Tuesday, 2025-11-25

gouthamrtc-members: a gentle reminder that our weekly irc meeting will be held here in ~52 mins16:08
gouthamrplease find the agenda here: https://wiki.openstack.org/wiki/Meetings/TechnicalCommittee#Agenda16:08
mnasiadkagouthamr: I’m off sick, so won’t be able to join16:09
gouthamrmnasiadka: oh :( sorry to hear that, get better soon!16:09
cardoeI'm going to have to miss the 2nd half of the meeting16:10
gouthamrack16:14
gouthamr#startmeeting tc17:00
opendevmeetMeeting started Tue Nov 25 17:00:31 2025 UTC and is due to finish in 60 minutes.  The chair is gouthamr. Information about MeetBot at http://wiki.debian.org/MeetBot.17:00
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.17:00
opendevmeetThe meeting name has been set to 'tc'17:00
gouthamrWelcome to the weekly meeting of the OpenStack Technical Committee. A reminder that this meeting is held under the OpenInfra Code of Conduct available at https://openinfra.dev/legal/code-of-conduct.17:00
gouthamrToday's meeting agenda can be found at https://wiki.openstack.org/wiki/Meetings/TechnicalCommittee17:00
gouthamr#topic Roll Call17:01
gtemao/17:01
noonedeadpunksemi o/ - will need to leave in ~20m17:02
gouthamrcourtesy-ping:  frickler, spotz[m], cardoe, bauzas17:03
cardoeo/17:03
gouthamrnoted absence: m n a sia d k a17:03
cardoeTalking in a different channel. Sorry.17:03
gouthamrno problem; i wonder if the meeting time's being an issue, or, its the end-of-year fever that gets us all :) 17:05
gouthamrour attendance is less than quorum, again.. lets get rolling17:05
gouthamr#topic Last week's AIs17:06
noonedeadpunkI think latter :)17:06
gouthamrReviews on TC resolution on os-net-config handover17:06
gouthamr#link https://review.opendev.org/c/openstack/governance/+/967463 17:06
gouthamrthis has a minimum number of votes and can merge anytime now, has been open 7 days.. 17:07
gouthamrif you've not looked yet, please do17:07
* bauzas joins late17:07
gouthamrah quorum!17:07
gouthamrfrom the last week's meeting, i surmised that you wanted to move the FIPS Goal Objectives to the tracker17:08
gouthamrit's already there :) i think i could do the patch today to move the goal back to "proposed" and start an ML thread with the actions we took at the PTG17:09
gouthamrspecifically, asking for help in redefining testing and completion of the goal, reviving failed testing, and stating that the TC is okay with project teams dropping failing jobs in the meantime17:09
gouthamrdid i miss anything?17:10
gouthamrthe next AI was regarding "PTL Appointment History", i think tonyb took this AI, but we didn't define it completely..17:12
gouthamra couple of weeks ago, we discussed a concern that PTL appointments must be preserved on projects.yaml.. and the AI was to write this down somewhere17:12
gtemayes, also last week we were not  able to recall details17:13
fungiright, the point was twofold17:14
fungifirst, a change in process so we stop wiping the prior list of appointments whena project switches from ptl to dpl model17:14
fungisecond, an audit of the git history to find past instances where that happened and restore the missing entries17:14
gouthamrah +117:15
gouthamrokay, tonyb has this detail in the scrollback now, ty fungi17:16
gouthamrnext AI was to finish working on the Monasca cleanup patches17:16
gouthamrthis looks green now, ty for the deep work on it, frickler: 17:17
gouthamr#link https://review.opendev.org/q/hashtag:%22monasca-retirement%22+(status:open%20OR%20status:merged) 17:17
gouthamri messed up the cleanup, i left the LICENSE file in each repo, bad regex :/ so our validation script is failing17:18
gouthamr#link https://review.opendev.org/c/openstack/governance/+/953671 17:18
gouthamrbut, i don't want to go and delete those files and adjust the README.rst files etc (because we call out that the repo contents can be restored by checking out HEAD^1) 17:19
gouthamrso i added the repos to an ignore list here:17:19
gouthamr#link https://review.opendev.org/c/openstack/governance/+/96826817:19
fungia more future-proof approach would be to stop explicitly referring to `git checkout HEAD^1` in the final readme and just use wording that indicates the last pre-retirement state is found in the branch's git history17:19
gouthamrthat would be nice17:20
fungiwe already have a ton of long-ago retired repositories for which it's no longer accurate anyway due to later changes merging17:20
gouthamrfair, that wording is here: https://docs.opendev.org/opendev/infra-manual/latest/drivers.html#step-2-remove-project-content17:21
fungii'm happy to review an adjustment to that17:22
gouthamrwe didn't need to follow that to the letter, its just a suggestion 17:22
gouthamrand was easy to copy-pasta17:22
fungialso openstack could just record a template of its own somewhere like https://docs.openstack.org/project-team-guide/repository.html#step-2-remove-project-content17:22
gouthamryeah.. 17:23
fungibut in this case i do think the wording in infra-manual would be fine to adjust to something more loose, as long as it stays project-neutral17:23
gouthamr+1.. so these patches could merge soon, if you'd like to review, please do.. 17:24
gouthamrthe next AI was regarding an openstack-discuss ML thread pointing to repositories using passlib to track its removal - was assigned to gtema17:25
gtemaoh yes, haven't done this yet, but improved analysis17:25
gtemaI updated the PQC wiki adding first actions17:25
gtemawill create 4 patches tomorrow agains requirements repo17:26
fungioh, that reminds me, someone reached out to me wanting to get more involved in security sig activities, and was specifically interested in helping with the pqc stuff17:26
fungii guess jp's ml thread would be a good starting point for them to get up to speed on it17:27
gouthamrnice, ty for the updates on this17:28
fungibut helping with stuff like passlib replacement would probably be a great place to jump in17:28
gouthamr++17:28
gtemathere is nothing to do with passlib since it is technically eliminated, just need to cleanup requirements17:28
gouthamr#link https://wiki.openstack.org/wiki/Post_quantum_openstack 17:28
gouthamr#link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/NH7SVJHLRA4ZLH4UINXULFX5SXKVDPZK/ (ML thread on PQC) 17:28
gouthamrprobably usage in projects that don't adhere to global requirements, mn asiadka mentioned needing to cleanup kolla-ansible17:29
fungioh, even better (re: passlib)17:29
gouthamrthere's this too: https://opendev.org/openstack/bifrost/src/branch/master/playbooks/roles/bifrost-ironic-install/tasks/install.yml#L6717:30
bauzasthanks for the work on that 17:30
cardoeI’ve got to step away17:30
gouthamrack cardoe.. ty for joining us17:31
gouthamrthat's a wrap on AIs17:31
gtemagouthamr - I know there are few cases of tooling still installing it, but we "agreed" last week to just get rid of it in requirements and send out info through the ML - which is what I'll do tomorrow17:31
gtemaunless the world explodes tomorrow again17:32
gouthamrty gtema 17:32
gouthamrgtema: no that's scheduled for day after tomorrow17:32
gouthamr#topic OS Manuals could use reviews/help17:32
gouthamr#link https://review.opendev.org/q/project:openstack/openstack-manuals+status:open17:32
gouthamrthink we had an impasse regarding this: https://review.opendev.org/c/openstack/openstack-manuals/+/90561417:33
gouthamrand a related item on oslo's plate: https://review.opendev.org/c/openstack/openstackdocstheme/+/90561317:33
gouthamris there a way to compromise on this? Can we advertise that we're tracking users to generate metrics and help us improve OpenStack; and suggest how to disable this tracking? 17:35
fungiwhile those are changes i pushed, i have no vested interest in either approach (accepting them or ripping that stuff out entirely)17:35
gouthamrmy view is that i'd accept the changes and then seek another change to rip them out.. 17:37
gtemathan leave such review17:37
fungitonyb also pointed out that we have similar outdated scripts in the openstack wiki theme which would need to be redone for one of the more recent versions we're planning to upgrade through17:37
gouthamris the wiki theme maintained on git too? 17:38
fungino17:39
fungiwe don't (yet) have any config management/deployment automation for the mediawiki server, that's something tonyb has been working on17:39
gouthamrso technically, there are aspects that can be changed, enabled/disabled without any of us needing to review 17:40
fungiright17:40
fungiit's something that was manually set up by a wmf staffer when we converted from moinmoin to mw, and they had committed to managing the server themselves back in the days before we insisted on having configuration management17:40
fungibut then they moved on to other things and the server has been semi-abandoned since17:41
fungibecause turning a giant pile of php into a configuration-managed system wasn't anything the rest of us had time for17:42
gouthamryeah :( yet its a significant resource for the community17:42
gouthamrare there any other opinions to note wrt this change?17:42
gouthamrtwo +2s would mean it can be merged; i'll add mine and close it out unless you tell me we need a bit longer to consider it17:44
gouthamr#topic Python PTI and test runner guidance17:45
gouthamr#link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/3V3CNPQLB77SKFVLZ6LXJ5NPNYWW4QFD/ (Request for guidance on improving Python PTI doc to include pytest for Horizon plugins testing)17:46
gouthamrthink this thread went in different directions and i learned a few new things 17:47
fungithe resolution seems to be where chandan says "Watcher team is going with unittests module for writing watcher-dashboard integration tests."17:48
gouthamri think we've established that there's no problem pursuing pytest based tests, and we're asking the horizon team to explore if those tests can be run with stestr 17:48
gtemanope - stestr is not part of the issue at all17:49
gtemanot the runner is the "question", but the test itself17:49
gtemawe have projects using pytest to run unittests already17:50
fungiright, pytest the runner is compatible with the unittests module's interfaces. pytest the module has additional test building blocks that don't work with a lot of other test runners17:50
clarkbthis situation is one where I think we should start from our goals first17:51
clarkbwhether or not anyone is using it already is beside the point.17:52
gtema+117:52
clarkband as I stated on the mailing list thread I don't think pytest is in opposition to the goals we have today17:52
bauzaswhy people want to use pytest ? 17:53
clarkbspecifically being able to run in parallel to take advantage of test resources and the ability to run in places like IDEs17:53
fungiprior discussions also uncovered that the horizon team's implementation was due to the developers who wrote it not being familiar with existing convention within openstack, and then the team simply didn't want to redo that work later once it was already written17:53
bauzasjust an open question 17:53
gtemabauzas - because it is usually simpler, and slimmer17:53
gouthamryes, so its reasonable to ask if they can run these tests in parallel, and produce a subunit o/p 17:53
clarkbbauzas: pytest the test runner has some generally nice features for developers when debugging tests. Pytest the library is a lot less verbose than unittest the library17:53
bauzaswell, I can pdb dito ;-)17:54
clarkbin the thread fixture creation management was specifically called out17:55
clarkbyou can totally write fixtures taht work with unittest and don't rely on pytest, but it sounds like people like the pytest fixture approach17:55
clarkbgouthamr: yes that seems reasonable: basically capture that our assumptions are correct then move on17:56
gouthamr#link https://governance.openstack.org/tc/reference/pti/python.html#python-test-running17:56
gouthamri think we can lead the discussion there, i agree that if this is encoded in the CI jobs, it'll catch the introduction of weird incompatible patterns/plugins that're possible with pytest17:57
gouthamranything else to round up this discussion? 17:59
gtemagouthamr - there are already quite a few projects using pytest instead of stestr, so this "guideline" is not fulfilled anymore17:59
gouthamryes; i think we did a code search to identify these18:00
frickleroops, I messed up meeting timing again. I blame downstream noise. reading backlog now18:01
gouthamrnp frickler; its time to wrap up18:01
gouthamranything to note for the minutes today?18:02
fungithe other projects brought up as existing examples of pytest use within openstack were cases in which the implementation was written before the projects became part of openstack, so essentially inherited/transplanted use18:02
funginot that it's necessarily a problem, just context18:03
gouthamr++ 18:03
gouthamrokay, thank you all for joining.. i'll punt the rest of the topics to next week18:04
gouthamrif there's anything pressing, please do continue the discussion here18:04
gouthamr#endmeeting18:04
opendevmeetMeeting ended Tue Nov 25 18:04:22 2025 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)18:04
opendevmeetMinutes:        https://meetings.opendev.org/meetings/tc/2025/tc.2025-11-25-17.00.html18:04
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/tc/2025/tc.2025-11-25-17.00.txt18:04
opendevmeetLog:            https://meetings.opendev.org/meetings/tc/2025/tc.2025-11-25-17.00.log.html18:04
sean-k-mooneygouthamr: the pytest topic was orginally a problem that i raised at the ptgu specific calluing out th use of pytest fixture or pytest as a lib as proiblematic18:23
sean-k-mooneypytest as a test runner is fine18:23
sean-k-mooneyon the watcher side we are plannign to do some pocs of diffent aprpoches, no descison has been made but we will be starting with unittest + (selenitum vs playwright) to compare the ux fo writing/reviewing tests with both18:24
sean-k-mooneywe will likely compare that ot the pytest based test as well18:25
sean-k-mooneybut im not generaly a fan of the pytest fixutre approch 18:25
sean-k-mooneylets boiler plate but also more magic that is harder to debug/think about18:26
sean-k-mooneywhich is the pros and cons of runtime depency injection and asert rewriting that pytest brings18:27
gouthamrsean-k-mooney: i see, i don't see our PTI preventing that though.. it's only clarified the "test runner" bit.. with any testing we'd like to ensure there is a "real time subunit output" and "support for parallel execution of tests". There's a note that the test runner we've chosen (stestr) "only runs tests conforming to the python stdlib unittest module".. 21:03
gouthamrwhen you diverge and use pytest, you can end up using arcane patterns that'll become a maintenance burden for the community 21:03
gouthamri.e., there's a risk.. but, the fixtures approach for one seems like a preference? help me understand :D ty for sharing the watcher team's approach to this. very glad you brought this up because, i, like the horizon folks, didn't really read teh PTI properly and understand the context for these past decisions!21:05
sean-k-mooneythe pti in my mind directly probiths any use of pytest as a lib21:07
sean-k-mooneywhich was somthign we reinfoced in teh rquriement repo for year by not allowing it ot be listed at all21:07
sean-k-mooneyhttps://github.com/openstack/governance/blob/master/reference/pti/python.rst#python-test-running21:07
sean-k-mooneyif we read that its say "stestr only runs tests conforming to the Python stdlib unittest model (and extensions on it like testtools)."21:08
sean-k-mooneyusing pytest as a libary creates testcases that do not conform to that requirement21:08
gouthamrif your research can show that this is possible without pytest, maybe you'll have converts21:12
sean-k-mooneygouthamr: my over all concer is not about usign it as a test runner21:12
sean-k-mooneyit about useing it to write tests21:13
gouthamryes, ty for clarifying - i do want to chase down the other aspects of the PTI though while we're discussing this21:15
sean-k-mooneyi brought this up when pytest was being dicsued in https://review.opendev.org/c/openstack/requirements/+/71231521:16
gouthamri.e., these integration tests currently don't conform to those report and parallelization requirements, so can they with relative ease? 21:16
sean-k-mooneyat the time i als dint really feel that using it as a test framework was a good thing21:16
sean-k-mooneybut i didnt obejct strongly since it was not going to directly affect me day to day21:16
sean-k-mooneygouthamr: https://pypi.org/project/pytest-subunit/ exits but has not been updated in a hile21:17
gouthamryes21:17
sean-k-mooneythe paralisation for those test by there nature cant alwsy be suprpoted but some of it could be21:18
gouthamri found it through your comment on that very change during the PTG, and sigh, i don't know if it works and is reliable to use 21:18
sean-k-mooneyto be clear if the parallisum and subunit reporting were adressed21:18
sean-k-mooneyit woudl not adresss my concern at all about using pytest21:18
gouthamrwe seem to be using https://libraries.io/pypi/pytest-html at the moment in these integration test jobs21:18
sean-k-mooneythat is focusing on the test runner part21:19
sean-k-mooneywhich i dont have an issue with21:19
gouthamrack21:19
sean-k-mooneyyes pytest-html sever a similar function for humans21:19
clarkbsean-k-mooney: gouthamr I don't think the PTI says anything about libraries does it?21:19
clarkbits just that some libraries may not be compatible with all test runners21:19
gouthamryes it doesn't 21:20
sean-k-mooneypytest test case adn fixture arnt runable with stestr21:20
sean-k-mooneyunless that has magiclly changes recently21:20
clarkbcorrect (nor unittest.main or nose)21:20
gouthamrmtreinish called this out in his comment on that requuirements change: "We've had this discussion of why not pytest many times over the years and wrote down the reasoning in the project testing interface" 21:20
sean-k-mooneyright and that is requird by the pti21:20
clarkbright but the reasons it is required by the pti are no longer at odds with pytest21:21
clarkbwhich is why I keep trying to get us back to thinking about goals and objectives first21:22
clarkbthe goal is to take advantage of the ci system while allowing developers to use whatever tool they prefer locally21:22
clarkbIDE or whatever21:22
sean-k-mooneywell i tough one of the goal was to ahve a commoen test framework 21:22
clarkbnot in the pti it doesn't say anything about the tests themselves21:23
sean-k-mooneythat now how i read "testr only runs tests conforming to the Python stdlib unittest model (and extensions on it like testtools)."21:24
sean-k-mooneyi read that as your test must be witen with the stdlib unitests and it exetions21:24
sean-k-mooneywhich pytest is not21:25
clarkbyes, because doing so enabled the other goals. I'm not sure it wqs ever a strict goal on its own21:25
clarkbits a consequence of the decisions made to achive the other goals21:25
sean-k-mooneyoh because i woudl condier that core to the pti21:25
sean-k-mooneynot a side effect21:25
sean-k-mooneywhich is why i rased this in the first place21:26
clarkbnote many people can and do use other testing tools that work fine becaus they are standard compatible21:26
clarkbfixtures for example21:26
clarkbthey are not part of the standard21:26
clarkbeven mock was at one time non standard iirc21:26
clarkb(but it again was compatible with the standard tooling)21:26
sean-k-mooneyright i knwo it an extention to the standard set21:28
sean-k-mooneywe use to have difent workdin tin the past https://github.com/openstack/governance/blob/91b7b8af4a4534fb23fb1348c3e9f4a1650af404/reference/pti/python.rst?plain=1#L72-L8121:28
sean-k-mooneythat was revisted by https://github.com/openstack/governance/commit/759c42b10cb3728f5549b05f68e826b1c62a968c21:29
sean-k-mooneywe used to be very explict that we buidl on top of https://opendev.org/openstack/oslotest and the depenciy it combines21:32
sean-k-mooneyi.e. fixutre and testtools21:32
sean-k-mooneyin the past we use mox instead of mock the lib and then eventually we moved to unittest.mock21:33
sean-k-mooneywhen we got rid of python 2.7 support21:33
clarkbright so I guess this goes back to what our goals are. Is it just ensuring we can adequately take advantage of ci resources and allow devs to use their IDEs too. Or do we also see value in consistency of actual test library tooling for the tests themselves21:34
clarkbthe current iteration of the docuemnt seemst o have relaxed the second thing21:34
sean-k-mooneyright the second thing is what i saw value in as tooling is easy to change btu leaning how a test framwork works i harder IMO21:35
sean-k-mooneyanyway i better run before stores start closing for the night o/21:36
sean-k-mooneyif the opion of the tc is that pytest as a lib is consitent with the pti then i can take that impact back to the watcher team21:37
sean-k-mooneyand we can decied seperatly fi the cost of learning pytest is worth the benift of the shared code with horizon21:37
* gouthamr writes for when you return21:53
gouthamrsean-k-mooney: i am leaning on the side of: if project teams think that there's benefits in pytest, and it can conform to consistency that we've built around test tooling, by all means, explore it - but please be careful not to violate the desire for maintainability. Ease of use and familiarity for one person, or a small subset of people shouldn't be the motive to drive this decision21:55
gouthamr it should be that we can sustain testing this way beyond our involvement with the project. 21:55
gouthamrwhich is why i'm supportive of the watcher team exploring our familiar toolkit: unittest + testtool extensions, to try the same sorta testing and showing us the way.. 21:57
gouthamrif that's a burden you're unwilling to bear, it's totally fair to collaborate with horizon and other UI folks that are going down the pytest path.. except we've to resolve these PTI/test runner capability questions21:58
sean-k-mooneyright the deisre for mantianablity and constisntsy is what intially tells me not to use pytest21:59
sean-k-mooneywe dont have expice with it so in addtion learning ot write test in playwright or selenium21:59
sean-k-mooneywe woudl have to also do so in a testing framework we do not have expsrince with22:00
gouthamrfair enough :) someone on the thread noted (i'm paraphrasing) that the wider python community seems to be gravitating towards pytest, and mayyybe we may get some non-openstack python contributors helping us in the long run 22:01
sean-k-mooneyi feel liket that unlikey but maybe22:01
clarkbI suspect that wher eyou might get value from that is less in attacting new contributors and more in using tools LLMs are more familar with22:12
sean-k-mooneyno that i cna belive22:13
sean-k-mooneythat one of the resons im interesting in playwright over selenium22:13
sean-k-mooneyas playwright seams to be winding there massivly in the llm space22:13
sean-k-mooney*playwright seams to eb winning...22:24
opendevreviewMerged openstack/openstack-manuals master: Update analytics tracking ID  https://review.opendev.org/c/openstack/openstack-manuals/+/90561422:45

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