[EKS] [Pod Eviction Order]: Pod Eviction during upgrades

When using ALB ingress controller with clusterIP it's important that alb ingress controller is not evicted along with the applications pods, this avoids any impact to the ongoing traffic. Either ingress controller needs to be evicted first or the last but not together, this is because alb ingress controller drains the pod ip from AWS ALB and when it's not running at the time of eviction causes disruption in service. It would be very useful to allow users to specify pod eviction order to deal with such scenarios.

Start by reviewing EKS upgrade behavior, Kubernetes pod eviction semantics, and the ALB ingress controller's pod-IP draining described in the issue. Define how users would express an eviction order and how to verify that ingress and application pods are not evicted together during upgrades.

Written by the indexing model from the issue text.