Self-Healing Kubernetes
summary
Describes running a Kubernetes control plane as pods on itself to survive failures, using kubelet static pod manifests for bootstrap, leader election for idle replicas, and a small checkpointer that snapshots API server manifests to disk and launches a bootstrap API server after master reboot or API-server death. Shows upgrades as image bumps, scaling masters with kubectl, and integration with Tectonic/CoreOS and Prometheus for observability — self-hosted control plane recovers automatically in tested reboots.
AI-written summary · about
chapters (8)
- 0:00Introduction to talk
- 2:08Kubernetes failure preparation
- 5:49Self-hosted Kubernetes basics
- 9:45Benefits of self-hosting
- 11:34Multi-master failure scenarios
- 14:26Single-master failure handling
- 17:57Future development ideas
- 20:25Closing and Q&A
related talks
28mThe Magic of Kubernetes Self-Healing Capabilities
34mSelf-Hosted Kubernetes: How and Why [I]
44mKubeCeption! A Story of Self-Hosted Kubernetes
37mSelf-Healing Applications with Docker and Kubernetes
19mKeynote: Kubernetes: Simple to Manage Anywhere
26mOperating kube-apiserver Without Hiccups
40mOperating a Self-Healing Bare-Metal Kubernetes Platform at Global Scale
37mManaging and Running Multiple Kubernetes Clusters in Hybrid Setups