| opendevreview | Kyuyeong Lee proposed openstack/neutron stable/2026.1: [OVN] Use poll(2) instead of select(2) for the OVSDB poller https://review.opendev.org/c/openstack/neutron/+/1008891 | 00:51 |
|---|---|---|
| opendevreview | Kyuyeong Lee proposed openstack/neutron stable/2026.1: [OVN] Use poll(2) instead of select(2) for the OVSDB poller https://review.opendev.org/c/openstack/neutron/+/1008891 | 00:58 |
| opendevreview | Merged openstack/neutron master: [OVN] Retrieve only the SG stateful flag in the logapi driver https://review.opendev.org/c/openstack/neutron/+/1008419 | 02:03 |
| opendevreview | OpenStack Proposal Bot proposed openstack/neutron master: Imported Translations from Zanata https://review.opendev.org/c/openstack/neutron/+/1008902 | 03:43 |
| opendevreview | Kyuyeong Lee proposed openstack/neutron stable/2026.2: [OVN] Retrieve only the SG stateful flag in the logapi driver https://review.opendev.org/c/openstack/neutron/+/1008911 | 04:27 |
| opendevreview | Kyuyeong Lee proposed openstack/neutron stable/2026.1: [OVN] Retrieve only the SG stateful flag in the logapi driver https://review.opendev.org/c/openstack/neutron/+/1008912 | 04:27 |
| opendevreview | Merged openstack/neutron stable/2026.1: Use idl.has_lock instead of idl.is_lock_contended for OVSDB lock checks https://review.opendev.org/c/openstack/neutron/+/1008759 | 05:44 |
| ralonsoh | lajoskatona, hi! if you have 1 min, please check https://review.opendev.org/c/openstack/neutron-lib/+/1006916 | 07:21 |
| ralonsoh | thanks! | 07:21 |
| ralonsoh | and a backport, no conflicts: https://review.opendev.org/c/openstack/neutron/+/1008758 | 07:21 |
| lajoskatona | ralonsoh: slaweq was faster :-) | 07:28 |
| slaweq | :) | 07:29 |
| opendevreview | Eduardo Olivares proposed openstack/neutron master: [OVN] Allow router interface ICMP on stateful security groups https://review.opendev.org/c/openstack/neutron/+/1008488 | 07:37 |
| *** amorin_ is now known as amorin | 07:45 | |
| ralonsoh | lajoskatona, slaweq thanks folks! | 07:45 |
| Mike-- | Question related to ovn load balancer health checks for backends. They run on the chassis/node that hosts the backend. If that chassis/host dies the backend stays up potentially forever if the host never comes back. Is this a known limitation? | 07:54 |
| ralonsoh | froyo, ^ can you check that? | 08:07 |
| froyo | sure! | 08:07 |
| froyo | Mike-- that is under the core ovn umbrella, ovn-octavia-provider doesn't implement the check itself, it is only receiving events from Service_Monitor table in OVN SB DB. But I agree on your conclusion, if the chassis where monitor es pinned dissapear looks there isn't any logic there to reassign it and last status of the backend will remain in the Service_monitor entry table. | 08:13 |
| froyo | I will check ovn doc trying to find if this is a known limitation or something is written there about this case, I will keep you update | 08:15 |
| Mike-- | froyo: thanks! most of the other parts on a chassis will get moved if the 'vm' moves (masakari / nova evacuate / manual) but this caught our eye. | 08:32 |
| Mike-- | this holds true for the backend as well once the vm moves. But the scenario the vm never moves there with be a % blackhole on the load balancer | 08:34 |
| *** Guest1419 is now known as ravlew | 08:36 | |
| opendevreview | Rodolfo Alonso proposed openstack/neutron master: ml2: replace tunnel segment physical_network NULL with '' https://review.opendev.org/c/openstack/neutron/+/1007844 | 08:36 |
| opendevreview | Merged openstack/neutron master: Drop ``sys_platform`` from requirements https://review.opendev.org/c/openstack/neutron/+/999595 | 09:01 |
| froyo | Mike-- Exactly — makes sense, since evacuation/masakari rebinding the logical port is what indirectly "fixes" this for every other case (health monitor gets re-created on the new chassis once the port moves). The gap is narrowly the stuck-VM-never-moves scenario: no fencing/evacuation trigger, or a backend that's never Nova-managed at all. In that case the LB has no way to know, and keeps routing a steady % o | 09:02 |
| froyo | f traffic (proportional to the backend's weight) into a blackhole indefinitely — it's a silent partial-outage, not a hard failure anyone gets paged for. I just confirmed it's a known class of issue: OVN upstream is currently fixing the exact same pattern for BFD status (ovn-org/ovn PR #321/#325, issue #320 — "NB/SB status stays 'up' forever nothing is left to update"), but that work explicitly scopes out | 09:02 |
| froyo | the hard-crashed-chassis case (left to an opt-in Neutron policy) and doesn't touch Service_Monitor/LB health checks at all. | 09:02 |
| opendevreview | Merged openstack/neutron-lib master: bgp: Add ``is_filter`` to API definition https://review.opendev.org/c/openstack/neutron-lib/+/1006916 | 09:12 |
| opendevreview | Merged openstack/neutron stable/2026.2: ovn: fix gateway scheduling for multi-segment networks https://review.opendev.org/c/openstack/neutron/+/1008758 | 09:16 |
| Mike-- | froyo: I did not find those ovn references (yet), thanks! if I read your messages correctly it would still require neutron level work? Does that make sense from an architecture level? | 09:25 |
| froyo | Mike-- Refs: | 09:26 |
| froyo | - ovn-sb(5): https://man7.org/linux/man-pages/man5/ovn-sb.5.html | 09:26 |
| froyo | - https://github.com/ovn-org/ovn/pull/321 | 09:26 |
| froyo | - https://github.com/ovn-org/ovn/pull/325 | 09:26 |
| Mike-- | read the github issue now, interesting read! Shared it with the team internally just now. | 09:31 |
| froyo | Mike-- So architecturally, yes — it needs a Neutron-side "confirmed dead" signal, with OVN/northd as the consumer that translates that signal into DB state (clear chassis_name, force status down). That's exactly the shape of the BFD fix: Neutron confirms silence → removes the chassis from scoped HA groups → northd reacts. The catch for your scenario: that signal doesn't really e in the Concept C work, sinc | 09:32 |
| froyo | exist today, even for Neutron's own purposes. Checked a related bug (RH BZ #1946179) — Neutron has no automatic dead-chassis detection at all right now. A stale chassis just sits in the SB DB until an operator manually runs openstack network agent delete for that host. | 09:32 |
| froyo | So IMO two things would need to happen: | 09:34 |
| froyo | 1. Neutron would need an automated "chassis confirmed dead" signal (today it's a fully manual operator step). | 09:34 |
| froyo | 2. OVN/northd would need a Service_Monitor-side consumer of that signal — which doesn't exist yet even in the Concept C work, since that's currently scoped only to BFD-tied HA chassis groups (gateway ports), not generic VIF-bound ports where LB members live. | 09:34 |
| Mike-- | Yes we foud the same state, although not that elaborately put. In our discussions, with our current knowledge of openstack services, we think the best course of action is to add a masakari flow to do said "openstack network agent delete" for the host masakari flags as down | 09:35 |
| Mike-- | reading your #2 our masakari route would not solve it either or am I misreading | 09:36 |
| Mike-- | ? | 09:36 |
| opendevreview | Merged openstack/ovsdbapp master: ovn-sb: Set default ``hostname`` in ``chassis_add`` https://review.opendev.org/c/openstack/ovsdbapp/+/1005207 | 09:54 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron master: Fix netlink test privsep entrypoint unwrapping https://review.opendev.org/c/openstack/neutron/+/1009010 | 09:56 |
| ralonsoh | ykarel, ^ | 09:56 |
| ralonsoh | I've executed `check experimental` | 09:57 |
| froyo | Mike-- Correction to my earlier point #2 — dug further and found this isn't as unsolved as I said. OVN northd has had a mechanism since 2022 (commit 23e203a3, "northd: set svc_mon status to offline if port_binding released", shipped in OVN v22.09.0) that forces a backend's Service_Monitor.status to offline when its Port_Binding.up goes false. But that fix had a gap: if up stays true while chassis gets cleared, th | 09:57 |
| froyo | e stale online status was still trusted forever — which is basically our exact scenario. | 09:57 |
| ykarel | thx ralonsoh | 09:57 |
| froyo | Mike-- Port_Binding.chassis is a weak reference to Chassis in the schema — so deleting the Chassis row automatically clears it at the DB level. With this new northd fix in place, that triggers northd to mark the Service_Monitor offline and clear chassis_name, which is exactly the outcome you want. | 09:59 |
| Mike-- | aah cool, then our perceived masakari route should work. That leaves the point if neutron wants to own some 'monitoring' or 'liveness' check for a chassis? | 10:00 |
| froyo | Mike-- commit e180d57 ("northd: Mark unbound ports' service monitors offline"), merged into ovn-org/ovn main on 2026-09-29 | 10:00 |
| Mike-- | nice nice | 10:01 |
| opendevreview | Fiorella Yanac proposed openstack/neutron-tempest-plugin master: Add additional Json Tempest api tests for indirect floating IP https://review.opendev.org/c/openstack/neutron-tempest-plugin/+/1009017 | 10:45 |
| opendevreview | Lajos Katona proposed openstack/neutron-vpnaas master: Rally scenario plugin API changed https://review.opendev.org/c/openstack/neutron-vpnaas/+/1000417 | 10:54 |
| opendevreview | Lajos Katona proposed openstack/neutron-vpnaas master: Fix IPsec device driver status reporting races https://review.opendev.org/c/openstack/neutron-vpnaas/+/1009019 | 10:54 |
| opendevreview | Merged openstack/ovn-octavia-provider master: Migrate setup configuration to pyproject.toml https://review.opendev.org/c/openstack/ovn-octavia-provider/+/1003948 | 10:58 |
| lajoskatona | slaweq, ralonsoh, mlavalle: do we have meeting today as haleyb is not around? For me it is fine to say that let's have everybody add topics to the etherpad and lets meet next week | 10:59 |
| slaweq | lajoskatona: personally I don't have any topics for today, we can probably wait week to discuss things during PTG next week | 11:00 |
| opendevreview | Elod Illes proposed openstack/networking-bagpipe stable/2026.1: DNM: gate health test https://review.opendev.org/c/openstack/networking-bagpipe/+/1009024 | 11:02 |
| lajoskatona | slaweq: thanks | 11:04 |
| opendevreview | Elod Illes proposed openstack/networking-bgpvpn stable/2026.1: DNM: gate health test https://review.opendev.org/c/openstack/networking-bgpvpn/+/1009028 | 11:09 |
| opendevreview | Merged openstack/ovn-octavia-provider master: Add --octavia-config-file to the neuron-ovn-db-sync plugin https://review.opendev.org/c/openstack/ovn-octavia-provider/+/999736 | 11:15 |
| ralonsoh | lajoskatona, slaweq good for me, we can skip todays meeting | 11:19 |
| ralonsoh | I've been adding some topics to the PTG etherpad | 11:19 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron-fwaas master: fwaas: skip L2 status when no local ports https://review.opendev.org/c/openstack/neutron-fwaas/+/1002590 | 12:44 |
| mlavalle | slaweq, ralonsoh no meeting today, right? | 13:02 |
| ralonsoh | no, we commented that before | 13:03 |
| ralonsoh | if there is any hot topic, we can discuss it here | 13:03 |
| ralonsoh | any other topic can be added to the PTG etherpad | 13:03 |
| ralonsoh | (let me find the link) | 13:03 |
| ralonsoh | #link https://etherpad.opendev.org/p/oct2026-ptg-neutron | 13:03 |
| mlavalle | ralonsoh, thanks! | 13:03 |
| mlavalle | have a nice day | 13:03 |
| slaweq | have a good one mlavalle :) | 13:04 |
| ralonsoh | So, for everybody, if you have something to discuss, you can raise any topic here in this channel. For PTG topics, use the ^^ provided link | 13:04 |
| ichen | Please review EVPN Type-2 route spec https://review.opendev.org/c/openstack/neutron-specs/+/1004908 . Thanks in advance. | 13:04 |
| * haleyb hopes someone got his email about missing meeting | 13:14 | |
| opendevreview | Merged openstack/neutron stable/2025.2: qos: Handle missing port binding in resource request https://review.opendev.org/c/openstack/neutron/+/1005520 | 13:17 |
| opendevreview | Fiorella Yanac proposed openstack/neutron-tempest-plugin master: Add additional Json Tempest api tests for indirect floating IP https://review.opendev.org/c/openstack/neutron-tempest-plugin/+/1009017 | 13:28 |
| *** elodille1 is now known as elodilles | 14:23 | |
| opendevreview | Rodolfo Alonso proposed openstack/neutron master: ml2: replace tunnel segment physical_network NULL with '' https://review.opendev.org/c/openstack/neutron/+/1007844 | 14:46 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron master: ml2: derive segment IDs from ``networksegments`` https://review.opendev.org/c/openstack/neutron/+/1007849 | 14:46 |
| opendevreview | Lajos Katona proposed openstack/neutron-vpnaas master: Rally scenario plugin API changed https://review.opendev.org/c/openstack/neutron-vpnaas/+/1000417 | 14:57 |
| opendevreview | Merged openstack/neutron stable/2026.1: [OVN] Use poll(2) instead of select(2) for the OVSDB poller https://review.opendev.org/c/openstack/neutron/+/1008891 | 15:23 |
| opendevreview | Merged openstack/neutron stable/2026.2: [OVN] Retrieve only the SG stateful flag in the logapi driver https://review.opendev.org/c/openstack/neutron/+/1008911 | 15:23 |
| opendevreview | Merged openstack/neutron master: Imported Translations from Zanata https://review.opendev.org/c/openstack/neutron/+/1008902 | 15:23 |
| opendevreview | Merged openstack/neutron stable/2026.1: [OVN] Retrieve only the SG stateful flag in the logapi driver https://review.opendev.org/c/openstack/neutron/+/1008912 | 15:23 |
| opendevreview | Mathieu GRZYBEK proposed openstack/neutron master: Restrict ``delete_port`` for the OVN metadata port https://review.opendev.org/c/openstack/neutron/+/1008762 | 15:38 |
| opendevreview | Marcin Wilk proposed openstack/neutron stable/2025.2: Use idl.has_lock instead of idl.is_lock_contended for OVSDB lock checks https://review.opendev.org/c/openstack/neutron/+/1009080 | 16:05 |
| opendevreview | Merged openstack/neutron master: ml2: Stop binding when a mechanism driver reports failure https://review.opendev.org/c/openstack/neutron/+/985732 | 16:06 |
| opendevreview | Lajos Katona proposed openstack/neutron-vpnaas master: Rally scenario plugin API changed https://review.opendev.org/c/openstack/neutron-vpnaas/+/1000417 | 16:07 |
| opendevreview | Maor Blaustein proposed x/whitebox-neutron-tempest-plugin master: Add EVPN-BGP FRR gateway HA failover test https://review.opendev.org/c/x/whitebox-neutron-tempest-plugin/+/1009097 | 17:15 |
| opendevreview | Maor Blaustein proposed x/whitebox-neutron-tempest-plugin master: Add EVPN-BGP FRR gateway HA failover test https://review.opendev.org/c/x/whitebox-neutron-tempest-plugin/+/1009097 | 17:31 |
| opendevreview | Merged openstack/neutron stable/2025.2: ovn: Add ip_substring_filtering API extension support https://review.opendev.org/c/openstack/neutron/+/1006667 | 18:29 |
| opendevreview | Brian Haley proposed openstack/neutron master: Add explicit return value to _bind_port_level() https://review.opendev.org/c/openstack/neutron/+/1009130 | 23:50 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!