Add docs for GHA self-hosted settings#2025
Conversation
|
I'm going to merge this for now. There's a chance we need to expand a bit on the details of GPU runners but for now this is valid information even if it's in a general way for any kind of self-hosted GHA runners. |
|
so can i donate a computer of mine to conda-forge to build some packages for us ;) or is this for non-conda-forge users to setup their own self hosted stuff? |
| github_actions_labels: | ||
| # use Microsoft free runners | ||
| - hosted # [osx or win] | ||
| # Use self-hosted runner with custom label | ||
| - - self-hosted # [linux and aarch64] | ||
| - custom-label # [linux and aarch64] | ||
| # Use self-hosted runner from cirun | ||
| - cirun-openstack-cpu-large # [linux and ppc64le] | ||
| # Use self-hosted gpu runner from cirun | ||
| - cirun-openstack-gpu-large # [linux and x86_64] |
There was a problem hiding this comment.
Could somebody explain how do I interpret this? how do I use this? how can I achieve this output:
- CONFIG: osx_arm64_python3.11.____cpython
UPLOAD_PACKAGES: True
os: macos
runs_on: ['macos-14']macos-14 are github's new arm macos runners
There was a problem hiding this comment.
You can't, this is only for self-hosted runners, or organizations that are not conda-forge. We haven't enabled native osx-arm64 runners with Github. See discussion at #1781
There was a problem hiding this comment.
I'm trying to do it for a not conda-forge feedstock. Would be nice if smithy woudl support that in such cases, loks like most of the building blocks needed for such functionality are there. It's a PITA manually changing back to the specific runner I want every time I rerender. I also want to specify the macos version for the x86_64 runner, my builds hang for a prolonged time sometimes until timeout on macos-latest but usually run without issue on macos-11
There was a problem hiding this comment.
Should we convert this to a conda-smithy issue?
Would help if we can capture what exactly is being tried, what happens now, and what the use case needs to work in the ideal case
Edit: Links to PRs, code, etc. would also be helpful context
6 participants
PR Checklist:
See conda-forge/conda-smithy#1703