16:00:55 #startmeeting OpenStack Ansible Meeting 16:00:56 Meeting started Thu Aug 13 16:00:55 2015 UTC and is due to finish in 60 minutes. The chair is odyssey4me. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:57 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:00 The meeting name has been set to 'openstack_ansible_meeting' 16:01:08 o/ 16:01:12 o/ 16:01:16 o/ 16:01:19 #topic Agenda & rollcall 16:01:20 o/ 16:01:27 #link https://wiki.openstack.org/wiki/Meetings/openstack-ansible#Agenda_for_next_meeting 16:01:40 \o/ 16:02:14 howdy all :) 16:02:52 o/ 16:04:03 o/ 16:04:22 Morning 16:04:31 hi 16:05:08 #topic Review action items from last week 16:05:20 * odyssey4me sigmavirus24 and cloudnull to put together a POC for pip and epoch's 16:05:29 ugh 16:05:35 stuck in upgrade hell right now 16:05:39 haven't had time for this 16:05:40 sigmavirus24 I take it that will need to carry considering that cloudnull is still on leave? 16:05:46 alright 16:05:47 yeah that too 16:05:55 #action sigmavirus24 and cloudnull to put together a POC for pip and epoch's 16:05:55 The idea is on the ML 16:06:09 * odyssey4me odyssey4me to release 11.1.0 on 7 Aug 2015, any blueprints not yet merged to be postponed to 11.2.0 16:06:36 so it was released on 10 Aug instead, but it was released with all the blueprints completed :) 16:06:52 well done everyone for getting that done, even with our busy days! 16:07:01 #link https://launchpad.net/openstack-ansible/kilo/11.1.0 16:07:15 there was a lot of work that went into there 16:07:45 * odyssey4me odyssey4me to set 11.2.0 release date to the (tentative) date of 4 Sep 16:08:23 Good job everyone 16:08:30 :round of applause: 16:08:47 I've set the date for 11.2.0 as 11 Sep to ensure that we have enough time to get two hotfix patches into 11.1.0, and another week for getting the 11.2.0 feature patches merged in. 16:08:54 https://launchpad.net/openstack-ansible/+milestone/11.2.0 16:09:00 #link https://launchpad.net/openstack-ansible/+milestone/11.2.0 16:09:07 from me a big thanks on helping meon the Ceph patch 16:09:38 thank you svg for doing that work and being patient with us in the reviews and testing 16:10:06 I thought you guys were the ones being patient :) 16:10:10 if any bugs are raised against that work, please target them to 11.2.0 for kilo so that we can track which ones need ot be backported 16:10:25 ack 16:10:48 alternatively patches can also be registered against the blueprint considering that it's still not released yet 16:11:04 not master? 16:11:19 ok my bad 16:11:36 evrardjp patches get initially added to master, but they won't be backported to kilo until the week before 11.2.0 16:11:49 so ceph patches into master are welcome 16:11:56 :) 16:12:10 is everyone happy with the dates for the milestones? 16:13:41 alright, moving on 16:13:45 #topic Blueprints 16:13:57 #link https://blueprints.launchpad.net/openstack-ansible 16:14:48 <-- late 16:14:52 I've added a liberty blueprint. For anything you see that may need follow up in liberty - whether it be an upstream bug, or perhaps settings that need changing due to deprecation - please associate them with that blueprint. 16:15:49 I'd like to propose adding the upstream liberty-related milestones to trunk so that we can target specific things and also have an easy reference for the dates to help with planning. 16:16:18 Are there any objections to that? 16:16:21 Nope 16:16:29 We need more structure around our Liberty stuff 16:16:42 Even though forward progress on that has been slow 16:16:42 #link https://wiki.openstack.org/wiki/Liberty_Release_Schedule 16:17:01 all we have left is liberty-3, then liberty-release 16:17:10 I have a feeling we'll miss liberty-release 16:17:18 At least any configuration changes 16:17:39 palendae I think it'll be easier than you think. :) 16:17:56 We'll see. I won't consider it working til we have working upgrades 16:18:25 now you're moving the goalposts 16:18:32 we will definitely have a functional release on the release day - we've been working through kinks as we've updated our sha's and have been fixing upstream bugs as we discover them 16:19:05 hughsaunders: Nope, not at all 16:19:09 palendae the upgrades will probably be a lot easier as we don't have as many structural changes as juno->kilo 16:19:22 hughsaunders: Function upgrades are more important than greenfield deployments IMO 16:19:27 odyssey4me: Hopefully :) 16:20:18 I'd like to ask everyone to take the time in the next week to look through the registered blueprints and to self assign where possible, also setting a goal milestone 16:20:47 it'd also be a good idea to get a starting review uploaded so that the design discussions can happen 16:21:19 palendae has a good one up already 16:21:23 #link https://review.openstack.org/207713 16:22:01 personally I think this one is fine, and I believe it's important for adding RHEL support later 16:22:04 #link https://review.openstack.org/203708 16:22:49 also if we could get some reviews on this patch to the specs repo so that we don't get so many merge conflicts, that'd be grand: https://review.openstack.org/208440 16:23:39 #action odyssey4me to add upstream liberty milestones to the trunk series 16:24:10 #action all to review blueprints, self-assign and target, then submit and review blueprints 16:24:58 I'd like to suggest that we spend most of next week's meeting discussing our roadmap and breaking down some of the steps into easy pieces to chew on. 16:26:00 Sounds reasonable 16:26:44 +1 16:26:45 part of the thought we need to put into it is how we can best serve our downstream consumers 16:27:35 We have different types of consumers currently - RPC is an example of a consumer which embeds openstack-ansible wholesale into something bigger. 16:28:18 Compass is a relatively new project building a UI around deployment, and is looking to consume our playbooks and roles. 16:28:46 We also have deployers consuming the playbooks & roles as they are, and perhaps doing their own bits around or over the top. 16:29:33 I'd like us to consider as part of our roadmap making the roles individually consumable, testable and reusable outside of our reference builds. 16:30:23 Do we have Compass contributors telling their opinion? 16:30:30 that could be nice 16:30:46 It'll make it easier for SME's to get involved in adding functionality without having to get too involved in the entire stack. 16:30:58 (or even commiting/improving the product) 16:31:35 #link http://lists.openstack.org/pipermail/openstack-dev/2015-August/071892.html 16:31:44 ^Compass call for contributors 16:32:34 I've been having conversations with various parties to get feedback about what holds them back from getting involved. 16:32:48 The feedback has been, so far: 16:33:12 1 - it'd be great if it was easier to get a dev/test environment going 16:34:17 so they ask to make openstack easier? βΈ® 16:34:19 2 - there is no support for implementing on the platform I use (usually CentOS) 16:35:17 ok 16:35:17 1) curl | bash 16:35:44 svg not openstack, but a dev/test environment to understand how to do operator things (add nodes, replicate a bug, etc) and also how to do dev things (add a role, feature or functionality to the stack) 16:36:04 hughsaunders: I was perhaps reading 1 differently: as a multi-node environment which is not that simple. 16:36:26 stevelle: fair 16:36:35 stevelle yes, an AIO is relatively easy in a VM or Cloud - but not everyone has a VM environment or Cloud 16:36:48 #link https://blueprints.launchpad.net/openstack-ansible/+spec/deploy-with-vagrant 16:36:52 So i'm not sure how we fix that 16:36:56 If they don't have access to VMs 16:37:07 ^ has been added and I'm liaising with the guy to add his Vagrant conf and some docs into the repo 16:37:12 maybe we make an installable iso, Apsu could you get working on that? 16:37:16 this opens another door 16:37:24 Or I'm misunderstanding what you're saying, because you just linked to deploying with vagrant after you said not everyone has access to VMs :) 16:37:25 andymccr: Right on it 16:37:27 we could also do better at docs 16:38:02 it could have like questions that stores vars like your public address range etc. 16:38:31 Compass handles the PXE booting so I guess it won't be hard to add osad on top of it when the management network is running 16:38:46 anyway - part of getting more host OS platforms supported are these blueprints: 16:38:48 andymccr: this wizard could be done with or without the iso 16:38:59 #link https://blueprints.launchpad.net/openstack-ansible/+spec/mariadb-upgrade-to-v10 16:38:59 but we are getting far from original topic, right? 16:39:06 #link https://blueprints.launchpad.net/openstack-ansible/+spec/add-support-for-systemd 16:40:45 evrardjp right, the point is not for us to change what we do which is deploy on existing hosts - the point is to make it easier for evaluators and developers to get to grips with how what we've built works and how they can contribute 16:42:07 most of the time when people raise their issues I find that they just don't know that we already do things that they thought we didn't 16:42:50 so I think we could also do better at sharing our successes and engaging with SME's in the various projects to assist when we're working on things that they could help with 16:42:53 how much of the friction is from the particular way that we use Ansible? 16:43:05 thinking of vars specifically 16:43:35 stevelle the 3rd objection that they have is to do with the opinionation - we have a pretty rigid way of deploying 16:43:40 though I know questions about inventory exist too 16:44:08 stevelle I think that's more of an issue for us than others at this stage. No-one's mentioned that just yet. 16:44:29 #link https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0CCEQFjAAahUKEwiwivvewKbHAhUGF9sKHdx-B9U&url=https%3A%2F%2Fgithub.com%2Fopenstack-ansible%2Fopenstack-ansible&ei=YsnMVbDRDIau7Abc_Z2oDQ&usg=AFQjCNEdKMLeFRncUo1F0xilQ5svqPrfCA&sig2=trguuB1XtNTgs3jHgn3LTg&bvm=bv.99804247,d.ZGU 16:44:48 #action odyssey4me to slap himself with a wet fish 16:44:55 #link https://github.com/openstack-ansible/openstack-ansible 16:45:16 ^ I've engaged with the guys who worked on these playbooks and roles. 16:45:20 odyssey4me: I've seen several questions from people about how the dynamic_inventory.py script works 16:45:43 They're currently on icehouse and have committed to evaluate the stack we have and provide feedback. It's early days though. 16:46:23 I think variables handling could prevent adoption: I was afraid myself at first 16:46:35 palendae agreed, those questions come from those who've actually deployed and tried doing hard things :) 16:47:01 palendae has suggested that he'd like to write up some reference documentation about it all 16:47:21 Yep, I have some things sitting on my hard drive to get pushed up 16:47:50 I'm of the opinion that we can break down a lot of barriers to adoption by making the roles independant units, consumable on their own. 16:47:56 Sure 16:48:06 It will allow those interested to tackle broader OS support in a smaller unit. 16:48:30 I agree 16:48:31 It will also allow people to deploy in whatever way they choose, instead of the way we've chosen. 16:49:05 It will also make it easier to get to grips with the pieces that people are interested in, instead of the monolithic large pile of code we have now. 16:49:12 That's gonna be a very long term road map, fwiw 16:49:24 well, maybe - but maybe not 16:49:26 I think Keystone's a great place to start, but this is not going to be all done by 12 16:49:39 Some of our roles are very intertwined 16:50:30 if we split the roles into their own repositories - which will be in the openstack namespace, then we use submodule references in os-ansible-deployment in order to handle a full cycle of deprecation, then it may be achievable 16:50:51 That's the high level, yeah.. 16:50:57 during the M cycle we can work on making them more individually functional and individually tested 16:51:05 Though submodules everywhere...we were down that road with the chef stuff, no? 16:51:39 yeah, not ideal - but after the one cycle we remove them and have them all as ansible galaxy roles 16:52:08 Ok, I think it's a reasonable high level plan. I just think the details may bog it down 16:52:10 or shall we say the submodules stay in the branches they need to stay in, but in master we switch 16:53:09 Anyway - let's put some thought into the future this week coming, and try to get our thoughts into specs. We can make the blueprints dependant on each other to help understand what follows what. 16:54:02 Our time is just about up, so I'd like to open the floor for comments, questions, suggestions and general banter... 16:54:29 * odyssey4me slaps himself with a wet fish 16:54:37 * odyssey4me ticks 'action complete' checkbox 16:54:57 I have the Rally ansible role, ready to go, https://github.com/meteorfox/ansible-rally 16:55:08 well, you can say is the mvp version 16:55:44 yep :) didn't realise you were here meteorfox 16:55:45 it could use a bit of love in the readme :) 16:56:13 I forgot about that - I'd like to submit the review to register our first role in the openstack namespace. 16:56:20 We need to figure out a naming convention. 16:56:39 yes, I'm aware of that, and there a few directories which are 'empty' and could be deleted 16:56:40 I was thinking something along the lines of 'openstack-ansible-' ? 16:57:21 once we've registered it, meteorfox can submit his code as a first review and we can collaborate on improving it 16:57:35 thoughts? 16:57:59 I'm fine with the name convention 16:58:18 palendae sigmavirus24 andymccr hughsaunders mattt ? 16:58:28 yeah that works for me 16:58:51 you're submitting that for Liberty? Prior to the blueprint being finalized? 16:59:07 so it would be the first role in its own repo? 16:59:25 with the expectation others follow the same pattern (are moved to that pattern) 17:00:04 andymccr yep 17:00:13 we're out of time - move to #openstack-ansible 17:00:17 #endmeeting