topic Re: LVM Problem in Operating System - Linux

LVM Problem

Jeff Ohlhausen — Tue, 19 Oct 2004 09:03:34 GMT

Hi,
Hopefully someone can help me with this. I am running SAP on SuSE Linux with Oracle and using LVM to manage the filesystems. I also have a Steeleye Lifekeeper cluster. Yesterday the system failed over and the only messages I see are:
/var/log/warn:
Oct 18 16:29:19 pd05yyz kernel: scsi4 (0,0,21) : RESERVATION CONFLICT
Oct 18 16:29:19 pd05yyz kernel: SCSI disk error : host 4 channel 0 id 0 lun 21 return code = 18
Oct 18 16:29:19 pd05yyz kernel: I/O error: dev 08:40, sector 6
Oct 18 16:29:19 pd05yyz kernel: scsi4 (0,0,21) : RESERVATION CONFLICT
Oct 18 16:29:19 pd05yyz kernel: SCSI disk error : host 4 channel 0 id 0 lun 21 return code = 18
Oct 18 16:29:19 pd05yyz kernel: I/O error: dev 08:41, sector 0

And in the lifekeeper log it complains about the logical volume not being active.

So my questions are:
1. how does a logical volume become inactive?
2. does anyone know the possible causes of this error?

EMC has check the disk and the switches and has said everything is fine. If anyone could help I would really appreciate it.

Jeff

Re: LVM Problem

Richard Allen — Tue, 19 Oct 2004 10:40:37 GMT

As far as I can see, somebody is reserving the disk/lun for use other than the node pd05yyz.

There is a small description of this and other scsi codes on http://www.tldp.org/HOWTO/SCSI-Programming-HOWTO-21.html

This also seems to be a problem many Steeleye users see and I think it's been fixed as pr:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=74672

Re: LVM Problem

Steven E. Protter — Tue, 19 Oct 2004 15:26:28 GMT

A volume group becomes inactive because a system administrator deactivates it.

I'd say there is a bad sector on the disk or a bad disk. verify as follows:

lvdisplay -v /dev/vg##/pd05yyz

Replace the ## signgs with the real volume group number. I'm pretty sure you'll see some stale sectors.

SEP

Re: LVM Problem

Jeff Ohlhausen — Wed, 20 Oct 2004 09:47:21 GMT

Hi,
I did perform the lvdisplay -v but there are no stale sectors. Are there other possible causes?
Thanks