| opendevreview | Andrew Bonney proposed openstack/openstack-ansible-rabbitmq_server stable/victoria: Bump rabbitmq version https://review.opendev.org/c/openstack/openstack-ansible-rabbitmq_server/+/801628 | 07:45 |
|---|---|---|
| *** rpittau|afk is now known as rpittau | 07:49 | |
| opendevreview | Andrew Bonney proposed openstack/ansible-role-python_venv_build master: Add distro/arch to requirements file path https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/801738 | 08:28 |
| masterpe[m] | Is it possible to speedup OSA with Mitogen or will it break OSA? | 08:53 |
| jrosser | masterpe[m]: i have a very very long outstanding patch to add mitogen support | 09:14 |
| jrosser | https://review.opendev.org/c/openstack/openstack-ansible/+/591236 | 09:14 |
| jrosser | but there are some outstanding bugs in mitogen which break with some of the delegated tasks | 09:15 |
| jrosser | i've raised issues on mitogen but that needs development effort to fix | 09:15 |
| noonedeadpunk | Also iirc we haven't see any significant performance bump with mitogen as well? | 09:58 |
| noonedeadpunk | *seen | 09:59 |
| noonedeadpunk | s/bump/improvement/ | 09:59 |
| opendevreview | Georgina Shippey proposed openstack/openstack-ansible-os_keystone master: Fix oidc scope misspelling in newer releases https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/801604 | 10:05 |
| jrosser | noonedeadpunk: i think that performance improvements from mitogen would be seen on real deployments rather than in CI, as it helps with parallelism and latency | 10:39 |
| jrosser | i suspect that the biggest improvement for AIO/CI would be the fastest possible storage backend | 10:40 |
| noonedeadpunk | yeah, agree. | 10:40 |
| noonedeadpunk | but I thin kwe're talking about like 10% anyway? | 10:40 |
| jrosser | we get at least 2x on our non OSA stuff here | 10:40 |
| noonedeadpunk | oh | 10:40 |
| jrosser | and if i used a mitogen_via host in the same rack as the target rather than via WAN i suspect we'd get more | 10:41 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: [WIP] Use the Mitogen connection method in OSA https://review.opendev.org/c/openstack/openstack-ansible/+/591236 | 10:41 |
| noonedeadpunk | let me rabes just out of interest | 10:42 |
| noonedeadpunk | *rebase | 10:42 |
| jrosser | theres a delegate bug in mitogen | 10:42 |
| noonedeadpunk | Is mitogen really maintained nowadays? | 10:43 |
| noonedeadpunk | like it's half a year from latest commit... | 10:43 |
| jrosser | oh and also the way it detects if python code is a module or not is also not liking the openstack collections https://github.com/bbc/mitogen/commit/cf6b6af250264eb820e1b53d75b3a8d311b15305 | 10:43 |
| jrosser | openstack collection has a base class rather than include the ansible module stuff directly | 10:44 |
| jrosser | heres the delegation issue https://github.com/mitogen-hq/mitogen/issues/584 | 10:46 |
| noonedeadpunk | Yeah, I bet it's not ready for ansible 4.0 as well... | 10:47 |
| noonedeadpunk | things are moving super fast now there... | 10:47 |
| opendevreview | Andrew Bonney proposed openstack/ansible-role-python_venv_build master: Add distro/arch to requirements file path https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/801738 | 10:54 |
| opendevreview | Merged openstack/openstack-ansible-os_keystone stable/wallaby: Improvements to federation packaging https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/801586 | 12:31 |
| jrosser | huh well the mitogen patch didnt totally fail :) | 12:32 |
| jrosser | and most interestingly it passes on openstack-ansible-deploy-aio_lxc-debian-bullseye, and LXC job | 12:33 |
| jrosser | *an LXC job | 12:33 |
| noonedeadpunk | it's weird really | 12:34 |
| noonedeadpunk | btw, focal fails now on `ansible -m setup -a 'gather_subset=!all,min' all` | 12:35 |
| noonedeadpunk | after setup_hosts | 12:36 |
| jrosser | yes https://zuul.opendev.org/t/openstack/build/62d0cb3ef8414128b5dc3e2f38efee57/log/job-output.txt#9354-9371 | 12:36 |
| noonedeadpunk | right before setup-infrastructure | 12:36 |
| jrosser | thats totally odd | 12:36 |
| noonedeadpunk | feels as a result of mitogen being unmaintained.... but like it's not failing on delegates right now! | 12:37 |
| jrosser | i guess part of the problem is that for 99% of things it kind of "just works" | 12:39 |
| jrosser | we don't have issues with it outside OSA really | 12:39 |
| jrosser | and we are unlucky in OSA with the connection plugin and funky delegation which is whats made it always fail | 12:40 |
| opendevreview | Andrew Bonney proposed openstack/openstack-ansible-os_keystone master: Fix shibboleth compatibility for ubuntu 18.04 https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/801849 | 13:07 |
| jrosser | andrewbonney: ^ this needs a backport I think | 13:12 |
| andrewbonney | Yes, I've noted. I'll need to do some V backporting too but thought it best to address the issue first | 13:13 |
| jrosser | yeah this too maybe? https://review.opendev.org/c/openstack/openstack-ansible-os_keystone/+/801586 | 13:13 |
| jrosser | as i think that the issues there might be the same for co-incident mod-auth-oidc and shib | 13:13 |
| andrewbonney | Yes | 13:14 |
| andrewbonney | Was SSO fully supported in U? Doesn't look like the Focal support for this goes back that far | 13:15 |
| andrewbonney | Ah, actually it may do but using the transitional package | 13:15 |
| opendevreview | Andrew Bonney proposed openstack/ansible-role-python_venv_build master: Add distro/arch to requirements file path https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/801738 | 14:36 |
| opendevreview | Andrew Bonney proposed openstack/openstack-ansible-rabbitmq_server stable/victoria: Bump rabbitmq version https://review.opendev.org/c/openstack/openstack-ansible-rabbitmq_server/+/801628 | 14:37 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_barbican stable/victoria: Allow to symlink barbican_user_libraries https://review.opendev.org/c/openstack/openstack-ansible-os_barbican/+/801878 | 15:19 |
| *** rpittau is now known as rpittau|afk | 16:02 | |
| noonedeadpunk | jrosser: regarding your comment on https://review.opendev.org/c/openstack/openstack-ansible/+/799800 (just noticed it). I'm not sure actually. I can imagine usecase where ppl want to continue using panko for some reason. If we want to fully drop it - we probably want also to clean up inventory and drop containers? | 16:16 |
| noonedeadpunk | (and drop db haha) | 16:17 |
| jrosser | its about haproxy config | 16:17 |
| jrosser | we need to make it be 'absent' for one cycle then we can delete the backend | 16:17 |
| jrosser | as it uses assemble the backend will still be present but then down forever otherwise | 16:18 |
| noonedeadpunk | I can imagine that if service is dropped from inventory - it will be removed from haproxy as well by default? | 16:18 |
| jrosser | https://github.com/openstack/openstack-ansible-haproxy_server/blob/abafc878c29a206414fc33b79883e926db8ee654/tasks/haproxy_service_config.yml#L33-L47 | 16:19 |
| jrosser | no | 16:19 |
| jrosser | haxproxy service config generates lots of files with_items then assemble module puts them all together for the whole config | 16:19 |
| jrosser | so they persist on the disk | 16:19 |
| noonedeadpunk | ah, indeed | 16:19 |
| noonedeadpunk | Ok, well, I mean maybe we want just to leave haproxy config as is then? | 16:20 |
| noonedeadpunk | and not mark as absent at all? | 16:20 |
| noonedeadpunk | but yeah, that's weird as welll | 16:20 |
| jrosser | seems like the role allows two ways to do this | 16:22 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Deprecate os-panko role https://review.opendev.org/c/openstack/openstack-ansible/+/799800 | 16:22 |
| jrosser | define a service with empty backup/backends defined | 16:22 |
| jrosser | or set service.enabled / service.state both false | 16:23 |
| jrosser | enabled=false state=absent, rather | 16:23 |
| noonedeadpunk | well, since there're "OR" I think just state=absent should be enough? | 16:49 |
| noonedeadpunk | or enabled=false | 16:50 |
| noonedeadpunk | empty backup/backends defined is kind of weird. As if you don't have haproxy_backup_nodes defined but haproxy_backend_nodes is empty list - that would still be false | 16:51 |
| noonedeadpunk | I guess I'd rather just dropped that condition then try to fix it. as haproxy_service_enabled/state should be enough I guess | 16:51 |
| jrosser | is it not (A or B) | 16:52 |
| jrosser | oh no it isnt | 16:53 |
| jrosser | (not A) or B | 16:53 |
| jrosser | thats kind of poor formatting on the conditionals, makes my head hurt | 16:54 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-haproxy_server master: Fix service removal condition https://review.opendev.org/c/openstack/openstack-ansible-haproxy_server/+/801910 | 16:59 |
| noonedeadpunk | ^ | 16:59 |
| noonedeadpunk | I dont think we need it that complicated | 17:00 |
| dmsimard | noonedeadpunk: btw in case you miss it, I just posted a WIP for tracking task delegation https://github.com/ansible-community/ara/issues/282, it'll be in the next release | 17:02 |
| noonedeadpunk | ٩(^ᴗ^)۶ | 17:04 |
| noonedeadpunk | that looks nice! | 17:04 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Deprecate os-panko role https://review.opendev.org/c/openstack/openstack-ansible/+/799800 | 17:10 |
| spatel | noonedeadpunk if i want to build aio-ovn in my lab then what are the easy step? | 19:59 |
| spatel | i did last time with applying bunch of patches.. | 19:59 |
| spatel | is there easy way or i should use same way | 19:59 |
| *** sshnaidm is now known as sshnaidm|afk | 20:10 | |
| spatel | just trying with export SCENARIO='aio_ovn_lxc' | 20:12 |
| spatel | lets see if it work | 20:12 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!