What are the differences between image used for GH Runners and Self-hosted runners?

[simpson]

I recently got the ARC runners set working with a lot of trial and error. The issue I'm facing now is that all of my builds work fine when using GH hosted runners (runs-on: ubuntu-latest), but when I shift to using my runner set the builds fail because it seems like there aren't the same packages installed (Ex. Python).

Are the differences in the GH image and the image that ARC runners use documented anywhere to understand the differences? Why not just use the same image? I'm using the latest action-runner image as far as I can tell.

Thanks!

  spec:
    containers:
      - name: runner
        image: ghcr.io/actions/actions-runner:latest
        command: ["/home/runner/run.sh"]
        resources:
          requests:
            memory: "12G"
          limits:
            memory: "16G"

[vpolikarpov-akvelon]

Hi @simpson. GitHub hosted runners are technically virtual machines that bundles a lot of different software in order to be as versatile as possible. You don't need most of it in your own self-hosted container-based runners. So it's up to you to configure containers in accordance with your requirements.

[simpson]

thanks @vpolikarpov-akvelon!

Is there any way to mimic a replica of the GitHub image, or examples of how to manipulate the above yaml to install packages in that container?

[vpolikarpov-akvelon]

We use Hashicorp Packer to build runner virtual machine images in Azure subscription. There is a manual in this repo on how to create machine image yourself.

Since there is a Docker builder for Packer you may try to adapt our workflow and templates to your needs. I can't provide exact instructions nor can guarantee success but in theory it may work.

But I think it's much easier to create your own image from ghcr.io/actions/actions-runner:latest adding only software you really need.

Once your runner image is ready, follow ARC documentation to deploy it to you cluster.

[tleerai]

You can build your own with this dockerfile: https://raw.githubusercontent.com/actions/runner/main/images/Dockerfile

Here is a good article on the subject:
https://some-natalie.dev/blog/kubernoodles-pt-5/