Once we've completed a multi-master failover stack, there's really only one task remaining to prove that it does as we claim. We need to both test the failover process and also modify one or two settings to get it as low as possible, while also doing so safely and avoiding disruptive flapping between nodes.
This recipe will explain what tests we can perform on our cluster to do this and which settings we can modify to wring as much failover performance out of a multi-master environment as possible.
