You do not need a cluster to watch a single server. A cron-driven probe, a handful of counters and an honest log will catch most regressions before users do. The trick is deciding what "healthy" means and writing it down.
Start with three checks: can the service complete a real request, does its certificate still validate, and is the error rate trending up. Everything else is decoration.