On Mon, Jul 27, 2009 at 08:17:20AM -0400, Jeff King wrote:
> So my problem has been resolved, but I am not sure whether there is a
> bug that will affect other users (hence this bug report). Why is it that
You're not alone; I had the same experience this afternoon. (This was
my first reboot since the upgrade.)
> Note that the UUIDs are the same in the first half, but totally
> different in the second half. One other thing to note is that
Same here.
Here are the UUIDs as they were before today:
/dev/md0 UUID=9900cea9:6a368469:d164b76c:a2dc5503
/dev/md1 UUID=e86dbf11:5b1671f1:d164b76c:a2dc5503
/dev/md2 UUID=c2a032c9:727b1dd3:2afa2a14:7e11b5d8
And here they are now:
/dev/md0 UUID=9900cea9:6a368469:2afa2a14:7e11b5d8
/dev/md1 UUID=e86dbf11:5b1671f1:2afa2a14:7e11b5d8
/dev/md2 UUID=c2a032c9:727b1dd3:2afa2a14:7e11b5d8
Notice how they switched to the same suffix, which was already present
in /dev/md2. (/dev/md2 is a one-drive RAID1 array, so it's always been
a bit funky.) Also notice how md0 and md1 used to share the same
suffix.
Before today, mdadm would appear to consider /dev/md/[01] as canonical
names, and complain that they were not in mdadm.conf, even though I had
/dev/md[01] in there. However, it *never* did that for /dev/md2, and
never created a corresponding /dev/md/2. Today, I find that
/dev/md/[01] are gone as well. Can't be a coincidence.
> Reinstalling (and rebuilding the initrd with) the version of mdadm from
> testing (2.6.9-3) works fine.
In my case, rebooting with the lenny CD yields the same UUID as with
3.0, so I'm guessing that it's not just a question of 2.x and 3.x
interpreting the UUID data differently, but that 3.x actually updated
the superblock UUID. I somehow doubt that reinstalling 2.6.9 would work
for me.
> -- Package-specific info:
Damn, I just filed another mdadm bug a few hours before rebooting, but
that info isn't included (and yet I could swear that it was). Too bad,
as we could've seen the old UUIDs for ourselves.
--
<liiwi> udp - universal dropping of an pigeon
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
