[linux-lvm] LVM snapshots causing SATA disk failures

[Joerg Sommrey]

Hello,

the subject sounds strange - I know.  However, all the tests I did lead
me to this conclusion.

The setup consist of:
- two identical SATA disk /dev/sdb and /dev/sdc connected to a Promise
  SATA150 TX4 controller
- both disks are identically partitioned into /dev/sd[bc][1-4]
- each pair of corresponding partitions are members of a MD-RAID0 or
  MD-RAID1 device /dev/md[0-3]
- each MD-RAID device is the only physical volume of a LVM volume group
- there are several logical volumes within these volume groups
  containing reiser filesystems.

So far everything works fine.

Now I create a snapshot of one of the logical volumes in a RAID-1 vg
(and mount it) and do some "heavy" memory-mapped i/o on either the
original or the snapshot filesystem (e.g. starting mutt on a large maildir
folder).  This causes the system to fail in a absolutely repeatable way.
Syslog gets flooded with errors and the RAID-1 continues operation on one
partition.  Need to reboot (reset!) at this point.  I am able to
raidhotadd the failed partition without problems after reboot.

This problem appeared between kernel 2.6.10 and 2.6.11 and persists
until recent 2.6.12-rc5.

Does anybody have an idea what's going on here?

TIA
-jo

syslog:

May 31 17:08:13 bear kernel: ata1: status=0x51 { DriveReady SeekComplete Error }May 31 17:08:13 bear kernel: ata1: called with no error (51)!
May 31 17:08:13 bear kernel: ata1: status=0x51 { DriveReady SeekComplete Error }May 31 17:08:13 bear kernel: ata1: called with no error (51)!
May 31 17:08:13 bear kernel: ata1: status=0x51 { DriveReady SeekComplete Error }May 31 17:08:13 bear kernel: ata1: called with no error (51)!
May 31 17:08:13 bear kernel: ata1: status=0x51 { DriveReady SeekComplete Error }May 31 17:08:13 bear kernel: ata1: called with no error (51)!
May 31 17:08:13 bear kernel: ata1: status=0x51 { DriveReady SeekComplete Error }May 31 17:08:13 bear kernel: ata1: called with no error (51)!
May 31 17:08:13 bear kernel: SCSI error : <1 0 0 0> return code = 0x8000002
May 31 17:08:13 bear kernel: sdb: Current: sense key: Medium Error
May 31 17:08:13 bear kernel:     Additional sense: Unrecovered read error - auto reallocate failed
May 31 17:08:13 bear kernel: end_request: I/O error, dev sdb, sector 80684237
May 31 17:08:13 bear kernel: ata1: status=0x51 { DriveReady SeekComplete Error }May 31 17:08:13 bear kernel: ata1: called with no error (51)!
May 31 17:08:13 bear kernel: SCSI error : <1 0 0 0> return code = 0x8000002
May 31 17:08:13 bear kernel: sdb: Current: sense key: Medium Error
May 31 17:08:13 bear kernel:     Additional sense: Unrecovered read error - auto reallocate failed
May 31 17:08:13 bear kernel: end_request: I/O error, dev sdb, sector 80684245
May 31 17:08:13 bear kernel: raid1: Disk failure on sdb3, disabling device.
May 31 17:08:13 bear kernel: ^IOperation continuing on 1 devices
May 31 17:08:13 bear kernel: raid1: sdb3: rescheduling sector 278912
May 31 17:08:13 bear kernel: ata1: status=0x51 { DriveReady SeekComplete Error }May 31 17:08:13 bear kernel: ata1: called with no error (51)!
May 31 17:08:13 bear kernel: SCSI error : <1 0 0 0> return code = 0x8000002
May 31 17:08:13 bear kernel: sdb: Current: sense key: Medium Error
May 31 17:08:13 bear kernel:     Additional sense: Unrecovered read error - auto reallocate failed
May 31 17:08:13 bear kernel: end_request: I/O error, dev sdb, sector 80684253
May 31 17:08:13 bear kernel: raid1: sdb3: rescheduling sector 278928
May 31 17:08:13 bear kernel: RAID1 conf printout:
May 31 17:08:13 bear kernel:  --- wd:1 rd:2
May 31 17:08:13 bear kernel:  disk 0, wo:1, o:0, dev:sdb3
May 31 17:08:13 bear kernel:  disk 1, wo:0, o:1, dev:sdc3
May 31 17:08:13 bear kernel: RAID1 conf printout:
May 31 17:08:13 bear kernel:  --- wd:1 rd:2
May 31 17:08:13 bear kernel:  disk 1, wo:0, o:1, dev:sdc3
May 31 17:08:13 bear kernel: raid1: sdc3: redirecting sector 278912 to another mirror
May 31 17:08:13 bear kernel: raid1: sdc3: redirecting sector 278928 to another mirror

[etc. until reboot]