WASM CI runs out of disk space #8355
Description
Re:
The scheduled WASM CI failed just now: https://github.com/agda/agda/actions/runs/21553501613/job/62105858259
Copilot says it is a disk space issue (see details below for its analysis).
I had seen the CI failure before recently and it went away by retrying.
Shots in the dark: Could be that one can get lucky with having more disk space by being assigned another machine, but this is just speculation.
Maybe missing caches could also be a problem? The job now only runs weekly, its caches could by then have been deleted by GitHub because we have a cap of 10GB total for caches.
@andy0130tw : Could you have a look?
(Btw, I send you an invitation to the @agda organisation just now...)
Details
The job failed due to a disk space issue:error: creating directory '/nix/store/tmp-3288-1709094913/x/pkgs/by-name/wh/whoogle-search': No space left on device
Solution:
- This is an out-of-space error in the GitHub Actions runner environment, not a code or configuration problem in your repository.
- Try the following:
- Reduce the size of dependencies or cached build artifacts if possible.
- Add cleanup steps early in your job to remove unnecessary files before disk-intensive steps.
- Split the job into smaller jobs, or use a self-hosted runner with more storage if this happens regularly.
Example: Add a cleanup step in your workflow YAML before the build step:
- name: Clean up unnecessary files
run: |
sudo rm -rf $HOME/.cache
sudo rm -rf /usr/share/dotnet
sudo rm -rf /opt/ghc
sudo apt-get cleanIf you control caching, review what is persisted between jobs and reduce the amount if possible.
Summary: The underlying issue is a full disk on the runner—optimize or clean up to free space, or request more resources from GitHub Actions if needed. No code changes in your repository are needed to fix this.