NUC 09.06.2010

Attendance:

Wolfgang Ehrenfeld, Angela Lucaci-Timoce, Andreas Gellrich, Yves Kemp, Kai Leffhalm

News

• new NUC member of ILC/Calice: Shaojun Lu will replace Angela
• should think about contributions to the HEP session at GridKa school 2010:

  Wednesday afternoon 08.09.10, please see http://gridka-school.scc.kit.edu/

• we should prepare a document detailing our wishes about future NAF resources

NAF status

• high load on ATLAS WGS, requires reboot:
  • comments form ATLAS: should shorten reaction time, download on batch?
  • new action items: download queue, reboot procedure
• dCache performance during the last month:
  • bottlenecks seen
  • action item: form for storage problem reports
• new Lustre space at HH and ZN
• need feedback from experiments for resource planning

Existing Action Items

• New action items:
  • NAF: present SL4 work group server usage
    • plots shown
    • repeat at next NUC
  • ATLAS/CMS: plans for SL4 work group servers
    • reduce to one shard WGS SL4, support ends in October
    • ATLAS/CMS talk to the users
  • NAF: check /scratch monitoring (instance full), email notification
    • all experiments want this
    • still open
  • NAF: technical constraints for Lustre extension
    • some boundary conditions, upcoming support issues
    • open
  • NAF: announcement of twitter service
    • done
    • closed
  • NAF/NUC: expert meeting on scratch deletion tool
    • no meeting yet
    • tried a new idea, alas it did not work out
    • open
• Still open action items:
  • NAF/ATLAS: CMT problem (NAF invite experts/ATLAS)
    • CMT developers looking into it
    • ATLAS@DESY can contribute
  • NAF: AFS scratch space creation (existing account migration for

    ILC/LHCb)

    • no migration needed for LHCb, ILC
    • new AFS servers ordered, over-subscription reduced
    • closed
  • NAF: deletion model for /scratch
    • working on it
    • see above
  • NAF: multicore batch job monitoring
    • plots shown, maybe add to accounting in the future.
    • show plot in every meeting

ATLAS report

• login/work group server problems (3-4)
• DB release installation
• install request (kcachegrind, no reply yet)
• ganga installation request
• ganga question
• lustre full (25 May, 3 June, now)
• policy question (2x)
• accounts (2x)

CMS report

• problems with access to dCache user directories (high load from Grid jobs/NAF)
• problems with access to dCache data directories (high load from Grid jobs/NAF)
• dCache user directory creation (not automatic)
• glite32 now used with crab
• need more Lustre space 10-20 TB
• Lustre unavailability on WGS
• would like well defined procedure to report storage problems
• briefly tested tcx053 (no problems seen)
• should formulate CPU and storage request for NAF in 2010/2011

ILC report

• Lustre space full

LHCb report

• No changes since the day 1, except:
  • REAL data are there
  • The total data size is increased
• We are in general happy with NAF, but...
  • dCache access a bottleneck (maybe limit number of parallel jobs per user?)
• problem with very slow data transfer: NAF -> Heidelberg
• need more Lustre space

New Action Items:

• CPU, dCache, Lustre resource requirements development
• download queue
• reboot procedure
• form for storage report
• accounting ideas via e-mail or next meeting or extra meeting
• dCache upgrade time line

Next meeting

- Wednesday 14 July 2010