Wednesday, 2026-05-20

opendevreviewOpenStack Proposal Bot proposed openstack/openstack-ansible master: Imported Translations from Zanata  https://review.opendev.org/c/openstack/openstack-ansible/+/98889702:39
opendevreviewMerged openstack/openstack-ansible-plugins master: Enable CI jobs again  https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/98811604:23
kleiniI found the issue with nova-novncproxy on aarch64: The /etc/pki/nova-novncproxy/*.pem files are not deployed on the LXC nova container. Digging through the playbooks, what causes this.06:37
kleininova_pki_console_condition is false on the aarch64 node due to https://opendev.org/openstack/openstack-ansible-os_nova/src/branch/master/defaults/main.yml#L749 and nova_console_type is serialconsole for aarch64. But I configured now nova_console_proxy_types: [novnc, serialconsole, spice] and all three proxies are deployed.06:55
kleiniDeployed nova-novncproxy pki and it works.06:56
kleiniI would try to provide a fix for nova_pki_console_condition.07:10
noonedeadpunkkleini: hm, what if you just do `aarch64: novnc` in old good https://opendev.org/openstack/openstack-ansible/src/branch/master/inventory/group_vars/all/nova.yml#L27-L39 ?07:25
noonedeadpunkas I would expect this should result in `nova_console_type: novnc` for aarch computes07:25
noonedeadpunkwhile now it;s a `serialconsole`07:25
noonedeadpunkbecause `nova_console_type: "{{ nova_console_type_mapping[ansible_facts['architecture']] }}"`07:26
kleiniAnd I think, serialconsole is correct for aarch computes. I don't have one yet. I just have a aarch infra node.07:26
noonedeadpunkah07:26
noonedeadpunkaaaaah07:26
noonedeadpunkok-ok-, right, I got the issue then07:26
kleinisorry, to not have mentioned that07:26
kleinimy infra nodes are two x86 and one aarch64, to have one repo container on aarch64. The aarch64 compute is ordered. I though, I can just replace one old and slow x86 infra node with an aarch64 node. Now I learn, I seem to be the first one doing this. Just having an Ampere One CPU with 144 cores only for repo container was a bit waste of resources for me.07:32
noonedeadpunkwell... it's almost working though? :D07:35
noonedeadpunkwe failed to add an arm64 testing back in the days, as CI nodes were too slow to build wheels back in the days, so they were just timeouting07:36
noonedeadpunkI think we might be having better/faster ways today, so could re-try adding arm CI 07:36
noonedeadpunkand also wheels should be mainly available in pypi already07:37
kleiniIt is working really well without testing.07:38
opendevreviewMarcus Klein proposed openstack/openstack-ansible-os_nova master: Fix nova-novncproxy deployment missing PKI certificates  https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/98933708:22
jrosserkleini: we have a hybrid arm/x86 deployment here, your not on your own :)08:58
jrosserthere’s a bunch of notes somewhere in the docs and surprising complexity with getting vm to actually land on the right architecture host and not get emulated08:59
jrosserwe used a 32C ampere altra for the controller as that was reasonably priced09:00
kleini1U Gigabyte R163-P32 here, with Ampere One A144, 128G mem, N225P. price was still cheap due to components still in stock. price for compute is now three times compare to first quote.09:24
kleiniabout VMs on correct node: I read https://docs.openstack.org/openstack-ansible/latest/user/multiarch/multiarch.html#example-adding-aarch64-nodes-to-an-x86-64-deployment. okay, there might be still a problem with uploaded images.09:35
WireLosthey guys, I'm trying OSA AIO `master` on Ubuntu 24.04, it seems almost there! For reproducibility, I'm using the following script: https://paste.openstack.org/show/bj463wqX0nVRyFYDjTzh/ - but there is an error: https://paste.openstack.org/show/bKYTk4nFVr6qOLpKHm9D/ - Any idea?13:08
noonedeadpunkThis rings a bell13:11
mgariepymigth want to bump the number of threads a bit 13:11
WireLostHmm, found erros on Neutron server: https://pastebin.com/aUUaDzRh13:12
noonedeadpunkdamn, it;s distro setup, is it?:)13:12
noonedeadpunkI think you are still having 2025.2 packages13:13
noonedeadpunkAs uca has 2026.1 only in proposed updates13:13
noonedeadpunkit's not ins the main one13:13
WireLostIt is a fresh machine, no 2025.2 in there =P13:14
WireLostHmm...13:14
noonedeadpunkI was actually about to submit bug report last friday to ubuntu, when realised that they never published 2026.1 packages to main13:14
noonedeadpunkWireLost: just do dpkg -l 13:14
WireLostUCA 'gazpacho' doesn't have Gazpacho in it?! O_o13:14
noonedeadpunkand then check versions against https://releases.openstack.org/gazpacho/index.html13:14
noonedeadpunkand then enable proposed and check again :)13:15
WireLostHere is the `dpkg -l` https://paste.openstack.org/show/b6QAyom1iGU5wHHMtUsO/13:15
noonedeadpunkneutron-common                2:27.0.013:15
WireLostLet me try to enable proposed13:16
noonedeadpunkit;s 28.0.0 for gazpacho13:16
WireLostShieet13:16
WireLostlol13:16
WireLostok13:16
noonedeadpunk27 is flamingo https://releases.openstack.org/flamingo/index.html13:16
noonedeadpunkright ;)13:16
noonedeadpunkeventually the issue triggered by https://review.opendev.org/c/openstack/openstack-ansible-os_neutron/+/97280813:17
WireLostDaammnn, you're absolutely right! lol13:17
noonedeadpunkas flamingo does not know about `project_network_types` option - it was introduced for gazpacho13:17
WireLostEnabling proposed brings gazpacho13:17
noonedeadpunkFeel free to propose bug report to UCA :D13:18
WireLostIs is possible to set OSA/AIO to enable proposed globally?13:18
noonedeadpunkyou can override `openstack_hosts_package_repos` I guess13:18
WireLostokay13:18
noonedeadpunkhttps://opendev.org/openstack/openstack-ansible-openstack_hosts/src/branch/master/vars/ubuntu.yml#L81-L9213:18
noonedeadpunkyou can also propose patch to actually control updates vs proposed in default defenition13:19
WireLostBaby steps... :-D13:20
WireLostIn that `ubuntu-yml`, there is a comment: NOTE: 26.04 contains OpenStack 2026.1 out of the box and UCA for 2026.1 won't be created13:20
noonedeadpunklike introduce apt_repo_suites or smth13:20
WireLostIs this official?13:20
WireLostFrom Ubuntu?13:20
noonedeadpunkUm, they always do like that?13:20
noonedeadpunkyou don't need UCA, as it's in native repos13:20
WireLostBut it is native only for 26.0413:21
noonedeadpunkright13:21
WireLostAnd in 26.04 it is Ceph 20, and Ceph Ansible fails, I think13:21
noonedeadpunkah, sorry, comment was scoped specifically for 26.0413:21
WireLostah ok, I see13:21
noonedeadpunk24.04 should be having it13:21
WireLostgot it13:21
WireLostSo, it seems the only blocker would be to ensure ceph-ansible works in 26.04?13:22
noonedeadpunkI am not sure why it's not out of proposed though, but I was going to wait for a week or so before submitting bug report, as they usually take like couple of weeks to promote from proposed to updates13:22
noonedeadpunkbut maybe smth didn't work this type13:22
noonedeadpunkI have not checked ceph specifically on 26.04 tbh13:23
WireLostOk, np13:23
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Update Ceph release to squid  https://review.opendev.org/c/openstack/openstack-ansible/+/98939414:06
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-ceph_client master: Switch default ceph release to squid  https://review.opendev.org/c/openstack/openstack-ansible-ceph_client/+/98939714:11
WireLostBut Ceph on 26.04 is 'tentacle', not 'squid'14:11
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-plugins master: Allow to reset peer during it's re-install  https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/96611714:24
noonedeadpunkwould be nice to get reviews for https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/987880 and https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/98814314:25
noonedeadpunkas the first one is actually a regression in 2025.214:26
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Update Ceph release to squid  https://review.opendev.org/c/openstack/openstack-ansible/+/98939414:40
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Update Ceph release to tentacle  https://review.opendev.org/c/openstack/openstack-ansible/+/98939414:42
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-ceph_client master: Switch default ceph release to tentacle  https://review.opendev.org/c/openstack/openstack-ansible-ceph_client/+/98939714:43
noonedeadpunkWireLost: tentacle passed in my CI on master14:47
noonedeadpunkbut on 24.0414:47
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Disable global_id_reclaim for Ceph  https://review.opendev.org/c/openstack/openstack-ansible/+/98940814:52
WireLostCool! Using upstream repost, right, not UCA?14:54
noonedeadpunknot uca14:55
WireLostok14:55
WireLostI'm trying OSA AIO on 26.04, but this `[lxc_hosts : Cached image preparation script]` is failing `[ERROR]: Task failed: object of type 'dict' has no attribute 'resolute'` - is there a branch for 26.04 or best one would be `master`?14:56
noonedeadpunkthere's the patch https://review.opendev.org/c/openstack/openstack-ansible-lxc_hosts/+/98521614:58
noonedeadpunkI hope it should be merged soonish14:58
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Install rabbitmq from distro for Ubuntu 26.04  https://review.opendev.org/c/openstack/openstack-ansible/+/98522214:58
WireLostawesome, it worked!15:26
WireLostQuick feedback, I'm also installing mariadb from Ubuntu 26.04 `distro`, not `external` (it failed)15:53
noonedeadpunkso for external you'd need to bump minor version of mariadb15:54
WireLostIt's okay.... I rather use 'distro', too many supply chain risks these days, better stick with Ubuntu's sources only (my two cents) hehe15:56
WireLostI don't even like 'pip' either, 'apt' is pretty  xD15:56
noonedeadpunkisn't it ubuntu packaging which is failing lately ? :D15:56
WireLostNo, it failed to 'apt update' in the mariadb container15:57
WireLostthen I set it to 'distro' for it to work15:57
noonedeadpunkexactly, as trhere's no external repo for 11.8.6, but there is for 11.8.715:57
noonedeadpunkhttps://review.opendev.org/c/openstack/openstack-ansible-galera_server/+/98657315:57
WireLostahh ok15:57
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible-galera_server master: Upgrade MariaDB to 11.8.7  https://review.opendev.org/c/openstack/openstack-ansible-galera_server/+/98657315:57
opendevreviewDmitriy Rabotyagov proposed openstack/openstack-ansible master: Install rabbitmq from distro for Ubuntu 26.04  https://review.opendev.org/c/openstack/openstack-ansible/+/98522215:58
noonedeadpunkbut I was thinking about the glance and uca15:58
WireLosttrue, that is a gross bug hihi16:01
WireLostnoonedeadpunk, I think I've found another small issue in Glance 2026.1 ('distro') on 26.04. The line `wsgi-file = glance-wsgi-api` in the file `/etc/uwsgi/glance-api.ini` should be changed to `module = glance.wsgi.api:application` for the service to start18:31
noonedeadpunkWireLost: yup: https://review.opendev.org/c/openstack/openstack-ansible-os_glance/+/98863718:32
WireLostCool! lol18:33
opendevreviewMerged openstack/openstack-ansible-os_glance master: Revert "Revert Swift retries value once fix is merged"  https://review.opendev.org/c/openstack/openstack-ansible-os_glance/+/98906519:25
opendevreviewIvan Anfimov proposed openstack/openstack-ansible-galera_server master: Remove galera_disable_privatedevices  https://review.opendev.org/c/openstack/openstack-ansible-galera_server/+/98431219:26
opendevreviewIvan Anfimov proposed openstack/openstack-ansible-galera_server master: Remove galera_disable_privatedevices  https://review.opendev.org/c/openstack/openstack-ansible-galera_server/+/98431222:08

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!