When you create a managed EKS nodegroup you must specify subnets, does it mean if I specify subnets from different AZs nodes from this group will be scaled across subnets' AZs? Or should I create separate node group with a single subnet for every AZ? What's the correct way to get cluster with multi-AZ nodes?
Also, when I create a EKS cluster it says you must specify at least two subnets from different AZs but what if I want to create a single-AZ cluster? What's the point of having subnets from two AZs in this case?
Unless you are doing something special you should probably keep it simple and run across AZ. This is a good blog post that goes into some details about pros and cons and reasons. As per the "at least two subnets" that is because of how the EKS control plane works: the EKS control plane is managed by AWS and it surfaces to the user by means of two ENIs that connect the API server to your own VPC. For HA reasons they are two and connected to two separate subnets. However, because of how VPC networking works ALL subnets can communicate with each others so regardless of where you deploy your worker nodes they will ALL be able to connect to the control plane. In other words, these are two orthogonal things (i.e. the two control plane subnets and the subnets for the worker nodes).
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With