Emit a warning message when t2.nano or t2.micro is set for *instanceType

[mumoshu]

Regardless of the fact that t2.medium being the minimum default instance type in kube-aws these days, I believe some people are still trying to use t2.micro for nodes by modifying cluster.yaml and resulting their clusters to not functional.

I've recently received a question in the k8s slack about why kube-aws up didn't create a working cluster and one of the causes was t2.micro.
I'm going to add validation to forbid using t2.micro to stop trapping new-comers and to reduce the amount of support required to learn k8s via using kube-aws.

I'll leave this open for a while to hear from the community before I actually add the validation.

Also see #138 and #184

[valichek]

I think it would be enough to have some warning in cluster.yaml. And maybe some description of possible issues.

[mumoshu]

Thanks for the comment @valichek!
AFAIK, using t2.micro almost consistently result in non-functional nodes.
So it isn't an "possible" issue but almost a persistent issue.
Just curious but are you still wanting to try t2.micro?

[valichek]

@mumoshu I'm running minimal cluster using t2.micro for controller/etcd nodes and t2.nano for workers and it's cheaper. I don't have any issues on startup (0.9.1). We use it as test/learning dev environment and don't run heavy jobs, so we still don't have such issues. And of course, we do understand that we cannot use it for serious tasks.

[mumoshu]

@valichek I'm surprised it is working for you! Thanks for the info.
As I'm not intended to break existing use-case as long as it is working, I'll just add warnings to somewhere then.

[mumoshu]

Btw warnings already exist in cluster.yaml as of today
https://github.com/coreos/kube-aws/blob/master/config/templates/cluster.yaml#L75
Probably I need to add louder warnings somewhere?