ser8 read-only watcher: alert on the lock-class crashloop signature on kai-server #315

Closed
opened 2026-06-12 11:20:02 +00:00 by coilysiren · 1 comment
Owner

Mitigation 2 from #311. The o11y outage ran 45 days because the alerting stack was the casualty. ser8 holds a read-only kubeconfig for kai-server and sits in a different failure domain, so it watches for the deterministic signature: recent k3s restart plus CrashLoopBackOff plus a lock-class log error (flock.lock, boltdb open, address already in use, Another server instance). The alert carries the diagnosis and reap commands, not a generic pod-unhealthy. Interim shape: a small poller until the full o11y stack on ser8 takes over the job.

Mitigation 2 from #311. The o11y outage ran 45 days because the alerting stack was the casualty. ser8 holds a read-only kubeconfig for kai-server and sits in a different failure domain, so it watches for the deterministic signature: recent k3s restart plus CrashLoopBackOff plus a lock-class log error (flock.lock, boltdb open, address already in use, Another server instance). The alert carries the diagnosis and reap commands, not a generic pod-unhealthy. Interim shape: a small poller until the full o11y stack on ser8 takes over the job.
Author
Owner

Backlog burndown 2026-06-17: closing low-priority (P3/P4) to bring the open count to a manageable level. Nothing lost — reopen if this resurfaces. Batch tag: burndown-2026-06.

Backlog burndown 2026-06-17: closing low-priority (P3/P4) to bring the open count to a manageable level. Nothing lost — reopen if this resurfaces. Batch tag: `burndown-2026-06`.
coilysiren 2026-06-17 08:22:32 +00:00
Sign in to join this conversation.
No description provided.