Thursday, 2026-11-19

opendevreviewHo Minh Quang Ngo proposed openstack/watcher-tempest-plugin master: Pass Tempest client defaults to Watcher clients  https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/101643308:35
opendevreviewVictoria Martinez de la Cruz proposed openstack/grian-ui master: Add shared helpers for the Telemetry dashboards  https://review.opendev.org/c/openstack/grian-ui/+/101631410:01
opendevreviewVictoria Martinez de la Cruz proposed openstack/grian-ui master: Add compute host health dashboard  https://review.opendev.org/c/openstack/grian-ui/+/101644510:06
opendevreviewJoan Gilabert proposed openstack/watcher master: Fix Hypervisor attribute access in saving_energy  https://review.opendev.org/c/openstack/watcher/+/99938711:51
opendevreviewJoan Gilabert proposed openstack/watcher master: Deprecate Ironic integration  https://review.opendev.org/c/openstack/watcher/+/101621111:51
opendevreviewWinicius Allan Bezerra da Silva proposed openstack/watcher-specs master: Add Preemptible Instances spec for 2027.1 cycle  https://review.opendev.org/c/openstack/watcher-specs/+/101425212:49
dviroelheads up team, watcher meeting will start in 512:55
dviroel#startmeeting watcher13:00
opendevmeetMeeting started Thu Nov 19 13:00:09 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
jgilabero/13:00
dviroelhi o/ - who is around for today's meeting?13:00
morenodo/13:00
dviroelcourtesy ping:  amoralej sean-k-mooney chandankumar13:01
dviroellet's start with today's meeting agenda13:01
dviroel#link https://etherpad.opendev.org/p/openstack-watcher-irc-meeting#L29 (Meeting agenda)13:01
dviroelfeel free13:02
dviroelto add your own topics to the agenda13:02
amoralej__o/13:02
dviroellets start with13:02
dviroel#topic Started working on ironic deprecation13:02
dviroelnice, jgilaber has some updates to share13:02
winiciusallan[m]o/13:03
dviroel#link https://review.opendev.org/c/openstack/watcher/+/1016211 (1016211: Deprecate Ironic integration)13:03
dviroelif we deprecate in 2027.113:03
dviroelwe could remove in 2027.213:03
dviroelsince indri is a slurp release13:04
jgilaberyes, amoralej__ pointed that out in the patch13:04
jgilaberI originally did not set any removal version13:04
amoralej__that was also my understanding, the requirement is to deprecate it in at least one slurp, right?13:04
dviroelyes13:05
jgilaberyes, deprecation should go in a slurp (current), so removal can be in next one13:05
dviroelsince deployments could skip the non-slurp one13:05
dviroeland miss the deprecation notice13:05
amoralej__I'm also fine if we prefer not to announce removal release, just wondering if there is any specific reason to not13:05
amoralej__in this particular case, actually, that is for a experimental feature that we know have bugs13:06
amoralej__I wouldn't be conservative about removing it tbh13:06
jgilabernot that I can think of, I just did not put it in the spec so the llm did not add it to the patch :)13:06
amoralej__;) 13:06
dviroelwe can just announce the removal in 2027.2 and later check if is viable to really remove13:06
dviroelsince it would remove strategies/actions together with it13:07
jgilaberspeaking of bugs, I chained the deprecation patch with a bug fix13:07
jgilaber#link https://review.opendev.org/c/openstack/watcher/+/99938713:07
amoralej__specially know that we now how watcher behaves on removals :) 13:07
jgilaberwanted to get your thougts on if you think it's worth fixing it13:08
jgilaberI've proposed because it's a relatively small patch13:08
jgilaberbut I've tried to be cautious since testing coverage is almost non-existent13:08
dviroelyes, i think we should fix13:08
jgilaberthanks for looking btw amoralej__ 13:08
dviroelit is going to be deprecated but, we don't know the future13:09
amoralej__+1 to fix it13:09
jgilaberack, that settles it then, thanks13:09
jgilaberanother point is, do we want to announce this deprecation together with the strategies13:10
jgilaber#link https://review.opendev.org/c/openstack/watcher/+/100632313:10
jgilaberor send separate announcements?13:10
dviroelhum13:11
amoralej__i think an announcement for all deprecations in this release is fine13:11
dviroelwe need to deprecate the strategies affected by ironic integratin, for sure13:11
dviroelif the team agree with deprecation all mention, fine by me13:11
dviroelwe may need to also deprecate the Action?13:11
jgilaberyes, the action is deprecated in my patch as well13:12
dviroelah ok, good13:12
dviroel+113:12
* dviroel didn't check any ironic deprecation patch13:12
* dviroel yet13:12
jgilaberdviroel, if you want I can send the announcement to the ML for your patch and mine13:13
amoralej__we agreed in deprecating the other non-supported in ptg?13:13
dviroeljgilaber: yes please13:13
amoralej__i think so13:13
jgilaberI think so too13:13
amoralej__in ether13:14
dviroeli don't really remember if we agree or not13:14
amoralej__in ptg etherpad it's "investigate airflow, temperature, host power metrics are available in current supported datasources/exporters"13:14
amoralej__i'm also not sure, tbh13:14
dviroelit is not in the Agreed list13:14
amoralej__wait, that's from previous13:14
dviroel#link https://etherpad.opendev.org/p/watcher-2027.1-ptg#L42013:15
opendevreviewDouglas Viroel proposed openstack/watcher master: Deprecate untested strategies and refresh status matrix  https://review.opendev.org/c/openstack/watcher/+/100632313:15
amoralej__it rings me a bell that we agreed, tbh, but i'm not totally sure13:15
jgilaberyeah I was looking for that, it's not explicitely listed13:15
dviroelWe can vote directly in the change then13:16
dviroeli just removed the W-1 and rebased13:16
amoralej__wfm13:16
jgilaber+1, I'll send the announcement later today13:16
dviroelthanks jgilaber 13:16
jgilaberso final thing on this topic, just an fyi I've also started working on adding watcher to the sdk13:17
jgilaber#link https://review.opendev.org/c/openstack/openstacksdk/+/101644813:17
jgilaberI will submit a series of patches13:17
jgilaberthe first one adds the base infr and Goals 13:17
jgilabergetting familiar with the sdk idioms and layout13:17
dviroelcool13:17
dviroelthat's a great start13:18
jgilaberI will ping in the future for a sanity check from watcher side13:18
dviroel+113:18
dviroelgreat, any questions wrt to jgilaber's topic?13:18
jgilaberI've already checked with the sdk team and their process is simple, just submit (no blueprints)13:18
dviroel++13:19
jgilaberthey are quite fast with review ime13:19
jgilaberthat is all I had for this, thanks13:19
amoralej__there is no specs repo iirc in openstacksdk :) 13:19
amoralej__thanks for taking care jgilaber 13:19
dviroel++13:20
dviroelok, so moving to the next topic13:20
dviroel#topic Proposal for splitting AZ Drain spec13:20
dviroelthe AZ drain spec is getting bigger13:21
dviroelwith the az drain functionality + zone migration improvements 13:21
dviroelso I started a new spec to isolate the zone migration improvements13:21
dviroel#link https://review.opendev.org/c/openstack/watcher-specs/+/101636013:21
dviroelwhich are: unlimited counter and counter's deprecation13:22
dviroelvolume migration between volume-types (src_type, dst_type)13:22
dviroelattached-only volumes option,13:22
amoralej__one question about counter's deprecation13:22
dviroeland disable source node service for src/dest host entry13:22
dviroelamoralej__: 13:22
dviroelyes13:22
amoralej__i'm total +1 to unlimited and disable hosts in the split13:23
amoralej__about counters, i always felt that the capacity of limiting the size is a nice feature, although poorly implemented in zone_migration13:23
amoralej__ideally, I'd propose to change default values to unlimited instead and deprecating/removing it only once we have an implementation of limits that can be common to all strategies13:24
amoralej__we discussed that in the past, i think13:24
amoralej__wdyt?13:24
jgilaberIIRC the proposal is to set to unlimited and deprecate it13:25
dviroelyeah, I do think that it should be implemented outside the strategy, in the base one13:25
jgilaberdoesn't mean we need to remove it next cycle13:25
amoralej__yep, with its own parameters out of the strategy parameter13:25
jgilaberwe can keep it around until we implement a better solution13:25
amoralej__ack, wfm13:25
jgilaberis that correct dviroel?13:26
dviroelyes, we can discuss the removal again in the next ptg13:26
amoralej__+1 then13:26
dviroelso the new spec just move these parts out of Az Drain one (which I will update once you folks agree)13:27
dviroelif you already read the az drain one, you will see that are similar13:27
dviroelwith that i would create a new bp too13:28
dviroelto link these patches13:28
jgilaberack, sounds good to have a second spec, I'll try to review it tomorrow, it should be quick13:28
dviroeldo you folks think that is reasonable?13:28
amoralej__wrt the policy stuff in the az_drain one13:28
dviroeland yeah, we would discuss more Az drain, pinning azs and policy in the az drain spec...13:28
amoralej__you update also that?13:28
dviroeli can update, not sure if we all agree in the etherpad13:29
dviroelbut it also depends on the Nova side atm13:29
dviroelso not sure if I will update until they agree13:29
amoralej__ok13:29
dviroelif nova will or not handle the pinning on their side13:29
dviroeljust in case someone is now aware13:30
dviroelthe az drain depends directly on https://review.opendev.org/c/openstack/nova-specs/+/101309413:30
amoralej__i think that applying new parameters include-az and policy to host based only when moving to different az is a good solution13:30
dviroelwhich implements the --availability_zone parameter as destination for migrate13:30
dviroeland others missing..13:30
jgilabermakes sense to wait until the nova spec, their spec freeze is usually before M213:30
amoralej__ok13:31
dviroelamoralej__: ack, i think that we all agreed with that then13:31
dviroel#link https://etherpad.opendev.org/p/watcher-migrate-pin-policy13:31
dviroelthis is the etherpad that we have been discussing the pinning policy...13:31
dviroelFYI13:31
sean-k-mooneyo/13:31
amoralej__my only doubt is if we could have better config option names that makes it more clear, but i feel that's a good way to manage it13:32
amoralej__and documenting it properly will also work with current names13:32
dviroelamoralej__: ack, yeah, long config names are also complicated to ready in the end, but we shall document it properly, and add some examples to the strategy docs..13:33
jgilaberok so the proposal is to have one default value, and document that the options will be ignored in some cases13:33
amoralej__not exactly ignored13:33
jgilaberi.e host based audits within the same AZ13:33
amoralej__it's the actual meaning of the config option :) 13:33
amoralej__but yeah13:33
sean-k-mooneyparmater13:33
sean-k-mooneyits not a config option13:33
dviroelyeah, parameter13:33
amoralej__yes, sorry, parameter13:34
dviroelconfig option is the watcher.conf stuff13:34
dviroelsean-k-mooney: so i proposed a new spec with zone_migration improvements only, not realted to the az drain13:35
jgilaberso tldr the defaults are 'include_az_*=True', 'az_pinning_policy=auto' 13:35
amoralej__but include_az... actually means allow_pinned_instances_cross_az_migrations13:36
amoralej__so for the host-based in the same az, it's not that we need to ignore that, it simply does not apply to it13:36
amoralej__no need to have different defaults13:36
dviroelalso note that, the Action itself defaults to None. So other strategies will continue to not change the pin.13:37
amoralej__yep13:38
dviroelalright, any other question in this topic, we need to cover the next ones13:39
dviroel#action dviroel will create a new blueprint for zone migration improvements13:39
dviroelok, we can move to reviews then13:40
dviroel#topic Reviews13:40
dviroelI added the one that we didn't covered last week13:40
dviroelAdd Preemptible Instances spec for 2027.1 cycle13:40
dviroel#link https://review.opendev.org/c/openstack/watcher-specs/+/101425213:40
dviroelwhich we already started reviewing13:40
dviroeland winiciusallan[m] updated based on comments13:41
dviroelI think that we just need to revisit and merge it13:41
dviroelafter the meeting13:41
dviroelfine?13:41
amoralej__yes13:41
jgilaber+113:41
dviroelack13:41
winiciusallan[m]amoralej__: I pushed a new patch this morning if you want to take another look :)13:41
winiciusallan[m]there is no major change, just addressing doug's comments13:42
dviroelwiniciusallan[m]: so you will be probably good to continue with the implementation :)13:42
amoralej__i did right before the meeting, just waiting for the ci 13:42
winiciusallan[m]awesome! so will start in the next days13:42
dviroel++13:43
dviroelok, so any other review to cover?13:43
dviroeli know that there are others open in gerrit13:44
dviroelwe should get back to them when possible13:44
jgilaberI have a couple minor ones13:44
jgilaber#link https://review.opendev.org/c/openstack/watcher/+/101130013:44
jgilaber#link https://review.opendev.org/c/openstack/watcher/+/101163313:44
jgilaberthey are CI changes that we already covered in the past13:44
sean-k-mooneydviroel: cool soory was responed to a ping in a diffenrt channel13:45
sean-k-mooneyill review that once im done wiht amoralej__  spec13:45
jgilabernot critical, but would be good to merge I think13:45
dviroeljgilaber: ack, i will check the logs of the aetos job and proceed with it yeah13:45
dviroelit was in my list13:45
jgilaberthanks!13:46
sean-k-mooneyjgilaber: speaking of bugs/reviews i was talking to stephenfin this morning about the api issue we have with version headers in watcher13:46
dviroelwe really need to discuss max_microversion at some point, maybe together with new features like az drain13:46
sean-k-mooneyand ya we shoudl eb remvoing ^ eventually13:47
sean-k-mooneyjgilaber: i see you mentioned adding watcher to the sdk above13:48
jgilaberyes, I started on that13:48
sean-k-mooneybut i dont see any mention of the microvsions header issue13:48
sean-k-mooneyhave you filed a bug for that13:48
jgilaberno, which issue exactly?13:49
sean-k-mooneywatchers api does not supprot the standard microversion header13:49
jgilaberah, I thought we had a bug for that13:50
sean-k-mooneyit currently requires the legay project specific hearders13:50
sean-k-mooneymaybe but we will need to fix that before the sdk work is merged13:50
jgilaberhmm I will double check, launchpad is not loading for me13:51
sean-k-mooneyi think its dwon again13:51
sean-k-mooneyfrom talkign to stephen microverion parse13:51
dviroelworking on my side13:51
sean-k-mooneycan handel this for us we just need to tell it the name of the legacy hearder13:51
dviroelat least this side of the ocean :)13:51
sean-k-mooneyits working for me as well13:52
sean-k-mooneyiu do not see a bug for it13:52
amoralej__it works, but pretty slow ...13:52
sean-k-mooneyi see https://bugs.launchpad.net/watcher/+bug/2161911 for the X-Show-Deleted header 13:53
sean-k-mooneybut i dont see any for the microverion, i think you might be thinking fo the cybrog bug that i fixed last cycle13:53
jgilaberI don't think we have one13:53
jgilaberit might be that13:53
jgilaberI'll look into it this afternoon and create one13:53
dviroel+113:54
dviroelthanks for raising this issue sean-k-mooney 13:54
amoralej__i wasn't aware of that ...13:54
sean-k-mooneyya so watcher is techinially not compliete with the api-sig spec on how microversion should work13:54
sean-k-mooneysince it never adopted the standard header that is ment to be common to all services13:55
amoralej__yep, make sense to fix that13:55
sean-k-mooneythat means the sdk woudl need special handelign and there are keyston auth implciation too13:55
sean-k-mooneyits simpler for use to jsut add the supprot and backprot it if we care too13:55
jgilaberthis is the cyborg commmit https://github.com/openstack/cyborg/commit/7faf80e098f189e6220ccbe5bfc9d179f482fa7613:55
sean-k-mooneyand for the sdk just assume its tehre13:55
amoralej__+113:55
sean-k-mooneyjgilaber: yep cybrog had supprot for the header just incoroect supprot13:56
sean-k-mooneyfor watcher its a larger issue13:56
sean-k-mooneybut the change will be similar13:56
sean-k-mooneyjust use microverion-parse  to handel it13:56
sean-k-mooneyand ther is a paremter fo the legacy project specifc header name13:56
jgilaberwill dig into that, thanks13:56
dviroel+113:57
dviroellets try to cover the bugs topic in these last 3 minutes13:57
dviroel#topic Bugs13:57
dviroel#link https://bugs.launchpad.net/watcher/+bug/217441113:57
dviroel"Audit scope: add exclusion by flavor (or flavor extra specs) to keep GPU / large instances out of optimization"13:57
amoralej__that's an interesting one ...13:57
dviroelnot an bug, but a rfe13:58
amoralej__yes13:58
dviroelbut indeed something that i would like to have in scope13:58
dviroelmore options for filtering based on extra_specs13:58
jgilaberyes, valid use case, we can keep it as wishlist13:58
amoralej__exclussions would be OR, right?13:59
dviroelwe would not need to exlucde gpus from scope if we implement the strategy filters13:59
dviroelsince we know that these would be invalid from watcher point of view13:59
dviroelbut still, the flavor exclusion is a use case13:59
sean-k-mooneythat is not a bug14:00
sean-k-mooneyits a feature14:00
dviroelshould we keep it as a RFE? 14:00
sean-k-mooneyif we do it we will likely need a spec14:00
sean-k-mooneywe coudl keep it as a whishlist bug as a rfe tracker14:01
dviroelright, maybe the reporter has a plan to implement it14:01
dviroelI can comment in the bug too14:01
amoralej__let's request to follow the new feature process14:01
dviroeland let they know that this would require an spec 14:01
amoralej__he may want to start working on it14:02
dviroelyeah14:02
dviroelwe would need to agree on how to model that in the existing scope14:02
sean-k-mooneyfor something that woudl affect the behvior of every stragiy it defintely needs a spec to discuss the potial upgrade impacts adn makign sure this work with the pipeline feature and the change too the syncer14:02
dviroel+114:02
sean-k-mooneyscopes are also not plugable14:03
sean-k-mooneyso arguable this shoudl be part fo the microverison contract14:03
* dviroel we are out of time 14:04
sean-k-mooneythat said we coudl allow parmater to be added to the scope without a microversion but scopes are not something that are implemtne by the stragey14:04
amoralej__are in the collectors actually14:04
amoralej__iirc14:04
dviroelwhile you folks discuss that, let me see if14:05
dviroel#topic Volunteer to chair next week14:05
dviroelsomeone wants to chair next one?14:05
dviroelotherwise I can take it14:05
jgilaberI can take it14:05
dviroeljgilaber++14:05
dviroelthanks14:05
dviroelalright14:05
dviroelwe are out of time14:05
dviroelwe can continue discussing it here or in the launchpad14:06
dviroelthanks all for joining14:06
dviroelwe will meet again next week14:06
dviroel#endmeeting14:06
opendevmeetMeeting ended Thu Nov 19 14:06:22 2026 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)14:06
opendevmeetMinutes:        https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-11-19-13.00.html14:06
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-11-19-13.00.txt14:06
opendevmeetLog:            https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-11-19-13.00.log.html14:06
jgilabersean-k-mooney, I did a quick check before going for lunch, and it seems in Watcher we already use microversion_parse https://github.com/openstack/watcher/blob/28e511b7f6a05ea6bfa66e3cc8e65c08b1a93fe0/watcher/api/controllers/base.py#L97 for the header14:08
jgilaberjust that we use the 'infra-optim' alias14:08
amoralej__also, it'd be great if we could close https://review.opendev.org/q/topic:%22blueprint-functional-test-infrastructure%22+status:open ... 14:08
amoralej__both have 1 +214:08
jgilaberit would not be too hard to change and keep 'infra-optim' as fallback so it's backportable14:09
jgilaberam I missing something else? I remember we talked about this a while ago14:09
jgilaberin my initial patch to the sdk I have a workaround for that, I haven't yet tested it14:09
dviroelamoralej__: ack, lets see if jgilaber can take a look on those and merge them. We can provide follow ups if needed too14:10
dviroeljgilaber: sorry, this is about https://review.opendev.org/q/topic:%22blueprint-functional-test-infrastructure%22+status:open14:10
jgilaberamoralej__, they are on my list, I had given priority to the specs14:10
sean-k-mooneyso infra-optim14:10
jgilaberbut I think I might be able to review them next week14:10
sean-k-mooneyis teh offcial name in the service athority i belive14:10
amoralej__sure, no problem, thanks in advance jgilaber 14:11
sean-k-mooneyoh no14:11
sean-k-mooneyhttps://github.com/openstack/service-types-authority/blob/master/service-types.yaml#L34-L3714:11
jgilaberhttps://github.com/openstack/os-service-types/blob/master/os_service_types/data/service-types.json#L14714:11
sean-k-mooneyyour right its resocue-optimization14:11
jgilaberoh thanks I forgot the exact repo name14:11
sean-k-mooneyjgilaber: one is generated form the other14:12
jgilaberyes, I rememember I check that  and filed a bug at some point, but it might have been downstream14:12
sean-k-mooneyi dont actully recall which one is the souce14:12
sean-k-mooneyso what we need to fix i think is makign sure both names work14:12
jgilaberok, yes I can do that before we merge the sdk changes14:13
jgilaberI'll file a bug once launchpad lets me14:14
jgilaberthanks for looking14:14
amoralej__that rings me a bell, we reported something related to that14:14
jgilaberyes, I'm sure I did, but maybe not upstream14:15
*** jgilaber is now known as jgilaber|lunch14:16
sean-k-mooneyjgilaber: so i just asked ^ about the actull issue14:16
jgilaber|lunchack thanks I'll read back in a bit14:16
amoralej__yes, it was downstream issue, probably a different one14:16
sean-k-mooneytldr wathcer shoudl accpet both the offial name and the alias14:16
sean-k-mooneybut it need the alias today14:16
stephenfinjgilaber|lunch: I don't have the scrollback but this is what I was discussing with sean-k-mooney https://review.opendev.org/c/openstack/service-types-authority/+/101645214:16
sean-k-mooney^ is a workaround for now 14:17
sean-k-mooneybut really we shoudl jsut fix watcher to accpet resource-optimization as well14:18
sean-k-mooneyeventully that is also what we shoudl docuemtn as the service type to use when isntalling/configuring watchers serview type in the keystone endpoint14:18
sean-k-mooneywhich would need changes to our downstream installer and devstack14:19
sean-k-mooneybut that can come later14:19
sean-k-mooneyamoralej__: im +1 on https://review.opendev.org/c/openstack/watcher-specs/+/1012696 see questions inline14:30
sean-k-mooneynothing major but just want to confirm a few things before merging14:30
dviroelok, so jgilaber|lunch should file a bug fopr that, we indeed need to fix and accept the oficial name14:30
*** jgilaber|lunch is now known as jgilaber14:36
jgilaberreading back, ok so it seems I understood the problem correctly, I'll file a bug and submit a fix, should not be too hard14:38
jgilaberlaunchpad is back for me, I filed https://bugs.launchpad.net/watcher/+bug/217505315:38
amoralej__sean-k-mooney, i will update the spec, thanks for your review. A question just came to my mind.What about AuditTemplates? they may also have parameters that are removed or renamed i.e. that would need to be upgraded. should we run upgrade_parameters on them?16:35
amoralej__dviroel, jgilaber ^ wrt https://review.opendev.org/c/openstack/watcher-specs/+/101269616:35
amoralej__for default_parameters 16:36
jgilaberdo we need to? Wouldn't the paramaters be upgraded once they are used to create an audit?16:37
amoralej__not if the default_parameters has an non longer valid parameter16:37
amoralej__i actually will fail to create the Audit, i think16:37
amoralej__it wouldn't affect the audits previously created from the AuditTemplate16:38
amoralej__those will be updated16:38
amoralej__but for the AT itself, if a new Audit is created from it, it'd fail i think16:38
sean-k-mooneydefault_parmaters is used to initalise pamters and hten you merge any actual auit parmater into that16:40
sean-k-mooneyi dont htink we need to upgrade the audit template16:40
sean-k-mooneythe expction16:40
jgilaberI was thinking the same, we don't validate the default parameters right?16:40
sean-k-mooneywoudl be the remvoal of a parmater form the relevent stragy16:41
sean-k-mooneybut we coudl do that when we are cretaign the actual audit form the templaste16:41
amoralej__so, i create an AT with parameter param1: foo. The strategy renames param1 to param2. An attempt to create the Audit would fail16:41
sean-k-mooneyand the admin coudl update the audit template default_parmaters after/beofre upgrade16:41
sean-k-mooneywell no16:41
sean-k-mooneybecause create an audit with parm1 shoudl still work after the rename16:42
sean-k-mooneybecause the upgrade on parmaters shoudl rename it right16:42
amoralej__well, in that case no migration is needed at all, that'd be a backwards compatible change16:42
amoralej__The whole point of the upgrade_parameters() is to cover non-backwards compatible changes16:43
amoralej__and would be only executed at upgrade (restart), not at audit creation16:43
amoralej__so, i rename param1 -> param2 (full rename, removing old name). Then i create an upgrade_parameters which renames param1 to param2 in the parameters dict of existing audits16:44
amoralej__on next restart16:45
sean-k-mooneyya so we will do soemthing like a = Autidt(parmaters=template.default_parmaters().update(request.parmaters)16:45
sean-k-mooneya.upgrade_paramaters16:45
sean-k-mooney()16:45
amoralej__yes16:45
sean-k-mooneywell wee need to make a copy since that would do bad things as written16:45
sean-k-mooneybut i think we agree over all16:45
amoralej__yes, but i'm not sure about AuditTemplates16:46
sean-k-mooneyi dont think we need to modify audit templates, we could16:46
amoralej__so, admins would need to update the AT, right?16:47
sean-k-mooneybut the point is that we are snapshoting the inputs wehn we are creating the audit16:47
sean-k-mooneyyes16:47
sean-k-mooneythey woudl need to update it everntully16:47
amoralej__wfm16:47
sean-k-mooneyif we were changign a name we shoudl suprot both for some period of time16:47
amoralej__yeah, that was just an example of non-backwards compatible change, which is like a last-resort16:48
sean-k-mooneywe can provide a helper method to migrate them as a watcher-db-manage commend if we want to16:48
amoralej__default policy is being backwards compatible ...16:48
sean-k-mooneyor they can manuallyupdate the template before upgrading to the version that does not supprot hte deprecated name16:48
amoralej__I'll be explicit in the spec16:49
sean-k-mooneythe audit parmaters actully have 3 levels of inputs for the snapshot16:49
sean-k-mooneythe lowest precendent is the stragey default16:49
sean-k-mooneythen the audit template default_parmaters16:49
sean-k-mooneyand then what ever you pass in the audit create request16:50
amoralej__yes16:50
amoralej__bad a wrong parameter in the AT will make the audit creation to fail, unless you are overriding that parameter in the audit16:50
sean-k-mooneywe could (not sure we should) recored the version of the stragey the template was created agasint16:51
amoralej__s/bad/but/16:51
sean-k-mooneybut i dont know if we need that16:51
amoralej__yeah, it's a extreme case tbh16:51
sean-k-mooneyso it shoudl not16:51
amoralej__it does16:51
amoralej__well, i can test it, but that's my expectation16:51
sean-k-mooneythat woudl be a bug16:51
amoralej__why?16:52
sean-k-mooneybecause the templates are version less and you are provideign an upgrade method16:52
sean-k-mooneyit shoudl only cause a failure if we removed the supprot16:52
sean-k-mooneywhich woudl need a deprecation period16:52
amoralej__yes, yes16:52
amoralej__i always mean in the context of non-backwards compatible chagnes, for the context of this conversation16:52
sean-k-mooneydo we currently validate and reject unknown parmater in a stragy today16:53
amoralej__yes, but if you upgrade the strategy and the parameter is not longer valid, that's the case i meant16:53
sean-k-mooneyok well in a non backwared compatable chagne16:53
sean-k-mooneythe process i described above was create teh obejct, set the parmaters filed and call upgrade_paramters before saving it16:54
amoralej__yes, but in the spec, i only do that for existing Audits in upgrace16:54
amoralej__so, pre-existing Audits will be fine16:55
sean-k-mooneyif we recored teh version of the stragey in teh audit templeate when its created we could include that in the parmater so the upgrade_parmateres use it to uplevel it16:55
amoralej__my question was about new Audits being created from the AT which is now outdated16:55
amoralej__yes, that was my question, if we would want to implement the exact same approach for AT default_parameters16:55
sean-k-mooneyi think the larger question is shoudl it only happne on upgrade or shoudl it happen on all future autid creatison as well16:55
amoralej__yeah, that's the other approach16:56
sean-k-mooneywe could allow you to optioanlly pass the version in teh parmaters16:57
sean-k-mooneyand then alwasy upgrade it perhaps if tis older then the server supprots16:57
sean-k-mooneybut i dont knwo if we are over thinking it16:57
amoralej__probably :) 16:57
sean-k-mooneyi think for now we could punt on this (AT migration)16:58
amoralej__given that that case won't affect to existing running Audits16:58
amoralej__which is the critical case16:58
sean-k-mooneyand just docuemnt that you shoudl update the template when we are droping supprot for a parmater16:58
amoralej__it's probably good to let it fail and document properly + deprecation period, etc...16:58
sean-k-mooneywhen we have a non backportabel change we can then dicsuss if we provide tooling to automate that for you16:58
amoralej__actually, it may have very tricky corner cases when mixing versions16:59
sean-k-mooneyhow do you mean?16:59
amoralej__I'm thinking ...16:59
sean-k-mooneyhere is a concreate example, we are thinking of removing the limits on action form the az migtion17:00
amoralej__actually, storing the strategy version in the default_parameters in AT, we should be able to do it17:00
sean-k-mooneywe are adding supprot for -1 for unlimited17:00
sean-k-mooneywe may then change the default17:00
sean-k-mooneyif we want to remvoe the parmater however that woudl be backwared incompataible17:00
amoralej__that is the case i was thinking of. Rename param1 to param217:01
sean-k-mooneybut in that case we woudl be removing the ablity to limit it via that approch. unless we have a replacment for the limit teh automated fix woudl be to drop the limits17:01
sean-k-mooneyfor the rename case old_name -> new_name its tricky only if we reuse old_name17:02
sean-k-mooneyif we dont then we can order the upgrades by version 17:02
sean-k-mooneyform oldest to newest17:03
sean-k-mooneyand even if you later rename new_name to new_name_217:03
sean-k-mooneyyou can determisticly upgrade form old_name->new_name_217:03
amoralej__i was thinking in the case where we have an AT with param1=X and in the Audit we set param2=Y (where Param2 is rename of param1). The upgrade_parameter should rename param1 to param2, if we do that in the AT and then merge, it'd be fine17:03
amoralej__so yeah17:04
amoralej__but we'd need to apply the upgrade in the default_parameters before merge with audit parameters17:04
amoralej__so, it's doable17:04
sean-k-mooneyack i think this is imprant but likely soemting we can resolve wehn we need too17:04
amoralej__but i doubt we need to do it now17:04
amoralej__ok +117:05
amoralej__btw, i replied some of your questions in https://review.opendev.org/c/openstack/watcher-specs/+/101269617:05
amoralej__the main two ones are https://review.opendev.org/c/openstack/watcher-specs/+/1012696/comment/7dde3736_732b27b0/ and https://review.opendev.org/c/openstack/watcher-specs/+/1012696/comment/490c9157_8b122b36/17:05
opendevreviewJoan Gilabert proposed openstack/watcher master: Accept resource-optimization API service type  https://review.opendev.org/c/openstack/watcher/+/101652417:07
dviroelamoralej__: hey, i was at lunch, and reading back now...17:23
amoralej__np17:24
dviroelI think that for AT we can just let if fail at Audit creation for now. AT is a stateless resource, and user will receive a failure when creating a new Audit from an outdated AT 17:24
amoralej__yeah, i think we are good with that17:24
amoralej__jgilaber, given that you +2 that in previous PS and last update was just a couple of agreed changes,I'm merging https://review.opendev.org/c/openstack/watcher-specs/+/101425217:35
opendevreviewVictoria Martinez de la Cruz proposed openstack/grian-ui master: Add instance health dashboard  https://review.opendev.org/c/openstack/grian-ui/+/101653117:37
jgilaberamoralej__, yes that's fine, I looked at it briefly and it's ok17:43
opendevreviewMerged openstack/watcher-specs master: Add Preemptible Instances spec for 2027.1 cycle  https://review.opendev.org/c/openstack/watcher-specs/+/101425217:58
opendevreviewMerged openstack/watcher master: Remove compute max_microversion set in ci  https://review.opendev.org/c/openstack/watcher/+/101130018:55

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