Dealing with the failure of a MySQL master is not simple.
The most common solution is to promote a slave as the new
master but in an environment where you have many slaves, the
asynchronous implementation of replication gets in your way.
The problem is that each slave might be in a different
- some could be very close to the dead master,
- some could be missing the latest transactions,
- and some could be far behind (lagging, delayed slaves, or slaves in maintenance).