8 Oct
2012
8 Oct
'12
10:33 a.m.
On 08/10/12 12:20 PM, Ethan Dicks wrote:
On Mon, Oct 8, 2012 at 11:56 AM, Toby Thain<toby at
telegraphics.com.au> wrote:
The IceCube datacenter at Pole keeps dozens of spare disks on hand for
replacements, but there, you can't just phone someone up in June and
get a disk shipped in overnight... well... it'll be there half-past
sunrise, but that means "November" in this case.
One advantage of a remote site is that you *know* you can't get parts
in when you want them so you have to plan in advance and maintain a
spares pile. No other place I've worked kept spare hardware around
When ex-employer put in an Xserve G4 + Xserve RAID (2003?) we had spares
of everything. It was a $40,000 system at that time (1.1TB online).
It's not just time-to-order; suitable disks often get EOL'd before the
originals fail.
I just find it amusing that the cargo cult practice is to configure
servers with RAID without thinking it through. (Or realising that RAID
isn't state of the art.) :) (Obviously this doesn't apply to actual
professional outfits.)
--T
RAID
!= backup != offsite backup!
I consider RAID as an availability measure - but how many installations have
you seen that have spare disks on-hand and procedures to monitor and replace
them? (I am sure the serious ones do. Elsewhere, every cowboy sysadmin
configures RAID on day 1... but where are the disks? :) like that - when it dies, call it in and they send one
(and you hope
that a second disk doesn't die before the first arrives). Had that
happen exactly once - disk 1 dies, call for replacement, 5 hours
later, disk 2 dies, replace 2 disks in moments then spend days
rebuilding the array from what backups there are.
-ethan