[nycbug-talk] Mysql clustering/failover

[Hans Zaunere]

> Yes I know that clustering and failover are two different things. I would
> like to know if any of the dudes or dudets out there know of any links to
> help me create a failover mysql server x 2. I don't quite like the
> replication method as I have lost data in the past, and the logs and
> overhead tell me that there are other methods out there.

There are of course other methods, but none of them really make sense.  Replication is the way to go - it's the poor man's clustering for MySQL, and in fact, not that poor.  With a well designed server and application architecture, it can be very robust, supporting some of the largest sites on the web.

One other method, albeit it full of caveats, is a multi-head server.  Meaning, a single shared data file (SAN for instance) with multiple MySQL instances.  This involves some work, however, as proper implementation requires OS specific know-how (and some OSs don't support it at all).

> This link is a decent start but not exactly what I am looking for.
> 
> www.davz.net/static/howto/mysqlcluster/

In addition to the MySQL flagship server, there's also MySQL Cluster.  It's implemented as a storage engine with the standard server, and can be used transparently with the other storage engines.  It can also be used with the standard MySQL replication, which makes for some very robust systems.

What types of applications are you looking to support?

H