| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-02-09 | |||
| 09:04:37 | bauzas | I assume we serve a 200 with an empty list | |
| 09:04:54 | bauzas | https://docs.openstack.org/api-ref/network/v2/index.html?expanded=list-subnets-detail,show-port-details-detail,show-subnet-details-detail,list-segments-detail#list-segments | |
| 09:05:17 | bauzas | it's just a filtering | |
| 09:05:21 | bauzas | filtering* | |
| 09:05:45 | bauzas | anyway, will upload and we'll see | |
| 09:05:50 | sean-k-mooney | the only error code it can return is 401 which is unautrised | |
| 09:05:56 | sean-k-mooney | so ya 200 with empty list | |
| 09:06:02 | bauzas | (saying this, because I'll stick with LOG.error on something else but a 200) | |
| 09:06:14 | bauzas | on the neutron method | |
| 09:06:26 | sean-k-mooney | https://docs.openstack.org/api-ref/network/v2/index.html?expanded=list-segments-detail#list-segments | |
| 09:06:43 | sean-k-mooney | looking at the doc its either a 200 or 401 and a 401 would not make sense here | |
| 09:07:36 | bauzas | sure | |
| 09:07:55 | bauzas | either way, will upload the patch once I'm done with fixing tests | |
| 09:10:19 | sean-k-mooney | yep its a list https://github.com/openstack/neutron/blob/master/neutron/services/segments/db.py#L167-L169 | |
| 09:22:54 | kashyap | Is it just me, or the (relatively) new Gerrit is much slower? | |
| 09:23:33 | kashyap | I've noticed it over the last couple of weeks -- when I hit "Send" on a comment, it takes more than 4-5 seconds to commit | |
| 09:23:45 | kashyap | s/commit/submit/ | |
| 09:24:05 | sean-k-mooney | its slower | |
| 09:24:22 | sean-k-mooney | its to do with storing the data in git apparently | |
| 09:25:10 | sean-k-mooney | its proably more the case with nova then other repos | |
| 09:25:21 | sean-k-mooney | since nova git repo is much bigger | |
| 09:40:40 | sean-k-mooney | we dont have the pci whitelist tags stored in the nova db... only in memory in the compute node | |
| 09:42:20 | kashyap | sean-k-mooney: I see, noted | |
| 09:42:27 | sean-k-mooney | i tought they were stored in extra_info ? did i imaging that? | |
| 09:48:00 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Add net and utils methods for getting routed netks and segments https://review.opendev.org/c/openstack/nova/+/773976 | |
| 09:48:10 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Add a routed networks scheduler pre-filter https://review.opendev.org/c/openstack/nova/+/749068 | |
| 09:50:13 | songwenping_ | sean-k-mooney: morning, i have manually set the [vnc]/auth_schema=["vnc"] and created a vm with a console without a password and it works in our product, but i have no idea about the tempest to test, would you give some specific guide, please? | |
| 09:51:42 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Add net & utils methods for routed nets & segments https://review.opendev.org/c/openstack/nova/+/773976 | |
| 09:51:52 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Add a routed networks scheduler pre-filter https://review.opendev.org/c/openstack/nova/+/749068 | |
| 09:51:56 | bauzas | gibi: uploaded the last revision, with just a cosmetic change on the fixtures (using the uuid sentinelà + setattr usage instead of copy/pasting 5 times compute create | |
| 09:54:39 | sean-k-mooney | songwenping_: if you can confirm it works manually i think that is fine for now | |
| 09:55:00 | sean-k-mooney | songwenping_: for the tempest test we woudl just set it in the zuul job we can do that as a follow up change | |
| 09:55:17 | songwenping_ | cool. | |
| 09:55:45 | sean-k-mooney | once the irc logs catch up ill link to this in a comment and change too a +1 once i re review | |
| 09:55:56 | sean-k-mooney | i think you had areassed my other comments however | |
| 09:56:29 | songwenping_ | thanks :) | |
| 09:57:29 | sean-k-mooney | bauzas: you didn tfix the exception in the case of networks | |
| 09:57:43 | sean-k-mooney | bauzas: https://review.opendev.org/c/openstack/nova/+/773976/5/nova/network/neutron.py#3510 | |
| 09:57:58 | sean-k-mooney | oh actully | |
| 09:58:27 | sean-k-mooney | will if resp: be true or false | |
| 09:58:34 | sean-k-mooney | i guess it will eveauate to true | |
| 09:58:49 | sean-k-mooney | even if it gets a 200 with an empty list | |
| 09:58:54 | sean-k-mooney | so ya that should be fine | |
| 09:59:56 | sean-k-mooney | bauzas: ah you have a test for it too cool https://review.opendev.org/c/openstack/nova/+/773976/5/nova/tests/unit/network/test_neutron.py#6293 | |
| 10:02:05 | bauzas | sean-k-mooney: yup | |
| 10:02:20 | bauzas | like I said earlier, I left the exception raise for networks | |
| 10:02:41 | bauzas | as this would be an unknown reason | |
| 10:02:45 | sean-k-mooney | bauzas: ya i got confused because i needed to look at v3 vs v5 | |
| 10:02:55 | bauzas | np | |
| 10:09:33 | sean-k-mooney | im not a huge fan of setattr but im +1 on both of those changes | |
| 10:10:37 | stephenfin | another "Multiple possible networks found" gate failure :-( | |
| 10:10:51 | stephenfin | will this compact DB migration series ever land | |
| 10:11:16 | sean-k-mooney | right before we swap to etcd :P | |
| 10:12:11 | sean-k-mooney | stephenfin: on a side note the pci filter is massivly less efficent then i tought it was. if i fix that is it a bug or a feature | |
| 10:12:24 | stephenfin | depends on the inefficiency | |
| 10:12:35 | stephenfin | and impact on end users | |
| 10:12:52 | sean-k-mooney | we dont store the physnet or other tags in the db just in memroy. i tought they were in the extra info column | |
| 10:12:54 | sean-k-mooney | but no | |
| 10:13:23 | sean-k-mooney | so when we are doing neutron sriov the filter does not know fi the VF is a nic, gpu, qat device | |
| 10:13:33 | sean-k-mooney | it just know its a vf | |
| 10:13:46 | sean-k-mooney | we only check that on the compute node | |
| 10:14:14 | sean-k-mooney | i could jut stick that in the extra info json blob and then we could check it in the filter | |
| 10:16:10 | sean-k-mooney | or we coudl track them in plamcnet next cycle but i kind of want to fix this for everyone not running master | |
| 10:17:39 | stephenfin | I'd say push the patch and let's see in the review what everyone else thinks | |
| 10:18:25 | sean-k-mooney | ya ok. not today though i was checking the db for my neutron port numa poicy seriese | |
| 10:18:56 | sean-k-mooney | just noticed this wehn i was looking at the pci devices table | |
| 10:24:30 | sean-k-mooney | the pci request has {"count": 1, "spec": [{"physical_network":"public"}] we have nothing to match that against in the pci device table so really that is all that is missing | |
| 10:28:31 | gibi | bauzas: ack, I will check | |
| 10:46:38 | sean-k-mooney | you know when you see something odd in nova and then you investiate and all you find its a world full of sad pandas | |
| 10:47:09 | sean-k-mooney | that is what the test coverage for this piece of the pci tracker is like | |
| 10:47:11 | sean-k-mooney | https://github.com/openstack/nova/blob/4827a90a02ef91e777b1fcaa946c20c54dd0a29a/nova/tests/unit/pci/test_stats.py#L164-L168 | |
| 10:48:30 | lyarwood | Sad Pandas, all the way down. | |
| 10:49:27 | stephenfin | gibi: It is not at all a priority, but while it's fresh in your mind I've addressed all but one of your comments on https://review.opendev.org/c/openstack/nova/+/676209 and am researching the final one now | |
| 10:49:47 | sean-k-mooney | lyarwood: indeed | |
| 11:12:15 | gibi | stephenfin: ack, will check this afternoon | |
| 11:15:36 | bauzas | for us, we say "I'm a sad clown" | |
| 11:29:57 | sean-k-mooney | bauzas: because pandas are cute so when they are sad its really bad. also its a meme | |
| 11:38:00 | sean-k-mooney | bauzas: sad panda https://i.kym-cdn.com/photos/images/newsfeed/000/092/706/Sad_Panda_Chibi_by_mongrelssister.png | |
| 11:39:30 | sean-k-mooney | happy panda https://i.pinimg.com/originals/99/2a/35/992a358dcff70ca9bfcb4b227c08333c.png | |
| 11:39:47 | sean-k-mooney | now back to unit test land... | |
| 11:42:18 | lyarwood | bauzas: also helpful reminder, sean-k-mooney isn't English, stephenfin is but that's another story. ./me runs | |
| 11:42:33 | sean-k-mooney | haha | |
| 11:44:30 | gibi | lol | |
| 11:45:45 | gibi | I feel there is an intersting story behind this | |
| 11:49:24 | sean-k-mooney | gibi: stephenfin lived in essex for a bit when he was young before moving back to live in clare | |
| 11:49:35 | lyarwood | and loves being called English | |
| 11:49:45 | lyarwood | /s | |
| 11:49:55 | gibi | :D | |
| 11:49:57 | sean-k-mooney | gibi: but calling an irish person english is like calling lyarwood welsh | |
| 11:50:31 | lyarwood | well tbh it's likely worse given recent history | |
| 11:50:47 | lyarwood | anyway | |
| 11:59:32 | gibi | bauzas: I'm +2 on the route net series | |
| 12:08:54 | sean-k-mooney | stephenfin: ok so the tags are not store in the pci_devices table but they are in the PciDevicePoolList in teh compute nodes table that is used to populate the host_state object | |
| 12:09:08 | sean-k-mooney | stephenfin: so its not a performcne issue | |
| 12:09:32 | sean-k-mooney | i was just expecting it to be in both | |
| 12:12:52 | kashyap | lyarwood: Reading back ... I parsed bauzas' "english speaking" as "anglosphere" | |
| 12:14:11 | lyarwood | thank you captain obvious ;) | |
| 12:15:21 | sean-k-mooney | kashyap: so did we :) but there was an opertunity for a joke to be made | |
| 12:16:23 | lyarwood | and as a new dad I need the practice | |
| 12:17:50 | kashyap | sean-k-mooney: Hehe, figured as much; don't want to rob you of a good tease | |