Monday, 2025-12-08

*** liuxie is now known as liushy02:15
*** mrunge_ is now known as mrunge06:32
opendevreviewMerged openstack/project-config master: Use ubuntu-noble for publish-openstack-artifacts  https://review.opendev.org/c/openstack/project-config/+/96336711:50
opendevreviewMichal Nasiadka proposed opendev/zuul-providers master: Add debian-trixie-nested-virt-8GB  https://review.opendev.org/c/opendev/zuul-providers/+/97011512:03
*** dhill is now known as Guest3313212:30
clarkbinfra-root I'll plan to approve https://review.opendev.org/c/opendev/system-config/+/969847 and https://review.opendev.org/c/openstack/project-config/+/969846 shortly if no one beats me to them as general cleanups after the gerrit upgrade. Then I'mgoing to see if I can get gerritlib's integration testing running against 3.11 before we land the acl update test change15:46
fungilooks like i already reviewed those, happy to approve them now15:47
clarkbfungi: thanks!15:47
fungiand done15:47
clarkband if I can't figure out the integration testing in a reasonable amount of time we can probably proceed with testing in production :/15:49
clarkbanyway something to eat and drink then I'm digging into that15:49
opendevreviewMerged openstack/project-config master: Update Jeepyb's Gerrit builds to Gerrit 3.11  https://review.opendev.org/c/openstack/project-config/+/96984615:51
opendevreviewMerged opendev/system-config master: Update infra-prod review and manage-projects deps for new Gerrit  https://review.opendev.org/c/opendev/system-config/+/96984715:56
opendevreviewMerged openstack/project-config master: Add SeaweedFS App to StarlingX  https://review.opendev.org/c/openstack/project-config/+/96952716:02
clarkbnote ^ will test manage-projects16:03
clarkbwe can probably abandon my bindep.config update change16:03
clarkbdeployment jobs are behind the hourlies so we have a minute to stop things if we want. But I guess its probably best to just rip the bandaid off and see what happens16:05
opendevreviewClark Boylan proposed opendev/gerritlib master: Update integration testing to test with Gerrit 3.11  https://review.opendev.org/c/opendev/gerritlib/+/97014216:08
clarkbmanage-projects is running now16:11
clarkbmanage-projects reports success. I'm going to look at logs now16:15
clarkbthe logs look ok to me and the project seems to exist with a config that looks correct: https://review.opendev.org/admin/repos/starlingx/app-seaweedfs,general (requiring signed off by is set to True for example which is not the default as far as I know)16:17
clarkbI've marked that item done on the etherpad and will abandon my noopy bindep config update change16:19
fungiyeah, i figured that was a more useful test16:20
clarkbdouble checking the All-Projects config for Blocked Users addition (or not) was on my todo list but that isn't super urgent so I'm going to focus on updating the gerritlib testing and filing the delete project deleting nonexistant file bug first16:21
clarkbhappy for someone else to do that one too. But I should be able to get to it eventually and the only change on our side is the docs update if it did change in prod so not really a big deal to get done right away16:22
opendevreviewClark Boylan proposed opendev/gerritlib master: Update integration testing to test with Gerrit 3.11  https://review.opendev.org/c/opendev/gerritlib/+/97014216:23
clarkbtrixie is python3.13 too right? Once this is working with gerrit 3.11 I should update the test platform for trixie to get python3.13 to better mimic prod and then update the unittests to include 3.13 as well16:28
fungiyes16:29
mnasiadkaclarkb: probably you’re busy with gerrit upgrade fallout - but it seems https://review.opendev.org/c/zuul/zuul-jobs/+/966187 is ready for another review16:29
clarkbthat is more about coverage for the java 21 move on trixie than anything else but may as well get it sorted while I'm looking at this stuff16:29
clarkbmnasiadka: ack I'll try16:29
mnasiadkaAnd if anybody can have a look in https://review.opendev.org/c/opendev/zuul-providers/+/970115 - I would be grateful (we use nested virt nodes for kvm based testing in Kolla-Ansible CI)16:30
clarkbheh and I immediately hit the docker hub rate limit16:33
clarkbour quay mirror only has gerrit:latest. But maybe that is good enough for now16:34
clarkbif ^ works then we can mirror 3.11, 3.12, 3.13 too and pull them from the mirror and cover them as well ro something16:34
mnasiadka970115 now only needs +w, because it has 2 * +2 ;-)16:36
opendevreviewMerged opendev/zuul-providers master: Add debian-trixie-nested-virt-8GB  https://review.opendev.org/c/opendev/zuul-providers/+/97011516:43
mnasiadkathanks16:44
opendevreviewClark Boylan proposed opendev/gerritlib master: Update Gerrit integration testing to test many Gerrit versions  https://review.opendev.org/c/opendev/gerritlib/+/97014216:44
opendevreviewClark Boylan proposed opendev/gerritlib master: Add Python 3.13 testing  https://review.opendev.org/c/opendev/gerritlib/+/97014416:46
opendevreviewClark Boylan proposed zuul/zuul-jobs master: Add base py313 jobs for tox and nox  https://review.opendev.org/c/zuul/zuul-jobs/+/97014616:52
opendevreviewClark Boylan proposed opendev/gerritlib master: Add Python 3.13 testing  https://review.opendev.org/c/opendev/gerritlib/+/97014416:53
opendevreviewClark Boylan proposed zuul/zuul-jobs master: Add base py313 jobs for tox and nox  https://review.opendev.org/c/zuul/zuul-jobs/+/97014616:58
opendevreviewClark Boylan proposed opendev/gerritlib master: Update Gerrit integration testing to test many Gerrit versions  https://review.opendev.org/c/opendev/gerritlib/+/97014216:59
opendevreviewClark Boylan proposed opendev/gerritlib master: Add Python 3.13 testing  https://review.opendev.org/c/opendev/gerritlib/+/97014416:59
clarkbthe integration testing against gerrit:latest passed so I think we are good16:59
clarkbI'll let reviewers decide if we want to do the full 3.11, 3.12, 3.13 matrix and whether we want to mirror all of those to quay before we rely on them17:00
opendevreviewClark Boylan proposed zuul/zuul-jobs master: Add base py313 jobs for tox and nox  https://review.opendev.org/c/zuul/zuul-jobs/+/97014617:12
clarkbinfra-root ^ I believe all three of those changes should be ready for review. The zuul-jobs update is the only one that ahsn't come back happy yet but the latest patchset should update bindep to make it happier17:13
clarkband none of those directly affect production gerrit so should be safe to land. They just help us ensure we're testing jeepyb, gerritlib, and gerrit together so that when we do upgrades there aren't surprises17:13
opendevreviewClark Boylan proposed opendev/system-config master: Drop Gerrit 3.10 image builds  https://review.opendev.org/c/opendev/system-config/+/97015017:17
opendevreviewClark Boylan proposed opendev/system-config master: Upgrade build and runtime for Gerrit to Java 21  https://review.opendev.org/c/opendev/system-config/+/97016017:36
clarkbI don't think we need to be in a huge rush for ^ the bigger thing is getting it working in CI so that I can get Gerrit 3.12 also working in CI17:37
clarkbwe probably want to schedule the java runtime swap for a time where we can pay attention to it etc17:37
opendevreviewClark Boylan proposed zuul/zuul-jobs master: Add base py313 jobs for tox and nox  https://review.opendev.org/c/zuul/zuul-jobs/+/97014617:48
fungiclarkb: that ^ makes flake8 unhappy because of importing unused modules18:01
fungii can think of a few ugly hacks to avoid that, or just silence its objection18:02
opendevreviewClark Boylan proposed zuul/zuul-jobs master: Add base py313 jobs for tox and nox  https://review.opendev.org/c/zuul/zuul-jobs/+/97014618:05
clarkbfungi: ya I went with just silence its objection. I think doing the imports is the thing that most closely matches the code we are trying to test so the rule just dones't make sense here18:05
fungithat's what i'd have done too18:05
fungii'm going to go grab lunch, bbiab18:06
cardoeSo what I am doing wrong with "git review"... I get "! [remote rejected]     HEAD -> refs/for/master (no new changes)"18:09
cardoeI've got 1 commit locally in my branch from master for keystone.git18:09
clarkbcardoe: (no new changes) means what you are trying to push already exists18:10
cardoeIt doesn't though18:10
cardoeI literally just rewrote the doc locally.18:10
cardoeUnless I managed a sha1 hash collision?18:11
clarkbor forgot to git add or commit ot something? We'd need more info I think18:11
clarkbgit log and git show HEAD may be useful.18:12
cardoehttps://opendev.org/openstack/keystone/commit/50f0f023bdd23bea8b9c21823a569a159af9ebeb18:12
clarkbbasically info that describse the staet of your local tree that you are trying to push18:12
cardoeThat's my commit18:12
cardoeIt's not in master I'm surprised its appearing on opendev.org though18:12
clarkbhttps://review.opendev.org/c/openstack/keystone/+/97016618:13
clarkbwe replicate unmerged commits18:13
clarkbis it possible you ran git review twice and the first one succeeded and the second failed?18:13
cardoeI mean I've run it a few times cause it's been erroring at me each time.18:13
* cardoe shrugs.18:14
clarkbok, I'd need a better account of what you did locally in terms of git review execution to understand if there was actually a problem or if it was maybe instigated by an expected error (missing signed off by for example) then dobule tapped after without realizing the first succeeded adn second failed with no new changes18:15
clarkbits up to you if you want to do that. It does seem to have worked18:15
clarkbwere you using claude to push?18:15
clarkbdid claude try multiple times and only report the last failure?18:16
cardoeNope. claude doesn't have that tool as permissions.18:16
cardoeThe very first time I got a timeout error actually.18:16
cardoeGoing back to the scroll back in my terminal.18:16
* clarkb suddenly wonders how claude permissions are enforced. but agree that seems like a thread not worth pursuing unless all others are exhausted18:17
clarkbok my hunch is that initial timed out connection did enough to create the change but maybe didn't get back the "here's your change number" response in time18:17
cardoeyeah that's the most likely situation.18:17
cardoeI just assumed it didn't exist and if it had I would get back a different error message.18:18
clarkband a likely explanation for that is: we upgraded gerrit yesterday and in the process deleted a bunch of caches because they grow unbounded and got too large. And we're hitting cold cache issues. Though it look like stephenfin pushed some keystone changes slightly before you18:19
clarkbno new changes is a telltale sign that what you're pushing exists fwiw. And gerrit is saying I don't want to bother updating something that has no updates18:19
clarkbcardoe: in the sshd_log for gerrit I see you connect from an ipv4 address when I think the timeout occurs. There are three connections around then one of which seems to work and one of which seems to be killed. I don't understand the other. Then your subsequent connections come over ipv618:23
clarkboh wait no there are four. Three look normal. The fourth via ipv4 is killed. Then you come in via ipv6 and all seems well from there18:23
clarkbnote git review may make more than one connection so this doen't mean you ran git review multiple times when this happened18:24
cardoeSo I assumed that my VPN was messing with it so I stopped the VPN and I bet that's when it switched to IPv6.18:24
clarkbMy new hunch is that git review ran and did enough to make the chagne but then some subroutine in git review failed beacuse its connection timed out/got killed18:24
cardoeuser error is a good enough answer for me. :-D18:26
cardoeI didn't think to check that it was created.18:26
cardoesorry for the noise.18:26
clarkbya the only traceback I can find is "channel already closed"18:27
clarkbcardoe: git review does have a verbose flag (-v iirc) which might eb useful if you find this happening frequently. That might help us narrow down where things go sideways18:28
clarkbbut also looks like it may have been on the order of ~15 seconds?18:28
clarkbgit can be slow18:29
cardoecorporate VPN also likes to fiddle with SSH18:29
clarkb15 seconds isn't necessarily a long time to wait depending on the repo. As both sides have to negotiate what commits they know about along the common dag path then construct and exchagne the missing data18:29
opendevreviewMerged opendev/system-config master: Drop Gerrit 3.10 image builds  https://review.opendev.org/c/opendev/system-config/+/97015018:30
clarkbhttps://zuul.opendev.org/t/openstack/build/3730086c8554461aa74ec28f7966c055/log/review99.opendev.org/logs/error_log#32 our first Gerrit 3.11 run on Trixie with Java 2118:31
clarkbLooking at the logs we're using java 21 to build a war that targets java 17 still18:33
clarkbwhich I think is fine18:33
clarkbjust want to note it as we aren't fully converted to 21 I guess until we get to 3.12 where java 17 is dropped entirely and they update the build target18:33
clarkbalso I don't think we've converted anything else to trixie first. It probably isn't the worst idea to convert something else first like irc bots maybe?18:44
clarkbthoughts on that?18:44
clarkbalso, I raelize I never got a clear answer on whether or not yall wanted me to followup with EMS to ask if we can use duplicate email addresses. Should I go ahead and do so?18:45
*** jgwentworth is now known as melwitt19:01
fungior we could just test the theory ourselves19:24
clarkbmy concern with doing that is associating an email address we'd want billing or whatever to go to with a specific account and breaking that system19:25
clarkbbut we could use a single plus addressing address for all the bots and test if duplicatse of that one are ok19:25
opendevreviewClark Boylan proposed opendev/system-config master: Add Gerrit 3.12 image builds and testing  https://review.opendev.org/c/opendev/system-config/+/97017319:36
clarkbof the pile of changes I've pushed so far I believe that https://review.opendev.org/c/zuul/zuul-jobs/+/970146 https://review.opendev.org/c/opendev/gerritlib/+/970144 and https://review.opendev.org/c/opendev/gerritlib/+/970142 should be mergable19:40
clarkband I just double checked that bookworm doesn't have openjdk-21 in backports (it doesn't appear to: https://packages.debian.org/search?suite=bookworm-backports&searchon=names&keywords=openjdk) So I think we do have to bump up to trixie19:41
clarkbok I rtfs'd that delete project exception we hit: https://gerrit.googlesource.com/plugins/delete-project/+/refs/tags/v3.11.7/src/main/java/com/googlesource/gerrit/plugins/deleteproject/fs/DeleteTrashFolders.java is the relevant code. The way that plugin works is it iterates through directories looking for directories that match one of three patterns. In this case the path that hit19:47
clarkbthe NoSuchFileException does not match the paths so it wouldn't try to delete anything. Instead something else mus've been writing and removing that file/dir and this plugin saw it and they raced each other19:47
clarkbPretty sure this should be a totally benign issue for us, but I'll file an issue with what I've learned after lunch because unnecessary exceptinos are annoying19:48
clarkb970173's build did build targetting java 21 so I think that confirms it20:12
opendevreviewClark Boylan proposed opendev/system-config master: Add Gerrit 3.12 image builds and testing  https://review.opendev.org/c/opendev/system-config/+/97017321:00
opendevreviewClark Boylan proposed opendev/system-config master: Test Gerrit 3.11 to 3.12 Upgrades  https://review.opendev.org/c/opendev/system-config/+/97018021:00
clarkbhttps://issues.gerritcodereview.com/issues/467103128 issue is filed21:22
clarkbI diagnosed it to the point where I could probably fix it, but I don't have my google cla signed yet so won't do that yet. However, I'm hoping that gets done soon as I have requested it. If it does happen I should be able to determine which of the two theories I posted in the issue is the correct one and fix it21:23
clarkbI'm going to work on putting together our meeting agenda shortly. Let me know if there is anything you'd like added or edited. I plan to recap the gerrit upgrade and cover the next steps. I'll add a note about fungi's plan for renaming that one starlingx project Friday at 2100 UTC (I think that was the proposed time). And then add in Matrix updates21:25
clarkbAnything else?21:25
funginothing on my end21:27
fungii do need to send out a (fairly short notice) announcement about the rename maintenance on friday, but shouldn't be a big deal since those are typically <5min downtime for gerrit21:28
clarkbThe last major task on the post upgrade todo list that I haven't done anything to start yet is checking the all-projects refs/meta/config log to see if we added Blocked Users to the acl. In the command I've got I run a git log -p against the git repo on the host. It occusr to me now that I'm not positive that is 100% safe to run against a git repo that gerrit is operating on. I think21:30
clarkbit is safe and I'm pretty sure I tested it with gerrit running on the test node21:30
clarkbany concerns with that? It is a read only command against the repo and we run git gc against the repos outside of gerrit which is read write so I think just checking the log should be fine21:30
fungino concerns on my part, seems like it should be perfectly safe yes21:31
clarkbok I'm going to run it now21:31
clarkb76473b67b78389fadef983291787e88cc1cb80f9 is HEAD and is the change fungi pushed in July to cleanup cla config21:32
clarkbthis means it did not update the acl for us which means we don't need to update our docs21:32
clarkbI'll leave a comment on my docs change indicating it wasn't automatic for us and if we want to apply that change we will need to manually update the acls too21:33
fungithanks!21:33
clarkbcomment posted to https://review.opendev.org/c/opendev/system-config/+/968228 and WIP is removed21:35
clarkbI could go either way. Since we allow anonymous access people would be able to logout and see/fetch everything so really all this would prevent is them writing things21:36
clarkband in that case maybe we're better off simply disabling their account which is what we've done for years21:37
clarkbgood news: looks like the entire java 21, gerrit 3.12 image, 3.11 -> 3.12 upgrade stack is going to pass. I think the main considering with merging those is doing so when we can restart to pick up the trixie + java 21 switch so not something for today though reivews are welcome any time21:41
clarkbI think personally I'd like to update something other than gerrit to trixie before switching gerrit21:41
clarkbjust as a sanity check that trixie as we build it in our container images is generally functional21:42
fungiyeah, that seems like a safe plan21:42
clarkbI need to go apply hashtags to changes as I've been bad about that. I'll link to that list here when done and then work on our meeting agenda21:42
clarkbhttps://review.opendev.org/q/hashtag:%22upgrade-gerrit-3.11%22+status:open21:44
clarkbok the agenda is updated. I'll send that later today in case there are any more updates we can think of22:06
opendevreviewGoutham Pacha Ravi proposed opendev/infra-manual master: Alter the repo retirement README template  https://review.opendev.org/c/opendev/infra-manual/+/97019122:17
clarkbinfra-root do you want to weigh in on https://review.opendev.org/c/opendev/zuul-providers/+/970042 and my comment there? specifically the proposal is that we stop booting alma on all our clouds and restrict it to the v3 capable clouds like rocky and centos stream. I'm not sure if that sort of redundancy makes sense if I'm honest22:19
opendevreviewMerged opendev/infra-manual master: Alter the repo retirement README template  https://review.opendev.org/c/opendev/infra-manual/+/97019122:30
opendevreviewGoutham Pacha Ravi proposed openstack/project-config master: Cleanup Monasca infra  https://review.opendev.org/c/openstack/project-config/+/97019322:30
*** melwitt is now known as jgwentworth22:31
*** jgwentworth is now known as melwitt22:32
*** dhill is now known as Guest3316922:35

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