Intro to Control Plane Engineering
summary
Kubernetes control plane needs deliberate sizing: etcd must run a three-node minimum for RAFT consensus, events should be separated from cluster state so split etcd for events is recommended, and API servers are fronted by an external load balancer. Scalability limits track pods more than nodes — expect trouble at high density (around 100 pods per node) and with high pod-creation or heavy status-polling rates. Self-hosted control plane patterns can automate upgrades but require careful bootstrap work.
AI-written summary · about
chapters (7)
- 0:00Start
- 5:18Building Blocks
- 13:30Approach...
- 15:43Starting point
- 16:15Cluster State
- 19:13Scheduler and Controller Manager
- 28:06Self Hosted Kubernetes
related talks
37mKubernetesception: How to run a 1000 control planes
34mScaling Kubernetes up to 1,000 Control Planes - Dario Tranchitella
34mPanel: Improving and Managing Kubernetes at Scale
34mBecoming an Expert of Diagnosing a Kubernetes Control Plane Problem
32mLaying the Foundation: Real World Kubernetes Deployment Patterns
41mKubernetes the Very Hard Way
28mThe Kubernetes Control Plane for Busy People Who Like Pictures
36mNot All Managed Kubernetes Are Created Equal: A Deep Dive into Control Plane Architectures