The most effective way to "divorce" it is to remove connectivity between
the datacentres. I would put in place firewall rules between the DC's to
stop them from communicating, and then rolling restart one of the DC's. You
should be left with 2 datacentres that see each other as down, and on each
one you can then go ahead and remove the opposite DC from replication, and
then remove the nodes from the other DC (via nodetool removenode). Before
changing RF or removing nodes make completely sure that there is *no way* for
the DC's to communicate.
