so while you can have failover, it is a single point of failure (however briefly) for writes