That is the intended behavior. When you import the EKS cluster as an EKS cluster, Rancher will take a snapshot of configuration. When it sees a difference in the intended state and the actual state, it will reconcile it. Rancher is supposed to “sync back” changes you make from the AWS console, but there is something happening here that is not doing that. I won’t speculate on that.
The easiest way to fix this would be the following two options:
1. Remove the node group from the Rancher UI. That will set the desired state from Rancher and it won’t recreate the node group.
2. Import the cluster as a “generic” cluster instead of an EKS cluster. Rancher won’t reconcile in the same way and wont recreate nodes.