My CI Runner Was Killed by My Own Script: The Dark Side of Cleanup

TL;DR AI
2 min readKey summary
A self-hosted GitHub Actions runner stopped taking jobs and showed offline, even though the host and other containers were still running.
There was no kernel crash or OOM event; the issue was a cleanup script that deleted files the runner still depended on.
The deletions hit the runner’s _work directory, causing the service to exit quietly and break CI availability.
The incident is a reminder that broad cleanup automation on CI machines can remove operational files and take runners offline.
