Just trying to get some information on the best route to setup replication for this DB I've got.

I found this and it seems like the idea we have in mind.

We're looking to have the Main Master be replicated by a Master/Slave and then a final machine replicating from the Master/Slave. That solution sounds wonderful and turning on --log-slave-updates on the Master/Slave machine sounds like what I want.

The question I have is when the Main Master(Master1) goes down we tell the services to write to Master/Slave(Master2) and then what happens with Master1? We rebuild it etc.. but where in the link does it go? Does Master/Slave(Master2) become the new Main Master? And does the Slave only machine get moved to Master/Slave?

Or does Master1 go down and gets rebuild and just becomes Master2(Master/Slave) keeping the Slave machine just that, a slave?

I feel like it's the latter of the 2 options. Just looking for some advice.

Thanks!