Node Pools not able to use t3.small in AWS EKS Auto Mode

[mridul-dazn]

We have facing issue while creating custom nodepool with t3.small in AWS EKS Auto Mode.

Working with t3.medium instance type however facing issue only with t3.small. We have also adjusted the CPU and Memory limits and requests for your daemon set and tried still it is picking up the t3.medium size.

Upon checking the nodeclaim that was created by EKS automode Karpenter while creating the node I have seen below config which does not contain small instance size could be the one of the possible reasons it is not picking up the small sized instances as Automode always make sures that enough resources are always available.

[mikestef9]

Auto mode only launches instance types that are:

• > 1 CPU
• not nano, micro or small

We are working on getting this info added to docs

[aslatter]

It would be helpful if this sort of failure condition were surfaced in some way. I created a node-pool which was restricted to a 1-CPU-instance-type by matching directly (using various combinations of conditions on eks.amazonaws.com/instance-cpu and eks.amazonaws.com/instance-size while I tried to debug things) and both the node-pool and node-class reported themselves as "ready" but never created any node-claims in response to scheduling pressure.

Maybe k8s events related to the node-pool object? Or some field in the "status" block of the node-pool.