Currently if specifying multiple recompute hosts, the requests are loadbalanced between them, which is fine for spreading the load on mutiple machines with automatic reduandancy if n-1 machines go down. It however wont support the usecase where one/multiple machines should only be used in case all others are unavailable. For example: a local recompute as primary and the gaming desktop as failover. It is undesirable to constantly be recomputing while gaming, but during maintenance/downtime of the primary recompute its acceptable.
Currently if specifying multiple recompute hosts, the requests are loadbalanced between them, which is fine for spreading the load on mutiple machines with automatic reduandancy if n-1 machines go down. It however wont support the usecase where one/multiple machines should only be used in case all others are unavailable. For example: a local recompute as primary and the gaming desktop as failover. It is undesirable to constantly be recomputing while gaming, but during maintenance/downtime of the primary recompute its acceptable.