The meeting is set, the Honourable Samuel Skipsey presiding. Meeting was run as a round table of issues initially. First topic that came up was the ATLAS spacetoken manipulation requests. HOTDISK not a problem for most people. The deletion of non-Rucio paths in PRODDISK lead to discussion, however, as there's actually quite a lot of stuff (mostly many, deeply recursed, but empty, directories, but also a not-tiny number of files) outside the Rucio paths at at least Oxford, Liverpool, Glasgow. Elena was going to look into this, but also we decided it might be safest to take syncat dumps, and move the problem to Brian to delete our excess data ;) Oxford/Ewan also looking at T2K spacetoken provision. Elena still waiting on DPM 1.8.7 and needing disk load mitigation advice. I promised a blog post on it for her (should make you happy too, Jens). We moved on to discuss look ahead for next procurement phase. It was noted that a lot of sites have at least some significantly old (out of warranty) disk, and that replacing this (or at least procurement to secure against lost capacity from its failure) will be a noticable fraction of spend on storage this time around. This may confound estimates of price/capacity made by VOs, as less "effective" spend will be made on extending capacity at sites. It was agreed to mention this at GridPP31. - Chat Log follows: 10:03:50] John Hill joined [10:03:51] Steve Jones Is anyone talking? [10:03:52] David Crooks joined [10:03:53] Steve Jones joined [10:03:54] Gareth Roy joined [10:03:55] Robert Frank joined [10:03:58] John Bland joined [10:04:55] John Bland I can hear you [10:05:05] Ewan Mac Mahon joined [10:08:37] Elena Korolkova joined [10:08:59] John Bland we've similarly just deleted hotdisk and are also pondering the contents of proddisk. The syncat+Brian process seems a good idea. [10:10:22] John Bland we're also doing some tests on some new non-grid storage, with SSD caching. No conclusions so far other than they don't seem to work in way I thought they would (at least with iozone tests) [10:10:26] John Bland no mic today, sorry [10:10:56] John Hill I think my mike isn't working either [10:11:19] John Hill I've got rid of HOTDISK [10:12:39] John Hill On PRODDISK I've found huge numbers of directories, but very few files - can these be removed by Brian? [10:12:50] John Bland we did have an Areca card fail the other week, luckily we had a spare, but we're getting a little concerned about some of our older kit. We hope the next funding round will allow us to replace some of it! [10:14:05] Ewan Mac Mahon I think that's going to be a common hope for the next funding round. I keep poking Pete to that effect, but I'm not sure whether or not the message has quite percollated up to the PMB and back to the VOs that some of the money isn't going to give them capacity increases. [10:17:14] Ewan Mac Mahon We should get a gridpp storage twitter account. That'd get the blog post count up. [10:20:32] Ewan Mac Mahon And I'm not just planning on telling Brian; I'm planning on him doing the deletions. [10:26:09] Ewan Mac Mahon We're going to be exceeding our MOU levels by a lot less, to say the least. [10:31:58] John Bland Even with new servers accidents happen, things fail. ATLAS can deal with lost data, and maybe federated storage will help. [10:32:40] Steve Jones Cost, Quality, Perfromance - pick any two - a classic trilema. [10:34:08] Elena Korolkova left [10:34:09] John Hill left [10:34:11] Gareth Roy left [10:34:14] John Bland t'ra [10:34:17] Robert Frank left [10:34:19] John Bland left [10:34:20] Ewan Mac Mahon Bye [10:34:22] David Crooks left [10:34:25] Ewan Mac Mahon left