Notes of US ATLAS Facility and Operation phone meeting. SLAC disk space availability is behind other T2s (item 3). -- Wei Yang | [log in to unmask] | 650-926-3338(O) ________________________________ From: [log in to unmask] [mailto:[log in to unmask]] On Behalf Of Popescu, Razvan Sent: Wednesday, February 21, 2007 12:21 PM To: [log in to unmask] Subject: [Usatlas-grid-l] F&O meeting notes 2/21 Attendance ... I don't see the point (am I missing something?) 1. DQ2 v0.2.12 upgrade: - Sites done so far: BNL, BU, UTA (2 instances), SLAC (on the new system; the current remains as is and will be phased out in few months) - Active work at UMich o New instructions and new package (Need more volunteers to test them. OU, interested?) o Pbs. from the 64bit OS: * Need 64bit MySQL and Python provided as local service and pointed at during the installation. * The creation of a new MySQL would consequently require the migration of the current LRC data. * If the above services are provided and the LRC data is migrated to the new catalog, the rest of the Dq2 installation should work. o Patrick and Shawn will work it (this afternoon). o Additional problem: repeated callbacks (since Jan). Why? Under investigation. - NEXT: OU - NEXT: MW 2. LRC upgrade: - Necessary to overcome the limitation to 256char in physical file names - Requires a new schema, which requires MySQL v5. Also requires a modified version of POOL, without an internal limitation of filenames to 256char. Available (at least) at BNL. - So the stack of upgrades is: o Upgrade to MySQL v5 o Apply new schema o Replace old POOL with the "enhanced" version o Migrate old LRC data o Recompile Python MySQL libs (to integrate the new MySQL v5 libs) o Recompile/Refresh DQ2 install(?) - The above assumes that no POOL usage exists outside PANDA/Production (otherwise the users will find themselves with the wrong POOL version). User access is assumed to be performed through the web interface. - BNL will setup a Twiki page/site to work together at the details of the upgrade (Dantong). - Hiro will place his instructions (from BNL's experience) on this page and all are encouraged to contribute. - When all the details, concerns and issues will be worked out, we should obtain a "community sanctioned" recipe... that we'll follow on a mutually agreed schedule. 3. AOD replication: - The US data management model anticipates the replication of 100% of the AOD dataset at each T2 site (except SLAC - 10%) to support analysis activities. - The recent substantial increase in AOD event size prompts concerns with the T2's ability to support such a large dataset. - Is this model still feasible? - A more accurate estimate of total needs will be available soon. - Currently available: o MW ~50TB o SW 12TB (addt'l 50TB+ in few weeks) o SLAC 4TB (+18TB mid March) o AGL ~ 35TB (+21TB on order) - The distribution of data will be handled automatically (given the subscription percentage) from T1 - The proper balancing of workload at split installations is a concern. - To be continued... 4. Site updates: - AGL: o Order placed (Dell), quad cores, 21TB (RAID50). Storage initially served via NFS. Later will try (grid)NFS over dCache. - MW: o "Proxy not valid" errors observed at MWT2_UC. Under investigation. - BNL: o Farm upgrade to SL4 o dCache upgrade yesterday. Still working at the optimization. o DQ2 backlog formed due to the disruption. Currently clearing up. - SW: o Work in progress at the new cluster installation. - SLAC: o Requisition out this week. Installation expected in early April. o Prototype xrootd installation in progress (gsiftp frontend), using small size federation. o DQ2 site functional. o Panda preparations in progress. 5. AOB: a. Don't forget the UCSD T2/T3 meeting travel arrangements. Have a nice week! R