Monday, 2022-01-10

*** ysandeep|out is now known as ysandeep05:35
fricklerany ansible cores around? this seems to affect at least kolla and tripleo
opendevreviewchandan kumar proposed zuul/zuul-jobs master: Introduce iptables_package local var
*** ysandeep is now known as ysandeep|afk07:05
fricklerzuul-jobs linting is also broken by that
*** ysandeep|afk is now known as ysandeep08:36
*** jpena|off is now known as jpena08:51
*** gibi_pto_back_on_10th is now known as gibi08:59
dtantsurfrickler: IIRC cirros did not have a real root partition for us to operate on. I'm looking at cirros-0.5.1-x86_64-rootfs.img now, and it does seem to have kernel/initramfs inside, as well as some grub files, but not grub-install or really any executables.10:23
dtantsurfrickler: and the UEC image has a completely empty root partition10:25
dtantsur(it even has "blank" in its name)10:25
dtantsurfrickler: for the context: Ironic needs to test (in the CI) writing a partition (= having only the root partition) image to the target disk and installing grub.10:39
dtantsurcurrently we only have one job that achieves that via a hand-crafted centos image. which is, at best, suboptimal.10:39
fricklerdtantsur: hmm. we could publish the rootfs for cirros, but that doesn't include grub. the grub install is only being executed external to cirros during the bundling. do you need grub itself within the image?10:57
dtantsurfrickler: yes10:57
dtantsurotherwise we cannot test installing the bootloader10:58
*** ysandeep is now known as ysandeep|coffee11:08
*** rlandy|out is now known as rlandy|ruck11:16
*** dviroel|afk is now known as dviroel11:22
*** ysandeep|coffee is now known as ysandeep11:54
fricklerdtantsur: did a bit of testing, to build grub, buildroot needs wchar enabled, which increases the whole image size by more than 50%. so I don't think that this is feasible for a generic cirros image. one might make that a variant, but then we'd likely have to build it ourselves.12:19
fricklerdtantsur: can you point me to how you are running these tests? I might want to try locally with some self-built cirros image12:19
dtantsurnot exactly trivial, you'll need a proper ironic devstack environment.. lemme see if I can find a suitable job12:20
dtantsurfrickler: so, basically but removing IRONIC_DEFAULT_BOOT_OPTION or setting it to "local"12:21
dtantsurfrickler: local.conf from it
dtantsur(again, IRONIC_DEFAULT_BOOT_OPTION=netboot is our workaround for not having local boot facilities, we're trying to remove that)12:24
fricklerdtantsur: o.k., so you don't actually use this centos image you mentioned in the CI currently? or did I misunderstand that? anyway I think I can set something up with those hints, thx12:28
dtantsurah, that, hold on, it's another job12:50
dtantsurfrickler: using
*** pojadhav is now known as pojadhav|dr_appt13:17
fungifrickler: it looks like the error you linked is consistent with what pr #1797 was fixing. do you know whether the error is continuing to surface? apparently we either need to use ansible_lint>=5.3.2 or pin rich<11 according to
fricklerfungi: 5.3.2 was tagged only just now, it solves the issue in my local test, will drop some rechecks now13:41
chandankumarhello infra, whom can be bug to get this job fixed?13:44
fricklerchandankumar: not sure if anyone is still interested in keeping opensuse builds working, possibly they should be disabled/made non-voting until that happens13:48
fungiagreed, it looks like the opensuse-specific job is already failing and set nonvoting, leading me to suspect support for it has been abandoned13:50
fungii guess it began breaking the functest job around the start of the year, judging from the build history13:51
chandankumarthanks fungi frickler, proposing a patch to make it nv13:51
fungichandankumar: dib-functests shouldn't be nonvoting, it should probably just stop testing the opensuste element13:51
chandankumarsure, will do that13:52
fungiand the readme for that element should probably be adjusted at the same time to say it's no longer tested and suspected to be broken, if we don't remove the element outright13:53
dtantsurwould that lead to eventually removing opensuse nodes from the CI?14:05
fungilikely, yes. i expect we're already unable to update them14:06
dtantsurokay, we've deprecated it in bifrost already (because of python 3.6 mostly)14:07
funginevermind, looks like we're still able to build opensuse-15 images in nodepool as of 10 hours ago14:08
fungihowever we don't have any booted14:08
funginot sure if that's because they're failing to boot or because nobody's got any jobs currently running for it14:08
funginodepool logs indicate we're booting them and jobs are using them still, so i guess not completely broken yet14:10
zigoHi there! One quick question: is it that it used to be that the refstack list contained test_name[id-2394872(smoke)] and now it's test_name[id-2394872,smoke] instead or something ?14:11
zigoI', seeing inconsistencies that leads to not found tests I believe...14:12
fungizigo: you might have more luck asking in #refstack or #openstack-qa14:31
zigoThanks, trying.14:32
dtantsur ran a suse job, which did succeed14:32
fungizigo: hopefully the tempest maintainers in #openstack-qa know how to identify test moves, but the refstack maintainers may also already be aware of that one14:32
fungidtantsur: thanks, so yes i guess it's still working, maybe some corner case has broken impacting how dib's functtest exercises the opensuse element14:33
zigoI had a hard time figuring out that refstack insist soooo much on using venv, it was very annoying, but got that part tackled out, and now tests are running! :)14:33
*** rlandy|ruck is now known as rlandy|ruck|brb14:36
chandankumarfungi: Do We also need to remove opensuse/opensuse15-build-succeed funtest?14:36
fungichandankumar: i'm not sure, hopefully the dib maintainers in #openstack-dib can provide better guidance on what they'll accept14:37
opendevreviewchandan kumar proposed openstack/diskimage-builder master: Remove opensuse/build-succeeds funtests
*** ykarel is now known as ykarel|away14:56
*** rlandy|ruck|brb is now known as rlandy|ruck14:59
*** dviroel is now known as dviroel|lunch15:02
*** dviroel|lunch is now known as dviroel16:07
opendevreviewchandan kumar proposed openstack/diskimage-builder master: Remove opensuse related funtests
*** ysandeep is now known as ysandeep|out16:17
*** Guest2 is now known as prometheanfire16:35
clarkbinfra-root if you have time for I think a change like that is a good one to start getting back into the swing of things in the new yaer (upgrade gitea to catch up on some bug fix relaeses)16:51
clarkbdtantsur: fungi: we've already committed to removing opensuse images baecuse no one has stepped into keep them16:51
clarkb(we said this almost a month ago now)16:51
fungioh, right, i forgot we announced that at the end of the year16:52
clarkbcentos 8, opensuse 15 and tumbleweed, and gentoo were all on the chopping block. We got a volunteer for gentoo so it will stick around. centos 8 is EOL so not much we can do and the o[pensuse images just aren't cared for16:52
clarkbI run opensuse at home and would love to keep them going but I just don't have the time and it isn't that important for my home usage16:52
clarkbRemoving tumblweed was on my "Hope to get that done this week" for this week list16:53
fungiyeah, loci folks noticed last week that something spontaneously broke in their centos-8 jobs, and concluded they should switch to stream16:53
clarkbI've still got some paperwork to do, but that is starting to get cleared off my desk so I can hopefully give time to things like the gitea upgrade above. There is also a gerrit project rename plugin that I'd like to add to our gerrit images and then test that it works with our CI stuff that already does renames16:56
clarkb(my major concern with that lpugin is it does a lot of stuff and does it online and I worry it might be very slow to rename individual projects)16:57
fungii wonder if there would be a possible replication race condition between renaming a project in gerrit live and renaming in gitea17:17
clarkbI think there is17:18
clarkbBut if we rename and then trigger replication we should be ok17:19
fungias long as nothing updates the repo in between those steps17:19
fungi(including something as simple as someone pushing a revision of a change in review, since we replicate those to gitea as well)17:20
opendevreviewClark Boylan proposed openstack/project-config master: Add gerrit rename-project plugin to Zuul
clarkbfungi: right we have to replicate after we rename both gerrit and gitea, then we should be fine17:21
clarkb(and I'm not sure the order between gerrit and gitea matter much if you do the explicit replication at the end17:21
fungimy main worry is that if we rename in gitea first and then in gerrit but a change updates gerrit in between those, then we end up recreating the old repo on the gitea side, but maybe instead it just errors because gitea needs an explicit create step anyway?17:24
clarkbno, gerrit cannot create repos in gitea17:26
clarkbya the push will be attempted and fail17:27
fungiseems fine in that case17:28
clarkball of this should be tested and I'm not sure we should bother with the next rename since people want that soon17:31
clarkbBut  Ifigure we can start looking at it for the future17:32
clarkband our CI can help a lot :)17:32
fungiyeah, i agree it's a great idea17:33
fungidoes the new rename plugin actually move review history? acls?17:33
clarkbacls are in the repo but yes it is my understand it does what we do but also triggers reindexing internally while online. It may even updates project watches for users (not sure about that)17:34
opendevreviewClark Boylan proposed opendev/system-config master: Better organize gerrit plugins in job defs
opendevreviewClark Boylan proposed opendev/system-config master: Use rename-project plugin to rename projects
clarkbworst case we find it doesnt' work for us and we can keep doing what we do17:34
fricklerzuul-jobs linter is still failing because ansible-lint is capped by guess we would need to cap rich for now, unless there will be stable backports of the fix. zbr any idea on that?17:36
fungioh, right, changes and acls are all trivially moved thanks to everything ending up in-repo with notedb17:37
fungiacls already were in fact17:37
clarkbfrickler: the capping isn't complete I guess so some new library causes us to break?17:38
fricklerclarkb: yes, some obscure dependency (rich) got an incompatible update17:40
fungiright, we'd have to add rich<11 to the test-requires17:41
*** jpena is now known as jpena|off17:53
*** odyssey4me is now known as Guest13517:59
clarkbinfra-root I've started putting our meeting agenda together. I trimmed out some stuff and added project renames to the list. Anything else to add today before I send that out?18:44
clarkbalso infra-root any reason I should go ahead and attempt the refstack bullseye image update: ?18:46
clarkb*any reason I should NOT go ahead?18:46
fungiclarkb: did they start testing with 3.9?19:06
clarkbnot that I see. Last change was in november19:08
clarkbwhich is older than my image update19:08
fungii don't object to trying anyway, it's _probably_ fine?19:09
clarkband if it isn't will anyone notice? :P19:10
clarkbI guess I can ask kopecmartin to sign off on it first. kopecmartin can you review and let us know if you have any objections to that image update for the refstack deployment?19:10
clarkbfrickler: good catch19:19
opendevreviewClark Boylan proposed opendev/system-config master: Update refstack image to bullseye
clarkbok with that largely out of the way I'm ready to look at the ansible lint rich thing. Do we have fixes yet or should I look into those?19:21
fungii think it's just zuul-jobs where ansible-lint is pinned?19:22
fungilatest ansible-lint release already includes a fix19:23
clarkbopenstack/project-config is similarly broken19:24
clarkbI'll try fixing openstack/project-config really quickly19:24
fungibut yeah, anywhere we pin ansible-lint and not rich we'll probably see the same error19:25
*** artom__ is now known as artom19:25
opendevreviewClark Boylan proposed openstack/project-config master: Add gerrit rename-project plugin to Zuul
opendevreviewClark Boylan proposed openstack/project-config master: Update ansible-lint to fix the rich dep
clarkbLets see if that works19:30
clarkbthen I can look at zuul-jobs next19:30
opendevreviewClark Boylan proposed opendev/base-jobs master: Update ansible-lint
opendevreviewClark Boylan proposed zuul/zuul-jobs master: Update ansible-lint pin to fix rich dep
opendevreviewClark Boylan proposed opendev/base-jobs master: Update ansible-lint
clarkbIf ^ doesn't work I'll do what zuul-jobs apparently did and just bump ansible ahead for linting19:57
opendevreviewJeremy Stanley proposed opendev/system-config master: Switch translate's IDP to OpenInfraID
clarkball of those ansible-lint update changes are green now. reviews appreciated20:25
fungiall three in that topic lgtm, thanks!20:28
clarkbfungi: is another. I'll update the topic on that one now20:29
clarkblast call on the infra meeting agenda. I'll get that setn out in the next 15 minutes or so20:53
fungiclarkb: i'm starting to suspect zanata doesn't store openids, and instead only checks the e-mail address they supply against its account table... do you happen to know?21:01
clarkbfungi: hrm I was pretty sure it did store them somewhere21:02
fungii'm digging through its db and not finding any obvious openid urls in any of the tables, but i'll keep looking21:02
clarkbI want to say when I debugged the user who had login problems I found the openid urls. But I don't recall what table or column they were in21:02
fungiaha, found it! HCredentials21:04
fungiokay, so i can do an etl on this similar to the one i did for the lp transition21:04
clarkband ya I do recall checking a lot of places :)21:04
opendevreviewMerged openstack/diskimage-builder master: Remove opensuse related funtests
opendevreviewJeremy Stanley proposed opendev/system-config master: Switch translate's IDP to OpenInfraID
opendevreviewJeremy Stanley proposed opendev/system-config master: Monitor the HTTPS cert for
opendevreviewJeremy Stanley proposed opendev/system-config master: Switch translate-dev's IDP to OpenInfraID
opendevreviewJeremy Stanley proposed opendev/system-config master: Switch refstack's IDP to OpenInfraID
fungii'll mark the last two in the series wip for the time being21:21
*** dviroel is now known as dviroel|out21:28
*** rlandy|ruck is now known as rlandy|dentist21:52
clarkbusing codesearch I think glean is the only project using opensuse-tumbleweed21:54
clarkboh and that job isn't even running on tumblweed21:55
clarkbits a job that builds a tumblweed image and boots it with nodepool21:55
opendevreviewClark Boylan proposed opendev/base-jobs master: Remove tumbleweed nodeset
clarkbI think ^ is our first step to removing tumbleweed21:57
opendevreviewClark Boylan proposed openstack/project-config master: Set tumbleweed min-ready to 0
opendevreviewClark Boylan proposed openstack/project-config master: Remove opensuse-tumbleweed from nodepool
opendevreviewClark Boylan proposed opendev/system-config master: Remove opensuse tumbleweed from our mirrors
clarkbok I think that is a complete set of changes to remove tumbleweed22:08
fungilgtm, though the base-jobs change will need the ansible-lint fix, which i've single-core approved just now23:04
clarkboh right. I got the order correct for the project-config changes but forgot about it on base-jobs23:05
opendevreviewMerged opendev/base-jobs master: Update ansible-lint
clarkb is green now. Thank you for the merge to fix that23:22

Generated by 2.17.3 by Marius Gedminas - find it at!