| *** haleyb_ is now known as haleyb | 13:30 | |
| opendevreview | Merged openstack/nova stable/yoga: Ironic: retry when node not available https://review.opendev.org/c/openstack/nova/+/868010 | 14:59 |
|---|---|---|
| melwitt | dansmith: fyi this is a fix for our subclass signature checker test that detects when volume drivers are not matching the base class https://review.opendev.org/c/openstack/nova/+/883217 I found it wasn't working when I was working on the cve stuff. it would have caught the issue with the wallaby patch | 17:56 |
| dansmith | melwitt: cool | 17:57 |
| sean-k-mooney | melwitt: i didnt know that was a thing | 18:21 |
| sean-k-mooney | i feel like there are proably better ways to detech that if we use mypy | 18:22 |
| sean-k-mooney | to do type checking now | 18:22 |
| sean-k-mooney | with that said im ok with fixing this as is | 18:23 |
| sean-k-mooney | we can likel just do it better with newer tools now | 18:23 |
| melwitt | sean-k-mooney: yeah it was new-ish to me too as I hadn't dug into the volume drivers too much before | 18:37 |
| melwitt | and yeah maybe there's a better way to do it now. the fix was pretty easy tho | 18:37 |
| sean-k-mooney | i havent looked into mypy too much but i think we caoudl annotation the base calses and have it detech if the signirues did not match | 18:40 |
| sean-k-mooney | but that would be a large change anyway | 18:40 |
| sean-k-mooney | espcially since we woudl ahve to enable mypi typeing for the entire file | 18:40 |
| dansmith | please no | 18:49 |
| sean-k-mooney | in the cld classes i think you jsut have to put @overload in the metods so its not that bad | 18:51 |
| sean-k-mooney | but its more work then we need | 18:51 |
| sean-k-mooney | if you declar it an overlaod/override and the signiture does not match it treats it as an error | 18:52 |
| dansmith | do you mean abc? | 18:52 |
| sean-k-mooney | basicaly ya | 18:53 |
| sean-k-mooney | that was how it sued to be done | 18:53 |
| dansmith | yeah, that's the better way, especially for things like this | 18:53 |
| sean-k-mooney | it might still be the way its currently doen | 18:53 |
| sean-k-mooney | i was looking at https://mypy.readthedocs.io/en/stable/class_basics.html#abstract-base-classes-and-multiple-inheritance vs https://mypy.readthedocs.io/en/stable/final_attrs.html#final-methods | 18:54 |
| sean-k-mooney | well https://mypy.readthedocs.io/en/stable/error_code_list.html?highlight=overload#check-calls-to-overloaded-functions-call-overload | 18:54 |
| sean-k-mooney | so there is some checking for overrides in bases https://mypy.readthedocs.io/en/stable/error_code_list.html?highlight=overload#check-validity-of-overrides-override but abstractg base classes is clearer | 18:56 |
| sean-k-mooney | i think the automtaic checking only applies if the types dont match | 18:56 |
| sean-k-mooney | anyway o/ | 18:57 |
| gouthamr | o/ dansmith: do you have logs of a multinode job that ran with changes here: https://review.opendev.org/c/openstack/devstack-plugin-ceph/+/882483 | 19:05 |
| dansmith | gouthamr: meaning other than the ones linked there? | 19:06 |
| dansmith | oh no, right, | 19:06 |
| dansmith | it wasn't running the multinode job | 19:06 |
| dansmith | gouthamr: but no, that was just me trying to sync up that job definition, I didn't get into any actual diagnosis of if the job was working | 19:07 |
| gouthamr | ah thanks | 19:07 |
| gouthamr | saw your comment: "There are errors from one of the nova computes that indicate that it's not getting access to the ceph cluster properly, which means there is more work to do here." | 19:08 |
| gouthamr | so assumed that was on the CI somewhere; ashrodri can enable the base job in the repo with her changes and debug that | 19:09 |
| dansmith | gouthamr: oh, that's probably from the nova test patch, which likely has some logs still because it has a multinode job | 19:09 |
| dansmith | hang on | 19:10 |
| dansmith | gouthamr: https://zuul.opendev.org/t/openstack/build/588de1c62e564a4e93d4e94d1e9a5216 | 19:10 |
| dansmith | yeah so all the migration tests fail which I assume is because the other node can't access the ceph cluster | 19:11 |
| dansmith | (sorry I had purged all this from my cache) | 19:11 |
| dansmith | ceph conf and keyrings are there on the subnode | 19:12 |
| dansmith | but nova can't see it: https://zuul.opendev.org/t/openstack/build/588de1c62e564a4e93d4e94d1e9a5216/log/compute1/logs/screen-n-cpu.txt#8038 | 19:13 |
| gouthamr | dansmith: it looks like we created two ceph clusters there :) | 19:18 |
| gouthamr | with REMOTE_CEPH set to true, we should skip the ceph install on the subnode | 19:18 |
| gouthamr | which we'll try with ashrodri's patch.. | 19:18 |
| dansmith | gouthamr: ack, makes sense | 19:19 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!