Attending: Jens, Gareth, John H, John B, Matt D, Wahid, Elena, Duncan, Brian, Ewan, Steve, Jeremy, Sam, Raul (and, briefly, Janusz) 0. Blogs and operations? DPM 1.8.9? DMLite libs 0.7.2 - should be available but Matt wasn't able to yum them (see chat) - maybe a mirror is stale. 1. Update from NA62 (either Dan or Janusz) - hopefully?! Janusz apologised that his train was delayed and he joined the meeting too late, so we will talk to NA62 next week. [Update: Dan sent a note after the meeting that he is on OSX which refuses to work with VidyoWeb, so we might try the phone bridge next week.] In general, Jens had tried to prod some of the non-LHC VOs - we generally hear fairly frequently from T2K and CERN@School and MICE, but less frequently from some of the others (or only indirectly): and attempts to prod some of them - Sam has tried, too - have led to nothing. Should we expect some level of responsiveness from the VOs we support? Small VOs are by definition small and have few people to respond to our queries (particularly non-urgent ones, perhaps), but we might want to talk to them not just to hear about their problems but also their success stories which after all are successes also for GridPP. Conversely, they may be interested in hearing from us about explorations and future directions. Another specific question is whether there is enough use of DIRAC to justify our putting more effort into understanding it, beyond "tried a tutorial and got the tee shirt." non-LHC VOs seem to be increasingly using, or at least interested in, DIRAC, so perhaps we should support data management in DIRAC - worth discussing with Janusz next week, too. The DFC may be a good alternative to the rapidly disappearing LFC, so now may be a good time. DIRAC's data management should be reasonably pluggable - eg LHCb have a file consistency layer - but as more people start using DIRAC it may need to meet more diverse requirements. Chris is working on the question of whether non-LHC VOs can migrate to DFC (but then someone may need to pick this up when Chris moves to greener pastures.) Jens used the T1 (internal) contact list for VOs supported by the T1 (most of whom also use T2s), but this may be out of date - Jeremy suggested checking the VOMS data. 2. ATLAS update, specifically questions for the Jamboree in December, and WebDAV, and RUCIO (Brian+other ATLAS) There is an ATLAS Jamboree next week, and Brian and Wahid are both going; from T1 also Gareth and - sans doute - Alastair. The agenda (see chat log) is secret, but will contain stuff relevant to us, about consistency and tools, RUCIO of course, "federation", storage protocols, FAX, and storage on S3. We expect there will also be a remote attendance opportunity. Also, as Chris highlighted in an email, WebDAV is beginning to be more available at sites, and ATLAS could be looking into extending their use of WebDAV. There are some general questions about WebDAV, such as whether data is "world readable" for "everyone" (meaning everyone who is able to authenticate with a suitable certificate.) 3. File deletion rates The old target of 1Hz is no longer sufficient, and we now need more like 10Hz. DPM, Sam reports, will accept the SRM-bulk and then work its way through the list sequentially, rather than doing them internally in parallel - which might lead the system to be unresponsive or lock transactions, but then sequential processing would take longer. Generally DPM does the Right Thing, or at least the Best Thing Under the Circumstances. The problem being that ATLAS is close to "full" at the T1; if they deleted 1% of their files they'd still need 450K files deleted and 450K seconds is on the order of four or five days. ALTAS could try DAV for deletion: Wahid says it worked well for TW but not for NL for whom SRM worked better. Proposed to (re)test deletion rates at Glasgow and T1. 4. AOB What would "rebalancing" mean to a sane person? Sane people think it means moving stuff around to spread it across filesystems. This big data day at Wellcome (Jens circulated agenda to the list): Jens volunteered Brian to go; Dave Colling expected to go. Is it a technical day or a talk day? Whoever gets nominated to speak for GridPP should have some useful slides from us, 2-3 slides with real information about our data transfer rates and processing rates. It would be useful to have a set of slides anyway, approved by the group, which we can all use to represent the group. Chat log Matt Doidge: (26/11/2014 10:02:35) I tried to upgrade to 1.8.9 yesterday afternoon, yum said no. John Hill: (10:04 AM) It's still refusing on SL6. I have the EMI-3 metapackage - is that blocking John Bland: (10:05 AM) I'm only just back from holiday. Are we supposed to bve upgrading to 1.8.9 now? wahid: (10:05 AM) what is the message John ? Matt Doidge: (10:05 AM) I'm having trouble too - yum can't see the 0.7-2 dmlite-libs John Hill: (10:05 AM) I see the same as Matt Duncan Rand: (10:05 AM) Janusz is not at his desk yet Matt Doidge: (10:05 AM) Only 0.6 in the repo - I'll see if they're fubar. wahid: (10:05 AM) can you try the main epel repo instead of a mirror John Bland - we don't have to upgrade but if you did then you need to reupgrade as there was a (overlogging) bug with the first version of 1.8.9 released John Bland: (10:06 AM) and are people using puppet or still relying on yaim and crossing their fingers? wahid: (10:07 AM) puppet for me (but I've only done disk servers ) Duncan Rand: (10:07 AM) Somebody has a bad mic Matt Doidge: (10:07 AM) Option number 3 - by hand! wahid: (10:07 AM) https://indico.cern.ch/event/276502/ John Bland: (10:07 AM) matt: we're not all as brave as you ;0) Duncan Rand: (10:08 AM) Please could they mute Matt Doidge: (10:08 AM) I'm not sure brave is the right word... wahid: (10:08 AM) its not me (the bad mike) Duncan Rand: (10:09 AM) I suspect Brian, he always has poor sound. wahid: (10:09 AM) Yaim will work - but actually not for this logging problem - for that you will have to manually add LogLevel 1 in /etc/dmlite.conf if you use Yaim but with puppet its done Matt; John - try core epel - I believe it should be there.. Matt Doidge: (10:11 AM) Not sure what was up for yum - I think the repo was misconfigured for the mirrors as Wahid suggested - manually installing the new dmlite-lib rpm removed the update blockage. John Hill: (10:12 AM) I'll have a look after the meeting - I'm not in a rush to upgrade Duncan Rand: (10:15 AM) http://sblunier.web.cern.ch/sblunier/lastTests/node1.html wahid: (10:16 AM) yes I know that - but its not production - I see its got better - but also worse UK seems missing now.. like a little scratch mouse Duncan Rand: (10:18 AM) Let me see if I can fix my mic Ewan Mac Mahon: (10:24 AM) Pilot system dirac. Jens Jensen: (10:24 AM) http://diracgrid.org/ wahid: (10:25 AM) OK - I thought we were talking about the other dirac - sigh happens every time anyway Jeremy Coles: (10:26 AM) DiRAC, DIRAC or Dirac? Jens Jensen: (10:26 AM) Not to be confused with http://www.stfc.ac.uk/1263.aspx Ewan Mac Mahon: (10:27 AM) Or is Duncan using the Imperial meeting rom conference phone? Jeremy Coles: (10:28 AM) Barely audible.... Duncan Rand: (10:28 AM) Sorry! Ewan Mac Mahon: (10:30 AM) Pretty much anything is a better bet than LFC and WMS now. They're doomed. Dooooooomed. Matt Doidge: (10:40 AM) I agree with Ewan. wahid: (10:41 AM) we can discuss in DPM dev mtg on Fri sam Matt Doidge: (10:41 AM) In a "modern" DPM this might include network bandwidth considerations as well as space. Ewan Mac Mahon: (10:44 AM) It seems a good opportunity to push our stuff as a good solution. That needs a degree of technical depth on the part of the speaker, but not screenshots of commandlines on the slides. wahid: (10:48 AM) It s on Monday 8th December btw based on Dave C's previous email Jeremy Coles: (10:50 AM) The 8th is correct. Earlier I was checking this was the same meeting I knew about! Ewan Mac Mahon: (10:50 AM) Other people could use the tools. Jens Jensen: (10:50 AM) You can access parts of files with HTTP... Ewan Mac Mahon: (10:50 AM) Plus webdav should be able to do the same thing in princple. ^ What he said :-) Jens Jensen: (10:51 AM) :-) Ewan Mac Mahon: (10:51 AM) It should be making arguments like "Use our stuff, it's the awesomest" Also: "Get yourself sodding X509 certs already" "They're not that scary" Biologists?