Created
February 17, 2016 21:59
-
-
Save gregdek/4ed5bd745881570a17db to your computer and use it in GitHub Desktop.
IRC Log, Ansible Contributor Summit, Feburary 2016, London
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| 4:25 AM <gregdek> Hey, can someone join the bluejeans call so we can do some level checks? | |
| 4:26 AM <gregdek> https://bluejeans.com/856479228/ | |
| 4:29 AM → crab joined (~ams@raven.toroid.org) | |
| 4:33 AM <gregdek> Hi crab :) | |
| 4:33 AM <gregdek> Will you be joining us for audio? | |
| 4:34 AM gregdek set the topic: Contributor summit audio: https://bluejeans.com/856479228/ | |
| 4:52 AM *.net ↮ *.split → alvaroaleman joined ↔ crab and rbergeron nipped out | |
| 4:59 AM <gregdek> OK, we're starting. | |
| 5:01 AM → jimi|ansible and resmo joined ← alvaroaleman left | |
| 5:02 AM <jimi|ansible> howdy | |
| 5:02 AM <resmo> hi | |
| 5:03 AM <gregdek> HI! | |
| 5:03 AM <gregdek> We're starting. :) | |
| 5:03 AM <gregdek> I will take copious notes. | |
| 5:03 AM ↔ crab nipped out | |
| 5:03 AM <vincent_vdk> hi resmo jimi|ansible | |
| 5:03 AM <gregdek> First topic: testing. | |
| 5:04 AM → j00bar joined (~j00bar@194.74.0.66) | |
| 5:04 AM <gregdek> jimi: testing is tricky. :) | |
| 5:04 AM <gregdek> Biggest problem: module testing. | |
| 5:04 AM <gregdek> YAY netsplit | |
| 5:04 AM → mordred joined (~mordred@openstack/mtaylor) | |
| 5:04 AM <gregdek> How to test without running all tests all the time? | |
| 5:04 AM <gregdek> Mocking module tests is going to be super important as we do python3 split | |
| 5:05 AM <gregdek> Making sure that modules stay compliant with both python 2.4 and python 3 | |
| 5:05 AM <vincent_vdk> how to test modules against an API you don't have? | |
| 5:05 AM <gregdek> what we have now: | |
| 5:05 AM <gregdek> * Unit tests and integration tests | |
| 5:06 AM <gregdek> * Integration tests cover core critical modules | |
| 5:06 AM <vincent_vdk> ah, you mean only core modules. k | |
| 5:06 AM <gregdek> One thing we're talking about doing: making a requirement of "core modules" have integration tests | |
| 5:06 AM <gregdek> +1s around the toom | |
| 5:06 AM <gregdek> room | |
| 5:08 AM <gregdek> Open question: should we completely block core module PRs that don't have tests associated? | |
| 5:08 AM <gregdek> And if so, how? | |
| 5:08 AM ⇐ mordred quit (~mordred@openstack/mtaylor) *.net *.split | |
| 5:09 AM <gregdek> jesse question: are there good examples of "making testing a hard requirement"? | |
| 5:09 AM → mordred joined (~mordred@openstack/mtaylor) | |
| 5:09 AM <gregdek> Current answer: the testing repository | |
| 5:09 AM <gregdek> Current tests are problematic, very serial | |
| 5:10 AM <gregdek> toshio: ziploader would also make it easier | |
| 5:10 AM → misc joined (~misc@nanardon.zarb.org) | |
| 5:11 AM <gregdek> matt: current infrastructure forces us to test everything, rather than the particular module the code is checked into | |
| 5:11 AM <gregdek> jesse: is it easy to run tests as a developer? | |
| 5:11 AM <jimi|ansible> (cd test/integration; make) | |
| 5:11 AM <gregdek> jimi-c: "it's easy for me" | |
| 5:11 AM <gregdek> :) | |
| 5:11 AM <jimi|ansible> (cd test/integration; TEST_FLAGS='--tags=test_copy' make) | |
| 5:12 AM <j00bar> once upon a time, a zope maintainer said to me about zope 3: “it _is_ easy to use. you’re just stupid.” don’t be that guy. :) | |
| 5:12 AM <jimi|ansible> (cd test/integration; make non_destructive) | |
| 5:12 AM <gregdek> LOL | |
| 5:12 AM <mordred> j00bar: ++ | |
| 5:12 AM <rbergeron> j00bar: yep, that | |
| 5:12 AM → abadger1999 joined (~abadger19@194.74.0.66) | |
| 5:12 AM <j00bar> in his defense, i'm pretty stupid. | |
| 5:13 AM <gregdek> jimi-c: I would like to start splitting tests to make them more parallelizable | |
| 5:13 AM <gregdek> Rather than having tags, make everything a make target and run more in parallel with make -j8 | |
| 5:14 AM <gregdek> You can do this now, but there are race conditions around the "prepare-test" command, so if two tests run and one finishes, it deletes the testdir, so fixing that issue will help | |
| 5:17 AM <gregdek> I am now asking if people are actually writing tests outside of the core team, and we're hearing "not a lot" | |
| 5:18 AM <gregdek> Can we move tests to alongside the modules? | |
| 5:18 AM <gregdek> (from willthames) | |
| 5:19 AM <gregdek> jimi-c has considered this | |
| 5:19 AM <abadger1999> gregdek: "Feel we all agree we need to do something to make testing better but pragmatically its always been the last thing on the priority list" | |
| 5:19 AM <gregdek> thanks toshio :) | |
| 5:20 AM <gregdek> jesse: "could we just have some deliverable that's 'optional, here's the best way to add testability to Ansible modules, it's optional now but will be mandatory at some point'" | |
| 5:20 AM <abadger1999> jimi-c has considerd putting them inline inside of the modules. toshio would like them to be in the same repo but doesn't like inline in the modules. | |
| 5:21 AM <gregdek> crab: want to make sure you're hearing ok | |
| 5:22 AM <rbergeron> crab: not sure if you talk if we can hear you but we'll find out :) | |
| 5:23 AM <crab> i'll wait until i have something useful to say ;-) (i did say "hi", but someone else was speaking too) | |
| 5:24 AM <gregdek> Tests need to be tagged "can run in travis" or "must run in jenkins" | |
| 5:24 AM <gregdek> Really, "needs an API to run" or "doesn't need an API to run" | |
| 5:25 AM → jlk joined (~jkeating@173.247.105.60) | |
| 5:25 AM <gregdek> So we currently have an internal Ansible jenkins that runs both Tower and Core tests. | |
| 5:26 AM <gregdek> One good thing to do: split out these tests and make Core run on a public jenkins instance | |
| 5:26 AM <crab> who's speaking now? | |
| 5:26 AM <gregdek> mordred is now talking about zuul and whether or not it actually exists | |
| 5:26 AM <jlk> "As a contributor, I want to be able to read/watch docs/screencasts on how to add tests to my modules, so that I can contribute tests to go with the modules I care about." | |
| 5:26 AM <crab> ah. | |
| 5:26 AM <gregdek> (monty taylor of openstack) | |
| 5:26 AM <j00bar> thoughts on ravello? it’s used elsewhere at red hat | |
| 5:27 AM <mordred> https://specs.openstack.org/openstack-infra/infra-specs/specs/zuulv3.html | |
| 5:27 AM <j00bar> https://www.ravellosystems.com/ | |
| 5:28 AM <jlk> "As a contributor, I want to be able to develop my tests for modules I care about along side the modules themselves, so that I can more easily track/submit changes to the code along with changes to the tests together." | |
| 5:29 AM <jlk> "As a contributor, I want to be able to indicate special requirements for my tests (such as Travis / Jenkins / Whatever) to prevent my tests from failing due to being in the wrong environment." | |
| 5:29 AM <gregdek> REQUIREMENT! "please make sure that you don't make every test run for every check-in." | |
| 5:30 AM <jlk> "As a test runner, I want to ensure the tests that run are scoped to the change being submitted, so that my test run is not prohibitively long and expensive and feedback is presented in a timely manner." | |
| 5:30 AM <gregdek> thank you jlk, please keep doing that :) | |
| 5:31 AM <gregdek> mordred says "hipster kidz don't use make" | |
| Load more backlog… | |
| 5:32 AM <crab> "how would you rate the overall quality of this meeting?" | |
| 5:32 AM <crab> heh | |
| 5:32 AM <gregdek> LOL | |
| 5:32 AM <gregdek> well? | |
| 5:32 AM <jlk> kwality | |
| 5:32 AM <crab> i'm going to reconnect. | |
| 5:32 AM <rbergeron> is that from bluejeans? | |
| 5:32 AM <crab> yes | |
| 5:33 AM — mordred hands jlk a cupcake | |
| 5:33 AM <jlk> "As a test developer, I want to be able to indicate dependencies for my tests, so that when changes happen in my modules or other modules I care about, testing will include all the relevant modules to the change." | |
| 5:33 AM <rbergeron> gregdek thanks jesse for helping to come up with concrete suggestions in the forms of user stories :) | |
| 5:33 AM → nitzmahone joined (uid132664@gateway/web/irccloud.com/x-jcedvmtlivhglisz) | |
| 5:34 AM <crab> fyi i don't think hipster contributors need to care about makefiles at all; just know which "make test-xxx" to run for whatever they're contributing to. | |
| 5:34 AM <jlk> How do we reward PRs with tests over PRs without tests, without implicitly making PRs with tests required? | |
| 5:36 AM <jlk> crab: they care when they want to influence how the tests are ran, or what the text matrix for a given code file is. Make was a pretty big barrier to contribution back when Fedora / Red Hat used it for the package repository | |
| 5:36 AM <crab> ok | |
| 5:37 AM <j00bar> https://www.evernote.com/l/AQxc1NsZjvlPlZiz-6_RyfatglQdU13ZsCwB/image.png | |
| 5:37 AM <gregdek> jimi-c: "I didn't realize we had tests in the modules directory already" | |
| 5:37 AM <gregdek> "So maybe we should keep doing that" | |
| 5:37 AM → newtMcKerr joined (~textual@194.74.0.66) | |
| 5:37 AM <gregdek> "In core and extras" | |
| 5:40 AM <gregdek> mordred asks: "is there a good way of going to giving resources to Ansible to test API stuff for cloud modules/etc?" | |
| 5:40 AM <gregdek> bcoca says "we'll figure it out" | |
| 5:41 AM <gregdek> resmo points out that many projects can use docker images for this kind of testing | |
| 5:41 AM <rbergeron> so can we get a list of the existing tests no matter where they are currently | |
| 5:41 AM <jlk> "As a contributor working on code that interacts with 'services', I want to be able to provide access to 'services' for the purpose of testing code contributions against said 'service', so that I can ensure contributions from others without direct access to said 'services' are properly tested" | |
| 5:42 AM <rbergeron> and put that list someplace so we can have ... things that people can look at? | |
| 5:42 AM <rbergeron> rather than "wait for things to move / change" because at the end of it... if only a few modules have tests, people will still have to hunt for them | |
| 5:42 AM <crab> is there a list of people participating in the meeting, btw? | |
| 5:43 AM <rbergeron> it doesn't have to be *forever* -- mostly at least for "while we figure out best practices" | |
| 5:43 AM <j00bar> in putting on my crazy architect’s hat… what i’d really love to see is: | |
| 5:43 AM <rbergeron> becasue right now we're in the land of figuring things out | |
| 5:43 AM <j00bar> 1) tests for modules vs. core are organized under directories with Makefiles; 2) for each “make test” jenkins/ravello/vagrant spins up VMs for supported relevant OSes and/or cloud environments and runs those tests; | |
| 5:43 AM <jimi|ansible> most are in this channel, with the exception of some Ansible employees | |
| 5:43 AM <jimi|ansible> ^ crab | |
| 5:43 AM <j00bar> 3) travis.yml uses $TRAVIS_PULL_REQUEST to look at the pull request diff and identify which directories are modified by a given pull request and runs the corresponding… (cont’d) | |
| 5:43 AM <j00bar> “make test” command for every directory in the pull request that contains a Makefile, otherwise running the top level “make test-all" | |
| 5:44 AM — j00bar is jag, architect @ ansible | |
| 5:44 AM <rbergeron> yeah, there are a few other folks here -- willthames, mscherer | |
| 5:44 AM <misc> rbergeron: I am here too | |
| 5:44 AM <misc> and we have 27 people in the room | |
| 5:44 AM <rbergeron> ohyes there are you are :) | |
| 5:44 AM <rbergeron> duhhh | |
| 5:44 AM <misc> and 16 in this channel | |
| 5:44 AM <j00bar> but ansible+redhat could on its travis/jenkins maintain creds for AWS, azure, rax, openstack, etc. to test those modules on our dime | |
| 5:46 AM <rbergeron> gregdek: what's the group responsible for moving this forward, what's the governance method, etc? | |
| 5:46 AM ⇐ newtMcKerr quit (~textual@194.74.0.66) Ping timeout: 272 seconds | |
| 5:47 AM <gregdek> ACTION! bcoca owns separating tests into the module repos | |
| 5:47 AM <gregdek> ACTION! bcoca owns creating the makefile to run those tests and discover new tests | |
| 5:48 AM <jlk> "BetaMax" -- record API integrations to a file, re-play them at wil | |
| 5:49 AM <jlk> Could be useful for testing API systems without having the system in question available to the test infrastructure | |
| 5:49 AM <rbergeron> who is responsible for continuing to poke at / wrangle the "improve testing stuff" process / group of interested humans? | |
| 5:49 AM <jlk> https://betamax.readthedocs.org/en/latest/ | |
| 5:49 AM *.net ↮ *.split ⇐ nitzmahone and drybjed quit ↔ sivel nipped out | |
| 5:50 AM <jlk> "Who mocks the mockers?" | |
| 5:50 AM <jlk> Mocked APIs can grow stale, need to ensure that they don't go /too/ stale. | |
| 5:51 AM <gregdek> jimi-c is "it". | |
| 5:52 AM → drybjed joined (~drybjed@helios.gumed.edu.pl) | |
| 5:53 AM <jlk> "As a community developer, I want to craft a 'Call to Action' to engage willing community contributors to help with the effort, so that jimi-c does not become a starved resource" | |
| 5:53 AM <gregdek> jimi-c says "let's reboot our community meeting" | |
| 5:54 AM <gregdek> "and what timeframe will this happen in?" | |
| 5:54 AM <crab> +1 to resuming community meetings | |
| 5:54 AM <gregdek> sounds like abadger1999 is volunteering to run the community meeting! | |
| 5:56 AM <crab> rbergeron: i can't hear what you say very well (but i can hear gregdek well) | |
| 5:57 AM → nitzmahone joined (uid132664@gateway/web/irccloud.com/x-henvjhbonblcvnnq) | |
| 5:59 AM <crab> who cramer? | |
| 5:59 AM <jimi|ansible> crab: tcramer (Tim Cramer) | |
| 5:59 AM <crab> ah, thanks. | |
| 5:59 AM <rbergeron> crab: yeah, sorry - i am not turning the mic on reliably, nd i think greg just has his laptop mic going | |
| 5:59 AM <jimi|ansible> ^ he's the director of engineering at Ansible/Red Hat | |
| 5:59 AM <abadger1999> vp for ansible/ansible and tower. | |
| 6:00 AM <jimi|ansible> oh doh yeah, VP now, not directory | |
| 6:00 AM <jimi|ansible> director... | |
| 6:00 AM <rbergeron> y'all are welcome to also help note things if you want :) | |
| 6:00 AM <jimi|ansible> https://github.com/ansible/ansible/blob/devel/ROADMAP.md | |
| 6:02 AM <rbergeron> jimi|ansible: thanks, didn't know where greg's irc window was :) | |
| 6:07 AM Connection closed unexpectedly | |
| 6:12 AM → gregdek and nitzmahone joined | |
| 6:12 AM <rbergeron> gregdek has yielded the floor to svrg, who is questioning the currenty pr process and tools in general, as referenced in the etherpad during the planning for this session. | |
| 6:13 AM <abadger1999> gregdek explainig how the current tools and processes came about. | |
| 6:14 AM <rbergeron> gregdek: for the future of PRs -- we'd like to be in a place where tehre is no manual commit, where things checked in have tests and code review and we don't need hands-on review and things just magically happen and get pushed in | |
| 6:14 AM <rbergeron> gregdek: no clue how to get there | |
| 6:14 AM <rbergeron> gregdek: in the meantime -- from a process perspective, we have ansibullbot | |
| 6:14 AM <jlk> gregdek also pontificating about a magical future of unicorns and faeries that tests the code and merges the submissions | |
| 6:14 AM <rbergeron> https://github.com/ansibull/ansibullbot | |
| 6:15 AM <jimi|ansible> ok jlk, NOW you're trolling... ;) | |
| 6:15 AM <rbergeron> mostly for PRs currently. | |
| 6:15 AM <rbergeron> and for extras/ core modules only currently. | |
| 6:15 AM <rbergeron> things have owners. we don't have anything quite like that in ansible/ansible. | |
| 6:16 AM <rbergeron> jimi|ansible: we used to, we had ansibot, but it was a fragile special snowflake. | |
| 6:16 AM <rbergeron> but ansibot wasn't designed to handle multiple repos, and so after core/extras split, we got away from using it. | |
| 6:17 AM <crab> i think ansibot isn't a sustainable solution, more people actually looking at *and closing* PRs is what's really needed | |
| 6:19 AM <jimi|ansible> crab: but we all spend an hour or two a day doing triage and requesting info and following up on those previous requests where we could be closing PRs or doing other things | |
| 6:19 AM <abadger1999> gregdek: +1's crab's suggestion above. | |
| 6:19 AM → nitzmahone_ joined (uid132664@gateway/web/irccloud.com/x-iihzsgdhjwzueacb) | |
| 6:20 AM <jimi|ansible> Having a bot request more info for those issues that are "X is broken, fix it" without any helpful details saves us time | |
| 6:20 AM <abadger1999> gregdek "problem with current github is that permissions are not fine-grained. You either have commit which gives you access to do a lot of things or you just have normal permissions. | |
| 6:20 AM <misc> bcoca: "there is also a issue of scope. you break a module, that's just a module, breaking ansible break all" | |
| 6:21 AM <crab> well, ansible/ansible is frequently broken and we're all still alive somehow | |
| 6:21 AM <rbergeron> issue triage is a subject for this afternoon, also. | |
| 6:22 AM <crab> rbergeron: mic! | |
| 6:23 AM <abadger1999> rbergeon says: "More about triage this afternoon. This topic is to figure out how to expand trust." | |
| 6:23 AM <rbergeron> i said that (a) issue triage is a problem, and we have time devoted to it this afternoon" | |
| 6:23 AM <rbergeron> and ... yes, what abadger1999 just said :) | |
| 6:23 AM <crab> i run devel in production | |
| 6:24 AM <crab> you're welcome | |
| 6:24 AM <rbergeron> :) | |
| 6:25 AM ← nitzmahone left ↔ nitzmahone_ nipped out | |
| 6:26 AM <rbergeron> devel vs. production is ... yes, a thing, but is that a reason to not at least expand the circle so that peopole can help out in areas where they'd like to help? :) | |
| 6:28 AM <abadger1999> when someone does work to triage how can we get them access to do this -- ex: list of PRs that are probably stale or bad - can be assembled by the community. But the only one that can close the PR is someone with github commit. | |
| 6:29 AM <abadger1999> Can we figure out how to give people the power to say no? - ( from gregdek) | |
| 6:33 AM <abadger1999> bcoca brings up sometimes there's a need that one segmnt of the community wants to fix via a PR but other community members check and think that it isn't a good change (doesn't seem to fit ansible's style, seems like there's another way to achieve it, very ugly hack...) For those someone may be needed to step in and arbitrate a solution. | |
| 6:35 AM <crab> i don't think we really need to optimise the process to protect against bad people | |
| 6:36 AM <jlk> There probably shouldn't be one set of rules for "Red Hat employees" and another set of rules for "Not Red Hat employees". | |
| 6:36 AM <jlk> (boy does that sound familiar) ^^ | |
| 6:36 AM <gregdek> :) | |
| 6:36 AM <crab> and as far as mistakes go, they *already* keep happening constantly. it's ok, things break. we can unbreak them. | |
| 6:36 AM <j00bar> either we need to accept that we’re going to have to use non-software/systems-based access control and use community norms, or we need to move issues out of github and into something with the granularity of control we want. | |
| 6:37 AM <j00bar> the hard part about using community norms is getting from “i want to contribute” to “i am trusted to commit” - there’s no middle ground. | |
| 6:37 AM <j00bar> but if we’re sticking with github issues, we’re using community norms. | |
| 6:37 AM <gregdek> +1 | |
| 6:38 AM <crab> i think a lot of people can be trusted to know where their limits are, and where to not commit stuff or ask for help or open a wider discussion etc. | |
| 6:39 AM <j00bar> and if we’re comfortable relying on social contract, awesome - we do need to talk about the process/policies about how one earns commit access and when commit access would be revoked and canonize those community norms. | |
| 6:39 AM <crab> and if they commit something that breaks, well, it can be reverted. | |
| 6:39 AM <crab> who are eric johnson and stan cutler? | |
| 6:40 AM <jimi|ansible> eric = Google/GCE | |
| 6:40 AM <abadger1999> eric johnson works on gce (google cloud) modules. | |
| 6:40 AM <rbergeron> crab: eric is a dude from google working on a kubernetes module, and sam ... used to be at red hat working on foreman, and now i think is a engineering mgr at digital ocean and i think he was working on DO modules, maybe. | |
| 6:40 AM <abadger1999> sam kottler is also working on some specific area but I don't remember | |
| 6:40 AM <crab> you can just *ask*, yes +1 to what gregdek said | |
| 6:41 AM <misc> yup, sam went to DO, but not sure for the manager stuff | |
| 6:41 AM <rbergeron> i have no idea how / why either of them got access (not that i am opposed, but) | |
| 6:41 AM <abadger1999> bcoca - we can't restrict people from doing just what we want them to do; gregdek: yes we can, just ask them not to do things that we don't want them to. | |
| 6:41 AM <jimi|ansible> was Sam DO? I couldn't remember which ones he was working on | |
| 6:41 AM <misc> ho, indeed, sam got a promotion | |
| 6:41 AM <rbergeron> jimi|ansible: i have no idea how / why | |
| 6:41 AM <rbergeron> misc: yeah | |
| 6:42 AM <crab> so people with commit access have to be careful, that's no surprise. | |
| 6:44 AM — misc wonder about saying "move to bugzilla" and see everybody scream at the same time | |
| 6:44 AM <rbergeron> misc: yes, there would be screaming | |
| 6:44 AM <jimi|ansible> from me, especially | |
| 6:44 AM <abadger1999> misc: bcoca liks bz ;-) He'd be screaming Yay! | |
| 6:45 AM <crab> i think the github issues question is peripheral, to be honest | |
| 6:45 AM <abadger1999> <nod> | |
| 6:45 AM <jlk> Moving to bugzilla doesn't automatically get you better bug input | |
| 6:46 AM <abadger1999> it's the landscape we live in... the real question is just - to trust or not to trust. | |
| 6:46 AM <mordred> jlk: ++ | |
| 6:46 AM <crab> the underlying question is whether "outside" people will ever get commit access/trust or not | |
| 6:46 AM <jlk> what github issues are missing are the management of the issues you have, not the input you get | |
| 6:46 AM <crab> abadger1999: exactly | |
| 6:46 AM <misc> jlk: indeed, there isn't any obvious good answer | |
| 6:47 AM <crab> so do i! so does everyone who actually contributes stuff! | |
| 6:47 AM <jlk> crab: right, I'm not saying he's unique, at all | |
| 6:48 AM <jlk> sival is just good at saying "SHUT UP AND TAKE MY CODE" | |
| 6:50 AM *.net ↮ *.split ↔ misc and vincent_vdk nipped out | |
| 6:50 AM <crab> i've lost audio | |
| 6:50 AM <crab> i'll try to reconnect | |
| 6:52 AM <gregdek> ok, sorry | |
| 6:52 AM <crab> i can hear now | |
| 6:53 AM ↔ j00bar nipped out | |
| 6:56 AM <crab> i've turned off my camera to see if that makes audio better (i'm getting only short bursts of audio at a time) | |
| 6:58 AM <rbergeron> crab: can you see things on the tablet camera | |
| 6:59 AM *.net ↮ *.split ⇐ mordred and svg quit | |
| 6:59 AM <crab> rbergeron: yes. | |
| 6:59 AM → svg joined (~svg@hydargos.ginsys.net) | |
| 6:59 AM <j00bar> do we care about archived, searchable history in that chat? | |
| 6:59 AM <crab> thank you so much. | |
| 7:00 AM → mordred joined (~mordred@openstack/mtaylor) | |
| 7:00 AM <misc> speaking of slack after how much we told that github limited us due to the close nature of the tool is kinda curious | |
| 7:00 AM <rbergeron> j00bar: which chat? | |
| 7:00 AM <rbergeron> sorry, i lost context | |
| 7:00 AM <j00bar> #ansible-devel on irc or slack or wherever | |
| 7:03 AM ⇐ newtMcKerr quit (~textual@194.74.0.66) Quit: My Mac has gone to sleep. ZZZzzz… | |
| 7:03 AM <rbergeron> gregdek: commit to making more confersations public somehow. continue the convo on the loose criteria for bringing more poeople into core commit. | |
| 7:04 AM <rbergeron> gregdek: don't want to wait on that. at a high level: time to commit to doing this stuff. | |
| 7:05 AM <crab> thanks! | |
| 7:05 AM <misc> #action item: get lunch | |
| 7:05 AM <rbergeron> crab: we're gonna eat and stuff :) | |
| 7:05 AM <rbergeron> back in an hour. | |
| 7:05 AM <crab> i'll try to join again later, but i have to go do some stuff. | |
| 7:05 AM <rbergeron> crab: ack. i'm gonna turn off my tablet for themoment so it can charge more | |
| 7:07 AM *.net ↮ *.split ⇐ resmo and abadger1999 quit ↔ drybjed and jlk nipped out | |
| 7:37 AM Connection closed unexpectedly | |
| 7:48 AM → gregdek joined ⇐ newtMcKerr quit ↔ j00bar nipped out | |
| 7:59 AM <rbergeron> okay, i think we're starting up again in short order | |
| 7:59 AM → nitzmahone_, abadger1999 and resmo joined | |
| 8:02 AM <gregdek> Hi y'all | |
| 8:03 AM → willthames_ joined (~willthame@194.74.0.66) | |
| 8:03 AM <jlk> Galaxy Team: https://pmcdeadline2.files.wordpress.com/2015/04/galaxyquestcropped.jpg | |
| 8:03 AM willthames_ → willthames | |
| 8:04 AM <gregdek> Topic: Galaxy! | |
| 8:04 AM <gregdek> Open issues at: https://github.com/ansible/galaxy-issues | |
| 8:04 AM <drybjed> hi, what's the current link to the stream? | |
| 8:04 AM <gregdek> Right now working on some issues and will release as 2.0.1 | |
| 8:04 AM <gregdek> https://bluejeans.com/856479228/ | |
| 8:04 AM <drybjed> gregdek: thanks | |
| 8:05 AM <gregdek> 2.1 release ideas: | |
| 8:05 AM <gregdek> * The Gold Star. How do we look at a role and assess it for "all the things that make a role good" and give a role a Gold Star as a result? | |
| 8:05 AM → Shrews joined (~nobody@compute-199-204-45-166.ymq-zone1.thenebulacloud.com) | |
| 8:05 AM <gregdek> We will have a way to give the Gold Star at upload time, and also at local testing time. | |
| 8:06 AM → masteinhauser joined (sid6069@gateway/web/irccloud.com/x-boauihmwxjyaufnp) | |
| 8:06 AM <gregdek> * Role Testing. What infrastructure can we use for this? Can we use containers? | |
| 8:06 AM <gregdek> Currently have lightweight Travis integration available, how can we extend? | |
| 8:06 AM <gregdek> Or do more/better? | |
| 8:07 AM <gregdek> Currently we only have syntax checking, but looking to do maybe Real Testing. | |
| 8:07 AM <gregdek> (i.e. asserts at end of roles.) | |
| 8:07 AM → bcoca joined (~bcoca@194.74.0.66) | |
| 8:08 AM <gregdek> Some use cases: | |
| 8:08 AM <gregdek> (from mordred) | |
| 8:08 AM <gregdek> 1. How to programmatically add a role to galaxy? | |
| 8:09 AM <gregdek> 2. Have official Galaxy roles point to Not Github Repos because everyone doesn't use Github | |
| 8:09 AM <gregdek> 3. A way to mark something as belong to An Organization (tm) | |
| 8:13 AM <jlk> "As a role producer, I want to be able to register a role with Galaxy (the website), that has the source for the role hosted somewhere that is not github, so that I do not have to change my development model to publish to github." | |
| 8:13 AM <gregdek> (from jlk) | |
| 8:14 AM <bcoca> iirc there is already an issue feature request with that | |
| 8:14 AM <bcoca> https://github.com/ansible/galaxy-issues/issues/96 | |
| 8:14 AM <gregdek> Struggling with ways of sharing roles in role requirements files, but galaxy cli tool is good for installing things once | |
| 8:14 AM <gregdek> (i.e. make galaxy installs idempotent) | |
| 8:16 AM <gregdek> Do we have a different directory structure to support multiple versions of roles? | |
| 8:16 AM <gregdek> Original idea: install roles in multiple role paths | |
| 8:17 AM <gregdek> Problem: if the same playbook is dependent on the same named role but has different versions installed (diamond dep problem) | |
| 8:18 AM <gregdek> * Have ansible playbook automatically pull in roles | |
| 8:18 AM <mordred> this is how I run ansible-galaxy: - shell: ansible-galaxy install --force -r roles.yaml | |
| 8:21 AM <gregdek> bcoca has proposed role include as part of standard client | |
| 8:21 AM — svg doesnt like the idea of pulling in roles automagically | |
| 8:22 AM <drybjed> well, imagine the fun you'll have when role has been redesigned ;> | |
| 8:22 AM <willthames> drybjed that's why you specify versions | |
| 8:22 AM <willthames> svg, what is your objection | |
| 8:22 AM <crab> svg: that idea scares me too, but in a restricted environment it might be convenient | |
| 8:23 AM <willthames> it's essential imbo | |
| 8:23 AM <willthames> imho | |
| 8:23 AM <jimi|ansible> the idea was to allow it to be controlled by the user, via cfg options | |
| 8:23 AM <willthames> having to tell people to run ansible-galaxy before running a playbook is insane | |
| 8:23 AM <jimi|ansible> so if you don't want it to download automatically (default off), it won't | |
| 8:23 AM <gregdek> bcoca: have a working group to work on this stuff | |
| 8:23 AM <misc> I also do like mordred but I only use my own role (which doesn't mean it doesn't break, but it is faster to discuss with myself) | |
| 8:23 AM <willthames> but I support a way to turn it on/off | |
| 8:23 AM <svg> I tend to prefer to manage code myself, and not to rely on external hosting | |
| 8:24 AM — rbergeron nods | |
| 8:24 AM <svg> But as long as that is only optional I wouln'd veto it of course :) | |
| 8:24 AM <gregdek> ACTION! bcoca set up working group to sort out versioning and download | |
| 8:24 AM <jlk> We "vendor" all our roles too | |
| 8:24 AM <jlk> and use Galaxy (the website) as inspiration. | |
| 8:24 AM <mordred> misc: https://github.com/openstack-infra/system-config/blob/master/roles.yaml | |
| 8:24 AM <mordred> misc: yah - me too :) | |
| 8:24 AM <rbergeron> i also wonder if we need to be more explicit about "what is galaxy for vs. not for" | |
| 8:24 AM <bcoca> svg, its optional and requires the origin to be whitelisted in config | |
| 8:24 AM <rbergeron> i mean people will do whatever they want with it :) but | |
| 8:25 AM <svg> lgtm | |
| 8:25 AM <gregdek> Galaxy is the Ansible role package manager. | |
| 8:25 AM <gregdek> :) | |
| 8:26 AM sivel_ → sivel | |
| 8:26 AM <gregdek> HI SIVEL | |
| 8:26 AM <gregdek> good morning :) | |
| 8:27 AM <sivel> hey! | |
| 8:27 AM <jlk> a wild sivel appears | |
| 8:28 AM <sivel> only here for another 15 minutes, then I have to take my wifes car to the body shop, then I'll be back | |
| 8:28 AM <sivel> fortunately you are talking about galaxy, and I have 0 input here | |
| 8:28 AM <gregdek> bye :) | |
| 8:29 AM <jimi|ansible> -s API_SERVER, --server=API_SERVER | |
| 8:29 AM <jimi|ansible> The API server destination | |
| 8:29 AM <jimi|ansible> ^ this has been a flag for ansible-galaxy since I wrote it | |
| 8:29 AM <jimi|ansible> we had always had the idea internally that we would have private galaxy servers | |
| 8:30 AM <vincent_vdk> can you have something in galaxy like prive repos in github? | |
| 8:31 AM <rbergeron> so the question is: who will own the beginnings of figuring out the open sourcing of galaxy and driving that down the road? | |
| 8:32 AM <gregdek> ACTION! gregdek will own the "figuring out open source galaxy" question. | |
| 8:32 AM <gregdek> s/galaxy/galaxy-server/ | |
| 8:32 AM <gregdek> ACTION! robyn will actually own the "figuring out open source galaxy" question. | |
| 8:32 AM <rbergeron> and will make sure to not make house drown in misery :) | |
| 8:33 AM <gregdek> vincent_vdk: the answer will very likely be "yes" once galaxy server is open source. | |
| 8:37 AM <gregdek> jimi|ansible says that galaxy cli needs to be cleaned up to be reusable, so that code can be accessed as libraries etc | |
| 8:38 AM <vincent_vdk> when I look at the galaxy docs, you can actually doe something with versioned roles? -> username.rolename[,version] | |
| 8:38 AM — willthames agrees, but with the caveat that we don't break galaxy for months on end again | |
| 8:38 AM <gregdek> "Do something" -- meaning "you can install a particular sha in git"? | |
| 8:39 AM <gregdek> bcoca and jimi|ansible are talking with willthames about his PR that does some of this stuff :) | |
| 8:39 AM <willthames> jimi|ansible tree-ish versions: https://github.com/ansible/ansible/pull/13203 | |
| 8:39 AM <rbergeron> so I think we ahve actions on this topic, yes? :) | |
| 8:40 AM <misc> +1 for treeish version | |
| 8:40 AM <willthames> vincent_vdk: http://willthames.github.io/2014/09/03/techniques-for-versioning-ansible-ii.html | |
| 8:40 AM <gregdek> NEXT: issue triage! | |
| 8:40 AM <rbergeron> gregdek: we should have a higher bar for what we require out of an issue. | |
| 8:40 AM <vincent_vdk> willthames: thx, interesting | |
| 8:41 AM <rbergeron> gregdek: not because we want people to give up, but because we have such a high volume of issues. | |
| 8:41 AM <rbergeron> "an issue should be ignored until it meets the requirement, or is filled out with this form, etc" | |
| 8:41 AM <crab> +1 for auto-closing and encouraging re-opening rather than leaving in "pending" state | |
| 8:41 AM <rbergeron> and once we have that -- then we can have workflow. | |
| 8:41 AM <jimi|ansible> willthames: merged | |
| 8:41 AM <willthames> jimi|ansible hooray, contributor summit was a success for me :) | |
| 8:41 AM ⇐ resmo quit (~resmo@194.74.0.66) Ping timeout: 265 seconds | |
| 8:42 AM <svg> willthames: Where are you seated? | |
| 8:42 AM <rbergeron> https://goo.gl/LhEGBn | |
| 8:43 AM <crab> no objections | |
| 8:44 AM <crab> but i think it's only part of the answer. we should also be encouraging people to investigate valid open issues. | |
| 8:44 AM <jimi|ansible> pending_closure <- the tag | |
| 8:45 AM <crab> i feel that's an important... step, or something, to increase the contributor pool | |
| 8:45 AM → newtMcKerr joined (~textual@194.74.0.66) | |
| 8:45 AM <gregdek> crab: yes, a big +1. | |
| 8:45 AM <abadger1999> crab: yeah | |
| 8:45 AM <crab> so maybe the core developers can say "this looks like a problem with x/y.py" and otherwise provide useful hints to people | |
| 8:46 AM <sivel> github should do this, but I have been working on something (very minimally) that could detect other similar issues, and link them, potentially via the CI mechanism in github | |
| 8:46 AM <sivel> but absolutely no promises on ever actually producing something | |
| 8:47 AM <crab> hm, inaudible mumbling. must be robyn :-) | |
| 8:48 AM <gregdek> "Do we create a triage project of some kind?" | |
| 8:49 AM ⇐ newtMcKerr quit (~textual@194.74.0.66) Client Quit | |
| 8:49 AM <willthames> http://docs.ansible.com/ansible/community.html doesn't mention triage at all | |
| 8:49 AM <sivel> I try to do triage all the time | |
| 8:49 AM <sivel> also, for anyone who needs it, https://ansible.sivel.net/pr/ | |
| 8:49 AM <sivel> that helps with doing triage | |
| 8:49 AM <gregdek> How do we make sivel's awesome tool an official thing that everyone knows about and uses? | |
| 8:50 AM <sivel> haha | |
| 8:50 AM <jlk> sival is now a tool. | |
| 8:50 AM <abadger1999> sivelity | |
| 8:50 AM <rbergeron> willthames: well, that may be because nobody else ... has permissions, at the moment, or any structure by which we can ... automatically make things that certain people trigger actions or github state changes, etc. | |
| 8:50 AM <gregdek> ACTION! gregdek will figure out how we make sivel's excellent pr/issue tool a more well known part of the ansible workflow | |
| 8:50 AM <misc> an-sival | |
| 8:51 AM <sivel> it also exists at https://github.com/sivel/pr-triage | |
| 8:51 AM <rbergeron> misc: sooooo bad | |
| 8:51 AM <misc> rbergeron: thanks, I am quite proud of that | |
| 8:51 AM <crab> no, i don't know anyone who loves doing triage. but that's sort of the problem i'm saying we should try to solve. | |
| 8:52 AM <crab> not just triage, but also code review. i mean, when i started out i didn't even know python, let alone ansible, but there was still lots of low-hanging fruit to review. | |
| 8:52 AM → newtMcKerr joined (~textual@194.74.0.66) | |
| 8:54 AM <crab> i think the idea of looking together at old issues in community meetings is great. | |
| 8:54 AM <rbergeron> crab: agreed | |
| 8:54 AM <willthames> the community page also doesn't mention code review | |
| 8:54 AM <bcoca> those work pretty well (when we did them) | |
| 8:54 AM <rbergeron> willthames: indeed | |
| 8:54 AM <crab> better developer documentation! \o/ | |
| 8:55 AM <crab> architecture diagrams! \o/ | |
| 8:56 AM <crab> actually, i can work on that. | |
| 8:56 AM <svg> better dev docs would also motivate more people via meetups | |
| 8:56 AM <rbergeron> crab: which part :) | |
| 8:56 AM <crab> architecture docs | |
| 8:56 AM <gregdek> ACTION: gregdek will set up a docs sprint led by docschick and involving crab and core folks | |
| 8:56 AM <rbergeron> svg: explain a bit more? | |
| 8:57 AM <svg> afaik, most ansible meetups are focusing on users, less on code | |
| 8:57 AM <crab> i have been paying attention to what jimi-c rewrote after all ;-) | |
| 8:57 AM <jimi|ansible> that's been the focus of most public things - using ansible rather than extending it | |
| 8:57 AM <jimi|ansible> ^ it's why i'm doing the talk i'm doing tomorrow | |
| 8:57 AM <crab> what talk are you doing tomorrow? | |
| 8:58 AM <jimi|ansible> writing modules | |
| 8:58 AM <masteinhauser> I can only speak from Ansible Boston meetup, and lots of more advanced users who *are* interested in leveling into developer and extension, but by far mostly just user-focused. | |
| 8:58 AM <svg> jimi|ansible: I thought you were going to tell a joke on screwing lightbulbs | |
| 8:58 AM <bcoca> sounds like i'll have to drop by and do that presentation .... | |
| 8:59 AM <bcoca> svg: hes is doing both | |
| 8:59 AM <jimi|ansible> they don't screw here, B22 just turns a little | |
| 8:59 AM <rbergeron> okay, greg is stepping out. and i'm going to buy a shitton of coffee. | |
| 9:00 AM <Shrews> Are any of the talks being live-streamed? | |
| 9:00 AM <bcoca> idk | |
| 9:00 AM <jimi|ansible> i don't believe so, we're going to record them, but assuming it'll be the same as SF where we record the slides output and audio over that | |
| 9:00 AM <jimi|ansible> not really the full stage | |
| 9:02 AM <bcoca> they are being recorded | |
| 9:08 AM <sivel> I just replicated the 'pending_closure' label to -core and -extras | |
| 9:09 AM <gregdek> sivel: we already have pending_action? | |
| 9:09 AM <sivel> was creating pending_closure so we could tag those issues we wanted to close | |
| 9:09 AM → resmo joined ⇐ newtMcKerr quit | |
| 9:10 AM <sivel> I feel like pending_action is too broad, but didn't realize the bot was the one using it | |
| 9:10 AM <jimi|ansible> sivel: go ahead and rename pending_closure -> pending_action on ansible/ansible | |
| 9:10 AM <sivel> jimi|ansible: got it | |
| 9:11 AM <jimi|ansible> i've noticed since using waffle.io that there are a few discrepencies there | |
| 9:11 AM → newtMcKerr joined (~textual@194.74.0.66) | |
| 9:11 AM <sivel> pending_closure updated to pending_action on ansible/ansible | |
| 9:11 AM <jimi|ansible> sivel: danke | |
| 9:12 AM <abadger1999> gregdek: there's issues with new modules in core and extras but to bite off something we can chew, let's concentrate on extras | |
| 9:13 AM <abadger1999> gregdek also says "We | |
| 9:13 AM <abadger1999> 're currently not clear on the split for users because pip intal ansible gets both | |
| 9:14 AM <abadger1999> but eventually we want extras to be a separate package so that people know who is supporting what. | |
| 9:14 AM <sivel> we will also be working on moving https://github.com/sivel/ansible-testing into the ansible org soon, and extending to core. I'd like to keep extending this, specifically into coding standards too | |
| 9:15 AM <sivel> hahaha | |
| 9:15 AM <abadger1999> jlk: "what constitutes correct?" gregdek: "define correctness and implement an automated linter to check it" | |
| 9:15 AM <abadger1999> gregdek: "sivel has linter at ansble-testing" | |
| 9:15 AM <crab> i agree with what dagwieers wrote on the etherpad: there are design questions involved in whether or not to merge a new module, not just syntactical correctness. | |
| 9:15 AM <abadger1999> bcoca: "current ansible-testing linter doesn't go far enough" | |
| 9:16 AM <rbergeron> coffee is on its way up | |
| 9:16 AM <rbergeron> and of course now they are setting up coffee outside the door. lol | |
| 9:16 AM <abadger1999> bcoca: "example: currently module exists that does basically the same thing as extras does" | |
| 9:17 AM <abadger1999> jlk: "Sounds like we need a human but they aren't necessarily checking that a module does the right thing" | |
| 9:20 AM <jimi|ansible> if nothing, abadger1999 has sold me on the name "waterfountain" for a new project | |
| 9:20 AM <bcoca> i want puddlejumper | |
| 9:22 AM → chouseknecht joined (uid132661@gateway/web/irccloud.com/x-pzjhkphveqowgloc) | |
| 9:23 AM <svg> How about disallowing the module: or action: shortcut again? | |
| 9:23 AM — svg ducks | |
| 9:23 AM <willthames> sivel, new linter feature: name should not conflict with other task arguments (become for example) | |
| 9:23 AM <drybjed> svg: action: setup, action: flush_handlers | |
| 9:23 AM <jimi|ansible> action: meta flush_handlers | |
| 9:24 AM <willthames> meta: flush_handlers, and just setup: alone | |
| 9:24 AM <bcoca> action: module=<modname> | |
| 9:24 AM <drybjed> ah, right | |
| 9:24 AM <svg> I meant the hortcut where 'action:' is being left out :D | |
| 9:24 AM <sivel> willthames: https://github.com/sivel/ansible-testing/issues/17 | |
| 9:27 AM <willthames> sivel: nice :) | |
| 9:27 AM <willthames> svg, I think that would break everybody's playbooks everywhere by now! | |
| 9:27 AM <svg> willthames: that was an old argument a couple of years ago :) | |
| 9:28 AM <crab> what project is waterfountain a name for? | |
| 9:28 AM <vincent_vdk> what about ditching key=value and start using plain YAML? | |
| 9:28 AM <gregdek> it is a theoretical project :) | |
| 9:28 AM <gregdek> ACTION: gregdek will set up the Extras New Module review process and ask for volunteers :) | |
| 9:28 AM <jimi|ansible> vincent_vdk: that is something we've talked about making an official best practice at the very least | |
| 9:28 AM <jimi|ansible> ditching that would break even more things than only allowing action: module_name | |
| 9:28 AM <vincent_vdk> jimi|ansible: ok | |
| 9:28 AM <abadger1999> crab: just needed to make up a pretend name to be a sstandin for "the next big thing we need a module for like docker today" | |
| 9:29 AM <sivel> we could start adding "DEPRECATED" messages in 2.1 :) | |
| 9:29 AM <rbergeron> COFFFEE IN THE BACK OF THE ROOM YO | |
| 9:29 AM <rbergeron> GO DRINK IT | |
| 9:29 AM <sivel> who is bringing me coffee? | |
| 9:29 AM <crab> ok! | |
| 9:29 AM <gregdek> sivel: expense it ;) | |
| 9:30 AM <misc> ansible all -i sivel, -m coffee -a present=yes | |
| 9:30 AM <drybjed> sivel: feeling adventurous? ;) | |
| 9:30 AM <sivel> I have coffee here, just haven't made it yet | |
| 9:31 AM <bcoca> who wants to be on the board? | |
| 9:32 AM <crab> again, i agree with dagwieers. we need a board to discuss and document design decisions. for example i think it's a real mistake to not make the hostname be a hostname in the inventory. | |
| 9:32 AM <crab> of course misc believes the exact opposite, possibly as strongly | |
| 9:33 AM <drybjed> crab: what is it about? | |
| 9:33 AM <gregdek> crab: since we have board discussion explicitly in 30m, I'll wait until then so other people can join | |
| 9:33 AM <gregdek> Now it's just sort of free-form talking about stuff :) | |
| 9:33 AM <gregdek> Right now, ansible-lint | |
| 9:34 AM <crab> drybjed: is "/foo/bar ansible_user=xxx" a valid line in an inventory? | |
| 9:34 AM <crab> gregdek: understood | |
| 9:34 AM <rbergeron> we basically have 15m until our break | |
| 9:34 AM <rbergeron> the actual coffee break :) | |
| 9:34 AM <crab> ansible-ssh! | |
| 9:34 AM <drybjed> crab: hmm, not sure what '/foo/bar' would be... | |
| 9:34 AM <misc> drybjed: crab point is "no", because we want to be more stringent (and I agree that checking is good) | |
| 9:35 AM <misc> drybjed: my point is that for non ssh connexion plugin, that's making them harder to use | |
| 9:35 AM <drybjed> misc: I see | |
| 9:35 AM <misc> I am sure we can find a way that preserve the check without making it harder however | |
| 9:35 AM <misc> cause indeed, things should be verified in advance as it prevent bugs | |
| 9:36 AM <misc> but we do not know is a waterfountain future plugin might requires, so I think flexibility could be required (but also, as we discussed on github, it might not be easy to do) | |
| 9:37 AM <sivel> alright, I have to head out. was supposed to be gone like 45 minutes ago. hopefully back in an hour | |
| 9:38 AM <bcoca> later | |
| 9:38 AM <crab> i agree with bcoca(?), i rewrote ansible-ssh to use the new cli stuff and it's pretty convenient (i have to do some contortions to add options, but that's understandable) | |
| 9:39 AM <crab> i am happy with ansible-ssh being an external thing. i mean, i'm maintaining it that way already. | |
| 9:39 AM <abadger1999> crab: yeah, that's bcoca talking about that (right behind me) | |
| 9:39 AM <bcoca> ansible-contrib and add authors as commiters there? | |
| 9:40 AM <rbergeron> crab: so how can we (or should we) be pointing people to it | |
| 9:41 AM <crab> i don't have any good answer | |
| 9:41 AM <rbergeron> or clarify that "there are tools, and hwere are some useful tools" | |
| 9:41 AM <crab> curated list sounds ok, but usually such things don't work too well | |
| 9:41 AM <gregdek> ACTION: documentation addition (getting started) of our "best friend" tools | |
| 9:41 AM <crab> "why didn't you list ansible-watermountain?!?!?!? YOU GUYS SUX!" | |
| 9:41 AM <gregdek> crab: yes, sadly, this will be an issue. :) | |
| 9:42 AM <bcoca> crab: send a pr | |
| 9:42 AM <rbergeron> yeah. then it becomes "what happens when maintainer of $tool moves on and it's not up to date", etc. | |
| 9:42 AM → ahaerpfer joined (~ahaerpfer@atlantic.v6x.org) | |
| 9:42 AM <rbergeron> but: maybe better than nothing and if it becomes problematic then .... we revisit | |
| 9:42 AM — svg wonders if Debian would have to rename watermountain to iceberg | |
| 9:42 AM <gregdek> now we're just talking rot | |
| 9:42 AM <jimi|ansible> yaml.safe_load("---\non: true") | |
| 9:43 AM <abadger1999> < ></ > < > </ > | |
| 9:43 AM <rbergeron> but someone has to be responsible for just curating otherwise ... it all gets sad. | |
| 9:43 AM <rbergeron> just like a wiki. | |
| 9:43 AM <crab> i would love to have the ansible core code in such a state that it's practical to have some code to build a Playbook object from JSON input and then have it just run as usual | |
| 9:43 AM <jimi|ansible> {True: True} | |
| 9:43 AM <rbergeron> "where infomation kills itself" | |
| 9:43 AM <jimi|ansible> yaml.safe_load("---\n!!str on: true") | |
| 9:43 AM <jlk> WAT | |
| 9:43 AM <misc> rbergeron: autocurated, if you do not update, it disappear itself, so people would have to make a PR on a regular basis ? | |
| 9:44 AM <crab> misc: haha, that is insane but i love it. | |
| 9:44 AM <rbergeron> misc: oh, yeah. :) | |
| 9:44 AM <jlk> that's right up there with NaNaNaNaNaN | |
| 9:44 AM <rbergeron> okay, it's break time for the next 15 minutes | |
| 9:44 AM <bcoca> !!str on: true | |
| 9:44 AM <svg> crab: starts sounding like ansible as a lib | |
| 9:44 AM <misc> ansilibrary | |
| 9:45 AM <misc> gosh, no more pastry :( | |
| 9:45 AM <rbergeron> and talking about technical board, and open floor, when we come back at 1500 UTC | |
| 9:45 AM <bcoca> crab: you currently CAN create a json playbook and it will run | |
| 9:45 AM <crab> svg: yeah, it's a step | |
| 9:45 AM <bcoca> crab: you currently CAN create a json playbook and it will run | |
| 9:45 AM <bcoca> crab: you currently CAN create a json playbook and it will run | |
| 9:45 AM <crab> bcoca: wow, i didn't know that | |
| 9:45 AM <crab> bcoca: wow, i didn't know that | |
| 9:45 AM <rbergeron> omg | |
| 9:45 AM <misc> is there tests for that ? | |
| 9:45 AM <rbergeron> omg | |
| 9:45 AM <rbergeron> :) | |
| 9:45 AM <bcoca> sorry, enter did not seem to work | |
| 9:45 AM <svg> yeah afaik all yaml can be json everywhere iirc | |
| 9:45 AM <bcoca> json == subset of yaml | |
| 9:45 AM <bcoca> so anaywhere you use yaml you can use json | |
| 9:46 AM <svg> yes | |
| 9:46 AM <crab> ! | |
| 9:46 AM <svg> been so since a while even | |
| 9:46 AM <crab> i can do: copy: {"src":"foo","dest":"bar"} and such? | |
| 9:46 AM <bcoca> yes | |
| 9:47 AM <crab> my head just exploded again | |
| 9:48 AM <mordred> crab: where is this ansible-ssh thing of yours people are telling me about? | |
| 9:48 AM → nirik joined (~nirik-fre@75.148.32.185) | |
| 9:49 AM <crab> mordred: the current version isn't anywhere because i got busy. i'll upload to 2ndQuadrant/ansible-ssh this weekend. | |
| 9:49 AM <crab> there's already something there, but that's old and sucks. | |
| 9:50 AM <crab> mordred: what's there is a slightly outdated version that just does exec("ssh <args that ansible would use> hostname") | |
| 9:50 AM <jlk> ansible-ssh drops you into a shell? | |
| 9:50 AM <jimi|ansible> crab: you've never used YAML short form? | |
| 9:51 AM <mordred> crab: neat | |
| 9:51 AM <crab> mordred: what i almost have now is something that can do: ansible-ssh -sK somehost and it can get the sudo password from a vault'ed inventory file and do privilege escalation for you and then give you a shell prompt | |
| 9:51 AM <crab> mordred: (and since i use ansible-vault with gpg, i don't ever have to type or cut and paste any passwords ever) | |
| 9:52 AM <crab> jimi|ansible: i didn't even know JSON was a YAML short form | |
| 9:52 AM <bcoca> crab: we are renaming to ansible-login and adding winrm support .... | |
| 9:53 AM <bcoca> crab: diff standards, but since they are based on common data structures in langauges, it just 'happened' that way | |
| 9:53 AM <bcoca> yaml has had 'shorthand for a long time' | |
| 9:53 AM <crab> bcoca: are you pulling my leg? | |
| 9:56 AM <bcoca> crab: 1/2 ... im sitting by the windoez guyz | |
| 9:56 AM <crab> how does winrm work? is it like ssh, where you run winrm somehost somecommand, or completely different? | |
| 9:57 AM <crab> and to do privilege escalation do you just change somecommand, or do you have to do something horrible? | |
| 9:58 AM <bcoca> pe does not currently wokr with winrm | |
| 9:58 AM <bcoca> it is a soap protocol that allows you to run powershell | |
| 9:59 AM — crab runs away screaming | |
| 10:00 AM <bcoca> crab: as you should | |
| 10:00 AM <crab> bcoca: you sound very different than i'd imagined | |
| 10:00 AM <bcoca> i sound very different than i imagine too | |
| 10:01 AM <nitzmahone_> he probably looks different than you imagine too. ;) | |
| 10:01 AM <crab> nitzmahone: i haven't actually been able to see how he looks because he was sitting behind abadger1999 | |
| 10:01 AM <crab> does james laska still work for ansible? | |
| 10:02 AM <jlk> crab: yes | |
| 10:02 AM <crab> jlk: do you work for ansible too? | |
| 10:03 AM <jlk> crab: no, but I used to work at Red Hat with a lot of these people. | |
| 10:03 AM <jlk> crab: I'm at Blue Box, an IBM Company. | |
| 10:03 AM <crab> ah | |
| 10:04 AM <misc> and do you have to wear a blue box on your head ? | |
| 10:05 AM <crab> who's currently taking a photo from behind gregdek's laptop? | |
| 10:05 AM <crab> well, a few seconds ago | |
| 10:05 AM <bcoca> noone | |
| 10:05 AM <crab> i see. | |
| 10:06 AM <gregdek> OK LET'S GET STARTED AGAIN | |
| 10:06 AM <crab> where's the gong? | |
| 10:06 AM <gregdek> that wasn't helpful | |
| 10:06 AM <bcoca> grg ate it | |
| 10:08 AM <gregdek> Current "governance": argue on IRC :) | |
| 10:08 AM <gregdek> At random times, with whoever happens to be around, and documented in our heads and not on paper | |
| 10:09 AM <gregdek> Once a decision gets made, record the decision | |
| 10:09 AM <willthames> irc is not sufficient as it's not recorded | |
| 10:09 AM <misc> it could be | |
| 10:09 AM <gregdek> It certainly could be. | |
| 10:09 AM <jlk> misc: I do not have to wear a blue box :) | |
| 10:10 AM <willthames> ansible-devel emailing list has a little more persistence | |
| 10:10 AM <misc> but most of the time, irc logs are hard to read. not sure who check meeting minutes, but each time i try, it is annoying | |
| 10:10 AM <crab> yeah, should hostnames contain spaces. ugh. | |
| 10:10 AM <gregdek> ACTION: have a bot on #ansible-devel (and #ansible-meeting) | |
| 10:10 AM <gregdek> misc: it is imperfect, but at least a bot allows some mitigation of that imperfection. | |
| 10:10 AM <drybjed> crab: shouldn't hostnames be basically compliant with DNS naming rules? | |
| 10:10 AM <crab> drybjed: yes! yes they should! | |
| 10:11 AM <misc> and you also need to be able to point to a decision, but I fear that in the end, we will end like openstack, and mordred will just throw something at me :) | |
| 10:11 AM → twaldrop joined (uid146940@gateway/web/irccloud.com/x-qftbpbffgefjwgzx) | |
| 10:11 AM <jimi|ansible> crab / drybjed: but we also allow things like /some/path/to/file for chroot hosts | |
| 10:11 AM <jimi|ansible> so that's definitely not compliant | |
| 10:12 AM → tksarah joined (~yaaic@123.76.239.49.rev.vmobile.jp) | |
| 10:12 AM <mordred> misc: I will do what I can to prevent that :) | |
| 10:13 AM <crab> jimi|ansible: personally, i think we shouldn't allow it. istm that a host is a fundamental abstraction in ansible, so if it happens to have some weird connection string that isn't a hostname, it should be set some other way. | |
| 10:13 AM <jimi|ansible> i don't disagree, i think the path should be contained in ansible_chroot_host and the hostname should be a compliant hostname | |
| 10:13 AM <jimi|ansible> but we allow it today so we'd have to deprecate it and move in that direction | |
| 10:13 AM <crab> jimi|ansible: but the current situation is the worst of both worlds, it basically does: if valid: add else: add anyway | |
| 10:13 AM <jimi|ansible> right | |
| 10:14 AM <bcoca> https://c1.staticflickr.com/1/1/8850_143dbaf5d4.jpg | |
| 10:15 AM <gregdek> potential agenda: | |
| 10:15 AM <gregdek> (repeating) | |
| 10:15 AM <crab> ponies don't have enough legs. octopodes are better. | |
| 10:16 AM <misc> octoponies | |
| 10:16 AM <gregdek> * Big Issues that have stacked up in the last week | |
| 10:16 AM <gregdek> * Open Modules (or something( | |
| 10:16 AM <gregdek> * etc. | |
| 10:16 AM <gregdek> And have a consistent format for this meeting every week | |
| 10:16 AM <crab> another potential "technical board" subject: looping over blocks | |
| 10:17 AM <crab> i think lots of people agree that we want it, but right now we don't know how to get there | |
| 10:17 AM <nitzmahone_> crab++ | |
| 10:17 AM <jimi|ansible> i have an idea for it | |
| 10:17 AM <willthames> crab, I think process is write an AEP, and then can discuss what it would look like and what would need to change to get there | |
| 10:17 AM <willthames> you could write AEP-001 | |
| 10:17 AM <jlk> kinda like OpenStack blueprints / specs | |
| 10:18 AM <jlk> they get PRd to the repo, so they can go through a review process with iterations. | |
| 10:18 AM <jimi|ansible> block+with_* -> create a Task and make block.task_include = that new task, then insert that task into the task list rather than the block | |
| 10:19 AM <masteinhauser> keeping track during the week seems like another item that the bot might be able to keep track of within irc logs too | |
| 10:19 AM <crab> jimi|ansible: geez, that's evil and would probably work fine | |
| 10:19 AM <jimi|ansible> and reassign the loop stuff to the new task | |
| 10:20 AM <nitzmahone_> would likely require the "name the loop var" construct that's been bandied about | |
| 10:21 AM <crab> yes | |
| 10:21 AM <nitzmahone_> (or at least "really benefit from") | |
| 10:21 AM <jimi|ansible> probably, because people wouldn't expect includes inside the block to break | |
| 10:21 AM <crab> +1 for blueprints or even plans | |
| 10:22 AM <gregdek> jlk sez "let's use PRs for text file proposals" | |
| 10:22 AM <willthames> ansible/ansible docs/proposals or something | |
| 10:23 AM <jlk> label the PR as "blueprint" or "proposals" ? | |
| 10:23 AM <jimi|ansible> AnsiPEP | |
| 10:24 AM <crab> PRs for text file proposals in a directory named proposals sounds fine | |
| 10:25 AM <gregdek> AGREED: for new proposals that are not yet "code-able", requester will add a PR with a textfile to ansible/ansible/docs/proposals | |
| 10:27 AM → geerlingguy joined (~geerlingg@pdpc/supporter/active/geerlingguy) | |
| 10:27 AM <jlk> "As a proposal reviewer, I want proposal PRs to never rebase, so that I can see the entire history of the proposal iterations." | |
| 10:27 AM <geerlingguy> Did I miss anything? :D | |
| 10:27 AM <bcoca> proposal/ => final/ | |
| 10:27 AM <gregdek> geerlingguy: only the whole day ;) | |
| 10:28 AM <gregdek> (will post the full log of IRC later) | |
| 10:28 AM <geerlingguy> haha, timezones-- | |
| 10:28 AM <geerlingguy> also huge projects wrapping up today that I had to stay up late for yesterday-- | |
| 10:28 AM ⇐ j00bar quit (~j00bar@194.74.0.66) Quit: j00bar | |
| 10:30 AM <crab> +1 to everything gregdek just said in summary | |
| 10:31 AM <abadger1999> jlk: mic on | |
| 10:31 AM <jlk> crap. | |
| 10:32 AM <gregdek> to recap: | |
| 10:32 AM <gregdek> * weekly meetings to address Srs Proposals | |
| 10:32 AM <crab> who's speaking now? | |
| 10:32 AM <gregdek> * give Srs Proposals real feedback, even if it's quick | |
| 10:32 AM <bcoca> james | |
| 10:32 AM <gregdek> (jimi-c) | |
| 10:32 AM <crab> oh, jimi-c | |
| 10:32 AM <gregdek> * Even if it's no, avoid the Slow No | |
| 10:33 AM <gregdek> OK, moving on: | |
| 10:33 AM <gregdek> jimi-c is talking about considering the Ansible Lieutenant Model | |
| 10:34 AM <gregdek> Some subsystems: | |
| 10:34 AM <gregdek> * Plug-ins or family of plug-ins | |
| 10:34 AM <gregdek> * Inventory | |
| 10:34 AM <gregdek> * Playbook parser | |
| 10:34 AM <gregdek> * Executor engine | |
| 10:34 AM <gregdek> * Galaxy | |
| 10:35 AM <gregdek> jlk: challenge the concept. If you're going to trust someone to merge a subsystem, why not just give them commit directly? | |
| 10:35 AM <crab> i agree with jlk | |
| 10:38 AM — svg likes inventory :) | |
| 10:38 AM <Shrews> linus just curses at everyone. please don't follow his model | |
| 10:38 AM <gregdek> me: we could use the lieutenant model that we essentially use for PRs in modules to do ansible/ansible PRs: each file has one or more "maintainer/lieutenant", that lieutenant has the ability to say "shipit", all shipits ultimately hit the core team for final review | |
| 10:38 AM <svg> we don't curse, we speak yaml | |
| 10:38 AM <gregdek> svg: same thing | |
| 10:38 AM <geerlingguy> lol | |
| 10:39 AM <bcoca> dammit ... i was looking forward to the cursing and belittleing | |
| 10:39 AM <willthames> gregdek I would like reviewers in addition | |
| 10:39 AM → defionsc_ joined (~defionsco@194.74.0.66) | |
| 10:39 AM <abadger1999> bcoca: I still have a mute button :-P | |
| 10:40 AM <abadger1999> don't make me use it on you | |
| 10:40 AM <bcoca> what? | |
| 10:40 AM <crab> honestly, i still don't understand why giving people commit access is so scary | |
| 10:40 AM <gregdek> ACTION: gregdek will send a detailed instruction guide for ansibullbot and that it actually exists | |
| 10:40 AM <bcoca> commit access is not scary, people are | |
| 10:40 AM <jimi|ansible> +1 | |
| 10:41 AM <bcoca> in teh past we've had people move the project w/o consulting the rest | |
| 10:41 AM <willthames> bcoca we can use social conventions to address that | |
| 10:41 AM <willthames> changes can be reverted | |
| 10:41 AM <misc> we can do like openshift, remove commit access to everybody but the bit | |
| 10:41 AM <willthames> and commit access can be revoked | |
| 10:41 AM <misc> bot | |
| 10:41 AM <bcoca> ^ not saying we should still be that wary, just explaining the history | |
| 10:41 AM <crab> bcoca: that sounds like a good way to make people mad and get thrown out | |
| 10:42 AM <bcoca> crab: exactly what happened | |
| 10:42 AM <Shrews> good testing makes commit access less scary | |
| 10:42 AM <rbergeron> shrews: ++++++++ | |
| 10:43 AM → j00bar joined (~j00bar@194.74.0.66) | |
| 10:43 AM <svg> Would we allow direct commit access or only the power to merge? | |
| 10:43 AM <misc> now, I think commit access should be temporary, like just for 1 year and removed by default unless you ask it again, so we avoid zombies maintainers with unprotected ssh keys (no, not what happened in some project I work on, and I didn't do a talk on it last week) | |
| 10:43 AM <bcoca> svg: github does not make distinction | |
| 10:43 AM <crab> svg: same thing | |
| 10:43 AM <bcoca> but the 'social convention' might direct one way or another | |
| 10:43 AM <svg> I know | |
| 10:44 AM → tksarah2 joined ⇐ tksarah quit | |
| 10:45 AM <crab> i think forced pushes can just be disabled | |
| 10:45 AM <bcoca> yes, they can | |
| 10:45 AM <gregdek> ACTION: we need a "thou shalt not" list for core contributors | |
| 10:45 AM <defionsc_> gregdek yep | |
| 10:47 AM — crab is crushed under the weight of social conventions | |
| 10:47 AM <bcoca> ^ all humanity | |
| 10:47 AM <gregdek> LOL | |
| 10:49 AM <sivel> We should look into the functionality provided by protected branches | |
| 10:49 AM <sivel> https://github.com/blog/2051-protected-branches-and-required-status-checks | |
| 10:49 AM <sivel> Something like what lgtm.co does | |
| 10:49 AM <geerlingguy> "mumbling around the room" | |
| 10:50 AM — svg likes th eidea of protected branches | |
| 10:50 AM <crab> what about "more cows"? | |
| 10:50 AM <jlk> Y'all should be doing branch protection, absolutely | |
| 10:50 AM <misc> crab: +1 | |
| 10:50 AM <geerlingguy> Was Galaxy covered earlier? Just wondering | |
| 10:50 AM <jlk> it won't solve everything, but yeah. | |
| 10:50 AM <jlk> geerlingguy: yes | |
| 10:50 AM <geerlingguy> will read logs after the fact :) | |
| 10:50 AM <crab> yes! introductions! | |
| 10:50 AM <geerlingguy> Dangit | |
| 10:50 AM <geerlingguy> Woohoo! | |
| 10:51 AM <geerlingguy> community summits++ | |
| 10:51 AM <Shrews> ++ as well | |
| 10:52 AM <geerlingguy> lol, this is an interesting perspective | |
| 10:53 AM <geerlingguy> Can you show more than four video displays at a time in BlueJeans? | |
| 10:54 AM <geerlingguy> Er, I guess there are only four | |
| 10:54 AM <abadger1999> Everyone at the table is famous by association with jlk ;-) | |
| 10:54 AM <geerlingguy> uh oh, muted | |
| 10:55 AM <jlk> lol | |
| 10:55 AM <jlk> I haven't done anything relevant in years, I just bring cool people to the table | |
| 10:56 AM <geerlingguy> ...and video's back! | |
| 10:56 AM <crab> bcoca killed the video feed from the tablet | |
| 10:57 AM <misc> geerlingguy: it can go up to 9, iirc | |
| 10:58 AM <geerlingguy> misc: thanks; never used bluejeans before, looks decent | |
| 10:59 AM <Shrews> thx all. have fun at the social | |
| 10:59 AM <geerlingguy> ++ wish we could be there! | |
| 10:59 AM <jlk> Max Headroom | |
| 11:00 AM <Shrews> here, but no mic! | |
| 11:00 AM <crab> thanks to everyone, and especially to gregdek and robyn for setting this meeting up. | |
| 11:01 AM <geerlingguy> See you guys! | |
| 11:01 AM <rbergeron> crab: glad you were able to join us remotely :) | |
| 11:01 AM <crab> it's weird to see gregdek's arm on his camera and also disembodied on robyn's camera | |
| 11:02 AM <svg> it's weird to see gregdek | |
| 11:02 AM <crab> it's weird to see | |
| 11:02 AM <gregdek> crab, geerlingguy, Shrews -- thanks for joining us remotely :) | |
| 11:02 AM <geerlingguy> sure thing, just wish I could've joined earlier :( | |
| 11:02 AM — geerlingguy back to the grind | |
| 11:02 AM ⇐ willthames quit (~willthame@194.74.0.66) Quit: My Mac has gone to sleep. ZZZzzz… | |
| 11:03 AM <crab> oh, we're back to "How would you rate the overall quality of this meeting?" | |
| 11:03 AM <geerlingguy> gregdek++ (invisible karma point) for spearheading it. and rbergeron++ too | |
| 11:03 AM <rbergeron> "amazinnnnnng" | |
| 11:04 AM <crab> protected branches look cool | |
| 11:05 AM ← crab left ⇐ newtMcKerr, bcoca, defionsc_, resmo, j00bar, jimi|ansible and abadger1999 quit ↔ sivel nipped out | |
| 11:36 AM <svg> I didn't even know github had them | |
| 11:43 AM ← Shrews left ⇐ nitzmahone_ and chouseknecht quit ↔ masteinhauser and sivel nipped out | |
| 3:44 PM → jimi|ansible and gundalow joined | |
| 3:58 PM <gundalow> Lots of interesting stuff on https://public.etherpad-mozilla.org/p/ansible-summit | |
| 3:58 PM <gundalow> Nice to see the roadmap | |
| 3:58 PM <gundalow> I'm at the conference tomorrow. Can people join in the sumit? | |
| 4:19 PM <svg> gundalow: the summit was only today | |
| 4:38 PM <jimi|ansible> ISSUE_TEMPLATE! | |
| 4:39 PM <jimi|ansible> we were so close to not having to change anything... | |
| 4:41 PM <gregdek> LOL... | |
| 4:42 PM <jimi|ansible> and it works | |
| 4:42 PM <jimi|ansible> go and try to open a new issue | |
| 4:42 PM <jimi|ansible> https://github.com/ansible/ansible/issues/new | |
| 4:43 PM <gregdek> Well, there we go! | |
| 4:43 PM <gregdek> Now we just need to write the bot to support it. | |
| 4:43 PM <gregdek> Or modify the bot. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment