| *** mhen_ is now known as mhen | 01:40 | |
| opendevreview | Takashi Kajinami proposed openstack/devstack master: Fix uninialized os_VENDOR https://review.opendev.org/c/openstack/devstack/+/956489 | 03:50 |
|---|---|---|
| opendevreview | Takashi Kajinami proposed openstack/devstack master: Drop old libvirt group detection for Ubuntu/Debian https://review.opendev.org/c/openstack/devstack/+/956525 | 03:57 |
| opendevreview | Takashi Kajinami proposed openstack/grenade master: Fix missing creator role for encrypted volume creation https://review.opendev.org/c/openstack/grenade/+/956536 | 07:42 |
| opendevreview | Takashi Kajinami proposed openstack/devstack master: Drop old libvirt group detection for Ubuntu/Debian https://review.opendev.org/c/openstack/devstack/+/956525 | 09:13 |
| opendevreview | Takashi Kajinami proposed openstack/grenade master: Fix missing creator role for encrypted volume creation https://review.opendev.org/c/openstack/grenade/+/956536 | 10:06 |
| *** ykarel_ is now known as ykarel | 14:01 | |
| stephenfin | sean-k-mooney: gmaan: Have you any suggested workarounds or this novnc issue, where installing python-novnc causes global venv packages to be ignore? | 15:35 |
| stephenfin | It's blocking the oslo.utils gate currently https://review.opendev.org/c/openstack/oslo.utils/+/952914 | 15:35 |
| sean-k-mooney | stephenfin: ya so the workaroudn is to swap to using the source install | 15:36 |
| sean-k-mooney | i tought we did that in devstack already i think i pushed a patch for that | 15:37 |
| sean-k-mooney | long term we may be able to stop using libvirt-python form the disto but that why we currently use site-packages in teh global venv | 15:37 |
| gmaan | I think we did it in job side but do nt know exactly maybe in destack | 15:37 |
| sean-k-mooney | right i think we did it in the job not devstack | 15:38 |
| stephenfin | oslo.messaging is failing on the tempest-full-py3 job, which it obviously doesn't define itself. Would it make sense to change the default in DevStack if this is a consistent issue? | 15:39 |
| sean-k-mooney | https://review.opendev.org/c/openstack/neutron/+/948561 | 15:39 |
| sean-k-mooney | so i fixed it in neutron | 15:39 |
| sean-k-mooney | i tought i also had a devstack patch | 15:39 |
| stephenfin | if you do I can't see it | 15:39 |
| sean-k-mooney | stephenfin: we confimied it as a upstream packaging bug | 15:40 |
| gmaan | it was fixed it in neutron only and I think nova jobs do source install form starting | 15:40 |
| gmaan | but why oslo.messaging is failing on the tempest-full-py3 job ? | 15:41 |
| sean-k-mooney | so installing novnc pulls in an olde version fo oslo vis distro packages | 15:41 |
| stephenfin | it's pulling the import from the system rather than the venv | 15:41 |
| stephenfin | and therefore getting a really old version | 15:42 |
| sean-k-mooney | yep wone that is now below our min requirements in many projects | 15:42 |
| sean-k-mooney | its missing the image inspection code too i think | 15:42 |
| stephenfin | I don't understand why that is happening. In the event of a conflict between the two, the venv should always win. Or so I thought | 15:42 |
| gmaan | yeah | 15:43 |
| sean-k-mooney | site packages has prefernce in the python path i bleive | 15:43 |
| sean-k-mooney | not the other way around | 15:43 |
| stephenfin | o rly? | 15:43 |
| sean-k-mooney | i could be wrong about that but in anycase that is the workaround. | 15:43 |
| sean-k-mooney | we reported it to zigo and they fixed it in debian | 15:44 |
| sean-k-mooney | not sure what going to happen for ubuntu | 15:44 |
| sean-k-mooney | stephenfin: are ye now doign a release? | 15:45 |
| sean-k-mooney | this was previosly only braking jobs that used oslo master | 15:45 |
| sean-k-mooney | but i guess it makes sense that it affect more thing now | 15:45 |
| stephenfin | not yet, but the oslo.utils gate is broken | 15:45 |
| stephenfin | I wonder if the editable install has anything to do with it | 15:46 |
| sean-k-mooney | so i dont know of a reason not to swap to source install by default | 15:46 |
| sean-k-mooney | we have it in the zuul tenatnt so i cna push a patch for folks to review if you want | 15:47 |
| gmaan | I agree to do it as default, no objection | 15:47 |
| sean-k-mooney | well or you can we just need to add novnc as a required project and flip the default then push an dnm to oslo.utils | 15:47 |
| gmaan | that is what odd to do in tempest-full-py3 | 15:50 |
| gmaan | oh you mean in devstack here? https://github.com/openstack/devstack/blob/master/.zuul.yaml#L372 | 15:50 |
| sean-k-mooney | yes | 15:50 |
| sean-k-mooney | although we coudl do it in devstack itself | 15:50 |
| sean-k-mooney | and remove the overrides in devstack-base | 15:51 |
| stephenfin | I think there's a bug in Python, pip, or setuptools... | 15:51 |
| sean-k-mooney | stephenfin: maybe but you not really ment ot mix the same dep via both site pacakages and a venv like this | 15:51 |
| stephenfin | I have a local reproducer. The DevStack deploy fails. If I attempt to import oslo_utils in the global venv, it imports the system package | 15:51 |
| sean-k-mooney | correct | 15:51 |
| sean-k-mooney | that was the issue we saw orginally | 15:52 |
| stephenfin | if I then re-install oslo.utils in non-editable mode, it imports the virtualenv version | 15:52 |
| stephenfin | finally, if I re-re-install oslo.utils in editable mode, it goes back to importing the system package | 15:52 |
| sean-k-mooney | right so the egg links are slightly diffent | 15:53 |
| sean-k-mooney | between editblae and not | 15:53 |
| sean-k-mooney | for non editiabel we put the code dirfectly into the venv site pacakge dir | 15:53 |
| sean-k-mooney | for editable it refence where the code is | 15:53 |
| sean-k-mooney | so the import logic is likely diffent | 15:53 |
| stephenfin | yes, but the import *behaviour* should be consistent, surely | 15:54 |
| sean-k-mooney | it depend on how they specified it or if its implementation defiend | 15:54 |
| sean-k-mooney | gmaan: since htis also breaks locally how woudl you feel ablout updating the default in devstack and removign the overried in teh zuul jobs | 15:55 |
| gmaan | sean-k-mooney: yeah I am ok with that. but keep the NOVNC_FROM_PACKAGE if any job want to override | 15:58 |
| sean-k-mooney | sure ill just change the default to false if tis not already false. i suspect it is and it just overreided in teh base job | 15:59 |
| gmaan | its false https://github.com/openstack/devstack/blob/c05e3f23004030035459a19d57c68e9e6fb404d2/lib/nova#L946C9-L946C27 | 15:59 |
| sean-k-mooney | ya it is https://github.com/openstack/devstack/blob/c05e3f23004030035459a19d57c68e9e6fb404d2/lib/nova#L946 | 16:00 |
| gmaan | we can just remove the override from job | 16:00 |
| sean-k-mooney | so if i just remove the overried in teh zuul jobs, that will fix it for oslo and most other. and any job that need it can opt back in | 16:00 |
| gmaan | ++ | 16:00 |
| stephenfin | I think that makes sense. Can you please add a note about why we're doing it though (and that it only affects editable installs) | 16:01 |
| stephenfin | I'll try draft a quick bug report for setuptools, since I think that's where the issue is | 16:01 |
| sean-k-mooney | i can see if i can find the previosu bug? adn link it | 16:01 |
| sean-k-mooney | stephenfin: the history is the nova novnc proxy used to actully live in the novnc package like 10 years ago | 16:02 |
| sean-k-mooney | or at least they had some helper files and test utils we used | 16:02 |
| sean-k-mooney | so they added a oslo dep to the novnc package but never removed it | 16:03 |
| sean-k-mooney | novnc has not had any python code for 8-10 years | 16:03 |
| sean-k-mooney | but it still "provieds" python3-novnc | 16:04 |
| sean-k-mooney | it was completely rewrrient into javascript as part of moving to supprotign html5 | 16:05 |
| zigo | FYI, Trixie has the fix indeed ! | 16:07 |
| zigo | Not sure about Ubuntu as well... | 16:07 |
| sean-k-mooney | zigo: i remember you pushed a patch while on pto | 16:07 |
| sean-k-mooney | stephenfin: https://bugs.launchpad.net/neutron/+bug/2109592 | 16:08 |
| zigo | Maybe yeah... when I was in China last sping. | 16:08 |
| sean-k-mooney | that the bug oslo.utils is on it but its still new | 16:08 |
| sean-k-mooney | ill reuse that for the devstack patch to maintin context | 16:09 |
| opendevreview | sean mooney proposed openstack/devstack master: Use novnc from source by default https://review.opendev.org/c/openstack/devstack/+/956595 | 17:07 |
| sean-k-mooney | stephenfin: ^ ill push a dnm to oslo to confirm but that should resolve the issue. hopefully that is also enough context for us to remember to consider reverting this in the future if we need too. | 17:16 |
| sean-k-mooney | stephenfin: this is the dnm to test it https://review.opendev.org/c/openstack/oslo.utils/+/956596 | 18:09 |
| sean-k-mooney | gmaan: where does c10s enablement stand in the qa team. i.e. devsteack adn first/third party ci vai rdo | 18:25 |
| sean-k-mooney | im asking because the rdo zuul is obvirouly failing on patches to devstack | 18:25 |
| sean-k-mooney | including on https://review.opendev.org/c/openstack/devstack/+/937251 which is adding devstack supprot for c10s | 18:25 |
| gmaan | there was some proposal on that to fix | 18:25 |
| sean-k-mooney | because of how there jobs are defiend | 18:25 |
| gmaan | yeah this one | 18:25 |
| sean-k-mooney | correct. so im wondering if i shoudl help... | 18:26 |
| sean-k-mooney | i dont realy want to but im wondiering if i shoudl go fix the conflicitng nodeset defintion and add novnc/novnc to the rdo tenenat definition while im at it. | 18:26 |
| melwitt | gmaan: this tempest test about volume snapshots on stopped instance is ready for review. I made it depend on the nova fix bc I wasn't sure how that's normally handled https://review.opendev.org/c/openstack/tempest/+/956302 | 18:27 |
| sean-k-mooney | on one hand i dont personally care about c10s and would perfer to use rocky/almalinux 10 isntead | 18:27 |
| gmaan | melwitt: ack, will check | 18:28 |
| sean-k-mooney | but having failign jobs report is also annoying os im debaing if i shoud just go fix that. | 18:28 |
| sean-k-mooney | melwitt: i think we normally merge the service patch first then the tempest one | 18:29 |
| gmaan | yeah, I think we should discuss it in PTG first, either 3rd party CI is ok or upstream | 18:29 |
| sean-k-mooney | so with a depend on like you have done | 18:29 |
| melwitt | saweet | 18:30 |
| gmaan | yeah, and nova change passing CI so we are good on this order and no extra things to be done on tempest side to merge nova fix | 18:30 |
| sean-k-mooney | ya its its fine either way im just not sure if we have the mirror and nodepool iamge building update to the new way of doing that | 18:30 |
| melwitt | I was thinking about your idea of configuring nfs on the hybrid plug job and changing the name. did you want to do that or did you want me to try it? | 18:30 |
| sean-k-mooney | i mean if you have time go for it | 18:31 |
| sean-k-mooney | i would like to do it next cycle but if you have time to do it before then then cool | 18:31 |
| melwitt | I think it's important in the context of this bug fix, thats why I was thinking about it | 18:32 |
| sean-k-mooney | doing it next cycle is just being pargmatic in that i have 2-3 things ahead of it on my mental todo list | 18:32 |
| melwitt | ok cool, maybe I'll chill then. we have nfs in the experimental queue already (I just didn't notice it before) | 18:33 |
| sean-k-mooney | the ci change that is before it is i want an experimatal debian 13 job to try testing with python 3.13 | 18:33 |
| sean-k-mooney | basically im going to pivot https://review.opendev.org/c/openstack/nova/+/951749 to a new weekly perodic/experimetnal job for this cycle | 18:34 |
| melwitt | ah cool | 18:34 |
| sean-k-mooney | and then i was considerign using the nova-alt-config or nova-goth :) to validate that next cycle when 3.13 is actully part of the testing runtime and debian trix is also added | 18:35 |
| sean-k-mooney | the hybrid plug job is alreqady using debian 12 to test with spice | 18:35 |
| sean-k-mooney | hench using it with debian 13 next cycle for spice and py3.13 coverage | 18:35 |
| melwitt | makes sense | 18:38 |
| opendevreview | Merged openstack/devstack master: Fix uninialized os_VENDOR https://review.opendev.org/c/openstack/devstack/+/956489 | 20:13 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!