CI on our runners sometimes throws out error regarding loop devices

Description

We are running several runners in parallel on bare metal. There are enough of loop devices, but still it fails.

https://paste.armbian.com/xebiladozu.yaml

Environment

None

Checklist

hide

Activity

Show:

Ricardo Pardini June 26, 2023 at 11:18 AM

Hmm I don’t think this is “done” at all… we’ve tried a bit, only

Igor Pecovnik March 13, 2023 at 9:34 PM

literally just write a .env file with FIXED_LOOP_DEV_NUMBER="$(( i + 0))" or such
[10:26 PM]
during deploy, although I can't grok what is START and STOP. (edited)
[10:26 PM]
but you see my meaning.

Igor Pecovnik March 13, 2023 at 9:23 PM

GH self-hosted runner supposedly supports a .env file in the root of the runner installation dir, that file is loaded and values passed as env vars to every job under that runner. So we could add FIXED_LOOP_DEV_NUMBER=1 (2, 3, etc for each runner) and build would use it instead of losetup -f if present.
[10:21 PM]
dunno if true / works / good idea
[10:22 PM]
also unsure if real problem is actually sharing of loop, I'd need to see logs from 2 builds actually starting at similar times to see