On Thu, Oct 11, 2012 at 04:17:32PM +0200, Oriol Martí wrote:
> I've to make a HA mysql, for me the scalability it's not important
> because it's not an environment with a lot of queries.
> I've thinked about, and I think that the best solution is to mount
> an master-slave arquitecture, and if the master goes down, promote
> the slave to master. Monitoring this case, after manually we can put
> the master with slave status and recover the environment.
How about not worrying about failover at all? If you have a multi-master
replication architecture, you don't need to worry about a machine going
offline. If a connection is dropped, reconnect to the database pool and
Mark Callaghan talks about this in his latest post. I've expanded on
this in the comments.
If I find the time I might write up a blog post and explain how we do
things at Opera Software.