There is some define in libc
so you might want to recompile everything related after modify kernel,
like LAM, SSH/RSH ...
that's my experience :-)
Qing
On Thu, 24 Mar 2005 14:06:50 -0500, Brian Barrett <brbarret_at_[hidden]> wrote:
> On Mar 22, 2005, at 12:11 PM, Clifton Kirby wrote:
>
> > Hello I am using lam 7.1.2b18 with the following configure line,
> >
> > /configure --prefix=/opt/mpi/lam-7.1.2b18 --with-fd-setsize=16384
> > --with-ro
> > mio-flags=-file_system=nfs --with-boot-tm=/opt/torque --with-boot=tm
> > --with-
> > memory-manager=darwin --disable-tv --disable-tv-queue
> > --with-lamd-boot=120
> >
> > Right now I can run reliably on 1008 processors but when I attempt
> > anything
> > larger than that some of the lamd daemons won't start up during
> > lamboot that
> > are always the last few nodes. I would like to scale bigger than this
> > but
> > would like to determine the max number of CPU's the lam community has
> > tested. I'm using the hello and pi programs included with the LAM
> > distribution and also Torque 1.2.0p2.
>
> To follow up on the list, we had some off list discussion about this
> issue. It looks like it might be a low disk space issue on a couple of
> the nodes, but the LAM team is awaiting confirmation of that. If
> anything interesting comes up from the off-list work, I'll post it
> here.
>
> Brian
>
> --
> Brian Barrett
> LAM/MPI developer and all around nice guy
> Have a LAM/MPI day: http://www.lam-mpi.org/
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
|
