Choose timezone
Your profile timezone:
PASTE meeting
→
Europe/Zurich
Description
CERN, bat 2 R-030
To join the telephone conference, click here: https://audioconf.cern.ch/call/0122882
To dial in to the conference:
a. Dial +41 22 7676000
b. Enter access code 0122882
Production and PASTE Meeting, 20th March 2007, CERN
Chair: P.Charpentier
Minutes: S. Paterson
CERN: M.Bargiotti, G.Castellani, P.Charpentier, C.Cioffi, J.Closier, E.Lanciotti, A.Smith, S.Paterson, R.Santinelli
Phone: N.Brook, A.Casajus, R.Graciani, G.Kuznetsov, M.A.S.Miguelez
Apologies: A.Tsaregorodtsev
1. Production Status Report ( J.Closier )
The last Monday meeting was cancelled since no production jobs were running until yesterday. Andrew Pickford
(current production manager), submitted the remaining 2M events. Joel reported that productions were submitted for
a new Gauss version with updated generators.
The issue of RAL storage being filled up by very active Tier-2 sites in the UK (QMUL and Manchester) was raised. It
was agreed to redirect output from these sites to NIKHEF where more storage space is available. ( ACTION: Marianne,
Ricardo )
Joel submitted several reconstruction jobs to test the DIRAC Stager which were successful. Stuart reported that with
the inclusion of Roberto's lhcb-bring-online tool (GridKa, Lyon), all Tier-1 sites are now supported. It was agreed to
put the stager in production after releasing v2r14 of DIRAC. ( ACTION: Stuart, Andrei )
The next step for the Stager is to support the stripping jobs where ancestor files must also be remotely staged.
Stuart reported that this requires a change to the Data Optimizer such that the genCatalog service is used to fill the
StagerDB. Tests will be performed this week. ( ACTION: Stuart, Joel )
Philippe raised the issue of bad quality data being reconstructed (phys-lumi2 and phys-lumi5). The production
number was used and a filter for the Processing DB was changed. It was agreed that any modification of filters should
be treated very carefully. The clean up procedure is now underway, Carmine reported that all affected DSTs have
been removed from the Bookkeeping. The Tier-1 site administrators should be contacted about physical removal of
the remainder of the files. ( ACTION: Marianne )
2. Production Operations Issues
Since several files have been found registered in the LFC without any replicas, Philippe suggested this should be
investigated. ( ACTION: Stuart )
Nick triggered discussion of disk usage at Tier-1 sites. A meeting will be held to discuss the requested and available
space at the Tier-1's at 10am on Tuesday. Also, some discrepancies were discovered between what Andrew's tool was
reporting and the usage claimed by the sites. It was suggested to discuss this at the meeting.
Stuart reported that the StagerAgent and StagerDB (A.Smith) has been committed to CVS pending the next release
of DIRAC. The performance of remote dccp staging has been very encouraging so far (e.g .20 files at RAL in less than
15 minutes). After sending emails to IN2P3 and GRIDKA last week requesting gsidcap and dccp remote staging there
has so far been no response. Andrew has since submitted GGUS tickets to both sites.
The CNAF SRM endpoint is set to change and Marianne attached an intervention report to the Agenda. The LFC
developers have been contacted about modifying the affected replicas but Marianne reported no reply was received so
far. Nick and Philippe reported that the issue has been raised at a higher level but there seems to be no consistent
view concerning existing files. ( ACTION: Marianne )
VOMS proxy problems were observed for one of the Production Managers. Even though the user was registered OK
with VOMSRS, the generation of a VOMS proxy on lxplus fails. Joel reported that it was possible to replicate the error
and Gianluca mentioned that voms-proxy-init fails to work on the Test system. Roberto confirmed that a VOMS
support ticket was recently submitted to report problems with the client installation.
An unresolved issue after temporarily putting the gLite Agent Director into production is that jobs run in the user area
instead of the shared area and subsequently run out of space. It was agreed several weeks ago to suspend
submission through the gLite Agent Director until this is resolved. Stuart suggested using the test system to check
the status at each site. Roberto confirmed it would be enough to check the present working directory of a pilot job.
Test jobs should therefore be submitted to each site. ( ACTION: Gianluca )
Joel reported that the lxgateYY machines are to be renamed to volhcbYY. An alias will be created to the old names to
smooth the transition. ( ACTION: Joel )
A discussion about the Processing DB was started (further to Joel's email).
1) We need to find way to continue the same transformation with the new filter (regular expression)
Dirac2 - The file filter replacement function will be checked
Dirac3 - Allow replication of transformations
2) Complexity of the regular expression:
Dirac2 - Gennady will provide Joel with the regular expressions
Dirac3 - It should be possible to create regular expression in humanly readable form and a testing
framework should be established for the regular expressions. Datasets should be treated
as an object.
3) Reprocessing of the failed files within the transformation:
Dirac2 - Make it possible to clear file assignment of jobs
( ACTION: Gennady )
3. DIRAC3 Tasks
Ricardo mentioned that the CS is still running on lxgate14. The VO-boxes are not picking up the alias. It was agreed
to change the default settings in the dirac-setup. (ACTION: Joel ) When this is done, lxgate14 can be renamed and
reinstalled with SLC4 for the development of DIRAC3.
Ricardo reported that the activity and logging services were almost there. In two weeks the framework will be ready.
( ACTION: Ricardo, Adria)
Gennady mentioned that there have been several iterations of the workflow libraries. Philippe suggested that a
document or presentation be made to describe the scope and functionality of the Workflow package and the proposed
changes for DIRAC3 (schedule for one or two weeks).
4. AOB
None
There are minutes attached to this event.
Show them.