Skip to main content

Validate Anyscale Runtime in custom Docker images

View Markdown

Validate Anyscale Runtime in custom Docker images

This article explains how to test whether the Anyscale Runtime is properly injected into your custom Docker image using an Anyscale job. Use this validation in your CI/CD pipeline or as a manual check after building a custom image.

Prerequisites​

Before running this validation, ensure you've built your custom Docker image following the requirements in Requirements for an Anyscale container image. This validation works with both uv and pip for managed dependencies.

Solution​

Create a Python script that checks for the Anyscale Runtime, then run it as an Anyscale job using your custom image.

1. Create the validation script​

Create a file named validate_runtime.py with the following content:

import ray

@ray.remote
def func():
    from ray import anyscale
    return hasattr(ray, 'anyscale')

if __name__ == "__main__":
    result = ray.get(func.remote())
    print(f"Anyscale Runtime installed: {result}")

2. Submit the validation job​

Submit the job using the Anyscale CLI with your custom image:

anyscale job submit \
  --image-uri <your-custom-image-uri> \
  --entrypoint "python validate_runtime.py" \
  --working-dir . \
  --runtime-env '{"pip": []}'

Replace <your-custom-image-uri> with your actual image URI. If the Anyscale Runtime is properly configured, the job outputs:

Anyscale Runtime installed: True

If the output shows False, the Anyscale Runtime isn't properly configured in your image.

Troubleshooting​

If the validation returns False, review your Docker image build process. Ensure you're following the requirements in Requirements for an Anyscale container image and copying the Anyscale Runtime files correctly. See Tutorial: Build a custom container image for a complete example Dockerfile.