Summary of storage phone conference, 2 Nov 2005 Present: Lancaster: Brian, Matt Edinburgh: Greig, Phil RAL Tier 1: Derek, Jeremy, Steve Queen Mary: Giuseppe Imperial: Olivier Manchester: Alessandra Durham: Mark Cambridge: Santanu Glasgow: Graeme Liverpool: Michael, Paul RAL Storage: Owen, Jiri, Chris K, Jens (chair+mins) Apologies: None As usual, if anyone joined after the roll call, and I did not recognise their voice (or didn't say anything), let me know. This is record attendance! 19 people! 11 Sites (RAL storage is not the same site as RAL Tier 1, there is a road between and some gates :-) Good stuff. 0. Quick site status and support-for-site and plans. Wasn't that quick, but I thought generally useful. 0.1. Durham (Mark) - DPM - nothing changed - planning to start this month (~mid). 0.2. Sheffield - no-one attended, Greig has been in touch with Andrew Beresford. Report that they have tried dCache install. => cc Alessandra on _all_ email NorthGrid correspondence until further notice. 0.3. QMUL (Giuseppe) - Alex has been working with Olivier on DPM install, but only has 40 GB capacity for now (planning 20 TB), and problems with internal hostname != external hostname (not multi-NIC). 0.4. Liverpool (Michael, Paul) - dCache. dCache running on test system, network improved, Owen will get in touch. Paul looking into performance and scalability when running dCache pools on WNs (100 GB disk per WN). 0.5. Manchester (Alessandra) - farm not installed yet. Alessandra has some experience with dCache pools on WNs, but on a smaller scale (for now). Reports that it works. 0.6. Cambridge (Santanu) - DPM. DPM running, Santanu feels confident that he can also reinstall if necessary. 1. LCG 2.6.0 and dCache Owen ran into a weird showstopper bug yesterday which he has now documented in the HOWTO. Owen is currently testing the upgrades $${1.5.\{2,3\} \atopwithdelims.\} 1.6.5} \rightarrow 1.6.6$$ The dCache in LCG 2.6.0 is $1.5.\{2,3\}$ (thus used by the Tier 2s) whereas Tier 1 is using 1.6.5. The most important new thing in 1.6.6 is that PNFS now uses postgres. 2. YAIM and the gLite install It was agreed that GridPP taking over dCache YAIM from LCG is a Good Thing(tm). Discussion followed about YAIM's future. The new gLite configuration tool uses python (the other scripting language :-) and XML input. You will (may) be interested in the gLite deployment experiences in PPS (click the slides link on rhs): http://indico.cern.ch/sessionDisplay.py?sessionId=28&slotId=0&confId=a0514#2005-10-27 And the deployment tools session: http://indico.cern.ch/sessionDisplay.py?sessionId=49&slotId=0&confId=a0514#2005-10-27 Alessandra is part of a triumvirate who's got the job to "ensure a smooth transition" to the new gLite deployment tools (the others are Oliver Keeble and A di Meglio). Graeme says Glasgow will join PPS so he should be able to report experiences with gLite configuration. 3. Hepsysman items. What do we need to cover and who's doing what. Thanks to the group's efforts and that of the coordinators', SRM deployment is progressing well. By 5 Dec, we should have nearly all sites running at least one SRM. Pete has asked for DPM information. *I* think that we should cover things such as the YAIM->gLiteConfig path, for the storage middleware. We should also cover things sysadmins need to know in preparation for SC4. Like how to get an SRM 2 installed, and how to test it. In any case, it is likely that Owen and Greig will [be volunteered to] give presentations on storage middleware. We'll see how big a slot we can get. As expected, we ran out of time, and we did not cover the information publishing. For those of you who attended EGEE, please send me additional storage middleware issues/outcomes that you think we should keep an eye on, or are otherwise worth knowing.