Error: other side closed

[iliapnmrv]

Run softprops/action-gh-release@v2
  with:
    tag_name: v2.5.1
    body_path: distribution/whatsnew/whatsnew-ru-RU
    files: android/app/build/outputs/apk/release/apk-v2.5.1-release.apk
 
👩‍🏭 Creating new GitHub release for tag v2.5.1...
⬆️ Uploading apk-v2.5.1-release.apk...
Error: other side closed

I have release v2.5.1-rc.0, then I want to create stable release v2.5.1
My release gets created in GitHub releases (without artifact) but pipeline fails with Error: other side closed

Do you have any ideas?

[paulo-ferraz-oliveira]

👋 I also use this action in nightly builds and have the following errors happen, aside from the one you mention:

• Error: write EPIPE (also mentioned in Upload fails with Error: write EPIPE #482)
• Error: write ECONNRESET

I wonder if all these are somewhat related, if there could be a retry in case of error, or if the issues should somehow be grouped together.

[AymericFerreira]

Hello, I also have this error since recently.
I think it is because I create 2 parallel jobs and they try to access this resource or something else at the same time.
Everytime I retry the failed job it works, and everytime my CI works it is because my jobs have difference in time.

Is there a more elegant way to do this other than serializing my jobs ?

Regarding the EPIPE error that Paulo mention I confirm that I also have this, but I am not sure they are related.

[paulo-ferraz-oliveira]

👋 fwiw, I don't run any of my jobs in parallel, and retries also seem to fix the issues consistently.

[mrutkows]

Also seeing this on matrix builds (i.e Error: other side closed)... is there a means to auto-retry (as re-running "failed jobs" manually kind of defeats the purpose of having the automation where each build takes 3-4 mins.)?

[vorporeal]

This has been happening increasingly frequently for us, and it's at the end of a 40-minute build, so re-runs are quite painful.

Some sort of internal retry logic would be great. At the very least, would love to know if anyone has figured out a good trick for running the step again if it fails the first time.

[manuelleduc]

Hello,
I'm experiencing the same with parallel builds too.