| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-09-12 | |||
| 16:38:54 | JayF | I'm not saying OpenStack projects should wait to move to new API; I'm saying we can't flip the requirement to actual-2.0 until distros are ready | |
| 16:39:11 | bauzas | cool, we are on page | |
| 16:39:12 | gmann | yeah, we need to wait for that | |
| 16:39:49 | bauzas | looks like we have a violent agreement, anything to add ? | |
| 16:40:16 | bauzas | if not, let's pretend I'm just giving you back 20 mins of your time | |
| 16:40:26 | bauzas | thanks all | |
| 16:40:28 | opendevmeet | Log: https://meetings.opendev.org/meetings/nova/2023/nova.2023-09-12-16.03.log.html | |
| 16:40:28 | opendevmeet | Minutes (text): https://meetings.opendev.org/meetings/nova/2023/nova.2023-09-12-16.03.txt | |
| 16:40:28 | opendevmeet | Minutes: https://meetings.opendev.org/meetings/nova/2023/nova.2023-09-12-16.03.html | |
| 16:40:28 | opendevmeet | Meeting ended Tue Sep 12 16:40:28 2023 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | |
| 16:40:28 | bauzas | #endmeeting | |
| 16:41:27 | sean-k-mooney | JayF: given SQLAlchemy is now up to 2.0.20 | |
| 16:41:53 | sean-k-mooney | i wonder how much fo the disto lag is becuase openstack and other poject have not moved yet | |
| 16:42:16 | JayF | sean-k-mooney: Like I said; 1.4 is the release that spans both 1.3 and 2.0 supporting projects, so I suspect that's going to be the 'safe place' distros remain for a while | |
| 16:42:32 | JayF | sean-k-mooney: especially since, unlike with some libraries, they won't be able to patch their way into newer versions for any stragglers | |
| 16:43:08 | JayF | sean-k-mooney: To be clear: Ironic is 100% SQLA 2.0 compatible and I am generally a neophile so I'm all about getting on the new version; just when I did the legwork to TC-resolution it gmann and others rightfully pointed out that we can't do it until the distros do it | |
| 16:43:56 | JayF | so I did the legwork and found that sadly, they are correct :( | |
| 16:44:15 | JayF | (despite some folks asserting otherwise at the forum session about it) | |
| 16:44:27 | sean-k-mooney | i think that feedback we shoudl be bringing to our downstream disto contacts | |
| 16:44:33 | opendevreview | Merged openstack/osc-placement stable/2023.2: Update TOX_CONSTRAINTS_FILE for stable/2023.2 https://review.opendev.org/c/openstack/osc-placement/+/894067 | |
| 16:45:21 | sean-k-mooney | currenly nova requires SQLAlchemy>=1.4.13 | |
| 16:45:42 | sean-k-mooney | but i am hopign we can get to and stay at 2.0 compatiblity soon | |
| 16:45:54 | JayF | sean-k-mooney: it's tough because it cuts both directions; we have to ensure *all* openstack projects can run on 2.0 when distros move, and I'm not sure all our projects are going to be able to get there without outside help | |
| 16:46:22 | JayF | so on one hand, faster is better, on the other, having more time to migrate is nice (not that I have real expectation those underresourced projects will prioritize it until it becomes breaky) | |
| 16:46:29 | sean-k-mooney | well i think stephenfin already spend a time moving a lot of projects | |
| 16:46:47 | JayF | get that man a cape with "OSLO" on the back :D | |
| 16:47:22 | JayF | sean-k-mooney: fwiw, I assume you likely know this, but you can make 1.4 warn about 2.0-breaky things in tests: https://github.com/openstack/ironic/blob/master/tox.ini#L17 | |
| 16:47:54 | sean-k-mooney | JayF: yep we already doo | |
| 16:48:02 | JayF | nice | |
| 16:48:19 | sean-k-mooney | https://github.com/openstack/nova/blob/master/tox.ini#L23 | |
| 16:54:59 | sean-k-mooney | JayF: for what its worth there is an experimal 2.0.19 build in debian https://packages.debian.org/source/experimental/sqlalchemy | |
| 16:55:27 | JayF | zigo: ^ do you have any idea for a timeline or liklihood for sqlalchemy hitting 2 in debian? | |
| 16:57:22 | sean-k-mooney | fedora have this tracker https://bugzilla.redhat.com/show_bug.cgi?id=2134559 but it does not seam to be active | |
| 16:58:12 | sean-k-mooney | the kogi build is failing however https://koji.fedoraproject.org/koji/taskinfo?taskID=104893237 | |
| 17:00:56 | sean-k-mooney | JayF: looks like tumbleweed are one of the first to ship it https://opensuse.pkgs.org/tumbleweed/opensuse-oss-x86_64/python311-SQLAlchemy-2.0.19-3.1.x86_64.rpm.html | |
| 17:02:36 | sean-k-mooney | ah they aslso ship python311-SQLAlchemy1 which is 1.4 | |
| 17:02:38 | JayF | That is brand new. I checked there before abandoning my TC resolution a weekish ago | |
| 17:02:52 | JayF | but I might have been looking for sqlalchemy named packages and missed it, with that in mind | |
| 17:03:39 | sean-k-mooney | they ship 2 packages now at least | |
| 17:04:29 | sean-k-mooney | the main package python311-SQLAlchemy is now 2.0 and they have python311-SQLAlchemy1 is 1.4.46 | |
| 21:42:45 | opendevreview | Jay Faulkner proposed openstack/nova master: [ironic] Query for shards with proper, plural key https://review.opendev.org/c/openstack/nova/+/894833 | |
| 21:48:12 | JayF | Sorry about ^ that, details are in the bug and on the change. Glad I hooked it all up in CI instead of relying on my manual testing. | |
| 23:12:24 | zigo | JayF: It's simply waiting for OpenStack to be released, as Antelope is not SQLA 2.x compatible, the maintainer is waiting on me ... | |
| 23:12:58 | zigo | If you guys need it now, I can ask for an upload to Unstable right away... | |
| 23:13:13 | zigo | Though I'd prefer for the release to happen. | |
| 23:13:57 | zigo | What's the situation on Bobcat? Are we 100% SQLA 2.x compatible already? | |
| 23:14:36 | zigo | If not, it's going to be huge troubles for me (unless there's patches I can apply on top of Bobcat). | |
| #openstack-nova - 2023-09-13 | |||
| 00:51:58 | JayF | zigo: bobcat is not 2.0. If you have a need on a timeline please put that on the list, although it won't change bobcat it can help decide future priorities | |
| 07:18:26 | zigo | JayF: I thought it was 2.0. Yes, I do need SQLA 2.0 support in Bobcat, and will write about it in the list, thanks for your care! | |
| 07:28:02 | bauzas | zigo: the situation is quite simple, most projects try to support the new DSL of 1.4 | |
| 07:28:22 | bauzas | which is considered a bridge release speaking old pre-1.4 and new 2.0 style | |
| 07:29:12 | bauzas | but then we need some guidance, as for the moment, none of the distros ship SQLA 2.0 | |
| 07:29:38 | bauzas | it's then a classic chicken and eggs situation, if everyone awaits the other :D | |
| 07:36:31 | frickler | also see https://review.opendev.org/c/openstack/requirements/+/879743 for a (possibly incomplete) list of projects that still struggle | |
| 07:38:20 | zigo | Well, for Debian, you should consider that Experimental represent the next OpenStack stable, so SQLA 2.x is the way ... | |
| 07:38:39 | zigo | frickler: Where do I see it in that patch?!? | |
| 07:41:58 | frickler | zigo: look at the failing cross jobs. I just saw https://review.opendev.org/c/openstack/requirements/+/887261/comments/6b1f2573_60f48f1c has some more detailed analysis of the issues. but this is getting a bit offtopic for the nova channel | |
| 08:04:57 | zigo | frickler: So, if I read it well, we are down to only 4 packages, the biggest concern being heat, but it's probably still working, right? | |
| 08:11:55 | frickler | heat seems very strongly broken, see https://review.opendev.org/c/openstack/heat/+/886549 and also https://08a8510b7f90952cf4e9-7691b2eb12cb7f63209fe7ead2f9c479.ssl.cf1.rackcdn.com/887261/2/check/openstacksdk-functional-devstack/f9d3c13/controller/logs/screen-h-api.txt | |
| 09:10:25 | bauzas | stephenfin: please, just one again needed modification for sg associations in https://review.opendev.org/c/openstack/nova/+/860829 | |
| 09:32:04 | sean-k-mooney | zingo so timeline wise caracal (2024.1) will still be based on ubuntu 22.04 and the D cycle 2024.2 shoudl be based on 24.04 | |
| 09:32:43 | sean-k-mooney | zigo: ^ for nova i woudl liek to merge the final patches so that we can run on 2.0 and im hoping the rest of openstack catches up in caracal | |
| 09:33:04 | sean-k-mooney | but im not exepcting it to become our default until D | |
| 09:33:41 | sean-k-mooney | althoguh i woudl be fine with nova runing most of our jobs with 2.0 provided we still had 1.4 coverage in at least one job | |
| 09:34:21 | sean-k-mooney | right now we have the opisite situation where we run most of our jobs on 1.4 but have a singel job testing 2.0 and we enabel all the wranings in yoru unit/funcitonal tests | |
| 09:34:55 | sean-k-mooney | once we are 2.0 comptaible it woudl be really nice if we could make that warning an error by the way | |
| 09:36:01 | sean-k-mooney | zigo: i dont know if you can take the same approch tumbleweed did nan have a second package for SQLA | |
| 09:36:24 | sean-k-mooney | they moved the main package to 2.0 and added a second package to track 1.4 | |
| 09:36:24 | zigo | sean-k-mooney: I can't, that's not how Debian works. | |
| 09:36:45 | zigo | There would be a path clash ... | |
| 09:37:08 | zigo | "import sqlalchemy" would need to be rewritten to "import sqlalchemy1" or something ... | |
| 09:37:16 | sean-k-mooney | sure on any one system | |
| 09:37:19 | zigo | Plus many many hacks. | |
| 09:37:19 | sean-k-mooney | you could only have one | |
| 09:37:37 | sean-k-mooney | but you could declare one package as conflicting with the other | |
| 09:38:01 | zigo | First, that's forbidden by the Debian policy to do that. Second, I'm not packaging for chroot or venvs... packages are supposed to be co-installable. | |
| 09:38:25 | sean-k-mooney | well tere are packages on debian that do that | |
| 09:38:33 | sean-k-mooney | and yes i know they are ment ot be co installable | |
| 09:38:58 | sean-k-mooney | i was just giving you options since bobcat wont be 2.0 compatiable in its entirity | |
| 09:39:04 | sean-k-mooney | althoguh caracal should be | |
| 09:39:51 | zigo | It's packages that are supposed to be different implementation of the same functionality, like gawk vs mawk. This is *NOT* for having multiple versions of the *SAME* package. | |
| 09:40:06 | zigo | So, your option is unfortunately not an option ... :/ | |
| 09:40:29 | zigo | sean-k-mooney: Really, my only viable option is to have patches backported from master ... | |
| 09:40:31 | sean-k-mooney | ya where i saw it was differnt pipewire backend plugins | |
| 09:42:17 | sean-k-mooney | zigo: so it was not a comunity goal to get this done this cycle | |
| 09:42:23 | sean-k-mooney | so they may or may not be written | |
| 09:42:49 | sean-k-mooney | that said several of us really wanted to get to that point since it was discussed as a possibel goal a few times | |
| 09:42:50 | zigo | This fact (that it was not a comunity goal) doesn't change anything to the situation, unfortunately ... | |
| 09:43:30 | zigo | I can live with a broken heat package in Unstable for some time, that's ok-ish, the backport will probably work. | |
| 09:43:35 | sean-k-mooney | well the question is do you need bobcat to run with 2.0 for debian or is the 2.0 change going to happen later | |
| 09:43:48 | zigo | But it has to be taken seriously and it'd be best if we have a patch soon. | |
| 09:44:05 | zigo | I'd have to discuss with the SQLAlchemy maintainer... | |
| 09:44:21 | zigo | Though he's pressing to have the package migrate to Unstable ... | |
| 09:44:23 | sean-k-mooney | what im really getting at is what was your orgianl plan and timeline | |
| 09:44:39 | sean-k-mooney | ya so i think it woudl be goood to get it in unstable | |
| 09:44:50 | sean-k-mooney | so that we can hopefuly get it in ubuntu 24.04 | |
| 09:44:53 | zigo | My original plan was to have Bobcat support 2.0, and upload it together with OpenStack when Bobcat is released. | |
| 09:45:02 | sean-k-mooney | ack | |
| 09:45:19 | zigo | I may follow that path even with a broken heat package in Unstable... | |
| 09:45:32 | sean-k-mooney | for what its worth i was really really hoping all openstack whoudl have got to 2.0 compatibliy in bobcat too | |