Thursday, 2024-11-21

opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581300:54
*** __ministry is now known as Guest38107:10
opendevreviewArtem Goncharov proposed openstack/pbr master: add python 3.12 support  https://review.opendev.org/c/openstack/pbr/+/92421611:04
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581318:04
clarkbfungi: ^ so I thought things weren't determinstic with grub kernel updates on arm for whatever reasno then looking at it with fresh eyes today I realized I updated grub post reboot not pre reboot (the shame)18:04
fungihah18:05
clarkbfungi: anyway I suspect that may actually be workable as a solution to our problem now18:05
fungihappens to all of us at some point18:05
clarkbthis is the ozj side. Going to check system-config which is what was actually holding up noble packages next18:05
clarkband then separately we should look into why git cache updates on nb04 are so flaky if that is still happening after I restarted nsd on ns0418:05
clarkbok the system-config chagne is hitting the issue whre upstream packages don't work for arm64 due to the invalid instruction in the kernel18:06
clarkbI think I may just mark those jobs non voting18:06
fungiis it still using older openafs then?18:07
clarkbfungi: yes it is the noble job that is failing18:07
clarkbso chicken and egg there18:07
fungiyeah, if only we could use our ppa build there18:08
clarkbbookworm also failed though which I thought was fixed18:08
fungiup to 350 uncaught bounce notifications to openstack-discuss-owner, i'm going to need to go through and manually disable these subscriptions (after i collate them all)18:09
clarkbI wonder if we unfixed the fix with the new updates18:09
clarkbthat would be annoying if so18:09
fungibookworm is not fixed in debian yet18:09
fungihttps://packages.debian.org/search?keywords=openafs-client18:09
clarkbthat rules out the security bugfixes regressing I guess. STill I thought upstream had fixed the cannot load the kernel module problem on bookworm18:10
clarkband I was pretty sure I tested that so maybe some other update in the interim broke us?18:10
fungibookworm is on 1.8.9, without any backported patches18:10
clarkbbut it had the patch for the modprobe error right? that was the whole bug I filed with debian and it got updated. I'm rpetty sure I checked this after the bug was closed18:12
fungiit was also removed from testing (trixie) in march, so needs to get reintroduced before it can migrate, and probably won't see any backports until that happens unless the debian security team intervenes: https://tracker.debian.org/news/1512372/openafs-removed-from-testing/18:13
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581318:13
fungireferences in the autoremoval are to https://bugs.debian.org/1060896 and https://bugs.debian.org/106282918:14
clarkbhttps://bugs.debian.org/cgi-bin/bugreport.cgi?bug=106978118:15
clarkbmaybe that bug is just saying the package updated in unstable?18:15
fungiyeah18:16
fungithe two bugs resulting in autoremoval are also fixed, i guess reintroduction to testing has just not been requested yet18:16
fungirelated, i got a reply from sarnold in #ubuntu-security that was essentially "patches welcome" (if we have time to backport security patches and initiate the sru process to each release)18:17
clarkboh wait those jobs are already marked non voting from when we first ran into this so the only failuer is still in getting the centos 9 arm64 job happy18:18
clarkbif we do that we should also backport the arm fix18:18
clarkbthat patch is really straightforward iirc18:19
clarkbok I think I have evidence now that we are install the 529 kernel and headers, doing a grub update, then rebooting and booting back onto the 527 kernel we were running before18:47
clarkbbut sometimes this seems to work and we boot onto 529? This is very weird18:48
clarkbseparately when you install the new kernel headers package for the kernel you will reboot onto dnf uninstalls the headers for the kernel you will boot onto18:48
clarkbI think this is the fundamental underlying issue and is why debuntu are fine they won't remove old kernels until you autoremove or similar to clean that stuff up and the dkms packages will build against all kernels on the system18:49
fungiwhich kernel we boot *could* be provider specific, depending on whether the boot routine reads the updated configuration18:54
fungilike grub config update vs efi config update18:55
clarkboh ya the arm machines boot efi18:59
clarkbso maybe there is an additional step we need? however I think the log says it updates all that stuff18:59
clarkbhttps://zuul.opendev.org/t/openstack/build/2abe1eabbbfd4fa3b5f3caafcc7c0e88/console19:00
fungiat least on debian derivatives, it's up to how grub is configured whether update-grub also updates efi configs19:00
fungibut yeah, it's also possible it thinks it's updating the config but then the config it updated isn't the one the boot process is ultimately reading (e.g. more than one copy in different partitions/locations)19:01
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581319:02
clarkbthis change is slowly becoming more debug info that actual effort :)19:02
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581319:12
clarkbI should probably hold a node at this point, but its a good exercise in debugging from afar (something we often ask others to do)19:14
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581319:19
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Don't explicitly install kernel-devel when building openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93598421:46
clarkbfungi: ^ I took a long lunch break and it occurred to me we may be able to get away with something as simple as that. We'll see21:47
clarkbif that works then I'm going to say stepping away from the problem for a bit was the genius move. If it doesn't work oh well it was worth a shot anyway21:47
clarkbnope that doesn't work because rpmbuild can see there is a newer version and it wants that newer version ugh21:51
clarkbI think that must come from the dnf builddep -y $SRPM step21:53
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581321:59
clarkbthe problem wtih log collection (or one of them anyway) is taht fetch-output and stage-output have different default expectations for where we stage the content on the test nodes22:00
clarkbBuildRequires: kernel-devel <- that is what the rpm spec file says so I guess its rpm/dnf that are seeing there is no other upstream version and demanding the latest and greatest22:04
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581322:57
opendevreviewClark Boylan proposed openstack/openstack-zuul-jobs master: Update system packages and reboot when building centos openafs  https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/93581323:55

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!