[Lustre-discuss] 1.6.4.1 - LBUG on MDS

Bernd Schubert bs at q-leap.de
Mon Jan 21 04:42:50 PST 2008


On Monday 21 January 2008 13:02:43 Niklas Edmundsson wrote:
> On Mon, 21 Jan 2008, Bernd Schubert wrote:
> >> Hmm, ext3-unlink-race.patch didn't apply at all, and looking manually
> >> I see no obvious place to apply it to.
> >>
> >> Diffing the ext3-trees between kernel.org 2.6.18.8 and debian 2.6.18 I
> >> see no patch that obviously touches ext3_link/ENOENT/i_nlink:
>
> <snip>
>
> >> So I think that this bug is most likely present when using vanilla
> >> kernel.org 2.6.18.8 too...
> >>
> >> Thoughts/suggestions?
> >>
> >> My gut feeling is that the MDS code is relying on some corner case
> >> behaviour of ext3, and that this behaviour is changing with newer
> >> kernels...
> >
> > Could you try this patch, this is what we are using and what should be in
> > debians lustre svn
>
> I'm building using the pkg-lustre/trunk repository, so I already had
> this patch when the MDS bug:ed out.

Have you checked the patch is really applied? I'm not always closely following 
debians svn and Goswin who is usually syncing our patches with debians was 
the last weeks rather busy with other things (as was I).

[...]

> In general it seems to me that a good chunk of the lustre code depends
> on specific versions of ext3 behaviour, which coupled with the fact
> that it's not shipped with lustre but patched based on the kernel
> version of ext3 only seems to produce surprises at regular intervals.
> I really can't wait until this gets into userspace, which if I have
> understood correctly will happen with the lustre 1.8/ZFS-thingie...

Sure, in userspace it probably will be much easier.

Cheers,
Bernd


-- 
Bernd Schubert
Q-Leap Networks GmbH



More information about the lustre-discuss mailing list