Tuesday, 2025-06-24

kyu0[m]Hello, could you please review my comment? https://review.opendev.org/c/openstack/neutron/+/94732401:25
opendevreviewKevin Carter proposed openstack/neutron master: feat(OVN): add new mode to only add when missing  https://review.opendev.org/c/openstack/neutron/+/94182403:50
*** ykarel_ is now known as ykarel04:18
ralonsohykarel, I'm checking https://4e87414b0fe3ca1de111-0dd3b493df614cc7e41a46754664ea83.ssl.cf5.rackcdn.com/openstack/a858fb193a21437987d557b6de06dbdc/testr_results.html06:01
ralonsohI'll open a LP bug06:01
ykarelralonsoh, ack this one took ~ 17 seconds to cleanup mac binding, and timeout is 15 sec06:02
ykarelnot sure why it took that much time06:02
ykarelralonsoh, another failure https://13105f8ef823650ec019-cb65abe58d87a1a010092a9adcbaff91.ssl.cf2.rackcdn.com/openstack/9f625bccccc54f9bb3b76dd44cc78bb9/testr_results.html in this it took ~ 16 seconds06:57
ralonsohI'll add it to the bug (still creating it)06:57
ykarelack i am reporting the unit test issue06:58
ykarelreported for unit test https://bugs.launchpad.net/neutron/+bug/211524107:06
opendevreviewRodolfo Alonso proposed openstack/neutron master: [FT] Wait for the FIP Port_Binding event before checking MAC removal  https://review.opendev.org/c/openstack/neutron/+/95319507:46
opendevreviewTakashi Kajinami proposed openstack/neutron master: Drop unused method  https://review.opendev.org/c/openstack/neutron/+/95297708:34
tkajinam^^^ seems like we are currently hitting 82% percent boundary and a small change can break it very easily...08:35
opendevreviewTakashi Kajinami proposed openstack/neutron-lib master: Remove constants for LBaaS  https://review.opendev.org/c/openstack/neutron-lib/+/95319809:20
opendevreviewTakashi Kajinami proposed openstack/neutron-lib master: Remove unused service type constant for FWaaS  https://review.opendev.org/c/openstack/neutron-lib/+/95319909:25
opendevreviewMerged openstack/neutron master: Remove compat logic for neutron-lib < 3.9.0  https://review.opendev.org/c/openstack/neutron/+/95311110:21
*** LarsErik1 is now known as LarsErikP10:30
zigohaleyb: FYI, I understood what happened. The issue is in both Debian and Ubuntu (I just checked the Ubuntu package): it's not Neutron that is killing the keepalived processes, but systemd !12:14
zigoI will add a killmode=process in the .service to fix it.12:15
haleybzigo: good to hear you figured it out12:56
haleyb#startmeeting networking13:00
opendevmeetMeeting started Tue Jun 24 13:00:26 2025 UTC and is due to finish in 60 minutes.  The chair is haleyb. Information about MeetBot at http://wiki.debian.org/MeetBot.13:00
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.13:00
opendevmeetThe meeting name has been set to 'networking'13:00
haleybPing list: bcafarel, elvira, frickler, mlavalle, mtomaska, obondarev, slaweq, tobias-urdin, ykarel, lajoskatona, jlibosva, averdagu, haleyb, ralonsoh13:00
mlavalle\o13:00
obondarevo/13:00
haleybhi everyone13:00
ralonsohhello13:00
slaweqo/13:00
rubasovo/13:02
haleybi'll get started13:02
haleyb#announcements13:02
haleybWe are currently in Week R-3 of Flamingo13:02
haleybOur next milestone in this development cycle will be Flamingo-2, on July 3rd13:02
haleybThat is next Thursday 8-o13:03
haleybFinal 2025.2 Flamingo release: October 3rd, 202513:03
haleyb#link https://releases.openstack.org/flamingo/schedule.html13:03
bcafarellate o/13:03
haleybThe next OpenInfra PTG[1] will take place October 27-31, 2025 and registration for the event is now open13:05
haleyb#link https://ptg.openinfra.org/13:05
haleybit's virtual and free as it has been for a while13:05
haleybReminder: If you have a topic for the drivers meeting on Friday, please add it to the wiki @ https://wiki.openstack.org/wiki/Meetings/NeutronDrivers13:06
haleyband last announcement13:07
haleybOpenstack is changing from CLA to DCO starting July 1st13:07
lajoskatonao/13:07
haleybi can't find the link to the ML thread, but it just means that signed-off-by will be mandatory in reviews13:08
haleybso change your workflow to 'git commit -a -s' or similar13:08
haleyband if anyone knows how to tweak .gitconfig to make that -s always take affect i'd love to know the trick13:09
haleybany other announcements?13:09
haleyb#topic bugs13:10
haleybmlavalle was the deputy last week, his report is at13:10
haleyb#link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/YNLD3OU44PAA5UOVKXMQDMYVUZHBIGUU/13:10
haleybfirst one is13:11
haleyb#link https://bugs.launchpad.net/neutron/+bug/211473213:11
haleybgrenade-skip-level-always job fails with: Security Group None cannot perform precommit_delete due to Callback13:11
ralonsohI commented there13:12
ralonsohwe can use WSGI in 2024.2, but it wasn't stable until we removed eventlet (that happened in 2025.1)13:12
ralonsohthis is why we don't move to wsgi until 2025.113:12
haleybralonsoh: yes, just reading your note13:12
ralonsohso with 2024.2 will work, but with some issues13:12
ralonsohin any case, skip jobs are not mandatory in non-slurp releases (as is now)13:13
ralonsohso I would move this job to periodic, as we do in neutron13:13
haleybdo we know what job this is failing with? i agree with your assessment13:13
ralonsohwe had multiple issues in wsgi and eventlet13:13
ralonsohonce we removed everything eventlet related from the API, everything started to work stable13:14
haleyboh grenade-skip-level-always i didn't know that was a job, wonder where it lives13:14
ralonsohin 2024.2+WSGI, the eventlet removal is not finished and we can't guarantee any stability, as we can see in the CI13:14
ralonsohin grenade repo13:15
haleybi see nova, swift and manila run it13:15
ykarelralonsoh, can't we remove "logger" mechanism driver from there to fix that job13:16
haleybmanila has it non-voting13:16
ralonsohykarel, I don't think this is the problem13:16
ralonsohin 2024.2+WSGI, we were still using eventlet13:16
ykarelack when i last checked that looked the issue but can cross check13:16
ykareliirc i mentioned that too 13:16
ykarelyes in ci meeting13:17
ykarel<ykarel> and here i see Loaded mechanism driver names: ['ovn', 'logger']13:17
ykarel<ykarel> so need to remove logger mechanism driver13:17
ralonsohwe can try that, for sure13:18
ykarel++ let's do13:18
ralonsohI'll propose a patch13:18
rubasovsidenote on git autosignoff: https://paste.opendev.org/show/828233/13:18
haleybthanks ralonsoh, and if that doesn't work can have them turn-off voting13:19
lajoskatonahttps://opendev.org/openstack/tempest/src/branch/master/zuul.d/integrated-gate.yaml#L46313:19
lajoskatonasorry13:19
haleybrubasov: thanks for that, autoSignOff=true is the trick :)13:19
ralonsohhaleyb, rubasov it is better if we integrate it with the openstack hook13:20
rubasovplus the git hook13:20
ralonsohfor example: https://paste.opendev.org/show/bgqKZ49CDIOpv4ieeIUv/13:20
ralonsoh^^ that is at the end of the openstack hook13:20
haleybralonsoh: i'm assuming someone will make a generic change to cover everything? we can worry when it's turned-on next week13:22
ralonsohno idea, this change can be done locally and copy/paste this hook to all repos you have13:23
ralonsohyou only need setup your name + mail in git, that is mandatory13:23
haleyback, i had done that a while back13:24
haleybnext bug13:24
haleyb#link https://bugs.launchpad.net/neutron/+bug/211505313:24
haleybconntrack flow created for stateless SG rule13:24
haleybslaweq: looks like you picked that up today13:24
slaweqyeap13:25
haleybubuntu noble fails :(13:25
slaweqI am triaging it but seems for me like ovn regression13:25
slaweqyes, I am now trying to reproduce it on Centos 9 also13:26
slaweqbut that's still in progress13:26
haleyback, i'll follow in case there is something that needs a downstream backport13:26
haleyball the other bugs have owners and/or patches13:27
haleybany others to discuss?13:27
haleyblajoskatona: i know we talked last week about you swapping bug deputy weeks with jlibosva, can you still take this week?13:28
haleybi forgot to update the calendar but will right after13:28
lajoskatonaYes, I am on it13:28
haleyblajoskatona: ack, thanks!13:28
haleyband next week will be obondarev - is that Ok with you?13:29
obondarevsure13:29
haleybgreat, thanks!13:29
haleybok, moving on from bugs13:30
haleyb#topic community goals13:30
haleyblajoskatona: any update on neutronclient?13:30
opendevreviewMerged x/whitebox-neutron-tempest-plugin master: [DVR] Add sleep after adding/removing FIPs  https://review.opendev.org/c/x/whitebox-neutron-tempest-plugin/+/95111113:31
lajoskatonayes, I pushed one patch for  horizon, for rbac API13:31
lajoskatonaand I checked this morning and it seems only neutron quotas remained to move to SDK (have to double check again)13:31
lajoskatonaso quite close to finish one project it seems :-)13:32
lajoskatonathe rbac policies patch (wip): https://review.opendev.org/c/openstack/horizon/+/95311613:32
lajoskatonathat's all for SDK vs neutronclient topic13:32
haleybgreat, thanks for all the patches13:33
haleybralonsoh: i'll let you talk about eventlet, it's made some progress as well13:33
ralonsohyes but we have a problem13:34
ralonsohfullstack CI13:34
ralonsoh1) Broken with OVS agent eventlet removal: https://review.opendev.org/c/openstack/neutron/+/93776513:34
ralonsoh2) Broken with the removal of eventlet server code: https://review.opendev.org/c/openstack/neutron/+/95265913:34
ralonsohI don't know how to replace the API part, to be honest13:34
ralonsohand the OVS agent doesn't work if executed in a eventlet process13:34
ralonsohso, at this point and in order to progress, I would suggest to deactivate the fullstack CI13:35
ralonsohand try to fix API part, if possible13:35
lajoskatonayes I think I run into the same for the dhcp agent patch also13:35
ralonsoh^^ right13:35
ralonsohso I would suggest to migrate the jobs, as much as possible, to n-t-p13:36
ralonsohthe fact the n-t-p and tempest work with the eventlet-less patches (dhcp, ovs, api) means these patches are working13:36
ralonsohbut this fullstack CI is specifically tailor-made for eventlet13:37
lajoskatonahard topic.....13:37
ralonsohI know13:37
ralonsohI think many fullstack jobs can be implemented in n-t-p13:38
haleybralonsoh: ack, i guess we don't have a choice if we want to complete the work, but it will leave a gap13:38
lajoskatonafew weeks ago I mentioned that migrating to SDK also hard in fullstack as sdk has no no-auth mode.....13:38
haleybwhat do others think?13:38
ralonsohhaleyb, yes but we can close it using neutron-tempest-plugin13:38
ralonsohto be honest, n-t-p is more accurate kind of testing that fullstack13:38
lajoskatonalet's check  if we can migrate those tests to n-t-p  and go fo it, we gain simpler gating perhaps13:39
ralonsohI know that for some API tests it was lighter13:39
lajoskatonayeah in fullstack we did things like service restart, in pure tempest that is no possible13:39
haleybralonsoh: understood, it's just it will take time is all13:40
ralonsohhaleyb, right13:40
ralonsohlajoskatona, we can have this with tobiko, for example13:40
ralonsohif you don't mind (and if you agree with this, of course)13:41
ralonsohI'll send a mail informing about this13:41
ralonsohare you all OK?13:41
haleybi am Ok with it13:42
lajoskatonayes, from side fine, for learning tobiko that's an interesting thing :-)13:42
ralonsohok, I'll do it and I'll send a patch to remove fullstack13:42
haleyb)-;13:43
ralonsohlajoskatona, send me what problems you had with sdk too13:43
lajoskatonaI can also spend some time on migrating some of the tests13:43
ralonsohjust for documentation purposes in the mail13:43
ralonsohI'll create a LP bug related ^^13:43
lajoskatonaralonsoh: ack, I will go back to that problem to collect my fading memories13:43
ralonsohthanks!13:44
haleybalright, moving to last topic13:44
haleyb#topic on-demand13:44
haleybralonsoh: i have a follow-on fix for the metadata-agent LoopingCall code in 2025.1 - https://review.opendev.org/c/openstack/neutron/+/95306413:45
haleybi broke it with my pep8 change13:45
ralonsohi didn't see it, let me check13:45
ralonsohwhat was the original change?13:46
haleybhttps://review.opendev.org/c/openstack/neutron/+/95256113:46
haleybthere was a pep8 error that i just fixed, but should have tested it13:47
haleybpep8 was happy but code silently failed, and it has no tests...13:47
ralonsoh^ but this is my patch13:47
haleybralonsoh: yes, i shouldn't have fixed the pep8 issue in PS313:48
haleybyour code was fine, i broke it13:48
haleybwell, fine except for failing pep8 job13:49
ralonsoh+2, let's merge it asap13:50
haleybeither way it needs that change, and i proposed a similar change for all of neutron's calls to that Looping class to make sure they don't get broken13:50
haleybsometimes pylint is not your friend :(13:51
haleybthat's all i had, any other topics?13:51
haleybok, have a good week everyone!13:52
haleyb#endmeeting13:52
opendevmeetMeeting ended Tue Jun 24 13:52:15 2025 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)13:52
opendevmeetMinutes:        https://meetings.opendev.org/meetings/networking/2025/networking.2025-06-24-13.00.html13:52
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/networking/2025/networking.2025-06-24-13.00.txt13:52
opendevmeetLog:            https://meetings.opendev.org/meetings/networking/2025/networking.2025-06-24-13.00.log.html13:52
haleybo/13:52
ralonsohbye13:52
mlavalle\o13:52
cbuggyo/13:52
lajoskatonao/13:52
slaweqo/13:53
ykarelo/13:53
obondarevo/13:54
opendevreviewTakashi Kajinami proposed openstack/neutron-fwaas master: Deprecate unused [fwaas] agent_version  https://review.opendev.org/c/openstack/neutron-fwaas/+/95321214:41
opendevreviewTakashi Kajinami proposed openstack/neutron-fwaas master: Avoid unnecessary iniset_multiline  https://review.opendev.org/c/openstack/neutron-fwaas/+/95321514:51
danfaiHi, I got a question regarding host_routes when using segments (e.g. in routed provider networks). It seems we get automatic host_routes for every single subnet. For us this means 100 host_routes, which all point to the same default gateway IP. I was wondering whether this is a bit redundant to have it or what the intended reason was to apply the default gateway only after.15:00
danfaiAs the API already supports using a supernet route, I was adding the default route via gateway IP, which reduces the routes to two identical default gateway routes, which some DHCP clients/init scripts did not like either15:02
danfai(relevant method: neutron/services/segments/plugin.py:_calculate_routed_network_host_routes )15:04
opendevreviewLajos Katona proposed openstack/neutron master: Delete tunnel endpoints if endpoint agent is deleted  https://review.opendev.org/c/openstack/neutron/+/95298815:09
lajoskatona1haleyb, ralonsoh (all): I forgot to add this governance discussion to the eventlet topic: https://review.opendev.org/c/openstack/governance/+/95290315:21
ildikovhaleyb: mlavalle: Hi, I'm checking in about the OpenStack Ops radio Hour session I mentioned last week. Is either of you available to attend the session and talk about Neutron/networking related topics?15:51
opendevreviewDai Dang Van proposed openstack/neutron master: Add dns forwarder l2 extension  https://review.opendev.org/c/openstack/neutron/+/95139017:18
opendevreviewDai Dang Van proposed openstack/neutron master: Add dns forwarder l2 extension  https://review.opendev.org/c/openstack/neutron/+/95139017:19

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!