Enable Kubenetes Node Autoscaling (and change Pod count calculations to match)
@f_santos requested we support Kubenetes Node Autoscaling on our Cloud Native Hybrid setups.
This is indeed something we should support as it goes to the very heart of Kubernetes. However our current GET approach is incongruent with this change.
GET currently attempts to calculate pod sizes based on the available resources in the cluster. With autoscaling the available resources is now a moving target.
As such we need to change how Pod counts are calculated and pin them on a different constant. One possible constant is Gitaly.
Task is to explore switching the pod counts to pin based on Gitaly specs and also confirm Node Autoscaling works as expected on GCP / AWS.