Skip to content

README.md

Latest commit

 

History

History
98 lines (63 loc) · 3.65 KB

README.md

File metadata and controls

98 lines (63 loc) · 3.65 KB

conda2docker

Build and publish a Docker image from a Conda environment YAML file

1. Setup

Fork this repository, or add all its contents into an existing repository. You will need a GitHub account to continue.

2. Customize the environment

env.yaml defines the Conda environment. Refer to Conda docs for examples on how to make changes to this file.

Any changes to env.yaml will automatically trigger a build of the image on GitHub Container Registry as an image tagged with the repository's name and owner.

Note

The first push in a repository may fail due to lack of GitHub permissions on a package that hasn't been created yet. Subsequent pushes should succeed.

3. Customize the automatic build

Trigger

The automatic build is defined in docker.yml, a GitHub Actions workflow. By default, it is set to run:

  1. on pushes to any branch that change certain files
  2. on manual invocation from the workflow webpage (repository > Actions > docker)

You can tweak this to use other workflow triggers.

Platform

By default, the build is done only for the native platform of the GitHub runner, linux/amd64. If you want to build for other platforms, you can pass --platform to ./build --platform or more directly by modifying the docker buildx build command within that script. If doing this, you may need to enable emulation by uncommenting the line for docker/setup-qemu-action@v3 in docker.yml.

Note that building for non-native platforms introduces complications and can slow the build process. See the Docker's official guide for more information: Building multi-platform images

Common issues

Error when pulling/running: docker: no matching manifest …

This happens when the image was built for a platform that differs from the machine you are running Docker on. For example, this would happen if the image is built for linux/amd64 and you are using a Mac running on Apple silicon (M1/M2/etc.). Some ways to address this:

  1. Build the image for your machine's platform using emulation (instructions). Typically this means a slower and more error-prone build process but faster run times.
  2. Pull the image with the image's intended platform explicitly (e.g. docker pull --platform=linux/amd64) and run using emulation. Typically this means a faster build process but slower run times.
  3. Build the image locally. Typically this is the fastest combination of build and run times, but the build process can be error-prone and it isn't automated via GitHub Actions.

docker pull permission denied

If you repository is public, log out and try again.

docker logout ghcr.io

If your repository is private, log in, follow prompts, and try again.

docker login ghcr.io

docker push permission denied

Configure GITHUB_TOKEN to have read and write permissions.

References

Based on work in https://github.com/nextstrain/docker-base.