Creating a new cluster and migrating workloads

As you can see, a typical upgrade will involve at least three steps:

1. Upgrading the control plane
2. Upgrading/replacing the worker nodes with more up-to-date AMIs and the kubelet
3. At least upgrading the core components, kube-proxy, coreDNS, and vpc-cni

In this approach, the Pods must first be drained and reallocated to worker nodes as they are replaced. This can lead to interruptions if not managed well. An alternative is to deploy a new cluster and then migrate workloads; this is sometimes referred to as blue/green cluster deployment.

Important note

This will be the least cost-effective approach as you will be paying for two control planes but may be suitable if you want to try to minimize disruption. We will only discuss this approach at a high level in this book as the most common approach is to upgrade the EKS control plane and then the worker nodes using managed worker nodes, greatly reducing cost and complexity...