rescue a broken LVM with a FC rescue disk

[morci]

For some reason (HDD or controller failure, I guess) the ext3 FSs of my FC3 box are gone.
I can't even mount the FSs from a rescue disk.
It turns out that the LVM itself has crashed. I tried to activate LVM volumes by hand but vlm complains about non-matching ids.
Is there anything I can try, booting from a recue disk, in order to mount at least the LVM volume on the second HDD and see if my postgres data are still there or lost forever?

Thanks,
Davide.

[tashirosgt]

I'm not a lvm expert but let me perform the service of clarifying what you want. Can you see the hard drives (fdisk -l) when you boot from a rescue cd? Did you format two hard drives as a single logical volume? Were they hda and hdb? Is it hda that failed and is your question whether you can access the information on hdb with a rescue cd ?

(If so, one pessimistic thread on such a topic is: http://fedoraforum.org/forum/showthr...&highlight=lvm )

[kg4cbk]

If it is a controller failure have you tried replacing the motherboard?

My understanding of LMV (which is not extensive) is that a logical volume defined under a volume group is treated as a single file system. If that logical volume is made up of several physical drives and one of those drives fails I don't think there is anyway to recover the data on the remaining drives short of using tools that read sector by sector and attempt to piece the files back together.

LVM is good for creating file systems that are larger than any one physical drive and to allow a file system to be grown or shrunk as needed.

LVM is NOT a form of RAID or way to protect your data. You can use LVM in combination with RAID. But I am not sure how that would complicate things if you did have a drive failure and went through the process of replacing the drive and rebuilding the RAID volume.

The extent of my experience with LVM aside from default FC installs is building one LVM volume that spanned 4 drives to create a 1TB file system using XFS. This was done in combination with LVM.

$ df -vh
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/VolGroup00-LogVol00 97G 5.7G 86G 7% /
/dev/sda1 99M 29M 65M 31% /boot
none 1014M 0 1014M 0% /dev/shm
/dev/mapper/VolGroup00-LogVol02 1018G 781G 238G 77% /video


I did not use raid as I wanted the largest file system possible for use in a mythtv box.

If I lose that file system or one of those drives then I lose the data. It will be a matter of rebuilding the system from scratch and I won't be trying to recover any of the data.

If I considred the data critical I would set things up using raid and have a backup scheme in place and tested so all data could be restored if needed.

[morci]

Originally Posted by tashirosgt

I'm not a lvm expert but let me perform the service of clarifying what you want. Can you see the hard drives (fdisk -l) when you boot from a rescue cd? Did you format two hard drives as a single logical volume? Were they hda and hdb? Is it hda that failed and is your question whether you can access the information on hdb with a rescue cd ?

(If so, one pessimistic thread on such a topic is: http://fedoraforum.org/forum/showthr...&highlight=lvm )

ok, let me try to clarify...
I can see both HDDs with fdisk -l.
In sdb I have a single partition that is allocated to the second logical volume of my main volume group.
This logical volume is the only thing in sdb and it's fully contained there (it does not span to other drives).
My guess is that sda failed so I should be able to rescue sdb (it's just a guess, the controller might have failed compromising data on both disks).
So I just need to mount the ext3 fs that's in this logical volume in sdb; the point is that I'm not able to activate the logical volume (lvm complains about a non-matching signature). That's it!

Davide.

[tashirosgt]

So the basic question is whether you should be able to mount sda1 (which is of type Linux LVM) as an ext3 partition? And if you do so, will this damage the data in it?

I don't claim to know the answers, but I would think that
mount /dev/sdb1 -t ext3 /mnt/my_mountpoint
would fail since the type is wrong. Your description of the error message from lvm didn't produce any useful hits on google. Perhaps if you post the exact output of
lvdisplay /dev/VolGroup00/LogVol01
someone will recognize the problem.

I did notice this page:
http://dcs.nac.uci.edu/~strombrg/EXT...y-in-LVM2.html

[morci]

Originally Posted by tashirosgt

So the basic question is whether you should be able to mount sda1 (which is of type Linux LVM) as an ext3 partition? And if you do so, will this damage the data in it?

Well, it is sdb1, but you got the point.

Originally Posted by tashirosgt

I don't claim to know the answers, but I would think that
mount /dev/sdb1 -t ext3 /mnt/my_mountpoint
would fail since the type is wrong.

In fact.

Originally Posted by tashirosgt

Your description of the error message from lvm didn't produce any useful hits on google. Perhaps if you post the exact output of
lvdisplay /dev/VolGroup00/LogVol01
someone will recognize the problem.

I'll try that tomorrow.

Originally Posted by tashirosgt

I did notice this page:
http://dcs.nac.uci.edu/~strombrg/EXT...y-in-LVM2.html

Yes, I noticed that too.
But as I get to

lvm vgchange --ignorelockingfailure -P -a y

I get the error about that wrong signature (I don't remember exactly what the error massage says, I have to try it again and write it down).

Davide.