21:04:19 <martial__> #startmeeting Scientific-SIG 21:04:20 <openstack> Meeting started Tue Oct 30 21:04:19 2018 UTC and is due to finish in 60 minutes. The chair is martial__. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:04:21 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:04:23 <openstack> The meeting name has been set to 'scientific_sig' 21:04:30 <oneswig> away we go :-) 21:04:31 <martial__> #chair oneswig 21:04:32 <openstack> Current chairs: martial__ oneswig 21:04:51 <janders> trandles: Blair would have a lot of those now :) 21:05:03 <martial__> Agenda at 21:05:05 <martial__> #link https://wiki.openstack.org/wiki/Scientific_SIG#IRC_Meeting_October_30th_2018 21:05:27 <martial__> #topic Berlin SIG BoF - Call for lightning talks 21:06:19 <martial__> Come-thee, come-thee, the Scientific SIG-ness has an opening for its participants to discuss the latest and greatest and the barely working during its Lighting Talk session 21:06:54 <martial__> we welcome the courageous, the brave and the crazy on our podium 21:06:55 <janders> barely as in bare metal? :) 21:07:02 <oneswig> oh, too good. 21:07:14 <oneswig> I shall think up a topic now. 21:07:32 <oneswig> Aha, money for old rope - I'm presenting at Ceph day on the Monday before. I can squeeze that into 3 minutes, sure 21:07:59 <martial__> be thee brave friend? be thee stupendous with knowledge to share? Well thee are perfect for us ... come and bedazzle 21:08:43 <oneswig> martial__: been at the communion wine again? :-) 21:08:43 <martial__> so ... add thy words to our pad of the ether and amaze the crowd 21:09:06 <martial__> (selling it :) ) 21:09:11 <oneswig> Sold! 21:09:22 <trandles> martial__: when are you auditioning for a job at Medieval Times? Just back from a stay at Excalibur in Vegas? 21:09:33 <martial__> :) 21:10:05 <martial__> so as the words above totally say in more words, come, be happy, present and ... profit 21:10:18 <trandles> Someone get in touch with Arkady. Best talk prize this go should be a sword-in-the-stone model. ;) 21:10:40 <martial__> who in this crowd is going beyond Sir OneSwig 21:10:53 <oneswig> I am lost for words 21:11:53 <martial__> well as long as you host the Joust :) 21:12:03 <janders> I'm sure that we can quickly write a sword-extracting ansible module 21:12:11 <martial__> :) 21:12:12 <oneswig> martial__: I hope you'll keep this up for your video message to address the SIG meeting 21:12:35 <martial__> 4am my time? I will be just in the right state of sleep :) 21:13:11 <janders> time is relative and only a function of space 21:13:16 <oneswig> Are preparations going well for SC? 21:13:24 <martial__> but yes I am proposing to host a GTM for people who are interested to view the meetings remotely 21:13:32 <martial__> (and maybe record it?) 21:14:17 <martial__> so if you are interested in participating, add your name to 21:14:23 <martial__> #link https://wiki.openstack.org/wiki/Scientific_SIG#IRC_Meeting_October_30th_2018 21:14:32 <martial__> oops wait no 21:14:43 <martial__> #link https://etherpad.openstack.org/p/Berlin-Scientific-SIG-BoF 21:14:55 <martial__> ^ that one 21:15:14 <martial__> in truth that is the agenda I have for Today 21:15:40 <janders> in which case let me propose this agenda item: 21:15:45 <janders> IBM acquisition of Red Hat 21:15:49 <martial__> ouch 21:16:05 <oneswig> janders: anything to avoid repeating your excellent presentation!? 21:16:11 <janders> what are your thoughts on what this means to OpenStack and Ansible? 21:16:16 <martial__> tim, are you staying until the end? (have to leave in 15 minutes) 21:16:20 <oneswig> I see no harm for Ansible. 21:16:53 <janders> oneswig: I don't mind repeating if there is interest, but curious what people think 21:17:10 <oneswig> IBM's interest in OpenStack is not what it once was (true for many larger and more corpulent corporations) 21:17:49 <oneswig> But perhaps that is why they covet Red Hat, who have clearly made a stellar success from their contribution 21:18:22 <janders> what's your feel regarding RHAT being left alone as promised as opposed to IBM making attempts of tighter integration of the two organisations? 21:18:52 <oneswig> I think it looks bleak for Red Hat's business-centric offerings. 21:19:12 <oneswig> And I'm not sure I'd back GlusterFS against SpectrumScale 21:19:21 <janders> I clearly wouldn't :) 21:19:46 <janders> if Ceph+CephFS didn't kill Gluster, Spectrum might 21:19:48 <martial__> did not think of that I must say 21:19:55 <oneswig> I assume there will be ever closer union. Perhaps a bit like Hong Kong and mainland China 21:20:12 <martial__> that said I truly saw this more as "we need cloud and open source" ... 21:20:38 <martial__> wait and see I guess 21:20:46 <oneswig> Who owns Pivotal (and RabbitMQ)? Was that Dell? 21:21:06 <janders> what are your bets on Microsoft buying Canonical now? :) 21:21:21 <oneswig> I'll bet you 500 beers that doesn't happen 21:21:28 <oneswig> (very small ones) 21:21:33 <janders> :D 21:21:44 <martial__> ah yes since we are already in that ... 21:21:47 <martial__> #topic AOB 21:22:11 <oneswig> martial__: will you have a beer night at SC for the Scientific SIG? 21:22:24 <janders> oneswig: why do you think Micosoft's acquisition of Canonical is unlikely? 21:22:32 <trandles> martial__: I'm here 21:22:58 <martial__> I can try to see who is around indeed (will send an email to the ML) 21:23:18 <martial__> the SC18 counter dinner 21:23:26 <oneswig> janders: I am not sure what to expect nowadays, but still 21:23:41 <martial__> janders: what next Microsoft buying Github? 21:23:53 <trandles> ah, well, I'm only at SC18 from Monday night until early Wednesday morning (read: I'm only available on Tuesday for anything at SC18) 21:24:02 <janders> ..and Apple buying Microsoft? :) 21:24:11 <janders> ..and then Amazon will take over the main Linux distro 21:24:31 <martial__> oh boy, let's conclude being on the record :) 21:24:46 <martial__> any AOB to discuss? 21:24:55 <jmlowe> Hi everybody 21:25:02 <trandles> Denver 19 Summit 21:25:05 <oneswig> Hi jmlowe 21:25:14 <trandles> I see on the foundation page that it's April 29-May 2 21:25:18 <martial__> Hey Mike, welcome 21:25:38 <oneswig> Thanks trandles, noted 21:25:41 <trandles> FWIW - The Rockies are playing a baseball game on May 3. I know it's the night after the summit ends, but I might try organizing a group to go. 21:25:59 <jmlowe> I'd be interested 21:26:17 <martial__> Stig, we are clearly on AOB, can you close? :) 21:26:31 <oneswig> martial__: I think so 21:26:37 <oneswig> bon voyage martial__ 21:26:46 <martial__> thanks, gotta go pick up the younglings 21:26:57 <trandles> later all 21:26:59 <martial__> team, see some of you at SC18 21:27:03 <janders> see you guys 21:27:05 <trandles> jmlowe: interest recorded ;) 21:27:49 <oneswig> trandles: are you still here? 21:27:59 <trandles> I am 21:28:07 <oneswig> smashing 21:28:28 <trandles> pumpkins (is this word association? ;) ) 21:28:37 <oneswig> trandles: jmlowe: did you see janders presentation from last week on mixing virt upon bare metal? 21:28:40 <oneswig> Highly recommended. 21:28:42 <trandles> oneswig: might want to endmeeting 21:28:52 <jmlowe> I didn't got a link? 21:29:01 <trandles> I saw janders link in the minutes but didn't click yet 21:29:11 <oneswig> pumpkins - very topical. I ate an entire pumpkin pie when it became clear my family are not fans. 21:29:12 <janders> https://docs.google.com/presentation/d/1fh6ZOq3DO-4V880Bn7Vxfm5I1kH5b0xKXHBLW2-sJXs/edit#slide=id.p 21:29:31 <trandles> +1 pumpkin pies - what's not to like?? 21:30:00 <jmlowe> Daylight savings ends this week for the US, makes participating a bit easier btw 21:30:22 <trandles> janders: thx for the link 21:30:25 <trandles> reading now 21:30:30 <jmlowe> for those of us in the eastern time zone 21:30:37 <oneswig> jmlowe: good to hear it, although in 2 weeks no doubt you'll have your hands full 21:30:43 <janders> no worries :) 21:31:11 <oneswig> janders: how close are you to production workloads at scale on this system? 21:31:54 <janders> we're ready for power-POC with users on but we're delaying this to focus on procurement 21:32:03 <janders> for full prod I'd like to do more work on resiliency and HA 21:32:31 <janders> we focused on capability first and we've got that and we're confident with that 21:33:00 <trandles> janders: we have a network stress test application (I think it's MPI-based) that we use for some early acceptance testing...if I can get my hands on it, would you be willing to give it a go on (Super)Cloud? 21:33:17 <janders> now we need to work on some implementation details that most people are not interested in 21:33:34 <janders> trandles: sure! jacob.anders@csiro.au 21:33:43 <oneswig> trandles: sounds like a quid pro quo arrangement there? 21:34:12 <trandles> janders: thanks, I'll get back to you 21:35:22 <oneswig> janders: what generation of IB was it? 100G? 21:35:29 <janders> FDR/56 21:36:05 <janders> but the final system will certainly have a newer interconnect :) 21:36:38 <janders> I do think we'll maintain significant investment in FDR - it's good enough for many things and there's potential of getting a fair bit of free FDR kit from other projects 21:37:47 <janders> we likely have some FDR14/40 in the mix too but it's transparent implementation wise so not sure what's the proportion 21:37:53 <oneswig> That 200G IB is taking a while to materialise 21:38:06 <janders> I meant FDR10/40 sorry 21:38:13 <janders> that's right! 21:38:32 <janders> I bet it's got something to do with PCIe4 taking sweet forever too 21:39:01 <janders> have you guys had a chance to do any testing on HDR/200? 21:39:15 <oneswig> not here 21:39:43 <janders> same here for now 21:40:50 <trandles> we're fully in OPA space on our latest machines, so no HDR here 21:41:07 <janders> from what I hear it should be very similar to drive to EDR as far as OpenStack is concerned, but I bet there'll be some catch somewhere :) 21:41:22 <janders> trandles: do you have any OpenStack on OPA? 21:41:42 <oneswig> I've been working on OPA and Ceph again - a last push to get it working well 21:42:09 <janders> OPA doesn't have virtual function support or fabric segmentation support, right? 21:42:18 <trandles> I do not have OpenStack on OPA :( 21:42:18 <oneswig> janders: you want to speak to rbudden on OPA+OpenStack 21:42:32 <trandles> AFAIK OPA doesn't support SR-IOV, no 21:42:38 <oneswig> I believe you're right 21:42:43 <trandles> but +1 what oneswig says 21:43:10 <trandles> PSC's Bridges cluster is OPA, albeit a very early OPA 21:44:42 <janders> I don't believe I've met rbudden yet 21:45:20 <jmlowe> janders: headed to sc'18? 21:45:29 <janders> no, OS Berlin 21:45:34 <trandles> janders: https://github.com/jlothian/sysconfidence 21:45:38 <trandles> that's the benchmark 21:46:04 <oneswig> thanks trandles, might bookmark that myself 21:46:26 <janders> thanks! :) 21:47:14 <oneswig> OK do we have any more AOB? 21:47:25 <trandles> I believe it was written by Oak Ridge originally. Meant to stress the fabric on a traditional cluster. It would be good to see what folks are getting on our scientific clouds, especially if you can test baremetal and from within VMs. 21:48:01 <oneswig> trandles: what does a good result look like? 21:48:53 <trandles> oneswig: I'll poke at our local expert and get back to you. He fired off the link quickly and said "talk later." 21:49:14 <janders> I'll probably need to audit the HCAs switches and cables to make sure there's no FDR40 to get a good number 21:49:25 <oneswig> OK thanks trandles. 21:49:29 <janders> for multinode workloads on IB any inconsistencies are asking for trouble 21:50:04 <trandles> it's focused on latency, so in theory we should know lowest from the vendor 21:51:03 <oneswig> Shall we close the meeting? Any more to add? 21:51:14 <trandles> none from me, thanks everyone 21:51:40 <janders> I'm good - thanks guys! 21:51:54 <oneswig> OK, it's a wrap... 21:51:59 <oneswig> #endmeeting