EKS Cluster Upgrades
What you'll learn: How to safely upgrade your EKS cluster's Kubernetes version while maintaining application availability.
The Upgrade Challenge
Think of upgrading your EKS cluster like renovating a building while people are still living in it. You need to upgrade the foundation (control plane) first, then carefully update each floor (node groups) without disrupting the residents (your applications).
Control Plane Upgrades
AWS manages the EKS control plane upgrade for you. When you initiate an upgrade through the AWS Console, CLI, or Infrastructure-as-Code tools, AWS performs a rolling update of the control plane components across multiple availability zones. This process typically takes 20-40 minutes and is non-disruptive to running workloads.
Critical rule: You can only upgrade one minor version at a time (e.g., 1.26 → 1.27 → 1.28). You cannot skip versions.
Node Group Upgrades
After upgrading the control plane, you must upgrade your node groups. This is where disruption can occur:
- Managed Node Groups: AWS handles the rolling update, cordoning and draining nodes before replacement
- Self-Managed Node Groups: You control the upgrade strategy (rolling update, blue/green deployment)
Node groups can run up to two minor versions behind the control plane, giving you flexibility in timing.
Version Compatibility Matrix
The kubelet on worker nodes must be at most two minor versions behind the API server. For example:
- Control plane: 1.28 ✓
- Nodes: 1.28, 1.27, or 1.26 ✓
- Nodes: 1.25 ✗ (too old)
Best Practices
Test upgrades in non-production environments first. Review Kubernetes changelogs for deprecated APIs. Update critical add-ons (VPC CNI, CoreDNS, kube-proxy) before node upgrades. Use PodDisruptionBudgets to protect critical workloads during node drains.
Key Takeaway: EKS upgrades require sequential version progression—control plane first, then node groups—with careful attention to version compatibility and add-on updates to ensure zero-downtime deployments.