- Minutes of the LT2 Technical Meeting -
- 31/05/06 -

O.van der Aa
D.Rand
G.Mazza
G.Sciacca
D.McBride
K.Septhon

By Phone:
G.Rybkine
H.Nebrensky


= Planning for the Glite 3.0 upgrade =

We have tried to squeeze the upgrade planning to be all upgraded 
before the 10July. 

UCL-CENTRAL: currently upgrading, Gianfraco told that william is stuck with the fact that there is no
combined glite and lcg tar ball for the worker nodes. See in http://glite.web.cern.ch/glite/packages/R3.0/R20060502/
there is: "combined gLite and LCG Standard Worker Node (WN)" with an rpm list but no tarball

ovda has send a mail on tb support but no answer. 
ACTION ovda: find out where the tarball is


QMUL: Could not fix a definitive date. Cabling of the new rack ongoining. 
Giuseppe proposed to setup at least the CE. OVDA agreed but need to fix a date. When the CE
is setup will need to put the site down to upgrade the remaining elements. Will try to fix a date for the CE upgrade.

IC-HEP: 05/06-09/06 if Olivier can be freed for doing it (Mona is on holiday up to 7/09). Depends also 
if olivier go to QMUL for the CE upgrade

BRUNEL: 05/06-09/06

IC-LESC: Start to upgrade on the 19/06, Mona and David. 
David Will be on holiday from 25/06 to beginning of july. 
David will not be involved in gridpp as from october (starting a thesis)

RHUL: Grigory/Duncan ok for 19/06. Expect one week

UCL-HEP: Gianfranco not there from the 25/june to 30/june. Will start upgrading 3/July. 
Would be ok to upgrade at that time since the new storage could be there. Hence we could improve the 
dpm performance by removing the nfs mount. Gianfranco announced that the Gb connection will be there soon. 


--> Important for the upgrade <--
- Check that the OPS vo is enabled with the yaim configuration given as in 
http://www.gridpp.ac.uk/tier2/london/Data/vo-lt2.txt
- Log the problems/solutions using the wiki log section of your site. 
I have created a placeholder in each site page:
http://www.gridpp.ac.uk/wiki/Brunel#Glite_3.0
http://www.gridpp.ac.uk/wiki/UCL-HEP#Glite_3.0
http://www.gridpp.ac.uk/wiki/UCL-CCC#Glite_3.0
http://www.gridpp.ac.uk/wiki/QMUL#Glite_3.0
http://www.gridpp.ac.uk/wiki/RHUL#Glite_3.0

The calendar can be made public (need gmail account)


= Improving DPM performance =

UCL-HEP: will get Gb connection soon and remove nfs mount
RHUL: nfs mount is readonly. update on the 100Mb connection ?
BRUNEL: Update on the 100Mb connection ?
IC-LESC: compiling dpm on solaris. Problems with the voms compilation. progressing. 


= AOB =

Duncan/Henry reporting high abort rate of lhcb jobs at Brunel (ggus ticket 8738) . Could be due to the limit in the number of 
queued jobs. Jobs submitted are scheduled and then directly aborted because pbs refuses to queue more jobs. 
Suggestion was to remove the queue limit and see what happens. 

Brunel/RHUL not clear why number of waiting jobs is sometimes very high. Could be qstat that is not responsive. 
Gianfranco told that they cached the qstat to avoid those problems. ovda thinks that it could also be diagnose that 
hangs. dwm explained that the information plugin log in syslog.