Go to the Indico Home Page Oldest event Older event Up to category Newer event Newest event Printable version
 
Focus on:  Hide Contributions
  • Compact style
  • Indico style
  • Indico style - inline minutes
  • Indico style - numbered
  • Indico style - numbered + minutes
  • Indico Weeks View
    • Switch to the management area of this event
    Login

    ADC Weekly

    → Europe/Zurich
    3162/2-E01 (CERN)

    3162/2-E01

    CERN

    20
    Show room on map
    I Ueda (Department of Particle Physics-University of Tokyo)
      • 15:40 → 15:45
        possible delay 5m
        ATLAS weekly
      • 15:45 → 16:05
        Hot topics
        • 15:45
          Pilot Factories 10m
          Speaker: Peter Love (Lancaster University (GB))
          • Four APFv2 factories are operating at CERN.
          • Factory queue configuration is generated from AGIS and documented on the AMOD twiki.
            • clouds are asked to confirm/update the queue info on AGIS
          • Initially each factory served all sites but load issues were seen, probably a condor issue, so the configs were split with across the factories. ie. some clouds on one factory, some on others. Things seem stable now.
            • probably the load is due to having more queues. e.g. analysis queues are associated with more CEs now with AGIS than before in the factory config
            • With the current configuration, there is no redundancy for some sites where no cloud-specific factories are serving.
            • it should be possible to have each site served by at least 2 machines.
          • There will be a new release of APFv2 next week which we'll roll out in due course. It has minor bugs fixes, a new ARC plugin, and improved monitoring.
          • There was a question about any difference in behavior between old and new factories. Different numbers of idle/running have been observed
            • the algorithm is different
        • 15:55
          MC production news 10m
          - Current situation on submitted tasks - MC12 AF-II production workflow -- simul.HITS written into TAPE without merge
          Speakers: Andrej Filipcic (Jozef Stefan Institute (SI)), Claire Gwenlan (University of Oxford (GB)), Wolfgang Ehrenfeld (Universitaet Bonn (DE))
          Slides
          Current situation on tasks
          • larger tasks have been submitted, hoping the sites being filled for ~month
          Workflow of MC12 AF-II
          • the output of 'simul' jobs are directly sent to T1_MCTAPE
            • cf. in mc11, the output was sent to T1_DATADISK and then merged before being sent to T1_MCTAPE
          • ND: when only AF-II was running, the dcache pool was occupied by the slow transfers from specific sites. In the end it was not a big trouble.
          • The other T1s are to be asked if they observe similar issues.
      • 16:05 → 16:20
        AMOD report 15m
      • 16:25 → 16:30
        AOB 5m
        • cvmfs 2.1
          there are issues reported on cvmfs 2.1
          need to be followed up
        • network reports (if any) 1m
          - for T2(D)s against T1s - for T1s against T2Ds
        • Analysis Availability Reports 1m
          ES report
          IT report
          UK report