Swiss WLCG Operations Meeting on 20100715

• Date and time: 2010/07/15 at 9:30
• Place: EVO, password: chipp
• External link / EVO: http://evo.caltech.edu/evoNext/koala.jnlp?meeting=MeMMMu2M2sDaDi9v9uDv9e

Agenda

• Site status
• WLCG Workshop in London last week
• CreamCE
• Report on German Atlas Cloud Meeting
• CMS vobox
• AOB

Attendants

• ATLAS: Marc Goulette
• CMS: Leonarda Sala
• LHCb: Roland Bernet
• CSCS: Pablo Fernandez, Peter Oettl

Minutes

• Site Status (PF)
  • mostly running fine
  • some Lustre errors which are mostly transparent to users
    • PBS at some point fails creating tmp directories; mds failover "solves" the problem
    • under inverstigation; also opened ticket to Sun support
  • Some Atlas jobs put high laod on mds which slows down other jobs
• WLCG Workshop in London last week. http://indico.cern.ch/conferenceOtherViews.py?view=standard&confId=82919
  • It's better to have a local BDII-top for job reliability
  • Is there a way to see if users gravitate out of our site? According to the VOs this is not happening
    • ATLAS
      • no user blacklisting
      • only during downtime or major problems
      • MG happy with CSCS; faster than other sites
    • CMS
      • cms users are quite happy
      • no user blacklisting at least from Swiss physics community
    • LHCb
      • users can't blacklist
      • centrally only during downtimes or major problems --> ticket
  • The storage model is under discussion, maybe the push model is not good enough and would be good to have some pull/caching mechanisms
  • Altas complained about the storage pledges in Tier2s, they only provide 75% of the pledges. We were one of the red sites, but all free space was assigned to the VOs and now we're fine.
  • The experiment software is creating troubles to sites. NFS is not the best place nor it is Lustre. Looks like CernVmFs could be a good solution.
  • Having 10 Gbps bandwidth is about three times more than we're required to, very good feedback.

• CREAM-CE (PO)
  • hostname: cream02.lcg.cscs.ch
  • installation was pretty smooth
  • up for a week
  • processed about 5000 jobs; mostly CMS
  • VO reps should tell their VOs that it's available for testing
• German Atlas Cloud Meeting (MG)
  • see attached slides

• CMS VO BOX (PF)
  • taken offline (e-mail)

Action Items

• VO Reps: announce cream02 available for testing