16:00:48 #startmeeting OpenStack Ansible Meeting 16:00:49 Meeting started Thu Nov 12 16:00:48 2015 UTC and is due to finish in 60 minutes. The chair is odyssey4me. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:50 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:53 The meeting name has been set to 'openstack_ansible_meeting' 16:01:04 \o 16:01:25 o/ 16:01:29 hey 16:01:29 o/ 16:01:32 o/ 16:01:45 o/ ? 16:01:54 o/ 16:01:55 * palendae waits for rollcall 16:02:14 BjoernT: not sure if you're here? 16:02:25 yes 16:02:59 lol, whoops 16:03:15 #topic Agenda & rollcall 16:03:35 Hello! 16:03:38 o/ 16:03:46 \o 16:03:47 o/ 16:03:49 o/ v2 16:04:11 welcome oneswig 16:04:18 \o 16:04:28 cloudnull: thanks, first tiem 16:04:40 we'll be gentle 16:05:07 welcome oneswig :) 16:05:17 right, let's get the show on the road 16:05:20 #topic Review action items from last week 16:05:32 except it wasn't last week, or the week before (summit week) 16:05:35 moo. 16:05:42 #link http://eavesdrop.openstack.org/meetings/openstack_ansible_meeting/2015/openstack_ansible_meeting.2015-10-22-16.02.html 16:06:00 no action items, so all is well 16:06:14 #topic Items in progress or otherwise in need of work 16:06:35 #topic Installation guide blueprint 16:06:49 KLevenstein Sam-I-Am ^ your topic :) 16:06:53 yup 16:07:04 what say you ?! :) 16:07:08 KLevenstein: moo? 16:07:29 ok... 16:07:32 so here's the thing 16:07:37 mostly I wanted to call attention to it 16:07:44 oh, there she is 16:07:46 since I think it’s important 16:08:13 it is :) 16:08:28 are we looking for volunteers to make it happen? or further input of some sort? 16:08:33 further input 16:08:34 it’s also relevant to various things we’re trying to manage from the rpc-openstack side 16:09:02 the direction it takes (and o-a in general) impacts rpc-openstack 16:09:39 the original plan was to more or less point to or consume the o-a docs for the generic installation bits... the rpc docs merely augment it with rpc-specific things 16:09:39 #link http://specs.openstack.org/openstack/openstack-ansible-specs/specs/mitaka/install-guide.html 16:10:07 however, it seems like o-a is taking a turn toward not being opinionated or fully deployable... just modules to consume. 16:10:43 not exactly, but continue 16:10:52 well thats the problem 16:10:59 docs doesnt really know whats happening with o-a 16:11:05 Sam-I-Am: no, OSA will break the modules out but the repo will still continue to be deployable 16:11:41 we could use input on the spec, then, for more clarity on what needs to be documented, and how. 16:11:59 's/modules/roles/' 16:12:01 we're trying to figure out if we should create (or continue to maintain) an entirely separate install guide for rpc that covers all of the things rpc needs... or expect that we can consume the o-a install guide. 16:12:12 cloudnull: submodules? 16:12:19 KLevenstein ah, I think when the spec was approved it seemed to cloudnull and I that the work was already in mind - it wasn't clear that there was a desire for input 16:12:23 #/kickban Sam-I-Am 16:12:45 :) 16:12:50 ok, so if I may I can clarify the general direction 16:12:52 odyssey4me: i wrote the spec when people were out on the assumption that o-a was keeping a conventional install guide that installs an opinionated-enough system that works 16:13:12 ^ Sam-I-Am that should continue 16:13:29 (fwiw I believe we should treat the OA installation documentation and the rpc-openstack documentation as completely separate documents.) 16:13:43 effectively the roles, playbooks and scripts in the repository(ies) are tools in a toolbox for deployers to use as they choose 16:14:17 Sam-I-Am: the role movement will have no effect on the deploy-ability of the OSA project. 16:14:20 as per the spec in http://specs.openstack.org/openstack/openstack-ansible-specs/specs/mitaka/gate-split.html the repository will still provide meta plays to showcase how to use everything to implement well known use cases 16:15:28 there will need to be pre-role documentation for OSA to describe the options for configuring the roles... I would guess that you would not want to duplicate these docs in rpc-openstack 16:15:33 *per role 16:15:50 odyssey4me: okay, the spec should capture that type of information 16:15:56 the idea was minimizing duplication and places to track changes 16:16:05 if its not specific to rpc, it resides in the o-a docs 16:16:06 there will also need to be documentation for how to use the tooling to setup your environment - essentially the basic install guide we now have 16:16:22 therefore if o-a docs change, the rpc docs follow 16:16:37 so I agree with cloudnull here, the existing install guide provides value to OSA and will continue too 16:17:15 ok. question is... can we expect it to be maintained? 16:17:21 and it would be awesome to have rpc-docs working on OSA docs as upstream and then only looking at specific RPC opinionations in a different docs location 16:17:30 there's still some rax technical debt in it (such as the architecture) 16:17:44 Sam-I-Am yes, thus far the maintenance has been done by the community, for the community 16:18:10 we could do with some help making it pretty, as it's become an organic growth mess 16:18:19 true. does it work still? 16:18:44 or even not pretty, but to keep it simple and validated on an ongoing basis 16:18:57 it sort of has fallen by the wayside 16:19:15 hence the spec... effort to figure out how to make it both useful and easier to maintain? 16:19:19 we've had many patches in the last few months from people working through them and validating them, so I expect that it still works 16:19:23 odyssey4me: I think if we (rpcdocs, and anyone else) have a clearer idea of what needs to be added/maintained in OA, it’ll—yes, what Sam-I-Am said 16:19:35 a large proportion of the specs since the liberty release has been docs 16:19:50 so I’d like to see the spec enhanced a bit with things like what you said: add pre-role documentation, etc. 16:20:17 basically i'm looking for more info on recommended structure 16:20:20 ok, perhaps we can set aside half of next thu's meeting to work through that in an etherpad, with discussion? 16:20:27 right now the docs are more or less a copy of what we had for rpc 16:20:47 basically if you guys can facilitate a structured discussion then we can focus energy on it with a productive outcome 16:20:47 or maybe we carry that convo into the channel ? 16:21:04 or that 16:21:12 odyssey4me cloudnull: I approve 16:21:33 but an etherpad to outline all the things would be great . 16:21:34 I can set up an etherpad 16:21:40 sweet 16:21:52 yeah, we can chat after the meeting a bit - but I think that a structured and prepared for discussion that's facilitated would be most productive - which needs the time to prep 16:22:05 here you go: https://etherpad.openstack.org/p/oa-install-docs 16:22:21 #link https://etherpad.openstack.org/p/oa-install-docs 16:22:59 yay 16:23:01 ok, so we can work up a framework and put Q&A's in during the next week - then can we have a volunteer to actually facilitate the finalisaiton of it in next week's meeting? 16:23:51 KLevenstein would you be happy to do so? 16:23:55 i may not be around for next week's meeting 16:23:55 sure 16:24:05 great, thanks! 16:24:32 #action KLevenstein to facilitate a discussion for documentation improvements in the next community meeting 16:25:09 #action all to contribute to https://etherpad.openstack.org/p/oa-install-docs with thoughts/ideas/questions ahead of the workshop 16:26:09 alright, we have no other specific topic at hand on the agenda but I did want to make sure that we have some open discussion 16:26:18 #topic Open discussion 16:26:38 I'd appreciate people's thoughts on https://review.openstack.org/#/c/242225/ 16:27:15 palendae: no immediate feedback, other than thank you 16:27:54 It's gonna require some work; I'd really like to get to the point where tests can just import the dynamic inventory, but I think that'll have to be done in steps 16:27:59 palendae I haven't looked in too much detail, but a few thoughts 16:28:27 one is that the general standard for tests in openstack projects appears to be to use the directory 'tests', not 'test' - so that'd be good to change 16:29:29 also, I think that we're going to end up with a variety of tests with other files needed, like this - so I'd suggest an extra subfolder with a name that matches the tox env name - does that make sense? 16:30:16 Yeah. We're also going to need different files for inventory tests, because testing with the AIO example files is just 1 case 16:30:24 ie put all files into tests/inventory/ 16:30:40 palendae: I think that change is great as a first step to generally improving our inventory bits 16:30:43 ah sure, but perhaps that can be an additional case added afterwards 16:30:46 so thank you ++ 16:30:54 odyssey4me: Right, I'm not adding those right now 16:31:02 But they'll be needed 16:31:42 I'm pretty sure that there could be a lot more tests implemented for the dynamic inventory. 16:31:56 Absolutely 16:32:14 did we hit any issues in the past w/ the inventory that we haven't accounted for yet? 16:32:21 i know the dupe IP thing was the obvious 16:32:31 mattt TDD doesn't work like that :p 16:33:06 mattt: One that may be a candidate https://bugs.launchpad.net/openstack-ansible/+bug/1512883 16:33:06 Launchpad bug 1512883 in openstack-ansible "Host name of haproxy in openstack_user_config.yml causes recursion error" [Undecided,Confirmed] 16:33:21 Though I've not really dug into that 16:35:00 one of the goals for this cycle is to increase test coverage to improve code quality 16:35:12 we should do this for every bit of python that we have in openstack-ansible 16:35:19 all the modules that we're carrying etc. 16:35:34 so it'd be great if those interested could spend some cycles considering how best to do some of that 16:35:35 so i think this is a great start 16:36:12 mattt: ++ i think once we find a place for the libs/plugins to live outside of the main repo we can embark on adding tests and such 16:36:33 this also comes into play for improving infrastructure upgrades - for example, before upgrading - run some tests to confirm that the environment is actually in good working order 16:36:40 also there's a lot of work happening in ansible for v2.0 interms of openstack libs to manage installs 16:36:50 then afterwards, do another test to confirm that stuff's not broken 16:36:51 I think ideally the inventory will become a lib that could stand alone, too 16:36:52 we might want to see what we can begin using there. 16:36:57 But that may be a long way off 16:37:34 so we need to think about tests in layers - unit, functional, etc 16:39:56 alright, anyone else got something they want to raise? 16:40:27 the neutron db migrations change, if someone has time to test 16:40:31 (pulling up the review now) 16:40:41 https://review.openstack.org/#/c/240560/ 16:41:13 ohai 16:41:16 ++ https://review.openstack.org/#/c/241483/ -- when people have time 16:41:18 sorry I am last 16:41:21 laste* 16:41:22 grr 16:41:28 o/ hi pabelanger 16:42:00 pabelanger o/ and welcome 16:42:49 we're currently in open discussion, so feel free to introduce yourself and present anything for discussion 16:43:03 Since it is open discussion, let me do intros. Name is Paul Belanger, work at Red Hat. Looking to help contribute to ansible more. Been working with openstack-infra for the last 3.5 years, so I want to help improving the testing infra for ansible (roles) 16:43:34 welcome ! 16:43:45 So, over the next few days. I'll likely be asking questions / helping to move all the ansible roles to a common test format. For example, tox -eansible-lint 16:43:52 pabelanger yeah, I've tried to engage on that topic with infra a few times but the interest is small, so your collaboration is welcome! 16:43:57 talking to cloudnull last night, seem we both share a common goal 16:44:07 ++ 16:44:50 odyssey4me: Ya, for sure. Personally, I have an unofficial project at https://github.com/openstack/ansible-role-nodepool and have plans to build out testing for it. If we can share testing stuff, I think it will be great 16:44:50 As a general question of direction, is the long-term aim to have modules that provide mechanism and configuration data to implement policy? It looks like this is the way the project is driving? 16:45:50 pabelanger agreed, the more we can help each other the better 16:46:19 oneswig if I understand what you mean - that is currently how everything's designed but we're driving to make that clearer 16:46:36 great. So, like I said. I'll likely be asking some questions in the coming days and help bridge the effort between openstack-ansible and openstack-infra :) 16:47:14 odyssey4me thanks I'm interested in using OSA to explore different configs and seeing the effect 16:47:29 oneswig right now it seems like everything's tightly coupled, even though all roles are usable independently - we're splitting the roles out into their own repositories to clarify that they're independently usable, but the playbooks also form two functions 16:47:41 But with that, I have to run. Need to walk daughter to school. 16:48:14 one function is that we have playbooks which do certain functions - like install keystone on multiple hosts, or build containers on multiple hosts 16:48:25 awesome, thanks pabelanger 16:48:53 those playbooks can be seen as kind-of an API within the project... they can get consumed by meta-playbooks 16:49:30 the meta-playbooks may be built by deployers to mix and match things as they want, but we also include meta-playbooks which cover the use-cases we test with and support 16:49:55 odyssey4me this sounds like just what I am looking for, thanks 16:49:59 then we have configuration, which we split entirely out of the code tree - and only provide samples and docs in tree 16:50:59 oneswig great! 16:51:17 anyone else wish to raise anything? 16:51:31 * Sam-I-Am raises a beer 16:54:21 alright, thank you all for your time :) 16:54:25 #endmeeting