Enable Hibernation for HyperShift Hosted Clusters

Summary 

Hosted Control Planes add value already by decoupling workloads from control-planes and allowing the control planes to run as payloads on the management clusters via Kubernetes native primitives (e.g., Deployments, pods,..).

Hibernation is really the next step in the game to use the benefits unlocked via the HyperShift topology to get costs down even more!

Background

Previously, we have documented how to scale to 0 workers, something that the use of CAPI makes possible by default with HyperShift. 

Goal 

In addition to scale to 0, we would like to study hibernation options, mainly answer the following: 

• How to scale the control-plane component. 
• What to do during Hibernation? Shut down, or delete underlying infra nodes?
• Should the API-Server stay reachable? 

Ideally, we would have options for Nodepool level hibernation, and cluster-wide hibernation, the semantic of what it would mean when either is enabled should be investigated in this effort.

Theme

Strategic Theme: Maximize the value of HyperShift’s core competencies