copying lvm with the same name

[Aldo Foot]

On Thu, Mar 19, 2009 at 11:39 AM, Rick Stevens <ricks at nerd.com> wrote:
> Aldo Foot wrote:
>>
>> On Thu, Mar 19, 2009 at 10:17 AM, Rick Stevens <ricks at nerd.com> wrote:
>>>
>>> Frank Cox wrote:
>>>>
>>>> On Wed, 18 Mar 2009 17:07:28 -0700
>>>> Rick Stevens wrote:
>>>>
>>>>> We have a serious conflict here.  The df command shows you as on sda,
>>>>> but LVM is reporting sdb.  My gut reaction is to have you do a:
>>>>>
>>>>>       vgreduce --test VolGroup00 /dev/sdb2
>>>>>
>>>>> and see if it would be successful.  If so, then remove the "--test" and
>>>>> cross your fingers.
>>>>
>>>> [root at mutt temp]# vgreduce --test VolGroup00 /dev/sdb2
>>>>  Test mode: Metadata will NOT be updated.
>>>>  Physical Volume "/dev/sdb2" not found in Volume Group "VolGroup00"
>>>>
>>>> This is consistent with the fact that the "active" Volume Group that I'm
>>>> using
>>>> is not on /dev/sdb2.  It's just the "lvdisplay -vm" command that shows
>>>> it
>>>> as
>>>> being in use.
>>>
>>> This is truly screwey.  The pvscan shows sdb2 as part of VolGroup00,
>>> lvdisplay shows the partition as in use, but vgreduce says sdb2 isn't
>>> part of the VG.  Hoo, boy.
>>>
>>> Frank, this is potentially dangerous, but you can try
>>>
>>>       # pvremove /dev/sdb2
>>>
>>> to wipe out sdb2's PV status.  Since you seem to be running on /dev/sda
>>> and vgreduce claims that /dev/sdb2 isn't part of the active VG, you
>>> should be able to do this without blowing things up.
>>>
>>> Damn this makes me nervous!
>>
>> The pvremove man page does say whether the UUID can be used to remove
>> a PV. Using the UUID would've come very handy in this kind of situation.
>
> True, Aldo, but the interesting thing is that we're seeing very
> inconsistent data.  The LV is running on /dev/sda2, but the status info
> show stuff as being on /dev/sdb2.  For example, vgreduce says that /dev/sdb2
> isn't part of the VG, while vgdisplay says it is.  lvdisplay
> shows extents on /dev/sdb2 as being in use when, in fact, they aren't.

And that's why I thought that the UUID would be of good use here.
I noticed that the OP has two VGs with the same name and two LVs
in each VG with the same names as well; and that's why pvremove
could yield unpleasant results.

Do you think that hard drive cable arrangement has anything to do with
how the drives are seen in this case? Frank has not indicated whether
he's using PATA or SATA drives.

> My brain is starting to hurt.

No kidding.

~af