21:03:50 #startmeeting scientific-sig 21:03:51 Meeting started Tue Mar 20 21:03:50 2018 UTC and is due to finish in 60 minutes. The chair is b1airo. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:03:52 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:03:54 The meeting name has been set to 'scientific_sig' 21:04:05 morning! 21:04:21 Hi b1airo - apologies, I am on a train, in a tunnel 21:04:34 #chair oneswig 21:04:34 Current chairs: b1airo oneswig 21:04:36 highly likely to be intermittently present :-) 21:04:48 same as stig 21:05:08 hey martial 21:05:12 oh come now oneswig, in Australia the wireless broadband service is generally better than fixed line! 21:05:19 Hi Stig, Blair 21:05:23 #chair martial 21:05:24 Current chairs: b1airo martial oneswig 21:05:34 I am not in a train in Australia... 21:05:40 ... alas 21:05:47 :-) 21:05:53 we are in the "Network Challenges" session of the conference with a presentation from Internet 2 21:06:19 ha 21:06:30 I've heard of such challenges 21:06:31 martial: did you get anything back from Wilfred/Paul ? 21:07:49 nothing unfortunately 21:07:55 or Enol for that matter 21:07:56 i've been up until past 3am this morning dealing with storage "challenges", so if no-one else shows up to this meeting in the next 5 mins i'll be happy to cut it short 21:08:19 *grumble* 21:08:29 but I got the slide deck from Zhipeng about Blockchain use in the Passport program 21:08:32 I'll be dropping off around then due to continuing my journey on bicyle 21:08:38 What's the news on storage? 21:09:09 b1airo if interested check goo.gl/nPXeMR 21:09:37 just the delicacy of a Ceph pool with a couple hundred OSDs over 95% full 21:09:57 oh Mike would have tons of things to say about that, he had that glitch 21:10:06 There's a chewy problem 21:11:02 we've added some more OSDs for the moment, but the iffy CRUSH distribution in our present setup means there's quite a period of yoyo-ing before things start to trend down for the most full OSDs 21:11:20 i saw one at 96.82% 21:11:44 somewhat too close to the 97% failsafe full ratio 21:12:49 we tried to copy a 100TB RBD into a new pool on different hardware, but a few OSDs in the source pool didn't seem to like the idea and stopped processing other work for some reason 21:13:41 alright, i'll run through the agenda for posterity 21:13:52 Someone I saw today was reporting a 30 minute commit latency for IO to their glusterfs. yikes 21:13:53 BTW b1airo: I thought your roundup on GPU passthrough was really helpful. Thanks for taking the trouble. 21:14:51 cool, i you thought so then it was probably ok :-) 21:15:40 i was reminded about NVIDIA's HGX-1 architecture afterwards too, but that's not generally available outside the big customers afaik 21:16:09 apparently they will have some big feature announcements around a 2nd gen of that platform soon 21:16:15 don't know it, sorry - not an nvidia fanboi 21:16:29 ok, train's in, gotta go 21:16:35 until next time 21:17:03 we are having a private meeting today indeed :) 21:19:39 b1airo: end of meeting? 21:20:31 sry, distracted by a small person 21:21:02 not too much happening today, so I would say let's skip? 21:21:05 #topic Forum Brainstorming 21:21:29 martial: yeah, i'll just write these things up for the logs but we can revisit next week 21:21:39 cool 21:21:56 The Forum process has started for Vancouver'18 21:22:26 we're are the brainstorming stage where the community starts throwing out ideas for topics and sessions 21:22:35 what is the link for the etherpad? 21:22:43 i've started an etherpad here: 21:22:50 #link https://etherpad.openstack.org/p/YVR18-scientific-sig-brainstorming 21:23:18 which is also linked from the Forum wiki page, #link: https://wiki.openstack.org/wiki/Forum/Vancouver2018#Etherpads_from_Teams.2C_Working_Groups.2C_and_SIGs 21:24:04 so if you have ideas for topics of interest to the Scientific OpenStack ops/user community then please throw them in 21:24:39 i've taken the liberty of kicking things off with a state of the art/practice on GPUs with OpenStack 21:24:51 #topic Discussion regarding state of GPU passthrough and interest in collaborating to resolve security issues 21:24:53 I see that 21:25:21 is that a follow up from the conversation you talked to us in Sydney? 21:25:54 partly, yes 21:26:16 not really much more to say about this right now beyond what I summarised on-list yesterday, happy to talk more if people have specific questions 21:27:21 i also know some scant technical details (for specific devices) about a couple of pci config space windows that should be closed to guests, e.g., VBIOS flash 21:27:53 so that could be a start to adding and testing pci-quirks in QEMU 21:28:38 i'd really like to find a register or something that does "a bad thing" (tm) that could be used as a demonstrable exploit - that would probably garner some more attention 21:28:53 and would be a useful test-case for quirks 21:29:09 #topic AOB 21:29:18 Anything else to mention? 21:30:04 not much from my side, the workshop is going very well 21:30:06 very busy 21:30:34 great 21:30:51 but we will have the pleasure of doing the second day partially on gotomeeting due to a fair amount of snow 21:31:48 the follow up from this meeting is expected to be at the Vancouver Summit 21:32:19 look forward to hearing the round-up from it then 21:32:34 yes Sir, will provide some information 21:32:38 we might have to defer that till next fortnight if we have Zhipeng next week 21:32:55 seems to be the case for next time indeed 21:33:09 good news still 21:35:10 ok all done then 21:35:21 catch you next week! 21:35:24 #endmeeting