Present: Elena, Jens (chair+mins), John H, Matt D, Tom, John B, Ewan, Steve, Govind, Winnie Special gues star: Andrew Sansum Apologies (implicitly) : everyone who is at CHEP/WLCG Summary of actions: * Ewan to prod Pete about including non-LHC-VO community support in GridPP34 agenda * Jens to kickstart the DiRAC backup stuff * Andrew to mail Lydia and Jens for Jens to kickstart the DiRAC backup stuff * Jens to follow up with GRIDPP-USERS-OWNERS about (ab)use of the list for non-LHC-VO community support. 0. Operational blog posts Reminder for everyone who is not at STFC to blog (STFC is subject to Purdah and can only blog after the election unless it is very strictly non-political, but communications did say "best not to blog") 1. DiRAC (not to be confused with DIRAC - see chat log) DiRAC have 5PB of "output" data (between custodial and replica), meaning it can be regenerated if needed but it would be time consuming. Could GridPP back it up? From the GridPP perspective it would be a good thing since it broadens further the use of GridPP, and collaboration between DiRAC and GridPP in general has been desirable in a while. DiRAC have no money for the backups this FY but should bid for the coming years. Luckily the T1 does have capacity for the extra 5PB this year, and it was included in the 100% scenario in the GridPP5 proposal. Durham has 2PB, EPCC 1200TB, Leicester and Cambridge 800TB each. Intend to start with Durham, Lydia Heck is the contact. Expected to identify measures in the next few weeks - should be "owned" by the storage group. Discussions also included JANET particularly if using OPNs but that is probably not worth it. Files would be O(1GB). The general suggestion (Ewan's) was to take this forward on the storage list as it is low volume and the extra Durham folks would probably not mind the extra traffic (~a few mails per week). On the list, we tend to be pretty good at chipping in to get a solution. In practical terms, our proposed way forward is to have a GridFTP endpoint at Durham (or use their SRM??) and then FTS files over to RAL using the GEN CASTOR instance. There, that's (almost) a plan. 2. Is it time to revisit (again) our data support for the "small" (ie non LHC) VOs? We've been in contact with LIGO and they're happy but there is the wider question about metadata and data discovery and data models for non-LHC VOs which generally have less regimented data models, and there is less distance between the data generation and the end user analysis. Ewan points out the DIRAC file catalogue may do what's needed, although we'd need some practical experience - it's not just tracking replicas, it's also making the data discoverable using the domain-specific metadata (in this case gravitational waves). So LIGO is a useful case study and it may be timely as well since there is talk of having gravitational waves "communities" in EU-T0. Ewan asks the question how we more generally support the users in the "small" VOs - for example, Paul came back to us with questions about workflows which are obviously outside the remit of the storage group but could nevertheless be solved by some of the same people who solve questions for the storage group... Having a separate list for small VOs' users may also help them feel part of a community instead of just being people we remember to talk to occasionally. We should figure out whether to set up a new list or to repurpose GRIDPP-USERS which currently only serves the lowish traffic of Tier 1 incidents... 3. AOB The Scientific Computing Department at RAL (of which the Tier 1 and the storage people are members) is having an away day next Wednesday and it is unlikely that we will be able to connect. Ideally Sam should be running next week's storage meeting and report from WLCG/CHEP. Jens Jensen: (15/04/2015 10:05:48) This is DiRAC not DIRAC :-) http://www.stfc.ac.uk/1263.aspx Also, working with DiRAC in general - not just for the back ups - would be a good thing. Ewan Mac Mahon: (10:10 AM) Indeed; we've tried and failed to get decent contacts going before and it's always fallen through. If this can get past the 'Jeremy Yates is enthusiastic' phase into something actually existing that might give us something substantial to build on. Arrgh, noooooo. No private links :-( (By absolute best case scenario here is that they set up a gridftp/something else that FTS can talk to on their end and just FTS the lot) Jens Jensen: (10:12 AM) Yes, we should use the normal production network. Ewan Mac Mahon: (10:12 AM) Er. s/By/My/ Jens Jensen: (10:12 AM) That was also the result at networkshop Ewan Mac Mahon: (10:15 AM) I can see the need for some practical lessons in the application of Ewan's first rule of firewalling grid services though. Jens Jensen: (10:15 AM) Lydia Heck:? https://www.dur.ac.uk/physics/staff/profiles/?username=dph0elh Ewan Mac Mahon: (10:16 AM) Also, I'm not sure I feel terribly comfortable about OUR credibility being on the line; we know we can do this literally in our sleep. We don't know that DiRAC can. Well, let's get her signed up to the storage list and conduct the discussions there. It's the best way to loop in all the gridpp expertise. John Hill: (10:22 AM) On the list is fine by me Jens Jensen: (10:24 AM) Thank, John. We have two people from Durham on the mailing list already Oliver Smith and Phil Roffe Makes sense to keep them in the loop Ewan Mac Mahon: (10:32 AM) It might be worth seeing whether there's any scope for getting Lydia to come to the GridPP meeting. It's way past the registration deadline, but I imagine we coul dmake it work if we assume she can sort her own accomodation. It'd be nice to be able to invite her, even if she can't come (though it would also be good if she actually could) Andrew Sansum: (10:34 AM) I'll check with QMW and if OK ask her if she would like to come. Jens Jensen: (10:34 AM) https://dirac.grid.hep.ph.ic.ac.uk:8443/DIRAC/ https://www.gridpp.ac.uk/wiki/Dirac https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=GRIDPP-USERS I can't see who is subscribed: I am not an owner of that one. Ewan Mac Mahon: (10:48 AM) Can we see who is the owner? Jens Jensen: (10:49 AM) There is a contact owner(s) button Ewan Mac Mahon: (10:51 AM) Send it something, I'm half expecting that it'll wind up in your inbox via a horribly convoluted path. Jens Jensen: (10:51 AM) :-)