Minutes of the storage EVO meeting, 29 Sep 2010 Present: Glasgow: David, Sam Edinburgh: Wahid QMUL: Chris RHUL: Govind Bristol: Winnie RAL: Brian, James, Jens (chair+mins) 0. Review of actions - quick, obviously (see below) 1. Overview of outcome of EGI in Amsterdam? Main question is UMD - DPM is in, StoRM is probably in, dCache is in. Sam will check whether StoRM is in (action). 2. Roadmap for DPM Main discussion is whether we should contribute officially to DPM. On the plus side: * It would be good to integrate toolkit closer; * Better access to "inside" knowledge; * Opportunity for wider dissemination of our contributions; * More collaboration (this is a project metric!) On the "minus" side, we are then obviously committing to doing the work, so the work will have to be planned into other activities we do. For example, if we need to commit physical resources, we need to understand how they are maintained and who commits them. Particular things we could get involved with: > toolkit integration - never b0rken toolkits again after upgrade; > Testing prereleases; > Performance testing (we're clearly already doing much here); > Monitoring, eg Nagios integration > Instrumenting - CERN may enable more "fine grained" instrumenting in the DPM code; if we contribute to this, the work could include writing code. Jens points out that from a project perspective we need to make sure GridPP is acknowledged; e.g. by adding a logo somewhere or something. Sam points out that our code is of course (C)'ed. NFS4 for DPM promised Really Soon Now(tm); deadline demo at CHEP. We could test this: Glasgow have a "spare" DPM. Disk deployment - making disks visible - we note that many sites are going through procurement, and disks should be published. It currently looks like T2s are short of pledges. 3. Testing in general revisited; see http://www.gridpp.ac.uk/wiki/Storage_and_Data_Management_Testing This page has been updated with lots of more stuff. Note that we also need a test schedule for NFS4 now. Brian has done some work with James T. Requires kernel 2.6.35 (or 33?) for the client. As regards experiment specific testing, CMS are getting involved in Hammercloud now. LHCb of course do not use storage at T2s, but may still be interested in providing details about testing. [Moreover, others may be using Ganga outside LHCb, and use resources at T2s.] 4. The end of the quarter is nigh! Repent! I mean report! (still needed!) * Interesting meetings attended? * Publications appeared? * Interactions with other WLCG (outside UKI) * Work with Users? (this is new but interesting) * Anything else interesting to report? The PMB seem to be interested in other stuff we do; particularly, I suspect, if it involves sharing knowledge from GridPP. Documents and ongoingments * dCache install report (Brian) - to be presented 6 Oct 2010. Quick report: dCache book replaced with wiki; and the wiki is helpful. More or less, dCache runs out of the box. (i.e., it runs when it's taken out of the box, not that it runs to get out of the box!) * Testing NFS4, plans - need to work on this. * pcache testing, status - new version, retest at T1? No, retest not needed; they are not that different. pcache enabled at Glasgow. Hit rate is not good. Maybe turn on selectively, the effect should be reduced load on network. Also studying RAID0 on tmpdir which effects pcache. There is a need to fine tune disk for WNs, particularly on manycore systems. pcache uses hardlinks so must stay within the same filesystem, which is not ideal for some circs. * Anything left to do with SSD? SSD testing on WNs done; still need to test databases. Need preliminary writeup to appease the managment. 5. Remaining interestingities from All Hands - proper summary?! Probably worth following up on the building-support-community stuff. Most of the AHM was about "bridging the chasm" (see http://nationalgridservice.blogspot.com/2010/09/mind-gap.html for references to the plenaries which are recommended reading) 6. AOB There was a longish discussion following the meeting about a "cernatschool" activity, and what should the baby^W VO be called? vo.cernatschool.org seems reasonable. Behold the GGUS ticket: https://gus.fzk.de/ws/ticket_info.php?ticket=62383 ACTIONS 402 02/06/2010 Clean up the wiki ALL Low Open Ongoing - but if you make a change in the wiki that you think people should know about, please let us know! Conversely if you find a really old and crusty page then either contact the maintainer - or fix it yourself. 406 04/08/2010 Complete T2K syncat for Lancaster Matt+Sam Med Open Done. 408 18/08/2010 Forward T2K Lancaster syncat to Chris W Sam High Open Done. Data sent to Chris. There may be a T2K discrepancy at QMUL. Sam is working on automated checking, not using LFC dump. Some problems with unicode characters appearing and syncat is (can be) XML which barfs on unescaped general unicode characters. [09:52:32] Sam Skipsey hello, all. Just off to coffee - hopefully will be back before 10. [09:58:16] Jens Jensen What would we do without caffeine [10:00:13] Wahid Bhimji joined [10:01:05] Queen Mary, U London London, U.K. joined [10:02:21] David Crooks joined [10:02:38] Brian Davies joined [10:04:15] James Thorne joined [10:06:57] Govind Songara joined [10:11:14] Winnie Lacesso joined [10:21:54] Jens Jensen http://www.gridpp.ac.uk/wiki/Storage_and_Data_Management_Testing [10:22:03] Jens Jensen http://www.gridpp.ac.uk/wiki/Storage_and_Data_Management_Testing [10:23:35] Wahid Bhimji https://www.gridpp.ac.uk/wiki/Local_File_Access_TestSuite [10:29:28] James Thorne Another meeting calls. [10:29:31] James Thorne left [10:40:08] Winnie Lacesso left [10:40:25] Govind Songara left