CMS Computing Hamburg

Wednesday 21 Mar 2007, 10:30 → 12:30 Europe/Berlin

SR 7A / Building 7

Bi-weekly meeting concerning computing issues in Hamburg

Minutes of the CMSHH-Computing-Meeting, March 21st 2007, SR 7A

(comments to florian.bechtel@cern.ch)

1. Commissioning of new workgroup servers
   • Hartmut has checked for missing software and all needed packages should be installed now. There are some instabilities with xemacs from time to time.
   • There is no CMSSW installation for SL3-64bit-architectures at the moment, but Hartmut has documented in the CMSHH wiki how to enforce the build of 32bit-libraries during compilation ! TO DECIDE: Do we want to include this already in the AFS installation? Note that this problem should go away for SL4. !
   • Setting up a new CMSSW project, compiling and running the code all works on the new machines
   • General scheme:
     • uhh-cms012-015 - login workgroup servers
     • uhh-cms011 - SL4 installation for Markus to run Millipede
     • uhh-cms016 - setup as 012-015
     • uhh-cms017-020 ongoing
     • the old fileserver (uhh-data01) is mounted on the new workgroup servers ! TO DO: export new disks to old workgroup servers (i.e. uhh-cms01-03) ! for disk space on the workgroup servers (and on the file server once it becomes available): contact cms-admin@desy.de (Hartmut, Alan, Birgit)
   • There is an incompatibility of kinit between the old and new workgroup servers. So at the moment code checkout only works anonymously.
   • The new file server is installed. Birgit has tried to get in contact with the responsible admin. Partition (20 TB total) : Proof 10 TB Scratch small... reserved to be decided
   • Birgit will ask IT to have standard nagios monitoring for the new workgroup servers
2. PROOF
   • Hartmut gave a very impressive demonstration of PROOF, see his talk for details
   • In the discussion, the point was raised whether Hamburg needs an additional batch system. We still need to understand the typical use-case for this (for example private Monte-Carlo production? Analysis on private ntuples? etc.). Birgit reported that DESY-IT would prefer SGE technology.
3. New hardware
   • Birgit has written a "BA" and sent it to Christoph who then gave it to Joachim Mnich.
   • There will be 2 new prod-agents which will be reserved for official Monte-Carlo production
4. Load Tests
   • Yves and Florian have installed PhEDEx 2.5.1 on the new CMS VO Box and started agents in the dev instance. We are waiting for CMS-PhEDEx-Admins to send us a small test sample. Transfers in the last 24 hours can be monitored at href=http://cmsdoc.cern.ch/cms/aprom/phedex/prod/Activity::QualityPlots?graph=quality_all&entity=dest&src_filter=FZK&dest_filter=T2&no_mss=true&period=l24h&upto= We still need to understand why the transfer quality is only about 60%.

10:30 → 10:50 Commissioning of new workgroup servers

o remaining issues o installation of CMSSW for 64bits ? o policy for data disks ?

CMSHH Wiki page on Local Computing Issues http://www-flc.desy.de/hamburgcms/cmshamburgwiki/Local_Computing_Issues

10:50 → 11:10 PROOF tests

CMS-Wiki on PROOF testing https://twiki.cern.ch/twiki/bin/view/CMS/ProofTesting

Slides proof.pdf

11:10 → 11:30 Plans / Need for / Status of further hardware (e.g. funded by DESY)