Construct a centralized observability platform for Apache Spark on Amazon EMR on EKS utilizing exterior Spark Historical past Server

Monitoring and troubleshooting Apache Spark purposes change into more and more advanced as corporations scale their knowledge analytics workloads. As knowledge processing necessities develop, enterprises deploy these purposes throughout a number of Amazon EMR on EKS clusters to deal with numerous workloads effectively. Nevertheless, this strategy creates a problem in sustaining complete visibility into Spark purposes operating throughout these separate clusters. Information engineers and platform groups want a unified view to successfully monitor and optimize their Spark purposes.

Though Spark supplies highly effective built-in monitoring capabilities by way of Spark Historical past Server (SHS), implementing a scalable and safe observability resolution throughout a number of clusters requires cautious architectural concerns. Organizations want an answer that not solely consolidates Spark utility metrics however extends its options by including different efficiency monitoring and troubleshooting packages whereas offering safe entry to those insights and sustaining operational effectivity.

This submit demonstrates the right way to construct a centralized observability platform utilizing SHS for Spark purposes operating on EMR on EKS. We showcase the right way to improve SHS with efficiency monitoring instruments, with a sample relevant to many monitoring options resembling SparkMeasure and DataFlint. On this submit, we use DataFlint for example to show how one can combine extra monitoring options. We clarify the right way to accumulate Spark occasions from a number of EMR on EKS clusters right into a central Amazon Easy Storage Service (Amazon S3) bucket; deploy SHS on a devoted Amazon Elastic Kubernetes Service (Amazon EKS) cluster; and configure safe entry utilizing AWS Load Balancer Controller, AWS Non-public Certificates Authority, Amazon Route 53, and AWS Shopper VPN. This resolution supplies groups with a single, safe interface to watch, analyze, and troubleshoot Spark purposes throughout a number of clusters.

Overview of resolution

Contemplate DataCorp Analytics, a data-driven enterprise operating a number of enterprise items with numerous Spark workloads. Their Monetary Analytics crew processes time-sensitive buying and selling knowledge requiring strict processing occasions and devoted assets, and their Advertising Analytics crew handles buyer habits knowledge with versatile necessities, requiring a number of EMR on EKS clusters to accommodate these distinct workload patterns. As their Spark purposes develop in quantity and complexity throughout these clusters, knowledge and platform engineers battle to take care of complete visibility whereas sustaining safe entry to monitoring instruments.

This state of affairs presents a super use case for implementing a centralized observability platform utilizing SHS and DataFlint. The answer deploys SHS on a devoted EKS cluster, configured to learn occasions from a number of EMR on EKS clusters by way of a centralized S3 bucket. Entry is secured by way of Load Balancer Controller, AWS Non-public CA, Route 53, and Shopper VPN, and DataFlint enhances the monitoring capabilities with extra insights and visualizations. The next structure diagram illustrates the parts and their interactions.

The answer workflow is as follows:

1. Spark purposes on EMR on EKS use a customized EMR Docker picture that features DataFlint JARs for enhanced metrics assortment. These purposes generate detailed occasion logs containing execution metrics, efficiency knowledge, and DataFlint-specific insights. The logs are written to a centralized Amazon S3 location by way of the next configuration (be aware particularly the configurationOverrides part). For extra data, discover the StartJobRun information to discover ways to run Spark jobs and evaluate the StartJobRun API reference.

{
  "identify": "${SPARK_JOB_NAME}", 
  "virtualClusterId": "${VIRTUAL_CLUSTER_ID}",  
  "executionRoleArn": "${IAM_ROLE_ARN_FOR_JOB_EXECUTION}",
  "releaseLabel": "emr-7.2.0-latest", 
  "jobDriver": {
    "sparkSubmitJobDriver": {
      "entryPoint": "s3://${S3_BUCKET_NAME}/app/${SPARK_APP_FILE}",
      "entryPointArguments": [
        "--input-path",
        "s3://${S3_BUCKET_NAME}/data/input",
        "--output-path",
        "s3://${S3_BUCKET_NAME}/data/output"
      ],
       "sparkSubmitParameters": "--conf spark.driver.cores=1 --conf spark.driver.reminiscence=4G --conf spark.kubernetes.driver.restrict.cores=1200m --conf spark.executor.cores=2  --conf spark.executor.situations=3  --conf spark.executor.reminiscence=4G"
    }
  }, 
  "configurationOverrides": {
    "applicationConfiguration": [
      {
        "classification": "spark-defaults", 
        "properties": {
          "spark.driver.memory":"2G",
          "spark.kubernetes.container.image": "${AWS_ACCOUNT_ID}.dkr.ecr.${AWS_REGION}.amazonaws.com/${EMR_REPO_NAME}:${EMR_IMAGE_TAG}",
          "spark.app.name": "${SPARK_JOB_NAME}"
          "spark.eventLog.enabled": "true",
          "spark.eventLog.dir": "s3://${S3_BUCKET_NAME}/spark-events/"
         }
      }
    ], 
    "monitoringConfiguration": {
      "persistentAppUI": "ENABLED",
      "s3MonitoringConfiguration": {
        "logUri": "s3://${S3_BUCKET_NAME}/spark-events/"
      }
    }
  }
}

2. A devoted SHS deployed on Amazon EKS reads these centralized logs. The Amazon S3 location is configured within the SHS to learn from the central Amazon S3 location by way of the next code:

env:
  - identify: SPARK_HISTORY_OPTS
    worth: "-Dspark.historical past.fs.logDirectory=s3a://${S3_BUCKET}/spark-events/"

3. We configure Load Balancer Controller, AWS Non-public CA, a Route 53 hosted zone, and Shopper VPN to securely entry the SHS UI utilizing an internet browser.
4. Lastly, customers can entry the SHS net interface at https://spark-history-server.instance.inside/.

Yow will discover the code base within the AWS Samples GitHub repository.

Stipulations

Earlier than you deploy this resolution, make certain the next stipulations are in place:

Arrange a standard infrastructure

Full the next steps to arrange the infrastructure:

1. Clone the repository to your native machine and set the 2 setting variables. Substitute with the AWS Area the place you wish to deploy these assets.

git clone git@github.com:aws-samples/sample-centralized-spark-history-server-emr-on-eks.git
cd sample-centralized-spark-history-server-emr-on-eks
export REPO_DIR=$(pwd)
export AWS_REGION=

2. Execute the next script to create the frequent infrastructure. The script creates a safe digital personal cloud (VPC) networking setting with private and non-private subnets and an encrypted S3 bucket to retailer Spark utility logs.

cd ${REPO_DIR}/infra
./deploy_infra.sh

3. To confirm profitable infrastructure deployment, open the AWS CloudFormation console, select your stack, and verify the Occasions, Assets, and Outputs tabs for completion standing, particulars, and checklist of assets created.

Arrange EMR on EKS clusters

This part covers constructing a customized EMR on EKS Docker picture with DataFlint integration, launching two EMR on EKS clusters (datascience-cluster-v and analytics-cluster-v), and configuring the clusters for job submission. Moreover, we arrange the mandatory IAM roles for service accounts (IRSA) to allow Spark jobs to jot down occasions to the centralized S3 bucket. Full the next steps:

1. Deploy two EMR on EKS clusters:

cd ${REPO_DIR}/emr-on-eks
./deploy_emr_on_eks.sh

2. To confirm profitable creation of the EMR on EKS clusters utilizing the AWS CLI, execute the next command:

aws emr-containers list-virtual-clusters 
    --query "virtualClusters[?state=='RUNNING']"

3. Execute the next command for the datascience-cluster-v and analytics-cluster-v clusters to confirm their respective states, container supplier data, and related EKS cluster particulars. Substitute with the ID of every cluster obtained from the list-virtual-clusters output.

aws emr-containers describe-virtual-cluster 
    --id 

Configure and execute Spark jobs on EMR on EKS clusters

Full the next steps to configure and execute Spark jobs on the EMR on EKS clusters:

1. Generate customized EMR on EKS picture and StartJobRun request JSON information to run Spark jobs:

cd ${REPO_DIR}/jobs
./configure_jobs.sh

The script performs the next duties:

• Prepares the setting by importing the pattern Spark utility spark_history_demo.py to a delegated S3 bucket for job execution.
• Creates a customized Amazon EMR container picture by extending the bottom EMR 7.2.0 picture with the DataFlint JAR for extra insights and publishing it to an Amazon Elastic Container Registry (Amazon ECR) repository.
• Generates cluster-specific StartJobRun request JSON information for datascience-cluster-v and analytics-cluster-v.

Overview start-job-run-request-datascience-cluster-v.json and start-job-run-request-analytics-cluster-v.json for extra particulars.

2. Execute the next instructions to submit Spark jobs on the EMR on EKS digital clusters:

aws emr-containers start-job-run 
--cli-input-json file://${REPO_DIR}/jobs/start-job-run/start-job-run-request-datascience-cluster-v.json
aws emr-containers start-job-run 
--cli-input-json file://${REPO_DIR}/jobs/start-job-run/start-job-run-request-analytics-cluster-v.json

3. Confirm the profitable era of the logs within the S3 bucket:

aws s3 ls s3://emr-spark-logs--/spark-events/

You will have efficiently arrange an EMR on EKS setting, executed Spark jobs, and picked up the logs within the centralized S3 bucket. Subsequent, we are going to deploy SHS, configure its safe entry, and visualize the logs utilizing it.

Arrange AWS Non-public CA and create a Route 53 personal hosted zone

Use the next code to deploy AWS Non-public CA and create a Route 53 personal hosted zone. This may present a user-friendly URL to connect with SHS over HTTPS.

cd ${REPO_DIR}/ssl
./deploy_ssl.sh

Arrange SHS on Amazon EKS

Full the next steps to construct a Docker picture containing SHS with DataFlint, deploy it on an EKS cluster utilizing a Helm chart, and expose it by way of a Kubernetes service of kind LoadBalancer. We use a Spark 3.5.0 base picture, which incorporates SHS by default. Nevertheless, though this simplifies deployment, it leads to a bigger picture measurement. For environments the place picture measurement is essential, contemplate constructing a customized picture with simply the standalone SHS element as an alternative of utilizing the entire Spark distribution.

1. Deploy SHS on the spark-history-server EKS cluster:

cd ${REPO_DIR}/shs
./deploy_shs.sh

2. Confirm the deployment by itemizing the pods and viewing the pod logs:

kubectl get pods --namespace spark-history
kubectl logs  --namespace spark-history

3. Overview the logs and make sure there aren’t any errors or exceptions.

You will have efficiently deployed SHS on the spark-history-server EKS cluster, and configured it to learn logs from the emr-spark-logs-- S3 bucket.

Deploy Shopper VPN and add entry to Route 53 for safe entry

Full the next steps to deploy Shopper VPN to securely join your consumer machine (resembling your laptop computer) to SHS and configure Route 53 to generate a user-friendly URL:

1. Deploy the Shopper VPN:

cd ${REPO_DIR}/vpn
./deploy_vpn.sh

2. Add entry to Route 53:

cd ${REPO_DIR}/dns
./deploy_dns.sh

Add certificates to native trusted shops

Full the next steps so as to add the SSL certificates to your working system’s trusted certificates shops for safe connections:

1. For macOS customers, utilizing Keychain Entry (GUI):
   1. Open Keychain Entry from Functions, Utilities, select the System keychain within the navigation pane, and select File, Import Objects.
   2. Browse to and select ${REPO_DIR}/ssl/certificates/ca-certificate.pem, then select the imported certificates.
   3. Develop the Belief part and set When utilizing this certificates to All the time Belief.
   4. Shut and enter your password when prompted and save.
   5. Alternatively, you may execute the next command to incorporate the certificates in Keychain and belief it:

sudo safety add-trusted-cert -d -r trustRoot -k /Library/Keychains/System.keychain "${REPO_DIR}/ssl/certificates/ca-certificate.pem"

2. For Home windows customers:
   1. Rename ca-certificate.pem to ca-certificate.crt.
   2. Select (right-click) ca-certificate.crt and select Set up Certificates.
   3. Select Native Machine (admin rights required).
   4. Choose Place all certificates within the following retailer.
   5. Select Browse and select Trusted Root Certification Authorities.
   6. Full the set up by selecting Subsequent and End.

Arrange Shopper VPN in your consumer machine for safe entry

Full the next steps to put in and configure Shopper VPN in your consumer machine (resembling your laptop computer) and create a VPN connection to the AWS Cloud:

1. Obtain, set up, and launch the Shopper VPN utility from the official obtain web page on your working system.
2. Create your VPN profile:
   1. Select File within the menu bar, select Handle Profiles, and select Add Profile.
   2. Enter a reputation on your profile. Instance: SparkHistoryServerUI
   3. Browse to ${REPO_DIR}/vpn/client_vpn_certs/client-config.ovpn, select the certificates file, and select Add Profile to save lots of your configuration.
3. Choose your newly created profile, select Join, and await the connection affirmation to determine the VPN connection.

Once you’re related, you should have safe entry to the AWS assets in your setting.

Securely entry the SHS URL

Full the next steps to securely entry SHS utilizing an internet browser:

1. Execute the next command to get the SHS URL:

https://spark-history-server.instance.inside/

2. Copy this URL and enter it into your net browser to entry the SHS UI.

The next screenshot reveals an instance of the UI.

3. Select an App ID to view its detailed execution data and metrics.

4. Select the DataFlint tab to view detailed utility insights and analytics.

DataFlint shows numerous useful metrics, together with alerts, as proven within the following screenshot.

Clear up

cd ${REPO_DIR}/
./cleanup.sh

Conclusion

On this submit, we demonstrated the right way to construct a centralized observability platform for Spark purposes utilizing SHS and improve SHS with efficiency monitoring instruments like DataFlint. The answer aggregates Spark occasions from a number of EMR on EKS clusters right into a unified monitoring interface, offering complete visibility into your Spark purposes’ efficiency and useful resource utilization. Through the use of a customized EMR picture with efficiency monitoring device integration, we enhanced the usual Spark metrics to achieve deeper insights into utility habits. In case your setting makes use of a mixture of EMR on EKS, Amazon EMR on EC2, or Amazon EMR Serverless, you may seamlessly lengthen this structure to combination the logs from EMR on EC2 and EMR Serverless in the same method and visualize them utilizing SHS.

Though this resolution supplies a sturdy basis for Spark monitoring, manufacturing deployments ought to contemplate implementing authentication and authorization. SHS helps customized authentication by way of javax servlet filters and fine-grained authorization by way of entry management lists (ACLs). We encourage you to discover implementing authentication filters for safe entry management, configuring user- and group-based ACLs for view and modify permissions, and organising group mapping suppliers for role-based entry. For detailed steering, check with Spark’s net UI safety documentation and SHS security measures.

Whereas AWS endeavors to use greatest practices for safety inside this instance, every group has its personal insurance policies. Please make certain to make use of the precise insurance policies of your group when deploying this resolution as a place to begin for implementing centralized Spark monitoring in your knowledge processing setting.

Concerning the authors

Sri Potluri is a Cloud Infrastructure Architect at AWS. He’s keen about fixing advanced issues and delivering well-structured options for numerous prospects. His experience spans throughout a variety of cloud applied sciences, offering scalable and dependable infrastructures tailor-made to every venture’s distinctive challenges.

Suvojit Dasgupta is a Principal Information Architect at AWS. He leads a crew of expert engineers in designing and constructing scalable knowledge options for AWS prospects. He makes a speciality of growing and implementing modern knowledge architectures to deal with advanced enterprise challenges.