[mythtv-users] 2 TB Hard Drive Recommendations
jra at baylink.com
Mon Dec 6 14:40:32 UTC 2010
----- Original Message -----
> From: "Simon Hobson" <linux at thehobsons.co.uk>
> Jean-Yves Avenard wrote:
> >so two drives out of 6 dying in 18 months, so close to each other
> >isn't that great.
> That's one of the problems where the idea of RAID breaks down. You
> buy a set of disks (same make and model, possibly even from the same
> manufacturing batch), stuff them in an enclosure, and run them for
> the same power on hours, doing the same usage profile - ie they are
> run under near identical conditions. That means the failures are
> liable to be less distributed temporally than you might be thinking.
> So you've massively increased the chance of two (or more)
> simultaneous failures.
> I saw that in one of our Unix servers at my last job. A drive failed,
> and it took our maintainers a while to get a replacement because they
> struggled to understand that they can't supply any old "9GB drive"
> and it must be at least the same number of blocks as the smallest in
> the array. By the time it did get replaced, two more were showing an
> increasing error count and I think we replaced 3 in an array of 4 !
> For that reason, I've always setup my arrays to use a bit less than
> the full disk to allow for a slightly smaller drive as a replacement.
If you buy *anything*, hard disks for a RAID, redundant power supplies,
or what have you, where "pull a unit and replace it" is a fix...
and you do not buy the spare unit at build time, test it, bag it, and
stick it on a shelf, you're just fooling yourself.
If you're really motivated about this, you can buy 6 drives for your RAID5,
and rotate one in every 3 months. That will spread your failures, too.
More information about the mythtv-users