bodhi-pungi
LOGS
15:02:20 <stickster> #startmeeting Bodhi/Pungi deliverables
15:02:21 <zodbot> Meeting started Fri Apr 28 15:02:20 2017 UTC.  The chair is stickster. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:21 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:02:21 <zodbot> The meeting name has been set to 'bodhi/pungi_deliverables'
15:02:26 <stickster> #meetingname bodhi-pungi
15:02:26 <zodbot> The meeting name has been set to 'bodhi-pungi'
15:02:33 <stickster> #topic Roll call
15:02:37 <stickster> .hello pfrields
15:02:38 <zodbot> stickster: pfrields 'Paul W. Frields' <stickster@gmail.com>
15:02:39 <jcline> .hello jcline
15:02:41 <zodbot> jcline: jcline 'Jeremy Cline' <jeremy@jcline.org>
15:02:55 <stickster> katec: if you do '.hello <FAS username>'
15:02:57 <acarter> .hello acarter
15:02:58 <bowlofeggs> .hello bowlofeggs
15:02:58 <zodbot> acarter: acarter 'Amanda Carter' <acarter@redhat.com>
15:03:01 <zodbot> bowlofeggs: bowlofeggs 'Randy Barlow' <randy@electronsweatshop.com>
15:03:04 <dustymabe> .hellomynameis dustymabe
15:03:07 <zodbot> dustymabe: dustymabe 'Dusty Mabe' <dustymabe@redhat.com>
15:03:11 <stickster> oops, if you do that, it adds you to the roll call (not a formal function but it's how things usually are done)
15:03:13 <kushal> .hellomynameis kushal
15:03:14 <zodbot> kushal: kushal 'Kushal Das' <mail@kushaldas.in>
15:03:31 <dustymabe> stickster: "that" == ?
15:03:36 <katec> .hello katec
15:03:37 <zodbot> katec: katec 'Kaitlyn Carcia' <kcarcia@redhat.com>
15:03:41 <stickster> dustymabe: that ^
15:03:43 <stickster> :-)
15:03:44 <kushal> There you go :)
15:03:48 <stickster> Now I'll chair everyone...
15:04:01 <stickster> #chair dustymabe kushal bowlofeggs katec acarter jcline
15:04:01 <zodbot> Current chairs: acarter bowlofeggs dustymabe jcline katec kushal stickster
15:04:02 * bowlofeggs sits down
15:04:11 <stickster> I didn't see lsedlar & he didn't respond to PM ping
15:04:18 <bowlofeggs> irc is just multi-player notepad
15:04:26 <acarter> he could be on PTO - long weekend for the CZ
15:04:36 <stickster> Oh right -- they have May 1 off as does much of EU and elsewhere
15:04:49 <stickster> Labor Day in many countries
15:04:49 <kushal> stickster, btw, I have applied for sick leave today.
15:04:54 <stickster> kushal: :-(
15:04:56 <kushal> stickster, Yes, even in India.
15:05:01 <bowlofeggs> kushal: i hope you feel better!
15:05:05 * stickster to
15:05:07 <stickster> too, even.
15:05:18 <dgilmore> hi
15:05:20 <kushal> bowlofeggs, stickster thanks :) I hope this meeting will help mentally :p
15:05:21 <stickster> OK, katec  -- to start the discussion just enter '#topic <name of topic to cover>'
15:05:43 <stickster> katec: you now have all the same powers as me since I #chair'd you (and everyone else)
15:05:46 <stickster> #chair dgilmore
15:05:46 <zodbot> Current chairs: acarter bowlofeggs dgilmore dustymabe jcline katec kushal stickster
15:06:05 * stickster turns the gavel over to katec
15:06:18 <katec> feel better kushal! and thanks stickster! let's get started with #multi-arch-ostrees
15:06:34 <dgilmore> katec: https://fedoraproject.org/wiki/Zodbot#Meeting_Functions
15:06:41 <stickster> katec: Here's a hint:
15:06:47 <stickster> #topic Multiarch ostrees
15:07:02 <stickster> The #topic needs to be at the beginning, and zodbot takes care of the rest
15:07:14 <stickster> All the commands start with # for meeting functions
15:07:28 <katec> got it, stickster. and thanks for the link dgilmore :) and in general thank you everything for your patience haha :D
15:07:37 <stickster> np, we're here to help
15:07:50 <katec> i sent an email out because essentially we will need to make a change to bodhi if we need to deliver multi arch ostrees
15:08:01 <katec> this change is part of removing mash from bodhi
15:08:13 <dgilmore> katec: its not part of it
15:08:20 <dustymabe> yeah, two separate work items
15:08:23 <dgilmore> katec: its a side effect of it
15:08:31 <bowlofeggs> yeah this change is about starting to use pungi
15:08:43 <katec> ahhhh, i see
15:08:51 <dgilmore> that suppoorting multiarch ostree's using pungi can enable us to also drop the mash dep
15:08:54 <bowlofeggs> getting pungi into bodhi is a bit of step towards removing mash, but it actually will keep us using mash for the time being
15:09:10 <bowlofeggs> kushal, dustymabe: would you like to update us on the ostree/pungi effort?
15:09:12 <kushal> bowlofeggs, yes.
15:09:36 <katec> dgilmore bowlofeggs dustymabe that you for the clarification :)
15:09:56 <kushal> So, dgilmore suggested to use a full scale pungi config (somehow dynamically we will have to update the pungi config to have the newly created repos somewhere in the config).
15:10:01 <dgilmore> katec: np. weird semantics possibly
15:10:15 <kushal> and then just call pungi from the ostree build par in the mash.
15:10:49 <dustymabe> there are more implementation details in https://github.com/fedora-infra/bodhi/issues/1182#issuecomment-297190959
15:11:03 <dustymabe> dgilmore: ^^ i wrote that from our discussion the other day
15:11:04 <bowlofeggs> kushal: is it possible to pass a config dictionary to pungi if we use pungi by importing it from python rather than subprocessing it, or is that too undocumented/not the golden path?
15:11:28 <dustymabe> bowlofeggs: i think we would clone pungi-fedora pagure repo
15:11:35 <dustymabe> and call pungi with one of those configs
15:11:37 <dgilmore> bowlofeggs: it may be, lsedlar would be best to answer that
15:11:38 <kushal> bowlofeggs, lubomir suggested pungi should not be used as a python module in the current state, and after reading the code for ostree build, I also suggest the same.
15:11:46 <bowlofeggs> oh i see - because the configs then could be dynamically updated
15:11:51 <kushal> lubomir suggested to only update fedora-atomic yum repos, and then execute compose like this example: https://pagure.io/fork/qwan/releng/blob/update-pungi-make-ostree-cmd/f/scripts/build-test-ostree#_22
15:12:06 <kushal> and you can see we can easily do multi arch following that example loop
15:12:16 <katec> dustymabe this also would address consolidating post ga and pre ga ostreee tree generation?
15:12:22 <dustymabe> katec: yes
15:12:26 <bowlofeggs> kushal: yeah i think lsedlar said something similar to me when i talked to him abotu this at devconf
15:12:38 <dustymabe> that is personally why i am pushing this work, multi-arch is a nice side effect
15:12:41 <dustymabe> :)
15:12:50 <kushal> bowlofeggs, which hopefully will keep things simple for us.
15:13:36 <dustymabe> bowlofeggs: kushal, so there are two approaches?
15:13:41 <bowlofeggs> kushal: cool this sounds reasonable to me
15:13:41 <dgilmore> kushal: we could go that route. we would need to write extra code to do things in parallel
15:13:48 <kushal> Our biggest problem to remove mash from bodhi: We have to first create similar dev-setup which can be replicated by us.
15:13:56 <dgilmore> the ostrees are all done in serial in that script
15:14:21 <bowlofeggs> i think we could parallelize it with python's multiprocess or subprocess modules
15:14:23 <dustymabe> so can we keep "remove mash from bodhi" and "pungi ostree generation" as separate topics?
15:14:27 <katec> yes
15:14:30 <kushal> dgilmore, Okay, that is something can be done.
15:14:40 <katec> so it seems like there are 2 things to be discussed here
15:14:42 <dustymabe> in my view, they are separate work items
15:14:48 <kushal> dustymabe, that is true for even any work on mash.
15:14:55 <katec> first 1. support for multi arch ostrees trees
15:15:02 <bowlofeggs> yeah let's keep the mash thing as a separate topic/effort
15:15:04 <katec> 2. removing mash from bodhi
15:15:15 <bowlofeggs> the ostree one is more laser focused and i think probably a lot easier too
15:15:17 <katec> so let's continue our discussion about the ostrees first
15:15:27 <kushal> Even for ostree generation for bodhi for multi-arch: We need to create dev-instances so that we don't blow up production for testing patches.
15:15:27 <bowlofeggs> i also see ostree as lower risk fwiw
15:15:28 <dustymabe> katec: ok taling about point #1
15:15:41 <katec> we had an email thread which dustymabe initiated with mike mcgrath
15:15:47 <kushal> bowlofeggs, yes, that is a good hope :)
15:15:59 <katec> from that email, it is still unclear to me if delivering ostrees for multi arch is a must have for f27
15:16:05 <katec> was anyone about to parse that out?
15:16:06 <dustymabe> katec: can i say something else about point #1
15:16:11 <katec> of course :)
15:16:21 <dustymabe> so for point #1 it is really multiple things
15:16:29 <bowlofeggs> kushal: yeah i would really like to get the dev environment capable of mashing
15:16:32 <dustymabe> A - multi-arch ostree support
15:16:49 <dustymabe> B - consolidations of preGA and postGA ostree generation workflow
15:16:54 <bowlofeggs> jcline recently made a patch that gets the dev environment using fedmsg-hub which is part of the way - we need a way to connect it to a koji
15:17:01 <dgilmore> kushal: the only way to test using koji as a dev is to stand up a koji instance
15:17:06 <dustymabe> C - allows for us to use more meaningful versions in our ostree generation
15:17:15 <dgilmore> kushal: or we do something to enable dev in stage koji
15:17:24 <dustymabe> can we talk about one thing at a time plz
15:17:25 * dgilmore really wishes that the dev koji was funded
15:17:40 <katec> yes, let's focus on dustymabe's points for a moment :)
15:17:49 <katec> he's breaking it down for us ;)
15:17:52 <katec> literally... haha
15:17:56 <katec> continue on dustymabe
15:17:57 <dustymabe> katec: so those are three items that all come from the same work
15:18:16 <dustymabe> so while multi-arch for ostree is not necessarily a requirement for f27
15:18:32 <dustymabe> the other pieces are important to me (and they are in our plan I believe)
15:18:45 <stickster> it's like a benefit that we derive from doing the work the right way IIUC
15:19:10 <dustymabe> and TBH it's probably not that much work as kushal is already working on it
15:19:16 <dustymabe> and we have a clear plan laid out
15:19:20 <kushal> dustymabe, if we start building ostree using pungi (pungi-ostree), we will get most of multi arch for free.
15:19:23 <dustymabe> https://github.com/fedora-infra/bodhi/issues/1182#issuecomment-297190959
15:19:36 <katec> dustymabe, correct in round 1 for planning, we did say we would do pre and post ostree considilation, and it's great the same fix would address other areas
15:19:40 <dgilmore> based on the issue that ksinny filed in releng I was under the impression that we wanted to fully support AtomicHost on multiarch in F27
15:19:48 <dgilmore> which implies being able to do updates
15:19:52 <dustymabe> dgilmore: yes, they do want that
15:19:58 <dustymabe> but it's hard to say requirement
15:20:09 <katec> and dustymabe makes a good point. kushal it sounds like you're already well under way with doing this fix?
15:20:11 <dustymabe> if they want to say it's a requirement, then fine
15:20:12 <dgilmore> if we are not going to, we need to be clear now that it will only be done in rawhide and maybe branched
15:20:47 <stickster> Well, and further than that, we need to be clear on *why* and what other work is prioritized ahead of it, so it's clear what the tradeoff is
15:20:57 <kushal> katec, I am working, at the end it may look like a very simple fix, but this may also blow up bodhi few times :)
15:21:28 <dustymabe> kushal: that's why we are going to special case the code and only run it in the new way if we ask bodhi to do so
15:21:34 <stickster> Think of it like changing the tire while the car is moving on the highway ;-)
15:21:35 <katec> well first, let me ask this
15:21:44 <dgilmore> if we use pungi-make-ostree by calling koji runroot, we limit the use to the ostree work.
15:21:46 <katec> who would be working on this?
15:21:51 <katec> it sounds like kushal, but who else?
15:21:53 <kushal> dustymabe, yeah :)
15:22:04 <kushal> dgilmore, which way?
15:22:12 <dgilmore> katec: I think it depends on what route we take
15:22:21 <dgilmore> kushal: because that all it can do
15:22:24 <dustymabe> katec: mostly kushal, and then some help from releng/me to create a new pungi config that can be used
15:22:34 <kushal> kalev, bowlofeggs will help me by providing bodhi guidance, and dusty with other important knowledge :)
15:22:40 <dgilmore> kushal: if we moved to calling pungi-koji we can then do more
15:22:49 <dustymabe> kushal: let's do it exactly how we described in https://github.com/fedora-infra/bodhi/issues/1182#issuecomment-297190959 unless we find some outstanding issue with it
15:22:53 <kushal> dgilmore, correct, I am putting pungi-koji as step 2.
15:23:16 <dustymabe> kushal: what is step 1 ?
15:23:17 <dgilmore> kushal: okay, It sounded like you were going down a different route
15:23:29 <kushal> dgilmore, or as dusty said above, help us with generating right pungi config, then we can use pungi-koji.
15:23:49 <dustymabe> kushal: that is exactly the work laid out in https://github.com/fedora-infra/bodhi/issues/1182#issuecomment-297190959
15:23:51 <dgilmore> kushal: I have already sai that releng will provide and manage the configs
15:23:51 <kushal> dgilmore, I am okay to follow your path, but will need help.
15:24:01 <dustymabe> if you do anything different than that please justify it in the issue
15:24:19 <katec> ok so let's review what we've said so far... this helps us with a. support for multi ostrees b. pre and post ga consildation for ostrees and c. more meaningful versions of ostree tree. the people working on this would be kushal and dustymabe
15:24:20 <dustymabe> we have a plan
15:24:22 <kushal> dustymabe, dgilmore Nope, I will then just follow the path as described in the bug.
15:24:50 <stickster> so if I understand right... everyone agrees that https://github.com/fedora-infra/bodhi/issues/1182#issuecomment-297190959 is the plan of record for now?
15:24:57 <stickster> kushal: dustymabe: dgilmore: ^
15:25:03 <dustymabe> stickster: that would be my opinion
15:25:05 <dgilmore> katec: mboddu, Kellin, and myself will provide configs and assistance
15:25:08 <katec> in terms of why we need this: we will need to support ostress for multi arch (even if it's unclear whether or not we need that now) but we do know we need to do best enablement for phoebe
15:25:18 <katec> and this is part of that
15:25:28 <katec> dgilmore thanks for pointing that out
15:25:41 <kushal> dgilmore, Thank you.
15:25:53 <dustymabe> yes, and i also don't think it's really going to be that much work in the grand scheme of things, mostly just wiring up things that already exist
15:26:34 <bowlofeggs> this plan sounds good to me. i really wish we didn't have to test in production, but that's another topic ☺
15:26:40 <kushal> dustymabe, yes, and I am hoping to help mboddu with more explanation of the configurations.
15:26:52 <kushal> For docs ^^^^
15:26:56 <stickster> bowlofeggs: yeah, that's a bigger can of worms for another day
15:27:03 <kushal> hehe
15:27:20 <stickster> katec: we're coming down to :30, what else do we need to settle here?
15:27:27 <bowlofeggs> it's a HUGE can of worms ☺
15:27:34 <katec> #topic removing-mash-from-bodhi
15:27:37 <dgilmore> bowlofeggs: we should be able to test in stage koji
15:27:47 <katec> with python 3
15:27:51 <katec> we have either 2 options
15:27:56 <bowlofeggs> dgilmore: afaik, it's not possible to mash on staging
15:28:04 <bowlofeggs> dgilmore: though i'd reaaaaally like to be able to
15:28:05 <dgilmore> bowlofeggs: course it is
15:28:06 <katec> dgilmore correct me if i'm wrong here but it's
15:28:06 <puiterwijk> bowlofeggs: mash: no. Pungi: yes
15:28:11 <katec> 1. remove mash from bodhi OR
15:28:31 <bowlofeggs> i have one more thought abotu the last topic
15:28:34 <katec> 2. port mash to dnf then move createrepo_c to python 3, port mash to python 3
15:29:03 <bowlofeggs> we're coming up on the beta freeze (may 17 iirc) and i'd realyl like to avoid deploying this change very close to it (like, i'd rather not deploy after may 10 ideally)
15:29:06 <dgilmore> katec: 2 is port mash to dnf and createrepo_c in order to be able to port to python3
15:29:15 <bowlofeggs> because we're likely to have problems since we can't test
15:29:32 <bowlofeggs> so we have not-much-time to make a release with this
15:29:35 <dgilmore> createrepo_c supports python3 today, createrepo never will
15:29:54 <kushal> bowlofeggs, problem to that, I will be traveling to USA starting my journey on 3rd super early morning from India, and will reach on 4th evening in Portland.
15:30:11 <dustymabe> kushal: bowlofeggs let's take to another channel
15:30:16 <dustymabe> #fedora-cloud
15:30:17 <kushal> yes
15:30:23 <dgilmore> katec: but yes in order to move to python3 for bodhi one of those 2 things needs to happen
15:30:33 <katec> correct, so the question is
15:30:36 <katec> what path do we take?
15:30:49 * stickster notes his time is up for this meeting, we have another one elsewher
15:30:52 <stickster> *elsewhere
15:31:00 <katec> ok, we can bring this to planning
15:31:06 <dgilmore> we have two paths to remove mash, we use pungi via pungi-koji or we port to dist-repos in koji
15:31:12 <katec> planning will be at 1pm
15:31:25 <bowlofeggs> yeah i have the same other meeting as stickster now
15:31:33 <dustymabe> dgilmore: is there a summary of pros/cons for those two?
15:31:50 <katec> if dgilmore can be prepared to tlak about the pros/cons during planning that would be great
15:31:57 <katec> we'll have time then to hash this out
15:31:59 <bowlofeggs> it sounds to me like removing mash and using pungi might be the path of least resistance?
15:32:02 <puiterwijk> katec: I'd say -1 to that, since this discussion should probably be public
15:32:06 <dgilmore> katec: sure can
15:32:10 <kushal> bowlofeggs, +1 from me.
15:32:29 <katec> dgilmore you guys can discuss it here, and then we can recap at planning
15:32:35 <katec> all notes from planning today will be pubic
15:32:42 <bowlofeggs> katec: why don't we just plan another meeting like this one to talk about mash vs. pungi vs. whatever else
15:32:55 <dgilmore> bowlofeggs: katec: I am good with that
15:33:01 <katec> ok
15:33:06 <katec> i'll get one on the cal for next week
15:33:10 <katec> :)
15:33:16 <bowlofeggs> katec: you can use #endmeeting to end the meeting
15:33:19 <dgilmore> gracias katec
15:33:23 <bowlofeggs> and then you will get automatic notes generated too!
15:33:29 <katec> thanks everyone for attending!
15:33:31 <katec> #endmeeting