| opendevreview | Steve Traylen proposed openstack/bashate master: Remove redundant import of reno for docs https://review.opendev.org/c/openstack/bashate/+/985796 | 10:56 |
|---|---|---|
| *** ykarel_ is now known as ykarel | 13:29 | |
| clarkb | sean-k-mooney: as a side note on python3.14 the tox (and nox) jobs have pyenv support built in and you can tell them to install arbitrary python versions too | 16:00 |
| sean-k-mooney | yep i have done that before | 16:04 |
| sean-k-mooney | geting it to work in dvstack is trickier. | 16:04 |
| sean-k-mooney | but pyvenv woudl techinall also be possibel | 16:04 |
| clarkb | yes I expect it would work jsut as well | 16:04 |
| clarkb | I just want to call out that you can test 3.14 today without changing anything but job config within the tox jobs. And often that coverage finds problems well in advance of devstack finding problems | 16:05 |
| sean-k-mooney | clarkb: ya i just used my uv patch becasue i wanted to also get some ci testing on that | 16:39 |
| sean-k-mooney | i would like to be able ot use uv for devstack in genreal but aht a seperate converstaions. | 16:39 |
| clarkb | I would probably be borderline -2 on that | 16:46 |
| clarkb | at least as a default | 16:46 |
| clarkb | (note I don't think pyenv should be a default either, its less about the tool and more about the fact that if we want to decuple devstack from the platform it runs on we should just go all the way and use containers) | 16:46 |
| clarkb | but we've been hesitant to decouple openstack from the platform like that for a number of reasons so I think that is the bigger question there | 16:46 |
| sean-k-mooney | i dont want it to be the default nessically | 16:47 |
| sean-k-mooney | so i think thre is a benifit to decoupling | 16:47 |
| clarkb | yes I think so too, but then there isn't really a reason to avoid containers anymore either and that better solves this problem | 16:48 |
| sean-k-mooney | but its more we have biforcation today with rpm disto not suprpoting the global venv | 16:48 |
| sean-k-mooney | but everything else is | 16:48 |
| clarkb | because then you and I can have the same python3.14 runtime in a straightfowrard manner with everything being consistent but our kernels | 16:48 |
| sean-k-mooney | yep | 16:50 |
| sean-k-mooney | that one of the reason why i was interested in usign uv to provide hte interperter and venv | 16:50 |
| sean-k-mooney | because i coudl have the same behvior on centos and or debian | 16:51 |
| sean-k-mooney | but i am not sayign it shoudl nessiary be the defualt | 16:51 |
| sean-k-mooney | clarkb: one interesiting thing of note is this is not using --site-pacakges. | 18:03 |
| sean-k-mooney | clarkb: so im thinking of add ign a ptach to make that configurble via USE_SITE_PACKAGES | 18:03 |
| sean-k-mooney | and maybe set that to false by default | 18:03 |
| sean-k-mooney | that does however mean 2 things. file injection in nova which we dont really test anyway may not work but the one job that would test that cloud just opt back in | 18:04 |
| sean-k-mooney | 2nd the way uwsgi is ued via apache need a slight twak | 18:05 |
| sean-k-mooney | tweak | 18:05 |
| sean-k-mooney | so regadless fo uv or pyenv or distop pakcages i think factoring that out of the series might be a nice improvment overall | 18:06 |
| sean-k-mooney | the use of --site-packeages has been problematic in the past and the fact that many things seam to work without it is nice to see | 18:07 |
| clarkb | sean-k-mooney: re uwsgi the issue with installing it from pypi into a venv is the compilation of uwsgi has a very high error rate | 18:36 |
| clarkb | you avoid that if installing it globally from the distro (though I can't recall if that is what devstack does or if it relies on wheels or what) | 18:36 |
| sean-k-mooney | true although with uv i think its pullign prebuilt wheels but i shoudl dboule check that in this case | 18:39 |
| sean-k-mooney | clarkb: i wonder if we shoudl also make the wsgi server more configurable long term | 18:40 |
| sean-k-mooney | or at least cenralise it | 18:40 |
| sean-k-mooney | unlike the database ro other lib files | 18:40 |
| sean-k-mooney | the wsgi stuff is not really behind a well defied interface and in its own file | 18:40 |
| clarkb | yes I think cardoe has been arguing for that. I think it is a good idea in part because uwsgi is on life support so the need to switch to something else seems like a likely outcome in the mid term future | 18:43 |
| cardoe | Yes! | 18:43 |
| opendevreview | James Parker proposed openstack/whitebox-tempest-plugin master: vTPM persistence testing https://review.opendev.org/c/openstack/whitebox-tempest-plugin/+/985731 | 18:52 |
| clarkb | sean-k-mooney: devstack install uwsgi from distro pacakges | 18:54 |
| clarkb | so ya that was the workaround to flaky uwsgi builds. And there aer no wheels on pypi | 18:54 |
| opendevreview | melanie witt proposed openstack/devstack master: DNM test host_ip_iface https://review.opendev.org/c/openstack/devstack/+/984605 | 20:34 |
| opendevreview | melanie witt proposed openstack/devstack master: DNM test host_ip_iface https://review.opendev.org/c/openstack/devstack/+/984605 | 21:10 |
| opendevreview | melanie witt proposed openstack/devstack master: WIP Fix IPv6 support for multinode devstack deployments https://review.opendev.org/c/openstack/devstack/+/984605 | 22:41 |
| opendevreview | melanie witt proposed openstack/devstack master: WIP Fix IPv6 support for multinode devstack deployments https://review.opendev.org/c/openstack/devstack/+/984605 | 22:47 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!