A demo application illustrating how to run a Java Spark job on EMR Serverless using LocalStack. The sample builds a Maven project, uploads the JAR to S3, creates an EMR Serverless application, and submits a Spark job — all locally without connecting to AWS.

• A valid LocalStack for AWS license. Your license provides a LOCALSTACK_AUTH_TOKEN to activate LocalStack.
• Docker
• localstack CLI
• awslocal CLI
• Java and Maven

make check

If not using awslocal, add the following profile to ~/.aws/config:

[profile localstack]
region=us-east-1
output=json
endpoint_url = http://localhost:4566

And to ~/.aws/credentials:

[localstack]
aws_access_key_id=test
aws_secret_access_key=test

Build the Java Spark application (java-demo-1.0.jar is included; you can also rebuild it):

make install

export LOCALSTACK_AUTH_TOKEN=<your-auth-token>
make start

Before creating the EMR Serverless job, we need to create a JAR file containing the Java code. We have the java-demo-1.0.jar file in the current directory. Alternatively, you can create the JAR file yourself by following the steps below.

cd hello-world
mvn package

Create an S3 bucket and upload the JAR file:

cd ..
make deploy

Creates an EMR Serverless Spark application (which will run Spark 3.3.0), starts it, and submits a job that runs the HelloWorld class:

make run

The Spark job submits with:

• Entry point: the JAR file in S3
• --class HelloWorld
• Spark logs written to s3://<bucket>/logs/ (specified in the logUri parameter)

To stop the EMR Serverless application after the job completes:

awslocal emr-serverless stop-application --application-id $APPLICATION_ID

This code is available under the Apache 2.0 license.