Self-Hosted Kubernetes: How and Why [I]
summary
Self-hosted Kubernetes runs the control plane as Deployments and DaemonSets so upgrades and autoscaling use native Kubernetes primitives; bootkube and kubeadm are used to bootstrap a temporary static control plane that pivots to self-hosted components, and Tectonic runs this in production. A lightweight checkpoint daemon stores local static manifests so a master node reboot or API-server outage can resurrect critical pods, and operators encode safe rollouts (API server → scheduler → controller manager) for zero‑downtime upgrades.
AI-written summary · about
chapters (13)
- 0:00Intro
- 0:28Who this talk is for
- 1:47A talk in three parts
- 3:14Desirable control plane properties
- 5:25Simplified node management
- 6:24So how do we select masters?
- 6:44Streamlined lifecycle management
- 7:47How does self-hosted Kubernetes work?
- 8:36How it works: Bootstrapping
- 15:17How it works: Upgrades
- 16:33How it works: Disaster recovery
- 20:09Bootkube Recover
- 23:38Automated operations
related talks
44mKubeCeption! A Story of Self-Hosted Kubernetes
21mSelf-Healing Kubernetes
24mGoing the Kubernetes Way: Self-hosted or a Managed Solution
19mKeynote: Kubernetes: Simple to Manage Anywhere
29mMinikube Intro
44mDemystifying Kubernetes internals
01h 01mDeploying Kubernetes, a Container Cluster Manager
32mLaying the Foundation: Real World Kubernetes Deployment Patterns