Skip to content

Replace Buildkite CI with GitHub Actions self-hosted runners#160

Merged
ChrisRackauckas merged 3 commits intoSciML:masterfrom
ChrisRackauckas-Claude:replace-buildkite-with-gha
Mar 19, 2026
Merged

Replace Buildkite CI with GitHub Actions self-hosted runners#160
ChrisRackauckas merged 3 commits intoSciML:masterfrom
ChrisRackauckas-Claude:replace-buildkite-with-gha

Conversation

@ChrisRackauckas-Claude
Copy link
Copy Markdown
Contributor

Summary

  • Migrate GPU tests from Buildkite to GitHub Actions using self-hosted runners with the gpu label (Tesla T4 on arctic1)
  • Add .github/workflows/GPU.yml for GPU tests on [self-hosted, Linux, X64, gpu]
  • Delete .buildkite/pipeline.yml

Part of the SciML-wide migration from Buildkite to GitHub Actions self-hosted runners. See https://github.com/ChrisRackauckas/InternalJunk/issues/16 for runner infrastructure details.

Changes

The new workflow:

  • Triggers on push/PR to master (ignoring docs/**)
  • Uses concurrency groups to cancel redundant runs
  • Runs on [self-hosted, Linux, X64, gpu] runners
  • Sets GROUP=GPU environment variable for julia-runtest
  • Includes coverage processing and Codecov upload
  • 60-minute timeout (up from Buildkite's 30min to account for setup time)

Test plan

  • Verify GPU workflow triggers on PR
  • Verify GPU tests pass on self-hosted runner
  • Verify coverage upload works

🤖 Generated with Claude Code

ChrisRackauckas and others added 3 commits March 18, 2026 14:59
Migrate GPU tests from Buildkite to GitHub Actions using self-hosted
runners with the `gpu` label (Tesla T4 on arctic1).

- Add .github/workflows/GPU.yml for GPU tests on [self-hosted, Linux, X64, gpu]
- Delete .buildkite/pipeline.yml

Part of the SciML-wide migration from Buildkite to GitHub Actions
self-hosted runners. See ChrisRackauckas/InternalJunk#16
for runner infrastructure details.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The test failed because it landed on demeter4 which has broken GPU
drivers (NVML version mismatch). Use gpu-t4 label to target only
arctic1 runners with working Tesla T4.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
test/GPU/Project.toml had compat `PreallocationTools = "0.4"` but the
package is at v1.1.2. The Pkg.develop call in the test setup fails
because v1.1.2 doesn't satisfy the 0.4 compat bound.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
@ChrisRackauckas ChrisRackauckas merged commit e1b8c96 into SciML:master Mar 19, 2026
13 of 16 checks passed
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.

2 participants