Unexpected errors without any logs

[scribejourdan]

What's going on

We've been experiencing this issue that's inconsistent but has been happening very frequently.

Frequent failing workflow	Less frequent failing workflow

I'd like to reiterate that there are no logs around that, absolutely nothing we could find that could point us to a root cause we can work on.

There are a few things we tried, though.

What we did so far

• We cleared GitHub Actions cache as it was 6 times the limit, surprisingly.
• We increased the timeout on the runner. It doesn't get to half of it.

I suppose I can't share a lot but I'll pick up as much as I can that I believe could help.

What we have

• Our runners are all runs-on: ubuntu-22.04 and its variations like ubuntu-22.04-16core.
• A few other spots make use of actions/cache/restore and actions/cache/save steps to handle cache, and that seems to work fine.
• The job that fails frequently has timeout-minutes: 30, and it doesn't get to half of it. The less frequent one has timeout-minutes: 15, and it doesn't top that.
• Not sure if that matters, but we have both Python and NodeJS cache definitions in the same job.
• Our Yarn cache is around 2.2GB last time I checked. Python one is around 41MB.
• The with key that was logged:

  with:
    path: ~/.cache/yarn/
    key: Linux-yarn-<files hash>
    save-always: true
    enableCrossOsArchive: false
    fail-on-cache-miss: false
    lookup-only: false

Etc.

I think it's likely an issue with GHA or even our setup, but we ran out of options almost a month ago and have been struggling without an answer.

Appreciate any help in advance!