← all case studies

Case study 09

Kubernetes upgrades became a checklist

Deprecated-API signals and a staged four-cluster sequence made upgrade risk visible before production.

reliability

The situation

Cluster upgrades are where platform teams get hurt: APIs get removed between versions, workloads break, and the fear of upgrading is how companies end up years behind on unsupported versions. (When I joined, our clusters were years behind — I ran four consecutive version upgrades in my first six weeks just to reach supported ground.)

What I did

I built the practice that turned upgrades from an event into a checklist:

The interesting part

The old scanner was only the start; once the API server itself began surfacing deprecated usage, the dashboarding practice simply moved to the native signal and kept the same flow.

What it changed

Deprecated-API discovery moved onto dashboards, and upgrades now follow the same rehearsed sequence from operations through production.