Present: Alessandra, David, Duncan, Jens, John B, Matt D, Matt RB, Steve, Tom, Wahid, Sam, Gareth, Ewan, Chris W, Jeremy, Brian, Govind 0. Blogerationial issuesand ops posts Disk server at Glasgow getting hot again; rfio metadata seems to be the problem: fixed now, but keeping an eye on it. 1. Updates (if any) on: - webfts - nobody tried the CERN one; still waiting for Andrew Lahiff to set up the RAL one. - "small" VO documentation.. - and VO policies (``thou shalt use space tokens?'') Probably more useful for Jens to nag people: several people ought to (be able to) contribute. Jeremy reports a lot of pages in the wiki are very old (see chat) and it may be worth just archiving them: OTOH, some should be revised. How do users even find us? Would they type "using storage on the grid" into their favourite search engine, or would they actually remember "ah, those gridpp guys and gals, they got it." There should be a common starting point but all VOs are different - most recently HyperK is different from everyone else as well, as they already have their established practices. VOs will expand from what they already have. How do the service levels work? Probably low level services could be granted at individual sites for the locals, and by GridPP T2s by some magic osmosis process, perhaps, for low level usage, but would then need signing off by the user board once they grow beyond a certain size (and payments may have to be considered, not just for sysadmin day.) Allocations are often considering CPU, and there are CPU heavy use cases which require very little data stored and in/out. For those that need data there is some merit in keeping it simple by just supporting them at a few sites, even if it goes against the original intention of the grid that data could be placed "on the grid" and you wouldn't have to worry about where it actually was. Similarly, a VO would need some sort of catalogueing and replica management which they most likely couldn't implement themselves so would need from the grid. Big VOs tend to "solve" their own problems. It is a problem that sites are not rewarded for awarding storage to non-LHC VOs. Sites have no incentive to do this other than the goodness of their hearts and the milk of human kindness. Steve Lloyd's monitoring should also be considered: we should create an LDAP query which would, er, query the resource usage - should not be hard. 2. Feedback session - I meant to do a survey to get some feedback on these weekly meetings; I might faf around a bit more with options, or we can do it directly in the meeting. A survey was sent to the list. 3. AOB Jens: (23/07/2014 10:00:27) sysadminday.com wahid: (10:05 AM) Brian ? Ewan Mac Mahon: (10:09 AM) I think they'd tweet @gridpp, wouldn't they? But practically, yes, Chris is right - no-one's got any chance of finding a way into this without some dedicated handholding effort anyway. I think we're a bit more Google+ aren't we? Jens: (10:13 AM) https://twitter.com/gridpp Samuel Cadellin Skipsey: (10:13 AM) I like to think that we're a G+ kinda consortium, yes. Ewan Mac Mahon: (10:17 AM) On a detail note, the gridpp contact page doesn't mention the twitter. It does mention the closed PMB-only mailing list, and gives a postal address (!) for Tom. Samuel Cadellin Skipsey: (10:17 AM) Defacto, they tend to be covered by GridPP policies, since we make people use the grid interfaces. Ewan Mac Mahon: (10:17 AM) It could probably do with a warm-over. Samuel Cadellin Skipsey: (10:17 AM) There is a *lot* of out of date info on the GridPP website, I've noticed. (I think the list of roles for the PMB might still be out of date) Ewan Mac Mahon: (10:18 AM) And there's this: http://www.gridpp.ac.uk/deployment/contact.html I don't even recognise all of those names. And I've been here a while now. Samuel Cadellin Skipsey: (10:20 AM) So, I think what we're saying is that the website is even more out of date than the wiki. Jeremy Coles: (10:21 AM) Ewan we've found a (large) number of pages that are out of date. We are deciding what to do with them... many will be archived. Jens: (10:23 AM) Maybe a review: would this still be useful if it were up to date Jeremy Coles: (10:24 AM) Agreed. Many pages came out of specific suggestions over the years and the drivers for them have since gone. We found 1725 .html files on the server and only 200 or so were updated or created in the last 2 years. Jens: (10:29 AM) We have also done syncats from time to time, with some VOs Ewan Mac Mahon: (10:29 AM) We'd actually improve the situation by replacing the front page with the words "DON'T PANIC' in large friendly letters, and a link to the wiki. Though arguably the instruction to not panic is itself just as misleading as a lot of the existing stuff. We need a friendly cartoon Tom and a link to tweet at him. Alessandra Forti: (10:38 AM) Gridpp and wlcg of course don't care but for me it is useful to sell to the university that we support also smaller VOs. It blocks any argument on the electricty and space bills Also for example atlas had a massive slump of jobs between April and May and biomed and ilc filled the gaps big time. Ewan Mac Mahon: (10:40 AM) Just to be clear, I'm not agasint supporting 'others', I'm pro the managemnet putting their money where their mouths are. Alessandra Forti: (10:41 AM) small VOs get all this for free wahid: (10:41 AM) Sorry gotto go - and on holiday next week ! Alessandra Forti: (10:41 AM) storage is the only thing I'm reluctant to share it's inflexible and is not spare storage in the sense of spare cpus is wahid: (10:42 AM) good luck solving all the small VOs problems - I'll stick with the big ones ;-) Jens: (10:42 AM) Have a nice holiday Wahid - don't install too many DPMs Alessandra Forti: (10:42 AM) I'm just going to carry on as now bye Jens: (10:43 AM) nudge effect Ewan Mac Mahon: (10:46 AM) I think the principle is that if you want me to put 10% into smll VOs, then I need to be paid for that 10% at the same rate as I am for ATLAS. If I run more minor VO jobs than that, drop the funding rate. Samuel Cadellin Skipsey: (10:46 AM) I was just about to suggest precisely what Ewan suggested. Ewan Mac Mahon: (10:47 AM) That's fine, but a site that's struggling for money (hi there) must not be put in the position of having to choose between a lucrative job/storage allocation and a non/less lucrative one. M M M