On 09/04/2008, Douglas A. Tutty <dtutty@porchlight.ca> wrote:
On Tue, Apr 08, 2008 at 11:08:11PM -0400, Mitchell Laks wrote:
> Someone recently talked about using
> ls -l /dev/disk/by-uuid
> So that must be the UUID for the individual /dev/sda1 and /dev/sdb1.
>
> What is the UUID for the raid1 itself?
Doesn't matter. The raid subsystem will search all disks for raid
members irrespective of /dev name when it reassembles the raid on boot.
/dev/md0 will always be /dev/md0 and not /dev/md3 or something.
FYI, LVM is the same way, only its /dev/mapper...
Doug.
Hey,
The raid subsystem writes headers to the disks themselves which include a uuid array identifier and a prefered device number (eg 0 for md0). It uses this to gather array members and then presents the array as /dev/md0.
This means, as Doug said, that you can put the disks back in any order, on any controller ( or over a network - check 'write-mostly' ) and the resultant block devices (eg /dev/sda) are in the search path (check mdadm.conf) the array will build.
So long as there aren't two or more arrays requesting the same dev number (eg for md0) then each should get their preference.
The array itself does have a uuid (check `mdadm --detail /dev/md0|grep UUID`), but I have never tried using it as the boot argument.
cheers,
Owen.
n.b. I'm on a windows comp at the moment so take above as all 'IIRC'.
--
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org