14:03:12 #startmeeting ansible-sig 14:03:13 Meeting started Fri Sep 13 14:03:12 2019 UTC and is due to finish in 60 minutes. The chair is cloudnull. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:03:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:03:16 The meeting name has been set to 'ansible_sig' 14:03:20 \o 14:03:21 welcome to the first ansible-sig meeting 14:03:22 o/ 14:03:42 how's everyone doing on this fine friday ? 14:04:45 I'm good thanks cloudnull. 14:04:59 sounds like folks need a ping :D 14:05:31 yeah, especially given I'm not really an active member right now, just curious :) 14:05:53 dmellado dmsimard EvilienM evrardjp guilhermesp gundalow jrosser logan- mwhahaha noonedeadpunk odyssey4me owalsh redrobot sshnaidm|off tiffanie yoctozepto zbr - ping 14:06:12 pong 14:06:14 fun with meetings 14:06:15 o/ 14:06:21 👋 14:06:27 fun with meetings indeed :D 14:06:36 * owalsh lurking 14:06:38 o/ 14:06:39 * cloudnull master of the out of band ping 14:06:41 * EvilienM hides 14:06:57 these firday evening meetings:) 14:07:11 so mnaser has a last minute conflict, so your stuck with me. 14:07:19 sorry in advance 14:07:39 it's been ages I haven't you lead a meeting, can you still do it? 14:07:47 I haven't seen you lead* 14:07:48 :D 14:08:00 evrardjp it'll be touch and go 14:08:05 but i'll fake it till i make it 14:08:14 so now that we're all here, we wanted to cover: 14:08:16 #topic What are going to do together 14:08:22 I am pretty sure it's like bicycle 14:08:50 * evrardjp go backs into lurking mode 14:09:00 evrardjp https://pics.me.me/%D0%BC%D0%B5%D0%BCey-ov-%D1%81om-image-result-for-bicycle-crash-meme-bicyclememes-50276296.png 14:09:17 ouch 14:09:27 so anyway. I think there are a lot of things we can all work on, spanning roles, modules, plugins, etc 14:09:28 anyway, your goals? 14:09:39 evrardjp one day, when I grow up 14:10:01 I also think there are patterns we can collaborate on 14:10:12 even if we end up using different roles, or plugins. 14:10:21 cloudnull: do you have something in mind already? outside the config_template and, for example, testing? 14:10:38 the os-tempest role is a good example between osa and tripleo 14:10:57 the connection plugin as well is another good one 14:11:20 wouldn't working on mitogen be better than carrying our own connection plugins? 14:11:26 we have a couple folks working on adapting the connection plugin so we can use it in tripleo 14:11:41 maybe 14:11:41 oh ok. That works too. 14:11:46 :D 14:12:00 mitogen is super cool, and works great, when it works 14:12:05 yeah, mitogen looks more prespective. And we'll be able to follow your's example:) 14:12:20 I think no matter what we do we'll end up having to maintian something, its just a matter of what we want to maintain 14:12:34 agreed 14:12:51 I think the overall trend inside OSA is to reduce what we have to maintain. 14:13:03 I think it'd be lovely if we could take the connection bits out of mitogen and using just that, at least initially 14:13:06 (it always has been though) 14:14:35 so what's your proposition on what can we do together? I am a little confused. 14:14:45 EvilienM mwhahaha would have a better opinion on the simplification matter, however, I think it makes sense for this group to collaborate on things to lower the tech-debt of all of our projects by focusing on collaborative points. 14:15:07 yup 14:15:14 that makes sense 14:16:04 Do you have some kind of roadmap for triple O, so that noonedeadpunk mnaser can compare it to OSA roadmap and find a series of "touch points" ? 14:16:11 from where I sit right now, the the plugins in tripleo and OSA are very complimentary. I also think Kolla has something things that we could all adapt and benefit from . 14:16:18 roadmap: survive 14:16:25 hahaha 14:16:29 EvilienM: oh. I see 14:17:12 a lot of our roadmap is around scale (req for Edge) and simplification (reduce tech debt) 14:17:29 and both touch ansible 14:17:30 kolla is in an interesting place where we have everything in-tree, and to do otherwise would add a little complexity for our users and cognitive burden to our devs 14:18:23 but enough useful stuff becomes shared, we might get to a point where it makes sense to take that hit 14:18:29 *but if 14:18:33 ++ 14:18:40 the docker connection plugin sounds neat 14:18:42 EvilienM: I see. Simplification is on the roadmap of OSA as far as I can tell too, so we are at least aligned on the goals :) Now I guess it's about how and the tech to be shared 14:18:56 cloudnull: hey 14:19:01 o/ 14:19:15 we have something similar for executing ansible in our kolla_toolbox container, but it's a bit of a fudge 14:19:20 isn't there already a docker connection plugin in upstream? 14:19:24 owalsh has been working on the docker plugin, which should easily support lxc, nspawn, docker, podman, etc. 14:19:34 evrardjp only for local connections 14:19:35 evrardjp: only for localhost 14:19:43 oh I thought this has changed 14:19:48 had* 14:20:00 o/ 14:20:10 is the connection stacking from mitogen something that could be extracted? 14:20:17 cloudnull: ack, well I was but I've been dragged into other stuff. Hoping to get back to it soon... 14:20:18 that seems like a more general solution 14:20:31 I like this indeed 14:20:52 I wonder if it's come up in ansible itself 14:20:54 we can reach out to dw and figure out if it is 14:21:06 maybe gundalow knows more? 14:22:04 https://github.com/ansible/proposals/issues/25 14:22:47 A single 'connection' does not always get us to where we want to. This is NOT an attempt to solve ssh proxying, but a more complex ssh + chroot or ssh + docker to access hosts. 14:22:57 is exactly what OSA has had for years 14:23:30 so maybe we can work on getting that RFE in 14:23:43 sounds rly nice 14:24:05 ya the way the mitogen connection stuff works is somewhat similar to osa last I looked. its more abstracted in that you can stack things arbitrarily but the way it works with strategy+connection plugin layout is not a ton different 14:24:07 which could be a whole sale import of the OSA plugin, pulling bits from mitogen, or something new? 14:25:01 I thought we tried bringing this kind of connection plugin in the past, is my memory serving me wrong? 14:25:11 we did 14:25:14 it was rejected 14:25:24 but maybe they changed their minds? 14:25:51 so here the idea would be to come with more ppl, and try to say "hey we need this" and if it doesn't work, try to make mitogen our "workaround" ? 14:26:10 more voices, from more projects 14:26:10 I am trying to understand how I can help here 14:26:12 from more companies 14:26:20 ok 14:27:11 do does anyone wnat to take lead on the connection plugin efforts ? 14:27:21 Assuming it doesn't work, we could also have a galaxy namespace and or a "collection" ? 14:28:11 i think mnaser has a namespace on galaxy ? 14:28:20 he's AFK so we'll have to circle back on that 14:28:24 ok 14:28:35 we do have the connection plugin from osa going into a stand-alone repo https://review.opendev.org/#/c/676421/ 14:28:36 * noonedeadpunk has no info about that 14:29:01 **we do have / we have a proposal to have 14:29:20 cloudnull: ok, so I suppose the idea would be to work on that project, and when it's ready, try to upstream it 14:29:24 is that your plan? 14:29:29 hi, actually here now 14:29:40 evrardjp thats always my plan 14:29:43 :D 14:29:49 ok 14:30:00 however, I'm happy to abandon that effort and pivot if we think it best 14:30:31 nope I think it's fine, it will naturally merge our ideas into one code base, so that it's ready for upstream when we think it's mature. 14:30:48 it seems mgoddard was interested too, so that looks like a nice place to reduce tech debt for at least 3 projects 14:31:04 +1 14:31:58 owalsh do you have yoru docker bits in a repo that we could push into the forked connection plugin ? 14:32:27 cloudnull: not yet, hopefully have time next week to look at this 14:32:27 I know you're a little occupied at the moment, but if you have a branch out there Id be happy to upstream what we have so far? 14:32:34 owalsh ++ 14:34:02 so I think the connection plugin is well covered. 14:34:07 what else? 14:34:44 OSA and TripleO already work together on tempest things, is there a way we could get Kolla into the mix? 14:34:53 Personally I am not too fond of our tempest in OSA. Even if we tried to collaborate on tripleO, I would prefer if tempest was just something I have on my client side, and say "test me this" 14:34:56 is there something we can do to better foster that collaboration ? 14:35:10 I thought for that a container would be best. Which leads to kolla tempest testing 14:35:23 +1 14:35:27 but I know nothing about it. And right now our things work 14:35:41 but I would love to hear about "just reusing a test container" :) 14:36:00 is there a tempest artifact we could consume? mgoddard 14:36:50 like something we could just pass a clouds.yaml, a few env vars like OS_CLOUD, a blacklist/whitelist, and then it would automatically test things? 14:37:00 kolla provides a tempest container image, if that's what you mean? 14:37:10 I don't know how widely used it is though 14:38:16 I did have a poke around in the new tempest role. Some of the config seems quite specific 14:38:51 which one are you talking about, and can we make that not specific so it's used by more than one project? 14:39:11 and... do you install python-tempestconf by any chance in the tempest container? 14:40:11 I was looking at https://github.com/openstack/openstack-ansible-os_tempest 14:40:29 cloudnull: other question, as you have now a view of both OSA and TripleO, is there a series of roles where OSA is too opinionated, and therefore can't be reused by triple O? Or the other way around, roles not opinionated so they can be reused? 14:40:47 mgoddard: yeah, I think it's not great for reuse -- which is why I am proposing to change that :) 14:41:01 https://github.com/openstack/openstack-ansible-os_tempest/blob/master/templates/tempest.conf.j2 14:41:31 only skimmed it and didn't use it though, probably missing sometihng 14:41:45 mgoddard: could you tell us more about how the template would be a problem? 14:42:16 cause this gets templated, this is just the normal file, after that you layer your overrides 14:42:23 flavor_ref = 200 14:42:25 example 14:42:32 (even those you don't see as var) 14:42:36 ok, I guess that helps 14:42:58 SO iirc flavor with this id is being created by role itself 14:43:03 evrardjp most of the OSA roles are OSA specific. especially when it comes to the os_* roles. that said, I'm sure we could make some of them work just fine, however, I've not tried to use any of them in TripleO at this time. 14:43:07 mgoddard: https://github.com/openstack/openstack-ansible-os_tempest/blob/85442eaed4e73eccb5c562d2601fd523a81ee9c5/tasks/tempest_post_install.yml#L40 14:43:40 noonedeadpunk: it's not necessary though, see also tempestconf 14:44:00 cloudnull: I am more thinking about the memcached, galera, etc. 14:44:27 I'd need to look at it properly to be sure 14:44:38 I've not tried, but I suspect those would be more compatible. 14:44:50 evrardjp and mwhahaha might have some thoughts there ? 14:45:20 I suppose it's for EvilienM :) 14:45:44 don't use templates? whaat? 14:46:01 wow that escalated quickly 14:46:05 hahaha 14:46:25 could be don't use ansible, but we're not there yet :D 14:46:31 soon 14:46:37 baby steps 14:47:06 opinionated is fine, it's just where you put that opinion 14:47:11 when it's hard coded in a template, it's a problem 14:47:35 putting it in the role would be better but then having to define everyone of thsoe options is aweful in ansible 14:47:57 this is why config_template was kinda neat here 14:48:19 yea i guess for reusability, adopting the config template would be beneficial 14:48:44 the other thing is, if the hard coded values are defaults in python should be actually define them at all? 14:48:45 although yes, it's less black box to have all those variables defined 14:49:03 in puppet we nuke things in the config file if they aren't different then service defaults 14:49:45 e.g. https://github.com/openstack/openstack-ansible-galera_server/blob/master/tasks/galera_post_install.yml#L97-L122 - everything in mariadb+galera is setup via config_template 14:49:52 technically we shouldn't carry the defaults 14:49:53 a slight aside but we had an email thread to support folks asking about comments in config. it seems like the overall feeling was people would prefer if the oslo config comments were there 14:50:08 if we do carry the defaults, then there is a bug in the template I would say 14:50:24 if they aren't the default, we should have some information indicating why they aren't the default as well 14:50:33 makes it easier to explain why things differ 14:50:33 but we are talking about os_tempest role, which is, IMO not a good place for the best collaboration 14:50:36 mwhahaha ++ https://github.com/openstack/ansible-config_template/commit/09c76e238026d7ba4134ee2b66a4e9fd2617b843 14:50:48 mwhahaha: fair 14:51:13 EvilienM asked me to make comment parsing better to preserve all of the comments and their structure 14:51:42 at last check I think we're at feature parity with how puppet does it 14:51:44 EvilienM? 14:52:29 i think so 14:53:45 #topic wrap-up 14:53:58 great conversation all 14:54:09 I think we have a couple things to get started on. 14:55:14 a quick take away, we will continue looking at the connection plugins and follow up with the mitogen folks to see if we can use / extract only the connection stacking parts. 14:55:20 maybe help work on the ansible RFE 14:55:49 mnaser noonedeadpunk mgoddard can you all huddle up on some of the ways we can make tempest testing better? 14:55:56 -cc evrardjp 14:56:01 cloudnull: yes we have a better parity now 14:56:08 cloudnull: i just didn't spend time on the next steps 14:56:17 pushups 14:56:19 :D 14:56:40 * EvilienM will do pushups when weshay|ruck do his pullups 14:56:46 ++ 14:56:59 OK Im going to call this meeting, as I have another meeting to get to. 14:57:05 thank you all 14:57:05 yep, sure :) 14:57:09 #endmeeting