Hi,

I have a server with an Areca 1230 raid card, running raid 6. I used lvm to create a volume group on the raid.
Yesterday though, the boot disk of this server died, and I was unable to recover anything from it using live cd's etc.

So then I decided to reinstall Centos x64, using 5.2 dvd, on a new hard disk. The Centos installer recognized my areca card, and loaded the driver for it. When the option was presented, where to install linux, I was given these choices:
- /dev/sda --> Areca raid
- /dev/sdb --> Samsung 500 GB
- /dev/sdc --> Samsung 500 GB

(/dev/sda was presented as FREE, I found this quite odd)

But being a bit scared to screw my raid up, I decided to boot from the 5.0 dvd instead, since this does not load the areca driver, just to make sure to mess nothing up. So that way I installed Centos on a new disk.

Then I loaded the areca module into the kernel, and my areca volume then showed up as /dev/sdc

But now though, when I use vgscan, it only detects the volume group on the boot disk, but not the one on my raid.

I don't understand what's gone wrong here. Did the 5.2 CentOS installer destroy my lvm? Is there any way I can regain access to the data on the lvm / raid?

Thanks.

I would say to boot into the raid bios and try to recover the raid lv. If you have not touched the disks, the bios should see the difference between the nvram and the disk headers, and if the nvram is corrupted, you can sometimes rebuild off headers, but it is not 100%.

If the disks you installed on are part of your raid set, you are hosed. You can also try a vgcfgrecover if you kept backups of your lvm configuration. Again, start with the RAID array, and get the lv back online. once that is up, work down to the VG, the LV, and then the filesystem.

I apologize if I am going "power cord" on you, but it you did not mention any attempts to recover from the raid bios / management screen.

Cheers!

Actually, it looks like you have done that. See if you have a copy of the lvm config anywhere, and check pvscan to make sure that the headers are truly lost. Sometimes checking via fdisk and making sure the partition type did not get changed off 8e is helpful too.

Sorry, it is late, time for me to go to bed. Good luck!

thanks for your reply.

actually what happened was (it's fixed now):

The CentOS installer had written a partition table at the beginning of the physical volume Using hexedit, I removed this table, and I also reconstructed the lvm config file, with the data I found using hexedit. The I did a vgcfgrestore, and my data was back. I did all of these actions first in a VM off course, since there was 6 TB of data at stake

it's all in detail here, but it is in dutch i'm afraid:

http://gathering.tweakers.net/forum/...ssages/1330489

translated page by google:

http://translate.google.com/translat...istory_state0=

P.S. this is quite wrongly translated though: "At this point the damage was already done by me, but I will return.", since the CentOS installer wrote the partition table WITHOUT asking me / I had done nothing