[gentoo-user] rescuing lvm data

[gentoo-user] rescuing lvm data

[Stefan G. Weichinger]

greetings ...

the server in my basement is driving me nuts ... and I start making
mistakes ... sigh

It had a LVM volume group ... consisting of two PVs ...

The question in short:

if I only have one PV at hand, which contains all the PEs of a given LV
... (how) am I able to extract/export/read the contents of that LV?

background:

when I searched for the reason of those crashes, I removed and swapped
and cloned hard disks all around.

Right now I am on a brand new machine and want to sort out harddisks etc.

One PV wasn't detected correctly ("unknown device") even though it had
the correct uuid. I did some "pvcreate -u" on it ... and now those LVs
on it aren't accessible anymore (can't mount or even fsck).

My luck (I hope!): I have a clone of that harddrive.

But I don't want to simply plug it in because I fear to ruin things as I
already messed up that VG further by trying stuff like vgcfgrestore etc.

Yes, my fault.

The clone contains all the blocks for an LV I would like to keep
somehow. So I assume it should be possible to access them?

(explanation why there is no backup: the data is mythtv-stuff, video
which is nice to have, but not critical.)

Thanks, Stefan

Re: [gentoo-user] rescuing lvm data

[Florian Philipp]

[-- Attachment #1: Type: text/plain, Size: 1685 bytes --]

Am 24.10.2012 22:08, schrieb Stefan G. Weichinger:
> 
> greetings ...
> 
> the server in my basement is driving me nuts ... and I start making
> mistakes ... sigh
> 
> It had a LVM volume group ... consisting of two PVs ...
> 
> The question in short:
> 
> if I only have one PV at hand, which contains all the PEs of a given LV
> ... (how) am I able to extract/export/read the contents of that LV?

So you still have the volume group metadata, just not one of the PVs, right?

> 
> background:
> 
> when I searched for the reason of those crashes, I removed and swapped
> and cloned hard disks all around.
> 
> Right now I am on a brand new machine and want to sort out harddisks etc.
> 
> One PV wasn't detected correctly ("unknown device") even though it had
> the correct uuid. I did some "pvcreate -u" on it ... and now those LVs
> on it aren't accessible anymore (can't mount or even fsck).
> 
> My luck (I hope!): I have a clone of that harddrive.
> 
> But I don't want to simply plug it in because I fear to ruin things as I
> already messed up that VG further by trying stuff like vgcfgrestore etc.
> 
> Yes, my fault.
> 
> The clone contains all the blocks for an LV I would like to keep
> somehow. So I assume it should be possible to access them?
> 

First things first: If you can, make a disk dump of every affected HDD
or at least back up the metadata with vgcfgbackup.

Have you tried activating the volume group with --partial (see `man
lvm`). Something along the lines of `lvchange --partial -a y`).

If that doesn't help, maybe `vgreduce --removemissing --force` will do
the job.

Hope this helps,
Florian Philipp


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]

Re: [gentoo-user] rescuing lvm data

[Stefan G. Weichinger]

Am 24.10.2012 22:23, schrieb Florian Philipp:

> First things first: If you can, make a disk dump of every affected
> HDD or at least back up the metadata with vgcfgbackup.
> 
> Have you tried activating the volume group with --partial (see
> `man lvm`). Something along the lines of `lvchange --partial -a
> y`).
> 
> If that doesn't help, maybe `vgreduce --removemissing --force` will
> do the job.

Florian, thanks for the hints ... "extracted" two smaller LVs already.
The big one seems to span both PVs  ... I will continue tomorrow after
some sleep.

Thanks once more, Stefan