Thursday, 2026-07-30

opendevreviewWinicius Allan Bezerra da Silva proposed openstack/watcher master: Add new WorkloadPreemption feature  https://review.opendev.org/c/openstack/watcher/+/99920203:26
opendevreviewWinicius Allan Bezerra da Silva proposed openstack/watcher master: Add new WorkloadPreemption feature  https://review.opendev.org/c/openstack/watcher/+/99920203:28
opendevreviewDavid proposed openstack/watcher-dashboard master: Fix audit list filter to use goal and strategy  https://review.opendev.org/c/openstack/watcher-dashboard/+/99904308:25
opendevreviewDavid proposed openstack/watcher-dashboard master: Switch to admin project before Playwright tests  https://review.opendev.org/c/openstack/watcher-dashboard/+/99796008:30
opendevreviewDavid proposed openstack/watcher-dashboard master: Fix audit list filter to use goal and strategy  https://review.opendev.org/c/openstack/watcher-dashboard/+/99904308:32
opendevreviewDavid proposed openstack/watcher-dashboard master: Switch to admin project before Playwright tests  https://review.opendev.org/c/openstack/watcher-dashboard/+/99796008:34
opendevreviewJoan Gilabert proposed openstack/watcher master: Change default database storage engine to InnoDB  https://review.opendev.org/c/openstack/watcher/+/99924108:50
opendevreviewDavid proposed openstack/watcher-tempest-plugin master: Update get_host_other_than to select hosts on the same host aggregate  https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/99925410:07
*** M00SE3 is now known as M00SE10:14
jgilaber#startmeeting watcher13:01
opendevmeetMeeting started Thu Jul 30 13:01:44 2026 UTC and is due to finish in 60 minutes.  The chair is jgilaber. Information about MeetBot at http://wiki.debian.org/MeetBot.13:01
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.13:01
opendevmeetThe meeting name has been set to 'watcher'13:01
dviroelo/13:01
sean-k-mooneyo/13:01
jgilaberhi! feel free to add topics to the agenda while we wait https://etherpad.opendev.org/p/openstack-watcher-irc-meeting#L3113:02
rlandyo/13:02
jgilabercourtesy ping list: amoralej chandankumar morenod13:02
morenodo/13:02
jgilaberok, let's start with the first topic from dviroel 13:03
jgilaber#topic Event Audit tests in watcher-tempest-plugin13:04
dviroelack13:04
dviroelyeah, I added this one13:04
dviroeljust to raise that I was looking at that13:04
dviroeland as we know, there isn't event type audit testing coverage in our plugin13:04
dviroeli was looking how feasible is to have it tested in our ci13:05
chandankumaro/13:05
dviroeli shouldn't be hard to get an aodh alarm to trigger an event audit, with a dummy strategy13:05
dviroelthat would give us a minimal test coverage, which we can expand afterwrds13:06
opendevreviewDavid proposed openstack/watcher master: Add strategy as filter on /v1/audits/detail endpoint  https://review.opendev.org/c/openstack/watcher/+/99909413:06
dviroelsounds interesting for you folks?13:06
dviroelI may work on next week or so...13:07
amoralejo/13:07
dviroelwhen I have some time13:07
dviroelbut in order to create the alarm with aodh13:07
dviroelwe would need to set up a client13:07
dviroelor create a dependency on telemetry plugin13:07
amoralejbut we have an api13:07
amoraleji thing the trigger based on alarms should not be in watcher13:08
amoralejbut the other way around13:08
amoralejit'd aodh or whatever external system who should call the watcher api to trigger the audit13:09
amoralejimo13:09
dviroelcorrect13:09
dviroelthe idea of the scenario test is to have the aodh triggering the alarm13:09
amoraleji think for the scenario test we should just trigger the alarm from api13:10
sean-k-mooneydviroel: we dont actully need a depency on the telemetry plugin13:10
sean-k-mooneywe can jsut use the tempest client ot create it ourselve13:11
sean-k-mooneyso i think that proably the better aprpoch13:11
dviroelamoralej: you mean to call our webhook POST api?13:11
amoralejyes13:11
sean-k-mooneywe can do that too13:11
sean-k-mooneywe shoudl test that as well13:11
dviroelsean-k-mooney: we would need to create ourselves then, since we don't have the client in tempest only in the plugin13:11
sean-k-mooneyyes13:12
sean-k-mooneyso i woudl start with just callign the webhook directly13:12
sean-k-mooneywe can alwasy add end to end testing with adoh after13:12
opendevreviewMerged openstack/watcher-dashboard master: Switch to admin project before Playwright tests  https://review.opendev.org/c/openstack/watcher-dashboard/+/99796013:12
dviroelyeah, calling directly the api would allow us to expand out tests13:12
dviroelbut in a scenario test i would also expect a service calling it, for integration purpose13:13
amoralejyep, for me that's second step, as we should start with the more open case and we can then run more specific and complex cases13:13
sean-k-mooneyas i said we can do both13:13
sean-k-mooneyevent bases audits are not inteded to only be called by openstack services13:14
sean-k-mooneythey can be called by external monitoring systems13:14
amoralejbtw, i think in aodh you can register webhooks urls directly in the alarms?13:14
sean-k-mooneywiht or without keystoen auth13:14
dviroelamoralej: yes13:14
sean-k-mooneyi.e. using alert manager in promethous or nagios alarms13:14
amoraleji don't know much about that tbh13:14
amoraleji mean we may not need a client, just to configure the alarm13:15
sean-k-mooneyamoralej: yep but you can also just have a cron jobs that dues a curl request13:15
amoralejyep13:15
sean-k-mooneythe event based triger is not exipclty for aodh13:15
sean-k-mooneyso we shoudl test it both ways13:15
amoralejexactly, that's why i think we should start just by calling the webhook directly13:16
sean-k-mooney+113:16
dviroelright, but I think that in scenarion test we should have a service validating it13:16
sean-k-mooneyno senairo test does not mean only via prject13:16
sean-k-mooneysenairo test just mean any test with multiple stages13:16
sean-k-mooneyi.e. a non trival api test13:17
amoralejwatcher-tempest-plugin should focus on validating watcher behavior13:17
sean-k-mooneyif you want to condier the drict call a api test that fine13:17
dviroelthat why i said we should13:17
dviroelnot must13:17
sean-k-mooneyso tempest plugins shoudl be generally frestandign13:17
amoralejadding the aodh layer is nice as integration between aodh and watcher, but as i said next step13:17
sean-k-mooneyi.e. not depend on each other13:17
opendevreviewDavid proposed openstack/watcher-dashboard master: Add Playwright tests to validate goals and strategies  https://review.opendev.org/c/openstack/watcher-dashboard/+/99606713:17
dviroelso we know that at least one other service could be integrated with watch13:17
sean-k-mooneyso we can add aodh related tests to ours13:18
sean-k-mooneybut we need to mage them check the service enabeld config option in tempst13:18
sean-k-mooneyand only run if aodh is deployed and enabeld there13:18
amoralejactually, adding it to aodh tempest plugin may be more appropiate :) 13:18
sean-k-mooneymaybe although im not sure13:18
sean-k-mooneythey woudl still have to build a watcher clinet13:18
dviroelso my point is: we can validate that our API works, but would be good to know which other services could be used with watcher13:19
dviroeli prefer to land in our repo, where we control things13:19
amoralejbut that may expand to whatever customers may decide. aodh may be a good and valid case, but not the only one13:19
dviroelcan be aodh/prometheus13:19
dviroelaodh i know that works :)13:20
dviroelit was the only one that I tested13:20
dviroeli may try with prometheus too as sean-k-mooney mentioned13:20
dviroelthat would work too13:20
dviroeli think that we agree then13:21
amoralejbut my point is, what if prometheus, i.e. doesn't work for any reason? if the api works as expected, that's our scope13:21
sean-k-mooneyfor promethious currently you woudl have to disabel keystone auth since i dont think its alarm magner thign can pass a keyton token13:21
amoralejit may be a problem in prometheus or whatever13:21
sean-k-mooneyso we woudl have to test that in a diffent job13:21
amoraleji doubt we should validate alertmanater in prometheus tbh13:21
dviroelamoralej: if doesn't work or if is broken doesn't mean that is an issue with watcher13:23
dviroelcorrect13:23
sean-k-mooneyi just mention it because the orgianl intent of event based audit13:24
dviroelyeah13:24
sean-k-mooneywas for an external syste to deciend a pre defiend audit shoudl be triggered13:24
sean-k-mooneyrhat then using simple cron triggrees13:24
sean-k-mooneyin tempst we can emulate that with a direct api call13:24
sean-k-mooneyso i htink that enouch for "external triggered" we can special case aodh13:25
sean-k-mooneybecause that is the native openstack consumer13:25
sean-k-mooneyone intersitng usecase long term for this13:25
amoralejwhat it'd be good is to clearly document how those can be called and requirements (i.e. how to manage auth). And yeah, as aodh is openstack, it deserves special treatment13:25
sean-k-mooneyis if we compelt the premetable intnsce work, i belive aodh can triger on a no valid host shcudling failr form nova13:26
dviroelyeah, i've agree with your points13:26
sean-k-mooneyso when nova raises no valid host, we coudl use aodh to tell watcher to reap premtable instnaces via an event trigger13:26
dviroelwhat bothers me is: do we know if that works with aodh? we are not sure...13:26
dviroelmaybe a ci test could periodically check that13:27
sean-k-mooneyim not agaisnt the aodh senario test just to be clear13:27
dviroelsean-k-mooney: yep13:27
sean-k-mooneyi think you can and shoudl write that13:27
sean-k-mooneybut if that was broekn it woudl likely be an aodh bug not a watcher one13:27
sean-k-mooneyhaving the direct call test tells use which it is13:28
dviroelyeah, and we would need to skip the test in that case in our ci13:28
dviroelbut still worth keeping it I think13:28
dviroelin case that happens13:28
sean-k-mooneyi.e. if direct call passes and aodh does not then it an aodh bug, if but break it our bug13:28
sean-k-mooneywe can readd aodh to one of our jobs13:28
dviroeland depending on the aodh issue, we may remove the job and call that aodh can't integrate with our api13:28
sean-k-mooneyit used to be in the job but we removed it because we were not using it13:28
sean-k-mooneyi think you shoudl just wrtie the patch and we can review them13:29
dviroelack, we can also evaluate that too13:29
dviroelack13:29
dviroeli think that we agree13:29
sean-k-mooney+113:30
dviroeltests directly calling the api13:30
dviroeland an additional test with aodh13:30
amoralejyep13:30
dviroeli may work on that next week, depending on how audit-pipeline goes13:30
dviroelthanks for the feedbacks13:30
jgilaberthanks for bringing this up dviroel, other questions/comments on this topic?13:31
dviroel++13:32
jgilaberok moving to reviews13:32
jgilaber#topic Reviews13:32
jgilaber#link https://etherpad.opendev.org/p/watcher-2026.2-status13:33
jgilaberwe've made some progress on the cidnerclient to sdk work, thanks dviroel amoralej ++ for the reviews13:33
jgilaberonly two patches left for that series13:33
jgilaberanyone wants to highligth anything else?13:33
dviroeli would ask another reviewer to look at https://review.opendev.org/c/openstack/python-watcherclient/+/99366113:34
dviroelmayve jgilaber since you reviewed the watcher patch, which merged13:34
dviroelthanks amoralej and jgilaber for reviewing the audit template default_parameters patch13:34
jgilabersure I'll review it13:35
dviroelother than that, i will submit new patches for audit-pipeline13:35
dviroelthere were some changes while testing the audit pipeline handler...13:35
dviroeli should bring them next week13:35
amoralejI'll review the patch adding the db and objects13:36
dviroelthanks amoralej++13:36
dviroelI saw that Winicius proposed a patch too, i will add to the review etherpad if not there yet13:36
dviroel#link https://review.opendev.org/c/openstack/watcher/+/99920213:36
dviroelI didn't check that patch yet too13:37
dviroelrelated to the preemtible instances feature13:37
amoraleji did a high level review and recommended to split in several patches13:37
sean-k-mooneyoh cool13:37
sean-k-mooneyyes that woudl help with review if nothign else13:37
dviroelthanks amoralej - yeah, we should assist since it is his first feature proposal in watcher 13:38
amoraleji have doubts about splitting goal and strategy, but for actions they should be first independen reviews imo13:38
sean-k-mooneyi woudl preoably do the appler chagne first to add the actions13:38
sean-k-mooneymaybe the planner change next and finally the addtion of the stragy and goal13:38
amoralejwhich applier change?13:38
sean-k-mooneythe applier chagne are adding delete and shelve actions13:38
sean-k-mooneythe goal/stragey defintion need to be last because that makes it viabale at the api13:39
amoraleji think that should be after actions are added imo, given that the refer the action name as registered13:39
amoralejyep13:39
sean-k-mooneyyep so i think we agree on the split13:39
sean-k-mooneyim just looking at the file ist for now but ill  try an loop back to this properly later13:39
dviroelnews actions could also be used/tested with actuator too, in case we don't merge the new strategy, new action are still valid13:40
dviroelbut yeah, we can split the review13:40
sean-k-mooneyyes that is true13:40
dviroelalso easier to review and update13:40
sean-k-mooneyalthough that would a sperte tempest change13:40
dviroelyeah13:40
sean-k-mooneyspeakign of we shoudl also wait for the tempest test before merging the sereis13:41
sean-k-mooneybut it s a good start13:41
amoralejyep13:41
sean-k-mooneyamoralej: i agree with your comment by the way and the docs13:41
sean-k-mooneyit also will need a release not in the last commit13:41
dviroel+113:41
amoralejyes13:41
sean-k-mooneyanyway ill note that in the review13:42
jgilaberanything else from the status etherpad?13:42
amoralejI'd like to mention https://review.opendev.org/c/openstack/watcher/+/993088/ , i finally implemented the cache in the model as suggested by dviroel 13:43
amoralejso, i also splitted in two reviews, one for the change in the model and one for the strategy13:43
dviroelyeah, I saw that in yesterday, but I still need to review in details13:43
amoralejsure, np13:44
dviroelthanks amoralej 13:44
amoralejin the one for the cache, note https://review.opendev.org/c/openstack/watcher/+/998883/2/watcher/decision_engine/strategy/strategies/base.py13:44
amoraleji'm restarting the cache when creating the strategy. tbh, it should be always be empty at that point as the cache is only populated by calls that are done by the strategies13:45
amoralejbut just as a defensive approach13:46
dviroelyeah, iirc the model is copied at every strategy execution, but i had to check that in the code again13:46
amoralejyes, exactly13:46
dviroelbut it is good to clear things that we know that should't be reused13:46
sean-k-mooneyim not sure that reseting it for every execution fo a stagy will be correct13:47
sean-k-mooneywhen it comes to the pipelining13:47
sean-k-mooneybut for normal audits yes13:47
dviroelin audit-pipeline I will reuse the cluster data model13:47
sean-k-mooneythe point of caching in the pipeline is that the metric for a give hsot for the rist stragy shoudl be ruesed for the second13:47
dviroelbut the strategy neves starts with a model == None in my case13:48
sean-k-mooneyi.e. the subsequet step in the pipelien soudl not get new raw values13:48
dviroelyeah, I need to carefully check that13:48
sean-k-mooneyif there are cased valeus13:48
sean-k-mooneyack13:48
amoralejeven in that case, this cache should be correct13:48
sean-k-mooneyok well we can revist when lookign at the pipeline part13:49
dviroelyes13:49
amoralejas i understand the operations on the model are done with the model methods migrate_instance, etc...13:49
dviroelamoralej: correct13:49
amoralejanyway, note that this cache is only for the used resouces by nodes13:49
dviroelthe simulation uses cdm migrate_instance13:49
amoralejit can be cleaned and automatically recalculated at any time13:50
dviroelamoralej: yes, i have a different cache for metrics13:50
amoralejit's only performance13:50
amoralejyep13:50
dviroel+113:50
jgilaberok to move to next review?13:50
dviroeljgilaber: yes13:50
jgilaberI have a quick one 13:51
dviroeldo we can cover the following ones13:51
jgilaber#link https://review.opendev.org/c/openstack/watcher/+/99924113:51
jgilaberit changes the default storage engine for mysql13:51
sean-k-mooneyi was wonderign if we shoudl be more explcity about hte default CHARACTER SET as well but tis proably fine13:52
dviroeljust checking that https://codesearch.opendev.org/?q=default_storage_engine%3DMYISAM&i=nope&literal=nope&files=&excludeFiles=&repos=13:52
dviroelwhich seems the default in many other projects13:53
sean-k-mooneyMYSIAM technially does not supprot atopic tranasctions13:53
sean-k-mooneyso it not somethign that we would ever use in proudction for watcher13:53
jgilaberit looks like the script was copied from project to project13:53
jgilaberwe actually noticed when I copied the script to cyborg13:54
sean-k-mooneyMYISAM is faster as a result but InnoDB is what we shoudl be using13:54
dviroelyeah, just wondering why other projects didn't  migrate 13:54
sean-k-mooneyit works so we just never got aroudn to changing it13:55
amoraleji though there may be something is oslo-db or something forcing to use innodb tbh13:55
sean-k-mooneythere is no reason beyond that13:55
dviroelok13:55
sean-k-mooneyamoralej: i dont think so but i think we can spcyfiy it vai alembic13:55
amoralejhttps://github.com/openstack/watcher/blob/a1b113ce54a5c4c1dae794488227af11e8e4499f/watcher/conf/db.py#L32-L34 ?13:55
sean-k-mooneyanyway this is a harmle test only patch, i asked jgilaber  ot use InnoDB for cybrog when teim suggeted it13:56
amoralejyep13:56
jgilaberthank folks, let's continue discussing in the patch so we can finish with the agenda13:56
sean-k-mooney yes13:56
dviroelack, jgilaber sure13:56
sean-k-mooneythat will default it for use too13:56
jgilaberlast review patch is from morenod 13:56
jgilaber#link https://review.opendev.org/c/openstack/watcher-dashboard/+/99606713:57
morenodthis is just to verify that we see the same information on horizon than api for goals and strategies13:57
dviroeli was reviewing this one atm13:57
dviroelchecking the snapshots and so 13:57
dviroeli will vote soon in this one13:58
dviroelthanks morenod 13:58
morenodthanks dviroel 13:58
jgilaberwe have one bug to triage, do we want to try quickly or leave it for next week?13:58
jgilaberI think we can leave it for next week, since we're out of time13:59
jgilaberlast thing13:59
jgilaber#topic Volunteers to chair next week13:59
jgilaberany volunteer? I'll be out next week13:59
dviroelI can13:59
jgilaberthanks dviroel ++14:00
jgilaberand thanks all for joining!14:00
jgilaber#endmeeting14:00
opendevmeetMeeting ended Thu Jul 30 14:00:26 2026 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)14:00
opendevmeetMinutes:        https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-07-30-13.01.html14:00
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-07-30-13.01.txt14:00
opendevmeetLog:            https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-07-30-13.01.log.html14:00
dviroelthank jgilaber++14:01
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Fix race condition in DefaultLoader._reload_config()  https://review.opendev.org/c/openstack/watcher/+/99484515:50
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Add functional test framework for Watcher  https://review.opendev.org/c/openstack/watcher/+/98838915:50
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Add Nova/Placement emulators and host_maintenance functional tests  https://review.opendev.org/c/openstack/watcher/+/99335215:50
amoralejdviroel, sean-k-mooney ^ i've sent new PSs with your feedback ^ I will not update the prometheus and cinder ones until you approve the approach15:53
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Add functional test framework for Watcher  https://review.opendev.org/c/openstack/watcher/+/98838916:11
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Add Nova/Placement emulators and host_maintenance functional tests  https://review.opendev.org/c/openstack/watcher/+/99335216:11
opendevreviewAlfredo Moralejo proposed openstack/watcher master: Add Nova/Placement emulators and host_maintenance functional tests  https://review.opendev.org/c/openstack/watcher/+/99335216:35
dviroelack amoralej 16:47
opendevreviewMerged openstack/watcher-dashboard master: Add Playwright tests to validate goals and strategies  https://review.opendev.org/c/openstack/watcher-dashboard/+/99606717:11
opendevreviewMerged openstack/python-watcherclient master: Add default_parameters support to Audit Template  https://review.opendev.org/c/openstack/python-watcherclient/+/99366117:50

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