| *** ysandeep|out is now known as ysandeep | 02:08 | |
| *** ysandeep is now known as ysandeep|afk | 03:40 | |
| *** ysandeep|afk is now known as ysandeep | 05:20 | |
| *** frickler_pto is now known as frickler | 05:35 | |
| *** akahat_ is now known as akahat | 05:39 | |
| *** ysandeep is now known as ysandeep|afk | 06:51 | |
| *** ysandeep|afk is now known as ysandeep | 07:18 | |
| jrosser_ | morning | 07:49 |
|---|---|---|
| *** ysandeep is now known as ysandeep|lunch | 07:50 | |
| noonedeadpunk | hey o/ | 07:55 |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible master: Remove neutron agents from ironic env.d file https://review.opendev.org/c/openstack/openstack-ansible/+/851699 | 07:57 |
| noonedeadpunk | I think we need to merge xena stuff and merge bump to unblock yoga https://review.opendev.org/q/parentproject:openstack/openstack-ansible+branch:%255Estable/xena+status:open | 08:37 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_manila stable/xena: Create backends when running against manila_share https://review.opendev.org/c/openstack/openstack-ansible-os_manila/+/847993 | 09:02 |
| *** ysandeep|lunch is now known as ysandeep | 09:40 | |
| opendevreview | Merged openstack/openstack-ansible stable/xena: Set minimum and maximum microversions for manila api https://review.opendev.org/c/openstack/openstack-ansible/+/850478 | 11:18 |
| *** dviroel_ is now known as dviroel | 11:45 | |
| *** ysandeep is now known as ysandeep|afk | 12:23 | |
| *** anskiy1 is now known as anskiy | 12:23 | |
| * jrosser_ stares hard at ironic role, kind of big mess here | 12:57 | |
| mrf | Hi! Can i use other node diferent than a controller for deploy openstack with ansible? | 13:28 |
| mrf | like a deployment node? | 13:28 |
| jrosser_ | mrf: absolutely yes | 13:28 |
| jrosser_ | i would recommend this too for a production deployment | 13:29 |
| noonedeadpunk | jrosser_: btw, do you recall stuff we did to pass requiest-id from haproxy and leave it the same to be able to track down request? | 13:29 |
| noonedeadpunk | I just hardly recall that part | 13:29 |
| jrosser_ | hmm | 13:29 |
| mrf | thanks @jrosser_ | 13:29 |
| jrosser_ | i can't remember doing that for haproxy - definatley in the ELK stack stuff though | 13:29 |
| noonedeadpunk | I either dreamed about it or clean forgot... | 13:30 |
| noonedeadpunk | As trying to map requests betweeen incoming to haproxy and to uwsgi | 13:30 |
| jrosser_ | mrf: the only thing you need for the deploy host is to be able to ssh to the targets, you can have a bastion or WAN link or whatever else in the way, and you also don't have to use br-mgmt for it, you can have a separate deplyoment/OOB network if you want | 13:31 |
| jrosser_ | noonedeadpunk: theres some stuff here about logging headers https://www.haproxy.com/blog/haproxy-log-customization/ | 13:33 |
| jrosser_ | i don't recall seeing anything like that in standard OSA | 13:33 |
| *** ysandeep|afk is now known as ysandeep | 13:33 | |
| noonedeadpunk | huh, where I have taken this from.... | 13:39 |
| noonedeadpunk | maybe some PR we never merged... | 13:39 |
| mrf | note: when you use a controller as deployment node the etc/hosts file got redundant entries (doubled) | 13:44 |
| noonedeadpunk | well, then basically you should set `openstack_host_manage_deploy_hosts_file: false`... But we likely can be smarter there | 13:45 |
| mrf | anyway im going to move the deployment to a dedicated host to prevent that , just info for you that duplicated entries happens | 13:46 |
| jrosser_ | noonedeadpunk: can you try something like `ansible -m debug -a var=group_names utility_all` for me? i'm surprised at what what prints in my lab env | 14:04 |
| jrosser_ | like this for one of them https://paste.opendev.org/show/brjquwupiONwR7GKnonJ/ | 14:04 |
| noonedeadpunk | sure, sec | 14:04 |
| jrosser_ | it's like it's printed the groups to do with everything on the LXC host, rather than just the `utility_all` grou | 14:06 |
| noonedeadpunk | huh, that's interesting indeed | 14:09 |
| jrosser_ | we have conditionals on group_names all over the place | 14:09 |
| noonedeadpunk | so I have that indeed https://paste.openstack.org/show/buQHhZoqOsUn82aiEKOz/ | 14:09 |
| noonedeadpunk | actually, I had spotted such weird behaviour when mentioning belongs_to several times on container_skel level | 14:10 |
| noonedeadpunk | like here https://opendev.org/openstack/openstack-ansible/src/branch/master/inventory/env.d/utility.yml#L25-L26 | 14:11 |
| noonedeadpunk | so basically, utility is added to all shared-infra_containers | 14:11 |
| noonedeadpunk | eventually, we kind of asked for this? | 14:11 |
| jrosser_ | taking the example of `operator_containers`, it looks like there are conflicting views from | 14:15 |
| jrosser_ | `ansible-inventory -i inventory/dynamic_inventory.py --list` and `ansible-inventory -i inventory/dynamic_inventory.py` | 14:15 |
| jrosser_ | ^ `--graph` | 14:15 |
| jrosser_ | --list shows something like i expect, --graph shows something that looks more like `group_names` | 14:15 |
| noonedeadpunk | I wonder if --list just do not iterate over groups where another groups are mentioned as children | 14:16 |
| noonedeadpunk | eventually I'm messing with inventory quite a bit to get AZs working and there was even idea to re-think how are we doing things, and isntead of using "dynamic inventory" just generate static one (as yaml?) and pass it instead... | 14:18 |
| jrosser_ | reason i started looking at this is i am messing with ironic | 14:18 |
| jrosser_ | and all of the 3 different types of ironic container are members of all the grous | 14:18 |
| jrosser_ | so the conditional logic in the role is just completely hopeless | 14:18 |
| jrosser_ | and then i was very surprised to see how many things were actually in group_names | 14:19 |
| jrosser_ | of course the ironic env.d could be just wrong, but then i looked at utility to sanity check | 14:19 |
| noonedeadpunk | well, it's highly likely that utility also wrong... | 14:23 |
| noonedeadpunk | Or we did some mistake in patching dynamic_inventory, or it was jsut like that for quite some time | 14:23 |
| opendevreview | Merged openstack/openstack-ansible stable/xena: Let git choose the branch when cloning a repo https://review.opendev.org/c/openstack/openstack-ansible/+/851552 | 14:26 |
| jrosser_ | well, dynamic inventory scripts are basically deprecated now | 14:43 |
| jrosser_ | i might investigate turning ours into a plugin | 14:44 |
| *** frenzyfriday|rover is now known as frenzyfriday|rover|lunch | 15:01 | |
| *** melwitt_ is now known as melwitt | 15:11 | |
| jrosser_ | from openstack_inventory.json this doesnt look right https://paste.opendev.org/show/b37rvrjn3An3UTlEz2rx/ | 15:23 |
| noonedeadpunk | basically, we just need inventory-manage.py, that would generate changes to openstack_inventory. And if we can switch json to yml.... | 15:29 |
| noonedeadpunk | but worth discussion indeed | 15:30 |
| noonedeadpunk | maybe jsut turn to the plugin | 15:30 |
| jrosser_ | yeah i am seeing super wierd things here in an AIO | 15:30 |
| jrosser_ | i deleted compute-infra_containers { ..... } section from openstack_inventory.json completely | 15:31 |
| jrosser_ | then re-ran the dynamic inventory script, and it is put back like this https://paste.opendev.org/show/bPKfAZuGZSvsBNlE7HFZ/ | 15:32 |
| jrosser_ | with children: being empty rather than `aio1-host_containers` | 15:32 |
| noonedeadpunk | whaaaat | 15:34 |
| noonedeadpunk | brrrr | 15:34 |
| jrosser_ | i know :( | 15:34 |
| noonedeadpunk | tbh this is last thing I wanted to touch, but it seems indeed we need to do smth with it | 15:39 |
| jrosser_ | it is wierd - some of these things come back with children: populated | 15:43 |
| jrosser_ | this is quite a good debugging angle i think, remove these sections and see why generate.py wants to put them back | 15:44 |
| noonedeadpunk | Well, for me it indeed was belongs_to in container_skel. So when I placed there more general group, hosts were added to all subgroups of this groups as well | 15:45 |
| noonedeadpunk | (or smth like that) | 15:45 |
| noonedeadpunk | and I ended up on having az1_cinder_volume as a part of whole az1 hosts plus their subgroups (like az1_nova_compute, az1_neutron_l3_agent, etc) | 15:47 |
| noonedeadpunk | (this doesn't make sense when I write it this way) | 15:47 |
| *** frenzyfriday|rover|lunch is now known as frenzyfriday|rover | 15:51 | |
| *** dviroel is now known as dviroel|lunch | 16:22 | |
| *** ysandeep is now known as ysandeep|out | 16:26 | |
| jrosser_ | this looks suspicious https://github.com/openstack/openstack-ansible/blob/master/osa_toolkit/generate.py#L217-L219 | 16:45 |
| jrosser_ | and perhaps duplicate of https://github.com/openstack/openstack-ansible/blob/master/osa_toolkit/generate.py#L397-L399 | 16:48 |
| jrosser_ | well not duplicate, just wrong | 16:48 |
| *** dviroel|lunch is now known as dviroel | 17:24 | |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible master: Do not create {hostname}-host_containers group as child of other containers https://review.opendev.org/c/openstack/openstack-ansible/+/851764 | 17:41 |
| jrosser_ | noonedeadpunk: maybe something like that ^ ...... i don't see a good case for that code to be there | 17:42 |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible master: Do not create {hostname}-host_containers group as child of other groups https://review.opendev.org/c/openstack/openstack-ansible/+/851764 | 17:43 |
| noonedeadpunk | jrosser_: won't it drop possibility to limit actions only against containers on provided controller? | 17:45 |
| jrosser_ | there is also a top level group aio1-host_containers | 17:45 |
| jrosser_ | which is pretty unclear atm how that is created in the inventory | 17:45 |
| jrosser_ | well thats not what i mean | 17:47 |
| jrosser_ | it's created here https://github.com/openstack/openstack-ansible/blob/master/osa_toolkit/generate.py#L397-L399 | 17:47 |
| jrosser_ | but how it is populated is less clear | 17:47 |
| jrosser_ | actually thats wrong - ignore me :( | 17:47 |
| jrosser_ | you could be right - the patch will show. It's hard doing this in an AIO because some things only happen kind of one-shot | 17:48 |
| jrosser_ | it looks OK tbh | 18:00 |
| noonedeadpunk | huh, interesting. I would likely need to try and see what's there. Btw another interesting thing, is that I inventory-manage doesn't really show all groups | 18:28 |
| noonedeadpunk | I occured on super strange things that were not shown there but debug play was doing what's expected... But I messed up with env.d there so wasn't really raising any disturbance | 18:29 |
| noonedeadpunk | feel quite sick today tbh... So really can hardly think... | 18:31 |
| noonedeadpunk | what I'm super concerned about is how these one-time things happen | 18:32 |
| noonedeadpunk | fwiw, all lxc jobs are red for 851764 | 18:32 |
| noonedeadpunk | https://zuul.opendev.org/t/openstack/build/053da6fd3614472c96a87086bc125651/log/job-output.txt#9273 | 18:33 |
| noonedeadpunk | seems we use host_containers for ssh plugin | 18:33 |
| noonedeadpunk | but don't see such direct reference in connection plugin... | 18:35 |
| jrosser_ | yeah i see how thats broken | 18:46 |
| jrosser_ | it's because all_containers references a list of things like `identity_hosts` | 18:47 |
| jrosser_ | oops | 18:48 |
| jrosser_ | it's because all_containers references a list of things like `identity_containers` | 18:48 |
| jrosser_ | before my change `identity_containers` has hosts: [] and children: [ "aio1-host_containers" ] | 18:50 |
| jrosser_ | so this all feels really messed up | 18:50 |
| jrosser_ | by cleaning up the stuff to do with child groups a bit, it seems to show that groups like identity_containers don't actually contain any hosts | 18:51 |
| jrosser_ | anyway it's late here and i'll look some more tomorrow | 18:51 |
| jrosser_ | those lxc jobs are failing because this returns no hosts https://opendev.org/openstack/openstack-ansible/src/branch/master/playbooks/containers-lxc-create.yml#L17 | 18:52 |
| *** dviroel is now known as dviroel|out | 21:26 | |
| *** anskiy1 is now known as anskiy | 21:45 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!