GHA- ISO Private Ephemeral Runners in K8s with image per step

[DrMerlin]

Hi there- I was really hoping that GHA would announce parity for self-hosted/private runners - to allow the specification of a container image for any given step- just like public runners, and like other competing CI tools allow for with private runners.

This is the biggest limiting factor to deciding to use GHA at scale instead of other products at my company

I realize it would require a kubernetes scheduling operator- but it seems like it shouldn't be too hard to implement.

I have seen a few community approaches to fix this- but think this really needs first party support.

Is this on the road map anywhere?

The recent "ephemeral" and "webhook" changes are nice, I guess, but don't really move the needle for my company. We need image per step.

[rjbell4]

@DrMerlin I was waiting for the "ephemeral" support, and didn't realize that is wasn't on par with an image per step. 😢

[karasowles]

Thanks @DrMerlin - could you post this on the feedback repo so we can keep track of the request post-Universe? Thanks 🙏
https://github.com/github/feedback

Here's the public roadmap, which we'll update as we start work on features:
https://github.com/github/roadmap