KubeCeption! A Story of Self-Hosted Kubernetes
summary
Demonstrates a self-hosted control plane where API server, scheduler, controller-manager and kube-proxy run as Deployments/DaemonSets, uses bootkube to bootstrap a replacement control plane, and shows creating HA from one to three nodes and upgrading from 1.4.4 to 1.4.5 using only kubectl-only upgrade commands. Explains that DaemonSets don’t yet do rolling updates so pods are manually deleted, outlines using an etcd operator to migrate seed data, and describes recovery techniques: force-scheduling a scheduler pod with nodeName and user-space pod checkpointing.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:33What is SelfHosted Kubernetes
- 5:47How SelfHosted Kubernetes Works
- 15:59Simplify Lifecycle Management
- 17:03How it Works
- 17:38BootCube
- 21:11Versions
- 27:21Control Plane
- 28:37Pod Network
- 30:47Ledet CD
- 33:16Disaster Recovery
- 33:59Loss of Control Plan
- 34:33Recovery
- 39:17Power Cycling
- 42:12Permanent Loss
- 43:10Whats Next
related talks
34mSelf-Hosted Kubernetes: How and Why [I]
21mSelf-Healing Kubernetes
44mDemystifying Kubernetes internals
37mkubeadm Cluster Creation Internals: From Self-Hosting to Upgradability and HA
01h 31mInside the Kube: A Guided Tour of Kubernetes Cluster Setup
39mIntroduction to Kubernetes
06mKubernetes as the Control Plane for the Hybrid Cloud
01h 01mKubernetes – going beyond the basics