Hi Kfir, very interesting article!

A couple of questions:

1. Why kops and not EKS? you need to upgrade k8s manually, patches, monitoring of controle plane ...

especially after they reduced price to 50%, it costs 77$ now I think.

2. How often interruption of spot instances happens on production cluster with 'capacity-optimized' strategy?

3. You using windows AMIs spot instances in production? I always had a headache with windows AMIs, CA not working correctly for this a lot of times.