[linux-lvm] disaster recovery

Rene Gallati lvm at gallati.net
Fri Jan 24 08:47:01 UTC 2003


Hello List,

two days ago a harddisk died (IBM imfamous click of death - after 2 
years of operation).

The harddisk was disk two (PV) of a two disk VG with several LV's inside.

The system continued to run fine, though I had to unmount some LVs. Some 
where still mounted from this VG and they continued to work well (those 
are probably on the remaining good disk)

Now I've ordered and recieved new disks promptly and since they are 
quite larger than those in use, I could remove all active disks, ie the 
damaged one AND the good one. I'm running with fresh disks now.

Still I'd like to get the valid data out of the good disk. Since it 
remained running (/usr is among one of them - and the system continued 
to execute programs underneath /usr) I am fairly sure that this data 
should be recoverable.

I had several LVs in the VG, most where quite small, the last one was 
large and spanned both disks but I'm quite sure it "began" on the good 
one also. Now I've read a bit in the archive and searched for howto's - 
but I've still got some questions on how to proceed.

Because I've needed to shut down the machine anyway, I just made a 
complete software upgrade. Now I'm running a SuSE 8.1 - with all 
relevant security patches applied.

I'm wondering if I have all the tools required to attemp to recover the 
remaining data from the good drive?

I keep reading about a "LVM2" - can anyone shortly tell me what exactly 
that is? Only new userspace utilities? complete new LVM subsystem?

Do I need LVM2 to recover my data or do I already have all I need? 
(Version of vgdisplay for example is) :
vgdisplay: Logical Volume Manager 1.0.5(mp-v6)
Heinz Mauelshagen, Sistina Software  15/07/2002 (IOP 10)

Do I need to build my own kernel if I need LVM2? I'd like to skip that 
at the moment. I always build my own kernel, that's not a problem per 
se, but I am still rebuilding the machine and I'd like to do that only 
once - at the end when everything is fine again.

Now once I have all the tools necessary, am I right that I basically 
need to execute vgscan and vgchange and they should be able to activate 
the VG even when one disk is missing, right? I tried that on the still 
damaged system and I couldn't get vgchange to activate the VG - it just 
didn't do it.

Also I ran vgcfgrestore once, but it segfaulted. pvdisplay though 
reported good looking data from the good drive. That was on the old 
system though (with old LVM tools, status at about SuSE 7.3 level 
(although security upgrades where always applied)).

But before I really try to get the data out I just want to be sure I do 
everything correct.

btw: Do I have a chance of getting part of the data from the LV out, 
that was spanning the disks? The filesystem was reiser, I think it 
should be able to handle that (and the beginning of the fs should be on 
the good disk). Or is that just out of the window anyhow?

Thanks for all answers in advance.

CU

René





More information about the linux-lvm mailing list