Node Image Upgrades
What you'll learn: How to update the operating system and container runtime on your Kubernetes nodes without taking down your applications.
The Challenge of Living Infrastructure
Your cluster's nodes are computers running an operating system and a container runtime (like containerd). Just like your laptop needs security patches and updates, these node images need regular refreshes. But here's the catch: you can't just restart them all at once without disrupting your workloads.
How Node Image Upgrades Work
Node image upgrades replace the underlying OS and runtime on your nodes while keeping your applications running. Think of it like renovating a hotel one floor at a time while guests stay in other rooms.
The process typically follows this pattern:
- Cordoning — Mark a node as unschedulable so no new pods land there
- Draining — Gracefully evict existing pods, which move to healthy nodes
- Updating — Replace the node's OS image and container runtime
- Uncordoning — Return the refreshed node to active duty
Managed Kubernetes services automate much of this. In GKE, you can enable node auto-upgrade tied to release channels. In AKS, node image upgrades happen automatically or manually per node pool. EKS managed node groups let you trigger image updates that perform rolling replacements.
Why This Matters
Node images contain critical security patches, kernel updates, and new runtime versions. Regular upgrades keep your foundation secure and compatible with newer Kubernetes features. Without proper upgrade procedures, you risk downtime or security vulnerabilities.
The key is rolling updates — upgrading a few nodes at a time while ensuring sufficient capacity remains for your pods to reschedule elsewhere, respecting PodDisruptionBudgets throughout.
Key Takeaway: Node image upgrades modernize your cluster's foundation (OS and runtime) using rolling replacements that drain nodes gracefully, keeping applications available while infrastructure updates happen beneath them.