r2 - 18 Apr 2008 - 12:11:25 - RobertGardnerYou are here: TWiki >  Admins Web > MinutesSRMApr18

MinutesSRMApr18

Introduction

  • Background: _As discussed during the facilities meeting yesterday, we will have a special meeting dedicated to SRM V2.2 issues.
  • Coordinates: Friday, 4/18, 11:00am Eastern, Phone: (605) 475-6000, Access code: 735188, Dial "*6" to mute/un-mute* (PLEASE mute when not speaking, thanks!)
  • Attending: John Brunelle, Wei, Paul, Bob, Rob, Charles, Marco, Shawn, Kaushik, We=njing, Tadashi, Wensheng, Michael, Hiro

Experience with SRM V2.2 at AGLT2

  • Started transition. AGLT_SRM site was defined in ToA already. Had been using this for ntuple and AODs. (Will transition this to ATLASDATADISK).
  • Had been using 20TB gridftp endpoint AGLT2 readonly.
  • New location in dCache - high performance storage pools. ATLASMCDISK area brought up with space tokesn - plan was to migrate the gridftp data. Wenjing used srmcp.
  • Afterwards would update LRC. And change the replica location in central catalogs - to use AGLT2_SRM from AGLT2.
  • MIgration time was long - srmcp picky and slow.
  • There was also a network problem as well - subnets at BNL not available.
  • Right now everything is pointing to an srm v2.2 compliant area, but without space tokens.
  • 100K files. uses srmcp and a list. during the copy srmcp would get "reset" every 7K files. gridftp server to one of 4 pools, so performance was good. failures could be recovered, but by hand. for the transfers - FTS may be used, though did not
  • Change site location in central catalogs. dq2-register-replica for datasets. dq2-delete-replicas to delete the old site registration.
  • PANDA will monitor only ATLASMCDISK.
  • For moving outputs - need to use a space-token-aware command.
  • Paul - pilot can only use lcg-cp site mover for space tokens.
  • glite 3.1.

Wei

  • Migration issues to ATLASMCDISK and ATLASDATADISK - re-subscribe data?
  • Keep the old site? Tadashi thinks Panda can handle this - both locations kept in the SE list. Gradual migration. Since all sites are using a single LRC, shouldn't be a problem.

Data migration issues

  • Discussed gradual migration.
  • No data presently needs to go into ATLASMCDISK.

Pilot/Panda issues

  • PandaMover will need to be fixed.
  • For reading, you can continue to use dccp.
  • PandaMover for xrootd - Wei can provide a tool, to avoid an srm interface and write directly.
  • In pilot3, there are two movers for reading and writing.
  • Paul - there are possible conflicts due to regional hard-coded settings.
  • Charles would like to review the lcg-cp dependencies.

Plans/schedule for completion of all T2 migration by April 25th

  • All comfortable with gradual migration
  • Sites setup atlasmcdisk, atlasdatadisk.
  • Sites should let panda team know when the migration is ready. Need to make switch in site services.
  • Panda mover needs to updated for space tokens.
  • Target - have endpoints ready by April 25.
  • But, need to get lcg-cp and panda mover and panda data mover changes.
  • glite-url-copy, deprecated.

Aside - pilot2/pilot3

  • Would like to switch to pilot3 for every site.
  • Since condor-g probs resolved, now ready to switch
  • aglt2, swt2, ou, slac running pilot3
  • mw, ne need to make switch
  • Torre needs to setup site in pilot3
  • pilot2/pilot3 jobs can run simultaneously.
  • What about ATLASENDUSER? Yes, set this one up too - where Pathena.

AOB


-- RobertGardner - 18 Apr 2008

About This Site

Please note that this site is a content mirror of the BNL US ATLAS TWiki. To edit the content of this page, click the Edit this page button at the top of the page and log in with your US ATLAS computing account name and password.


Attachments

 
Powered by TWiki
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback