run on `CORE` nodes, because managed scaling scales task nodes to accommodate executor demand. The following examples demonstrate the scenario of scaling cores nodes only. [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html) **Scenario 2: Scale task nodes only ** To scale task nodes only, the managed scaling parameters must meet the following requirement: + The maximum core node must be equal to the minimum boundary. The following examples demonstrate the scenario of scaling task nodes only. [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html) **Scenario 3: Only On-Demand Instances in the cluster ** To have On-Demand Instances only, your cluster and the managed scaling parameters must meet the following requirement: + The On-Demand limit is equal to the maximum boundary. When the On-Demand limit is not specified, the parameter value defaults to the maximum boundary. The default value indicates that Amazon EMR scales On-Demand Instances only. If the maximum core node is less than the maximum boundary, the maximum core node parameter can be used to split capacity allocation between core and task nodes. To enable this scenario in a cluster composed of instance groups, all node groups in the cluster must use the On-Demand market type during initial configuration. This scenario is not applicable if you use managed scaling with node labels and restrict your application processes to only run on `ON_DEMAND` nodes, because managed scaling scales `Spot` nodes to accommodate executor demand. The following examples demonstrate the scenario of having On-Demand Instances in the entire cluster. [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html) **Scenario 4: Only Spot Instances in the cluster** To have Spot Instances only, the managed scaling parameters must meet the following requirement: + On-Demand limit is set to 0. If the maximum core node is less than the maximum boundary, the maximum core node parameter can be used to split capacity allocation between core and task nodes. To enable this scenario in a cluster composed of instance groups, the core instance group must use the Spot purchasing option during initial configuration. If there is no Spot Instance in the task instance group, Amazon EMR managed scaling creates a task group using Spot Instances when needed. This scenario isn't applicable if you use managed scaling with node labels and restrict your application processes to only run on `ON_DEMAND` nodes, because managed scaling scales `ON_DEMAND` nodes to accommodate application process demand. The following examples demonstrate the scenario of having Spot Instances in the entire cluster. [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html) **Scenario 5: Scale On-Demand Instances on core nodes and Spot Instances on task nodes ** To scale On-Demand Instances on core nodes and Spot Instances on task nodes, the managed scaling parameters must meet the following requirements: + The On-Demand limit must be equal to the maximum core node. + Both the On-Demand limit and the maximum core node must be less than the maximum boundary. To enable this scenario in a cluster composed of instance groups, the core node group must use the On-Demand purchasing option. This scenario isn't applicable if you use managed scaling with node labels and restrict your application processes to only run on `ON_DEMAND` nodes or `CORE` nodes. The following examples demonstrate the scenario of scaling On-Demand Instances on core nodes and Spot Instances on task nodes. [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html) **Scenario 6: Scale `CORE` instances for application process demand and `TASK` instances for executor demand.** This scenario is only applicable if you use managed scaling with node labels and restrict application processes to only run on `CORE` nodes. To scale `CORE` nodes based on application process demand and `TASK` nodes based on executor demand, you must set the following configurations at cluster launch: + `yarn.node-labels.enabled:true` + `yarn.node-labels.am.default-node-label-expression: 'CORE'` If you don't specify the `ON_DEMAND` limit and the maximum `CORE` node parameters, both parameters default to the maximum boundary. If the maximum `ON_DEMAND` node is less than the maximum boundary, managed scaling uses the maximum `ON_DEMAND` node parameter to split capacity allocation between `ON_DEMAND` and `SPOT` nodes. If you set the the maximum `CORE` node parameter to less than or equal to the minimum capacity parameter, `CORE` nodes remain static at the maximum core capacity. The following examples demonstrate the scenario of scaling CORE instances based on application process demand and TASK instances based on executor demand. [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html) **Scenario 7: Scale `ON_DEMAND` instances for application process demand and `SPOT` instances for executor demand.** This scenario is only applicable if you use managed scaling with node labels and restrict application processes to only run on `ON_DEMAND` nodes. To scale `ON_DEMAND` nodes based on application process demand and `SPOT` nodes based on executor demand, you must set the following configurations at cluster launch: + `yarn.node-labels.enabled:true` + `yarn.node-labels.am.default-node-label-expression: 'ON_DEMAND'` If you don't specify the `ON_DEMAND` limit and the maximum `CORE` node parameters, both parameters default to the maximum boundary. If the maximum `CORE` node is less than the maximum boundary, managed scaling uses the maximum `CORE` node parameter to split capacity allocation between `CORE` and `TASK` nodes. If you set the the maximum `CORE` node parameter to less than or equal to the minimum capacity parameter, `CORE` nodes remain static at the maximum core capacity. The following examples demonstrate the scenario of scaling On-Demand Instances based on application process demand and Spot instances based on executor demand. [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-allocation-strategy.html) # Understanding managed scaling metrics in Amazon EMR Amazon EMR publishes high-resolution metrics with data at a one-minute granularity when managed scaling is enabled for a cluster. You can view events on every resize initiation and completion controlled by managed scaling with the Amazon EMR console or the Amazon CloudWatch console. CloudWatch metrics are critical for Amazon EMR managed scaling to operate. We recommend that you closely monitor CloudWatch metrics to make sure data is not missing. For more information about how you can configure CloudWatch alarms to detect missing metrics, see [Using Amazon CloudWatch alarms](https://docs.aws.amazon.com//AmazonCloudWatch/latest/monitoring/AlarmThatSendsEmail.html). For more information about using CloudWatch events with Amazon EMR, see [Monitor CloudWatch events](https://docs.aws.amazon.com/emr/latest/ManagementGuide/emr-manage-cloudwatch-events.html). The following metrics indicate the current or target capacities of a cluster. These metrics are only available when managed scaling is enabled. For clusters composed of instance fleets, the cluster capacity metrics are measured in `Units`. For clusters composed of instance groups, the cluster capacity metrics are measured in `Nodes` or `vCPU` based on the unit type used in the managed scaling policy. | Metric | Description | | --- | --- | | [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-metrics.html) | The target total number of units/nodes/vCPUs in a cluster as determined by managed scaling. Units: *Count* | | [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-metrics.html) | The current total number of units/nodes/vCPUs available in a running cluster. When a cluster resize is requested, this metric will be updated after the new instances are added or removed from the cluster. Units: *Count* | | [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-metrics.html) | The target number of CORE units/nodes/vCPUs in a cluster as determined by managed scaling. Units: *Count* | | [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-metrics.html) | The current number of CORE units/nodes/vCPUs running in a cluster. Units: *Count* | | [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-metrics.html) | The target number of TASK units/nodes/vCPUs in a cluster as determined by managed scaling. Units: *Count* | | [\[See the AWS documentation website for more details\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/managed-scaling-metrics.html) | The current number of TASK units/nodes/vCPUs running in a cluster. Units: *Count* | The following metrics indicate the usage status of cluster and applications. These metrics are available for all Amazon EMR features, but are published at a higher resolution with data at a one-minute granularity when managed scaling is enabled for a cluster. You can correlate the following metrics with the cluster capacity metrics in the previous table to understand the managed scaling decisions. | Metric | Description | | --- | --- | | `AppsCompleted` | The number of applications submitted to YARN that have completed. Use case: Monitor cluster progress Units: *Count* | | `AppsPending` | The number of applications submitted to YARN that are in a pending state. Use case: Monitor cluster progress Units: *Count* | | `AppsRunning` | The number of applications submitted to YARN that are running. Use case: Monitor cluster progress Units: *Count* | | ContainerAllocated | The number of resource containers allocated by the ResourceManager. Use case: Monitor cluster progress Units: *Count* | | `ContainerPending` | The number of containers in the queue that have not yet been allocated. Use case: Monitor cluster progress Units: *Count* | | ContainerPendingRatio | The ratio of pending containers to containers allocated (ContainerPendingRatio = ContainerPending / ContainerAllocated). If ContainerAllocated = 0, then ContainerPendingRatio = ContainerPending. The value of ContainerPendingRatio represents a number, not a percentage. This value is useful for scaling cluster resources based on container allocation behavior. Units: *Count* | | `HDFSUtilization` | The percentage of HDFS storage currently used. Use case: Analyze cluster performance Units: *Percent* | | `IsIdle` | Indicates that a cluster is no longer performing work, but is still alive and accruing charges. It is set to 1 if no tasks are running and no jobs are running, and set to 0 otherwise. This value is checked at five-minute intervals and a value of 1 indicates only that the cluster was idle when checked, not that it was idle for the entire five minutes. To avoid false positives, you should raise an alarm when this value has been 1 for more than one consecutive five-minute check. For example, you might raise an alarm on this value if it has been 1 for thirty minutes or longer. Use case: Monitor cluster performance Units: *Boolean* | | `MemoryAvailableMB` | The amount of memory available to be allocated. Use case: Monitor cluster progress Units: *Count* | | `MRActiveNodes` | The number of nodes presently running MapReduce tasks or jobs. Equivalent to YARN metric `mapred.resourcemanager.NoOfActiveNodes`. Use case: Monitor cluster progress Units: *Count* | | `YARNMemoryAvailablePercentage` | The percentage of remaining memory available to YARN (YARNMemoryAvailablePercentage = MemoryAvailableMB / MemoryTotalMB). This value is useful for scaling cluster resources based on YARN memory usage. Units: *Percent* | The following metrics provide information about resources used by YARN containers and nodes. These metrics from the YARN resource manager offer insights into the resources used by containers and nodes running in the cluster. Comparing these metrics to the previous table’s cluster capacity metrics provides a clearer picture of the impact of managed scaling: | Metric | Associated releases | Description | | --- | --- | --- | | `YarnContainersUsedMemoryGBSeconds` | Available to release label 7.3.0 and higher | The consumed container memory \$1 seconds for the publishing period. **Units:** GB \$1 seconds | | `YarnContainersTotalMemoryGBSeconds` | Available to release label 7.3.0 and higher | The total yarn container \$1 seconds for the publishing period. **Units:** GB \$1 seconds | | `YarnContainersUsedVCPUSeconds` | Available to release label 7.5.0 and higher | The consumed container VCPU \$1 seconds for the publishing period. **Units:** VCPU \$1 seconds | | `YarnContainersTotalVCPUSeconds` | Available to release label 7.5.0 and higher | The total container VCPU \$1 seconds for the publishing period. **Units:** VCPU \$1 seconds | | `YarnNodesUsedMemoryGBSeconds` | Available to release label 7.5.0 and higher | The consumed node memory \$1 seconds for the publishing period. **Units:** GB \$1 seconds | | `YarnNodesTotalMemoryGBSeconds` | Available to release label 7.5.0 and higher | The total node memory \$1 seconds for the publishing period. **Units:** GB \$1 seconds | | `YarnNodesUsedVCPUSeconds` | Available to release label 7.3.0 and higher | The consumed node VCPU \$1 seconds for the publishing period. **Units:** VCPU \$1 seconds | | `YarnNodesTotalVCPUSeconds` | Available to release label 7.3.0 and higher | The total node VCPU \$1 seconds for the publishing period. **Units:** VCPU \$1 seconds | ## Graphing managed scaling metrics You can graph metrics to visualize your cluster's workload patterns and corresponding scaling decisions made by Amazon EMR managed scaling as the following steps demonstrate. **To graph managed scaling metrics in the CloudWatch console** 1. Open the [CloudWatch console](https://console.aws.amazon.com/cloudwatch/). 1. In the navigation pane, choose **Amazon EMR**. You can search on the cluster identifier of the cluster to monitor. 1. Scroll down to the metric to graph. Open a metric to display the graph. 1. To graph one or more metrics, select the check box next to each metric. The following example illustrates the Amazon EMR managed scaling activity of a cluster. The graph shows three automatic scale-down periods, which save costs when there is a less active workload. ![\[Graph managed scaling metrics\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/images/Managed_Scaling_Decision.png) All the cluster capacity and usage metrics are published at one-minute intervals. Additional statistical information is also associated with each one-minute data, which allows you to plot various functions such as `Percentiles`, `Min`, `Max`, `Sum`, `Average`, `SampleCount`. For example, the following graph plots the same `YARNMemoryAvailablePercentage` metric at different percentiles, P10, P50, P90, P99, along with `Sum`, `Average`, `Min`, `SampleCount`. ![\[Graph managed scaling metrics with different percentiles\]](http://docs.aws.amazon.com/emr/latest/ManagementGuide/images/Managed_Scaling_Metrics.png)