00:02:24 #startmeeting azure_working_group 00:02:24 Meeting started Thu Mar 1 00:02:24 2018 UTC. The chair is Kylie_. Information about MeetBot at http://wiki.debian.org/MeetBot. 00:02:24 Useful Commands: #action #agreed #halp #info #idea #link #topic. 00:02:24 The meeting name has been set to 'azure_working_group' 00:02:25 nitzmahone: Error: Can't start another meeting, one is in progress. 00:02:34 #chair nitzmahone 00:02:34 Current chairs: Kylie_ nitzmahone 00:02:40 Hi Matt 00:03:35 Three topics: 1) Bug fixes for 2.5 (requested by top customers) - Zim; 2) branch strategy - Matt; 3) F2F meeting - Kylie 00:03:40 Any other topic? 00:03:49 #chair zikalino 00:03:49 Current chairs: Kylie_ nitzmahone zikalino 00:03:54 @Kylie_ Remember that the meetbot admins have asked us to include "Ansible" in the names of the meetings, and you always want to use the same name for the meeting or they won't be easy to find on the zodbot site 00:04:29 (the bot is shared with Fedora infrastructure) 00:04:35 O, sorry, I missed it. 00:05:06 Ansible Azure Working Group. 00:05:31 yep, that's what I've always used (since they asked us to include "Ansible" in there somewhere, anyway) 00:06:01 It looks good. I forgot it, sorry. Keep in mind. 00:06:27 It seems Zim's connection has some problem. Yuwei is out for US VISA interview. 00:06:58 Zim, are you ready for #topic 1 bug fixes for customers? 00:07:11 yes, just had some connection problems, but i am back 00:07:30 go ahead. 00:07:37 You submitted 5 PRs. Right? 00:07:49 Anything you need Matt's help here? 00:08:10 so i have 5 small prs fixing issues, i will paste them here in a moment 00:08:26 i guess we need some help merging, our shipit doesn't work 00:08:32 Hey 00:08:38 sorry I'm late 00:08:45 (another topic to add: do we know Yuwei's status on the API versioning/kwargs fixes across the rest of the module stuff?) 00:08:58 Hi Jordan:) 00:09:09 hi 00:10:18 Matt, thank you for asking. It is in the plan of this sprint (by 3/16). 00:10:32 OK, not going to make 2.5.0 then. 00:10:44 I noticed 2.5 is Beta now. 00:10:49 (which is fine, but good to know) 00:11:03 We can continue to backport those for 2.5.x as they get done 00:11:06 Bug fix can be in 2.5? Or 3/16 is too late? 00:11:09 #36826 #36824 #36818 #36768 #36766 00:11:14 these are the numbers 00:11:46 When is date for 2.5 code freeze and official release schedule? 00:11:47 We're at 2.5.0rc1, rc2 will probably be next week, targeting 2.5.0 final on 3/15 00:12:09 o, ok. 00:12:41 https://github.com/ansible/ansible/pull/36826 00:12:47 #36824 #36818 #36768 #36766 00:13:13 Matt, Jordan, can you help review them and cherry pick back to 2.5 if pass review? 00:13:20 So the way we want to do these going forward is once they're merged to devel, have the submitter create a PR to cherry-pick back to stable-2.5 and ensure that tests are passing and that cherry-pick occurred cleanly, rather than relying on the release manager to do it 00:13:33 Zim, two shipit do not work. Is it because changes target networking and vm (certified)? 00:13:48 That would certainly do it 00:14:05 certified requires commit to merge 00:15:20 only community is allowed bot merge 00:16:01 Then #action item @nitzmahone @jborean93 review #36826 #36824 #36818 #36768 #36766. Once merged, #action item @zikalino create PR to check-pick back to 2.5. 00:16:24 Does the PR for cherry-pick request review and approval again? 00:17:15 yep- I'll add the ones that are eligible for cherry-pick to the 2.5.x blocker project, so if Zim can link to the cherry-pick PR underneath the original devel PR via comments, that'll help me find them and make sure they get into the first available 2.5.x release 00:17:21 well, it's pr, so somebody must approve and merge there :-) 00:18:12 Understand for certified. Before the holiday, we also met the issue two shipit against a module without certified do not work. I remember Matt mentioned it may be a bot issue. Does it fix? 00:18:19 Is it fixed? 00:18:38 Thank you Matt. 00:19:03 If no objection for following #action, we could move to topic 2. 00:19:29 #topic branch strategy 00:19:43 The problem was that test changes weren't being properly connected to the module, so the bot thought the PR touched files that required commit to merge- I can look to see if it's been fixed 00:20:07 The 2.6 branch strategy is still WIP- I don't have the answer yet 00:20:27 We'll have to revisit that for next week. Just do business as usual for now 00:20:29 ok, i think i can check with our community modules 00:21:21 @zikalino nope, looks like the bug is still open: https://github.com/ansible/ansibullbot/issues/808 00:23:27 So ... next topic? 00:23:29 Could you please help push it? 00:23:57 I can poke the bot maintainer, but he's got a lot going on 00:24:23 Otherwise, as maintainers Zim and Yuwei could not anything - assign bug to themselves, or review/approve some azure community (non-certified) modules... 00:24:31 Thank you. 00:24:37 #topic F2F2 meeting 00:24:55 I sent one thread to you guys. I remember Matt, you locate at Oregon. Right? 00:25:03 Zim/Yuwei will travel to Seattle. 00:25:12 During 3/19-4/1. 00:25:14 Yes, but there's no Red Hat site here- I'm remote. 00:25:32 I'm also in Australia so not near Oregon at all 00:26:11 Dylan locate at San Francisco. Right? 00:26:30 Any chance for you guys to travel to Seattle? :) 00:26:35 I'll be in the middle of a bunch of travel from 3/12-3/29, so probably wouldn't work out too well for me... :( 00:26:41 I also could see any MSFT site at Oregon. 00:27:14 if you find the money I may be interested but it is a long distance :) 00:28:23 Seems very far :( 00:28:33 Matt, how about 3/30? Will you be at Oregon? 00:29:11 Yes, I should be back by then 00:29:50 You know, you guys work so closely but not see each other in person. I believe a F2F is a good opportunity to let both side developers know each other better and work more closely. 00:29:55 There's a small Microsoft site nearby, but after a couple weeks of travel I'm probably not going to be keen on 6 hours of driving... ;) 00:30:20 So if they *really* want to come this way, I can head to the Microsoft office nearby 00:30:48 you mean nearby is 6 hours? :-) 00:30:54 6 hours means driving to Seattle? I believe they could go to Oregon. 00:31:10 No, I just mean round-trip to Seattle (more like 7-9h with traffic) 00:31:15 i know there's some site in oregon working on surface hub 00:31:32 Ok. then let us say 3/30, Oregon MSFT site. 00:31:34 Hrm, I heard they killed that one (Wilsonville) 00:31:49 The one I'm talking about is in Tigard 00:32:44 Noted. 00:32:52 10260 SW Greenburg Rd #600, Tigard, OR 00:33:00 (about 5 minutes from my house) 00:33:38 Nice. 00:34:30 Hi Matt, Jordan, We got more and more feature requests about Ansible on Azure. It is good. But the challenge is how to move fast for code review and merge. 00:34:55 If branch strategy of 2.6 can help, that is great. If no, we need to work together to figure out a way. 00:35:27 Any good idea? 00:36:40 For example, if you compare the feature list (https://docs.microsoft.com/en-us/azure/ansible/ansible-matrix), Zim already got a bunch of SQL/MySQL/PostgreSQL modules ready to create PRs. It definitely heavy workloads for you guys. So far we hold on and push critical fixes first. 00:36:54 Any idea to fasten this process is appreciated! 00:37:56 If you have the modules in a role for consumption by others, maybe the best process would be to tackle the currently issues before adding in more modules 00:38:30 that way the Ansible modules are on a good standing in terms of support and reliability before we start adding more and more modules 00:40:21 yes, that's what we are focusing on right now -- we want to clean up backlog of issues 00:40:29 Yes, that is why we focus on PRs for bug fixes and refactoring first. 00:40:55 Then I'd suggest working on new modules out of band and allowing them to mature there first 00:40:58 Instead pushing more module PRs at this moment. 00:41:29 I've felt pressured from the business to let things through that I wouldn't normally (eg, keyvault stuff is not ready for primetime yet and IMO should not have been shipped for 2.5) 00:41:52 Many problems exist with using these modules in the real world 00:42:23 A member of our team has been trying to work with the Openshift team to get Openshift deployable on Azure, and he's hitting all sorts of problems 00:42:56 Do the Azure docs link to the GH page for the out of band modules that you maintain? Having people know where to place bug requests for those would help to shape and form those modules into a more Ansible-like fashion 00:43:16 Rather than increasing the surface area and tech debt by shipping modules that have not been proven in the real world, I think we should work on improving what's there. 00:43:23 Harold is working on enabling openshift on Azure through Ansible. I guess he is working with your member together. 00:44:31 This would probably be a hotter issue, except he just went on paternity leave, but I'm sure it'll come up again in 2.6 00:46:23 Any issue related to key vault, let us know. 00:47:32 The primary issues were known before it was merged- they're not bugs, just that there are missing pieces to make it usable in the real world (principal lookup) 00:47:57 modules into a more Ansible-like fashion:). I guess a training or sharing from Matt during F2F will help Zim/Yuwei grow up to learn more Ansible style. 00:48:25 i actuall agree with @nitzmahone and @jborean and we could try to test every module in real world through our role 00:49:53 regardning principal lookup for instance -- i was wondering whether such functionality should be implemented inside module or maybe we could just provide external lookup plugin 00:50:58 I would probably do it in the module first, if it is used by multiple locations then the azure module_utils would be a good place 00:51:15 I have no objection for that:) In fact, I like the idea to use the role test the maturity in real world. But finally they will go back to upstream, still need effort for review (of course, maturity test in role could help) which means to catch up 2.6 we will still have a long list. 00:53:25 Yeah, that's kinda my point- I feel like a lot of the things we're spending time on in review would be much more solid if there were real-world use of it ahead of time (and also that after a bit of that, you'll be able to anticipate the kinds of criticisms you're getting both from us and community usage ahead of time) 00:53:54 btw, i was just wondering whether there should be similar module for Azure, like this: https://docs.ansible.com/ansible/2.4/aci_rest_module.html 00:54:23 this would allow patching any missing functionality before modules in particular area are available 00:54:53 zikalino: there is the uri module you can use for raw requests 00:55:07 of course there is azure_rm_deployment but it's not perfect 00:56:45 also i have one question, has any kind of module "inheritance" been considered in ansible? 00:56:57 Having a module that could do raw requests from dicts while sharing the authentication layer and stuff could be a good "escape hatch" for that, but it wouldn't be possible to do idempotence generally... 00:57:17 what do you mean module inheritance? 00:57:57 in general i mean using existing module from new module. 00:58:26 so calling another module from a module or something? Or is this in terms of development of modules? 00:58:32 sorry I don't fully understand 00:58:40 Not really; module_utils is the only sharing mechanism that's supported 00:59:18 The Azure base class is actually one of the most complex ones in Ansible (probably only the networking modules are more complex) 00:59:37 there is action plugins where the code is run on the controller and that can run 1 or more modules on the host. You only really need that if you need to do work on the controller though 01:00:00 So it'd theoretically be possible to do even more complex class hierarchy stuff for true "inheritance", but I suspect the complexity would override the utility in most cases. 01:03:25 Thank you all for great discussion here. We have the same goal - let real customer successfully automate their workload through our Ansible modules in real world. The focus of our March is to clean up backlogs (review open bugs, customer request). One scenario we received recently is - create a VM (without public ip) in a new resource group but use existing vnet. 01:03:33 yes, i was just wondering what is general thinking, thanks :-) 01:04:47 Any other backlog we should pay attention on or typical real scenario we should look into? Matt/Jordan, feel free send note to us. We are still learning how to getting Ansible better run on Azure. Thank you. 01:05:21 no worries 01:06:28 Thank you:) 01:06:32 If no other open, I will end the meting. 01:07:00 #endmeeting Ansible Azure Working Group