Thursday, 2025-11-13

noonedeadpunkhey folks12:08
noonedeadpunkso, I think I will be slowly starting with our bug figthing activity12:08
noonedeadpunkjust in case - the etherpad is here: 12:08
noonedeadpunk#link https://etherpad.opendev.org/p/osa-nov-2025-bug-triage12:08
noonedeadpunkso if you're taking part - please fill yourself in the participants list12:09
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible stable/2025.1: Bump SHAs for 2025.1  https://review.opendev.org/c/openstack/openstack-ansible/+/96649612:22
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_keystone master: Ensure shibboleth utils are installed  https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/96701812:33
noonedeadpunkthere are already 2 quite good bugs actually12:42
noonedeadpunkquestion - do we actually want to do anything wrt https://bugs.launchpad.net/openstack-ansible/+bug/2048659 ?12:55
noonedeadpunkAs the bug kinda sounds gross, but in fact I feel lazy about doing smth there12:55
noonedeadpunkas then what are we kinda going to even if we find it, but there are no other repos available12:58
noonedeadpunkI wonder how can I reproduce this one: https://bugs.launchpad.net/openstack-ansible/+bug/212277813:22
noonedeadpunkas I totally saw it during upgrades internally as well13:22
noonedeadpunkaha, ok, just drop facts...13:25
noonedeadpunkand this is related to https://opendev.org/openstack/openstack-ansible/src/branch/master/inventory/group_vars/all/nova.yml#L27-L2913:27
noonedeadpunkI actually wonder if this is correct at all... what are wetrying to achieve here? find out if there are arm computes at all?13:35
noonedeadpunkjrosser: maybe you have some idea here? as I can recall we've discussed it quite a bit in past ^13:36
jrosserin the past arm vm only did serial consoles13:37
jrosserthat might not be the case any more13:37
noonedeadpunkbut like...13:37
noonedeadpunkhow it makes sense in context of haproxy endpoints? https://opendev.org/openstack/openstack-ansible/blame/commit/2d66e80f111758c63838dd690ce6efdfe8467bfa/inventory/group_vars/nova_all/haproxy_service.yml#L7313:38
noonedeadpunkas I assume that nova_console_proxy_types gonna be unique per compute host? or at least intended to be?13:38
noonedeadpunkbecause it should not be resolved as intended when it's defining haproxy service13:40
noonedeadpunklike we need to get from all hostvars for the nova_compute group[13:44
noonedeadpunkshould not these be 2 separate tasks? https://opendev.org/openstack/openstack-ansible-os_nova/src/branch/master/tasks/nova_install.yml#L78-L9313:48
noonedeadpunkas we might need to run both?13:49
noonedeadpunkand again nova_console_type is kinda irrelevant...13:49
noonedeadpunkor incorrect in context of setup13:50
noonedeadpunk /o\13:50
* noonedeadpunk thinking where to start unpacking this13:51
jrosserwell13:52
jrosserthe console proxies run in the control plane13:52
jrosserand currently we have logic to decide which of those proxies get run (spice/novnc/serial)13:53
jrosserbut tbh this is a very awkward set of logic to deal with13:53
noonedeadpunkso if your all controllers are x86_64 - what will be this one? nova_console_type: "{{ (ansible_facts['architecture'] == 'aarch64') | ternary('serialconsole', 'novnc') }}"13:53
jrosseras it has been the case that "if arm hosts you need serial"13:53
noonedeadpunkor what if one controller is arm6413:53
jrosser"if ironic consoles you need serial"13:53
jrosserand so on13:54
noonedeadpunkbut now we take fact for current host13:54
noonedeadpunkand assume we verify all computes13:54
noonedeadpunkand take decisions on that13:54
jrosserso, it might just end up being simpler to install all the proxies, i'm not sure that there is a good answer (other than some kind of optimisation) to do that selectively13:54
jrosserthe logic in the haproxy role was previously wrong and horrible13:54
jrossertbh i think the best answer here is to simplify13:55
noonedeadpunkit's still wrong on all levels... I'd assume you just having override for nova_console_proxy_types13:55
jrosseryes we do, in user_variables so it's global for *13:56
noonedeadpunkright. as if you don't current code will likely screw things13:56
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_nova master: Include both novnc and spice tasks if needed  https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/96703614:01
mgariepyhmm. any issue with https://github.com/openstack/openstack-ansible-os_nova/blob/stable/2024.1/tasks/nova_post_install.yml#L126-L133 ?14:10
mgariepyi'm deploying new compute and it fails because the files don't exist.14:10
noonedeadpunkum, nothing I'm aware of?14:10
noonedeadpunktry -e venv_rebuild=true14:10
mgariepyit's copied from venv to /etc/nova/ ? 14:11
noonedeadpunkI think that /etc/nova is a symlink to venv14:17
mgariepyyep indeed.14:17
mgariepynot sure what happen to the first time it did install.14:18
mgariepybut whatever.14:18
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Improve nova_console_proxy_types logic  https://review.opendev.org/c/openstack/openstack-ansible/+/96704714:53
noonedeadpunkI hope I am thinking in the correct direction here^14:53
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Improve nova_console_proxy_types logic  https://review.opendev.org/c/openstack/openstack-ansible/+/96704714:56
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Improve nova_console_proxy_types logic  https://review.opendev.org/c/openstack/openstack-ansible/+/96704714:57
jrosseri do wonder if more modern arm/libvirt supports graphical console15:09
jrosseri'm not sure though15:10
jrosserit should be obvious though with a vga device in the vm?15:10
noonedeadpunkum. I have nowhere to check that15:12
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_nova master: Ensure nova_console_proxy_types respects all computes  https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/96705215:12
noonedeadpunkbut I'd say that these patches, ideally, should be getting us to behavior which I think was intended15:14
noonedeadpunkby extracting arch vars from all computes and making `nova_console_proxy_types` and deployment-wide variable15:15
noonedeadpunkand also closing bug :)15:16
andrewbonneynoonedeadpunk: I can't test exhaustively, but I've tried that on one of our deployments and it appears to behave correctly15:28
noonedeadpunknice, thanks15:30
jrosserheres the arm pci devices i have currently https://paste.opendev.org/show/bj9NRPY3Z6YJ0dj0RPGs/15:37
jrosserso i have serial console on that, which is working15:40
noonedeadpunkyeah, I was more about default logic, which should not work, unless you override it to be correct15:45
noonedeadpunkanyway. if we wanna change default console type for arm - we can do that as a follow-up15:46
damiandabrowskio/15:55
*** jizaymes_ is now known as jizaymes16:04
noonedeadpunkso...16:06
noonedeadpunkwhat are we going to do with this one https://bugs.launchpad.net/openstack-ansible/+bug/211255916:06
noonedeadpunkI've just confirmed it's still valid16:07
noonedeadpunkwe can kinda drop https://opendev.org/openstack/openstack-ansible-os_nova/src/branch/master/tasks/nova_post_install.yml#L86-L9416:07
noonedeadpunkbut I'm afraid we don't have any good way to say that clean up file content if it already exists16:08
noonedeadpunkThe simplest thing - to always have an empty policy.yaml regardless16:08
noonedeadpunkNext what we can do - add notify to this task 16:08
noonedeadpunkAnd we have that for pretty much each and every role: https://codesearch.openstack.org/?q=Remove%20legacy%20policy&i=nope&literal=nope&files=&excludeFiles=&repos=16:09
noonedeadpunkHard way could be try to fix the oslo_policy....16:10
damiandabrowskitbh, I'm fine with both(empty policy.yaml file and adding notify)16:15
noonedeadpunkI will probably switch now to looking through really old bug reports to mainly close them16:21
damiandabrowskinoonedeadpunk: I wonder what to do with https://bugs.launchpad.net/openstack-ansible/+bug/208775916:24
damiandabrowskiso if I understand correctly, OSA will never show up in "deployment projects" section in https://docs.openstack.org/2025.2/projects.html for the development release16:25
damiandabrowskiit's really surprising that tripleo is there :D 16:25
damiandabrowskibut I wonder if we can do something to always be there, even for releases that are in development16:25
damiandabrowskii think it's okay to not show in "deployment guides" (https://docs.openstack.org/2025.2/deploy/) until we do OSA release for the particular openstack version16:26
noonedeadpunkyes, tripoleo is weird16:26
noonedeadpunkI'm not really sure if we can do anything wrt16:26
damiandabrowskibut it would be nice to always be listed in deployment projects section16:26
noonedeadpunkwhenever we release a new version, I am going and patching openstack-manuals16:26
noonedeadpunkto ensure osa is shown there16:26
noonedeadpunkAnd kolla does pretty much the same today16:27
noonedeadpunkas we are not part of coordinated release, and kinda no point to show up for the verison, which we don't have yer16:27
noonedeadpunkand actually we show up for development branch: https://docs.openstack.org/2026.1/projects.html16:28
noonedeadpunkjust 2025.2 is not development anymore :)16:28
damiandabrowskiahh right...16:28
damiandabrowskishould be close that bug report then?16:28
noonedeadpunkSo I think we should say we probably can't fix it...16:28
damiandabrowskiack, will do that16:30
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-repo_server master: [doc] Add documentation around repo_server role  https://review.opendev.org/c/openstack/openstack-ansible-repo_server/+/96705916:46
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_ceilometer master: Add coordination support to the role  https://review.opendev.org/c/openstack/openstack-ansible-os_ceilometer/+/96706417:10
opendevreviewMerged openstack/openstack-ansible stable/2025.1: Bump SHAs for 2025.1  https://review.opendev.org/c/openstack/openstack-ansible/+/96649617:12
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_ceilometer master: Add coordination support to the role  https://review.opendev.org/c/openstack/openstack-ansible-os_ceilometer/+/96706417:13
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_ceilometer master: Add coordination support to the role  https://review.opendev.org/c/openstack/openstack-ansible-os_ceilometer/+/96706417:14
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_aodh master: Add coordination support to the role  https://review.opendev.org/c/openstack/openstack-ansible-os_aodh/+/96706717:20
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_aodh master: Add coordination support to the role  https://review.opendev.org/c/openstack/openstack-ansible-os_aodh/+/96706717:21
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Test coordination for Telemetry stack  https://review.opendev.org/c/openstack/openstack-ansible/+/96707017:24
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: [doc] Fix link to supported distributions  https://review.opendev.org/c/openstack/openstack-ansible/+/96707617:39
damiandabrowskinoonedeadpunk: any ideas what to do here? https://bugs.launchpad.net/openstack-ansible/+bug/207419617:58
damiandabrowskihttps://review.opendev.org/c/openstack/keystone/+/930589 is cherry-picked to 2024.1 and 2025.117:58
damiandabrowskiso if you were right about that, the problem is fixed. But hard to fully confirm it17:58
noonedeadpunkI am not 100% sure, but I think it was related17:59
noonedeadpunkBut also there are probably multiple reports regarding the same issue17:59
noonedeadpunkalso backprot to 2024.2 was not merged18:00
noonedeadpunkI think I'd suggest to mark as fix released18:00
noonedeadpunkas no idea what else we can do there anyway18:01
noonedeadpunkand iirc there were actually migrations in 2024.1 which were related to the issue18:01
damiandabrowskiyeah....I was also leaning towards marking it as fixed. Will do that18:01
noonedeadpunk++18:01
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_keystone master: Replace set_fact with vars defenition  https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/96708818:06
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-os_keystone master: Use dynamic includes for conditional tasks  https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/96708918:13
noonedeadpunkjrosser: need your Ironic knowledge to evaluate https://bugs.launchpad.net/openstack-ansible/+bug/1896355 18:21
noonedeadpunkdoes it make sense to have ironic-conductor to run separately from API?18:21
jrossercentos7 / train -> invalid :)18:22
jrosserbut yeah18:22
noonedeadpunkI'm just thinking it still might make sense to fix it?18:23
noonedeadpunklike looking at current env.d and there's no good way to do that today: https://opendev.org/openstack/openstack-ansible/src/branch/master/inventory/env.d/ironic.yml18:23
jrosseratm it’s kind of simplistic18:24
jrosser but there is a concept in ironic of conductor groups that I don’t think we support at all18:25
jrosserso there is one thing there is we want to be able to split conductor/api apart, and for scale out that might totally be worth it18:25
jrosserand then a second is how to have conductor groups and what that would mean18:26
noonedeadpunkthat is potentially doable with overrides... but it's gonna be quite confusing18:27
jrosserso first step is certainly to improve the existing env.d18:27
noonedeadpunkyeah18:27
noonedeadpunkok, thanks, so it;s still valid then18:28
jrosseryeah, particularly if you don’t use swift for images18:28
jrosserie the conductor has to locally host and serve them18:28
noonedeadpunkok, cool, I wrote that down to etherpad18:33
noonedeadpunkkinda wonder wtf is this one https://bugs.launchpad.net/openstack-ansible/+bug/183306618:57
jrosserthis could be just something unrelated to the lock dir19:03
jrosserlooks very odd tbh19:03
noonedeadpunkany guesses wtf is that? https://opendev.org/openstack/openstack-ansible/src/branch/master/inventory/env.d/glance.yml#L29-L3019:53
noonedeadpunkor well, I know it's size of cow or lvm volume for glance19:53
noonedeadpunkbut like... wtf it's doing there....19:54
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Remove hardcoded container_fs_size for glance  https://review.opendev.org/c/openstack/openstack-ansible/+/96710720:01
noonedeadpunkso we went down from smth around 154 bugs to 12920:24
noonedeadpunkwhich is not great, but given it was only me and Damian - it's not that bad either :)20:24
noonedeadpunkAnd we totally need to handle found issue with policy removal. As this sounds like a nasty thing to have...20:25
-opendevstatus- NOTICE: The OpenDev team will be restarting Gerrit at approximately 2130 UTC in order to pick up the latest 3.10 bugfix release.20:30
noonedeadpunkwill check out for today20:44

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