Skip to contentSkip to conversation

Recovering data from an LV/VG w/ a crashed PV

linux-lvm Ken Bass 2 participants 2 messages Nov 16, 2009

lore
Share & tools
lore

Ken Bass

author
#0 lore raw

(I know this topic has probably been brought up before, but I couldn't find
anything in the archives - if there is, please give me a link.)

I have an LV made up of 4 SCSI drives (PVs). One of those just crashed. Specifically, the SCSI controller can't get anything but a not ready error from it. Ergo, it is D-E-A-D. (And no, it wasn't part of a RAID :-( ).

I've come to accept that, but now I want to know if I can recover anything from the rest of the LV. Right now, at boot time, the LVM sees the VG with an LV that is missing a PV, so it drops me to maintenance mode, where I really can't do much. I can get the system to boot by physically removing the VG (it's not my system VG), then disable it in fstab, and reboot again. So, theoritically, I should have access to the VG/LV, missing the one dead PV.

I can't migrate the extents off of the dead PV (duh), but can I remove that PV from the LV/VG, and maybe recover some of the data on the other PVs?

FWIW, the filesystem is XFS. It is mainly used as a store and forward server, and fortunately, most of the data on it had been saved. The rest of it can be recreated, but it would be a very timely process (several days at least). So any data that I can get off of the "amputated" LV would definitely help.

I'm just about at the point where I will just rebuild the VG/LV from scratch
- that is, recreate a new VG/LV from the remaining PVs. But I just wanted to know - before I commit - if I can recover anything.

Also - minor gripe - why won't the system boot if a VG has a bad LV, and that VG is not the system VG? I'd like the LV to mount if it is okay, that is why I have it in fstab. But, as above, when the system boots, and the LVM sees that crippled VG/LV, the system goes to maintenance mode. Is there some way to tell the system at mount time to continue w/o mounting that VG/LV?

As always, TIA

ken

#1 lore raw

Something you might want to try is replacing the controller board on the disk. On two occassions, I had friends with disk that died, but we were able to get the data completely off the disk by replacing the controller board on the disk itself. This does require that you get exactly the same board as is on the drive, and this only works if the failure was due to the disk controller, and not internal hardware. From a Defcon presentation, this is the case in I recall 85% of disk failures.

I found the exact same disk as my friends disk on ebay for $20, and just replaced the board. May or may not work, but it is worth a shot, and will either work or not.

Be carefull when disconnecting the boards, to make sure you don't damage the cables that connect to the internal hardware.

Good Luck.

+----------------------------------------------------------+
  Michael D. Setzer II -  Computer Science Instructor      
  Guam Community College  Computer Center                  
  mailto:[email protected]                            

mailto:[email protected]
http://www.guam.net/home/mikes
Guam - Where America's Day Begins
+----------------------------------------------------------+

http://setiathome.berkeley.edu (Original)

Number of Seti Units Returned:  19,471
Processing time:  32 years, 290 days, 12 hours, 58 minutes

(Total Hours: 287,489)

BOINC@HOME CREDITS
SETI 8,839,360.988 | EINSTEIN 3,255,438.231 | ROSETTA 1,455,251.513