Re: What will be with cluster if network problems between nodes?

[Mikael Ronström]

It will be decided by an arbitrator. The cluster have choosen an arbitrator
for the cases when both sides can survive. In this case each side will try
to contact the arbitrator, the first one to succeed will survive, the one
not succeeding will be stopped.

This assumes that we have equal number of data nodes in each DC.

For proper operation of NDB it is important that DCs are not too far apart.
Latencies of 1-2 milliseconds should be ok, but not 30-40 milliseconds. In
case of long latencies it is advisable to use MySQL Cluster Replication between
DCs (can be advisable even in cases with lower latencies).

Failover in MySQL Cluster Replication is manual (meaning normally scripted).