Friday, 2025-10-31

amoralejHi, i'm seeing an issue in our devstack jobs with prometheus datastores. While the controllers are configuring notifications.notification_format = both in nova.conf, for compute1 node it is setting notifications.notification_format = unversioned08:56
amoralejthat's making to miss some messages for instance.create.end i.e.08:57
amoralejin gnocchi jobs, it's configured correctly, but i don't see where that difference comes from08:57
amoralejfor example logs https://storage.gra.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_452/openstack/4523166e9ebf4b2da307666d68ef216d/09:00
*** jgilaber_ is now known as jgilaber09:09
jgilaberamoralej, I think the difference is that the gnocchi job deploys a decision engine in the compute09:09
amoralejno, i found similar behavior in previous jobs09:09
jgilaberand triggers https://github.com/openstack/watcher/blob/master/devstack/override-defaults to be used which configures both versions09:09
amoralejsorry, i missuderstood your message, lemme check09:10
jgilabersee https://storage.gra.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_a88/openstack/a88231bb6d3f4dd3af2e9096447e4754/compute1/logs/devstacklog.txt the logs for devstack deploy on the compute of the gnocchi job09:10
jgilaberlook for 'NOVA_NOTIFICATION_FORMAT'09:10
jgilaberin the prometheus job it has different https://storage.gra.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_452/openstack/4523166e9ebf4b2da307666d68ef216d/compute1/logs/devstacklog.txt09:11
amoralejyou are correct09:12
amoralejactually https://review.opendev.org/c/openstack/watcher/+/964546/ is fixing it in prometheus jobs09:12
jgilaberthat is surprising though, I didn't know we were deploying a decision engine in the compute https://github.com/openstack/watcher/blob/74efcbf9992b0ffee1fcd5bc72b8b4f7963a4166/.zuul.yaml#L4209:12
amoralejtbh i was not aware that gnocchi was deploying decision-engine in the compute before09:12
amoralejI may make https://review.opendev.org/c/openstack/watcher/+/964546 the first one in the series09:13
jgilaberwe can either do that or set 'NOVA_NOTIFICATION_FORMAT=both' in the subnode devstack config in the prometheus job09:15
jgilaberbut since we want to enable it anyway your patch seems like a cleaner solution09:15
amoralejthanks jgilaber++ !09:42
amoralejthat should fix some flakiness in the prometheus jobs09:44
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Add second instance of watcher-decision-engine in the compute node  https://review.opendev.org/c/openstack/watcher/+/96454609:55
opendevreviewAlfredo Moralejo proposed openstack/watcher master: APISchedulingService migrate audits also on first discovery of services  https://review.opendev.org/c/openstack/watcher/+/96398109:55
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Move decision-engine monitoring service to the decision-engine  https://review.opendev.org/c/openstack/watcher/+/96325209:55
jgilaberHi! Can someone tell me if my understanding of SLURP releases is correct? I'm working on the spec to rename strategy parameters and thinking how many releases we should keep both parameters10:22
jgilaberGazpacho is a slurp release, so we need to support upgrades from both epoxy and flamingo, but the next H relase we only need to support upgrades from Gazpacho, is that correct?10:23
amoralejhttps://releases.openstack.org/#releases-with-skip-level-upgrade-release-process-slurp10:23
amoralejyes, that's correct10:24
jgilaberI had not seen that image, that's really clear, thanks!10:25
amoralejactually, upgrade from non-slurp to slurp is marked as optional10:25
jgilaberso technically we need to support any new parameters and deprecate the old versions in gazpacho and could remove them in the next 2026.2 release, right?10:26
amoralejhttps://docs.openstack.org/project-team-guide/deprecation.html#guidelines10:28
amoraleji think we can not deprecate in G and remove in H10:31
jgilaberI see we need to wait until I to remove10:31
amoralejAt the very minimum the feature (or API, or configuration option) should be marked deprecated (and still be supported) in at least one “tick” release branch, and for at least 12 months of releases.10:32
amoraleji understand 12 months of releases means two releases ?10:32
jgilaberI understood the same10:32
amoralejanyway, for me the main discussion is how to manage the old parameters? if we want to maintain some form of backwards compatibility in strategies parameters10:33
amoralejafter old-name is not accepted for new audits10:34
jgilaberI'm working under the assumption that we will support both forms internally, mainly for audits existing before the upgrade10:36
amoralejyep, that's the main case10:36
amoralejso, if the strategy keeps accepting the old name as parameter, i'm not sure if that's technically considered parameter removal10:37
opendevreviewOpenStack Release Bot proposed openstack/watcher-dashboard master: reno: Update master for unmaintained/2024.1  https://review.opendev.org/c/openstack/watcher-dashboard/+/96567712:03
opendevreviewOpenStack Release Bot proposed openstack/watcher master: reno: Update master for unmaintained/2024.1  https://review.opendev.org/c/openstack/watcher/+/96569212:04
chandankumarHello jgilaber sean-k-mooney , I checked with openstack-release team regarding unmaintained branches removal. Based on this https://governance.openstack.org/tc/resolutions/20230724-unmaintained-branches.html, we need to keep multiple unmaintained branches in the repo. But it is not the responsability of team to maintain it. If other interested people want to maintain it and they can maintain it.12:23
sean-k-mooneychandankumar: no12:26
sean-k-mooneywe a can keep multipe but we are not required too12:26
sean-k-mooneyfor the branch to be kept after each new upstream release a unmainteind branch liason has to explcity request it to be kept, and at that poin tin time the ci need to be passing12:27
sean-k-mooneyso we shoudl propose the removal of the unmaintained 2024.1 branch and move it to eol. and if someone want to maintain it they can request to keep it and continue to do that each release12:28
sean-k-mooneyif there is no one who is requesting to maintain it now there is no reason to keep the brnach as the default is ot retire it and an explcit opt in is requried to keep it12:29
sean-k-mooneychandankumar: jgilaber  https://github.com/openstack/governance/blob/master/resolutions/20230724-unmaintained-branches.rst?plain=1#L83-L92 are the relevnet sections12:30
sean-k-mooneyjgilaber: in terms of upgrade testing yes for .1 release we are required to test upgrade form the previous .1 adn the previous .212:31
sean-k-mooneyfor .2 release testing upgrades form the previous .2 is not required and optionally done by some teams12:31
jgilaberack thanks, that was what I understood, but wanted to make I got it right12:33
jgilaberfor the unmantained branch pov iiuc we can choose to keep unmantained 2024.1 until the next slurp release or remove it now, is that right?12:37
jgilaberhmm I see  we still have unmaintained/2023.1 https://opendev.org/openstack/watcher/src/branch/unmaintained/2023.1/ that one we can remove for sure12:38
chandankumarThe PTL or Unmaintained branch liaison are allowed to delete an Unmaintained12:40
chandankumar  branch early. Since we follow DPL model, not sure who has the power to delete the branch.12:40
chandankumarchecking the same with release team.12:40
sean-k-mooneyjgilaber: i think we should remove it12:52
sean-k-mooneychandankumar: the release liasons do12:52
sean-k-mooneythe way the tooling is set up release liasons are delegate the ptl approval power 12:53
sean-k-mooneyso we delete the branch by tagging it as eol12:53
sean-k-mooneyit can be recreeated from the tag at a later date if really needed for some reason12:53
chandankumarsean-k-mooney: yes correct, I got the similar reply from release team.12:54
sean-k-mooneyi did it before :)12:54
sean-k-mooneyso i i know that a release liason can do it12:54
sean-k-mooneyi.e. i removed everythin pre 2023.1 12:55
jgilabersean-k-mooney, so we need to create a git tag eol/2023.1 and push it?12:59
sean-k-mooneyno you do it via the release repo12:59
sean-k-mooneylet me get an example13:00
sean-k-mooneyhttps://review.opendev.org/c/openstack/releases/+/94811813:00
sean-k-mooneyhttps://review.opendev.org/c/openstack/releases/+/93598313:00
sean-k-mooneythe second is where i moved all the older brnaches to eol13:01
jgilaberack thanks, I can do that later today, or do we want to raise it in the next IRC meeting?13:02
sean-k-mooneysure we can you could also send a mail to the list saying that since no one has requested it be maintained and there for no one has opted in to it we plan to remove unmaintied/2023.1 next week13:03
sean-k-mooneythen we can review it in the irc meeting and proceed with the release team if no objects13:03
jgilabersure, I can do that13:03
sean-k-mooney*objections13:03
sean-k-mooneythere has been no chang to unmaintained/2023.1 in 11 months since it was created13:04
sean-k-mooneyso that to me say no one cares13:04
sean-k-mooneystable/2024.2 shoudl go eol too by the way13:05
sean-k-mooneyit is no more then 18 month old and its not eligable to become unmainted/2024.213:05
sean-k-mooneyactully no im off by one13:05
sean-k-mooney2025.2 2025.1 and 2024.213:06
sean-k-mooneyare stil supproted until 2026.1 is release13:06
sean-k-mooneythen 2024.2 will go eol after the 2026.1 release13:06
sean-k-mooneywhat confused me is the stable/2024.1 branch has not been removed yet13:06
sean-k-mooneyalthoguh unmaintained/2024.1 now exists13:07
sean-k-mooneyso that should get fix soon i guess13:07
jgilaberthe patch https://review.opendev.org/c/openstack/releases/+/963601 was supposed to be merged last week so it only got in today13:08
jgilaberI guess that is why13:08
chandankumarelodilles is unmaintainer branch core and his employer is interested in maintaining unmaintained branches. keeping it open in upstream help to do less work downstream.13:14
chandankumar*help them13:14
opendevreviewDavid proposed openstack/watcher-tempest-plugin master: Unify wait_for_instance_in_model order execution on real data tests  https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/96577913:14
chandankumarBut when eol review will go if any one interested they can object it.13:14
jgilaberI'll write an email today and a second one with the patch next week after the IRC meeting, that should give people plenty of type to object13:15
chandankumar+113:16
opendevreviewDavid proposed openstack/watcher-tempest-plugin master: Unify wait_for_instance_in_model order execution on real data tests  https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/96577913:16
dviroeljgilaber: tks, i just added a topic to next meeting agenda13:40
jgilaberthanks dviroel!14:00

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