21:00:35 #startmeeting scientific-sig 21:00:36 Meeting started Tue Jun 23 21:00:35 2020 UTC and is due to finish in 60 minutes. The chair is oneswig. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:37 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:39 The meeting name has been set to 'scientific_sig' 21:01:02 Greetings 21:01:24 No set agenda for today 21:02:58 lurking 21:03:16 Hey b1airo, how's things 21:03:19 #chair b1airo 21:03:19 Current chairs: b1airo oneswig 21:04:20 a bit fuzzy this morning... up a few times with nosebleed dramas - Eva is growing too fast at the moment apparently :-) 21:05:13 In her sleep? Ouch. 21:05:54 I was up late but putting together some work stuff. Not much to show for it. 21:06:05 the joys 21:06:48 Having fun now though - adding new telemetry sources to a deployment. Watching Elastic and co eat all the RAM in the world. 21:07:19 i had a colleague from Melbourne get in touch recently regarding GPU P2P in an OpenStack KVM cloud 21:08:03 With questions or answers? 21:08:22 mainly questions 21:08:45 he appears to have a 4-way T4 box 21:09:31 he's working with Aptira. i pointed him in the direction of the experimental GPU clique stuff in KVM and he was going to see if they could try it out 21:09:57 though who knows if that even works - Google doesn't provide much hope 21:10:10 There have to be some developments in this area by now... 21:10:43 and looks like NVIDIA wants you to pay again for the privilege now and do it via GRID/vGPU (which recently got support for P2P by the looks of things) 21:10:56 It's pretty grunty stuff to get working though. All those layers of indirection in PCI bus addresses. 21:11:19 Interesting. Got a link to it? 21:11:54 https://docs.nvidia.com/grid/latest/grid-vgpu-user-guide/index.html 21:12:14 https://www.irccloud.com/pastebin/tGKYpyzo/ 21:13:01 might be some devil in the "ifs" though 21:14:10 NVlink it says - I was hoping for PCI for generality. 21:14:49 yeah, there's that too 21:14:57 sorry to be late 21:14:59 Progress of sorts. 21:15:03 Hi martial 21:15:06 #chair martial 21:15:07 Current chairs: b1airo martial oneswig 21:15:55 What's new with you martial? 21:16:38 not much 21:16:52 been fighting Tensorflow compilation issues lately :) 21:17:43 ie fun times 21:18:05 How is the meeting? 21:18:35 I don't think an agenda was posted so it's pretty quiet :-) 21:19:01 ah well :) 21:19:06 martial: I saw Mike has been active on Kolla-Ansible Ceph this last week 21:19:53 indeed, we are trying to design our core installation scripts 21:20:14 Somewhat interesting Next Platform discussion on Ceph usage at Sandia - https://www.youtube.com/watch?v=BEfZ3FyN0IQ&feature=youtu.be&t=1118 21:21:28 martial: watch out, I believe the long-term plan for Kolla-Ansible is to use Ceph-Ansible or CephAdm rather than reinvent that work. 21:22:05 will pass along the information 21:22:40 eg the warning here: https://docs.openstack.org/kolla-ansible/train/reference/storage/ceph-guide.html 21:24:05 urgh - liberal use of red warning boxes on that page 21:24:27 still, external looks eassy enough 21:24:35 Thank you 21:24:54 shared it with him (now just a matter for him to see it :) ) 21:24:59 i'm afk for 5 to boost caffeine levels, gonna be a long day 21:25:14 Caffeine power :) 21:25:53 how was HPCW, did you participate? 21:26:38 I listened in to some of the conversations 21:27:03 but I did not present this year 21:27:08 it was good and busy 21:27:18 the videos are on you tube already 21:27:34 and there were quiet a few conversations on slack 21:27:44 let me find the link 21:27:55 I wonder - what counts for busy with online workshops? 21:28:18 #link https://www.youtube.com/playlist?list=PLfE3_wJGw9KTx8t6byn5bCWpk4L5bRORY 21:28:25 I'm wondering how ISC will go next week 21:29:28 Looks like Christian did a lot of legwork to get all those videos done 21:29:42 I know Christian posted a number of views ... looking for it 21:32:46 I can not find it (lost in all the slack messages that day) but I remember something in the 70 concurrent viewers 21:33:22 That's good going I'd say. 21:34:15 it was very interesting 21:34:22 (wish I could have been there the entire time) 21:35:11 Good conversation on Saurus, Buildah and other topics (as last year actually :) ) 21:36:23 Not heard of Saurus 21:36:58 https://sarus.readthedocs.io/en/stable/ 21:37:38 lmgtfy :-) 21:37:53 it's CSCS's container engine 21:38:08 still haven't worked out why they need their own though 21:39:41 I remember one of the authors was doing a lot with Shifter a couple of years ago. 21:40:00 they had a good slide deck that explained the use case (looking for it) 21:41:19 it was not posted in the channel it looks like but there is a similar one in the HPC container google drive 21:41:21 let me check 21:43:20 found a PDF 21:44:04 martial: What's going on with Singularity? Does it still have good uptake? 21:45:12 Christian was running a few live polls, 21:46:26 14 docker / 7 shifter / 10 singularity / 3 sarus / 5 charliecloud / 10 podman / 3 KataContainers / 6 LXC 21:46:42 so yes still very much used 21:46:49 keep in mind this was an HPC crowd 21:47:57 Image building was high on Dockerfile with Docker (10) with distant (3) singularity 21:49:10 Christian in particular had a good idea to add to the "machine" type conversation, see https://github.com/opencontainers/image-spec/pull/761 21:49:23 What a fascinating split between so many contenders. 21:50:05 podman keeps getting higher in popularity 21:50:59 easy to understand too .. their tagline is alias docker=podman 21:51:24 i guess as deployments get it "for free" with RHEL/CENTOS 8 it will continue to get greater uptake 21:53:40 Sarus was "push to docker" "deploy to HPC" without any job 21:54:01 the link I have was too permissive so I found another PDF 21:54:04 #link http://www.hpcadvisorycouncil.com/events/2019/swiss-workshop/pdf/030419/K_Mariotti_CSCS_SARUS_OCI_ContainerRuntime_04032019.pdf 21:55:26 no news on Berlin? 21:57:21 Not that I've heard. At the end of this month there is an OpenDev online conference on infrastructure scaling. Should be interesting 21:57:34 got a link? 21:58:02 #link https://www.openstack.org/events/opendev-2020/opendev-schedule 21:58:29 thanks - it's a lousy name because in browser history it hits on so many other things 21:58:51 so that is the one? 21:59:13 yes - I'm going to participate in the Tools session. 21:59:41 cool! 22:00:03 We are nearly out of time... 22:00:10 final comments? 22:00:21 thanks for the update on HPCW martial 22:00:33 happy to do so 22:00:41 #endmeeting