Thursday, 2025-12-04

opendevreviewDavid proposed openstack/watcher master: [DNM] Testing nodeset with three nodes (two computes + 1 controller)  https://review.opendev.org/c/openstack/watcher/+/96733109:44
jgilaberhello! just a reminder that we'll have our irc meeting in ~ 20 minutes, feel free to add topics to the agenda https://etherpad.opendev.org/p/openstack-watcher-irc-meeting11:42
jgilaberwatcher meeting strating in two minutes11:59
jgilaber#startmeeting watcher12:01
opendevmeetMeeting started Thu Dec  4 12:01:54 2025 UTC and is due to finish in 60 minutes.  The chair is jgilaber. Information about MeetBot at http://wiki.debian.org/MeetBot.12:01
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.12:01
opendevmeetThe meeting name has been set to 'watcher'12:01
jgilaberhi all o/, how is around today?12:02
rlandyo/12:02
amoralejo/12:02
dviroelo/12:02
jgilabercourtesy ping: sean-k-mooney chandankumar morenod12:03
morenodo/12:03
sean-k-mooneyo/12:03
jgilaberthanks for joining, let's start with today's agenda12:03
jgilaber#link https://etherpad.opendev.org/p/openstack-watcher-irc-meeting#L26 (Meeting agenda)12:03
jgilaberfeel free to add your own topics to the agenda12:04
jgilaber#topic new blueprint for openstacksdk migration12:04
jgilaberthis was carried from last week12:04
jgilaberI created a blueprint and a spec to move watcher to use openstacksdk as discussed during the PTG and other meetings12:05
jgilaber#link https://blueprints.launchpad.net/watcher/+spec/use-openstacksdk12:05
jgilaber#link https://review.opendev.org/c/openstack/watcher-specs/+/96802312:05
jgilaberthis is the first time I created a blueprint, so I'm not sure if it's missing anything12:05
jgilaberdo we need to do anything with the blueprint? if not I'll just use the opportunity to ask for reviews on the spec12:08
sean-k-mooneyjgilaber: that looks ok12:08
sean-k-mooneywhen we ahve a spec the bluepirnt is often just the first paragah or a cut down version of the problem description12:09
jgilaberack, thanks, if there are no further questions/comments we can move on the next topic from dviroel 12:10
jgilaber#topic spec proposal freeze/spec freeze12:10
dviroeljust added this one 12:10
sean-k-mooneythe default would be milestone 212:10
dviroelso we can agree on a deadline to get our specs for this release merged12:10
dviroelwhich is Jan 05 - Jan 09 (m2)12:11
sean-k-mooneywhich is january 8th ish12:11
dviroel#link https://releases.openstack.org/gazpacho/schedule.html12:11
sean-k-mooneyhttps://docs.openstack.org/watcher/latest/contributor/release-guide.html#milestone-212:11
jgilaberI think that sound reasonable, we have currently 3 open specs12:11
dviroeldue to holidays we may want to get most of them merged asap12:11
jgilaber#link https://review.opendev.org/q/project:openstack/watcher-specs+status:open12:11
dviroeljgilaber: yeah, I plan to send another one in a couple of minutes12:12
dviroeldo we want to have a proposal freeze too?12:12
sean-k-mooneyyep sooner rahter then later. nova pulled theres in to today to see if that can help with apporving less and compelting more12:12
sean-k-mooneydviroel: no12:12
sean-k-mooneymost teams dont12:12
dviroelto have some time to review them..12:13
sean-k-mooneywe coudl ahve one but i dont know that we have the volume to do it12:13
amoralejI may need to create a new one for instrumenting the decision engine with osprofiler12:13
amoralejin this cycle12:13
amoraleji guess I'll need a spec for that, right?12:14
sean-k-mooneywhat might be more intersting is do we want an early Q1 "mid cycle" or check in for what we want to complete this cycel12:14
sean-k-mooneyamoralej: that might qualify for a specless blueprint12:14
sean-k-mooneyprovide tehre is no upgrade impact (its off by defualt)12:14
amoralejright12:14
sean-k-mooneyas it not really user facing12:14
amoralejok, I'll prepare the blueprint for next meeting hopefully12:15
jgilabersean-k-mooney, that "mid cycle" would be like a sort of PTG session to go over the status?12:15
sean-k-mooneyyep 12:16
jgilaberI like the idea rather than trying to impose a hard freeze12:16
sean-k-mooneywell im suggeeting it after the hard freeze12:16
sean-k-mooneywe acutlly have more approved already then i think we can complete12:16
dviroelthey are 2 different things I think12:17
jgilaberoh ok, I misunderstood then12:17
sean-k-mooneythis is our action item list form the ptg https://etherpad.opendev.org/p/watcher-2026.1-ptg#L5612:17
sean-k-mooneyin theory we shoudl be tryign to compelte thos as well this cycle12:17
sean-k-mooneyso one of the things i was thinking we woudl do is review how much fo that we got done and what to compelte12:18
dviroelok, just wanted to be clear on deadlines for specs12:18
dviroelbecause there are holidays right before12:19
sean-k-mooneyright so in our contibutor docs is docuemneted spec freeze as mileston 212:19
sean-k-mooneyso unless w eexpclity opt out of that that woudl be the freze for every cycle12:19
dviroeland we may want to accept/merge specs asap12:19
dviroeli may be on pto the first week of jan, so I would review all open in 2025 :) 12:20
jgilaberso we have roughly 2 full weeks before people start going on PTO12:20
sean-k-mooneyeffectivly yes12:20
jgilaberwe should try to prioritize reviewing the existing specs, but it seems doable12:21
sean-k-mooneyi will likely finish december 20th and return january 5th12:21
sean-k-mooneythe exact dates are tbd12:22
dviroeljgilaber: the idea was to make deadlines clear to contributors and reviewers, so I don't have anything else in this topic12:23
jgilaberack thanks for bringing that up dviroel!12:24
jgilaberif there are no more comments we can move to reviews12:24
jgilaber#topci reviews12:24
jgilaberfirst one12:24
jgilaber#link     https://review.opendev.org/c/openstack/watcher/+/968025 12:24
jgilaberamoralej, do you have anything to cover for this patch?12:25
sean-k-mooneyoh the action docs cool ill add that to my review list12:25
amoralejjust to mention it as i expect it to be quick to review and is blocking other stuff on top12:25
jgilaberack, thanks we can move to the second one then12:26
sean-k-mooneywell you can rebase the other on top12:26
jgilaber#link https://review.opendev.org/c/openstack/watcher/+/96398112:26
sean-k-mooneybut ya i think we can land that soon12:26
jgilaberthe second one has +2 from dviroel and me12:26
sean-k-mooneyyep so dviroel shoudl ahve added +w unless they were askign for a change12:26
amoralejthat one already has 2 +2, it'd be great of i can also get +w12:27
jgilaberI was looking if we were waiting for anything in particular12:27
sean-k-mooneyi just added it12:27
dviroelI still reviewing 963252, have some comments to sed12:27
dviroelsean-k-mooney: i gave a chance for more people to review it, but yeah, we can +w12:27
amoralejack12:28
amoralejso, that was it from me12:28
jgilaberthanks, we can move to bug triage then12:28
jgilaber#topic Bug Triage12:28
jgilaber#link https://bugs.launchpad.net/watcher/+bug/212969212:28
opendevreviewDavid proposed openstack/watcher master: [DNM] Testing nodeset with three nodes (two computes + 1 controller)  https://review.opendev.org/c/openstack/watcher/+/96733112:29
jgilaberI opened this bug a while back for zone migration12:29
jgilaberI'm not sure if we've covered it before12:29
jgilaberbut in short after we added support for the 'src_type' parameter, I realized the strategy was not taking it into account when selecting the destination for the migration12:30
sean-k-mooneytis not really a bug in the normal sense so at most i would triage it as wishlist12:30
sean-k-mooneywell sorry12:30
sean-k-mooneyim mixing linkgs 12:30
sean-k-mooneythat bug is not for davids ci change12:31
jgilabernope, it's https://bugs.launchpad.net/watcher/+bug/212969212:31
sean-k-mooneyso "Zone migration does not select volumes properly according to user input"12:31
sean-k-mooneyi dont thinkn i have looked at that before12:31
jgilaberright12:31
sean-k-mooneythis was somethign you found when doing the other fixes12:31
jgilaberI found it after we landed that series of patches actually12:32
jgilaberbasically if a user wanted to migrate some volumes that are in the same pool but have different types to different destination depending in the type12:32
sean-k-mooneyyou have already started on 2 patchs, so i think we can consier it validi and you shoudl assign it to yourself12:33
jgilaberit would not work properly12:33
sean-k-mooneydo you have a feelign on the prioty 12:33
jgilaberright, I though I had assigned it already12:33
jgilaberI think it's somewhat between medium and high12:33
sean-k-mooneyoh right this is about doing a retry and a migrate at the same time12:33
sean-k-mooneywell i dont think that is supproted in cinder12:33
sean-k-mooneyyour can rety btu i dont think you can specify the strage backend when you do a retry can you?12:34
jgilaberno, this is about selecting the destination pool12:34
sean-k-mooneywe dont supprot have differnet voluem go to diffent destionats in the same audit today12:34
amoralejso, the problem is that, if ths src_pool is repeated with different src_type values, it does not work12:35
jgilaberwhy not?12:35
amoralejright?12:36
sean-k-mooneydoesnet the volume migrate part refer to all voluems?12:36
jgilaberyes, amoralej that is a good summary12:36
jgilaberin zone migration we evaluate each volume migration individually https://github.com/openstack/watcher/blob/e427fa68a379cbff4641c937b71d353d9e6387a0/watcher/decision_engine/strategy/strategies/zone_migration.py#L40912:37
amoralejyep, looks like a bug12:37
jgilaberso for each volume to consider we get the dst_pool and dst_type and crate the appropiate action (retype or migrate) according to that plus the volume type12:38
sean-k-mooneyhttps://github.com/openstack/watcher/blob/master/watcher/decision_engine/strategy/strategies/zone_migration.py#L89-L14712:38
sean-k-mooneyso if you look at the schem today12:38
sean-k-mooneyare you saying you think we shoudl supprot haveing12:38
sean-k-mooney2 entries with the same srouce oool12:39
sean-k-mooneyand different soruce types and dest pools12:39
jgilaberyes, I think there could be use cases for that12:39
sean-k-mooneyyou can technically express that in the schdme but i didnt thinkw e had any documented exampels or test to supprot that12:39
jgilaberI put one in the bug description, not sure how realistic it is12:40
sean-k-mooneyi mean i guess its reasonabel if you wanted to split or merge you high performace and mass storage voluem 12:40
amoralejyep ^ that's a valid case12:40
sean-k-mooneyi.e. start with a san that is use for both and hten split them12:40
sean-k-mooneyor do the reverse12:41
jgilaberthat right now passes the schema validation12:41
jgilaberbut the strategy does not do what the user I think would expect12:41
sean-k-mooneyyes but the input validation in the api is not great12:41
sean-k-mooneywe coudl supprot this but it feel more like an incompelte feature12:42
sean-k-mooneyi.e. they never finished implementing it12:42
sean-k-mooneydid you check fi this was described in the spec for zone migration12:42
jgilaberyes, I feel the same12:42
dviroelso maybe we should work on a schema validation of what is supported? and then improve the strategy to support different combinations?12:42
sean-k-mooneywell not everything shoudl be checked in the schema12:42
jgilaberI don't think we can validate this case in the schema12:42
sean-k-mooneythe schmea is for syntax not semantics12:43
sean-k-mooneysemantics are validate after the schma validation12:43
dviroelthat would be validate in python then12:43
sean-k-mooneyso i think we coud supprot this, im not sure we would backprot this however12:43
sean-k-mooneyi.e. this feels like something that is valid to fix on master but not valid to include in stable branches12:44
jgilaberthe way the strategy works I think it's kind of implicit that it should be supported, but it was not tested, like many other things12:44
jgilabersounds good to me to make it master only12:44
sean-k-mooneyif its not tested and there are no docs i default to implecit is not enough 12:44
sean-k-mooneylets see how invaisve it end up being12:44
sean-k-mooneyand we can think about the backportablity then based on that12:45
amoralejjgilaber, is https://review.opendev.org/c/openstack/watcher/+/964717 enough to make that use case work fine?12:45
jgilaberthe fix is quite small, just 14 lines of code12:45
sean-k-mooneyoh right you have 2 patches up12:45
jgilaberamoralej, no in that patch I only added tests to reproduce12:46
amoralejsorry, i mean in the serie12:46
amoralejhttps://review.opendev.org/c/openstack/watcher/+/964718/12:46
jgilaberyes, the second patch in the series is the fix12:46
jgilaberI don't expect anything else would be needed12:46
sean-k-mooneyoh well lets mark it as medium i guess, jgilaber can you assing it to your self and then we can deciend based on teh review how to proceed once its fixed on master12:46
jgilabersure, mediums seems fine12:46
jgilaberthanks for the discussion! I think we're done with this one12:47
jgilaberif there no more questions/comments we can continue with the other bugs12:47
sean-k-mooneywe have been bit by pelople abusing input validation bugs in the past to achive something that we never inteded to supprot creating persitent tech debt12:47
sean-k-mooneyso im alwasy causious with enablign supprot for a fucntionality if it was not documented as intended as a result12:47
jgilaberI can see how that could become a problem12:48
sean-k-mooneyhttps://bugs.launchpad.net/nova/+bug/191528212:48
sean-k-mooneythat was orginally a secuity bug that we could not fix becuase folks built a feature in neutron that depended on it12:49
jgilaberthat sounds like quite a headache12:50
jgilaberif there no more comments, we can cover the second bug12:50
jgilaber#link https://bugs.launchpad.net/watcher/+bug/213350512:50
jgilaberthis one is from dviroel 12:50
dviroelah12:51
dviroelI opened as a RFE12:51
dviroelto track the limitation on applier running in native thread mode12:52
sean-k-mooneyah so as a tracker to note we need to eventually fix that12:52
dviroelsince we merged the support for native threads in applier already12:52
dviroelso yes, we need to fix that before removing eventlet code12:52
sean-k-mooneyi dont reacall if we left a fixme or todo comment for that in the code but ya we should eventrually fix that12:52
jgilaberso we should mark that as high right?12:53
sean-k-mooneyi would say maybe hight ya12:53
sean-k-mooneyalthough the actul fix migh tnot be somethign we end up tracking as a bug depending on what ti becomes12:53
dviroelhigh should be fine, since it is really important to fix12:53
jgilaberso for now we mark as triaged+high, agreed?12:53
sean-k-mooney+112:53
dviroelyeah, it may touch many places in the code12:53
dviroel+112:53
jgilaberok, done, thanks for tracking that dviroel 12:54
jgilaberwe have a couple more bugs12:54
jgilaber#link https://bugs.launchpad.net/watcher-tempest-plugin/+bug/213318112:54
dviroelthe next one is a tempest plugin 12:54
dviroelwhich sporadically appears in ci12:54
dviroelaction plan is created as defailt as Recommended 12:55
dviroeland then goes to the correct state12:55
dviroeli.e.: if there is no actions, it is moved to Succeeded12:55
sean-k-mooneywell recommend its not invalid12:55
sean-k-mooneyi assume we are using the auto apply12:55
sean-k-mooneythe tempest test shoudl handel that and keep checkign if its recommend until it get to succeeded ro canceled12:56
jgilaberit seems that that is precisely what the fix does12:57
jgilaber#link https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/96875012:57
dviroelthere is a fix propose yes12:57
sean-k-mooneyyep more or less12:57
sean-k-mooneyits wiating for it to be in a terminal state 12:57
sean-k-mooneywhic avoid the intermidary value12:57
sean-k-mooneyso that look correct to me12:58
sean-k-mooneyalthough im unsure about the valueeror12:58
dviroelyeah, and after that there is the real check for the expected state12:58
sean-k-mooneyhow/when is that raised12:58
dviroelhum, that's a copy and paste from other places, but I can double check if is correct 12:59
sean-k-mooneyi hope we are not rasing ValueError to signal that call_until_ture timeed out12:59
sean-k-mooneywell its ok to raise an exepcitn of we time out12:59
sean-k-mooneyit just shoudl be StopIteration or simialr12:59
sean-k-mooneyor we shoudl just return false13:00
sean-k-mooneyin either case i dont thnk we need the try excpet13:00
sean-k-mooneyif we rase any expction it will fail the test13:00
sean-k-mooneyif we return false we will faill the assertTrue13:00
sean-k-mooneythis give a slightly better errror message but that all13:01
dviroelyeah ack13:01
* dviroel time check13:01
dviroelso yeah, pls review this one13:01
jgilaberthanks dviroel 13:01
sean-k-mooneywe can probaly proceed as it is and look at this laster13:01
dviroelan chandan's one, that is the next on the list13:01
jgilaberwe can continue the discussion on the patch13:02
jgilaberwe have one final bug13:02
jgilaber#link https://bugs.launchpad.net/watcher-tempest-plugin/+bug/213377713:02
jgilaberthat is also an issue seen in ci13:02
jgilaberwhich also has a fix open 13:02
jgilaber#link https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/96936613:02
jgilabersince we're out of time we can discuss them in the reviews13:02
dviroelyep13:03
jgilaberlast topic13:03
jgilaber#topic Volunteers to chair next meeting13:03
amoraleji can take it13:03
chandankumari can take it13:03
jgilaberamoralej is our lucky winner :) thanks!13:03
* dviroel can take it too :) 13:03
jgilaberthat's all for today, thanks everyone!13:03
dviroelthanks jgilaber o/13:04
jgilaber#endmeeting13:04
opendevmeetMeeting ended Thu Dec  4 13:04:04 2025 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)13:04
opendevmeetMinutes:        https://meetings.opendev.org/meetings/watcher/2025/watcher.2025-12-04-12.01.html13:04
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/watcher/2025/watcher.2025-12-04-12.01.txt13:04
opendevmeetLog:            https://meetings.opendev.org/meetings/watcher/2025/watcher.2025-12-04-12.01.log.html13:04
morenodthanks jgilaber++!13:04
amoralejthanks jgilaber++ 13:04
opendevreviewDavid proposed openstack/watcher master: [DNM] Testing nodeset with three nodes (two computes + 1 controller)  https://review.opendev.org/c/openstack/watcher/+/96733116:14
opendevreviewDouglas Viroel proposed openstack/watcher-specs master: Add spec for Strategy Pipeline feature  https://review.opendev.org/c/openstack/watcher-specs/+/96984019:23

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