asked on

Buffer I/O error on device sdb, sdc, sdd, sde, sdf, sdg, sdi - Oracle VM 3.2.8

Version 3.2.8 of OVM. Get these errors when rebooting VM Server. Rebooting takes about 20 mins.

Clip from /var/log/messages:-

Jan 7 10:45:04 svr440 kernel: Buffer I/O error on device sdi, logical block 31457251

Jan 7 10:45:04 svr440 kernel: Buffer I/O error on device sdi, logical block 31457252

Jan 7 10:45:04 svr440 kernel: sd 9:0:0:1: [sdi] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE

Jan 7 10:45:04 svr440 kernel: sd 9:0:0:1: [sdi] Sense Key : Illegal Request [current]

Jan 7 10:45:04 svr440 kernel: sd 9:0:0:1: [sdi] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1

Jan 7 10:45:04 svr440 kernel: sd 9:0:0:1: [sdi] CDB: Read(10): 28 00 0e ff ff 18 00 00 08 00

Jan 7 10:45:04 svr440 kernel: end_request: I/O error, dev sdi, sector 251658008

Any ideas?

Mazdajai

Have you run check disk on your hard disks to confirm they are not failing or have bad sectors?

David

Do you have a PERC RAID controller? If so the error indicates a problem with multipathing configuration when trying to read 512KB from the indicated offset. Did something physically get moved, and/or are you running multiple paths?

You need to report more of the log. It does NOT indicate a HDD failure. If you had a HDD read error then the sense key would not be "Illegal request", and ASCQ would not be 0x94. It is as if you're trying to read from a location that doesn't exist, either because the disk is not addressed where the controller expects it to be, or the offset is greater than the capacity of the drive.

P.S. the LAST thing you want to do is a check disk (fsck) as another author suggested. It will likely result in 100% data loss.

paul williams

ASKER

Yes, I was a little dubious about running an fsck for now.

Good news that its not a HDD failure.
VM server is running on a SunX3-2 (X86) and connected to an Oracle Storage 2540-M2 storage array.

Im wondering whether something has changed on the set up. Its not been working for a while apparently but I've only just picked this up.

ASKER CERTIFIED SOLUTION

noci

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

paul williams

ASKER

Hmm. Got reply from oracle on this :-

I can see the "Buffer I/O error" are generated by active ghost devices.
===========
$ grep "Buffer I/O error" var/log/messages | sort -k11| awk '{print $11}'|uniq
sdb,
sdd,
sdf,
sdi,
sdk,
sdm,

$ grep ghost sos_commands/devicemapper/multipath_-v4_-ll | sort -k2
`- 7:0:0:0 sdb 8:16 active ghost running
`- 7:0:0:2 sdd 8:48 active ghost running
`- 7:0:0:4 sdf 8:80 active ghost running
`- 9:0:0:1 sdi 8:128 active ghost running
`- 9:0:0:3 sdk 8:160 active ghost running
`- 9:0:0:5 sdm 8:192 active ghost running
7:0:0:0 sdb 8:16 -1 undef ghost SUN,LCSM100_F running
7:0:0:2 sdd 8:48 -1 undef ghost SUN,LCSM100_F running
7:0:0:4 sdf 8:80 -1 undef ghost SUN,LCSM100_F running
9:0:0:1 sdi 8:128 -1 undef ghost SUN,LCSM100_F running
9:0:0:3 sdk 8:160 -1 undef ghost SUN,LCSM100_F running
9:0:0:5 sdm 8:192 -1 undef ghost SUN,LCSM100_F running
===========

Which is not harmful so, can be ignored. Refer Doc ID 1464587.1 for more information.

paul williams

ASKER

I've requested that this question be closed as follows:

Accepted answer: 0 points for paul williams's comment #a40539683

for the following reason:

Advice from Oracle support.

noci

Those messages were not included in the snippet.
So how could we check for that fact?

still the advice you need to review the multipath setup is a valid one.
In this case there were more messages indicating that those were just warnings.

So it would be fair imho that you award some points.