One assumes that you are implementing an FFT=2, which implies all your VMs survive this double failure.
Another host in the cluster will take the responsibility of master, and another will take responsibility of backup.
This may take a little longer than normal as the new master and backup have to learn about the configuration and rebuild the directory contents (this is what the backup node already has for speedy conversion to master should the master fail).
Other than that, no adverse affects.
http://cormachogan.com
