Recently I’ve set up Percona xtraDB cluster on Rackspace (3 nodes cloud servers). All communication between nodes was directed
through our internal private network (not Rackspace service network). The problem was that whole cluster (without any load) was degraded
from time to time (usually a few times a day) and then re-sync without any action from my side. Since we are going to go big (in number of nodes)
I’ve set up multicast communication according to:
I wasn’t able to identify problem (no OS/firewall limitation) and then I’ve turn off multicast communication. Now cluster
is stable. I’m waiting for Rackspace answer is there any multicast limitation on their network. Is there anyone who has
experience with xtraDB cluster multicast communication? I think that is Rackspace network fault but maybe I’ve missed
something what could cause instability of the cluster in case of multicast communication.