If this is production consider a configuration with a hot spare and
always keep a spare disk available on the shelf. If one fails from a
batch another cannot be far behind (remember Murphy).
Bob
Robert O'Connor
AIS Infrastructure
Penn State University
On Nov 26, 2003, at 11:12 AM, fedora-list-request@xxxxxxxxxx wrote:
To: fedora-list@xxxxxxxxxx
Subject: Re: RSYNC Fedora
From: "Edward C. Bailey" <ed@xxxxxxxxxx>
Date: Wed, 26 Nov 2003 10:59:40 -0500
Reply-To: fedora-list@xxxxxxxxxx
"Brian" == Brian Fahrlander <Brian@xxxxxxxxxxxxxx> writes:
Brian> I once asked a woman that worked at Veritas in tech support.
Brian> She said that on all these RAID arrays, loss of a second disk is
Brian> _ALWAYS_ fatal. I was stunned.
Yup, she was right.
Brian> I mean, a 30 drive array...and if the second drive goes
out, the
Brian> entire thing is toast?
If you choose to set up a single array with 30 drives in it, that is
correct.
Brian> There gets to be a point where this is a problem. In 1-5
disk
Brian> arrays, what are the chances of a second drive going out? Right:
Brian> almost nil.
Brian> But when you have 50 drives in a special bay, the chances
of a
Brian> second drive going out while you locate a vendor, find out it's
been
Brian> obsoleted since it was installed, order a new part....
Exactly, so as the old joke goes when the guy tells his doctor it hurts
when he moves his arm, then don't do that! :-)
Instead, keep the number of drives in any given array low. If you need
more space, look at getting bigger drives. If you need even more
space,
consider "stacking" arrays (make a RAID 0 array out of as many 3-drive
RAID
5 arrays as you need to reach your storage requirements, for example).
By
doing this you still are exposed to the risk of multiple drives
failing at
any given time, but you reduce the exposure of any one array losing
more
than one drive.
Ed
--
Ed Bailey Red Hat, Inc. http://www.redhat.com/
