John R Pierce wrote:
> Johnny Hughes wrote:
>> The easiest way to make that happen is to use a product called DRBD
>
> do be aware, drbd replicas are not 'safe' for things like transactional
> databases, unless they are configured to be synchronous (such that a
> fsync doesn't return until its written on both the local disk AND the
> replica), which slows everything way down.
If you can tolerate a small amount of downtime, a more simple-minded
approach is to get servers with swappable drive carriers, set up your
production servers with all partitions on raid1, and keep a spare
similar chassis around. Then if a single drive dies (the most likely
failure), you just swap in a new one and resync the mirrors. If the
motherboard or power supply dies, you swap the drives into the spare and
come up in the time it takes to reboot (you'll probably have to fix the
NIC setup for the different hardware addresses, though.).
--
Les Mikesell
lesmikesell at gmail.com