Attending: Jens (chair+mins), John B, Steve, Matt D, Brian, Sam, Winnie, Dan 0. Operational blog posts (new quarter ahoy!) More excellent stuff published from Brian - great stuff! Doesn't have the plot that inspired lots of people though (the kraken/freddy/anemone plot) In operational issues, Govind reported a CVMFS issue but did not attend to discuss it. Something about limitations on repositories that can be mounted. Weirdness in ATLAS space token accounting due to their using WebDAV to delete data, which , as was discussed in previous weeks, leads to accounting being incorrect. Is this a PEBKAC? Or because WebDAV deletion is there, should it be expected to work (a la Scott Meyers on object design: Birds can fly. The Penguin is a Bird. Penguins can fly, but it is an error for them to try to do so.) Alastair is doing an accounting thing for SRM; does it comply with any of the standards, though? We also need them for APEL (meaning standard ones). 1. Compare notes from last week's GridPP (storage and data management related) Brian - Scotgrid distributed/federated xrootd; ARC cache; lots of ATLAS only sites Sam - changes is hard, hence the Power Rangers Proposal. ATLAS seem to prefer fewer sites with more storage. ARC cache is ok, not difficult to manage. xrootd used in expectation of rapid access to files a la Hadoop; better visibility at this GridPP of what experiments want to do and priorities. "cloud easier than grid" as it has less backend diversity? Might be advantage in sharing resources between grid and non-grid eg with HTcondor. Discussion - cloud is sometimes not simple either, some platforms are complicated and some services - not one size fits all, that's also why we need to run different things. Also we have no sanctions against sites that decide to go the path less travelled. What would a site end up doing if it weren't a storage site? Perhaps its storage services would just degrade and of course become relatively smaller as storage sites were upgrading their stuff. Eventually might hit a threshold of how much storage is required to be a "site". Funding model does not "necessarily encourage" the "right" kind of distribution of services. Lancaster and Glasgow will be doing more testing stuff on the requirements arising from the improved experiments roadmaps, particularly caching and suchlike. Less effort at sites also leading to the Power Rangers model. 2. Events, dear boy, events. Who's submitting/going (again SaDM related) * hepsysman 13-15 June (at RAL) o Corrected URL (browser memory error perhaps?) http://hepwww.rl.ac.uk/SYSMAN/June2017/main.html * WLCG workshop in Manchester which Brian says is the week after hepsysman o Corrected URL: https://indico.cern.ch/event/609911/ * HEPiX spring meeting in Budapest, 24-28 April o https://indico.cern.ch/event/595396/ * Networkshop - now! o https://www.jisc.ac.uk/events/networkshop45-11-apr-2017/programme Networkshop of course is too late, and it is not an attractive event due to the organiser's refusal to issue tickets for anything less than the whole three days (after some moaning they did introduce some tickets but they do not apply to all possible participants.) Some of the Usual Suspects will be attending the HEPiX but there didn't seem to be much interest in submitting anything. Hepsysman and the WLCG workshop seem more interesting but the latter may be expensive? However, given that it is in Manchester, it seems like a good opportunity. Sam and Matt will try to write up their stuff on caching, but it may be worth submitting more. 3. AOB Happy Easter...! 1234: (12/04/2017 10:00:49) 1234 is Ste Jones. OK? Jens Jensen: (10:00 AM) ...ok https://indico.cern.ch/event/601969/contributions/ Daniel Peter Traynor: (10:19 AM) it was noticed that the software stack mix in the UK is large: dpm, xrootd, dcache, storm, castor; hdfs, nfs, lustre, glustrefs, ceph; htcondor, sge, torque/mauie, slurm; cream, arc, htcondorce, vac/vcycle. have i missed anything? some openstack and opennebula stuff. Samuel Cadellin Skipsey: (10:20 AM) I agree, Dan, simplification would be easier the problem here partly being that "Grid" means "diversity" to some Matt Doidge: (10:26 AM) Same at Lancaster - it's also the devil we know. John Bland: (10:27 AM) If we were going to combine our resources with Liverpool CSD I could save a lot of bother by just throwing it all out the window, cos it ain't gonna happen. Matt Doidge: (10:28 AM) All my moving plans from DPM start with "If an EMP goes off in my machine room and wipes all my data". Talk to your kids about SE choices before someone else gets to them first. Jens Jensen: (10:30 AM) @Matt you could build an EMP device during the Easter break? Matt Doidge: (10:31 AM) I think that might get me put on some kind of government list... But everyone needs a hobby Jens Jensen: (10:32 AM) that's right maybe Dell can sell you one? Like the hard drive wiper tool Daniel Peter Traynor: (10:35 AM) http://www.wikihow.com/Build-an-EMP-Generator Matt Doidge: (10:36 AM) Where the heck do you get a disposable camera in this day and age... Jens Jensen: (10:37 AM) ... you'd probably want that on http*s* ;-P 1234: (10:37 AM) I'd use a sledgehammer. It's just as good as an EMP, and far more fun. Yeah... they might be right.... it's too soon to tell. It's too soon to tell if the Roman Empire was good thing, too! Matt Doidge: (10:41 AM) I'm definitely up for trying out the caching stuff Sam. Both arc (when we have an arc) and xroot Samuel Cadellin Skipsey: (10:44 AM) https://indico.cern.ch/event/609911/ WLCG Workshop Jens Jensen: (10:45 AM) https://indico.cern.ch/event/518392/registrations/participants Matt Doidge: (10:46 AM) http://hepwww.rl.ac.uk/SYSMAN/June2017/main.html Jens Jensen: (10:46 AM) https://indico.cern.ch/event/592622/ Matt Doidge: (10:53 AM) I miss getting bags and t-shirts and things.