It would be nice to have multiple upstreams for replicas in a stack, A simple RR would be fine. There should be less downtime (or none) on a container restart if you did a rolling upgrade this way.
It would be nice if it could do upstream health checks and exclude things from the RR, but that feels like a separate issue.
It would be nice to have multiple upstreams for replicas in a stack, A simple RR would be fine. There should be less downtime (or none) on a container restart if you did a rolling upgrade this way.
It would be nice if it could do upstream health checks and exclude things from the RR, but that feels like a separate issue.