Automate updating GitHub Action runners (datasource/github-runners)

[HonkingGoose]

Tell us more.

Can we automate updating the source of truth for the GitHub Action runners?

Current situation

Right now, our source of truth is the actions/runner-images repository's README.md file, that has a table of all supported versions (beta/stable).

Renovate can't read/understand the Markdown file, and Markdown formatting is easy to break by accident.

Here are our Renovate docs to add/promote/drop a GitHub runner.

Upstream creates new source of truth JSON file

Would it help if we had a runner-images.json file on the upstream repository? I'm thinking of something like the node-js-schedule.json file that we update with Renovate. Here's a update PR from Renovate:

• fix(data): automatic update of static data #25487

Edit: I'm asking the upstream maintainers if they want to create a source of truth file for us here:

• Add a JSON file listing the images with beta/stable/EOL dates actions/runner-images#9484

Work for upstream maintainers

The upstream maintainers would need to:

• decide on a JSON format
• update the JSON file when they're releasing a new beta version
• update the JSON file when they're marking a version as stable
• update the JSON file when they're removing a version

Work for us

• Read the JSON format
• Make Renovate configuration to update the JSON file on our side

Related issues/PRs:

• Mark macOS 14 runners as stable in our datasource/github-runners when out of beta #27228
• feat(datasource/github-runners): add macos 14 beta #27292

What do you think?

What do you maintainers think?

[rarkins]

Can it be part of a custom datasource? Not sure if we quite support enough right now but ideally this type of thing can be programmatic without needing to maintain a new datasource and/or repo

[HonkingGoose]

Thank you for the explanation! 😄

---

Edit: even if we drop the datasource, or rewrite it, we probably still need a JSON file upstream as source of truth?

[rarkins]

In theory we could fetch and parse markdown into results as long as it's structured. If it's unstructured then the only way is manual maintenance of it

[HonkingGoose]

Hmm, here's the edit history for the upstream. Markdown file. I see the last two commits alone made big changes to the ordering and naming of the versions. I suppose that's too unstructured for Renovate to rely on.

[rarkins]

Could be hard to automate then?

[HonkingGoose]

I think automating around the upstream Markdown table is going to be hard.

[viceice]

if they have such a file, then I would simply download it on the fly at runtime, like we do for the azure workflow stuff.

[HonkingGoose]

They don't have a JSON file like that Node.js version JSON file. Should I ask them to make such a file? It would make our life as maintainers easier, vs more work for upstream maintainers...

[viceice]

sure 🤗

[HonkingGoose]

I'm asking the upstream maintainers about creating a JSON file that we can use here:

• Add a JSON file listing the images with beta/stable/EOL dates actions/runner-images#9484

Let's see what they think. 😄