Present: Jens, Brian, David, Adam, Chris, Elena, Ewan, Gareth, John H, Matt, Robert, Sam, Steve, Wahid 0. More blog posts!! Pore blog mosts! Plore mog bosts! There is relatively few in number but actually with really really good content. Maybe we need a new metric which uses content * posts. 1. Sharing puppet scripts - didn't see much happening...? There are links to some snippets on the wiki, one of which is 'dpm'. We need to understand a bit more what the benefits are of sharing (if there are any.) Mark S was in Glasgow, and wanted to tidy up the puppet scripts. Maybe needs another nudge nudge - he seems to be further ahead, SL6 may be a good time to move. Also stuff that is not fully polished. Durham may also have stuff, Sam wrote a fair bit of this stuff. Rob Appleyard from the CASTOR team is going to the puppet labs camp in London. Same time as GridPP. We could offer to host the puppet camp some time in the future. Extremely low cost. Docs, link to context to things. Dedicated page on the wiki or KeyDoc, or is it too early? Steve suggests it would be useful. 2. "Big data" update. Cf. http://www.gridpp.ac.uk/wider/interdisc.html Chris and Sam providing input to "small VOs" for EGI CF. Advice on how to use data properly. T2K, NeiSS, and EPIC. Suggesting posters: Dave, Small VOs, GridPP storage. Both Brian and Alessandra have PerfSonar. 3. Poster preparation for EGI CF (and opportunity for demos!?) Other events, like HEPiX? 4. We still have this Savannah group: https://savannah.cern.ch/project/memberlist.php?group=srmsupportuk which contains this: https://savannah.cern.ch/task/?group=srmsupportuk Getting stuff done without an issue tracker, so we don't need it that much. 5. Anyone doing something with DIRAC (ie the other one, not the LHCb one?) I need to do something with them but I am not sure what. Sam has had precisely one meeting with DIRAC where Sam pointed out there were other transfer methods than rsync. Their bandwidth is made up of channel bonded links between Durham and Glasgow; they are playing with GridFTP now. Also EPCC link. GlobusOnline, as they are not terribly certificate savvy. 6. Update on redirectors?? (GLasgow test? Docs?) Which reminds me, did we need more docs on xroot itself? We've got them, and they are kind of boring, but not being used much. Works when you test it. Not running on StoRM - Chris eagerly awaiting EMI3. Works with HTTP but doesn't do the certificate juju. Both snoplus and T2K can use WebDAV mount (with certificates, but with no further checks) Maybe worth getting them to talk to us (again) about the data model. Utility which is pointed at the data - part of DPM? Which would do certificates stuff as well (Chris). 7. AOB DPM - problem with disk load at Sheffield. Still need better ways to control load balancing. This is a problem which is raised as a task in the DPM collaboration. Due for 1.8.7 a way of minimising number of connections, cascade-to-failure limit Elena is encouraged to tell Ricardo that there is a problem. xroot may be a bit more efficient so will push the problem point a bit further off but doesn't solve the problem. [09:53:40] Sam Skipsey Getting Tea. [09:59:23] Wahid Bhimji morning [09:59:29] Elena Korolkova morning [10:00:13] Jens Jensen tea is good [10:01:47] Elena Korolkova there is no problem for me to hear both of you [10:02:00] Steve Jones By "both", does that mean Steve? [10:02:23] Steve Jones I can hear Jens. [10:03:23] Christopher Walker Morning. [10:03:27] Steve Jones Morning [10:03:32] Adam Huffman Morning [10:03:41] Steve Jones I said something [10:09:04] Ewan Mac Mahon It is. [10:09:05] Wahid Bhimji yes - it is [10:09:09] Ewan Mac Mahon But we haven't. [10:09:22] Ewan Mac Mahon I don't have puppet snippets [10:09:25] Christopher Walker Blog posts: you should have a prize awarded at GridPP for best quality blog post (a box of Quality Street perhaps). [10:09:32] Adam Huffman It is worth sharing, though I would say perhaps making things available in Puppet forge is better [10:09:43] Ewan Mac Mahon Mark has been poked about this from a southerly direction too. [10:10:10] Wahid Bhimji people always want to clean first - but I think just having an unclean repo might be good [10:10:45] Wahid Bhimji indeed - there is "official" DPM stuff - that I am testing and that is almost ready - but it will just do the basics instead of yaim [10:11:04] Wahid Bhimji so we want some other examples [10:11:38] Ewan Mac Mahon Indeed. [10:11:41] Wahid Bhimji and liverpool use it right ? [10:11:53] Ewan Mac Mahon If Durham is ground up puppet, that would be ideal. [10:12:25] Ewan Mac Mahon So, action Sam - upload Durham. [10:12:30] Ewan Mac Mahon [10:12:41] Wahid Bhimji please put whatever you have perhaps with a flag for its "completeness" [10:12:41] Ewan Mac Mahon And blog about it, obviously. [10:12:57] Sam Skipsey I'll check with Mike that he's okay with it, and do so. And blog about the puppet config sharing idea. [10:13:24] Wahid Bhimji there are puppet meetings at other times [10:14:58] Adam Huffman I was at a Puppet Camp in Edinburgh last year and it was very good [10:15:14] Wahid Bhimji puppet fundementals training indeed- fkn expensive: [10:15:15] Wahid Bhimji http://puppet-fundamentals-training-london-april-2013.eventbrite.com/?mkt_tok=3RkMMJWWfF9wsRokvqzIZKXonjHpfsX67uosW6G0lMI%2F0ER3fOvrPUfGjI4DRMRnI%2FqLAzICFpZo2FFID%2FCFeZRM%2B%2FdO [10:17:32] Jens Jensen http://www.gridpp.ac.uk/wider/interdisc.html [10:19:58] Wahid Bhimji Can you also cover (or raise) issues such as [10:19:59] Wahid Bhimji s what impact will a gradual reduction in use of SRM have on the smaller VOs? [10:20:13] Sam Skipsey that was going to be in the slides, yes [10:20:23] Sam Skipsey In the "things we need to do for small VOs" bit [10:20:51] Wahid Bhimji OK - that would be a good discussion to have with devlopers etc there - cheers [10:22:48] Brian Davies http://cf2013.egi.eu/programme/posters_demos.html [10:23:44] Christopher Walker We need, I think, a generic version of lcg-cp that works for things other than srm. [10:23:49] Jens Jensen https://savannah.cern.ch/task/?group=srmsupportuk [10:26:05] Jens Jensen Can you still hear me? Brian says we lost network [10:27:27] Ewan Mac Mahon So, anyone actually want an issue tracker? [10:27:31] Wahid Bhimji I think we all agree to drop it then [10:27:34] Brian Davies sorry lost my network fo abotu 2 mins [10:27:49] Ewan Mac Mahon RAL's tinkering with the network today IIRC [10:27:52] Christopher Walker The wiki for ops meetings is a bit of a pain. Whether learning JIRA is worth it is anohter matter. [10:28:17] Jens Jensen JIRA is like other request trackers - you forget about the tasklists afer a while [10:30:28] Wahid Bhimji we've now picked it up as one of our draft tasks in the "collaboration" .. haven't we - so soon you can be grumpy at yourself or me [10:34:22] Wahid Bhimji I think a post to dpm-users-forum complaining about it would add some weight to them fixing it [10:34:40] Elena Korolkova ok. I'll do that [10:34:43] Wahid Bhimji but also I will follow up with him in person next week when I'm at cern [10:34:54] Elena Korolkova thanks [10:37:33] Wahid Bhimji yes - ps if you are meeting in edinburgh maybe I could come along if they let me [10:38:48] Wahid Bhimji what about xrdcp or http transfering (perhaps not the former) [10:39:03] Sam Skipsey I mentioned http, but they weren't convinced. [10:39:11] Wahid Bhimji why not [10:39:13] Sam Skipsey and they were unkeen on using xrootd. [10:39:33] Sam Skipsey They didn't believe that http streaming would maximise their bandwitdh [10:39:49] Sam Skipsey ... mostly what they want to do is *stream lots of data* continuously for ever. [10:40:15] Wahid Bhimji still going on ... quite boring testing process on atlas side. I will try and put together smthg for gridpp meeting [10:40:50] Ewan Mac Mahon Streaming is awkward. [10:41:06] Ewan Mac Mahon FTS style transfers between buffering storage systems is much more resilient [10:41:24] Ewan Mac Mahon Streaming always has the issue of trying to catch up after a breakage. [10:41:36] Wahid Bhimji you can do FTS style transfers with http [10:41:42] Sam Skipsey Yes, the resilience issue has been pointed out to them. [10:41:53] Wahid Bhimji FTS3 specifically does it - even 3rd party [10:42:55] Ewan Mac Mahon I do think there's something of a meta-argument we can make about all of our infrastructure along the lines of "It may look complicated, but it's all there to solve real problems, and it actually does - use it, or be condemned to reinvent it badly at great cost" [10:43:17] Sam Skipsey Essentially, they want to end up with a complete backup copy of tens of TB of data, continuously updated (with an acceptable lag, obviously) [10:44:41] Sam Skipsey (I am slightly unclear on the rate of production, but it might be on the order of 100s of TB + a year, maybe up to 1PB) [10:45:31] Wahid Bhimji bye