Failure to save the cache should not fail the entire run

[thetorpedodog]

Description:

When running a workflow, there may be an issue which causes the cache not to be saved. This should not cause the entire run to be reported as a failure.

Action version:

v4.5.0

Platform:

• Ubuntu
• macOS

Runner type:

• Hosted

Tools version:

Independent of Python version.

Repro steps:

1. Include a setup-python step in a workflow with caching enabled.
2. Have it happen to run at a time when the cache servers are acting flaky.

Example failed run: https://github.com/single-cell-data/TileDB-SOMA/actions/runs/3960808578/attempts/1

Expected behavior:

If the run finishes successfully, but the cache cannot be uploaded to the cache server, the failure to create the cache entry should be ignored.

Actual behavior:

When the cache fails to upload, the run is failed despite completing successfully in all other respects.

[e-korolevskii]

Hello @thetorpedodog! Thank you for reporting this!
We will investigate the issue and get back to you as soon as we have updates

[thetorpedodog]

For comparison, a failure to upload with the github/cache action does not fail the run: https://github.com/single-cell-data/TileDB-SOMA/actions/runs/3961705977/jobs/6787460842

Post job cleanup.
/usr/local/bin/gtar --posix -cf cache.tzst --exclude cache.tzst -P -C /Users/runner/work/TileDB-SOMA/TileDB-SOMA --files-from manifest.txt --delay-directory-restore --use-compress-program zstdmt
Warning: Failed to save: uploadChunk (start: 0, end: 33554431) failed: Cache service responded with 503
Cache saved with key: libtiledbsoma-build-dist-macos-12-3.10-79892bce4ca84bea3b28c1364692c753f17e1b0711f57db30aef7a6297fb0420

It seems like the Python setup action needs to just ignore the failure to cache (and probably report it) the same way that the github/cache action does.

Thank you for the quick response.

[e-korolevskii]

Thanks for the clarification, I'm sure it will help.

[dmitry-shibanov]

Hello @thetorpedodog. We released a new version of the action with changes to prevent the step from falling if there are issues with the cache services.

For now I'm going to close the issue.