| opendevreview | Oleksandr Kozachenko proposed openstack/openstack-helm master: octavia: add helm hook annotations in jobs https://review.opendev.org/c/openstack/openstack-helm/+/965887 | 02:36 |
|---|---|---|
| cardoe | ricolin_: hey can we talk about your OVN change for Octavia? | 15:54 |
| cardoe | mnaser: I see that ricolin_ is marked as away. mind poking him for me? | 15:55 |
| cardoe | I posted it on the OSH Slack | 15:55 |
| cardoe | But I'm looking at https://github.com/openstack/openstack-helm/commit/d617712e689585800a5e85b67fd53d5ee58d5fe2 and I have ZERO clue how that's suppose to work. | 15:55 |
| cardoe | I looked at your guys https://github.com/vexxhost/atmosphere and I see you don't even run that patch | 15:55 |
| cardoe | The octavia-api expects that its gonna talk to the driver agent via a UNIX socket but now that you moved it to a separate deployment and didn't share that directory its not possible. | 15:56 |
| mnaser | Since he is based in Taiwan I dont think hell get a chance to answer soon :p | 15:56 |
| mnaser | I think there's an assumption that octavia api runs on thr same node as driver agent.. usually | 15:57 |
| cardoe | Yep. Except both deployments mount it as an emptyDir and not a host path | 15:58 |
| cardoe | I could get behind the host path thing or a PVC | 15:58 |
| mnaser | Id argue it should be a sidecar at that point in octavia api | 15:59 |
| cardoe | That's how we implemented it locally. | 15:59 |
| mnaser | But it seems that indeed maybe this didnt work properly | 16:00 |
| mnaser | We got excited about ovn driver and then | 16:00 |
| cardoe | Before I submit a patch to change it around entirely I just wanted Rico's feedback on how he wanted it to go. | 16:00 |
| mnaser | The amount of pitfalls and things that broke in terms of accessibility paths was a huge headache. | 16:00 |
| cardoe | ah okay. So it's not actually in use for you guys? | 16:00 |
| mnaser | We have a whole variety of edge case bugs for accessibility | 16:00 |
| cardoe | So if I tinker some more it won't break you. | 16:01 |
| mnaser | Heck we had to implement this https://bugs.launchpad.net/neutron/+bug/2045415 | 16:01 |
| mnaser | https://bugs.launchpad.net/neutron/+bug/2110225 | 16:01 |
| cardoe | oh yes I very much understand. | 16:02 |
| cardoe | I spent HOURS trying to make the neutron folks understand how VXLAN works. | 16:02 |
| cardoe | Trying to land https://cardoe.com/neutron/evpn-vxlan-network/admin/data-center-networks.html in upstream neutron docs. | 16:03 |
| cardoe | We're using Ironic and Neutron driving a VXLAN overlay in the DCs and need to bind those to ports on physical servers. | 16:03 |
| cardoe | But good to know that rico is in Taiwan so that I know his timezone | 16:05 |
| opendevreview | Nicholas Kuechler proposed openstack/openstack-helm master: fix(octavia): Fixes typos for driver agent mounts in helm templates https://review.opendev.org/c/openstack/openstack-helm/+/965970 | 19:13 |
| opendevreview | Merged openstack/openstack-helm-images master: Remove 2024.1 jobs https://review.opendev.org/c/openstack/openstack-helm-images/+/965834 | 20:05 |
| opendevreview | Vladimir Kozhukalov proposed openstack/openstack-helm master: [keystone] Use keystone native wsgi script https://review.opendev.org/c/openstack/openstack-helm/+/965316 | 22:35 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!