Present: RAL: Brian, Jens Edinburgh: Wahid Lancaster: Matt Liverpool: Stephen, Mark Oxford: Ewan, Pete Glasgow: Sam, Gareth Sheffield: Elena RHUL: Govind RALPP: Chris QMUL: Chris 1. DPM upgrade paths - we still have some oldish systems out there that ought to upgrade. gLite 3.1, 3.2, EMI. Recommended version is still 1.8.0 or 1.8.1 depending on whether you are on gLite 3.1 or 3.2. SL4 now being unsupported for security. We discussed whether if the site is on SL4 or 3.1, they should upgrade to SL5 with gLite 3.2 first, before making the jump to EMI or UMD, or whether they should go to UMD or EMI directly. Sites may just go to UMD or EMI directly, since they are reinstalling anyway. Do we recommend EMI or UMD? Apparently there is supposed to be more testing done in UMD, but not necessarily fixing it. UMD is supposed to contain the packages that will be working together, but this may not be required for DPM. It is not clear what the value added testing is? We should probably just use EMI. DPM bug circulated to tb-support about not using EPEL? You need the priorities right. The EMI install relied on packages from EPEL but EPEL got upgraded to something that wasn't compatible. Steve Traylen maintains the packages. 2. There is at least one ancient StoRM out there, too... Publishing hacked at Bristol; maybe get Bob Cregan to do it. He is no longer GridPP funded. 3. Loose ends: wiki, WebDAV, NFS4, TEGs WebDAV - Sam talked to Ricardo and we now know what has been tested for DPM. Matt expecting to do work next week. NFS4 - low level testing for DPM. For dCache, need to get the test instance back at RAL. Can we support clients directly with SL6? MAybe in SL6.3, possibly not in SL6.2. Definitely in SL7, RedHat at least. TEG - trying to get stuff done by the end of March, discussed IO last week (see chat). Aiming to strengthen the recommendations. Maybe a presentation in three weeks. 4. Other storage stuff - Terena Storage TF, see links below. Jens contributing to ASPIRE task force for LCG. 5. AOB Brian looked into slow transfers to TRIUMF and ASGC. Seem close to getting good rates compared to Westgrid T2 - QMUL getting half which is good, some are getting much less. QMUL have use of a 10 gig link which did increase the rate (do we have "before" and "after"). So maybe networking is an issue. Sites should let us know if they change the networks so we can see if it makes a difference. Ewan suggested getting some perfsonar install instructions customised to us - deploy it before the network upgrade so we can see the difference. Also note the DPM workshop on the 27th, should be video. With EVO as backup. Monday Opening Monday - Bonac (ARNES) Monday - Nexenta Monday 15:30-17:00 Monday - ToR Discussion 1 Tuesday - 9:00-11:00 Tuesday - 11:30-13:00 Tuesday - 13:30-15:00 [10:01:22] Wahid Bhimji joined [10:01:27] Matthew Doidge joined [10:01:59] Mark Norman joined [10:02:01] Stephen Jones joined [10:02:04] Ewan Mac Mahon joined [10:02:26] Sam Skipsey joined [10:02:51] Gareth Roy joined [10:03:15] Elena Korolkova joined [10:04:07] David Crooks joined [10:05:04] Elena Korolkova should we upgrade to glite dpm 1.8.2 if we are forced to move emi? [10:05:17] Elena Korolkova in some time [10:06:01] Govind Songara joined [10:06:09] Ewan Mac Mahon I'm certainly planning to go straight to the 'latest' on my SL4 boxes. [10:06:21] Ewan Mac Mahon Where latest is either EMI or UMD. [10:06:40] Chris Brew joined [10:07:22] Ewan Mac Mahon Argh a dcache user! [10:07:24] Wahid Bhimji I think EMI [10:07:26] Wahid Bhimji http://www.eu-emi.eu/emi-1-kebnekaise-updates/-/asset_publisher/Ir6q/content/update-10-24-11-2011#DPM_v_1_8_2_task_22248 [10:07:27] Stephen Jones Daniella indicated that UMD was "not working out". [10:09:22] Stephen Jones If you have V&V after dev, then bugs go back to dev. If you have V&V after V&V, then where do the bugs go! [10:10:37] Ewan Mac Mahon I think is is rather in scope for the storage group - this isn't policy, it's a question of what works best. [10:10:55] Pete Gronbech joined [10:10:59] Stephen Jones It's in scope - what software do we use! [10:11:30] Ewan Mac Mahon I think my formulation of this would be "I'm going for EMI - anyone any objections?" [10:13:05] Stephen Jones It make sense to me, alright [10:14:34] Ewan Mac Mahon Well, that's another alternative. We've pulled directly from upstream DPM before - should we go back to that? [10:14:53] Ewan Mac Mahon (I'm thinking probably not) [10:17:28] Ewan Mac Mahon That's all based on my understanding of the emails - I've not actually tested this. [10:19:59] Wahid Bhimji so ... for this EPUL globus issue [10:20:02] Wahid Bhimji https://ggus.eu/tech/ticket_show.php?ticket=79354 [10:20:14] Wahid Bhimji what do we recomend people to do - any ideas... [10:20:45] Wahid Bhimji they say they will put a fix in dpm in 1.8.3 but with current versions we will need a workaround [10:22:41] Wahid Bhimji I can ask in person to the DPM peeps at the Community Forum on Monday - but I think we need a plan even before that. PS - reminded there is a Community forum on Monday [10:22:41] Wahid Bhimji https://indico.cern.ch/conferenceDisplay.py?confId=172988 [10:22:48] Matthew Doidge Should be able to get stuck in next week [10:23:09] Matthew Doidge (Fingers crossed nothing explodes) [10:23:14] Queen Mary, U London London, U.K. joined [10:23:45] Wahid Bhimji yeah sounds like its a bit far away from really possible from a DPM perspective. [10:23:51] Wahid Bhimji (NFS 4.1 that is ) [10:24:12] Ewan Mac Mahon Wahid - sounds like a special update of this 'dpm-dsi' package would solve the EPEL issue for now, so I'd suggest we get a copy of that package somewhere. [10:24:43] Ewan Mac Mahon Alternatively it's fun and games with yum version locking and so forth, and that's not going to be pleasant. [10:24:58] Sam Skipsey I tend to concur with Ewan. The easiest thing to do is to not get those packages from EPEL. [10:25:33] Wahid Bhimji (thats different to what Ewan says? - he says do get those packages from EPEL - and get new dpm-dsi ? [10:25:46] Ewan Mac Mahon I think RHEL6 is supposed to have pNFS as a 'preview'. [10:26:02] Ewan Mac Mahon i.e. works but not supported. [10:26:07] Wahid Bhimji https://twiki.cern.ch/twiki/bin/view/LCG/TopicsDataStorageTEG [10:26:25] Wahid Bhimji https://twiki.cern.ch/twiki/bin/view/LCG/SMIOLanStorageEvolution [10:26:47] Ewan Mac Mahon Yes; I'm still expecting to get the bulk of the globus stuff from EPEL - I don't think there's much choice there (is it available anywhere else?) [10:27:17] Ewan Mac Mahon The options I can see are: a) get the latest, and the new dpm-dsi package to work around the EPEL breakage, [10:27:37] Ewan Mac Mahon b) get the older EPEL packages from EPEL using a list of specific package versions. [10:27:45] Ewan Mac Mahon I'd incline towards 'a'/ [10:28:36] Sam Skipsey Right. [10:30:04] Ewan Mac Mahon I also care about this in a fairly pointed fashion because I'm going to want to install some new boxes with SL5/EMI quite soon, and it would [10:30:13] Ewan Mac Mahon be useful to be able to install them in a working state. [10:30:40] Ewan Mac Mahon So "don't upgrade" isn't completely going to help me. [10:32:32] Wahid Bhimji I emailed ricardo and cc'd you in - will forward reply to the list [10:33:54] Wahid Bhimji Maybe it is good to see "before and after" [10:35:47] Chris Brew left [10:37:24] Wahid Bhimji https://indico.cern.ch/conferenceDisplay.py?confId=172988 [10:38:13] Govind Songara left [10:38:14] Elena Korolkova left [10:38:15] Gareth Roy left [10:38:16] Sam Skipsey left [10:38:16] David Crooks left [10:38:16] Mark Norman left [10:38:17] Wahid Bhimji left