Technology · Ebook

Kured: Safe Automatic Node Reboots for Kubernetes

by Shriira Press

Kured — the KUbernetes REboot Daemon — is a CNCF sandbox project that solves one sharp operational problem: when a node's OS needs a reboot to finish a patch, who reboots it, and how do you do that across a fleet without taking the cluster down? Kured runs a DaemonSet on every node that watches for the reboot signal your package manager already leaves behind, then cordons and drains the node, takes a cluster-wide lock so no two nodes reboot at once, reboots, and uncordons on the way back. This free book covers the reboot sentinel, the lock that serializes reboots, the cordon-drain-reboot-uncordon cycle, guardrails like reboot windows and Prometheus alerts, and running Kured well in production.

Contents

1. 1Preface
2. 2Chapter 1 — What Kured Is
3. 3Chapter 2 — The Reboot Sentinel
4. 4Chapter 3 — Coordinating Reboots with the Lock
5. 5Chapter 4 — Cordon, Drain, Reboot, Uncordon
6. 6Chapter 5 — Guardrails: Windows, Alerts, and Blocking
7. 7Chapter 6 — Kured in Practice