| opendevreview | Merged openstack/neutron master: metadata: Fix haproxy IPv6 bind address format https://review.opendev.org/c/openstack/neutron/+/997348 | 05:33 |
|---|---|---|
| opendevreview | Terry Wilson proposed openstack/neutron master: Use Neutron DB for OVN agent health https://review.opendev.org/c/openstack/neutron/+/818850 | 06:16 |
| ralonsoh | sean-k-mooney, ping me today to talk about what would be needed in Neutron during this release and the next one | 06:56 |
| opendevreview | Merged openstack/neutron-vpnaas master: ovn: Replace removed ``BaseEvent`` with ``row_event.RowEvent`` https://review.opendev.org/c/openstack/neutron-vpnaas/+/997387 | 07:49 |
| *** Jeff is now known as Guest13498 | 07:57 | |
| *** Guest13498 is now known as JeffYang | 08:00 | |
| opendevreview | Rodolfo Alonso proposed openstack/neutron-vpnaas master: db: Create VPN tables in start migration if missing https://review.opendev.org/c/openstack/neutron-vpnaas/+/997343 | 08:40 |
| sean-k-mooney | ralonsoh: so i think, and i say think, if i use the sriov nic agent it could work today. the problme is on the cyborg side this is only implemnted for intel nics so i will need to factor this fucntionatly into the pci driver refactor im planning for next cycle, if we wanted it to work for ovn i might jus tneeed to add it here | 08:59 |
| sean-k-mooney | https://github.com/openstack/neutron/blob/master/neutron/common/ovn/constants.py#L496-L497 and then just make sure the ovn mech driver treats portbindings.VNIC_ACCELERATOR_DIRECT, and portbindings.VNIC_DIRECT the same. by adding it to https://github.com/openstack/neutron/blob/master/neutron/common/ovn/constants.py#L463 EXTERNAL_PORT_TYPES | 08:59 |
| opendevreview | Merged openstack/neutron master: Save socket fd to be cleaned up in functional test https://review.opendev.org/c/openstack/neutron/+/997151 | 09:00 |
| sean-k-mooney | ralonsoh: i think my first step is oging to be trying to deploy the integration taht was done in xena | 09:00 |
| sean-k-mooney | and see if that works end to end following the spec | 09:01 |
| sean-k-mooney | but i proably wont get around to that untile august or september | 09:01 |
| ralonsoh | sean-k-mooney, how VNIC_ACCELERATOR_DIRECT works? Is that a HWOL OVS port? or a SR-IOV kind port? | 09:05 |
| sean-k-mooney | its just an sriov port no hardware offload | 09:05 |
| sean-k-mooney | its just that the vf comes form cyborg instead of nova | 09:06 |
| ralonsoh | ok, so that makes sense to have it in the sr-iov and ovn mech drivers | 09:06 |
| ralonsoh | in OVN, as external port | 09:06 |
| ralonsoh | and in SR-IOV as suppoted port, right? | 09:06 |
| sean-k-mooney | ya its alreayd htere in the sriov nic agent | 09:06 |
| sean-k-mooney | you added it https://github.com/openstack/neutron/commit/c33f0588cb0fdb52cf279fea68f827ce993658b8 | 09:06 |
| sean-k-mooney | i think that may have been before you joined redhat when you were sill at intel | 09:07 |
| ralonsoh | no no, that was in RH | 09:07 |
| sean-k-mooney | oh really | 09:07 |
| sean-k-mooney | i assumed this was related to the fortvil profile stuff | 09:07 |
| ralonsoh | I don't know why I participated in the LP bug, to be honest | 09:08 |
| ralonsoh | I don't remember the context | 09:08 |
| sean-k-mooney | no worries | 09:08 |
| sean-k-mooney | anyway the fact that the pluming seams to be there on the nova and nduton cide | 09:08 |
| sean-k-mooney | basiclly mean the ask is on me to test and document this end to end and set up ci | 09:08 |
| ralonsoh | if you have time to test it, we can do a follow-up of this LP bug and add this VNIC to ovn and sriov | 09:09 |
| sean-k-mooney | if i can do that we can thes discuss if/how to extend it | 09:09 |
| sean-k-mooney | ralonsoh: so until yesterday i tought there was part fo this work missing in nova | 09:10 |
| sean-k-mooney | but apprnetly not | 09:10 |
| sean-k-mooney | so ya what im going to try and do before the end of 2026.2 is test if there is a gap and document it | 09:10 |
| sean-k-mooney | and then next cycle, assuming pm ectra care about this figure out how to make this work for more usecses | 09:11 |
| ralonsoh | sean-k-mooney, open a LP bug for Neutron, referring to the old LP bug | 09:13 |
| ralonsoh | this is the best way to make the whole team aware. Next week we'll triage it in the Neutron team meeting | 09:13 |
| ralonsoh | and I'll take care of it | 09:13 |
| sean-k-mooney | well so far i dont know of work for ye to do in the sriov-nic-agent path, which shoudl be enouch for inital testing. why this is interesting is newer nvidia gpus have a pcie bridge per gpu that has a Connectx 7 or connectx-8 nic per gpus in principal this fucntionalty would allow you to request a the assignemt fo both the nic and the gpu or a partion of both i.e. a VF form | 09:15 |
| sean-k-mooney | the nic and a vgpu slice form the gpu. | 09:15 |
| sean-k-mooney | if this still work for that usecase that actully a net new thing that nova cant already do better | 09:17 |
| sean-k-mooney | and well cyborg does not have a lot of those at the moement so every one helps :) | 09:17 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron master: ovn: Allow ``get_datapath_type`` to look up Chassis by UUID https://review.opendev.org/c/openstack/neutron/+/995296 | 11:47 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron master: ovn: Wait for ovn-controller OpenFlow rules during live migration https://review.opendev.org/c/openstack/neutron/+/994296 | 11:47 |
| cardoe | ralonsoh: if you have any suggestions on https://bugs.launchpad.net/neutron/+bug/2160719 that change to explicitly call the check on get_flavor policy was me just trying to let Zuul go and convince me I wasn't crazy | 12:48 |
| ralonsoh | cardoe, the patch https://review.opendev.org/c/openstack/neutron/+/997444 does nothing but declare again the same method that is already inherited by the upper class | 12:54 |
| cardoe | Yes I'm aware. | 13:02 |
| cardoe | I was just using Zuul to make sure I wasn't going crazy at the end of my day. | 13:02 |
| cardoe | I expected it to do nothing and when I got back to my desk it had passed. | 13:02 |
| cardoe | I'm not seeing what's causing that issue. | 13:03 |
| cardoe | Any ideas are welcome. | 13:09 |
| ralonsoh | cardoe, sorry, I'll check that tomorrow. Now I have a customer call | 13:10 |
| opendevreview | alisafari proposed openstack/neutron stable/2026.1: bgp: Always assume just one provider switch https://review.opendev.org/c/openstack/neutron/+/997550 | 13:13 |
| opendevreview | alisafari proposed openstack/neutron stable/2026.1: bgp: Create new InterconnectBridge https://review.opendev.org/c/openstack/neutron/+/997551 | 13:15 |
| opendevreview | alisafari proposed openstack/neutron stable/2026.1: bgp: Implement peer-to-peer flows on Interconnect Bridge https://review.opendev.org/c/openstack/neutron/+/997552 | 13:15 |
| opendevreview | Doug Goldstein proposed openstack/neutron master: validate get_flavor working for role:reader users https://review.opendev.org/c/openstack/neutron/+/997444 | 14:23 |
| opendevreview | Terry Wilson proposed openstack/neutron master: Use Neutron DB for OVN agent health https://review.opendev.org/c/openstack/neutron/+/818850 | 15:45 |
| opendevreview | Terry Wilson proposed openstack/neutron master: Use Neutron DB for OVN agent health https://review.opendev.org/c/openstack/neutron/+/818850 | 17:20 |
| opendevreview | Terry Wilson proposed openstack/neutron master: Use Neutron DB for OVN agent health https://review.opendev.org/c/openstack/neutron/+/818850 | 18:46 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!