15:00:03 #startmeeting Ansible Core Meeting 15:00:03 Meeting started Thu May 11 15:00:03 2017 UTC. The chair is thaumos. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:03 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:00:03 The meeting name has been set to 'ansible_core_meeting' 15:00:08 #chair thaumos 15:00:08 Current chairs: thaumos 15:00:15 Good day everyone! 15:00:28 * mattclay waves 15:00:30 'lo 15:00:37 #chair mattclay branko 15:00:37 Current chairs: branko mattclay thaumos 15:00:40 * nitzmahone lurks 15:00:46 #chair nitzmahone 15:00:46 Current chairs: branko mattclay nitzmahone thaumos 15:01:00 * gundalow waves 15:01:05 #chair gundalow 15:01:05 Current chairs: branko gundalow mattclay nitzmahone thaumos 15:02:21 I'll give folks another minute before we get started 15:03:12 howdy 15:03:16 hey sivel :) 15:03:33 #chair sivel 15:03:34 Current chairs: branko gundalow mattclay nitzmahone sivel thaumos 15:04:04 #topic ansible/ansible#24463 15:04:08 #link https://github.com/ansible/ansible/pull/24463 15:04:37 I just wanted to see if we're cool with merging it. 15:04:52 thaumos: I'm happy with it 15:04:57 Fine with me. 15:05:03 cool 15:05:06 abadger1999: will https://github.com/ansible/ansible/pull/24463/files need backporting to stable-2.3? 15:05:10 +1 15:05:13 should I backport? 15:05:21 I assume METADATA needs to be kept in sync across branches? 15:05:28 thx gundalow 😃 15:06:22 #action thaumos to follow up with abadger1999 offline to confirm backport 15:06:32 :) 15:06:44 I'd assume so as well, unless we want to show history in some way. 15:06:57 #topic Open floor 15:07:06 * gundalow assumes so, though the first one, so time to ask akward questions 15:07:12 :P 15:07:18 heh 15:07:47 Just hijack topic, or...? :) 15:08:13 Pretty much, @branko. If you have something bring it up! 15:08:56 #topic https://github.com/ansible/ansible/pull/23015 (DConf + D-Bus daemon spawning/cleanup within module) 15:09:10 \o/ 15:09:18 So, not sure if people have read this one, I'll try to write small description. 15:09:31 #link https://github.com/ansible/ansible/pull/23015 15:09:50 Essentially, wrote a module for doing configuraiton changes via dconf tool. This is used by recent Gnome and Gnome/GTK+ apps. 15:10:07 However, you need running D-Bus user session for this (dbus-daemon binary). 15:10:24 And you should try very hard not to spawn multiples, because things can step over each-other. 15:10:57 The problem is that there is no offical standard way to ensure you have just one daemon running, and there is no official way to get its address. 15:11:05 The address needs to be set via environment variable. 15:11:29 Depending on distro you are using, you may or may not get things running in some standard way. 15:11:35 if there is no sure way to do it, just make it a requirement of using the module 15:11:45 if env var is not set, user must set it for module 15:11:48 Yes, but that removes a lot of usability from it. 15:11:57 i dont disagree, but i dont see other way forward 15:12:29 aside from finding a spare tardis and getting dbus/systemd to standarize in a discoverable way 5yrs ago 15:12:37 Heheh 15:12:56 So, is daemon spawning/managing a big no-no in Ansible modules? 15:12:57 That was roughly my recommendation as well. 15:13:47 Not everyone would be happy if a daemon spawned randomly. 15:13:57 It is somewhat mixing the functionality of the module. Is it to start dbus, is it to configure things via dbus? 15:14:05 * thaumos shakes head at own poor attempt at demon joke 15:14:32 It is to configure things via d-bus. 15:14:40 branko: not normally, service modules are used for that, most other modules have no need 15:14:47 d-bus is a special monster 15:14:47 Maybe break the module into 2? The dconf stuff in one, another module that can manage whether dbus is running 15:14:56 in another 15:15:20 Hm... That's an option, I guess. The user would still need to clean-up afterwards, though. 15:15:50 What if I would make the spawning an option? 15:15:57 With big red warning in docs? 15:16:08 I'd still prefer it not in the dconf module 15:16:33 not sure that would work well either 15:16:40 Hm... Would it be acceptable to have this separate module strictly for dbus-daemon management? 15:16:48 diff ssh sessions, are you going to get teh correct env for reusing the daemon? 15:17:10 For me the main thing is to know what course to take in terms of trying to make this pull request work for everyone, or just abandoning it. 15:17:29 I think it would be for everyone if it just configured, and that's it. 15:17:30 bcoca: Maybe :) 15:17:44 And that's also one of the problems. There is no reliaable way to get connecting address. 15:18:16 i think at thsi point simplest approach is bets, make module require 'workable environment' , it trying to spawn services to make this work seems very wrong 15:18:49 think of it similar to ssh-agent 15:19:09 ssh-agent will not conflict with other copies 15:19:18 dbus will 15:19:29 i meant in just the sense that if you don't have it running, it don't work 15:19:59 good point, thought you meant its pretty safe to start if needed (which it is) 15:20:26 I'm just thinking on how useful the module would be to other people if they need to understand all about D-Bus daemons etc in order to use it. 15:20:32 here the issue is intrinsic to the tools involved, dbus has no good way to detect itself running nor safe way to run multiple user sessions of it 15:20:50 It's probably safe to assume that if they're gonna use the module, they might know it. 15:21:01 branko: if you are using dconf batch confiuration, you SHOULD know about dbus 15:21:25 bcoca: Well, I can tell you I didn't when I started off with this :) 15:21:36 Wasn't even aware dconf needed it, actually. 15:21:55 branko: i dont doubt, neither did i first time i delved into it, was eye opener (also reason i avoid dconf) 15:22:27 I think eye opener is an understatement :) 15:22:28 branko: most people aren't, which for me forces exposure on how badly designed this is for batch/central management 15:22:31 I think my jaw drapped ;) 15:22:49 Ok, I guess the conclusion is that no D-Bus daemon spawning of any form should be in the dconf module, and it should be replaced with user providing address for communication? 15:22:56 so branko, to close out, just keeping it simple and only have it configure. 15:22:58 yep 15:23:25 branko: you can try default (see if var is there) but otherwuse, yeah, onus is on user to discover proper setting 15:23:40 Hm... 15:23:43 just document it clearly 15:23:51 While at it, should user pass this as an option, or environment parameter? 15:24:00 either works for me 15:24:04 and probably for module 15:24:19 i think option is nicer 15:24:26 Ok, will do that then. 15:24:50 I'll open the floor if noone has any extra input? 15:24:54 #action branko to add option to module for dconf module 15:25:10 #action branko to remove D-Bus daemon handling from dconf module 15:25:15 While at it 15:25:25 thanks! 😃 15:25:28 #topic Open floor 15:25:32 Thanks for the input :) 15:26:54 any other topics to discuss today? 15:27:07 patch 7.2.5 15:27:39 lol 15:29:11 if no topics at the half, I'll end meeting 15:30:13 https://i.makeagif.com/media/12-30-2015/zuZvki.gif 15:30:19 #endmeeting