Thursday, 2026-08-20

opendevreviewMerged openstack/watcher stable/2026.1: Add policy enforcement to webhook trigger endpoint  https://review.opendev.org/c/openstack/watcher/+/100150700:36
opendevreviewMerged openstack/watcher stable/2025.2: Add policy enforcement to webhook trigger endpoint  https://review.opendev.org/c/openstack/watcher/+/100150807:16
opendevreviewMerged openstack/watcher stable/2025.1: Add policy enforcement to webhook trigger endpoint  https://review.opendev.org/c/openstack/watcher/+/100150910:36
opendevreviewMerged openstack/watcher master: Add Nova/Placement emulators and host_maintenance functional tests  https://review.opendev.org/c/openstack/watcher/+/99335211:17
opendevreviewDouglas Viroel proposed openstack/watcher stable/2026.1: Fix allowed_nodes corruption in ComputeScope  https://review.opendev.org/c/openstack/watcher/+/100164411:20
opendevreviewDouglas Viroel proposed openstack/watcher stable/2025.2: Fix allowed_nodes corruption in ComputeScope  https://review.opendev.org/c/openstack/watcher/+/100164511:52
opendevreviewDouglas Viroel proposed openstack/watcher stable/2025.1: Fix allowed_nodes corruption in ComputeScope  https://review.opendev.org/c/openstack/watcher/+/100164812:05
dviroel#startmeeting watcher13:00
opendevmeetMeeting started Thu Aug 20 13:00:20 2026 UTC and is due to finish in 60 minutes.  The chair is dviroel. Information about MeetBot at http://wiki.debian.org/MeetBot.13:00
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.13:00
opendevmeetThe meeting name has been set to 'watcher'13:00
* dviroel forgot to send a reminder message :)13:00
dviroelhi all o/ - who is around today?13:01
morenodo/13:01
winiciusallan[m]o/13:01
dviroelcourtesy ping: amoralej sean-k-mooney chandankumar rlandy13:01
amoralejo/13:02
dviroelalright, lets start 13:03
dviroelstarting with today's meeting agenda13:03
rlandy_o/13:03
dviroel#link https://etherpad.opendev.org/p/openstack-watcher-irc-meeting#L34 (Meeting agenda)13:03
*** rlandy_ is now known as rlandy13:03
chandank`o/13:03
dviroelfeel free to add any topics that you want to cover/highlight13:03
dviroelwe should be covering today the patches that need to merge before FF13:04
dviroelfirst13:04
dviroel#topic Announcements 13:04
sean-k-mooneyo/13:04
dviroelthere are a few13:04
dviroel1) PTL/TC Elections13:04
dviroelso nominations ended, we have the candidates list here https://governance.openstack.org/election/#ptl-candidates13:05
dviroelas we can see, no ptl elections will happen13:05
dviroeland now13:06
dviroelthe TC Election Campaigning Period starts13:06
dviroelyou can find more info on13:06
dviroel#link https://governance.openstack.org/election/13:06
dviroelor in the kickoff email sent 13:07
dviroel#link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/LTBWW3CEIPYG6QPQAUUQHGJKPPV7GPSC/13:07
dviroeland that cover the election announcements13:07
dviroelnext one is13:08
dviroel2) Feature Freeze is next week13:08
dviroel#link https://releases.openstack.org/hibiscus/schedule.html13:08
dviroelwe should cover the pending reviews/blueprints in the following topics13:08
dviroellets see how we are with them13:08
dviroelany other announcement that you folks want to cover?13:09
dviroelif not, we can move to our Review topic13:09
dviroelack13:09
dviroel#topic Reviews13:10
dviroelthere is a few already listed in the meeting etherpad13:10
dviroellets start with the feature listed in the review etherpad13:10
dviroel#link https://etherpad.opendev.org/p/watcher-2026.2-status13:10
dviroel1) Openstacksdk migration13:10
dviroelwe already discussed last week13:10
dviroelwe are godd there13:11
dviroel2) Audit Pipeline13:11
dviroelAudit template extension already merged, only tempest tests are still pending , which is ok13:11
dviroelwe can merge them later too13:11
dviroelAbout the audit pipeline itself13:12
dviroel#link https://review.opendev.org/q/topic:%22bp/audit-pipeline%22+is:open13:12
dviroelthere is already a lot of changes open13:12
dviroelwe merged the first one about strategy metric cache13:13
dviroelthere are some updates since last week13:13
dviroelI included an additional change just to cover the efficacy indicators piece13:13
dviroelit ended to be a little bit different from the spec13:14
dviroelI had to create a indicator grouping for similar indicators: e.g: zone_migrations provides both planned_live_migrate_instance_count and planned_cold_migrate_instance_count, while workload_stabilization provides instance_migrations_count13:14
amoralejwhat combination of strategies may be a good one to test?13:14
dviroelI have been using zone_migration/host_maintenance + workload_stabilization13:15
amoralejhardware_maintenance + workload_stabilization? 13:15
dviroeldepends how many compute nodes you have13:15
sean-k-mooneyhum ok we may want to cosnier if we shoudl normalise those indicator in the future13:15
sean-k-mooneyi think today the have mostly been adhoc rahter then standarised names across all the stragies for the same thing13:16
dviroelamoralej: since hardware_maintenance will disable one node, you will likely want 2 others to balance13:16
amoralejyes13:16
amoralejah, right, good point13:16
amoralejactually, it's hard to test with two, i guess...13:16
sean-k-mooneyya13:16
amoralejzone_migration +  workload_stabilization would send instances back to the origin one?13:17
dviroelbut in CI i proposed zone_migration (which does not disable the source) to have it moveing an instance back (which also tests the audit pipeline collapsing migrations)13:17
sean-k-mooneyyou can use consolidation and then stabelisation13:17
sean-k-mooneybut it may be hard to seee the final result13:17
sean-k-mooneyrealy you need 3 computes13:17
amoralejyep13:17
sean-k-mooneythe funciotnal test would be good for testign this later13:18
dviroelwe currently don't have a CI job running no check/gate with 3 nodes..13:18
sean-k-mooneywe can create more complex sensiors without the actull resouce need13:18
dviroelright13:18
amoralejyep, i was also thinking in the functional ...13:18
sean-k-mooneydviroel: that prety trival to add 13:18
sean-k-mooneyjust need to defien a new nodeset13:18
dviroelsure, we can discuss if we want to  move one job to 3 nodes13:19
dviroelwe already have the nodeset 13:19
morenodim going to need to add a job with 3 computes13:19
sean-k-mooneyfor now we can test with the actuator13:19
sean-k-mooneywe can compose a number of actuator audit templeates toghether13:19
morenodfor audit scope: https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/99579913:19
morenodhttps://review.opendev.org/c/openstack/watcher/+/99579713:19
morenodwe can use that13:19
sean-k-mooneyfor example have one live migrate a vm and another cold migrate it13:20
sean-k-mooneyput both in a pipelien13:20
dviroelgood idea 13:20
sean-k-mooneyand see that all of the actions are applied 13:20
sean-k-mooneyin the event log13:20
sean-k-mooneyit can be 2 live mgiration was well or any other to actions13:21
sean-k-mooneyi.e. disabel in one and enable in the other13:21
sean-k-mooneyas long as we can assert the sideffects13:21
dviroelso for now, there are one thing that is still missing in the changes13:22
dviroelthat  is the storage model mutation, current patches only works with compute model13:22
dviroelI plan to propose as an additional change13:22
dviroelto not have bigger changes13:22
amoralej+113:23
dviroeland see what we can have merged in this release13:23
dviroelI also have a spec update to made, which I just need to push13:23
amoralejhonestly, we have just one supported strategy using the storage model, right?13:23
dviroelwe also have StorageCapacityBalance 13:24
dviroelbut we don't even have a ci job running for it13:24
dviroelonly for zone_migration13:24
amoralejyep13:24
dviroelso yeah13:24
dviroeldepending on the amount of work to get all in, we may need to consider defering some specific funcitonalties to the next release13:25
amoraleji think it's fine to do storage mutation in separated review13:25
dviroelack13:25
sean-k-mooneyit would be ok for the storage support to come next cycle13:25
sean-k-mooneyprovide we docuemnt that adn call it out as a limiation13:25
dviroelsean-k-mooney: teim-ci has been very useful and got a lot of issues in the audit pipeline patches13:25
sean-k-mooneyim glad :)13:25
dviroeli proposed updats based on its output13:26
sean-k-mooneyjust an fyi to everyone i made some changes to it on monday to improve its output after some regressiosn 3 weeks ago13:26
dviroelso for now, feel free to ping me for any audit pipeline related question13:26
sean-k-mooneyso let me know how that is workign for folks13:27
dviroel++13:27
dviroeldo you guys have any concern or question about this feature? otherwise we can move to the next one13:27
sean-k-mooneyi think you have highlight the main risks13:27
sean-k-mooneyso i am not sure we realsiticly will get everythign merged i.e. the storage supprot13:28
sean-k-mooneybut we can try13:28
dviroelack, lets see how we move forward during next weel13:28
sean-k-mooneyand as i asid if we docuemnt that as a limitation i dont think that is a blocker for the basic supprot13:28
dviroelwe can go with a proper documentation with limitations if needed13:28
dviroelack13:28
dviroelok, next one in the list 13:29
dviroelFunctional Tests Infrastructure13:29
dviroeli think that we are on track here, we recently merged some patches13:29
dviroelit will be very useful for us to expand our tests13:29
amoralejyes, thanks for merging13:29
dviroelI think that we can move forward in the list 13:30
amoralejI'm working in adding the prometheus one asap13:30
dviroel+113:30
amoralejFYI, i'm creating some helpers functions for the topology to populate metrics based on mean or standard deviation13:30
sean-k-mooneysound useful13:31
dviroelthat change will be very helpful for audit pipeline, mainly for strategies that we also simulate the metrics13:31
amoralejso, after adding the compute and instances using the builder pattern something like ..distribute_cpu(mean=X, sd=Y)13:31
amoralejI'm trying to get that working while keeping predictability13:31
amoralejwhich i think will be helpful when creating topologies13:32
dviroelyeah, for more specific test cases13:32
amoralejyep13:32
dviroelnice amoralej++13:33
amoralejthat's it wrt functional tests13:33
dviroelcheck if makes sense splitting changes, depending on the sizing13:33
sean-k-mooneythe builder pattehr can have its limits13:33
amoralejyes it can :) 13:33
sean-k-mooneyits nic but we may also want to be abel to tweak spcrif comptue ectra13:34
amoralejwe will never lose the option of defining specific compute, instance, etc...13:34
amoralejhelper is just something optional on top13:34
amoralejif i can make it work :) 13:34
dviroelack13:35
dviroelok, lets then move to the next feature in the list13:35
amoralejactually, it may also be follow up13:35
dviroelamoralej: sometimes make easier to merge13:35
dviroelnext one is13:35
dviroelPreemtible Instances13:35
dviroeli think that winiciusallan[m] is with us13:35
winiciusallan[m]yes, I'm around13:36
dviroellet me update the etherpad13:36
dviroelsince you merged the 2 actions in one patch13:36
dviroel#link https://review.opendev.org/c/openstack/watcher/+/100016813:36
winiciusallan[m]in general, the actions patch are ready to review13:36
winiciusallan[m]I need to push tempest tests for those actions, will try to work on it this week13:37
amoralejthanks for adding the docs winiciusallan[m]!13:38
dviroelack, as I mention last week, let me know if you need help with that13:38
winiciusallan[m]I will probably need some help to setup an environment so I can test myself quickly13:38
winiciusallan[m]but yeah, I will let you guys know13:38
dviroelack, you can ping me or someone else in the irc, no problem13:38
winiciusallan[m]do you have any concerns regarding the strategy patch? I saw that you left a few comments13:39
amoralejquick question, instance deletion in nova is synchronous or asynchronous ?13:39
winiciusallan[m]IIRC it's async13:39
winiciusallan[m]sean can confirm13:39
amoralejyou are waiting for it in the helper13:41
amoralejso it should be finbe13:41
dviroelwhile we wait for sea n to confirm13:41
amoralejnp13:41
dviroel#link https://review.opendev.org/c/openstack/watcher/+/100017013:41
* dviroel got distracted13:43
dviroeli have been reviewing the second patch13:43
dviroelthat adds the new goal/strategy13:43
sean-k-mooneyinstance deleteion i belive returns a 20413:43
sean-k-mooneyso i think its async yes13:43
sean-k-mooneywe poll for the instnace to not be in server show in tempetst normally13:44
sean-k-mooneyactully idf definlty ascyn13:44
dviroelI saw an issue with sort_candidates method in the strategy13:44
dviroelsince in the spec was mentioned that it would be using placement resource allocations13:44
dviroelbut in the change it is using flavor resources already available in the model13:45
sean-k-mooneybecause we have a feature call "local delete" where if the compute agent is offlien when you ask to delete a vm it will be marked as deleted in teh db and then actully deleted when the compute agent is started again13:45
sean-k-mooneydviroel: ah yes13:45
sean-k-mooneyi have not looked at the yet13:45
dviroelin order to keep in sync with the spec13:45
sean-k-mooneybut we do not curently capture the allocation13:45
dviroelwe would need to also extend the cdm13:45
sean-k-mooneyyes13:45
winiciusallan[m]yes, it'd be needed a chnage in the CDM13:46
sean-k-mooneywe shoudl do that13:46
sean-k-mooneybut if we dont have time just remove the resouce match on the allcation for now13:46
sean-k-mooneyand readded it later13:46
sean-k-mooneyi think that will be simple to add13:46
dviroelwe are 1 week from feature freeze, so we need to be careful on what we still plan to propose13:46
dviroelsince we may not have time to review all13:47
dviroelyeah, we can move that for next release13:47
sean-k-mooneyi would split the reouce matching part into its own comiit13:47
sean-k-mooneyadn we can merge it if we have tiem and if not continue it into next cycle13:47
winiciusallan[m]++13:48
winiciusallan[m]but I'll do what's currently missing in the already opened patches, and then can owrk on this ^13:48
dviroelwiniciusallan[m]: so you can keep  only the SORT_KEY_AGE in the sort_candidates13:49
dviroeland propose a new change on top of 1000170 adding the resource sorting piece + cdm13:50
dviroelas an extension of preemption_preference 13:50
dviroelthis additional change will have more risk of not making into hibiscus13:51
dviroeland focus first on solving all issues/comment in current strategy implementation + tempest test addition13:51
winiciusallan[m]so wouldn't be the CDM change an independent patch?13:51
amoralejsorry but, the cdm changes are to have the resource usage allocations in computes ?13:53
sean-k-mooneyit woudl be a singel patch for resouce and cdm13:53
sean-k-mooneyyou coudl do it cdm and then reouce sort in that order13:53
sean-k-mooneyas 2 patches13:53
sean-k-mooneybut useing the flavor is not correct today13:53
amoralejwe have logic in the model to calculate resources usage13:54
sean-k-mooneyit devites form teh design in the spec13:54
amoralejbased on flavor13:54
amoralejwhy is not correct sean-k-mooney ?13:54
sean-k-mooneythe spec wasy this will be based on the placement allcoation so it works for all resocue classes13:54
sean-k-mooneyi.e. cybrog device 13:54
sean-k-mooneyor anything else beyond cpu ram and disk13:55
amoralejack13:55
sean-k-mooneywe coudl have a limited form for cpu ram and disk13:55
sean-k-mooneyprovided that does not use resouce:13:55
amoralejbut for cpu, ram and disk, we'd rely on existing logic to calculate used resources ?13:55
sean-k-mooneythe `resouce:` syntx in the spec was specificly to be exlictvly based on placmente allcoation13:55
amoralejotherwise we need to check placement on instance notifications 13:56
sean-k-mooneythe placement allction only change in repsocne to api action like spawn or live migrate13:56
sean-k-mooneybut yes they are not in the notifcaitons13:56
amoralejin my understanding, when discussing about cpu / mem / disk i was always thinking on the methods we already have13:56
sean-k-mooneyso we woudl have to call placmenet on notificao13:56
sean-k-mooneyamoralej: we had 2 diffent realted features13:57
sean-k-mooneyone was supproting placeement allcaotion based resouces sorting13:57
sean-k-mooney`resouces:` constatins13:57
sean-k-mooneyand the we had the concpe of a generic waying by usage13:57
sean-k-mooneyevenully i think we will end up deprecating and removing the curren cpu/ram/disk methods 13:58
sean-k-mooneyin favor fo placmeent only beut that off topic13:58
amoralejyep13:59
sean-k-mooneyhttps://opendev.org/openstack/watcher-specs/src/branch/master/specs/2026.2/approved/preemptible-instances.rst?display=source#L189-L20113:59
sean-k-mooneythat is what we are discssiong now and that is excpive based on placment allcoation13:59
sean-k-mooneynot using any of the existign fields13:59
amoralejyep, that requires placement, right14:00
sean-k-mooneyif we want to have a generic `size` that uses the exitng option we could btu that was not part of the spec dezie14:00
amoralejack14:00
sean-k-mooneyso size:cpu or  usage:cpu14:01
dviroelit seems more feasible to get preemptible intances without including this, in this cycle14:01
winiciusallan[m]yeah. what you all suggest so we can make it mergeble until the end of the deadline?14:01
dviroelsince it would require more discussion around the approach14:01
* dviroel time check14:02
sean-k-mooneywiniciusallan[m]: move the resouce: related changes into its own patch at the end14:02
sean-k-mooneywe will merge the feature with just age initally14:02
sean-k-mooneyand then add resouce: with teh cdm and notifciaotn change slater14:02
dviroelwiniciusallan[m]: let us know if you have any question while working on that14:02
dviroelyou can ping us here in the irc14:03
dviroelthis is the last feature in the list14:03
dviroelwe are out of time now14:03
winiciusallan[m]ok. thanks!!14:03
dviroelnote that morenod added some changes in the Review topic14:03
morenodjust backports14:04
morenodnor urgents14:04
dviroelwe can get into them afterwards14:04
dviroelmorenod: ack, I was holding the backport yesterday due to another bugfix that was more critical14:04
dviroelbut it got merged now14:04
dviroeland we can continue with our backports14:04
dviroeland also with your tempest changes14:04
dviroelbut now we need to wrap up for today14:05
dviroelwe will meet again next week14:05
dviroelthank you all for participating14:05
dviroel#endmeeting14:05
opendevmeetMeeting ended Thu Aug 20 14:05:34 2026 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)14:05
opendevmeetMinutes:        https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-08-20-13.00.html14:05
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-08-20-13.00.txt14:05
opendevmeetLog:            https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-08-20-13.00.log.html14:05
amoralejthanks dviroel for chairing!14:05
morenoddviroel++14:06
opendevreviewMerged openstack/watcher-dashboard master: Add Playwright integration test for skip action workflow  https://review.opendev.org/c/openstack/watcher-dashboard/+/97659420:04

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!