| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2022-06-21 | |||
| 16:24:26 | gibi | your turn | |
| 16:24:44 | elodilles | #info gates are mostly not blocked | |
| 16:24:56 | elodilles | #info stable/train is blocked - melwitt's fix: https://review.opendev.org/c/openstack/nova/+/844530/ | |
| 16:25:04 | elodilles | #info stable branch status / gate failures tracking etherpad: https://etherpad.opendev.org/p/nova-stable-branch-ci | |
| 16:25:13 | elodilles | release patches proposed (yoga, xena, wallaby): https://review.opendev.org/q/project:openstack/releases+is:open+intopic:nova | |
| 16:25:27 | elodilles | the release patches are on the way now | |
| 16:25:34 | gibi | \o/ | |
| 16:25:58 | elodilles | (sorry i've re-used last week's lines as there are not that much news o:)) | |
| 16:26:36 | gibi | elodilles: thanks | |
| 16:26:41 | elodilles | np | |
| 16:26:43 | gibi | anything else about stable? | |
| 16:26:51 | elodilles | nothing from me | |
| 16:28:03 | gibi | #topic Open discussion | |
| 16:28:10 | gibi | (artom) Can we revisit stable func test backport policy? Specific patch stack: https://review.opendev.org/c/openstack/nova/+/791480/1 Previously we didn't want to backport func test infrastructure because it just offloads the backport debt onto whoever is doing backport for older than train releases. Some time has passed now, are there still operators running < stable/train and needing backports? | |
| 16:28:16 | gibi | https://etherpad.opendev.org/p/r.ea2e9bd003ed5aed5e25cd8393cf9362 indicates a majority of "train or older", but how many are on the "older" half of that? | |
| 16:29:02 | artom | Bringing this up again because at this point there are 3 separate bugfixes depending on those test refactor patches | |
| 16:29:08 | artom | But basically $topic :) | |
| 16:29:28 | gibi | personally I'm OK to bring back any test refactors to stable branches | |
| 16:30:17 | artom | IIRC last time we talked about this, elodilles was worried that anyone backporting to rocky and older would get the "refactor" debt unloaded onto them | |
| 16:30:36 | elodilles | IF there are enough reviewers then maybe it could be OK, though it's best to keep things on the safe side and backport less risky things | |
| 16:31:11 | artom | Are func test refactors really risky though? | |
| 16:31:14 | sean-k-mooney | well test code is less risky in general since it does not affect the runing code | |
| 16:31:27 | elodilles | artom: yepp, if we backport mass amount of functional test refactors, then it makes the backport harder for older branches | |
| 16:31:27 | gibi | it does not risk the production code, it risk the CI stability | |
| 16:31:36 | sean-k-mooney | and in some cases are not actully installed with the production code | |
| 16:32:11 | artom | elodilles, yep, agreed on that. So in practice, bauzas and gibi were at summit, is anyone still doing backports for < stable/train? | |
| 16:32:34 | artom | As in, Red Hat will have to care about stable/train for a long time | |
| 16:32:46 | sean-k-mooney | :( | |
| 16:32:53 | gibi | I tend to propose backports to stable/pike while I were in E/// | |
| 16:32:54 | sean-k-mooney | its true but :( | |
| 16:32:55 | artom | Yeah, sad face indeed | |
| 16:33:16 | gibi | I assume E/// still uses stable/pike | |
| 16:33:24 | gibi | but I don't think we will see much backports there | |
| 16:33:48 | elodilles | i see that there are less and less backports pushed toward old branches, though if we make it harder for developers, then it will not help the situation as well | |
| 16:33:56 | artom | gibi, elodilles, so I can buy the gate stability argument for integration tests, but when was the last time we had an issue with func tests that wasn't about versions of things like tox? | |
| 16:34:20 | sean-k-mooney | elodilles: well right now its hard to backprot to train because once you get past about victoria you are missing the helpers | |
| 16:34:25 | gibi | we have still open a bug where nova funct test leaks notifications between tests :) | |
| 16:34:26 | elodilles | gibi: unfortunately my pike patches are hanging there without reviews, so.... o:) | |
| 16:34:34 | gibi | elodilles: I know :) | |
| 16:34:56 | gibi | artom: so func test could be problematic | |
| 16:35:11 | gibi | as they run eventlets | |
| 16:35:26 | gibi | and sometimes depends on extrenal things like sysfs :) | |
| 16:35:44 | gibi | still I think we should backport func test infra | |
| 16:36:27 | sean-k-mooney | its the integrated_helpers that are most useful | |
| 16:37:05 | artom | gibi, I feel like sysfs should be poisoned in func tests... | |
| 16:37:15 | gibi | artom: I have a patch | |
| 16:37:26 | gibi | https://review.opendev.org/c/openstack/nova/+/844627 | |
| 16:37:40 | sean-k-mooney | artom: for the most part its mocked already modulo bugs | |
| 16:37:56 | sean-k-mooney | but yes the poison is also good to do | |
| 16:38:31 | gibi | does anyone here strongly disagree to backport func test infra? | |
| 16:38:36 | elodilles | also note, i'm not completely against backporting func test refactors, but i still think it is best to keep it in a low level and we should not backport massive refactors :/ | |
| 16:39:12 | gibi | elodilles: it is a tradeoff, either you take the risk by backporting the refactor or take the risk when you backport a fix that needs to be changed due to the missing refactor | |
| 16:39:33 | gibi | artom: 3 lines! come on! :) | |
| 16:39:33 | sean-k-mooney | well or we drop the functest on backport | |
| 16:39:43 | elodilles | if a refactor breaks something then we don't have the bandwidth to keep it maintained i think. stable should be stable :( | |
| 16:39:45 | artom | The one on top is a bit bigger ^_^ | |
| 16:39:46 | gibi | I strongly against droping the func test on backport | |
| 16:39:56 | artom | elodilles, so that was the crux of my argument | |
| 16:40:09 | artom | Red Hat *will* maintain stable/train for literally years, we have no choice | |
| 16:40:28 | sean-k-mooney | well at least 2.5 more | |
| 16:40:29 | artom | But we don't want to inflict pain on anyone maintaining older than stable/train | |
| 16:40:38 | artom | So: do those folks... well, exist? :) | |
| 16:41:01 | artom | At Summit, what release did operators say they were on? | |
| 16:41:05 | sean-k-mooney | technially we still maintain queens downstream too for a while more | |
| 16:41:42 | sean-k-mooney | but i would prefer to have the backports of the func infra as that makes backporting simpelr in the long run | |
| 16:41:45 | gibi | artom: there were no specifics other than what is in the etherpad | |
| 16:42:01 | artom | So only "train or older" with no info if it's train... or older :( | |
| 16:42:43 | elodilles | should have been added 'train' + 'stein and older' :D | |
| 16:43:24 | artom | Yeah :S | |
| 16:44:34 | artom | So compared to stable/train, there are 4 patches last updated this year, compared to train's ~50 | |
| 16:44:53 | elodilles | why not https://review.opendev.org/q/project:openstack/nova+branch:stable/stein ? | |
| 16:45:04 | artom | Because I suck at alphabet :P | |
| 16:45:09 | elodilles | o:) | |
| 16:45:36 | artom | Seems to be mostly Felix and Vlad Gusev... | |
| 16:45:50 | artom | But similar level of involvement drop-off | |
| 16:46:21 | elodilles | a bit more patch but without reviews, yes :/ | |
| 16:46:24 | gibi | so what if we say, func infra backport are OK to stable/train as there are maintainers there but not further backwards | |
| 16:46:44 | gibi | due to lack of maintainers | |
| 16:47:04 | elodilles | gibi: that is good for RH but not really helps to encourage backporting for older branches | |
| 16:47:26 | sean-k-mooney | well even train is in em now right | |
| 16:47:34 | artom | I think elodilles's point is that if we rewrite the fixes to not need func test refactors *before* train, it helps maintainers of older branches, such as they are | |
| 16:47:48 | sean-k-mooney | given the peopel we have its hard to keep maintianing older branches | |
| 16:47:57 | keerthi | can some one help on review this https://blueprints.launchpad.net/nova/+spec/define-max-volume-limit-at-flavor ? | |
| 16:47:58 | sean-k-mooney | train is 5 releases old currently | |
| 16:48:03 | artom | If we backport func test refactors, we're offloading that rewriting work onto whoever is still working with rocky and older | |
| 16:48:14 | gibi | keerthi: we are in a meeting right now | |
| 16:49:04 | artom | OTOH, why would it be wrong of facilitating the work of the majority? | |
| 16:49:10 | gibi | so we say no func test backport as it there is no maintainers but also we say if we dont backport func infra then we dont have maintaniers, this is contradiction now | |
| 16:49:12 | sean-k-mooney | keerthi: we can proably discuss it after the meeting or when we are done with this topic | |
| 16:49:31 | keerthi | sure Sean, I will wait for it | |
| 16:50:23 | elodilles | anyway, i have said my preference, and i'm only one of the stable maintainers o:) | |
| 16:50:52 | gibi | either we don't have stein maintainers and then I don't feel back about not helping them, or we have maintainers and the I would ask them to backport the func refactor from train to stein | |
| 16:51:01 | gibi | s/back/bad/ | |
| 16:51:45 | gibi | and I would help them by backporting the refactor up until train | |
| 16:51:56 | gibi | that would be a win-win | |
| 16:52:00 | artom | gibi, hard to argue with that | |
| 16:52:37 | gibi | elodilles: would you be -1 on a func infra backport to stable/train? | |
| 16:52:46 | gibi | (or even -2?) | |
| 16:53:01 | elodilles | 'func infra'? | |
| 16:53:12 | gibi | functional test infrastructure backport | |