Skip to main content

Set Up Starlake with Airflow or Dagster

Starlake supports Airflow and Dagster as orchestration engines. The starlake-docker repository provides Docker Compose files that install and configure Starlake, the web UI, and the chosen orchestrator in a single command. This setup is designed for developers and data engineers who want a ready-to-use pipeline environment without manual orchestrator configuration.

The core project is open source: starlake-ai/starlake.

Deploy Starlake with Airflow or Dagster via Docker Compose​

Prerequisites​

  • Docker and Docker Compose installed and running
  • Git installed (to clone the repository)

Step-by-Step Setup​

  1. Clone the starlake-docker repository
git clone https://github.com/starlake-ai/starlake-docker.git
  1. Navigate to the docker directory
cd starlake-docker/docker
  1. Start the stack

For Airflow:

docker compose up

For Dagster:

docker compose -f docker-compose-dagster.yml up
  1. Open the Starlake UI -- Navigate to http://localhost in your browser.

To run on a different port, set the SL_UI_PORT environment variable:

SL_UI_PORT=8080 docker compose up

Stop the Starlake Docker Compose Stack​

To stop all services, run in the same directory:

docker compose down

Mount External Starlake Projects in Docker​

If you have existing Starlake projects and want to access them from the Docker setup, mount their parent folder as an NFS volume.

Steps for macOS​

  1. Run setup_mac_nfs.sh to expose your folder via NFS. Modify the root folder to share if needed (default: /user).
  2. In docker-compose.yml, comment out - external_projects_data:/external_projects.
  3. Uncomment - starlake-prj-nfs-mount:/external_projects.
  4. At the bottom of the file, modify the volume path to point to your parent folder.

Expected Folder Structure​

The mounted parent folder must contain one or more Starlake project directories, each with a metadata/ subdirectory:

my_parent_folder/
├── sl_project_1/
│   └── metadata/
│       └── ...
├── sl_project_2/
│   └── metadata/
│       └── ...

If you have multiple parent folders, create a separate volume for each one.

Known Limitations​

  • You cannot mount Starlake projects directly under /external_projects. Place them inside a subfolder.
  • This feature has been tested on macOS and Linux only.

Deploy Starlake on Kubernetes​

Kubernetes deployment is available via Helm charts. Contact the Starlake engineering team on Discord for the latest Helm chart and deployment instructions.

Deploy Starlake on AWS, GCP, or Azure​

Cloud deployment via Terraform is available. Contact the Starlake engineering team on Discord for the latest Terraform scripts targeting AWS, GCP, or Azure.

Frequently Asked Questions​

How do I set up Starlake with Airflow?​

Clone the starlake-docker repository, navigate to the docker folder, and run docker compose up. Airflow is automatically installed and configured. Access the Starlake UI at http://localhost.

How do I set up Starlake with Dagster instead of Airflow?​

Use the Dagster-specific compose file: docker compose -f docker-compose-dagster.yml up. Dagster replaces Airflow as the orchestrator in this configuration.

Can I deploy Starlake on Kubernetes?​

Kubernetes deployment is available. Contact the Starlake engineering team on Discord to get the latest Helm chart.

What orchestration tools does Starlake support?​

Starlake supports Airflow and Dagster. Both are pre-configured in the Docker Compose setup.

How do I mount external projects in the Starlake Docker setup?​

On macOS, run setup_mac_nfs.sh to expose your folder via NFS. Then modify the docker-compose.yml to use the NFS volume mount instead of the default volume. External project folders must contain the Starlake project structure (metadata/ etc.).

How do I change the port for Starlake UI?​

Set the SL_UI_PORT environment variable before running Docker Compose. Example: SL_UI_PORT=8080 docker compose up.