15:01:43 <nils> #startmeeting modularity_wg
15:01:43 <zodbot> Meeting started Tue Mar 21 15:01:43 2017 UTC.  The chair is nils. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:43 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:01:43 <zodbot> The meeting name has been set to 'modularity_wg'
15:01:43 <nils> #meetingtopic Meeting of the Modularity Working Group (once every two weeks)
15:01:43 <nils> #chair dgilmore langdon mikedep333_ sct tflink
15:01:43 <zodbot> Current chairs: dgilmore langdon mikedep333_ nils sct tflink
15:01:51 <sct> .hello sct
15:01:53 <nils> #topic Roll Call
15:01:54 <linuxmodder> .fas linuxmodder
15:01:58 <contyk> .hello psabata
15:01:58 <nils> .hello nphilipp
15:02:07 <zodbot> sct: sct 'Stephen Tweedie' <sct@redhat.com>
15:02:09 <jkurik> .hello jkurik
15:02:10 <zodbot> linuxmodder: linuxmodder 'Corey W Sheldon' <sheldon.corey@openmailbox.org>
15:02:13 <zodbot> contyk: psabata 'Petr Šabata' <psabata@redhat.com>
15:02:15 <zodbot> nils: nphilipp 'Nils Philippsen' <nphilipp@redhat.com>
15:02:20 <langdon> .hello langdon
15:02:21 <zodbot> jkurik: jkurik 'Jan Kurik' <jkurik@redhat.com>
15:02:26 <zodbot> langdon: langdon 'Langdon White' <langdon@fishjump.com>
15:02:27 * threebean 
15:02:37 <sgallagh_> .hello sgallagh
15:02:38 <zodbot> sgallagh_: sgallagh 'Stephen Gallagher' <sgallagh@redhat.com>
15:03:06 * asamalik_ had some networking issues..
15:03:10 <asamalik_> .hello asamalik
15:03:11 <zodbot> asamalik_: asamalik 'Adam Samalik' <asamalik@redhat.com>
15:03:29 <karsten_> .hello karsten
15:03:30 <langdon> threebean, you here for the meeting? want to know if we should add jkurik's fesco ticket comment to the agenda
15:03:30 <zodbot> karsten_: karsten 'Karsten Hopp' <karsten@redhat.com>
15:03:43 <threebean> ah, yes.
15:03:47 <contyk> it's on the agenda
15:03:51 <nils> #topic Agenda
15:03:51 <nils> #info Deferring Modularity Changes in F26 scope to F27
15:03:51 <nils> #link https://pagure.io/fesco/issue/1688#comment-432448
15:04:04 <jkurik> langdon: it is there: http://piratepad.net/modularity-wg-agendas line #26
15:04:22 <nils> ^^
15:04:29 <nils> is that it?
15:04:29 <langdon> cool.. but thought we needed appropriate participants too :)
15:04:38 <jkurik> sure :)
15:04:41 <threebean> I haven't had a chance to respond on the fesco ticket yet, but in short I don't think we yet have to defer any compose work for f27 yet.
15:05:06 <nils> before we start discussing individual things, anything else for the agenda`
15:05:07 <nils> ?
15:05:11 <langdon> threebean, NO punchlines yet!
15:05:11 * threebean waits
15:05:12 <langdon> :)
15:05:45 <nils> Nothing else? Okay.
15:05:57 <contyk> it will be all part of this point
15:06:04 * threebean nods
15:06:13 <nils> #topic Deferring Modularity Changes in F26 scope to F27
15:06:30 <nils> .chair threebean jkurik contyk
15:06:30 <zodbot> threebean jkurik contyk is seated in a chair with a nice view of a placid lake, unsuspecting that another chair is about to be slammed into them.
15:06:34 <jkurik> so, I put the comment there as formally we are in the Alpha freeze now, when we should not change our infrastructure
15:06:45 <jkurik> the MBS Change needs modification of Pungi, which we should not be doing IMO at this stage of F26
15:06:58 <nils> #chair threebean jkurik contyk
15:06:58 <zodbot> Current chairs: contyk dgilmore jkurik langdon mikedep333_ nils sct tflink threebean
15:07:02 <threebean> jkurik: ok, not totally correct.  allow me to clarify:
15:07:03 <jkurik> that is basically my only concern
15:07:26 <langdon> when did zodbot start getting salty?
15:07:34 <threebean> jkurik: there is a "simpler" modular compose that can be done with no modifications to the pungi software.
15:07:38 <nils> langdon: when I used a dot instead of a hash sign
15:07:57 * nils hi-fives zodbot in the face, with a chair
15:08:04 <threebean> jkurik: it involves (simply) adding a new cronjob and configs that tells pungi to run nightly to pull in the base-runtime module, and do a single-module compose from that.
15:08:05 <langdon> nils, tomas & contyk may be adding to the agenda
15:08:18 <threebean> it's not what we want in the long run, but it is sufficient for alpha
15:08:56 <jkurik> ok, so may I ask you please to comment it on the FESCo ticket ?
15:09:09 <contyk> langdon: I haven't read the #fedora-modularity backlog yet so I'm not entirely sure what I should be adding :)
15:09:16 <threebean> at the moment, we're partly blocked on the content of the base-runtime module for that.  we think we can build for the "primary" arches, but some work needs to be done to provide a few packages to enable secondary arches which is required by releng.
15:09:31 <threebean> jkurik: I will.  but let's discuss it here so the group can help figure out if we're missing anything.
15:09:37 <langdon> contyk, ohh.. request for base-container-image automated testing.. so we don't split the conversation
15:09:50 <contyk> the current package set in koji lacks shim, so I wouldn't be so sure about ix86 either
15:09:58 * threebean nods
15:09:59 <contyk> but there will be another build of the base-runtime module later this week
15:10:21 <threebean> contyk: is there anything blocking you with respect to shim?
15:10:27 <contyk> threebean: nope
15:10:30 <threebean> awesome.
15:10:59 <contyk> I'm just in the middle of putting the new list of binary RPMs together
15:10:59 <threebean> ok - so that's the lowest hanging fruit for f26 modular compose.  just run pungi as is with the base-runtime as input -- a single-module compose.
15:11:17 <contyk> that compose won't contain the modulemd file, though
15:11:21 * threebean nods
15:11:23 <threebean> that's right.
15:11:55 <contyk> threebean: which reminds me (off topic) -- we'll want MBS to fill in the list of binary RPMs at some point
15:12:06 <threebean> contyk: yes, let's take it up later.
15:12:10 * contyk nods
15:12:13 <threebean> so, the second milestone:
15:12:24 <threebean> we have a patch to pungi prepared and proposed that allows for:
15:12:31 <threebean> a) multi-module composes
15:12:43 <threebean> b) baking the modulemd into the resultant compose repo.
15:13:10 <threebean> and a lot of the anxiety seems to be about the fate of that patch.
15:13:54 <threebean> my understanding at the moment is that the pungi maintainers are "ok" with including it in the next release.
15:13:57 <sct> threebean: Do you know what format the modulemd will appear in in the compose?  Or is that a detail still to be worked out with the client tools?
15:14:16 <threebean> sct: I can't cite it off hand, but the details can be figured out from the existing patch.  I'll link to it shortly.
15:14:35 <sct> threebean: Thanks!
15:14:56 <contyk> I think the modulemd files will be part of the repodata, listed as "modules"-type content
15:14:57 <threebean> there's a second hurdle to clear -- when we do have a release of pungi that includes our module-module/modulemd patches, will release engineering *use* it for F26 deliverables?
15:15:17 <contyk> sct: kinda like here -- https://fedorapeople.org/groups/modularity/repos/base-runtime/26/repodata/
15:15:18 <threebean> the latest I've heard is that, even if we get a release of pungi today, it will not be used to produce F26 Alpha.  it's just too late in the alpha cycle for that.
15:15:33 * ttomecek waves
15:15:36 <threebean> however, the infra/tooling freeze *lifts* after alpha goes out, and we have another window to upgrade pungi.
15:15:58 <threebean> releng is the ultimate authority on this, but I do not think we are blocked from having multi-module/modulemd enabled composes for F26 Beta.
15:16:07 <jkurik> threebean: so this comment https://bugzilla.redhat.com/show_bug.cgi?id=1397141#c6 is not valid then ? .. and we do not need the pungi modification ATM ?
15:16:10 <contyk> threebean: sgallagh_ says you would still need FESCo approval, even after the "unfreeze"
15:16:26 <contyk> if I understood him correctly
15:16:35 * threebean clicks https://bugzilla.redhat.com/show_bug.cgi?id=1397141#c6
15:16:35 <jkurik> threebean: releng is fine to modify the pungi for rawhide (as I was told today)
15:16:54 <sgallagh> Not necessarily
15:17:04 <sct> threebean: Yeah --- that drops the module md straight into the repo, gzipped
15:17:04 <sgallagh> Mostly if there's serious reservation about getting this in at this point
15:17:12 <threebean> jkurik: on https://bugzilla.redhat.com/show_bug.cgi?id=1397141#c6, it is out of date.  the BRT module is built now.
15:17:34 <contyk> \o/
15:17:38 <threebean> jkurik: there are two milestones, we would very much like to have the pungi modification in.
15:17:41 <sct> threebean: There's a scalability question about whether we want to do that for all modules, resulting in a lot of separate metadata downloads; or whether we can tar/zip or concat the yaml files into a single file
15:17:45 <threebean> but without it, we can still produce "something".
15:18:13 <contyk> threebean: YAML files can contain multiple documents in one file
15:18:16 <langdon> jkurik, "we" in threebean's comment == modularity team as well as factory
15:18:18 <contyk> sct: ^
15:18:30 <langdon> sct, let's address that later
15:18:32 <contyk> we could do that... and teach modulemd (the library) to handle that
15:18:54 <langdon> contyk, geppetto is testing that right now actually.. although on PTO at the moment
15:19:02 <sct> langdon: Will do, I'll send email --- shouldn't be hard (as contyk says, yaml should be able to help here anyway)
15:19:19 <langdon> sct, we asked geppetto to look at it as part of his prototype
15:19:22 <sct> langdon: as long as we make the call soon so we don't block threebean's work
15:19:28 <threebean> contyk: on the topic of "fesco approval even after unfreeze", I'll follow up with releng before commenting on the fesco ticket.
15:19:41 * contyk nods
15:19:49 <sct> langdon: Right --- I just want to avoid threebean doing a different implementation while geppetto is doing that
15:19:50 <sgallagh> ack
15:19:56 * threebean nods
15:19:57 <nils> okay, before it gets too much: anything you guys want to #info, #link, #action up to now?
15:20:22 <contyk> hmm
15:20:23 <threebean> #action threebean to follow up with fedora releng about what changes can be made after alpha freeze lifts.
15:20:33 <langdon> yes please.. request threebean to summarize status.. request jkurik to comment on if he is "good with status"
15:20:55 <contyk> nils: we have a build of Base Runtime in koji now, that's newsworthy :) even though it's not perfect
15:21:05 <langdon> contyk, \o/
15:21:10 <threebean> #action threebean to report status (including releng conversations) to the fesco ticket.
15:21:13 <nils> contyk: but that's another topic, right?
15:21:17 <asamalik_> contyk: yay!
15:21:22 <nils> and, yay!
15:21:27 <jkurik> langdon: I am just point on it as it is somehow breaking the process, but the decision is on FESCo, not on me
15:21:29 <contyk> nils: there's no separate topic :)
15:21:47 <nils> contyk: there's always open floor, and we can just tack it on after we're done with this one
15:21:48 <contyk> nils: basically threebean's already mentioned all there is to say about it
15:21:48 <langdon> jkurik, ack.. but does the explanation thus far make sense to you?
15:21:53 <nils> and/or
15:22:01 <jkurik> langdon: yes. it does
15:22:04 <langdon> cool
15:22:18 <jkurik> thanks threebean for the explanation
15:22:21 <jkurik> threebean++
15:22:22 <zodbot> jkurik: Karma for ralph changed to 11 (for the f25 release cycle):  https://badges.fedoraproject.org/tags/cookie/any
15:22:27 <threebean> of course!  can I answer any other questions on the topic?
15:22:49 * langdon notes things are starting to come together for real with modularity.. which == everything coming apart ;)
15:23:20 <langdon> ok.. thanks threebean & jkurik .. . nils next topic?
15:23:20 <nils> langdon: first, things have to be broken to be mended again? ;)
15:23:25 <nils> no next topic
15:23:34 <nils> I mean I can shoehorn BRT in of course
15:23:38 <langdon> its only real when you see the cracks things are falling throuhg
15:23:41 <nils> or we can go to open floor directly
15:23:51 <langdon> ttomecek & contyk had items to add
15:23:51 <nils> haha
15:23:59 <langdon> or they could just be open floor.
15:24:02 <contyk> it's more of an open floor item anyway
15:24:06 <nils> okay
15:24:10 <nils> #topic Open Floor
15:24:17 <contyk> right
15:24:18 <nils> bring it on :)
15:24:20 <langdon> contyk, i wanted your test request in there too.. not just the brt build
15:24:50 <contyk> so as threebean and I have mentioned, there's a koji build of base-runtime available now, so you can reference it in your modules -- https://koji.fedoraproject.org/koji/taginfo?tagID=478
15:25:02 <threebean> \ó/
15:25:14 <langdon> \o/++ ?
15:25:32 <nils> contyk: #link it?
15:25:47 <contyk> once we can compose it, we will be creating a kickstart for a base-runtime-based docker base image which should be available in parallel to the standard Fedora docker base image
15:25:49 <asamalik_> \ò/
15:26:06 <contyk> nils: never sure how these commands work
15:26:24 <nils> contyk: #link http://foo/bar some description (optional)
15:26:26 <sgallagh> contyk: literally "#link <URL>"
15:26:48 <contyk> didn't that happen automatically for posted URLs?
15:26:50 * contyk shrugs
15:26:55 <nils> AFAIK no
15:27:00 <contyk> #link https://koji.fedoraproject.org/koji/taginfo?tagID=478 the first ever Base Runtime build in koji
15:27:08 <nils> 👍
15:27:10 <contyk> anyway
15:27:42 <contyk> I was wondering -- once we have the compose and a docker base image kickstart, does anybody know what happens next?
15:27:59 <contyk> how is it tested, if at all? does that happen before we push the image to Fedora registry?
15:28:32 <contyk> if not, what needs to be done to make that happen?
15:29:01 <contyk> we could always create an empty layer and test Base Runtime through that but I'd much rather test it before we push it to the registry
15:29:14 <contyk> and gate on the test results
15:29:50 <contyk> tflink: ? ^
15:30:18 <contyk> langdon: is there anything else you had in mind?
15:30:22 * langdon notes maxamillion isn't here
15:30:31 * contyk noticed
15:30:41 <tflink> I think that's the idea, assuming I'm following the question correctly
15:30:42 <langdon> ttomecek, wanted to also ask about tests for containers, i believe
15:31:03 * ttomecek is waiting for the right moment
15:31:06 <nils> langdon: he's on IRC, just not here
15:32:01 <ttomecek> any notes, actions, links?
15:32:23 <nils> ttomecek: contyk linked the BRT build in koji
15:32:29 <langdon> need to follow up with maxamillion on the base-continaer testing
15:32:46 <langdon> volunteer? threebean was gonna but he is pretty busy :)
15:32:53 <ttomecek> nils, yes, I meant the brt image testing
15:32:59 <nils> ahh
15:33:14 <threebean> :P
15:33:21 <ttomecek> contyk, you want to test it as a docker image?
15:34:05 <threebean> fwiw, tflink said something in person the other week about just writing a taskotron task that tries to start it with docker.
15:34:13 <ttomecek> you can always write a test to do `dnf install --installroot ...` && tar && docker import
15:35:17 <contyk> Sorry, lost network completely
15:35:49 <contyk> ttomecek: I want to test the base image contents
15:35:56 <tflink> if we enable the triggering on docker build (I assume that's what we're talking about), it's possible to run a container from a task
15:36:06 <contyk> Preferably before pushing the base image to the registry
15:36:42 <ttomecek> contyk, how do you plan to create the base image? via image factory?
15:37:10 <contyk> I think you build those with kickstarts
15:37:25 <tflink> imagefactory takes ks as input
15:37:35 * contyk nods
15:38:54 * ttomecek brb
15:38:56 <langdon> ok.. so who wants to do the follow up?
15:39:00 <langdon> contyk, ?
15:39:58 <contyk> sure; I'll discuss that with maxamillion
15:40:09 <langdon> contyk, cool.. thanks
15:40:18 <langdon> ok.. ttomecek wandered off .. but i think he is next
15:40:48 <langdon> basically, he wanted to know "where on the roadmap is triggering taskotron tests for the docker namespace in dist-git"
15:40:57 <langdon> related.. do we have that for module namespace?
15:41:13 <langdon> tflink, ^^ mostly, i think? maybe threebean ?
15:41:23 <ttomecek> langdon, thank you
15:41:38 <ttomecek> so 2 things: we want to test our container image sources (Dockerfiles)
15:41:39 <tflink> assuming that I understand where things are, it's just a config change and making sure that we're enabling the right thing
15:41:52 <ttomecek> this should be triggered by dist-git pushes
15:42:16 <tflink> that's going to take more than a config change but shouldn't be too hard
15:42:27 <ttomecek> second thing is pretty much the same thing as contyk said: testing built images
15:42:51 <tflink> that -> triggering on dist-git pushes. specifically, telling the difference between tasks that should be run on build vs. on push
15:43:15 <langdon> well.. in the meantime.. a fedpkg build triggers it wwhne the build is done
15:43:31 <contyk> hm?
15:43:31 <langdon> ttomecek, so.. it is "mostly" automated.. just not on commit.. but rather build
15:43:37 <langdon> tflink, right?
15:43:40 <contyk> no
15:43:47 <langdon> ??
15:43:51 <langdon> now i am confused ;)
15:43:51 <contyk> fedpkg build doesn't trigger any tests, it just initiates builds
15:44:00 <ttomecek> langdon, are you saying we should lint our dockerfiles after a build is done?
15:44:10 <tflink> tests are triggered on fedmsgs. right now, we're mostly listening for build messages
15:44:39 <tflink> the thing missing to enable post-push checks is a method for declaring "X is run on git push" and "Y is run on build"
15:46:05 <langdon> i wonder if we need a voice call to walk through the planned infra.. and see what is done.. i wonder if you could put that work on the module itself.. in other words, taskotron runs all tests on every event, the module indicates, based on process stage, which tests must pass
15:46:15 <ttomecek> tflink, that should be defined in runtask.yaml, right?
15:47:10 <tflink> ttomecek: still undefined but that is one option, yes
15:47:32 <contyk> langdon: we run a modulemd sanity check on git push to modules/*, we then run various global rpm-level tests on rpm build completion
15:47:32 <langdon> tflink, ttomecek couldn't we just fire a manual "go test this now" message on the bus?
15:47:41 <ttomecek> contyk, OSBS is capable of initiating a koji task (imgfac) to create a filesystem tree, then fetch it from koji and finally produce a base docker image
15:47:47 <contyk> langdon: we will also run rpm package-specific tests upon build completion soon
15:47:52 <langdon> contyk, currently? or we would like too?
15:48:05 <contyk> langdon: currently
15:48:21 <langdon> contyk, automatically? like.. "how?"
15:48:44 <tflink> langdon: we could also put some assumptions in place that would eventually go away
15:49:00 <tflink> something like "everything in dir X is post-push, everything else is post-build"
15:49:04 <contyk> langdon: git push produces a fedmsg which is caught by a jenkins listener or whatever it is
15:49:18 <contyk> it then runs check_modulemd in jenkins
15:49:56 <langdon> tflink, contyk i think we need a doc(s) saying "if you want to setup a test right now, do this" .. and the "eventually, it will work like this"
15:49:56 <contyk> langdon: tflink and threebean will know more about where the global rpm-level tests are
15:52:41 <langdon> contyk, anyone on your team have bandwith to write up what you are doing to exec tests? even if it is an email?
15:53:03 <langdon> tflink, anyone in your world have time to write up what we would like to see? and/or what you think we can do right now?
15:53:06 <contyk> merlinm: ^?
15:53:44 <tflink> langdon: I can work on that but I'd prefer to have some input from the brt and/or other modularity folks
15:54:35 <merlinm> contyk: sorry, let me read and catch up
15:54:39 <langdon> tflink, a strawman is all i meant
15:54:54 <langdon> merlinm, just asking for a doc saying how all the brt tests work in fedora-infra
15:54:57 <tflink> I can at leasta start on that
15:55:10 <langdon> like how they get triggered/run.. more than "uses avocado"
15:55:21 * tflink is also happy to work with the brt folks to get things enabled in the short term WRT running things on push vs build
15:55:29 <langdon> tflink, threebean may already have an early / high-level version
15:56:08 <langdon> tflink, merlinm can i give you "#actions"?!?!?!
15:56:10 <langdon> :)
15:56:14 <langdon> or give yourself ones?
15:56:19 <tflink> langdon: go for it
15:56:39 <tflink> as an aside, I have a hard stop at the top of the hour
15:57:04 <langdon> #action tflink to create a more detailed version of "fedora testing infra: future-state" as a strawman for discussion
15:57:04 <merlinm> I can't say I know the latest on it all works. But it's a good idea to summarize, so I can find out and do that.
15:57:06 * ttomecek is fine with the outcome
15:58:03 <langdon> #action merlinm to write up how the testing works for BRT (specifically the chain of workflow) so that other modules can replicate in the near term
15:58:25 <jkaluza> Hi, have you discussed modular compose? I have joined quite late.
15:58:35 * jkaluza will read the meeting log in case you did :)
15:58:35 <ttomecek> #link https://github.com/projectatomic/atomic-reactor/blob/master/atomic_reactor/plugins/pre_add_filesystem.py OSBS is able to create filesystem via koji task and then produce a base image
15:58:46 <langdon> #action phracek & jscotka to work in modularity-testing-framework in to the doc by merlinm
15:58:58 <langdon> jkaluza, we did
15:59:06 <jkaluza> langdon: great, thanks.
15:59:19 <langdon> anything else for open floor?
15:59:34 * tflink needs to head out. will check minutes when he gets back
16:00:15 <langdon> merlinm, the action make sense?
16:00:55 <merlinm> yup
16:01:38 <ttomecek> so what about the other action? *the wished* state
16:01:51 <langdon> ttomecek, tflink said he is gonna put a strawman together
16:01:54 <langdon> see above
16:02:04 <ttomecek> langdon, but also said he wants some input from us
16:02:38 <ttomecek> so maybe tflink can create a strawstrawman and we'll help him create a strawman
16:02:40 <langdon> ohh.. i figured he would do a strawman.. then we could all discuss.. much more useful to talk based on something than just totally abstract (IMO)
16:02:55 <ttomecek> deal
16:02:58 <langdon> and maybe it will be perfect out of the gate :)
16:03:27 <langdon> and.. i am fairly sure threebean already had the high level in the f-2 plan..
16:03:52 * langdon digs for link
16:04:10 * langdon also x-d in the ether with ttomecek's strawman remark :)
16:04:25 <contyk> should we close this meeting?
16:04:25 * jkurik needs to leave
16:04:29 * threebean nods
16:04:30 <contyk> the longest open floor ever
16:04:31 <threebean> over time.
16:04:34 <langdon> ha
16:06:44 <langdon> nils, ? close?
16:06:56 <langdon> link is currently broken.. hassling asamalik about it now
16:07:04 <langdon> https://pagure.io/modularity/issue/15
16:07:39 <asamalik> langdon: ah... will fix this tomorrow
16:07:45 <langdon> ok.. im closing
16:07:49 <langdon> we seem to be missing nils
16:07:53 <langdon> #end-meeting
16:07:56 <langdon> #endmeeting