15:00:32 <ttx> #startmeeting large_scale_sig 15:00:32 <opendevmeet> Meeting started Wed Apr 23 15:00:32 2025 UTC and is due to finish in 60 minutes. The chair is ttx. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:32 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:32 <opendevmeet> The meeting name has been set to 'large_scale_sig' 15:00:48 <amorin> hello! 15:00:56 <ttx> amorin: high! 15:01:02 <ttx> hi! 15:01:26 <dmsimard[m]> hello o/ 15:01:46 <ttx> Hi dmsimard good to see you! 15:02:10 <ttx> felix.huettner: around? 15:02:39 <ttx> Welcome to the monthly Large Scale SIG meeting... 15:02:46 <ttx> Our agenda lives at: 15:02:59 <ttx> #link https://etherpad.opendev.org/p/large-scale-sig-meeting 15:03:12 <ttx> Feel free to add topics there while we wait for stragglers 15:03:57 <ttx> #topic Next Large Scale SIG virtual meetup 15:04:49 <ttx> Unfortunately not much progress to report on that front. We were waiting for Nexgen to confirm (as we wanted an AI infrastructure episode) but they seem too busy buying GPUs right now 15:05:23 <ttx> I'm now looking at plan C, after our two main targets did not follow up 15:06:00 <ttx> Let me know if you have suggestions of who we could reach out to! 15:06:25 <ttx> Definitely want to do one before summer hits 15:07:29 <amorin> ack, I dont remember the name of the company, but they were doing baremetal with openstack 15:07:43 <amorin> dmsimard[m]: I think we talked about it once 15:07:48 <dmsimard[m]> CERN is always interesting to hear about but they already talk a lot about what they do :) 15:08:15 <amorin> yes, cern is a good idea, they do use GPU as per talks I heard about on kubecon 15:08:18 <ttx> yes and we had them several times already :) Would like to feature a recent/new user ideally 15:09:23 <ttx> I'll come up with suggestions and probably discuss it over email so that we make progress before next month meeting 15:09:27 <dmsimard[m]> agreed 15:09:52 <ttx> #action ttx to suggest plan C solutions in a ML thread ahead of next meeting 15:10:41 <ttx> but yes, if we can't convince a recent user to talk about their stuff, we'll reach out to older users, maybe one that's been growing their footprint a lot 15:11:30 <ttx> Anything else on that topic? 15:12:14 <ttx> #topic Large scale doc 15:12:15 <ttx> #link https://docs.openstack.org/large-scale 15:12:30 <ttx> Nothing to review on https://review.opendev.org/q/project:openstack/large-scale+status:open 15:12:49 <ttx> Does anyone have new doc changes cooking? 15:12:58 <amorin> I remember, it was openmetal! 15:13:04 <amorin> sorry for late answer :( 15:14:00 <ttx> OK, next topic... 15:14:04 <ttx> #topic Anything to highlight from the PTG ? 15:14:27 <ttx> Good question! I haven't been attending that many sessions so I can't tell 15:14:38 <ttx> Anything relevant in sessions you attended? 15:15:03 <dmsimard[m]> I added this topic, I also haven't been able to attend but did browse a couple etherpads after the fact 15:16:57 <dmsimard[m]> I didn't come prepared with a list but was hoping to know if there was noteworthy things to put on our radar to follow 15:17:22 <ttx> amorin: did you attend anything relevant? 15:18:02 <dmsimard[m]> I saw in keystone there is an attempt to rewrite it in rust for improved performance, for example 15:18:56 <dmsimard[m]> https://etherpad.opendev.org/p/apr2025-ptg-keystone line 99 15:19:08 <ttx> I guess taht's a good question -- would you consider that Keystone performance is such a bottleneck that you would just both feet into a Rust rewrite (and the inevitable regressions it would initially bring?) 15:19:14 <amorin> I did attend few stuff around GPU, autoscaling 15:19:17 <ttx> jump* 15:19:29 <amorin> I attented some very interesting talks about telemetry / monitoring of applications 15:19:41 <amorin> which is something we lack in OpenStack IMHO 15:20:09 <ttx> I would agree that telemetry is a bit lacking right now 15:20:45 <dmsimard[m]> @ttx, I don't personally have strong opinions and know what happened to past attempts at rewriting things in golang or rust :) 15:21:27 <dmsimard[m]> I am curious and interested in claims of 10x to 100x performance improvements, though 15:21:34 <amorin> we are trying to implement sentry inside neutron downstream, but this is still wip 15:23:19 <dmsimard[m]> We have something on the order of 100 million keystone API calls in 24h so even 10x improvement is significant 15:24:03 <ttx> dmsimard: interesting to know that at least the struggle is real (not sure that rewrite is the solution, but there is a problem to solve) 15:24:44 * dmsimard[m] nod 15:25:01 <ttx> Any other interesting insight from the PTG week? 15:26:20 <ttx> if not, moving on 15:26:23 <ttx> #topic Next meeting(s) 15:26:40 <ttx> Our next meeting is scheduled for May 28, 15utc, here as usual 15:26:45 <dmsimard[m]> I noted a challenge in ARM testing from the nova pad since Linaro has stepped away 15:27:05 <ttx> My personal goal is to have a guest and a date for our next virtual meetup by then 15:28:17 <ttx> #topic Open discussion 15:29:17 <dmsimard[m]> Any plans for operators and this SIG at this upcoming France summit ? 15:31:12 <ttx> We'd like to have something yes 15:31:57 <ttx> either a ops feedback session, or something Large-Scale-SIG specific 15:32:34 <dmsimard[m]> That would be great, let us know if we can help :) 15:34:24 <ttx> Will sure do! We should see clearer in a month 15:34:33 <ttx> Alright, anything else to discuss? 15:34:51 <dmsimard[m]> nothing from me 15:36:02 <ttx> OK then... Thanks for joining. Looks like we really need a new high-profile meetup to promote the monthly meeting and get more attendance :) 15:36:09 <ttx> #endmeeting