LVM seems to be causing SCSI errors.

This is a real odd problem and my searching of the mailing lists and docs has not shed any light on it.

I created 3 logical volumes (with lvm 1.0.1r2, r3 and r4) across 4 disks on a dual 300 MHz PII running 2.4.6 (same behavior under 2.4.9) and readhat 7.1. After this I noticed IO errors while reading some files and SCSI errors when accessing lvol2. The errors were similar to the following

Sep 28 17:11:21 gar kernel: scsi0: ERROR on channel 0, id 11, lun 0, CDB: Read (10) 00 00 ba e1 97 00 00 80 00 
Sep 28 17:11:21 gar kernel: Info fld=0xbae1e6, Current sd08:21: sense key Medium Error
Sep 28 17:11:21 gar kernel: Additional sense indicates Unrecovered read error
Sep 28 17:11:21 gar kernel:  I/O error: dev 08:21, sector 12247456

lvol2 spanned two disks and all the errors were on only one of them. I naturally thought that the problem was a hardware problem with the disk. The only odd thing was that lvol1 which spanned the same two disks did not have any errors.

The first thing I tried was an mke2fs with bad block checking on lvol2. It found several errors and when I copied the files back on new errors cropped up and more files were lost.

The next thing I did was to move all the data off of that disk (pvmove) and then did a low level format on that disk. No errors were detected during the low lever format nor the verify I did after that. I then did a mke2fs with bad block checking on the disk without LVM. No errors. I then did a pvcreate and re-created lvol2 and then did a mke2fs with bad block checking. Once again the SCSI errors. Removing that drive from LVM and formatting yet again without LVM showed no errors.

Moving lvol2 to a different physical volume results in the SCSI errors moving to that volume.

I am now convinced that the problem is with LVM and not with the hardware. I still have two other logical volumes (lvol1 and lvol3) that are not having ANY problems at all. Does anyone have any suggestion on how to solve this one?

Thanks.

-----------------------------------------------------
Protect yourself from spam, use http://sneakemail.com

People have recently reported similar "hardware" errors on IDE drives as well. Can you verify if the sector in question is really within the boundaries of the device (this appears to be /dev/sdc1, sector 12247456)? If not, it is possible that LVM is doing math incorrectly somewhere.

Cheers, Andreas
--
Andreas Dilger \ "If a man ate a pound of pasta and a pound of antipasto,
\ would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/ -- Dogbert

Karl,

did you check the filesystem.
Ages ago there was a couple of problems like this caused by filesystem corruptions being fixable by the checker.

Please report, if this helped.

Regards,
Heinz -- The LVM Guy --

*** Software bugs are stupid.
Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen Sistina Software Inc.
Senior Consultant/Developer Am Sonnenhang 11
56242 Marienrachdorf Germany
[email protected] +49 2626 141200

                                                       FAX 924446

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-