14:00:43 <haleyb> #startmeeting networking 14:00:43 <opendevmeet> Meeting started Tue Jan 21 14:00:43 2025 UTC and is due to finish in 60 minutes. The chair is haleyb. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:43 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:43 <opendevmeet> The meeting name has been set to 'networking' 14:00:54 <haleyb> Ping list: bcafarel, elvira, frickler, mlavalle, mtomaska, obondarev, slaweq, tobias-urdin, ykarel, lajoskatona, jlibosva, averdagu, amotoki, haleyb, ralonsoh 14:00:55 <mlavalle> \o 14:01:00 <ralonsoh> hello 14:01:00 <svinota> ralonsoh, I couldn't login to edit the wiki page, so if there will be one-two minute for an ad-hoc announce in the end of the meeting, would be nice. 14:01:20 <ralonsoh> ralonsoh, sure, there will be time 14:01:22 <slaweq> o/ 14:01:38 <rubasov> o/ 14:01:58 <lajoskatona> o/ 14:01:59 <haleyb> svinota: there will be an on-demand time at the end, thanks for attending 14:02:06 <svinota> thanks 14:02:13 <haleyb> #announcements 14:02:35 <haleyb> We are in week R-10 if my math is correct 14:02:41 <haleyb> #link https://releases.openstack.org/epoxy/schedule.html 14:03:15 <opendevreview> Rodolfo Alonso proposed openstack/neutron master: WIP == [eventlet-deprecation] Remove the usage of eventlet in the Neutron API https://review.opendev.org/c/openstack/neutron/+/938659 14:03:20 <haleyb> in R-6 we will have non-client library freeze 14:03:32 <opendevreview> Rodolfo Alonso proposed openstack/neutron master: DNM - Test "neutron-ovn-tempest-ipv6-only-ovs*" with WSGI https://review.opendev.org/c/openstack/neutron/+/932601 14:03:33 <haleyb> in R-5 will be E-3 milestone 14:03:38 <bcafarel> late o/ 14:05:07 <haleyb> Reminder: If you have a topic for the drivers meeting on Friday, please add it to the wiki @ https://wiki.openstack.org/wiki/Meetings/NeutronDrivers 14:05:18 <haleyb> And of course a reminder to use the priorities dashboard for patches in the "ready to merge" state 14:06:05 <haleyb> I sent the Linuxbridge removal email to the ML last week, and it seems people have accepted is was the right thing to do 14:07:01 <slaweq> great :) thx haleyb for doing this finally 14:07:02 <haleyb> there was a related email from Cern regarding knowledge sharing 14:07:06 <lajoskatona> good question if heores will move the code to x/ namsspace, let's see:-) 14:08:52 <haleyb> i will not hold my breath on that, i would hope we can spend any time helping with migration questions and OVN adoption 14:09:12 <lajoskatona> +1 for pushing toward migration 14:09:47 <haleyb> i did not have any other announcements 14:10:16 <haleyb> #topic Bugs 14:10:21 <haleyb> lajoskatona was deputy last week 14:10:26 <haleyb> #link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/JIRUHN7HMJUDB7KB55UYUZ5Q2IK2MC6X/ 14:10:36 <haleyb> there were quite a few 14:11:19 <opendevreview> Sahid Orentino Ferdjaoui proposed openstack/neutron master: async_process: fix potential race condition with respawn https://review.opendev.org/c/openstack/neutron/+/939627 14:11:20 <opendevreview> Sahid Orentino Ferdjaoui proposed openstack/neutron master: async_process: remove usage of eventlet for AsyncProcess https://review.opendev.org/c/openstack/neutron/+/939348 14:11:20 <opendevreview> Sahid Orentino Ferdjaoui proposed openstack/neutron master: ovs: reimplement signals handling https://review.opendev.org/c/openstack/neutron/+/939321 14:11:21 <opendevreview> Sahid Orentino Ferdjaoui proposed openstack/neutron master: common: fix wait_until_true to support native thread https://review.opendev.org/c/openstack/neutron/+/937843 14:11:21 <opendevreview> Sahid Orentino Ferdjaoui proposed openstack/neutron master: ovs: remove the usage of eventlet in the OVS agent https://review.opendev.org/c/openstack/neutron/+/937765 14:11:33 <haleyb> lajoskatona: were there any bugs you wanted to highlight? 14:12:31 <lajoskatona> not really there were few without owners 14:12:55 <lajoskatona> but as I remember the high prio ones are under control 14:13:15 <haleyb> ok 14:13:18 <lajoskatona> https://bugs.launchpad.net/neutron/+bug/2094842 14:13:27 <lajoskatona> this one in the high list without owner 14:13:37 <lajoskatona> "ovs.db.error.Error: ovsdb error: 0 values when type requires between 1 and 9223372036854775807" when router is attached to network without subnet 14:13:40 <haleyb> that looks familiar, or at least we just fixed something related in quota 14:14:10 <haleyb> at least it looks like the quota one, but it wasn't in ovsdb of course 14:14:40 <ralonsoh> the cause is explained in the description 14:14:41 <haleyb> oh, maybe this is related to a fix ralonsoh had regarding gw_port without an IP address? 14:14:51 <ralonsoh> we can't create a GW port wihtout subnet in the GW network 14:15:13 <ralonsoh> haleyb, I don't have anything yet 14:15:25 <ralonsoh> there is a quick attempt https://review.opendev.org/c/openstack/neutron/+/939253 14:15:32 <ralonsoh> but this is something that needs to be tested 14:16:12 <haleyb> ralonsoh: ah, that was it, it was ihrachys not you, sorry i mis-remembered 14:17:03 <lajoskatona> it is visible in zuul logs also not just for master 14:17:13 <lajoskatona> see the opensearch logs I added as comment 14:18:16 <haleyb> ack, i guess we can wait until that change is updated to see if it fixes it 14:18:33 <lajoskatona> +1 14:18:56 <haleyb> another one ihar filed was 14:19:02 <haleyb> #link https://bugs.launchpad.net/neutron/+bug/2094846 14:19:22 <haleyb> trunk driver functional failure 14:20:26 <haleyb> according to codesearch there was only 1 failure so far, but it looks like something that shouldn't have happened 14:20:32 <ralonsoh> it happened only once 14:20:40 <ralonsoh> and this is a functional test 14:20:53 <ralonsoh> I would reduce the importance, but I'll keep it in the list 14:21:17 <haleyb> yes, let's see if it happens again first 14:21:21 <lajoskatona> it is medium at the moment due to low frequesncy, but we can dereease 14:22:19 <haleyb> sure 14:22:38 <haleyb> there was another medium i started to take a look at 14:22:42 <haleyb> #link https://bugs.launchpad.net/neutron/+bug/2095097 14:22:49 <haleyb> 500 error creating resource with emoji for name 14:23:11 <haleyb> it happens for any string field, for example, description too 14:23:28 <lajoskatona> that's a funny one, but really happens 14:23:50 <haleyb> my original thought was to just fix the network db code, but now i think we should change neutron-lib to add a validator for such fields 14:23:59 <haleyb> but i wanted to get others thoughts on that 14:24:16 <ralonsoh> an extra check for printable chars is ok 14:24:23 <lajoskatona> haleyb: yes a validator sounds good 14:24:30 <ralonsoh> it won't break anything using "normal" chars 14:24:50 <ralonsoh> and with the n-lib validator, we'll have a printable exception, not error 500 14:24:50 <haleyb> and i'm not sure it's a medium as i originally tagged it, since it only affects the caller 14:25:21 <ralonsoh> the only problem here is to return a valid exception 14:25:22 <haleyb> yes, a quick hack of the network db code and i could get a 400, 14:25:23 <ralonsoh> not 500 14:25:33 <ralonsoh> so I would lower the priority 14:25:37 <haleyb> i.e. InvalidInput such exception 14:25:43 <ralonsoh> exactly 14:26:26 <lajoskatona> nova gives back 400 for such input for example 14:26:54 <ralonsoh> as it should be, this is a bad request, not a server error 14:27:19 <lajoskatona> exactly, so good to have simlar response from Neutron API also 14:27:53 <slaweq> ++ for some 4xx error in such case 14:28:23 <haleyb> i took it and can work on a validator 14:28:40 <haleyb> next unowned bug 14:28:45 <haleyb> #link https://bugs.launchpad.net/neutron/+bug/2095185 14:29:03 <haleyb> ihrachys was busy last week 14:29:33 <haleyb> test_legacy_router_conntrack_helper failure 14:29:35 <lajoskatona> even on Friday I got a bunch of bugs from Him :-) 14:29:51 <haleyb> he gave a possible solution, but would need more investigation 14:29:58 <lajoskatona> this one was seen in zuul logs only on stable/ 14:30:07 <ralonsoh> last sentence of the description is a clue for a fix 14:30:19 <ralonsoh> so I would give a try 14:30:26 <lajoskatona> but as he wrote can happen on master 14:30:34 <opendevreview> Ihar Hrachyshka proposed openstack/neutron master: QinQ implementation for the ML2/OVN backend https://review.opendev.org/c/openstack/neutron/+/937633 14:30:51 <haleyb> right, just needs an owner if someone can take it 14:31:38 <haleyb> the last bug without owner is 14:31:44 <haleyb> #link https://bugs.launchpad.net/neutron/+bug/2095091 14:32:28 <haleyb> ykarel_: you filed this, does it happen frequently enough to debug it further? 14:32:39 <lajoskatona> yesterday we discussed on the CI meeting with ykarel and slwaq and mlavalle 14:32:49 <ykarel_> haleyb, happened only once 14:32:50 <haleyb> ah 14:33:02 <ykarel_> mlavalle, will be working on alternative 14:33:02 * haleyb was out yesterday as it was a holiday 14:33:32 <haleyb> ok, as long as someone is working on it 14:33:45 <haleyb> any other bugs to discuss? 14:34:19 <haleyb> ykarel is the deputy this week, mtomaska next week 14:34:26 <haleyb> is that ok? 14:35:33 <haleyb> i will follow-up later so we can move along with other items 14:35:37 <haleyb> #topic community goals 14:36:04 <mlavalle> haleyb ykarel I'll keep an eye on that failure 14:36:09 <haleyb> ack 14:36:22 <haleyb> lajoskatona: seems forward progress on neutronclient deprecation? 14:36:28 <haleyb> #link https://review.opendev.org/q/topic:%22bug/1999774%22 14:36:43 <ykarel> haleyb, ack for deputy, thx for the reminder 14:36:44 <lajoskatona> not this week, I had no time for it 14:37:16 <haleyb> well, at least both open changes have zuul +1, i will take a look at the second in the chain 14:37:26 <lajoskatona> thanks 14:38:22 <haleyb> and evenlet deprecation 14:38:35 <ralonsoh> this is not going well 14:38:47 <ralonsoh> I can't identify what is happening 14:38:59 <ralonsoh> initially the hash ring manager was a possible culprit 14:39:22 <ralonsoh> but since last week (we merged a patch), the timeouts are increased and we don't have idle nodes 14:39:49 <ralonsoh> now the problem is, sometimes, that the events are processed in different order, DB locks and other problems I can't explain 14:40:02 <ralonsoh> I'm trying to open the related bugs 14:40:09 <ralonsoh> and I'm trying with this: 14:40:16 <ralonsoh> https://review.opendev.org/c/openstack/neutron/+/938659 14:40:42 <ralonsoh> there are many places in the code (python-ovs, ovsdbapp, Neutron, etc) where the libraries have workarounds for eventlet 14:40:44 <ralonsoh> so I' 14:40:46 <ralonsoh> sorry 14:41:01 <ralonsoh> so I would prefer testing Neutron API without eventlet 14:41:07 <opendevreview> Merged x/whitebox-neutron-tempest-plugin master: Add config dict dataplane_podified_services https://review.opendev.org/c/x/whitebox-neutron-tempest-plugin/+/939519 14:41:10 <ralonsoh> because this is actually the expected behaviour 14:41:29 <ralonsoh> and that's all (I'm overwhelmed right now) 14:43:20 <haleyb> ralonsoh: what else can we do to help? i know ihrachys was helping last week, and i can look at things this week 14:43:36 <haleyb> but i realize it is very complicated 14:43:47 <ralonsoh> haleyb, first of all, trying to check what is happening in CI patches like https://review.opendev.org/c/openstack/neutron/+/932601/12?tab=change-view-tab-header-zuul-results-summary 14:44:17 <ralonsoh> even without eventlet, I see common errors (ports not active, PG not present, etc) 14:44:53 <ralonsoh> so, for example, to take one CI log and try to squeeze it to find what was the problem 14:45:30 <haleyb> ack, and i would ask anyone with spare cycles to take a look so we come up with some ideas 14:46:17 <lajoskatona> I ry to allocate some time to it, anyway the n-d-r issue is also there with the same root 14:46:24 <haleyb> thanks 14:46:41 <haleyb> i want to move to on-demand as svinota is here 14:46:44 <haleyb> #topic on-demand 14:46:52 <ralonsoh> svinota, hello! 14:46:56 <svinota> hello 14:47:06 <svinota> and thanks for your time :) 14:47:17 <haleyb> thanks for pyroute2 :) 14:47:26 <lajoskatona> Welcome :-) 14:47:38 <svinota> so shortly, the pyroute2 project is undergoing a big (to say the least) shift in the core 14:48:07 <svinota> I continue to support the old synchronous API, it will not be dropped 14:48:28 <svinota> but from 0.9.1 the core is based on asyncio 14:48:49 <svinota> related docs: https://docs.pyroute2.org/asyncio.html 14:48:58 <svinota> also: https://docs.pyroute2.org/iproute_intro.html 14:49:15 <svinota> thanks to ykarel , I start to monitor the zuul as well 14:50:02 <svinota> so I hope not to miss major compatibility errors; beside of that I run own integration tests 14:51:15 <svinota> I don't know how this shift can affect the neutron core, so as I said — I keep the sync API anyways 14:51:54 <svinota> but if the project can provide something more within the async paradigm — let me know 14:52:01 <haleyb> is there a point in time where we should start migrating things? will there be gotcha's, etc? 14:52:05 <svinota> that's all folks :) thanks 14:52:12 <ralonsoh> this is a bug change, of course. We use pyroute "only" in the agents (luckily, because asyncio is not compatible with wsgi) 14:52:35 <ralonsoh> and most of the pyrouter calls are executed inside a root daemon (privsep) 14:52:59 <ralonsoh> so we'll need to refactor the non-admin calls and the admin calls independetly 14:53:08 <ralonsoh> can both coexist right now? 14:53:29 <svinota> 0.9.1 is scheduled for the beginning of February, but as I see, the current neutron code works with the master branch 14:53:42 <svinota> and the master branch is already async-enabled 14:53:46 <ralonsoh> perfect 14:53:53 <svinota> so I believe no migration is required 14:54:07 <ralonsoh> but eventually this is the desired implementation, right? 14:55:05 <svinota> async is the first class citizen now, so the sync code is simply a set of wrappers around it; so you can use either variant, totally up to you, both will coexist 14:55:24 <ralonsoh> perfect 14:55:32 <svinota> in the best case the library users should not notice any change 14:55:59 <slaweq> svinota thx for sharing all of that with us 14:56:05 <svinota> (it took like some months to fix it in that way, but I hope it pays back) 14:56:14 <slaweq> but neutron is not the only user of the pyroute2 in openstack 14:56:25 <ralonsoh> os-vif, right 14:56:27 <slaweq> did you maybe had any chance to test it with other projects as well? 14:56:34 <ralonsoh> and other projects in networking 14:56:38 <slaweq> octavia I see is using it a bit 14:57:06 <svinota> I tested kuryr. Octavia should work as before. 14:57:08 <ralonsoh> so this should be on us to create pyroute-master CI jobs 14:57:11 <slaweq> https://codesearch.opendev.org/?q=pyroute2&i=nope&literal=nope&files=&excludeFiles=&repos= 14:57:16 <slaweq> I see many of them really 14:57:16 <svinota> I forgot os-vif, thanks 14:57:33 <slaweq> I don't want you to test them all on your own 14:57:58 <haleyb> ralonsoh: we do have functional and fullstack pyroute2 master jobs 14:58:05 <svinota> slaweq, thanks for the link. I will try to go through the list and see if we can include them in the integration testing 14:58:07 <slaweq> but maybe you could send email to the openstack-discuss ML to raise awarness of this change in different teams 14:58:14 <ralonsoh> haleyb, yes but for all projects? 14:58:43 <haleyb> ralonsoh: i only know of neutron 14:58:50 <lajoskatona> perhaps some common tempest integration job? 14:58:59 <lajoskatona> that an cover a lot of projects 14:59:25 <slaweq> svinota thx a lot, if you will have any questions about it, please ping me :) 14:59:36 <svinota> slaweq, 10x 15:00:01 <ralonsoh> ok, let's create a bug in LP to handle this 15:00:11 <ralonsoh> to create the needed CI jobs for pyroute-master 15:00:16 <ralonsoh> functional or tempest 15:00:37 <ralonsoh> and, of course, svinota thanks a lot 15:00:55 <lajoskatona> and advertise it on mail list to have a flag for all users not just for us 15:01:09 <svinota> thanks, colleagues 15:01:22 <lajoskatona> svinota thanks for the good cooperation and for the news :-) 15:01:38 <haleyb> yes, thanks a lot svinota! 15:02:01 <haleyb> does anyone want to file the bug, etc? or is this on the PTL :) 15:02:32 <lajoskatona> always the one who asks :-) 15:02:43 <slaweq> haha 15:02:58 <haleyb> ok, i have to drop for another meeting 15:03:07 <haleyb> thanks everyone for attending, have a good week 15:03:12 <slaweq> o/ 15:03:14 <haleyb> #endmeeting