Choose timezone
Your profile timezone:
LCG Service Coordination Meeting
Description
Present
Jamie, Gavin, Harry, Maite, Piotr, Ulrich, Olof, Nicholas, Simone, Patricia, Andrea, Joel Closier (for Roberto), Miguel dos Santos, Ian, Louis, Miguel Anjo
Introduction
• No points
AAS (I.Neilson)
• No additional points to the status report.
WMS (U.Schwickerath)
• The report of problems with volhcb01, the LHCB job log collector, running out of memory lead to a long discussion with J.Closier. He suspected a memory leak in gridftpd . H.Renshall thought the architecture was wrong since once a backlog of transfers builds up, for whatever reason, an avalanche effect occurs. It was agreed H.Renshall would organize an urgent technical meeting.
DMS (G.Mccance)
• J.Closier added that lhcb are seeing many (10%+) slow transfers using the srm end-point but not with the castorgrid end point. These are direct transfers of Montecarlo event files from remote worker nodes to CERN. J.Shiers pointed out that the supported computing model is for these to be first stored at an SE on a ‘close’ T1 then transferred to CERN, if needed, via FTS. Studies to continue.
IS (P.Nyczyk)
• PN emphasized the need for support for the ops VO. MB added that gLite 3.0.2 will go to the pre-production service next week for a week then, if all is well, be rolled out to sites before the mid-September service challenges.
Experiments
• Alice : (P.Mendez) Still have no end point at RAL nor NDGF. Had some srm_put problems at Sara. IN2P3 is giving preference to CMS transfers this week but ALICE will carry on. They reached an aggregate peak transfer rate of 150 MB/s but only for a short time.
• Atlas : (S.Campana) Have had a post-mortem analysis of their July Tier0/Tier1 exercise. Overall it was good with no major issues for CERN but some problematic T1 sites.
• CMS : No special issues. Their export exercise reached 350 MB/s this morning (the target is 500 MB/s).
• LHCb : Their important issues had already been raised – job logger crashing and slow transfers to srm.cern.ch.
AOB
• M.Anjo announced that they needed to install a new Oracle security patch and J.Shiers advised him to carefully follow the agreed procedure for such changes.
• G.Mccance announced they were preparing a new FTS software.
• P.Nyczyk asked if they should continue their SFT test which copies directly from a remote worker node to a CERN SE. J.Shiers said this was still a good use case for what is expected to be an infrequent operation.
There are minutes attached to this event.
Show them.