Getting Started with Apache Spark and Apache Polaris With Delta and Iceberg

June 17, 2026 ยท View on GitHub

This getting started guide provides a docker-compose file to set up Apache Spark with Apache Polaris using the new Polaris Spark Client.

The Polaris Spark Client enables manage of both Delta and Iceberg tables using Apache Polaris.

A Jupyter notebook is started to run PySpark, and Polaris Python client is also installed to call Polaris APIs directly through Python Client.

Build the Spark Client Jar and Polaris image

If Spark Client Jar is not presented locally under plugins/spark/v4.0/spark/build/2.13/libs, please build the jar using

  • ./gradlew assemble -- build the Polaris project and skip the tests.

If a Polaris image is not already present locally, build one with the following command:

./gradlew \
  :polaris-server:assemble \
  :polaris-server:quarkusAppPartsBuild --rerun \
  -Dquarkus.container-image.build=true

Run the docker-compose file

To start the docker-compose file, run this command from the repo's root directory:

docker-compose -f plugins/spark/v4.0/getting-started/docker-compose.yml up

This will spin up 2 container services

  • The polaris service for running Apache Polaris using an in-memory metastore
  • The jupyter service for running Jupyter notebook with PySpark

NOTE: Starting the container first time may take a couple of minutes, because it will need to download the Spark 4.0.2. When working with Delta, the Polaris Spark client requires delta-io >= 4.2.0, and it requires at least Spark 4.0.2.

Run with AWS access setup

If you want to interact with S3 bucket, make sure you have the following environment variables setup correctly in your local env before running the docker-compose file.

AWS_ACCESS_KEY_ID=<your_access_key>
AWS_SECRET_ACCESS_KEY=<your_secret_key>

Access the Jupyter notebook interface

In the Jupyter notebook container log, look for the URL to access the Jupyter notebook. The url should be in the format, http://127.0.0.1:8888/lab?token=<token>.

Open the Jupyter notebook in a browser. Navigate to notebooks/SparkPolaris.ipynb

If the above url doesn't work, try to replace 127.0.0.1 with localhost, for example: http://localhost:8888/lab?token=<token>.

Run the Jupyter notebook

You can now run all cells in the notebook or write your own code!