On Fri, 26 Sep 2008, Marc Skinner wrote:
> if ha is important and you can't have any downtime, you will need to 
> pony up the cash for a 2 node RAC deployment.  if you are ok with the 
> time needed to replay transaction logs (typically in the hours - on big 
> db's) you can use an active/passive cluster.

You can also deploy MySQL in a shared-nothing cluster, but there are 
disadvantages to doing so.

Another option is to use DRBD and Heartbeat to handle failover; by doing 
that, you are actually failing over the same data disk, so no need to 
replay logs.

------------------------------------------------------------------------
| nate carlson | natecars at natecarlson.com | http://www.natecarlson.com |
|       depriving some poor village of its idiot since 1981            |
------------------------------------------------------------------------