16:00:06 <tiffanie> #startmeeting openstack_ansible_meeting
16:00:09 <openstack> Meeting started Tue Aug 27 16:00:06 2019 UTC and is due to finish in 60 minutes.  The chair is tiffanie. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:10 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:12 <openstack> The meeting name has been set to 'openstack_ansible_meeting'
16:00:42 <tiffanie> #topic office hours
16:00:47 <noonedeadpunk> o/
16:01:25 <mnaser> henlo
16:01:28 <guilhermesp> o/
16:01:37 * mnaser hasn't been around as much, been busy doing magnum-y things
16:02:59 <guilhermesp> we have murano in wip process though https://review.opendev.org/#/c/677830/
16:03:09 <noonedeadpunk> So, https://review.opendev.org/#/c/676410/ is basicaly ready and there are bunch of patches that already use that role (like https://review.opendev.org/#/c/678216/)
16:03:41 <noonedeadpunk> guilhermesp: I guess I've left another bunch of comments there
16:03:49 <openstackgerrit> Merged openstack/openstack-ansible-os_panko stable/queens: Remove openSUSE 42.3 job  https://review.opendev.org/677949
16:03:49 <guilhermesp> also I've been seeing discussions about the POST_FAILRES for rocky branch
16:03:59 <guilhermesp> yeah noonedeadpunk I will take a look in a while
16:04:48 <evrardjp> o/
16:04:50 <noonedeadpunk> yeah, rocky post failures has been fixed thanks to jrosser and infra folks
16:05:04 <jrosser> Hopefully post failures are sorted out now, team effort there fixed it
16:05:06 <guilhermesp> that's wonderful
16:05:09 <jrosser> Ah snap :)
16:05:50 <noonedeadpunk> but I have some "issue" with linters check for uwsgi role
16:06:22 <noonedeadpunk> So we used to check linters with script from tests repo https://opendev.org/openstack/openstack-ansible-tests/src/branch/master/test-ansible-lint.sh
16:07:25 <noonedeadpunk> Since we're migrating to the integrated tests, necessity of usage of tests repo reduces. And this makes possible to drop tests folder in repos, run_tests.sh and more things
16:10:03 <noonedeadpunk> So if we want to continue doing ansible-lint checks, we should probably move it to the main repo
16:10:42 <noonedeadpunk> What I was thinking about is to add another action for gate-check-commit.sh, like check, to launch all linters with it
16:11:33 <noonedeadpunk> But right now I don't know how to limit it to the exact role, since scenario is not designed right now for passing role names as is
16:12:52 <noonedeadpunk> or playbook name, like os-nova-install.yml
16:13:44 <evrardjp> the alternative would be to implement it directly with ansible-lint from tox?
16:14:04 <evrardjp> It's a little repeat ourselves, but that doesn't sound like a bad idea, does it?
16:14:12 <noonedeadpunk> but as I said, we will be missing plugins, ike config_template
16:14:45 <noonedeadpunk> and this returns us back to usage of openstack-ansible-tests
16:14:48 <evrardjp> isn't that working with dependencies?
16:15:05 <evrardjp> yeah let's not head that way
16:15:38 <noonedeadpunk> so yes, than we should carry dependencies, install them, etc - so run_tests.sh actually does the same...
16:16:18 <jrosser> If we run an integrated test, all the things we need must be present, what do I miss?
16:17:09 <noonedeadpunk> jrosser: you're right, and that's the thing I was thinking about. But in this case we should somehow provide what should be tested by linters
16:17:25 <noonedeadpunk> ofc we may provide setup-everything.yml for the check
16:17:54 <jrosser> The scenario is converted into a list of enabled services iirc
16:18:08 <noonedeadpunk> But I'd say that number of found problems would be enormous
16:18:42 <noonedeadpunk> yeah.. but it's not correlated neither with the name of playbooks nor with the name of the roles.
16:19:05 <noonedeadpunk> ofc, we can probably generate scenario in other way when action is check
16:19:55 <evrardjp> sorry let me clarify:
16:20:12 <noonedeadpunk> but I'm not sure that I'm leading the right way
16:20:48 <evrardjp> I meant that the roles, per se, are not reusable because they don't contain dependencies in meta/main.yml. And that's something we can nowadays fix. Maybe by doing that, ansible-lint would be smart enough and work (but I am not sure there)
16:21:57 <noonedeadpunk> I guess it may work only for dependant roles but not plugins/modules
16:22:14 <evrardjp> The other way would be to implement another job, that ensures all the right repos are present in zuul, and for which the parent folder is a role folder in an ansible configuration. This would allow to test a role (current patchset folder) with a complete view of other roles
16:23:05 <evrardjp> dependencies of roles shipping plugins work inside ansible. Maybe not in ansible-lint, but in ansible it totally works. But that's a feature gap that could be fixable
16:23:56 <evrardjp> I am not saying we should go for that absolutely
16:24:14 <evrardjp> I am just saying we should avoid rewiring things in tests now that we are getting away with it
16:24:48 <evrardjp> after that, it's up to the person implementing things. I am happy to help PoC things if you want
16:24:56 <evrardjp> and help you figure that out
16:25:38 <mnaser> i'm just wondering what is the path to success to at least get us uwsgi refactored out
16:25:46 <mnaser> rather than block out all that work to fix a long standing issue with linting
16:26:11 <evrardjp> mnaser: you're right there. Maybe we can just merge it without linting, and implement linting on top?
16:26:30 <mnaser> i think indirectly it is being linted through roles that consume it
16:26:44 <evrardjp> yeah that too
16:27:00 <evrardjp> so that kinda helps on the story to not care about linting right now
16:27:06 <noonedeadpunk> So using another job which will prepare things sounds like a moving things from tests repo to each role, which might extend the maintenance costs
16:27:24 <mnaser> i think long term, we should move towards using collections
16:27:31 <noonedeadpunk> But if dependencies inside meta really works, that would be fantastic
16:27:38 <evrardjp> noonedeadpunk: it can be inside one repo, or inside the main repo
16:28:14 <evrardjp> The idea with that new job, is that it simply applies to all repos without issue.
16:28:49 <evrardjp> (even outside OSA I mean)
16:29:05 <evrardjp> for our tripleO folks :)
16:29:07 <evrardjp> anyway
16:29:16 <evrardjp> I can check the dependency real quick today
16:29:27 <noonedeadpunk> it would be great tbh
16:33:46 <jrosser> For bind-to-mgmt I am stuck on rabbit and galera - maybe we need to look at what other install tools do there, not had time to progress that
16:35:21 <jrosser> I also have Octavia on my todo list but also lack of time - the stable branches are all in a bad way due to changes in the amphora test image which are now only usable on master
16:36:23 <jrosser> I think we need some new code to build our own images in test jobs now
16:36:32 <guilhermesp> oh... I was about to test octavia in stable/stein... good to know I may face bad things in the course of testing it on metal
16:36:39 <guilhermesp> maybe I could try to help with it
16:37:07 <jrosser> You can deploy it but it won’t work with the image we lost in defaults/main.yml
16:37:27 <guilhermesp> oh dear...
16:37:34 <jrosser> If you can write some tasks to build the image that would be great, I can dig out some notes later
16:37:40 <noonedeadpunk> btw, I'm worried that neutron role is broken as well since it doesn't pass calico test https://review.opendev.org/#/c/677311/
16:38:19 <noonedeadpunk> probably we should ask logan- to check it if he can?
16:38:48 <johnsom> Ping me in #openstack-lbaas if you have questions about the image builds.
16:39:50 <guilhermesp> great johnsom thanks!
16:46:29 <jawad_axd> Please help with this playbook error http://paste.openstack.org/show/765664/. Ceph integration works fine with glance and nova. For cinder,  it creates container for cinder_volume on ceph node but that container seems to have no internet access. Any suggestions?
16:47:28 <jawad_axd> Using Stein.
17:00:15 <tiffanie> #endmeeting