My AI Server Wouldn’t Restart. So Another AI SSH’d In and Fixed It.
An upgrade yesterday uncovered tasks from seven months ago still running on my AI server. Not causing problems. Just never cleaned up.
When AI software upgrades, it needs to drain: finish pending work before restarting. Those old tasks blocked the drain. 5 cascading restarts. Broken config. Server that couldn’t stay up.
The fix came from a second AI server I run. It SSH’d into the broken one and repaired the config. Two AI instances, one fixing the other.
After sorting it out, I documented the full procedure: how to detect stale tasks before upgrading, what order to run things, what to do if the drain gets stuck. That document became a reusable procedure — all 4 of my AI servers can call on it for any future upgrade.
The next server upgrade, an hour later, went perfectly.
This is what I didn’t expect about running AI long-term: it builds institutional knowledge. The failure became the procedure.
OpenClaw is a self-hosted AI assistant you run on your own server.
I teach a class on setting up and getting the most from OpenClaw — details at themeperks.com/openclaw-course/.
