Aggregated API Server Pattern overview

A Kubernetes extension mechanism where a custom API server registers itself with the main kube-apiserver via an APIService resource. Requests to the custom API group are proxied to the aggregated server, which can use any backing store (Postgres, Redis, etc.) instead of etcd. This enables high-volume mutable state (e.g., pull request comments, pipeline logs) to live in a relational database while still being accessible through the standard Kubernetes API surface — kubectl, RBAC, admission webhooks, and watch semantics all work transparently. Krate uses this pattern for its 23 data-plane resource kinds.