Say I have the following scenario:
A fly app deployed with the Balanced autoscaling strategy, min 3, max say 10. There are three regions in the pool, let’s say NA, EMEA, and APAC.
Starting out I understand the 3 minimum replicas are spread across the regions, one each. Now, what happens when there’s an enormous usage spike in EMEA, and the 10 maximum replicas have been exhausted. The docs say Balanced will move instances from other regions. Does this run the risk of draining the other two regions completely? I’d want to maintain at least one replica per region to avoid large spikes in latency.