Wahid, Alessandra, John, Brian, John, Shaun, Stephen, Ewan, Matt, Chris, Govind, Jens, Dave, Pete 1. DPM Incredibly exploding DPM issues? CASTOR also had CGSI_gSOAP problems. UCL problem (https://ggus.eu/ws/ticket_info.php?ticket=85467) May need some more manpower at UCL? Try to avoid formal processes. Does Duncan remember his DPMology? What can we do remotely? Also problems with Durham, but Glasgow have access to it (ie the ScotGrid support infrastructure.) Also lack of manpower. This is a problem for the PMB. And Bristol. Funding was cut because they didn't deliver with the manpower; within CMS, it was discussed not to have storage at Bristol. Would the ScotGrid support model help? A good fraction of LT2 does have the same model, Duncan can log in. There are also hosting issues, rather than just technical storage issues. CGSI-gSOAP errors frequent, but errors can arise from lots of different things, not necessarily just gSOAP. Segfaulting issues - all known about? At Manchester, hit by the problem that disk server being down can cause a segfault of the DPM. Liverpool applied a patch - not a simple update, quite a few different things. John B will look back to find the yum commands (see chat). Matt did a yum clean; so did Alessandra but it didn't seem to work? Maybe we need a "proper" recipe from Ricardo. This is the gLite DPM. gLite 3.2 upgrade path - very ancient and slightly old DPMs - is gLite support going away? https://indico.egi.eu/indico/conferenceDisplay.py?confId=1096 There is a recipe for moving from gLite to EMI without reinstall. It was in the broadcast, with a link, which didn't seem to have the right information. For a DPM pool server there is no need not to reinstall, as long as you don't destroy the data obviously. Alessandra will have to upgrade everything... when does gLite 3.2 _actually_ finish? Would be bad to postpone till the end of the year. Do we need to coordinate so sites are not all down at the same time? It's in the GLUE schema, if you trust what it publishes; also "unset" for DPM in Wahid's monitoring. The "falsely" publishing gLite 3.2 problem... (CASTOR at least - the BDII helpfully added some attributes...) Maybe have sites pick the weeks they want to use for upgrading; or could do it as Ewan did with just one pool at a time. 2. Quick task review https://savannah.cern.ch/task/?group=srmsupportuk Could GridFTP into DPM's head node. 3. Remaining storage-related tickets (review) - we don't do this as often as we should(?) or is it sufficient to do it in dteam? https://ggus.eu/ws/ticket_info.php?ticket=68853 - SL4 DPM at RAL?? https://ggus.eu/ws/ticket_info.php?ticket=68859 - gLite 3.1 @ D'ham https://ggus.eu/ws/ticket_info.php?ticket=80155 - Bris https://ggus.eu/ws/ticket_info.php?ticket=83627 - RHUL negative value https://ggus.eu/ws/ticket_info.php?ticket=85438 - file lost? https://ggus.eu/ws/ticket_info.php?ticket=84461 - Lancs timeout RHUL problem - pool node being down for 1-2 weeks, with hardware problems. Data not declared lost... used tool from Ricardo to move data from the pool node. Still publishing negative values. Brian has access to the database and has been investigating. Lancaster timeout - weird network problem, networking upgrade got delayed, but should solve the problems. 4. AOB [10:00:15] Shaun de Witt Anyone speaking? [10:00:50] Stephen Jones joined [10:00:52] Stephen Jones left [10:00:59] Stephen Jones joined [10:01:28] Ewan Mac Mahon joined [10:01:49] Shaun de Witt left [10:03:47] Shaun de Witt joined [10:03:56] Shaun de Witt OK - sound sorted [10:05:44] Matt Doidge joined [10:06:13] PPRC QMUL joined [10:06:24] Govind Songara joined [10:06:49] Ewan Mac Mahon There comes a point where you can't run a decent site without the manpower. [10:07:00] Ewan Mac Mahon Whether that's Durham or Bristol or whereever. [10:07:11] Ewan Mac Mahon This really is the PMBs problem now. [10:07:55] Ewan Mac Mahon My main plan is to just grumble at the PMB at the next GridPP meeting. [10:08:03] David Colling joined [10:08:21] Ewan Mac Mahon And London..... [10:12:40] Ewan Mac Mahon It's harder than you'd think to make this work - we've tried it a bit with Bristol, and it's difficult. [10:12:41] Pete Gronbech joined [10:12:58] Ewan Mac Mahon Also, arguably a bit pointless - you're better off putting the resources somewhere decent. [10:15:43] Pete Gronbech someone typing [10:17:41] Jens Jensen It's me typing, writing notes from the meeting, sorry. [10:17:46] John Bland don't laugh: yum update "DPM*" "dpm*" python-dpm perl-dpm [10:17:52] John Bland that seemed to do it [10:18:52] John Bland fixed our segfaults straight away [10:19:16] Ewan Mac Mahon If you're still running gLite I wouldn't rule out the possibility of dumping the whole thing and moving to EMI - in my experience it's less painful than you might think. [10:19:58] Ewan Mac Mahon I still wouldn't do that though. [10:20:02] Ewan Mac Mahon It's such a bad idea. [10:20:09] Ewan Mac Mahon Baaaaad. [10:22:09] Govind Songara when SL5 OS end of support? [10:22:33] Ewan Mac Mahon 30th Nov is a nightmare date [10:22:40] Matt Doidge Is DPM EMI2 viable? (I lose track of these things) [10:22:42] Wahid Bhimji having to do a reinstall of OS -is clearly putting people off.... [10:22:51] Wahid Bhimji yes I have DPM EMI1 [10:22:55] Wahid Bhimji 2 I meant [10:23:19] Govind Songara I mean OS, if it is nearing early next year then I would rather wait for it [10:23:44] Wahid Bhimji SL5 EOL is _way_ in the future [10:24:01] Ewan Mac Mahon @Matt mine is EMI1, but I believe so. [10:24:06] John Hill SL5 support is currently ending 31 March 2017 [10:24:32] Ewan Mac Mahon Including EMI2 on SL6 - DPM is really pretty portable. [10:25:11] Ewan Mac Mahon @Wahid - yes, but those people are probably going to be better off just getting over it - a re-install will likely be less hassle than a non-reinstall. [10:25:44] Mark Norman joined [10:26:25] Ewan Mac Mahon Given that DPM nodes all run gridftp, you can just look at their filesystems directly and see what they've got installed. [10:26:46] Ewan Mac Mahon And for the interesting StoRM sites, just email Chris and ask him what he's running. [10:26:48] Wahid Bhimji indeed [10:27:19] Wahid Bhimji I think Sam has "fixed" the unset thing [10:27:26] Mark Norman left [10:27:27] Wahid Bhimji by manual hack for glasgow [10:27:56] Wahid Bhimji so it is not reliable - but 1.8.3 is only EMI I believe [10:28:32] Mark Norman joined [10:28:32] Mark Norman left [10:28:39] John Bland we're planning to upgrade at the same time as our major network reshuffle [10:28:46] John Bland whenever that is [10:30:17] John Bland gotta go [10:30:19] John Bland left [10:30:25] Wahid Bhimji Um I have to leave reasonably promptly. But I think if we make the list we may find there is not so much (for the storage) to do [10:31:41] Shaun de Witt left [10:31:49] Ewan Mac Mahon DPM does [10:31:56] Ewan Mac Mahon Yup. [10:32:02] Brian Davies sorry , had no sound, can the plan for how to schedule SE upgrade to EMI from glite3.2 be pu tinto the chat window [10:32:02] Ewan Mac Mahon Wierd but true. [10:32:31] Ewan Mac Mahon there isn't so much a plan, as an intention to create one. [10:33:01] Ewan Mac Mahon Slightly vague stuff about trying to give each site a week to do their downtime and making sure they don't overlap. [10:33:16] Ewan Mac Mahon I think we'll be taking it to email. [10:40:02] Wahid Bhimji thanks ... bye