Autoscaled MacOS VMs runner manager deployment POC

This issue is to track the POC deployment of the new autoscaler with Orka.

Current idea:

Deploy two runner managers in Orka's Kubernetes cluster
Handle capacity limit by setting concurrency = capacity/2 on each manager
Doesn’t require VPN between runner & VM
Need VPN for Ops, but easy to connect manually
Takes some resources away from builds (but only 2 containers so not a lot). Need to see how to setup k8s limits and how we can bin-pack a runner and VMs, and the impact from those 2 containers on available CPU

Edited Feb 22, 2021 by Adrien Kohlbecker

To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information