Workflow for building container image

[gvegayon]

This pull request introduces a new GitHub Actions workflow for building Docker images and refactors the Dockerfiles to separate dependencies from the main application. The key changes include the creation of a workflow file, modifications to the Containerfile, and the addition of a new Containerfile.dependencies.

Workflow and Dockerfile changes:

• New GitHub Actions workflow:

  • Added .github/workflows/containers.yaml to automate the building and pushing of Docker images. This workflow includes steps to checkout code, retrieve commit messages, determine branch names, and handle Docker caching and login.

• Refactoring Dockerfiles:

  • Modified Containerfile to use a pre-built dependencies image from the container registry, allowing for faster builds and better separation of concerns.
  • Added a new Containerfile.dependencies to define the base image with all necessary dependencies, including Python, R, and CMake installations.

Important

This depends on the cfa-cdcgov runners, which are currently not set for this repository. If you don't want to depend on the cdcgov runner (for now), we can change this to use ghcr.io instead.

[jkislin]

@gvegayon @damonbayer you've been setup with two instances of cfa-cdcgov for this repository.
Please let me know if you experience issues

Important

Please do not invite non-cdc users to be admins of this repository, or to allow non-cdc users to push directly to branches that trigger CI/CD.

[gvegayon]

@gvegayon @damonbayer you've been setup with two instances of cfa-cdcgov for this repository. Please let me know if you experience issues

Important

Please do not invite non-cdc users to be admins of this repository, or to allow non-cdc users to push directly to branches that trigger CI/CD.

@damonbayer and @dylanhmorris, aren't there explicit rules in the repo for that?

attn @zsusswein and @natemcintosh

[gvegayon]

The thing is runnning! But there's an issue with pyarrow. Will investigate.

[gvegayon]

Hey @damonbayer! It turns out that python 3.13 breaks the container (b/c of some stuff with pyarrow). Downgrading to Python 3.12 works.

[natemcintosh]

Hey @damonbayer! It turns out that python 3.13 breaks the container (b/c of some stuff with pyarrow). Downgrading to Python 3.12 works.

I also had some issues with pyarrow when upgrading to 3.13. I was able to fix them by updating from pyarrow 17 to 18.

[damonbayer]

@gvegayon I update pyarrow. Please try Python 3.13 again, if you don't mind.

[gvegayon]

@gvegayon I update pyarrow. Please try Python 3.13 again, if you don't mind.

Running again!

[gvegayon]

Yeah @damonbayer and @natemcintosh, docker is not happy with the change. I would recommend going back to python 3.12 (if you can't think of a straight forward solution) and adding a new issue indicating the need to move to 3.13.

[natemcintosh]

If Damon merged the change to 3.13 into main, did that change also get merged into this branch?. Otherwise this branch will still be trying to use pyarrow 17, no?

[gvegayon]

If Damon merged the change to 3.13 into main, did that change also get merged into this branch?. Otherwise this branch will still be trying to use pyarrow 17, no?

I don't think so. GHA workflows triggered by PRs run on a merge (see here). I can try actively merging main into the Branch, but I think it will be the same thing.

[gvegayon]

@natemcintosh and @damonbayer, it's passing now! I'm still learning how PR-triggered workflows work, but at least is working!

[damonbayer]

Thanks @gvegayon!