Re: [Xen-users] lvm as raw disk and barriers

["Fajar A. Nugraha" <list@xxxxxxxxx>]

On Mon, Sep 5, 2011 at 3:21 PM, Alexandre Chapellon
<a.chapellon@xxxxxxxxx> wrote:
> Hello,
>
> I was previously using LVM2 volumes as partitions for my PV guests.
> For some reasons I now have to switch to a model where an LV is a raw disk
> and not a partition.
> This works flawlessly at first look. But today while searching in dmes logs
> I found the following errors:
>
> ##############
> ...
> [    8.515912] EXT4-fs (xvda3): mounted filesystem with ordered data mode
> [   12.961268] blkfront: xvda: write barrier op failed
> [   12.962398] blkfront: xvda: barriers disabled
> [   12.963434] end_request: I/O error, dev xvda, sector 3410072
> [   12.964285] end_request: I/O error, dev xvda, sector 3410072
> [   12.965438] JBD2: wait_on_commit_record: sync failed on xvda1-8 -
> disabling barriers
> [   13.820957] JBD: barrier-based sync failed on xvda3-8 - disabling
> barriers
> ...
> ##############
>
> I didn't have any trouble with fs consistency sor far, but I know not using
> barriers is more prone to falling into FS problems (even with modern
> journalised fs).
> Why am I getting this issue? is it related to using LV as raw disks? Is
> there anything I can do to re-enable barriers?

I'd say don't bother. Most likely it never worked in the first place.

>
> Am using Xen 3.2.1 on Debian Lenny system, kernel 2.6.26-2-xen-amd64 as
> dom0, debian squeeze , kernel 2.6.32-5-amd64 as guests. backend storage
> devices are lvm volumes from multipathed FCs SANs.

LVM doesn't fully support barriers until 2.6.33
(http://en.wikipedia.org/wiki/Logical_Volume_Manager_(Linux)#Caveats),
so again, most likely you never had it working in the first place,
regardles whether you use it directly as raw block device or put a
partition on it. Add to that the old dom0 kernel version, if
everything works then the best you can do is just ignore it.

FWIW, even with newer kernels (e.g. 3.0 dom0, 2.6.38 domU) and
partitions, I still get "blkfront: xvda: barriers disabled". The
difference is that the message appears early during boot and without
any "I/O error" messages before that.

-- 
Fajar

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users