14:00:23 <haleyb> #startmeeting networking 14:00:23 <opendevmeet> Meeting started Tue Mar 18 14:00:23 2025 UTC and is due to finish in 60 minutes. The chair is haleyb. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:23 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:23 <opendevmeet> The meeting name has been set to 'networking' 14:00:26 <haleyb> Ping list: bcafarel, elvira, frickler, mlavalle, mtomaska, obondarev, slaweq, tobias-urdin, ykarel, lajoskatona, jlibosva, averdagu, haleyb, ralonsoh 14:00:28 <haleyb> o/ 14:00:33 <obondarev> o/ 14:00:34 <slaweq> o/ 14:00:35 <ralonsoh> hello 14:00:35 <bcafarel> o/ 14:00:37 <mlavalle> \o 14:01:01 <rubasov> o/ 14:01:28 <haleyb> #announcements 14:01:32 <lajoskatona> o/ 14:01:39 <haleyb> We are currently in week R-2 of Epoxy 14:01:59 <haleyb> RC1 was tagged and stable/2025.1 branch created 14:02:26 <haleyb> so anything critical should be cherry-picked to get into RC2 14:02:44 <haleyb> i know there is the IPv6 metadata one 14:03:15 <ralonsoh> the dhcp eventlet removal 14:03:23 <ralonsoh> https://review.opendev.org/c/openstack/neutron/+/944826/2 14:03:28 <haleyb> Final RC deadline: March 28th, 2025 (R-1 week) 14:04:09 <haleyb> ralonsoh: oh, the revert 14:04:29 <ralonsoh> yes, both patches 14:04:34 <ralonsoh> must be in 2025.1 too 14:04:45 <lajoskatona> +1 14:05:07 <haleyb> glad we figured that out 14:05:40 <lajoskatona> after half openstack collapsed, but in time anyway :-) 14:06:04 <haleyb> https://review.opendev.org/c/openstack/neutron/+/944634/2 is the parent change, got it 14:06:31 <haleyb> Final 2025.1 Epoxy release: April 2nd, 2025 14:07:20 <haleyb> So please ping people here if you need reviews on critical gate fixes, and/or get them on the priority board 14:08:00 <haleyb> Project Teams Gathering (virtual): April 7th-11th, 2025 14:08:06 <haleyb> #link https://ptg.openinfra.dev/ to sign up 14:08:23 <haleyb> #link https://etherpad.opendev.org/p/apr2025-ptg-neutron 14:08:49 <haleyb> #link https://ptg.opendev.org/ptg.html for room assignments 14:09:34 <haleyb> I added two slots Tuesday, 4 on Wednesday and Thursday 14:09:41 <haleyb> did not want to conflict with eventlet 14:10:22 <haleyb> please add topics to the etherpad, think we should just push all RFE discussions until then 14:11:16 <haleyb> that was all i had for announcements today, any others? 14:11:29 <lajoskatona> thanks for PTG management :-) 14:12:02 <haleyb> that's what you pay me for, oh wait... :) 14:12:17 <sahid> o/ late 14:12:29 <haleyb> o/ 14:12:36 <haleyb> #topic bugs 14:12:45 <haleyb> mlavalle was deputy last week, report is at 14:12:49 <mlavalle> \o 14:12:51 <haleyb> #link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/UA5Q7RUKIIWHZLYIEERYHB4T5NQANEPP/ 14:14:02 <haleyb> so the high have been addressed 14:14:14 <haleyb> and the dhcp one is in flight 14:15:09 <haleyb> #link https://bugs.launchpad.net/neutron/+bug/2101871 14:15:10 <haleyb> Use of revision_number for router update on 2023.2 Bobcat cause error 500 14:15:29 <haleyb> that was an interesting one, still waiting for the commands to reproduce 14:15:33 <mlavalle> we are waiting for a response from submitter 14:16:36 <haleyb> and this one is incomplete as well 14:16:39 <haleyb> #link https://bugs.launchpad.net/neutron/+bug/2102044 14:16:46 <haleyb> Nova-API problem - cannot list instances anymore 14:16:58 <haleyb> might be config, i'm assuming it's not packages 14:18:01 <ralonsoh> but do we have the Neutron APi exception? 14:18:12 <haleyb> ralonsoh: no he never pasted it 14:18:23 <ralonsoh> ok, we need it, for sure 14:18:56 <haleyb> yes. i'm watching as i want to make sure it's not a distro problem since it was ubuntu, but don't think it is 14:19:47 <haleyb> i did start seeing random batch_notifier unit test failures, one was in the metadata patch 14:19:50 <haleyb> #link https://bugs.launchpad.net/neutron/+bug/2100806 14:19:58 <haleyb> saw them again i meant to say 14:20:14 <haleyb> ralonsoh: can you see my comment in https://review.opendev.org/c/openstack/neutron/+/943212 ? maybe we can discuss after meeting 14:20:24 <ralonsoh> sure 14:20:31 <opendevreview> Candido Campos Rivas proposed x/whitebox-neutron-tempest-plugin master: Increase burst window size to avoid unstabilities in slow envs https://review.opendev.org/c/x/whitebox-neutron-tempest-plugin/+/944616 14:21:14 <haleyb> any other bugs to discuss? 14:21:28 <ralonsoh> yes, two 14:21:33 <haleyb> sure 14:21:50 <ralonsoh> https://bugs.launchpad.net/neutron/+bug/2102090 14:22:03 <ralonsoh> I initially set it as low and no RFE 14:22:15 <ralonsoh> because that could be a bug, not a RFE 14:22:44 <ralonsoh> the problem: with eventlet module, all workers (API, maintenance, rpc, periodic) were executed in the same process 14:23:12 <ralonsoh> now this is not happening, but the OVN placement initial config is in the maintenance worker 14:23:32 <ralonsoh> but anyone will expect that, if you restart the Neutron API, that will be executed 14:23:44 <ralonsoh> so the question is: should we move this to the Neutron API code? 14:24:08 <ralonsoh> now, if we want to re-execute this initial config, we need to restart the maintenance process 14:25:18 <haleyb> ralonsoh: it seems it was almost an unintended change and should be fixed 14:25:48 <ralonsoh> perfect, I had the same idea 14:25:53 <ralonsoh> this is why I opened the bug 14:26:17 <ralonsoh> anyone else if free to comment on the LP bug, if needed 14:26:25 <ralonsoh> the second one: 14:26:31 <ralonsoh> this is the upper patch: https://review.opendev.org/c/openstack/neutron/+/944053 14:26:37 <ralonsoh> it needs to be backported 14:26:53 <ralonsoh> but the first patch (singleton one) could be intrusive 14:27:13 <ralonsoh> I implemented that for the tests, but I can re-implement the tests for stable branches without it 14:27:41 <ralonsoh> so please, review these patches and I'll backport them (2nd and 3rd patch) without the 1st one 14:27:56 <ralonsoh> that's all 14:28:02 <frickler> about https://review.opendev.org/c/openstack/neutron/+/944763 / https://launchpad.net/bugs/2103417 14:28:19 <frickler> a) it is blocking kolla and we have no feasible workaround 14:28:35 <frickler> b) I wonder how that issue could have been detected within neutron 14:29:29 <ralonsoh> we use metadata in the L3 agent 14:29:43 <ralonsoh> I don't know if we have any test using the dhcp one 14:30:29 <frickler> well the issue in kolla isn't that metadata is broken, but that the dhcp agent is failing and thus instances don't get a dhcp response 14:30:33 <ralonsoh> the patch needs a respin, but looks fine. Could be merged today and backported 14:30:43 <frickler> I'm not sure why that doesn't happen in a devstack setup 14:31:25 <ralonsoh> we would need to check all our tempest tests 14:31:31 <haleyb> the changes to the unit test would trigger it from waht i remember from an early PS 14:31:39 <lajoskatona> the dhcp failure is not related to the eventlet change? 14:31:45 <lajoskatona> or that is a parallel one< 14:31:48 <ralonsoh> not this one 14:31:52 <lajoskatona> ack 14:31:56 <haleyb> it's from my metadata ipv6 change 14:32:12 <haleyb> i fixed one bug but caused another 14:32:25 <lajoskatona> that's life :-) 14:32:55 <ralonsoh> haleyb, will you push the new PS? 14:33:01 <lajoskatona> the dragon has endles new heads, and we had them just before release this time it seems 14:33:10 <haleyb> ralonsoh: sure i can do right after meeting if i don't see an update 14:33:22 <frickler> I'm not sure whether running a (n-v) kolla job on neutron changes would be an option, or do you think that that would be too much? 14:34:16 <ralonsoh> you can have your own neutron-master CI job 14:34:27 <ralonsoh> with a daily execution 14:34:42 <frickler> well we consume neutron from master anyway, so that would be just our normal jobs 14:35:34 <frickler> the idea would be able to recognize issues before they are merged into master 14:36:05 <ralonsoh> yes but this is expensive, just when we are removing jobs 14:36:18 <ralonsoh> when was the last time Neutron introduced an error in kolla this big? 14:36:46 <ralonsoh> also, as you said, you have your own jobs and we responded quick to this problem 14:37:24 <lajoskatona> like we have master oslo or main pyroute2 jobs to have quick alert 14:38:10 <slaweq> frickler we had in the past such kind of cross project non-voting jobs 14:38:30 <slaweq> but in order to save some infra resources we moved those jobs to periodic queue 14:38:53 <slaweq> so maybe kolla-neutron job can be in periodic line too? 14:39:14 <slaweq> and we are checking those periodic jobs weekly 14:39:21 <slaweq> so IMO it is good tradeoff 14:39:49 <lajoskatona> +1 for periodic weekly and let's check them on weekly CI meeting 14:41:10 <frickler> I think we have those already, but querying zuul is a bit slow for me right now 14:41:51 <frickler> anyway, fine for me for now 14:42:49 <haleyb> ok, we can discuss in CI meeting to see if there is any work here to help 14:43:08 <haleyb> that was all the bugs i think 14:43:26 <haleyb> this week jlibosva is the bug deputy, but i don't see him 14:43:35 <ralonsoh> I'll ping him later 14:43:41 <haleyb> can someone ping him on slack channel, thanks 14:43:59 <haleyb> next week is obondarev, is that good with you? 14:44:08 <obondarev> yep 14:44:15 <haleyb> thanks! 14:44:29 <haleyb> #topic community goals 14:45:25 <haleyb> lajoskatona: i don't see any neutronclient changes in-flight, so that's good :) 14:45:34 <haleyb> is it something to continue next cycle? 14:46:18 <haleyb> we can come back to that 14:46:27 <lajoskatona> yes, all open things merged, I have to go and find the next api topic to change to sdk 14:46:45 <haleyb> lajoskatona: ack, thanks 14:47:15 <haleyb> ralonsoh: evenlet removal is next, i'll let you give status 14:47:24 <ralonsoh> yes, in bullet points 14:47:28 <ralonsoh> SRIOV: merged 14:47:43 <ralonsoh> DHCP: reverted due to missing backend in oslo.service 14:47:58 <ralonsoh> MacVTap: I need to check, but same ^^ problem 14:48:13 <ralonsoh> L3: no activity (I had bug fixes to push) 14:48:16 <ralonsoh> OVS: https://review.opendev.org/q/topic:%22bug/2087939%22+status:open 14:48:20 <ralonsoh> that's all 14:48:30 <lajoskatona> yes we said macvtap also uses oslo.services, so have to wait with it 14:48:44 <ralonsoh> yes 14:49:40 <haleyb> ralonsoh: in your opinion, should be merge any/all of the ovs changes and get into 2025.1? i did have some comments out to sahid on one of the changes at least 14:50:07 <haleyb> #link https://review.opendev.org/c/openstack/neutron/+/940983 - i need to look again 14:50:23 <ralonsoh> no, we don't need these changes in stable 14:50:55 <haleyb> ack, that's what i thought, plus i don't want to cause breakeage 14:51:01 <ralonsoh> right 14:51:42 <haleyb> ok, anything else for community topics? 14:52:00 <opendevreview> Pierre Riteau proposed openstack/neutron master: Spawn IPv6 metadata proxy only when required https://review.opendev.org/c/openstack/neutron/+/944763 14:52:05 <haleyb> #topic on-demand 14:52:20 <haleyb> we already discussed the OVN placement bug 14:52:53 <ralonsoh> ah yes, I added it in this list 14:53:24 <haleyb> ralonsoh: regarding the DB bug you fixed (i can't find the patch) - is that something we need to backport? 14:53:41 <ralonsoh> which one? 14:53:51 <haleyb> https://review.opendev.org/c/openstack/neutron/+/944803 14:54:24 <haleyb> would that affect a db migration? 14:54:44 <ralonsoh> pffff to be honest, I don't know why this was not affecting the migration 14:55:09 <ralonsoh> I'll push stable only patches for the affected ones 14:55:17 <ralonsoh> that should be 2025.1 and 2024.2 14:55:30 <ralonsoh> and 2023.1 14:55:41 <haleyb> ack, thanks 14:56:11 * haleyb was thinking no one uses openstack? :-p 14:56:34 <haleyb> s/newer openstack 14:56:51 <haleyb> but then i'm not a comedian 14:57:03 <haleyb> anyways, any other topics to discuss? 14:57:05 <ralonsoh> that should be affecting the migration in the CI 14:57:16 <ralonsoh> but as I said, I have no idea why is not... 14:57:25 <haleyb> i don't either 14:58:58 <haleyb> ok, thanks for attending everyone, and ping here if there are critical things for RC2 14:59:01 <haleyb> #endmeeting