5

I have a CentOS 6 box with LVM setup and one of the PVs is a USB disk (I know). One of them is getting the error:

Oct 30 10:57:07 alpha01 kernel: lost page write due to I/O error on dm-3
Oct 30 10:57:07 alpha01 kernel: Buffer I/O error on device dm-3, logical block 4

Which is causing problems with all of the LVs on it. pvs shows the PV as unknown device. I can ls to the logical volumes and they show up in lvdisplay, but first I get a bunch of IO errors. I made sure the cables are secure between the USB drive. What should I do to get this back up and running for the meanwhile? Should I unmount each LV and run an fsck.ext4 on each one like fsck.ext4 -y /dev/vg1/lv_logvolname ?

Gregg Leventhal
  • 7,590
  • In addition to fsck, if the external drive is SMART capable, checking the drive status/health and running the drive self tests may be useful. Backing up all the data may also be important. – rickhg12hs Oct 30 '13 at 15:44

1 Answers1

3

I usually don't go the route of running an fsck and assume the disk is failing or has bad sectors. I definitely wouldn't run the fsck using the -y, since this will give fsck to attempt to start moving blocks which may exacerbate the problem.

Instead I'll run a tool such as Spinrite (Commercial) or HDAT2 (freeware) on the disk to do the analysis & potential repair.

What else?

See my other answers to these questions for additional methods:

The 2nd link provides details about another tool, safecopy for attempting to recover data from a failed/failing drive. It doesn't attempt to do any repairing of hte HDD.

Community
  • 1
slm
  • 369,824
  • Is there a linux spinrite that can be run from a normal runlevel 3 without a boot cd, and simply analyze specific filesystems? – Gregg Leventhal Oct 30 '13 at 17:18
  • @GreggLeventhal - no you have to reboot the system to do this analysis. I typically pull the drive out of the system and run it standalone. Another tool that's been used to do this is badblocks. – slm Oct 30 '13 at 17:33