GKE supports multiple versions of nodes and today began offering release channels to further simplify version management. There pool and node limits are far greater. Why does AWS require yet another CLI "eksctl" other than to vendor lock-in?
I still have yet to see anything on EKS that can match GKE.
To address your other comments: * EKS has automatic updates on the roadmap, which I imagine is similar to release channels on GKE. * GKE Limits. Yes, they are much larger than EKS, which I guess matters a lot if you have a requirement to scale horizontally to thousands of nodes. * eksctl isn't necessary to use EKS, but they do encourage it. I personally use Terraform and kubectl. It's essentially a wrapper over the AWS CLI. For GKE you would use the gcloud tool. For EKS you can use the AWS CLI or eksctl. It is an open-source tool built by Weaveworks btw: https://eksctl.io/
They've had auto updates for nearly two years. Release channels enable you to pick a minor version and stay there.
Other nice features from GKE
- gVisor is a checkbox
- istio is a checkbox
- knative is a checkbox
- Cloud Run is a checkbox
- Custom CPU/Mem/SSD/GPU/TPU (this is more EC2 v GCE)
- extra hardened OS
- automatic credentials for GKE from Cloud Build
- service account binding between IAM and k8s
- Anthos!
- Labels for understanding costs and multi-tenant billing (internal or external)
GKE is free compared to the nearly $150 / month AWS fee that no one else is charging.
Oh, and the Google Cloud runs on 100% renewables. When will AWS be able to say that?