From: Jeff Hanson (jhanson_at_sgi.com)
Date: Mon Aug 13 2007 - 14:22:23 CEST
Message-ID: <46C04CFF.6030809@sgi.com> Date: Mon, 13 Aug 2007 08:22:23 -0400 From: Jeff Hanson <jhanson@sgi.com> Subject: Re: [suse-sles-e] sles10sp1: xfs + lvm + snapshot = Input/output error
Norbert Paschedag wrote:
>
> On Fri, 10 Aug 2007, Dr J Pelan wrote:
>
>> On Wed, 8 Aug 2007, Norbert Paschedag wrote:
>>
>>> we just stumbled over a problem with lvm snapshotting of xfs
>>> filesystems. Just creating a snapshot of a _mounted_ volume and then
>>> accessing the original filesystem creates an I/O error. Has anyone else
>>> seen this and knows a way to prevent this ?
>> Good catch. I'd just like to confirm that this problem is reproducable
>> here on a fully patched SLES10+SP1 test system (x86_64). It seems that
>> after a remount subsequent snapshots do not cause problems so there is a
>> workaround of sorts, but I would not deem this acceptable for production.
>
> My colleague investigated further and found that the error appears only
> on xfs filesystems that have barriers enabled.
> In the presence of a snapshot, mounting the original filesystem will
> disable barriers (also in ext3). This explains why the error goes away
> after doing the recommended umount/mount.
>
> File systems can be mounted without barriers using the undocumented
>
> mount -o nobarrier <dev> <mntpt>
>
> Can anyone with more knowledge of the SLES kernel explain what xfs patches
> made it into the kernel since SLES10 ?
> Officially (I think) barriers are on by default only in kernel >= 2.6.17.
Looking over the diff of the changelogs in the kernel rpms from 10sp1
and 10ga for x86_64 I only see this clue
< * Mon Mar 26 2007 - neilb@suse.de
< - patches.fixes/xfs-barrier-test: Relax tests for write-barrier
< by XFS (256097).
None of the other comments on xfs changes that refer to SGI PV (our
bug tracking system) seem to apply. A quick search of our internal database
doesn't reveal the specific code.
Note we are suggesting to our customers that xfs filesystems be mounted
with nobarrier EXCEPT when the drive has write caching on. This is
SGI PRB 201509 which is available to all customers.
-- ----------------------------------------------------------------------- Jeff Hanson - jhanson@sgi.com - Field Technical Analyst You can choose a ready guide in some celestial voice. If you choose not to decide, you still have made a choice. You can choose from phantom fears and kindness that can kill; I will choose a path that's clear I will choose freewill. - Lee/Lifeson/Peart --------------------------------------------------------------------- To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com For additional commands, e-mail: suse-sles-e-help@suse.com
This archive was generated by hypermail 2.1.7 : Mon Aug 13 2007 - 16:26:29 CEST