Skip to content

Fix docs build: update dead URLs, fix GPU OOM, fix runner labels#326

Merged
ChrisRackauckas merged 3 commits into
SciML:mainfrom
ChrisRackauckas-Claude:fix-docs-build
Mar 26, 2026
Merged

Fix docs build: update dead URLs, fix GPU OOM, fix runner labels#326
ChrisRackauckas merged 3 commits into
SciML:mainfrom
ChrisRackauckas-Claude:fix-docs-build

Conversation

@ChrisRackauckas-Claude

Copy link
Copy Markdown
Contributor

Summary

  • Update old *.sciml.ai subdomain URLs to docs.sciml.ai/* format across all comparison pages, inverse problems page, index, and PINN showcase (17 URLs fixed across 7 files)
  • Add QuadDIRECT.jl to linkcheck_ignore for transient connectivity issues from self-hosted runner
  • Fix PINN GPU showcase: use coarser grid (0.1 vs 0.05) and add CUDA.reclaim()/GC.gc() to reduce GPU memory pressure (previous build failed with OOM at 99.94% GPU usage)
  • Fix aggregate docs runner label from large to gpu — no runner matched large, causing a 24h timeout with no runner pickup
  • Update README badge from Buildkite to GitHub Actions

Context

PR #325 migrated from Buildkite to GitHub Actions, but the first CI run failed with:

  1. :linkcheck errors — many old *.sciml.ai subdomain URLs (e.g., diffeq.sciml.ai, sensitivity.sciml.ai) are dead
  2. :example_block error — PINN GPU example ran out of GPU memory (31.7 GiB / 31.7 GiB)
  3. Aggregate job timeout[self-hosted, Linux, X64, large] runner label matched no available runner

Test plan

  • Documentation CI build passes (linkcheck + example blocks)
  • Aggregate documentation build picks up a runner and completes
  • Verify all updated URLs resolve correctly

🤖 Generated with Claude Code

ChrisRackauckas and others added 3 commits March 25, 2026 07:18
- Update old *.sciml.ai subdomain URLs to docs.sciml.ai/* format
  (benchmarks, diffeq, mtk, sensitivity, linearsolve, nonlinearsolve,
  optimization, neuralpde, catalyst, datadriven, diffeqflux)
- Add QuadDIRECT.jl to linkcheck_ignore (transient connectivity issues)
- Fix PINN GPU example: use coarser grid (0.1 vs 0.05) and add
  CUDA.reclaim()/GC.gc() to reduce GPU memory pressure
- Fix aggregate docs runner label from `large` to `gpu` (no runner
  matched `large`, causing 24h timeout)
- Update README badge from Buildkite to GitHub Actions

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The aggregate docs build succeeded but failed at `aws s3 sync` because
the self-hosted GPU runner (arctic1-1) doesn't have the AWS CLI
installed. Add a step to install it if not present.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The aggregate build succeeds but S3 deploy fails on PRs because
secrets aren't available. Split into separate build and deploy steps
with deploy conditional on push events only.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@ChrisRackauckas ChrisRackauckas merged commit d67a5e2 into SciML:main Mar 26, 2026
4 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