[linux-lvm] Recovering from a hard crash

[Rechenberg, Andrew]

Also, can anyone see any harm in me modifying the source for vgscan to
skip /dev/md0 since it will never actually be used in a volume group
outside of the RAID0 stripe on top of it?  Would I have to modify any
other commands to make sure that I don't run into any trouble?

Thanks,
Andy.

-----Original Message-----
From: Christian Limpach [mailto:chris at pin.lu] 
Sent: Monday, February 24, 2003 12:49 PM
To: linux-lvm at sistina.com; Rechenberg, Andrew
Subject: Re: [linux-lvm] Recovering from a hard crash


"Rechenberg, Andrew" <ARechenberg at shermanfinancialgroup.com> wrote:
> Well, unless I'm reading this wrong, it looks as if /dev/md0 and
> /dev/md10 have the same pvdata for some reason.  /dev/md0 is the first
> part of /dev/md10.  Any ideas as to what's going on and how to resolve
> this issue?

md0 is the beginning of md10 and the LVM metadata is located at the
start of
the PVs.  This is why vgscan/pvscan sees the same PV on md0 and md10.  I
think (untested...) that the ugly quick fix is to "mv /dev/md0
/dev/notmd0",
this should work because vgscan/pvscan then won't see the device node.
The
less ugly fix is to use LVM2 with a devicename filter which excludes
md0.
In the long run (and since this is a test system), I'd suggest that you
recreate your VG from PVs on each /dev/md[0-9] instead of creating and
using
/dev/md10.  This also gives you better control of where your
snapshot-copy-on-write-space will be located (best not on the same
disks...).

    christian