Fixed an outage caused by the cluster database filling its disk and taking the management console down with it

What happened

The cluster's configuration database ran out of disk space. When it has no room to write it stops accepting changes entirely, which took the management console offline along with it — so the outage also removed the usual tool for diagnosing an outage.

Impact

Platform management was unavailable on 2025-10-14 and changes could not be applied until the database was recovered.

Resolution

Recovered directly on the cluster nodes rather than through the console, then given headroom and routine maintenance so the disk cannot silently fill again.