P14 Reprocessing: Meeting of 12-Nov-2003: 10:00-11:00 VRVS Desert.

Agenda

1. News from CPB.
2. Site Certification (Intro by Phil?).
3. Transport speed (Intro by Rod).
4. Status of Sites.

Minutes

Participants

Topics

1. News from CPB:
   • Restart reprocessing using p14.05.02, which fixes a bug found in certification. Remaining problems in p14.05.02 will be repaired by TMBfix.
   • At the same time switch to non-selective reprocessing.
   • Once the reprocessing is running some site might switch to MC (reprocessing or production) depending on the efficieny which we can reach.
2. Site Certification:
   • Phil: FNAL plots are missing.
   • Five partitions were made at Cluedo (non-selective) compared to non-selective from Lyon. Differences remaining can't come from Processor differences. Maybe it's due to possible DB access on Cluedo (mag field).
   • Remote sites differ only by processors. All differences are considered small.
3. Transport speed:
   • Rod: Central router was getting very low percentage of transfere requests. Reconfigured in SAM: New station rp-router with higher priority to enstore.
   • All site should switch to rp-router. In case of problems we can setup another station.
   • 65GB cache of the new station shouldn't be a problem. It only needs to hold as many files as sim. transports are going on.
   • New datasets will be assigned by Mike which should be used for stressing the new station. All sites should start the transport on Thursday 10am CST.
4. Storing DSTs: should be done by the sites. We'll provide an update of the copy scripts which check the number of events in the TMBs and DSTs before transportin/storing them.
5. Site Reports:
   • GridKa: Th. Nunnemann: All datasets assigned are processed. In selective reporcessing a file might have no selected event which makes it look like not being processed. (No problem in future)
   • Lyon: Stopped by data transfer. 8.4M transported, 8M processed, using 10CPUs only. LHC experiments are expected to start the challenge soon.
   • Nikhef: Running on NIKHEF local farm (not EDG) and storing files to d0mino, but permissions are such that files are only readable for SAM.
   • NPACI: -
   • SAR: Expect another 1 or 2 weeks before ready to go.
   • UK: Currently running RAL only, Imperial and Manchester are ready to go, but due to the transport speed not used at the moment.
   • WestGrid: Both sites up. Transport is working (after fixing IP-routing tables). Tracing problems arising from worker nodes being on a private network.

Action Items:

• Prepare an update of the copying script which checks the number of events before transporting/storing it (Mike and Daniel).
• Start transport of newly assigned datasets using the new routing station "rp-router" on 13.11.2003 at 10am Chicago time (All sites).
• Swtich to p14.05.02 as soon as it appears an redo the certification dataset (All sites).
• Create Tarball for p14.05.02 asap (Ian)

Next Meeting