16:00:21 <gthiemonge> #startmeeting Octavia
16:00:22 <opendevmeet> Meeting started Wed May 24 16:00:21 2023 UTC and is due to finish in 60 minutes.  The chair is gthiemonge. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:22 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:22 <opendevmeet> The meeting name has been set to 'octavia'
16:00:27 <gthiemonge> Hi
16:00:31 <johnsom> o/
16:01:01 <jdacunha4> Hello
16:01:09 <tweining_> o/
16:01:38 <QG> o/
16:01:46 <matfechner> o/
16:02:20 <gthiemonge> #topic Announcements
16:02:46 <gthiemonge> No announcement from me... Does anyone have any announcements?
16:02:54 <johnsom> Nope
16:04:24 <gthiemonge> ack
16:04:30 <gthiemonge> #topic CI Status
16:04:40 <gthiemonge> we need a followup commit for the configuration of the CI jobs
16:04:54 <gthiemonge> last week, I split the scenario jobs in 2 jobs in the check pipeline
16:05:06 <gthiemonge> but I didn't update the gate pipeline :/
16:05:17 <gthiemonge> here is the new patch
16:05:21 <gthiemonge> #link https://review.opendev.org/c/openstack/octavia/+/883999
16:05:43 <gthiemonge> as it unblocks the gate, it will be approved right after this meeting, unless someone raises concerns about it
16:06:28 <gthiemonge> then I will squash this patch with https://review.opendev.org/c/openstack/octavia/+/883131
16:06:40 <gthiemonge> and propose backports to the stable branches that use Ubuntu Jammy
16:07:22 <gthiemonge> FYI FIPS jobs are broken, it is not an octavia issue but a global issue
16:07:34 <gthiemonge> there's a bug in the Octavia LP
16:07:38 <gthiemonge> #link https://bugs.launchpad.net/octavia/+bug/2020434
16:07:47 <gthiemonge> I know that some folks are investigating it
16:09:36 <gthiemonge> #topic Brief progress reports / bugs needing review
16:10:27 <gthiemonge> I was on PTO, not much activity
16:10:46 <jdacunha4> I need some review/replies to the ACMEv2 implementation
16:10:59 <jdacunha4> here is the latest doc generated: https://426a74fa4eaec8e077a0-814491f7a8e968b50eeaf14b0c7f115f.ssl.cf5.rackcdn.com/877281/11/check/openstack-tox-docs/8486822/docs/contributor/specs/version1.1/letsencrypt.html
16:11:11 <johnsom> I have made some progress with my SRIOV investigation. I was able to manually attach a VF to an amphora.
16:11:22 <johnsom> Next step for me is putting together a spec for review
16:11:26 <gthiemonge> jdacunha4: ack, I'm adding it to my todo list
16:11:29 <jdacunha4> some questions from johnsom are not replied, i need some more inputs/discussions
16:12:23 <tweining_> I am very interested in seeing that SRIOV spec.
16:12:59 <jdacunha4> #link https://review.opendev.org/c/openstack/octavia/+/877281
16:13:02 <jdacunha4> thx
16:14:30 <tweining_> I got DPDK working with amps in my devstack environment without any code modification. It seems as long as the compute node is configured for DPDK nova will automatically configure the guest to use a vhost user mode port, which is nice.
16:16:28 <gthiemonge> tweining_: do you know if the CPUs of the VMs are pinned automatically by Nova?
16:17:35 <tweining_> no, that is not Nova's job.
16:17:44 <gthiemonge> (I mean on the compute node)
16:18:02 <tweining_> I did that manually using ovs-vsctl
16:18:35 <tweining_> it's the job of the deployment tool basically
16:19:43 <QG> #link https://docs.openstack.org/nova/pike/admin/cpu-topologies.html
16:20:48 <QG> you can pin, but it's not really praticle
16:20:49 <gthiemonge> ok it's in the flavor
16:21:39 <tweining_> pinning cpus in the VM is unrelated to running DPDK on the compute node though
16:22:10 <QG> like nova is not able to spread out between numa, it will first full fill the first one and then do the second one
16:22:34 <QG> tweining_  yep
16:22:51 <johnsom> Typically with DPDK you will want the hypervisor CPUs pinned for the VM too, not just processes inside the guest
16:24:41 <tweining_> mmh, ok. I didn't think about that yet.
16:26:34 <gthiemonge> there's no rush ATM :)
16:30:04 <tweining_> anyway, I think there are quite a few patches that are waiting for review now
16:30:18 <gthiemonge> +1
16:30:35 <gthiemonge> #topic Open Discussion
16:31:50 <gthiemonge> any other topics, folks?
16:32:32 <tweining_> just for my understanding: do you expect Octavia to configure Nova to pin cpus for amp VMs?
16:33:08 <gthiemonge> if we can do it with the flavor, we don't need to do anything else in Octavia
16:33:40 <tweining_> yeah, hugepages need to be enabled in the flavor (or image itself) anyway
16:33:47 <gthiemonge> I think it is the "hw:cpu_policy=dedicated" property in QG's link
16:34:27 <johnsom> Yeah, that is a common setting
16:34:56 <johnsom> As also mentioned, NUMA can become an issue too. But we need to dance the line of not becoming a nova scheduler....
16:36:35 <tweining_> ok, thanks. I think that's all from me
16:37:07 <gthiemonge> ok, I think that's it for this meeting
16:37:11 <gthiemonge> thank you guys!
16:37:13 <gthiemonge> #endmeeting