Hello Maniel,
for PBS, if you want the PBS nodes to use "cp" instead of "scp", you have
to modify the file $PBS_HOME/mom_priv/config (PBS_HOME usually is
/var/spool/pbs). Here you have to use $usecp to tell pbs_mom to use cp:
$usecp hostname:path local_path
path and local_path are different if you use different mountpoints on
pbs_mom and pbs_server. E.g:
$usecp ce.hp.com:/foo /bar
is being used if you have the same filesystem mounted under /foo on you CE
and mounted under /bar on every other machine
You can also use wildcards, e.g.
$usecp *.hp.com:/home /home
Best regads,
Andreas
--
Andreas Unterkircher
IT Department
CERN
CH-1211 Geneva 23
http://cern.ch/openlab
On Mon, 28 Mar 2005, Sotomayor, Maniel wrote:
> Hello Folks,
>
> Thanks for helping!
>
> Right now my our farm does not share the home directories with the CE. I installed the site using YAIM. Which steps are involved in changing from a scp to an nfs shared home dirs environment ?
>
> Regards,
> ./MS
>
> -----Original Message-----
> From: [log in to unmask] [mailto:[log in to unmask]]
> Sent: Fri 3/25/2005 2:45 PM
> To: Charles Loomis
> Cc: Sotomayor, Maniel; LHC Computer Grid - Rollout
> Subject: Re: [LCG-ROLLOUT] aborted jobs
>
> On Fri, 25 Mar 2005, Charles Loomis wrote:
>
> > [...]
> >
> > The problem looks to me more like a problem with the job manager and
> > non-shared home directories. Maniel is using the "pbs" job manager as
> > the "lcgpbs" job manager doesn't work with MPI. I seem to remember that
> > the pbs job manager doesn't work unless the home directories are shared.
>
> Exactly! Maniel, your CE must share the grid home directories with the WNs!
>
|