50 likes | 75 Views
This meeting in Lyon on February 3, 2009 discusses the problem of accessing data at T1 for users of T2 and proposes solutions from CERN, FNAL, and CCIN2P3.
E N D
CMS T1-T2 CCIN2P3Data Access CMS-France-CCIN2P3 Meeting February 3, 2009 Lyon Problem CERN Solution FNAL Solution T2_FR_CCIN2P3 ? Tibor Kurča Institut de Physique Nucléaire de Lyon T.Kurca - CMS-CC
Problem Definition Access to data in T1 for users of T2 - data stored at T1 only - non productions jobs to be run at T2 - avoid transferring the same data 2x - avoid T1T2 intra CCIN2P3 transfers - access to T1 data by local users status –Site configuration :CE - different for T1 & T2 SE - one for both T1&T2 PhEDEx – only T1 nodes status – jobs: temporaryhack inCRAB_2_4_4(Jan 23,2009) users jobs can access T1_CCIN2P3 data without show-prod = 1 option …. all T1 are masked in DLS by default except CCIN2P3 should be solved at PhEDEx level with different T2_FR_CCIN2P3 node T.Kurca - CMS-CC
CERN Configuration dCache: the same for T1 & T2 ???? Disk pools : ?????? PhEDEx nodes: T1_CH_CERN_Buffer, T1_CH_CERN_MSST2_CH_CAF T2 end point diskonly but having access to T1-tape (CASTOR) ? SE: srm-cms.cern.ch (T1) - caf.cern.ch (T2) -real, not only alias CE: No CE registered (T1&T2) ???? Data subscription: - T2 subscription – if data already at T1 then no actual PhEDEx transfer again …. just stageing to the right disk - developed dedicated T1CAF local download agents to ensure replication to the correct service class - using space tokens to separate T1T1_CH_CERN from T1T2 transfers - a special CAF agent to register download data in the local CAF DBS T.Kurca - CMS-CC
FNAL Configuration 1 dCache: the same for T1 & T3 (independently configured in SiteDB & PhEDEx) all files accessible for everyone Disk pools : common for T1&T3 PhEDEx nodes: T1_US_FNAL_Buffer, T1_US_FNAL_MSST3_US_FNALLPC SE: cmssrm.fnal.gov (T1) - cmsdca2.fnal.gov (T3) – alias for cmssrm ! T3 end point diskonly is a CNAME to FNAL’s CMS-SE cmssrm.fnal.gov CE: production jobs run on T1; users run on T3 Data subscription: - T1 subscription doesn’t mean automatically also data at T3 ! - T1 data are fully accessible via CRAB to T3 users (no blacklists…) - user data are subscribed to T3– track kept by the T3-manager - caveat: don’t subscribe the same data to T1 & T3 ! ….. No mechanism to prevent the data deletion from both PhEDEx endpoints if the data subscribed to the both - data subcribed to T3 is still archived by dCache but PhEDEx doesn’t know that ! (???old PhEDEx) T.Kurca - CMS-CC
CCIN2P3 Configuration dCache: the same for T1 & T2 ???? Disk pools : only for T1 ? create specific for T2 ? PhEDEx nodes: T1_FR_CCIN2P3_Buffer,T1_FR_CCIN2P3_MSS no T2node for the moment ! create T2 node T2_FR_CCIN2P3 real or only alias to T1 node ? (no subscription synchronization should be needed …) SE: ccsrm.in2p3.fr (T1) - ccsrm.in2p3.fr (T2) create T2 specific …. alias to ccsrm.in2p3.fr ? Wouldn’t this be enough ? CE: different CE for T1 and T2 ! SE is CloseSE to both the T1 & T2 CEs; if the jobs cannot go to T1 why they don’t got to T2 ??????????????? Isn’t T1-SE close to T2-CEs ? - problem that there is no T2_PhEDEx node ? Data subscription: T.Kurca - CMS-CC