 |
 |
Hi, about the globus setup: in the past I (or any user) had to follow https://hypernews.cern.ch/HyperNews/Atlas/get/SLACGridJobs/4/3.html Is this not necessary anymore and we should just use Wei's script? (or not even that if the atlas login takes care of this now) -- Dieter Stephen J. Gowdy wrote: >ATLAS SCCS Planning 01Mar2006 >----------------------------- > >Present: Dieter, Charlie, Steffen, Wei, Chuck, Len, Gary, Stephen, >Bob, Richard > >Agenda: > >1. DQ2 Status > >I think is running. DQ2 client can tell it to import. Not clear if it >actually did the import. Discussion if we can put the certificates in >the standard location as all applications want them there and have >different ways of being told where they are. > >Have we tested what the ATLAS productions will use? No, suspect they >use dq2_put. Should ask if there is some dataset names that will not >be exported. > >Need to make sure the grid user that gsiftp will be can write to the >data space. > >Final machine in the queue to be allocated. > >2. mysql replica Status > >In the queue to be allocated. > >3. Trigger Farm Status > >Need to find out more about when and how much. 50 Dell 1650s. Need to >talk to John W still probably. > >The duration is expected to be around two years at maximum. Could be >classified permanent. The racks will not come with them. They are >stacked on rack shelves (two at a time just now, could be three). > >They will be available shortly after the end of the run, which is >expected to be 1st August at the moment. > >4. Questions from ATLAS Grid Meeting: LCG Accounting, VOMS, OSG Version? > >We're exempt as we are obviously not a Tier-2 yet. They are talking >about OSG accounting, Mateo may be working on it. Not sure if or how >that maps to LCG accounting. > >We have a VOMS. We have 0.4.0 running. > >5. AOB > >Need to provide feedback back to ATLAS on the large size of path >elements. Moreover LD_LIBRARY_PATH can cause performance problems (for >example the limited BaBar one doubles how long a kernel takes to >compile). > >Discussion of certificates. Anyone who wants to be in the ATLAS V0 >needs to be in the CERN HR database. Do system folk need to be? There >is a Dteam V0 which is setup to run tests, and for that you still need >to be in the CERN HR database. The CERN HR database doesn't know which >experiment you are on explicitly but that info is used later when you >apply to join the ATLAS V0. Don't know how non-collaborators get onto >the ATLAS V0. The US ATLAS V0 should disappear. We currently use it. > >Action Items: >------------- > >060301 Stephen Ask about test dataset names > >060301 Stephen Ask about long term multiple element path issue > >060301 Systems Determine best way to distribution certificates to all > machines (/etc/grid-security/certificates) > >060224 Stephen Enquire with ATLAS about certificates for DQ2 machine > 060301 Not heard back from ATLAS folk, Wei pointed out a DoE > place to get service certificates. > >060224 Stephen Open ticket for DQ2 machine > 060301 Done. RT 43380 > >060224 Stephen Open ticket for mysql replica machine > 060301 Done. RT 43381 > >060224 Systems Provide DQ2 machine > 060301 In the queue. > >060224 Systems Provide mysql replica > 060301 In the queue. > >060224 Chuck Will check on web server request for DQ2 machine > 060301 Waiting for web server request information from Stephen. > >060224 Stephen Enquire about correct LRC settings for DQ2 with ATLAS > 060301 Done. Now seems to be working. > >060224 Wei Email setup script for globus > 060301 Done. /afs/slac/package/globustools/vdtclient/v1.3.9/setup.*sh > >060224 Stephen Add Globus setup to ATLAS login > 060301 Done. (Some issues with size of environment) > >060224 Richard Discuss ATLAS trigger machines with others in SCCS > 060301 Only limited response from John W was resigned > acceptance... need to work on an actual deployment plan as > there are real issues to be solved. > >060224 Chuck Check on actual availability of SCCS rooms for this meeting > 060301 Only constraint is the Security meeting every > month. Will assume that this ends at 11am. Will update > calendar. > > >