Platform SIG
summary
Community decides to stop relying on Halyard because it blocks installs and is hard to maintain, and Cleat is effectively unmaintained; the Kubernetes operator still embeds Halyard so work aims to remove halyard or instead ditch the operator and use kustomize or simple kubectl/Helm for day‑one installs. A Go CloudDriver shows ~12 GB → ~3 GB RAM for ~500 accounts, Gradle remote cache cuts CI builds from >60 minutes to ~15–20 minutes, and a mono-repo approach simplifies releases.
AI-written summary · about
chapters (10)
- 0:00Introduction and agenda
- 1:05Halyard and its challenges
- 7:55Operator and Halyard replacement
- 18:06Day one vs day two experience
- 24:52Configuration and validation needs
- 38:28Closing install method discussion
- 40:23Cloud driver refactoring
- 48:21SIG landscape and responsibilities
- 54:00Mono repo proposal
- 1:01:09Summary and next steps
related talks
22mCreating a Micro Open-Source Community with Helm
26mAre We Stuck With Helm Forever? A Landscape of Cloud Native CNF Manag...
01h 03mWorkshop: Migrating from Halyard to Armory OSS Operator for Kubernetes
30mThe state of Package Management on Kubernetes
34mKustomize vs Helm - Civil War
36mOptimizing Resource Usage in Kubernetes
45mBuild Your Kubernetes Operator with the Right Tool!
43mLessons learned writing 10+ Kubernetes Operators