Hi,
Another gotcha when moving from LCG-2_2_0 to 2_3_0.
We'd previously defined queues per VO by hacking the static part of the info
provide object as recommended in varoius places.
After the upgrade when I did a list-match for dteam|alice|atlas|cms|lchb all
(but one) of the queues appeared:
heplnx103 - ~ $ edg-job-list-match --vo dteam helloworld.jdl | grep
heplnx131
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-S
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-aliceL
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-atlasL
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-babarL
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-cmsL
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-dteamL
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-honeL
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-lhcbL
heplnx131.pp.rl.ac.uk:2119/jobmanager-lcgpbs-zeusL
but for any on the "extra" VOs no match was made:
heplnx103 - ~ $ edg-job-list-match --vo babar helloworld.jdl | grep
heplnx131
heplnx103 - ~ $
So there are three problems in the lcginfo-cepbs-cfg.h file that needed
solving:
1) Why did the 10th queue not show up?
The file only supports a maximum of nine queues.
2) Why did all the queues show up for all the "CERN" Vos?
The file has:
#ifdef SE_VO_ALICE
EXTRA(lcginfo.values_GlueCEUniqueID1_GlueCEAccessControlBaseRule) VO:alice
#endif
for each of the CERN VOs for each of the queues.
3) Why did none of the extra VOs match?
The file doesn't add ControlBaseRules for them.
I guess this might all be fixed if one uses the new script to add extra VOs
but since this is the last LCFGng upgrade I plan to do and I didn't want to
start the extra VO stuff from scratch, I've hacked the file by hand to match
the setup here.
I've attached it so anyone interested can diff it with the original and work
out what I had to do.
Yours,
Chris.
--
Chris Brew ([log in to unmask]) +44 1235 446326
Particle Physics Department
Rutherford Appleton Laboratory
Chilton, Didcot. Oxfordshire.
OX11 0QX. United Kingdom.
|