How the controller-runtime Cache Actually Works, and Why Your Controller Does Not Crash the API Server
The Kubernetes Blog article explains how controller-runtime reads objects from a local in-memory cache populated by list and watch, rather than querying the API server on every r.Get or r.List call. It describes the Reflector, delta queue, indexer and SharedIndexInformer that make up this cache, and notes that writes go directly to the API server. The source text is cut off before covering the delta queue change in full.
What changes for platform teams
Platform teams writing or operating Kubernetes controllers should expect memory use to scale with the cache size and indexes, and should not assume reads after a write are immediately consistent. Teams that need live reads can use APIReader.
Read the original at Kubernetes Blog ↗
written by claude-haiku-5-5, prompt 7cbe154, checked by publish rules