15:02:47 <noonedeadpunk> #startmeeting openstack_ansible_meeting
15:02:47 <opendevmeet> Meeting started Tue Aug 30 15:02:47 2022 UTC and is due to finish in 60 minutes.  The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:47 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:02:47 <opendevmeet> The meeting name has been set to 'openstack_ansible_meeting'
15:02:52 <noonedeadpunk> #topic roll call
15:02:57 <noonedeadpunk> o/ hello everyone
15:03:34 <anskiy> hello :)
15:04:51 <ebbex> o/
15:07:15 <noonedeadpunk> #topic bug triage
15:07:27 <noonedeadpunk> I know that couple of bugs has landed during last previous weeks
15:08:02 <noonedeadpunk> #link https://bugs.launchpad.net/openstack-ansible/+bug/1987405
15:09:41 <noonedeadpunk> hm, so it's about network-generic-switch which seems to be an ml2 driver
15:12:30 <noonedeadpunk> While I kind of see reason for this bug report, it feels also a bit off
15:13:06 <anskiy> this one was discussed a bit on last meeting
15:13:11 <noonedeadpunk> ah
15:13:22 <anskiy> it's the bug from jrosser_ team
15:13:47 <noonedeadpunk> yeah, I know that. Then I guess I just should to wrap bug triage if it was already discussed
15:13:57 * noonedeadpunk forgot to read trough meeting notes
15:14:07 <noonedeadpunk> #topic office hours
15:14:48 <noonedeadpunk> I think I should create another bug report about the way how we configure cinder
15:15:41 <anskiy> there were actually no resolution on how to proceed with fixing it: https://meetings.opendev.org/irclogs/%23openstack-ansible/%23openstack-ansible.2022-08-23.log.html#t2022-08-23T15:07:09
15:16:59 <noonedeadpunk> yeah, I already found that, but still thanks
15:17:44 <noonedeadpunk> so for cinder, there's issue with what we do for ceph backend, as we configure active/active without using tooz, which is not supposed to be that
15:18:09 <noonedeadpunk> And the problem that it's quite hard to rollback from a/a since in DB all volumes have cluster already set
15:18:26 <noonedeadpunk> and fixing that qould require deploying etcd or zookeeper
15:19:23 <noonedeadpunk> I've already touched that slightly previously, but this needs proper thinking through
15:20:08 <noonedeadpunk> though in fact in production envs we don't see issues with current setup, but I can imagine some nasty race conditions to happen though
15:21:58 <noonedeadpunk> regarding neutron - probably we should jsut set default to not use uwsgi and backport it to Y
15:22:29 <noonedeadpunk> jsut to check that neutron-rpc-server will get stopped as expected with that switch, as I can imagine it won't
15:23:46 <noonedeadpunk> as it will be just filtered out instead of being stopped/disabled/masked
15:25:51 <noonedeadpunk> other then that tbh I don't have much to discuss, as still trying to get in sync with all I've missed for 2 weeks
15:26:41 <ebbex> yeah, just add ml2.genericswitch to the exeption in neutron_use_uwsgi.
15:28:25 <noonedeadpunk> or that
15:28:36 <noonedeadpunk> that is even faster
15:29:06 <noonedeadpunk> though I can imagine there can be more things that does fail
15:31:10 <ebbex> how? They'll be back to running plain non-uwsgi neutron-server.
15:32:05 <ebbex> it probably worked for them before the uwsgi split.
15:41:37 <noonedeadpunk> I meant there can be more ml2 plugins that does not work with uwsgi
15:41:48 <noonedeadpunk> so except list can be wider in fact
15:42:27 <noonedeadpunk> but yes, it's kind of safe and simple thing to do right now
15:42:46 <ebbex> yeah, hehe :) i think pretty much everyting except lxb and ovs would have problems.
15:43:21 <noonedeadpunk> Oh. One more thing. After switching to cloudsmith, it seems it was not that perfect decision, since they tend to clean-up repos quite freqently and in a bit weird manner
15:43:50 <noonedeadpunk> (we get rabbit/erlang from there now)
15:45:12 <noonedeadpunk> so if anybody have ideas of how or where to get rabbit/erlang from would be great. Eventually we switched to cloudsmith because native erlang repos for ubuntu are not stable, which causes CI and deployments to fail
15:45:29 <noonedeadpunk> and with cloudsmith there's another problem in versions that we can get from there
15:45:53 <noonedeadpunk> maybe, we should just pin major version, and get whatever minor version is present...
15:46:09 <noonedeadpunk> but I don't really like that approach
15:47:24 <anskiy> by native, you mean ppas?
15:53:46 <anskiy> checked git log, sorry. I've found this thing: https://launchpad.net/~rabbitmq/+archive/ubuntu/rabbitmq-erlang, but I don't know if it's any better, and it seems, it's only Erlang too
15:54:41 <ebbex> major, minor or patch?
15:55:55 <ebbex> i think patch is ok, but minor doesn't change that often?
16:01:19 <noonedeadpunk> ebbex: I think I meant patch
16:01:53 <noonedeadpunk> anskiy: nah, ppa contains only single version, while what we want is to be able to pick and stick to the version
16:02:07 <noonedeadpunk> anyway
16:02:10 <noonedeadpunk> #endmeeting