16:00:16 <noonedeadpunk> #startmeeting openstack_ansible_meeting
16:00:16 <openstack> Meeting started Tue Oct 20 16:00:16 2020 UTC and is due to finish in 60 minutes.  The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:17 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:19 <openstack> The meeting name has been set to 'openstack_ansible_meeting'
16:00:26 <noonedeadpunk> #topic office hours
16:01:19 <noonedeadpunk> well, seems we can't discuss today our current status and patches that are due because of the gerrit situation
16:02:16 <noonedeadpunk> but we can probably discuss plan for ptg that will take place next monday
16:02:22 <noonedeadpunk> #link https://etherpad.opendev.org/p/osa-wallaby-ptg
16:02:51 <noonedeadpunk> I've just placed topics in the doc we took for V
16:03:40 <noonedeadpunk> we didn't do much since we had really lot's of stuff on our hands with centos 8, ubuntu focal, new ceph and galera and etc
16:04:43 <jrosser> o/ hello
16:06:19 <noonedeadpunk> I think we still can do several during this cycle:) like mark nspawn for removal for example
16:06:35 <jrosser> i get kind of fed up trying to fix the roles there
16:07:11 <jrosser> we also have to do some migration on ansible group names
16:07:32 <noonedeadpunk> so they have only underscore?
16:07:47 <noonedeadpunk> `TRANSFORM_INVALID_GROUP_CHARS`?
16:08:22 <noonedeadpunk> well yes, this seems like important part
16:08:47 <jrosser> yes was just looking at https://docs.ansible.com/ansible/latest/reference_appendices/config.html#transform-invalid-group-chars
16:09:41 <noonedeadpunk> well, I see nothing about it's deprecation....
16:09:56 <jrosser> iirc it's in the warnings on the zuul logs :/
16:10:02 <jrosser> maybe 2.12? cant remember
16:10:49 <jamesdenton> o/ hi
16:11:22 <jrosser> hello
16:11:35 <noonedeadpunk> well, there would be deprecated_in option then...
16:11:39 <noonedeadpunk> hey!
16:11:49 <noonedeadpunk> s/option/field
16:11:54 <jamesdenton> i see ironic listed as a broken role. what's broken? besides inspector...
16:13:21 <noonedeadpunk> I think it was result of haproxy tbh
16:13:35 <noonedeadpunk> and probably it's fixed nowadays - dunno....
16:13:43 <jamesdenton> oh, the baremetal issue?
16:13:44 <noonedeadpunk> hard to say without gerrit
16:13:48 <jamesdenton> hah yeah
16:14:11 <jrosser> oh hold on
16:14:12 <jamesdenton> the one day i have some patches to upload...
16:14:15 <jrosser> the logs still work
16:14:30 <jrosser> if you've got emails from gerrit the links to job results are still working
16:14:58 <noonedeadpunk> well, I have 2700 emails from gerrit at the moment....
16:15:23 <jamesdenton> i don't have them.
16:15:24 <jrosser> heres what happened in the last os_ironic job http://paste.openstack.org/show/799220/
16:15:35 <noonedeadpunk> https://zuul.opendev.org/t/openstack/build/f65e2435af0f448a8a13bfbcbea0de23
16:16:50 <noonedeadpunk> ok
16:17:01 <jrosser> looks like db setup?
16:17:19 <noonedeadpunk> looks like swidft issue
16:17:42 <noonedeadpunk> Account POST failed: http://172.29.236.101:8080/v1/AUTH_3e09bb30f84c4e2d98c249da5a48bb14 503 Service Unavailable  [first 60 chars of response] b'<html><body><h1>503 Service Unavailable</h1>\\nNo server is av'"
16:18:14 <nurdie> Does anyone know of any writeups, additional help, or examples for deploying Magnum on an existing OSA cluster? I find the existing "openstack-ansible-os_magnum/latest" to be mildly daunting
16:18:51 <jamesdenton> jrosser noonedeadpunk might be easier for me to pull down master and run tests locally
16:19:04 <noonedeadpunk> oh, and swift can;'t work with bind-to-mgmt at the moment
16:19:05 <noonedeadpunk> https://zuul.opendev.org/t/openstack/build/9e9e951822754be9b544bbce8066141a/log/logs/host/swift-proxy-server.service.journal-11-21-58.log.txt#63
16:19:32 <noonedeadpunk> well, seems we've missed it
16:19:38 <jrosser> oh no! missed one
16:19:39 <jamesdenton> is there no var set for that?
16:19:41 <jamesdenton> kk
16:20:22 <noonedeadpunk> well, if it wasn;'t swift, I'd say we will easily fix that...
16:20:39 <noonedeadpunk> but at least we understand the issue:)
16:22:19 <jrosser> https://github.com/openstack/openstack-ansible-os_swift/blob/master/templates/proxy-server.conf.j2#L6
16:22:50 <noonedeadpunk> yeah. the thing is it has more bindings:)
16:23:04 <noonedeadpunk> but probably we don't use haproxy for them....
16:24:01 <jrosser> https://github.com/openstack/openstack-ansible-os_swift/blob/master/defaults/main.yml#L231-L237
16:24:11 <jamesdenton> there's a similar issue i found with the spicehtml5proxy service. i defaults to 0.0.0.0 and you need to set html5proxy_host (!= server_listen)
16:24:41 <jamesdenton> i have a patching waiting
16:25:00 <noonedeadpunk> awesome
16:25:28 <jrosser> thats kind of interesting defauilting the swift ports to ansible_host
16:25:59 <jrosser> that means storage traffic on the mgmt network even though theres a storage interface? wonder what the intention is here
16:26:08 <noonedeadpunk> yeah.... but anyway, the only problem should probably be swift-proxy, as we use only 8080 in haproxy
16:26:40 <noonedeadpunk> jrosser: well, in aio we have storage traffic over mgmt anyway - ceph does exact same thing
16:26:48 <jrosser> it does
16:27:10 <noonedeadpunk> nothing to be proud of but we have what we have
16:28:17 <noonedeadpunk> but agree that super weird
16:28:35 <jrosser> theres storage and storage i guess
16:28:45 <jrosser> iscsi != swift != ceph
16:30:27 <noonedeadpunk> but all of them are supposed to use storage net by default anyway? except swift proxy or rgw, which is under haproxy
16:32:03 <noonedeadpunk> ah, another question - have anyone played with new pip resolver?
16:32:37 <noonedeadpunk> except placing patch which has shown that we don't fit
16:33:06 <noonedeadpunk> as I wanted to look into it during this week
16:36:09 <jrosser> i've not done anything beyond the patch to test it
16:36:46 <jrosser> i think i figured some new defaults/overrides need adding to python-venv-build
16:37:21 <jrosser> as the existing var to enable things like the new resolver gets passed to the system python before you upgrade the one in the venv
16:37:40 <jrosser> and the system python doesnt understand the new flags, so that needs splitting into two vars
16:38:11 <jrosser> sorry system pip
16:40:56 <noonedeadpunk> sounds reasonable despite I didn't have much dive into the topic yet
16:41:04 <noonedeadpunk> well, moving to it might be challenging
16:41:32 <jrosser> seems to be what we put in constraints for source git repos now needs to be in requirements, or similar
16:41:43 <noonedeadpunk> yep
16:42:11 <noonedeadpunk> that the only thing I brought out of this actually
16:42:23 <jrosser> and i guess we have to be careful to keep python2 support there for the os_tempest role
16:43:01 <noonedeadpunk> I'm wondering if tripleo uses py3 for centos7 there
16:43:16 <noonedeadpunk> as probably we shouldn't be so carefull:)
16:47:14 <noonedeadpunk> arxcruz: have any insights?:)
16:54:46 <noonedeadpunk> well, will ask them later then
16:54:53 <noonedeadpunk> #endmeeting