Production and PASTE Meeting, 17th November 2009, CERN
Chair: A.Tsaregorodtsev
Minutes: S.Paterson
CERN: P.Charpentier, J.Closier, Z.Mathe, S.Paterson, M.Rashid, R.Santinelli, A.C.Smith, F.Stagni
Phone: V.Fernandez, R.Graciani, E.Lanciotti, R.Nandakumar, A.Tsaregorodtsev, A.Zhelezov
Apologies:
0. Ongoing Action List
#10919 - Joel has started to prepare the request for new machines to replace those with expiring warranty. Ricardo
mentioned that there were some issues with the externals compilation that look to be resolved. For the new
development machine it is necessary for all files to have g+rw privileges to allow concurrent development with SVN.
Kerberos tokens are not currently supporting the mapping to actual users causing problems for the development
machine.
#12524 - The time estimate for DIRAC releases via SVN is the end of this week / next week.
1. Production Status Report
- Report from the Production meeting
The L0HLT stripping made it to 95% of the 10^9 sample without triggering The main concern at the moment is
automatic recovery of failed jobs, Stuart reported that a new agent is almost ready to do the work. Operationally we
are still suffering from the 134 error and the Gaudi finalisation crash (affecting ~10% of jobs). The former is very
difficult to debug but has become more prevalent with SLC5 and the latter has been affecting us for some months now.
- DIRAC services ports
Andrei commented that the ports were changed over last Sunday and this caused a problem for the running jobs that
do not refresh their configuration. These kind of updates will only be carried out during work days in the future. Now
all DIRAC services are in the appropriate port range except for one on 9300, Andrew will update this one.
Philippe mentioned that there will be two new online streams to support events that crash the HLT and the lumi
events. The ERROR stream will not require offline processing but there may be some processing in the future of the
LUMI events stream. Both streams will be transferred to Castor but to a different pool. There was a suggestion for
using the proposed histogram space token to also store these lumi stream files, this should be pursued with IT. (
ACTION: Joel )
The Brunel release for real data is expected in the next day or two, Stuart is in close collaboration with Marco in order
to test the reconstruction workflow.
2. DIRAC Release Update
- v4r20p4 release
A new file type was added to the workflow library to support the L0HLT1 stripping.
3. DIRAC Tasks
- Infrastructure monitoring (SAM,SLS,Dashboard)
No new updates from Federico. Roberto commented that a new plugin to the dashboard was defined to produce some
overall metrics by job type and site in a similar way to Alexey's page. This should allow site admins to get an overall
picture of the activity on their site. Federico mentioned this may be a suitable candidate for cross-linking from the
DIRAC portal.
- Framework status
Adri migrated the monitoring and logging systems to the Framework system.
- Transformation System
Andrei has moved the transformation components out of Core.
- DMS status
Andrew will remove all LHCb specific stuff and add an LHCb system DMS.
- Bookkeeping status
Zoltan commented that several urgent changes are pending release. A proposal will be circulated regarding updates
to the BK processing pass. Another important requirement is that the concatenation of the processing passes for
stripping productions should be resurrected. A new DIRAC release will be prepared as v4r21 with these features.
- Production Management
Stuart commented that an agent is being developed to achieve 100% processed files. In order for this to be safe the
'reschedule' operation should be disabled for all clients for data processing productions (the JobWrapper and the
JobAgent should still be allowed to reschedule jobs however).
4. AOB
None
There are minutes attached to this event.
Show them.