Koping with Change: What kops Learned Adopting etcd3, cluster-api and CRDs
summary
kops documents a concrete migration path: move clusters from etcd2 to etcd3 using a new etcd-manager that adds automated backups, cluster resizing and TLS-by-default and survives master roll restarts; replace fragile aggregated API servers with CRDs and controller-runtime to remove vendored api machinery; shim kops instance groups into machine deployments so existing configs keep working while adopting cluster‑api; cut alphas/betas early to speed releases at the cost of up to four cherry‑picks per bugfix.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:17Introductions
- 1:54kops as metaphor
- 2:22kops release - philosophy
- 3:56kops release - reality
- 6:30kops release - the fast-forward
- 10:59kops & etcd3
- 14:21API aggregation - CRDS
- 17:01Cluster API
- 21:29Addons - Bundles & Operators
- 23:52kops as components
- 25:50What we learned!
related talks
35mSIG Cluster Lifecycle (kops)
06mKeynote: Distributed Systems Simplified Using Kubernetes
22mkOps: Recent Advances and the Journey to an OSS Kubernetes Distribution
32metcd V3.6.0 and etc...
43mSelf-hosted, Scale, and Federation with Kubernetes v1.4 and Beyond
36mMigrating from kOps to EKS with Minimal Developer Involvement
24mAutomating Stateful Apps with Kubernetes Operators
29m11 5 years of kubernetes and looking to the future