Package Operator: Large-Scale Day-2 Operations
summary
SREs at Red Hat build Package Operator to run day‑2 ops inside customer OpenShift clusters at scale (thousands, approaching 10,000) without exfiltrating cluster data or granting broad permissions. It introduces ObjectSet/ObjectDeployment CRDs that aggregate heterogeneous resource status, stores releases as a package as OCI image, executes phased probe-driven rollouts with incremental canaries and automatic rollback history, and can scale sets to zero to archive revisions. Demo on kind shows blocked rollouts for bad images and a readable success/failure history; it is running in limited production use.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:02SRE update failure scenario
- 5:15Kubernetes deployment problems
- 7:20Why tools fail at scale
- 11:37Package Operator concept
- 17:59Design goals
- 18:45Live demo
- 27:01Production use and plans
- 27:24Q&A
related talks
19mHow to build an operator that doesn't break prod
39mBuilding Production-Ready Kubernetes Operators
24mExclude Human – Continuous Deployment and OpenShift
28mOpenShift for the Enterprise or What Could Possibly Go Wrong
22mOpenShift, or Kubernetes on steroids
30mGitOps: Keeping It Light, Bright, and Deployed Right with OpenShift!
44mPlatform Engineering: Raise the Level of Abstraction using Kubernetes Operators
43mLessons learned writing 10+ Kubernetes Operators