Thursday, 2025-08-28

*** mhen_ is now known as mhen01:15
opendevreviewmelanie witt proposed openstack/tempest master: WIP Add tests for vTPM live migration  https://review.opendev.org/c/openstack/tempest/+/95747503:30
opendevreviewmelanie witt proposed openstack/tempest master: WIP Add tests for vTPM live migration  https://review.opendev.org/c/openstack/tempest/+/95747503:37
opendevreviewGhanshyam proposed openstack/devstack master: Configure glance user in cinder conf  https://review.opendev.org/c/openstack/devstack/+/95871804:07
gthiemongeHi folks, octavia-grenade is broken on 2025.1, it appears that it's related to the setuptools version installed in 2024.2 and some new python modules installed during the update07:36
gthiemongeI added more details in https://bugs.launchpad.net/octavia/+bug/212157807:36
gthiemonge(designate-grenade-* is also affected)07:37
gthiemongeany ideas? shouldn't we update setuptools in grenade?07:37
fricklerin general it would be good if upgrades can work with only updating openstack code, keeping dependencies untouched. not sure if that is possible with what setuptools maintainers are doing08:05
*** ralonsoh_ is now known as ralonsoh08:43
oschwartHey folks, the same errors that broke grenade in Octavia break Designate grenade jobs too09:08
oschwartI am not sure, but maybe this patch could fix those (?)09:09
oschwarthttps://review.opendev.org/c/openstack/keystonemiddleware/+/95401809:09
gthiemongeoschwart: pip is already pulling oslo.cache 3.10.2, so I don't think it will fix it09:19
oschwartIt looks like grenade gets a Segmentation fault when it runs sudo usermod -a -G libvirt stack12:26
oschwartFrom a quick search I found instances of that error on tempest, devstack, neutron, designate, cinder, manila, and others12:26
clarkbgthiemonge: frickler: we shouldn't be pinning setuptools at all...14:25
clarkbI'm not sure I understand the "setuptools version installed in 2024.2" assertion. It should use the latest version of setuptools supported by the python version there14:25
clarkbthis changes if you start using pyproject.toml everywhere to control the setuptools version but until openstack does that setuptools isn't something we can really manage the version of reliably so we've tried to stick with the latest14:26
gthiemongeclarkb: ack, I will check which setuptools version is installed in the initial deployment14:29
clarkboschwart: any idea if the segfault is in sudo or usermod? I would probably start there to determine if sudo or usermod is the source of the bug14:32
clarkbdo you have a link to the failure? the link about has passing tests14:34
clarkb*the link above14:35
gthiemongeclarkb: during the deployment, devstack does something like "pip install -U pip 'setuptools[core]' -c https://raw.githubusercontent.com/openstack/requirements/refs/heads/stable/2024.2/upper-constraints.txt" which pulls setuptools 75.3.2, this version appears to be broken with recent modules, and we don't update it before/during grenade14:36
clarkbwow I would personalyl consider that a bug in devstack. For this very reason14:37
clarkbas a general rule trying to control setuptools is a bad idea unless you can control the entire build environment14:37
clarkb(in this case I'm guessing that PBR is too new for that old setuptools but I'm not certain)14:38
clarkbso by only controlling a portion of the build environment there is a risk of these sorts of failures14:38
oschwartclarkb here is an example: https://zuul.opendev.org/t/openstack/build/ece7d0bfffd34c1e90cdec378dc7f94315:50
clarkboschwart: https://zuul.opendev.org/t/openstack/build/ece7d0bfffd34c1e90cdec378dc7f943/log/controller/logs/syslog.txt I think this indicates that usermod is segfaulting not sudo.15:52
oschwarthmm right, I see it now15:54
clarkblibc6 (the ld-linux-x86-64.so.2 source) doesn't seem to haev been updated since May. And passwd (where usermod comes from) is even older15:59
clarkbits also curious that nothing else seems to be tripping it (everything links to libc so you'd expect a lot more fallout)16:00
clarkbI can put a hold request in so that you can inspect the node directly. Maybe the package is corrupted or something?16:00
clarkbjust let me know what change and job you'd like to do that for if you want ot do that16:01
clarkbbut then you can run it under gdb and check the files match what you see upstream in the repos16:03
oschwartThat same error happened also to tempest, devstack, neutron, cinder, manila16:08
oschwartI don't think it is related to a specific patch or to designate16:08
clarkbright its happening in usermod and ld-linux16:11
tkajinamis anyone aware of the problem that all of the "installed" flake8 plugins are now enabled regardless of flake8:local-plugins ?16:24
opendevreviewmelanie witt proposed openstack/tempest master: WIP Add tests for vTPM live migration  https://review.opendev.org/c/openstack/tempest/+/95747517:54
sean-k-mooneyanyone know how depenent the rpm disto support is on rdo in devstack. i know we used to use eple in the past but started depenign on rdo  more over time19:51
clarkbI think that packges like ovs come from rdo19:52
sean-k-mooneyon c9s at leat we coudl have pulled form form the centos fastdatapath sig i think19:53
sean-k-mooneybut ok19:53
clarkbin devstack all of the openstack components should be from source. Then we get things like the database and libvirt from the distro. On centos/rhel ovs did not exist in the base os and it comes from rdo19:53
clarkbso I think its basically whatever non openstack systems aren't available in the os (ovs, ceph, rabbit? etc)19:54
sean-k-mooneyhttps://sigs.centos.org/nfv/#packages-and-centos-stream-versions-matrix19:54
clarkbyou can probably get a sense for those items by listing packges in rdo and removing all the openstack stuff19:54
sean-k-mooneyya the reaon im asking is redhat has massivly pulled back it investment in rdo of late and no one has really stepped in to supprot it in the rdo comunity19:55
sean-k-mooneyits not as if there is 0 maintaince19:56
sean-k-mooneybut long term im not sure whats going to happen there19:56
sean-k-mooneyso im wondering if we could pull more form upstream distos ranter then deping on rdo in the future19:57
sean-k-mooneyto be fair its many of the same people inovled in the centos sigs and rdo19:57
sean-k-mooneyclarkb: ceph we take form upstream anyway and installe ti with cephadm as contaienr i think19:58
sean-k-mooneyat elast i dont think we use distro ceph as our defaults even if its an option in the devstack plugin19:59
sean-k-mooneyhttps://github.com/openstack/devstack-plugin-ceph/blob/master/devstack/lib/cephadm#L3220:00
sean-k-mooneyi would be very surpised if rabbit or mysql is not alos packged elsewhere 20:00
sean-k-mooneyim not going to start looking into this now20:01
sean-k-mooneybut i am going to wait and see if RDO does an offical 2025.2 release20:01
sean-k-mooneyand if so what distos they release packages for20:01
clarkbya I don't know concrete details just giving examples of what may be20:03
clarkbbasically its the tertiary things that are not part of centos/rhel proper20:03
sean-k-mooneyworst case we could continue down the ceph route and use containers if there were missing native packages at least for rabbit ectra20:04
sean-k-mooneyalthough it woudl calll into qustion the utility of testing with centos at all20:05
opendevreviewMerged openstack/devstack master: Configure nova 'service' role in tempest  https://review.opendev.org/c/openstack/devstack/+/95861220:05
sean-k-mooneyi see takasi asked the same thing https://meetings.opendev.org/irclogs/%23rdo/%23rdo.2025-08-22.log.html and 20:08
sean-k-mooneyno responce20:09
sean-k-mooneythe mailing list have been dead sicne 202120:09
opendevreviewmelanie witt proposed openstack/tempest master: WIP Add tests for vTPM live migration  https://review.opendev.org/c/openstack/tempest/+/95747520:53

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!