setting VG UUID

Hi!

After figuring out that my all my PVs were lacking UUIDs, I used this hack from Heinz to correct the situation:

http://lists.sistina.com/pipermail/linux-lvm/2002-January/010521.html

(Alasdair also kindly provided a patch to handle the situation in LVM2.)

Now my VG itself is still lacking a UUID and this causes problems such as vgcfgrestore refusing to continue. Is there any way I can update my VG UUID?

I also have this problem where my PV's are using different system ids:

VG: localhost.localdomain1073237462
/dev/hda6: localhost.localdomain1073237462 (same)
/dev/hda7: localhost.localdomain1073237466
/dev/hda8: localhost.localdomain1073237470
/dev/hda9: localhost.localdomain1073237474

Is there a way of fixing this?

Thanks,
Navin.

For the record, I'm absolutely crazy, but I fixed this by doing:

echo "deadbeefdeadbeefdeadbeefdeadbeef"|dd of=/dev/hda6 obs=1 seek=4096 count=32
echo "deadbeefdeadbeefdeadbeefdeadbeef"|dd of=/dev/hda7 obs=1 seek=4096 count=32
echo "deadbeefdeadbeefdeadbeefdeadbeef"|dd of=/dev/hda8 obs=1 seek=4096 count=32
echo "deadbeefdeadbeefdeadbeefdeadbeef"|dd of=/dev/hda9 obs=1 seek=4096 count=32

So much for universal uniqueness... but now finally LVM2 is happy to accept my LVM1 partitions! Woohoo!!

For the record, I created my LVM1 partitions with Linux Mandrake 9.2. So I anticipate other people will be having this kind of fun. :-)

One inconvenience I still seem to have is that /dev/bytepool/ is now filled with symlinks and when I boot back to Linux 2.4 with LVM1 (no device-mapper), it fails to mount the partitions because of this. Cleaning /dev/bytepool/ and rebooting seems to fix this.

Should "vgchange -a n" maybe clean up these symlinks?

Cheers,
Navin.

The VG UUIDs *are* meant to all be the same on devices that are part of the same VG. It's the PV UUIDs that differ.

Could /dev/urandom have been missing when pvcreate was run? We think that would explain the symptoms.

It's supposed to do.

Alasdair

[email protected]

Oh, I realise that. Just a small joke about my non-random choice of UUID... :)

Good thought! It's very possible since of course only a crippled environment is available at install time. I'll try to verify it later.

Hmmm. I've hacked my init scripts to do it in the meanwhile, but I will pay closer attention to what happens next time I boot and report back.

Happily, so far I don't see any reason to go back to Linux 2.4 though and a quite a few reasons to stay in 2.6....

Cheers,
Navin.

It's probably trying to do it AFTER the kernel unmounts / (remounts RO) Watch your shutdown, that's where I see "unable to BLA, read-only
filesystem" messages. Since /dev gets cleaned up on boot I'm not too
worried about it.

--Dan

is it possible to have a copy of this patch?
L.

Luca Berra -- [email protected]
       Communication Media & Services S.r.l.
/"\
\ /     ASCII RIBBON CAMPAIGN
 X        AGAINST HTML MAIL
/ \

Up to Alasdair... have you seen this problem too?

Remember, if your PV is lacking UUIDs, your VG probably is too and that's going to cause problems as well.

At least it did for me.

Cheers,
Navin.

not personally, but i had some bug reports of similar problems.

yes, thanks for the useful info.

L.

Luca Berra -- [email protected]
       Communication Media & Services S.r.l.
/"\
\ /     ASCII RIBBON CAMPAIGN
 X        AGAINST HTML MAIL
/ \

It's already in LVM2 CVS, together with patches to *change tools to let you change the UUIDs to new random ones (not to be used while any LVs are active).

Alasdair

[email protected]

thanks a lot
L.

Luca Berra -- [email protected]
       Communication Media & Services S.r.l.
/"\
\ /     ASCII RIBBON CAMPAIGN
 X        AGAINST HTML MAIL
/ \