[tmikaeld]

Automatic failover is not the same as High Availability.

You should have a completely redundant node standing by to take over if the first one fails if you are running critical services - although, few clients seem to want to pay the +double cost of having it.

There is even the possibility to have a node in sync in a different datacenter, although the bandwidth costs can become high.

[coolboykl]

Yes, lesson learn from us.. We do have a scripts to auto cut over our Slave to Master DB once it detect our Master DB "death"..

Guess it's best to setup Master to Master replication..

[stevekemp]

> Guess it's best to setup Master to Master replication..

Master-Master has issues of its own, if you use shoddy applications that assume next-id=max(id)+1.

In the worst case situation you can find data on one master, and not the other, and vice-versa. So be prepared to reconcile things if you get into a split-brain scenario.