Summary: you need to modify /opt/lcg/etc/lcg-info-dynamic-
scheduler.conf to map your new pool groups for ATLAS sgm and prd
users onto the correct VOView of your batch system.
1. Make a copy of lcg-info-dynamic-scheduler.conf to work on (e.g.,
lcg-info-dynamic-scheduler-new.conf).
2. Edit lcg-info-dynamic-scheduler-new.conf, modifying the current
mapping to a VOMS Role based VO view to a simple VO name view, e.g.,
atlas:/VO=atlas/GROUP=/atlas/ROLE=production -> atlasprd:atlas
On the LHS you have the primary group of the pool account (warning
this may be different at your site - use _yours_), on the RHS you
have the name of the VO in the VO view.
It's probably worth doing this for all VOs, however it's in fact the
ATLAS production role that it's absolutely essential that you do.
3. Test the new configuration, running the plugin by hand, e.g.,
/opt/lcg/libexec/lcg-info-dynamic-scheduler -c /opt/lcg/etc/lcg-info-
dynamic-scheduler-new.conf
Check that the GlueVOViewLocalID=VONAME figures are now correct.
ATLAS should now look like:
dn: GlueVOViewLocalID=atlas,GlueCEUniqueID=svr016.gla.scotgrid.ac.uk:
2119/jobmanager-lcgpbs-atlas,mds-vo-name=local,o=grid
GlueVOViewLocalID: atlas
GlueCEAccessControlBaseRule: VO:atlas
GlueCEStateRunningJobs: 215
GlueCEStateWaitingJobs: 0
GlueCEStateTotalJobs: 215
GlueCEStateFreeJobSlots: 109
GlueCEStateEstimatedResponseTime: 0
GlueCEStateWorstResponseTime: 0
4. When you get it correrct (should be pretty easy!) then move then
backup the old configuration and copy in the new one to /opt/lcg/etc/
lcg-info-dynamic-scheduler.conf.
Some other notes:
1. If you have a DNS VO name, then give the full name in the RHS:
snemo:/VO=supernemo.vo.eu-egee.org/GROUP=/supernemo.vo.eu-egee.org/
ROLE=lcgadmin -> snemosgm:supernemo.vo.eu-egee.org
2. You cannot run the current YAIM config_gip_scheduler_plugin or it
will put the file back the way it was. Wise idea to backup the new
file, in case it gets trashed by mistake (there are a few other
obvious tricks: change /opt/lcg/var/gip//plugin/lcg-info-dynamic-
scheduler-wrapper; override the YAIM function).
3. If you have no jobs running for a VO, so cannot tell if the
changes have been effective, then su to an atlas production account
and qsub in a sleep job to the batch system and check you get a 1 for
the VOView.
4. If you haven't moved to pool accounts for prd and sgm accounts
then I'm pretty sure you don't need to do this - but please check the
VOView is right.
5. These instructions are for torque/pbs. For SGE sites I have no
idea what plugin you're using to provide this information!
OK, any questions or problems, please ask me. It's really important
for ATLAS that this change gets made correctly - ATLAS have a lot of
production jobs right now and fixing this will help the right numbers
of jobs to your site.
Sites we know need to do this are:
Edinburgh
IC-HEP (hep-ce.cx1.hpc.ic.ac.uk and ce00.hep.ph.ic.ac.uk)
Birmingham
Lancaster
Liverpool
RAL PP
Sheffield
Bristol
Cambridge
Glasgow (DONE)
Oxford (2ce02.physics.ox.ac.uk and t2ce03.physics.ox.ac.uk)
Thanks
Graeme
--
Dr Graeme Stewart - http://wiki.gridpp.ac.uk/wiki/User:Graeme_stewart
ScotGrid - http://www.scotgrid.ac.uk/ http://scotgrid.blogspot.com/
|