pre-GDB on Batch Systems

Europe/Zurich
31/S-028 (CERN)

31/S-028

CERN

30
Show room on map
Description

Monthly meeting of the WLCG Grid Deployment Board See also Twiki GDB area for actions and summaries. This pre-GDB will be co-organized with HEPiX.If you plan to attend the meeting in-person at CERN, please register to help organize logistics.

This meeting will probably focus on HTCondor experiences as this is what has been adopted by most sites in the last year. But it is by no mean restricted to HTCondor and I welcome any input from sites that used another approach/product. In addition to the batch system migration itself, the day will also cover ARC CE experiences at several (non NDGF) sites and the status of CREAM CE with HTCondor. And for every batch system we'll discuss, we'll discuss advanced topics  like multicore job support and memory limitation. 

The goal of this meeting is not to repeat again the (very good) session that we had during last HEPiX in Oxford (https://indico.cern.ch/event/346931/timetable/#20150327.detailed) but rather to have an interactive meetings where sites in the process of migrating can share their experience and their problems and get input from sites with a more advanced expertise. Sites who have not yet started to look at a Torque/MAUI alternative but are considering to do it are also welcome to participate and will certainly get a lot of useful information!

The proposed agenda reflects this goal. The first hour  and an half of the meeting will be for very short reports from sites who are in the process of migrating (or just completed it) to identify issues to be discussed in more details in the second part of the meeting. For this to work, I'd like any participating site who wants to report about where they are to contact me and to prepare NOT MORE than 2 slides (solution/configuration adopted, main issues) for this initial presentation. You are welcome to have more slides that could be used later in the discussion but it should remain very short on each topic to let time for discussion and experience sharing.

 

Participants
  • Almudena Montiel
  • Andrea Sciaba
  • Andreas Petzold
  • Andrew David Lahiff
  • Catherine Biscarat
  • Dal Pra Stefano
  • Denis Pugnere
  • Gianni Mario Ricciardi
  • Guillaume Philippon
  • Götz Waschk
  • Helge Meinhard
  • Iain Bradford Steers
  • Ian Peter Collier
  • Jerome Belleman
  • John Gordon
  • Michel Jouvin
  • Nadia LAJILI
  • Pepe Flix
  • Sebastien Gadrat
  • Stephen Jones
  • Ulrich Schwickerath
  • Vanessa HAMAR
  • Vincent Brillault
  • Wadenstein Mattias
    • 10:30 10:40
      Introduction 10m
    • 10:40 12:00
      Report from recent site migration experience

      Including migration in progress
      Share choices and identify issues

    • 12:00 12:30
      BDII publication 30m
      BDII use cases GIP development needs ? How/who?
    • 12:30 14:00
      Lunch 1h 30m
    • 14:00 15:30
      HTCondor
      • 14:00
        Initial configuration and monitoring 30m
        Interaction with MW Security: account mapping Accounting groups/fair-share Python bindings
      • 14:30
        Memory limitation 30m
      • 15:00
        Multicore jobs 30m
        Draining: defrag daemon, DrainBoss... Backfilling: status of passing job parameters
    • 15:30 15:50
      Coffee Break 20m
    • 15:50 16:50
      CE: alternative to CREAM
      • 15:50
        ARC an HTCondor CE 40m
      • 16:30
        Accounting 20m
        New CE issues with APEL accouting Direct publication vs. APEL box CERN approach
    • 16:50 17:00
      Wrap-Up 10m
      Further coordination