Thursday, 2025-10-02

opendevreviewOpenStack Proposal Bot proposed openstack/project-config master: Normalize projects.yaml  https://review.opendev.org/c/openstack/project-config/+/96255702:13
mnasiadkaMorning, finally rocky10 is passing in https://review.opendev.org/c/opendev/zuul-providers/+/962564 (hopefully it’s not just that lucky run)05:12
fungicongrats mnasiadka! i know that was a lot of work12:55
mnasiadkaWell, I’ve done majority of zuul-providers work, added Rocky 10, so now Alma was not that complicated ;)12:59
mnasiadkaNow I can go post a devstack job when that lands12:59
opendevreviewMerged openstack/diskimage-builder master: Remove testing for f37  https://review.opendev.org/c/openstack/diskimage-builder/+/95295413:36
opendevreviewPiotr Parczewski proposed openstack/diskimage-builder master: Fix GRUB arm64 package names  https://review.opendev.org/c/openstack/diskimage-builder/+/96279313:37
opendevreviewPiotr Parczewski proposed openstack/diskimage-builder master: Fix GRUB arm64 package names  https://review.opendev.org/c/openstack/diskimage-builder/+/96279313:39
opendevreviewJan Gutter proposed zuul/zuul-jobs master: Update ensure-helm role to add more functionality  https://review.opendev.org/c/zuul/zuul-jobs/+/96279413:44
opendevreviewJan Gutter proposed zuul/zuul-jobs master: Update ensure-helm role to add more functionality  https://review.opendev.org/c/zuul/zuul-jobs/+/96279413:46
corvusclarkb: fungi if it helps (probably not in this case, but just fyi) zuul uses a sphinx config setting to generate canonical link urls in the html14:00
corvushttps://opendev.org/zuul/zuul/src/branch/master/doc/source/conf.py#L73-L7414:00
fungithanks, that's an excellent example14:01
fungibut yeah, not directly portable to the gitea backends vs proxied name discrepancy. i think we'll just set it in the apache vhost config14:01
corvusthat was added mostly to tell search engines to ignore all the old versions of docs and focus on the latest14:01
corvusfor gitea, i think the header is the way to go14:02
corvus(because gitea is in control of the canonical location there, it's generating the content (and apache is the next best thing))14:02
opendevreviewJan Gutter proposed zuul/zuul-jobs master: Update ensure-helm role to add more functionality  https://review.opendev.org/c/zuul/zuul-jobs/+/96279414:21
opendevreviewClark Boylan proposed opendev/system-config master: Set canonical Link paths for gitea resources  https://review.opendev.org/c/opendev/system-config/+/96282614:59
clarkbinfra-root ^ that is almost certainly incomplete and buggy but I figured we can start there as our testinfra test cases give us some decent feedback15:00
clarkbit isn't clear to me for example if the query string should be included (I think so?) and if so does teh query string variable include the ? prefix like the request_uri includes the root /15:01
fungipopping out for lunch, should be back around 16:3015:19
clarkbrax-dfw continues to look good15:25
clarkbone concern I have is that s we use it more it may slowly snowball itno sadness but there is no evidence of that yet15:26
clarkbdoes the opendev zuul installation no longer have streaming logs that work for anyone else? cc corvus 15:30
clarkbI'm using an up to date firefox installation and there are characters emitted but not coherently in anything that resembles a stream. I know there was an update to the ansi stuff recently maybe that is related?15:31
clarkbthe job-output.txt file looks fine so it must be in the processing on the web side?15:32
opendevreviewClark Boylan proposed opendev/system-config master: Set canonical Link paths for gitea resources  https://review.opendev.org/c/opendev/system-config/+/96282615:33
melwittdoes anyone know if there is a way to avoid using a hardcoded node name from hostvars to get a subnode name? I tried this but it seems to not work https://review.opendev.org/c/openstack/nova/+/957477/16/roles/run-evacuate-hook/tasks/main.yaml my problem is that I am trying to use a nodeset that has not named the subnode 'compute1'15:47
clarkbmelwitt: if you look in the resulting job's zuul-info dir that should include the ansible inventory used to run the job playbooks15:49
clarkbmelwitt: I would start with that datastructure and see if any lookup path from there looks workable15:49
melwittohh ok15:49
melwittthank you15:49
clarkbbut generally I suspect the ansibly solution to problems like this is to rely on groups more than hostnames15:50
clarkbso you might have a compute group and then expect that ndoes that fill that role are in that group15:50
melwittyeah that makes sense. that's basically what I want to do15:50
clarkbmelwitt: https://zuul.opendev.org/t/openstack/build/065d156f3a0c4db7a39d7021072132d3/log/zuul-info/inventory.yaml it does look liek you have subnode and compute in there already so you can in theory lookup from there15:52
clarkbyou may need to convert the dict keys into a list to 0 index them?15:52
melwittyeah I'm pretty ansible ignorant so I didn't even know where to look heh15:53
clarkbmelwitt: https://www.reddit.com/r/ansible/comments/ozopzi/list_hosts_in_group_and_filter/15:55
clarkbI think the select(match) filter is replacing the indexing there. So ya something along those lines or using a filter to get keys as a list then indexing them that way15:56
clarkbmelwitt: I find this page is often helpful https://docs.ansible.com/ansible/latest/playbook_guide/playbooks_filters.html15:56
melwittthanks this all helps a lot :)15:56
clarkbI do wish that page more prominently linked to https://docs.ansible.com/ansible/latest/playbook_guide/playbooks_filters.html as well15:57
corvusclarkb: https://zuul.opendev.org/t/zuul/stream/9ea0e70a8fec43228404c209068dcfc5?logfile=console.log looks like a legit stream to me15:57
clarkbbasically you can use any of those built in filters or the ansible provided filters15:57
clarkbcorvus: huh it doesn't for me. Let me try another browser15:57
clarkbcorvus: it works in chrome and chromium but not firefox15:58
clarkblet me try firefox without any extensions15:58
corvusi will restart my firefox and try another15:58
clarkbfirefox incognito mode also does not work so I think it is specific to firefox15:58
corvusff 143.0.1 and ZOMG i can not copy and paste the text "canonical-002 - 1.0"15:59
corvusi can copy/paste the upstream version number, but not the downstream one.  nice.16:00
clarkbI'm on 143.0.1 but there is new update available so I'll update and retest16:00
corvusi have another one on 143.0.3 (same downstream) and it works too16:01
clarkbcorvus: ok 143.0.3 works for me now16:01
clarkbI wonder if the pending update installation broke some shared library stuff and preventing rendering16:01
clarkbits weird because everything else seemed to work except for when I tried to stream a log out of that gitea change16:02
corvusoh maybe.  i've seen things get weird when that happens.16:02
clarkbsorry for the noise, I suspect this was an issue with the pending update impacting the old installation16:02
opendevreviewMerged opendev/zuul-providers master: Add Alma Linux 10 image build jobs  https://review.opendev.org/c/opendev/zuul-providers/+/96256416:06
mnasiadkaYay16:07
opendevreviewClark Boylan proposed opendev/system-config master: Set canonical Link paths for gitea resources  https://review.opendev.org/c/opendev/system-config/+/96282616:11
opendevreviewClark Boylan proposed opendev/system-config master: Set canonical Link paths for gitea resources  https://review.opendev.org/c/opendev/system-config/+/96282617:31
opendevreviewClark Boylan proposed opendev/system-config master: Set canonical Link paths for gitea resources  https://review.opendev.org/c/opendev/system-config/+/96282618:28
clarkbI missed that it looks like you need the e suffix on the variable substitution stuff there18:29
clarkbthough still not sure those variables are considered environment variables...18:29
clarkbok that last run confirms that those variables are null in that context. I found https://andreas.scherbaum.la/post/2024-12-26_add-canonical-headers-for-files-in-apache2/ which uses mod rewrite to define variables with this sort of information for every request19:48
clarkbdo we think ^ that is reasonable cc fungi 19:49
corvusclarkb: i give up, what's with the `e`s near query_string?20:06
clarkbcorvus: that is apparently the syntax you use in the mod_headers Header directive values20:10
clarkbcorvus: you can use %{foo}e to lookup env vars or %{bar}s to look up special ssl vars20:10
corvusoh i see it in the docs there now.  i was looking on the general substitution page20:12
clarkbya I tripped over this quite a bit too. Its iteresting that tehy use their own syntax20:12
corvusclarkb: looks like the test is failing because it's getting Link: <https://testgitea.opendev.org(null)(null)>; rel="canonical"20:14
corvusoh and that's your question: should we use mod_rewrite to fix that20:14
corvusi think that's reasonable20:15
clarkbya I guess I wasn't sure if setting env var values for every request is something we might expect to have a big impact on server operations20:15
clarkbwe're essentially shuffling the data from mod rewrite vars to generic env vars that mod headers can read20:16
clarkbI'll work on making that update shortly20:18
clarkbI'm finally getting over my writers block and have some content going into a reveal.js html file finally20:18
corvusi think that's fine, mod_rewrite is made for that20:21
opendevreviewClark Boylan proposed opendev/system-config master: Set canonical Link paths for gitea resources  https://review.opendev.org/c/opendev/system-config/+/96282621:25
clarkb962826 passes now. Our testinfra tests do test some git operations in addition to normal http requests for browser traffic so in theory this isn't breaking anything22:22
clarkbbut probably still work some careful consideration of what the potential impacts are here and whether or not we need more test cases22:22

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