On Tue, Oct 28, 2014 at 06:53:44PM -0400, John Morrissey wrote:
> On Thu, Oct 23, 2014 at 06:38:32PM -0400, John Morrissey wrote:
> > On Tue, Oct 21, 2014 at 06:05:32PM +0200, Bastian Blank wrote:
> > > On Fri, Oct 17, 2014 at 12:48:55PM -0400, John Morrissey wrote:
> > > > Bastian, I'm happy to help, so please let me know if there's anything I
> > > > can
> > > > do.
> > >
> > > As the lustre client support in 3.16 is broken, could you review the
> > > changes in 3.17 and see if a backport looks reasonable?
> > It seems the client is broken on systems with more than one CPU, which I
> > kludged around with 'cpu_npartitions=1' - is that the breakage you're
> > talking about? I haven't had a chance to look at or try 3.17 yet.
> I looked through the changes in the Lustre client between 3.16 and 3.17,
> there were a few bugfixes, but mostly stylistic changes. It's a clean
> backport with the addition of the attached patch, to account for the
> hlist_add_ API changes.
> I built backported 3.16 kernel packages and did some basic tests on a
> mounted Lustre filesystem, which seemed fine.
> > > Also a test setup with some rhel/centos 6 server would be helpful.
> > Will do. I'm actually working with Intel to do some benchmarking with
> > their Amazon Web Services offering for Lustre, which runs CentOS 6 on the
> > server.
> Are you looking for access to a test setup with CentOS 6-based Lustre
> servers, or someone to test the Debian kernel's Lustre client with CentOS
> Like I said, I did some basic smoke tests with the Lustre client in 3.17
> (and that version of the client backported to 3.16). I might be able to
> make a Lustre server setup available for a short period (a week or so), or
> I could run some other tests.
Hi Ben, Bastian, is there anything else I can help with on this?
It'd be great to have the mainline Lustre client working out of the box on
Pkg-lustre-maintainers mailing list