Self-hosted runners stuck in "active" state despite not doing anything

[mhx]

Select Topic Area

Bug

Body

This happened a few hours ago, for the first time in years of using self-hosted runners.

My workflow got stuck with most x86 jobs still in the queue. However, both x86 runners (on two different machines) were idle, i.e. they had successfully finished a job and had not picked up a new job yet.

The Settings/Actions/Runners page was showing both runners as "active". So I stopped the runners, which made them change to "offline", and started them again, which made them "active" again, but on the command line, they were both just saying "Listening for Jobs".

So as a last resort, I tried to set them up from scratch. However, trying to remove the runners yields:

$ ./config.sh remove --token *****

# Runner removal

Failed: Removing runner from the server
Runner ***** is currently running a job and cannot be deleted.

Trying to force remove the runner from the web UI yields a rather unhelpful:

Sorry, there was a problem deleting your runner.

So, at this point, I've run out of ideas.

[micheloliveira-com]

Same happened to me today.

[lingster]

same here, I have tried to uninstall but haven't figured out how to reset everything...

[micheloliveira-com]

After ~20h I was able to remove the stuck runner, but the pipeline still shows as “Started 21h ago” with a failed status. It seems the state wasn't properly updated even after the failure.