CHIPP + CSCS Face to Face Meeting on 2016-03-11

• Date and time: 10:30
• Place: Zurich Zentrum (CLA J 1)
• External link / EVO: No

Agenda

• 10:30 -- Coffee, presentation and agenda
• 10:45 -- Tier-2 status and plans
  • CSCS Dario/Gianni (45')
  • UNIBE-LHEP Gianfranco (30')
• 12:00 -- ATLAS compute model
• 12:30 -- Lunch
• 14:00 -- Tier-3 status and plans / Open discussion
  • PSI (Fabio 15') photo
  • UNIBE-ID (?)
  • UNIGE (Luis 15')
• 15:00 -- NGI_CH Sigve (60' or shorter)
• 16:00 -- A.O.B.
• 16:30 -- END of the meeting

Attendants

• CSCS: Stefano, Dario, Dino, Gianni, Miguel, Pablo, Michele
• ETHZ: Joosep Pata, C.Crab
• PSI: Fabio Martinelli
• UNIBE-LHEP: Gianfranco Sciacca
• UNIBE-ID: xx
• UNIGE: Luis March
• UZH: Roland Bernet

Minutes

• Decommissioning of CreamCEs can be done now according to CMS and ATLAS but needs to wait for LHCb to confirm it (action on Roland) (it seems to work, but only since very short time to be sure). It was agreed to do it in 3-months time unless CSCS is told not to.
• Storage space for Swiss users need to be given to the international Tier2 side. LHCb and CMS confirm this can be done (if not done already) but ATLAS needs to confirm it within the next two weeks (Action on Roland, then action on CSCS).
• After correcting with the new HepSpec value from CSCS, ATLAS still sees 15% under-pledge in average for last year. CSCS is working on it for some time, but still did not find an answer to the under-utilization. Work in progress.
• The important source of accounting information for SNF and other reviewers is REBUS: therefore accounting needs to be correct there. We (action on CSCS) need to check published values are correct and report to Catherin Noble. APEL accounting records (for CreamCE) cannot be decommissioned until all the accounting records are published up.
• (action CSCS and Gianfranco) Check with Andrea Sciaba why EGI accounting is not precise and correct it.
• Tickets can be automatically routed from the grid-rt queue to the ops-rt queue (where they are not visible from the VO Reps through the dashboard) (action on CSCS) need to check if an automatic notification/CC can be done to a mailing list with all the VO Reps, otherwise CC's need to be set manually by the submitter if they want to make anyone else aware of the issue.
• Downtimes should only be declared in GOCDB and central factories should know what to schedule. No need for reservations or disabling submissions: jobs can be killed when inside the declared downtime. This way resources can be used up to the very last minutes by sort-of-skavenging jobs. (action on CSCS) need to see how this fits with the standard procedures at CSCS for the LHConCRAY project
• Need to check if setting the queue defaults to 5-day jobs is OK for all VOs (action on VO Reps), or in other words, if VOs properly set the length of their jobs instead of using the queue defaults.
• Pledged values should appear in Tier2 statistics next time (action for CSCS and Bern)
• (action) Luis should talk to Gianfranco to see if they could run some grid jobs from outside Geneve.
• (action) Everyone is asked to provide Gianfranco feedback about the procedure to obtain grid certificates.
• Support from NGI unclear (including fees and certificate generation), funded by Sigve's project that finises soon.