Initially, I was planning to use AWS Managed Node Groups with Cluster Autoscaler, but then I started exploring Karpenter.

It sounds promising, but I also care about the simplicity of overall architecture. I know if I use AWS Managed Node Groups the process of upgrading to the new EKS version will be straightforwart, but how it will look like when worker nodes will be managed by Karpenter?

CodePudding user response：

As of now Karpenter deal with this by setting a node expiry. Here's a good discussion about that approach and what it may looks like in the future.