16:00:18 <noonedeadpunk> #startmeeting openstack_ansible_meeting
16:00:19 <openstack> Meeting started Tue Mar 16 16:00:18 2021 UTC and is due to finish in 60 minutes.  The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:20 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:23 <openstack> The meeting name has been set to 'openstack_ansible_meeting'
16:00:29 <noonedeadpunk> o/
16:00:44 <gshippey> o/
16:02:10 <openstackgerrit> Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_gnocchi stable/victoria: Provide different package lists for install and wheel build  https://review.opendev.org/c/openstack/openstack-ansible-os_gnocchi/+/780878
16:02:40 <jrosser> o/ hello
16:02:46 <ebbex> o/
16:02:53 * noonedeadpunk trying to look through new bugs
16:04:04 <noonedeadpunk> I think we don't have nothing outstanding there, except gnocchi
16:04:23 <noonedeadpunk> except maybe this https://bugs.launchpad.net/openstack-ansible/+bug/1917644
16:04:24 <openstack> Launchpad bug 1917644 in openstack-ansible "OpenStack os-horizon-install failled at task "Ensure static files are collected and compressed"" [Undecided,Confirmed]
16:04:43 <jrosser> thats the race condition i think
16:04:50 <openstackgerrit> Merged openstack/openstack-ansible-os_swift master: Use ansible_facts[] instead of fact variables  https://review.opendev.org/c/openstack/openstack-ansible-os_swift/+/780729
16:05:06 <noonedeadpunk> yep, might be actually
16:05:37 <noonedeadpunk> should we increase number of retries?
16:05:46 <noonedeadpunk> or make them configurable for so slow envs?
16:05:50 <jrosser> thats on an older branch
16:06:01 <jrosser> and i think i may have looked at a backport?
16:06:38 <noonedeadpunk> yep, you did
16:06:51 <noonedeadpunk> well, folk claims to apply patch, but who knows
16:06:58 <jrosser> https://review.opendev.org/c/openstack/openstack-ansible-os_horizon/+/775592
16:07:30 <noonedeadpunk> btw, I think I should look into tempest test for horizon and why it's not running...
16:08:21 <noonedeadpunk> ci looks pretty bad :(
16:08:53 <jrosser> yeah i've been trying to get the stable branches going again
16:09:29 <noonedeadpunk> yeah and tempest for centos 7 worked nicely actually
16:10:07 <noonedeadpunk> mean https://review.opendev.org/c/openstack/openstack-ansible-os_tempest/+/779696
16:10:35 <noonedeadpunk> ok, what I wanted to discuss today is upcoming PTG schedule
16:11:56 <noonedeadpunk> It's on week 16:  April 19 - April 23
16:11:59 <jrosser> time seems to be moving very quickly
16:12:47 <noonedeadpunk> yeah :(
16:12:54 <noonedeadpunk> and let's use this etherpad
16:12:56 <noonedeadpunk> https://etherpad.opendev.org/p/osa-xena-ptg
16:13:08 <noonedeadpunk> I think that we can get done in a single 4hour time slot?
16:13:38 <noonedeadpunk> schedule is pretty empty at the moment, so we can pick whatever day we want https://ethercalc.net/oz7q0gds9zfi
16:14:01 <jrosser> gshippey: andrewbonney ^^ PTG timetable
16:14:25 <noonedeadpunk> we can split session for 2 hours on 2 different days not to get too tired as well
16:14:50 <noonedeadpunk> I'd say let's pick between Tuesday?
16:15:15 <noonedeadpunk> or monday even
16:17:13 <gshippey> I know I will be busy the evening of Monday 19th
16:17:32 <gshippey> But obv go ahead if that works for everyone else
16:17:47 <noonedeadpunk> I'm ok with any day eventually
16:18:12 <noonedeadpunk> Probavbly Tuesday tends to me the most free one in terms of recurring meetings
16:18:32 <jrosser> we want to be late afternoon onward EU time for jamesdenton to join?
16:19:43 <noonedeadpunk> you mean 21utc+?
16:20:58 <jrosser> oh well sort of "now" ish i guess
16:22:15 <noonedeadpunk> Oh, yeah.  I was thinking about 13UTC-17UTC as we did 2ptgs in a row
16:23:02 <noonedeadpunk> I really miss there kind of 18-19UTC timeslots
16:23:37 <noonedeadpunk> we eventually can take 15UTC-17UTC both on Tuesday and Wednesday
16:24:15 <gshippey> works for me!
16:24:16 <jrosser> ok, lets do that
16:24:44 <gshippey> We'll be on BST @jrosser just to remember!
16:24:56 <jrosser> grrr
16:25:49 <noonedeadpunk> uh... I actually don't know at what time I will be lol.... Goverment is about to cancel summer time here, but they have kind of 2 weeks left for this and it's not finally approved yet
16:26:10 <noonedeadpunk> kind of ridiculous
16:26:31 <gshippey> 😂 that is pretty crazy
16:26:48 <gshippey> good thing we have UTC!
16:26:54 <noonedeadpunk> indeed
16:27:53 <jrosser> i have a couple of things if we are done with PTG stuff
16:28:30 <jrosser> good news + bad news on centos-8 stream
16:28:31 <noonedeadpunk> another thing I wanted an opinion on are policies... I've started pushing  policy-json-to-yaml and realizied that every role has kind of different implementation
16:28:45 <noonedeadpunk> what are bad news?
16:28:52 <noonedeadpunk> is it discontinued? :)
16:29:01 <jrosser> lol
16:29:13 <noonedeadpunk> or these are good ones?)
16:29:44 <jrosser> i did a patch to convert all the base images to locally built https://review.opendev.org/c/openstack/openstack-ansible-lxc_hosts/+/778713
16:30:03 <jrosser> which is all nice
16:30:11 <jrosser> but look at the runtime on the centos job :(
16:31:48 <noonedeadpunk> and what's wromng with it?
16:32:08 <jrosser> well, thats just the infra integrated job, and it takes 2 hours +
16:32:17 <noonedeadpunk> oh.
16:32:20 <noonedeadpunk> this part
16:32:31 <noonedeadpunk> and it's stream image laready?
16:33:02 <noonedeadpunk> so it has the same issue I was seeing on images from lxccontainers are all other ones
16:33:14 <jrosser> here it is with all the patches stacked up for the integrated repo
16:33:19 <jrosser> nearly 3hours for the centos jobs
16:33:20 <jrosser> https://review.opendev.org/c/openstack/openstack-ansible/+/776226
16:34:25 <noonedeadpunk> ok, so ther issue there is defenitely with lc performance according to what I see...
16:35:02 <jrosser> so something is common between the lxccontainers image and building the rootfs with dnf
16:35:07 <jrosser> but i have no idea what
16:35:21 <noonedeadpunk> iirc lxccontainers image was slow on metal as well
16:35:29 <noonedeadpunk> and it's not the case here
16:36:03 <noonedeadpunk> but that's really frustrating...
16:36:13 <jrosser> it is, becasue the changes to lxc_hosts are really nice
16:36:22 <jrosser> it removes tons of code and makes it really simple
16:38:34 <noonedeadpunk> yep...
16:39:17 <jrosser> anyway, i don't really know what to do about that
16:39:18 <noonedeadpunk> I can recall you have tried out lxd as well is some patch? I'm wondering if there's any difference in performance..
16:39:35 <jrosser> i did a small amount of work but it needs a big effort really
16:39:48 <noonedeadpunk> gotcha
16:39:50 <jrosser> patches exist for that
16:40:16 <jrosser> anyway, just wanted to flag that we are not done on centos-8 stream yet even though the patches are passing
16:40:30 <jrosser> and really introducing a 3 hour job is -------
16:40:36 <noonedeadpunk> maybe I should register for free rhel and question support about wtf is going on with lxc on rhel?:)
16:40:49 <jrosser> "its not supported kthx'
16:41:09 <jrosser> "have you tried podman?"
16:41:22 <noonedeadpunk> fair
16:41:37 <jrosser> second thing was, are we at the point to remove nspawn code
16:41:49 <jrosser> given i am doing massive tidy up patches......
16:42:17 <noonedeadpunk> I wasn't able to find a difference in the docker image we're retrieving atm vs linuxcontainers ones. I think there won't be much with dnf one as well...
16:42:47 <jrosser> did you diff the contents of /etc ?
16:42:50 <noonedeadpunk> regarding that - we use nspawn tests in ops repo
16:43:11 <noonedeadpunk> um, no, I guess I'm not. I was comparing package versions mostly
16:43:35 <jrosser> i wonder if it's a config thing, thats baked into the image after the pacakges are installed
16:43:46 <noonedeadpunk> not only versions, but also set of packages as well
16:43:49 <jrosser> rather than actual differences in the binaries
16:43:57 <openstackgerrit> Merged openstack/ansible-role-systemd_service master: Remove references to unsupported operating systems  https://review.opendev.org/c/openstack/ansible-role-systemd_service/+/780747
16:44:01 <noonedeadpunk> smth like sysctl...
16:44:08 <jrosser> yes something like that
16:44:28 <jrosser> perhaps something has to be set to be container friendly
16:44:35 <noonedeadpunk> hm, yes, might be worth having extra round...
16:45:07 <jrosser> so on nspawn, the roles have been broken the whole of W cycle so far
16:45:22 <noonedeadpunk> but ops passing :p
16:45:38 <jrosser> where are they in the ops repo?
16:47:11 <mgariepy> lxc container restart, the stop part seems to take a miintes each time
16:47:37 <mgariepy> so it waits for 1 minutes before kill -9.
16:47:40 <noonedeadpunk> for osquery
16:47:51 <noonedeadpunk> https://opendev.org/openstack/openstack-ansible-ops/src/branch/master/osquery/tests/inventory/test-container-inventory.yml#L39
16:48:24 <jrosser> hrrm
16:50:37 <openstackgerrit> Merged openstack/ansible-role-uwsgi master: Remove references to unsupported operating systems  https://review.opendev.org/c/openstack/ansible-role-uwsgi/+/780748
16:50:38 <noonedeadpunk> so I was about to push deprecation patch but found that
16:50:44 <openstackgerrit> Merged openstack/ansible-role-systemd_networkd master: Remove references to unsupported operating systems  https://review.opendev.org/c/openstack/ansible-role-systemd_networkd/+/780746
16:50:49 <openstackgerrit> Merged openstack/openstack-ansible-os_tacker master: Use ansible_facts[] instead of fact variables  https://review.opendev.org/c/openstack/openstack-ansible-os_tacker/+/780730
16:53:37 <jrosser> ok well perhaps we need to think a little, but iirc we did talk at the last PTG about removing nspawn
16:53:56 <jrosser> and nothing stops us cleaning up the openstack-ansible repo and leaving other stuff i guess
16:55:03 <jrosser> and last i had was i started a to-do list for what needs finishing for W https://etherpad.opendev.org/p/osa-wallaby
16:55:08 <openstackgerrit> Merged openstack/openstack-ansible-os_ceilometer master: Move ceilometer pip packages from constraints to requirements  https://review.opendev.org/c/openstack/openstack-ansible-os_ceilometer/+/779668
16:55:24 <openstackgerrit> Merged openstack/ansible-role-systemd_mount master: Remove references to unsupported operating systems  https://review.opendev.org/c/openstack/ansible-role-systemd_mount/+/780745
16:55:30 <openstackgerrit> Merged openstack/openstack-ansible-os_heat master: Use ansible_facts[] instead of fact variables  https://review.opendev.org/c/openstack/openstack-ansible-os_heat/+/780641
16:55:53 <openstackgerrit> Merged openstack/openstack-ansible-os_octavia master: Use ansible_facts[] instead of fact variables  https://review.opendev.org/c/openstack/openstack-ansible-os_octavia/+/780712
16:56:50 <jrosser> i think we are ready to move to the new pip resolver now https://review.opendev.org/c/openstack/openstack-ansible/+/770284
17:03:36 <openstackgerrit> Merged openstack/openstack-ansible-os_sahara master: Use ansible_facts[] instead of fact variables  https://review.opendev.org/c/openstack/openstack-ansible-os_sahara/+/780716
17:05:27 <noonedeadpunk> #endmeeting