There is a default variable here that sets nodepool health checks to EC2.

I am running into an issue when cycling server nodes where the autoscaling group thinks the node is healthy, but the kube apiserver won't come up (or comes up after ec2 determines the node healthy). I believe this could be fixed by setting the server autoscaling group health check to ELB. This would ensure the kubeapi server is up and running on the node before determining the node health in AWS.

I believe this should be the default behavior and this comment makes me think that this was the original intention to make ELB the default for the server nodes. However, flat out changing this default value could be considered a breaking change? Or maybe this would be a bug fix?

Either way, I'd like to submit a PR to make this at least configurable by the top level module. If it is deemed that we can change the default value, I'd be happy to just submit that change as well.