Minutes 15 December 05

  • VRVS/phone: Lee(CMS), Stefan(ATLAS), Iris(BNL), Eric(FNAL), Elisabetta(CNAF), Barbara(CNAF)
  • @CERN: Radovan(POOL), Birger(ARDA), Marco Clemencic(LHCb), Dirk(3D), Luca(CERN DB), Jacek(CERN DB)
  • Apologies: Gordon(RAL), Doris(GridKA), Eva(3D), Pierre-Etienne(IN2P3)

  • General points:
    • DB readiness workshop dates 6 Feb (Mon) at CERN * try to limit to a 1 day workshop in the IT Auditorium/VRVS * looking for volunteers to host next workshop at a tier 1
    • The weekly meeting slot is confirmed for all of 2006 (same time Thu 16-18 - same CERN room 513-1-024)
    • GridKA and IN2P3 still problems with testbed replication setup * IN2P3 is applying patch for Oracle problem * GridKA - waiting for results from OS upgrade
    • Kuba left the 3D project after finishing his technical student appointment at CERN - many thanks to him!
    • GridKA confirmed they will setup 2 2-node RACs for ATLAS and LHCb
    • as only BNL and CNAF were represented in the meeting it was agreed to shift Luca's presentation about the CERN RAC setup into a meeting in January

  • the minutes of the last meeting were accepted with minor modification: ARDA test so far covered only the functional tests.
  • Roundtable Tier 1 production service setup with invited Tier 0 DBAs
    • BNL : Iris reported that the db server hardware setup BNL is progress well. Next step will be Oracle software setup
      • Q IRIS: BNL currently use only one cluster interconnect interface (CERN T0 uses two). Is this a problem?
      • A Luca: This affects load balancing and failover, but BNL could add additional card later if this should be required to increase perfromance/availability.
      • Q Dirk: what hardware/ software setup is used at BNL for the storage network?
      • A Iris: using a switched fibre channel setup. May use Oracle ASM, but not clear yet.
    • CNAF : Barbara reported that the server installation is progressing also at CNAF
      • Currently testing ocfs as shared file system for data
      • Luca pointed out that ocfs2 (which comes with significant improvements) would be coupled to the 2.6 linux kernel (meaning RH4)
      • CNAF plans to move from SLCx to RHES deployment before production to insure Oracle support

  • Roundtable experiment/project replication tests
    • ARDA-AMGA tests
      • Birger: ARDA tests on hold for the moment as replication problems with IN2P3 affect the replication timing.
      • Dirk: should continue as soon as the IN2P3 problems are fixed
    • LHCb-COOL tests
      • Marco: LHCb tests ready for replication and waiting for connection string from RAL
      • (Gordon setup the required databases after the meeting and will provide Marco with the necessary information)
    • ATLAS-COOL tests
      • Stefan reported first results comparing local (CERN-CERN) to remote database(Oxford-CERN) access
      • significant slow down (factor 10) for remote connections (ping latency difference some factor 30)
      • this is not unexpected and just confirms that Tier 1 replicas will be required for some workloads (eg Conditions data)
      • Stefan will try to further minimise the number of server round-trips by using bulk retrieval functionality in COOL
    • CMS-Conditions tests
      • Lee summarised on the recent discussions between the FroNtier and POOL teams
      • several improvements to the frontier client and pool were discussed including handling of target schema, specification of proxy servers and failover and improvements to the error handling/propagation are being defined by Radovan and Sinisa.
      • Squid performance test being done by Lee:
        • frontier access showed no significant improvements when passing through a local or remote squid. Why?
        • after the meeting Rado pointed to some older email exchange - caching had been disabled during the earlier tests to make schema changes visible to frontier clients. He has in the meantime changed the implementation of the POOL frontier plugin to only disable caching for schema queries, but enabled it for payload of CMS application queries.
        • cache invalidation policy needs to be discussed in one of the next 3D meetings.
      • Lee also reported about the recent test in the validation setup which were performed with the help of Jacek


This topic: PSSGroup > MinuteS14December05
Topic revision: r1 - 2005-12-21 - DirkDuellmann
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback