Hi Dave,
Sounds intresting, the thing you absolutley have to have on a WN for EDG is
globus-url-copy, the WN fetches and delivers the I/O sandboxes this
way to and from the RB.
Simple test for its failiure.
globus-job-run hepbf4.ph.qmul.ac.uk:2119/jobmanager-pbs -q redhat9 \
/opt/globus/bin/globus-url-copy file:/etc/group \
gsiftp://gppse05.gridpp.rl.ac.uk/tmp/junk
of course at the moment the binary is not there so there's little point.
There are various ways to acheive this:
+ Install the globus rpms for an WN, - dependency nightmare but possible.
+ Fake a globus-url-copy with a shell script that scp's the files back on to
the CE and then copies of from there. Dave Martin is problably the expert with
that one.
+ nfs mount /opt/edg , /opt/globus from somewhere.
This can be to an alternate location, Lyon the tier1a queue here and scotgrid all
do this. You then have to play lots of PATH games.
For all the above you will need access to up to date CRLs on the WNs so the
jobs trust other resources. There is environment variable though to make these
relocatable.
Steve
On Fri, 6 Jun 2003, D.Kant wrote:
> Hi Steve,
>
> We were testing whether the QMUL EDG frontend can succesfully submit jobs
> to a private farm configured according to Andrew Mcnab's description given
> at: http://www.gridpp.ac.uk/tb-support/existing/
>
> Local job submission using "globus-job-run" work fine.
>
> [kant@ui]globus-job-run hepbf4.ph.qmul.ac.uk:2119/jobmanager-pbs /bin/hostname
> cn001.esc.qmul.ac.uk
>
> cn001 is a worker node on this farm running RH9.0 and is completely
> standalone and has nothing in the way of EDG software installed on it.
>
> However, after successful job submission via a RB using the "dg-job-submit"
> method, the job aborts:
>
> dg_JobId =
> https://gm03.hep.ph.ic.ac.uk:7846/138.37.50.249/14334867785261?gm03.hep.ph.ic.ac.uk:7771
> Status = Aborted
> Last Update Time (UTC) = Fri Jun 6 15:36:53 2003
> Job Destination = hepbf4.ph.qmul.ac.uk:2119/jobmanager-pbs-S
> Status Reason = Failure while executing job wrapper.
> Job Owner = /O=Grid/O=UKHEP/OU=ph.qmul.ac.uk/CN=D.Kant
> Status Enter Time (UTC) = Fri Jun 6 15:36:53 2003
> *************************************************************
>
> Cheers, Dave.
>
>
>
> On Fri, 6 Jun 2003, Steve Traylen wrote:
>
> > It's kind of up, You don't actually need logging and bookeeping to
> > submit a job and get the output. That's intresting bookeeping.
> >
> > Steve
> >
> > On Fri, 6 Jun 2003, Stephen Burke wrote:
> >
> > > On Fri, 6 Jun 2003, D.Kant wrote:
> > > > Is gm03 playing up??
> > >
> > > You might like to check this web page for the status of the three
> > > general-use brokers - this suggests that IC is the only one which is up,
> > > but the last update was 50 minutes ago ...
> > >
> > > http://marianne.in2p3.fr/datagrid/WEB/RB_STATUS.html
> > >
> > > Stephen
> > >
> >
> > --
> > Steve Traylen
> > [log in to unmask]
> > http://www.gridpp.ac.uk/
> >
>
> --
> --------------------------------------------------------------
> Department of Physics | Dr Dave Kant
> Queen Mary College | TEL/FaX: +44 (0)20 7882 5054
> Mile End Road London E1 4NS | e-mail : [log in to unmask]
> --------------------------------------------------------------
>
--
Steve Traylen
[log in to unmask]
http://www.gridpp.ac.uk/
|