Failover, or not Failover, that is the question.
Building a High-Availability database system is hard. It's not enough to just provide a pair of redundant servers, you also need to make the data redundant, while protecting data integrity. A long standing discussion item in this context is the difficulty of implementing the failover. Should you use automated failover or is it wiser to use manual failover? If you use an active-active cluster then you have no failover, how about that?
In this talk we will cover this problem space from all angles:
Do you failover or not failover?
Fast failover or slow failover?
Automated failover or manual failover?
With automated failover, are you being careful or careless about it?
In some systems there is no failover!