Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Use Ubuntu 24.04 in CI workflows. #248

Open
wants to merge 3 commits into
base: branch-24.12
Choose a base branch
from
Open

Use Ubuntu 24.04 in CI workflows. #248

wants to merge 3 commits into from

Conversation

bdice
Copy link
Contributor

@bdice bdice commented Oct 2, 2024

This PR replaces ubuntu22.04 with ubuntu24.04. This allows us to continue testing endpoints of our support matrix (we still support Ubuntu 20.04). The CI matrix has been reduced enough that I do not see a significant value proposition to keep any jobs with Ubuntu 22.04. It seems more valuable to have extensive testing of Ubuntu 24.04.

Depends on rapidsai/ci-imgs#195.

Contributes to rapidsai/build-planning#74.

Copy link
Member

@jameslamb jameslamb left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for getting this started. I left one suggestion for your consideration.

.github/workflows/wheels-test.yaml Outdated Show resolved Hide resolved
…untu 22.04 instead of 24.04 on CUDA versions older than 12.5.1 (not supported by CUDA).
@bdice
Copy link
Contributor Author

bdice commented Oct 4, 2024

Some notes:

  • I realized we can't use Ubuntu 24.04 in workflows with CUDA versions older than 12.5.1 because no nvidia/cuda images exist (we already excluded those in the ci-imgs matrix, I just hadn't considered the downstream effects).
    • I migrated the affected test jobs to Ubuntu 22.04.
  • We should do our conda builds (C++ and Python) on Rocky Linux 8 because it's safer and better supported.
    • Rocky Linux 8 has the oldest glibc we support, which makes builds safer.
    • This will align with conda-forge as they migrate to glibc 2.28.
    • This also aligns RAPIDS conda builds with RAPIDS wheel builds, which already use Rocky Linux 8.
    • I think we have an old issue or thread about this somewhere -- we have meant to do switch conda builds to the oldest supported glibc in the past.

@bdice
Copy link
Contributor Author

bdice commented Oct 18, 2024

CI testing is complete! Here is the link: rapidsai/build-planning#74 (comment)

Are there any other repositories we should test? I think I got sufficient coverage here, and am happy to help iron out any additional issues that arise after merging.

@bdice bdice self-assigned this Oct 18, 2024
@bdice
Copy link
Contributor Author

bdice commented Oct 18, 2024

For future record, there were a couple of minor hiccups:

Copy link
Member

@jameslamb jameslamb left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks! I'm happy with the matrix changes (they look identical to what I'd previously reviewed) and with the extent of the testing that was done.

I support merging this as-is, will defer to your judgment on the timing.

# arm64
- { ARCH: 'arm64', PY_VER: '3.11', CUDA_VER: '12.0.1', LINUX_VER: 'ubuntu20.04', GPU: 'a100', DRIVER: 'latest', DEPENDENCIES: 'latest' }
nightly:
# amd64
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '11.4.3', LINUX_VER: 'rockylinux8', GPU: 'v100', DRIVER: 'earliest', DEPENDENCIES: 'oldest' }
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '11.8.0', LINUX_VER: 'ubuntu22.04', GPU: 'v100', DRIVER: 'latest', DEPENDENCIES: 'latest' }
- { ARCH: 'amd64', PY_VER: '3.10', CUDA_VER: '11.8.0', LINUX_VER: 'ubuntu20.04', GPU: 'v100', DRIVER: 'latest', DEPENDENCIES: 'latest' }
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why has this been downgraded from 22.04 to 20.04?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I rebalanced the OS version matrix a bit but this is open to change. As usual, our matrix should focus on the endpoints (oldest/newest). There is one job for Ubuntu 22.04 to provide additional coverage.

Would you prefer to make this 24.04?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually, I'm sorry. We cannot make this 24.04 because Ubuntu 24.04 only supports CUDA 12.5+.

Our options are 22.04 or 20.04. Rocky is already covered on the line above so we should keep this on Ubuntu.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For now, I'll revert to 22.04. I think that should be fine.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ergh. I read this again on my terminal (somehow it escaped me when reading the GitHub diff).

If we change this to 22.04, we will have only ARM coverage for 20.04. We should test this endpoint ("earliest supported Ubuntu") on amd64 as well.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants