source-git-sig
LOGS
13:01:48 <ttomecek> #startmeeting Fedora Source-git SIG
13:01:48 <zodbot> Meeting started Thu Apr 22 13:01:48 2021 UTC.
13:01:48 <zodbot> This meeting is logged and archived in a public location.
13:01:48 <zodbot> The chair is ttomecek. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:48 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
13:01:48 <zodbot> The meeting name has been set to 'fedora_source-git_sig'
13:02:04 <ttomecek> #meetingname source-git-sig
13:02:04 <zodbot> The meeting name has been set to 'source-git-sig'
13:02:21 <ttomecek> #chair csomh lbarczio mfocko Bennett flachman jpopelka Eighth_Doctor
13:02:21 <zodbot> Current chairs: Bennett Eighth_Doctor csomh flachman jpopelka lbarczio mfocko ttomecek
13:02:29 * Eighth_Doctor waves
13:02:34 <ttomecek> hello everyone!
13:02:38 * csomh o/
13:02:53 <jpopelka> hello
13:03:02 <Bennett> Hello all
13:03:29 <ttomecek> I'm wondering if I got the invitation right
13:03:48 <ttomecek> malmond, welcome! sorry to wake you up so early
13:03:59 <malmond> ttomecek: no problem, I made coffeeeee
13:04:21 <ttomecek> #topic Introduction and Welcome
13:04:46 <ttomecek> please use #chair to let us know that your participating :)
13:05:01 <ttomecek> I already chaired a few people above which I knew that want to participate
13:05:31 <Eighth_Doctor> .hello ngompa
13:05:31 <zodbot> Eighth_Doctor: ngompa 'Neal Gompa' <ngompa13@gmail.com>
13:05:33 <ttomecek> let's start with some introductions so that we get to know each other better
13:06:14 <ttomecek> #meetingname source-git-sig
13:06:14 <zodbot> The meeting name has been set to 'source-git-sig'
13:06:22 * King_InuYasha waves again
13:06:41 <ttomecek> #chair King_InuYasha
13:06:41 <zodbot> Current chairs: Bennett Eighth_Doctor King_InuYasha csomh flachman jpopelka lbarczio mfocko ttomecek
13:06:49 <malmond> hey Neal, fancy meeting you here
13:07:17 <malmond> #chair malmond
13:07:23 <malmond> I guess?
13:07:35 <Eighth_Doctor> malmond: Yo! Well, I figured I should be here since I've implemented this whole thing for work already, and so I can probably help avoid bad things from happening with this. :)
13:07:41 <ttomecek> #chair malmond
13:07:41 <zodbot> Current chairs: Bennett Eighth_Doctor King_InuYasha csomh flachman jpopelka lbarczio malmond mfocko ttomecek
13:07:50 <jforbes> .hello2
13:07:50 <zodbot> jforbes: jforbes 'Justin M. Forbes' <jforbes@redhat.com>
13:08:04 <ttomecek> hmm, I thought that everyone can do that - perform #chair
13:08:10 <Eighth_Doctor> nope
13:08:10 <ttomecek> #chair jforbes
13:08:10 <zodbot> Current chairs: Bennett Eighth_Doctor King_InuYasha csomh flachman jforbes jpopelka lbarczio malmond mfocko ttomecek
13:08:14 <Eighth_Doctor> only chairs can
13:08:24 <jforbes> Can only stay for a bit, was supposed to leave for a trip this mornin
13:09:18 <ttomecek> jforbes, oh, thanks for hopping in then! please give us a short intro
13:10:04 <ttomecek> Eighth_Doctor, nice! I almost forgot you did that, it was a long time since I saw you last time :)
13:10:21 <malmond> .hello malmond
13:10:22 <zodbot> malmond: malmond 'None' <malmond@fb.com>
13:10:30 <jforbes> Oh, I am Justin Forbes, Fedora kernel maintainer
13:10:49 <malmond> yikes, I don't get out much, malmond = Matthew Almond
13:11:08 <Eighth_Doctor> ttomecek: yeah, I think we spent half a day at Summit a few years back just talking about what I did and the pros and cons of the source-git approach
13:12:07 <ttomecek> intro from me: I'm Tomas and I'm a product owner for project packit ( https://packit.dev ), our current focus right now is to create a source-based downstream maintenance workflow for CentOS Stream (both versions, 8 and 9) and after getting feedback from multiple places, we wanted to pursue the same thing in Fedora Linux to that we can have one source-based workflow across all Red Hat distros
13:13:18 <ttomecek> Eighth_Doctor, yes! I recall you have in on your private gitlab, but back then we did not have any support for gitlab in our project
13:13:24 <Eighth_Doctor> Yes
13:14:36 <ttomecek> anyone else wants to introduce?
13:15:00 <flachman> Hi all! I'm František. I am another member of the Packit team and was there from the very beginning and I am mostly responsible for the upstream part -- github/gitlab application.
13:16:25 <ttomecek> oki, thank you for your introductions!
13:16:29 <malmond> intro: I'm Matthew and I'm part of the team that work on packaging software/infrastructure at Facebook. We use CentOS and rpm, like to contribute to upstream first, but then do backports to CentOS Stream. We (well, dcavalca) set up the Hyperscale SIG so we can publish and share our backports. Doing backports is my main interest here, and expecially patch management. I'll admit I've not dug deep into this
13:16:35 <malmond> yet but I think packit/source-git could help reduce toil.
13:17:37 <ttomecek> malmond, wow, you seem to arrive to the right place :) is your current patch solution public? can we take a look?
13:18:51 <ttomecek> #topic The purpose of the SIG
13:18:55 <malmond> Our current patch solution is manually rebasing patches and working using cbs. We've only been using it for a few months. Previous to this we have an internal tool called 'yummy' which automated some stuff, but was entirely proprietry
13:20:03 <ttomecek> malmond, ah, okay; because I recall that you had an organization on github with maintenance branches, all tracked as sources; so I was wondering if that's still the case
13:21:32 <ttomecek> as for the next meeting topic, the purpose of this SIG: we would like to build a minimal viable product for the source-based workflow: https://fedoraproject.org/wiki/SIGs/Source-git#MVP
13:21:33 <malmond> We do have a bunch of stuff on github. I think you might be referring to https://github.com/facebookincubator/rpm-backports/ - this is pre Hyperscale SIG and most things are now elsewhere (see https://github.com/facebookincubator/rpm-backports/blob/master/MIGRATION.md)
13:22:01 <ttomecek> #link https://fedoraproject.org/wiki/SIGs/Source-git#MVP
13:22:08 <ttomecek> #link https://github.com/facebookincubator/rpm-backports/
13:22:16 <ttomecek> #link https://github.com/facebookincubator/rpm-backports/blob/master/MIGRATION.md
13:22:39 * ttomecek is using zodbot for the first time as a meeting host
13:22:53 <malmond> one more time than me :D
13:24:03 <ttomecek> malmond, thank you! do you already have a custom kernel at facebook?
13:25:38 <ttomecek> wow, this is pretty beefy: https://git.centos.org/rpms/systemd/commits/c8s-sig-hyperscale
13:26:19 <malmond> Sort of: We have much "newer" kernels available for CentOS. The source work in upstream, but the patches are backported to point releases. I'm not super familiar with the process used by the kernel devs. I'd like to see the kernel included in the Hyperscale SIG, but I don't know about timelines or even if we're committing to it yet
13:27:10 <Eighth_Doctor> We're definitely planning on a kernel for Hyperscale, we don't know which one yet
13:27:47 <Eighth_Doctor> at this point in time, we're planning to maintain split source packaging (dist-git-style) rather than merged source packaging (source-git-style)
13:28:15 <Eighth_Doctor> that was mostly from my experience contributing to kernel-ark, which kind of sucked :(
13:28:53 <ttomecek> Eighth_Doctor, what was that didn't work for you?
13:29:00 <Eighth_Doctor> oh god, pretty much everything
13:29:05 <Eighth_Doctor> even cloning the repo was terrible
13:29:16 <Eighth_Doctor> it timed out so often, and keeping track of commits is just impossible
13:30:05 <ttomecek> yeah, cloning kernel is not easy, a lot of git forges just can't handle a repo that big ( flachman: hi :)
13:30:16 <csomh> Eighth_Doctor, what do you mean by "keeping track of commits is just impossible"?
13:30:22 <Eighth_Doctor> and the weird merge workflow means nobody outside of the kernel team can understand it
13:30:39 <Eighth_Doctor> csomh: I cannot figure out how the kernel changes between upstream and downstream
13:31:11 <Eighth_Doctor> I cannot figure out what is for what, how they landed, and importantly, what order they landed in
13:31:50 <ttomecek> #info Complex merge workflows of kernel-ark make it hard for external contributors to perform correct contributions.
13:31:56 <jforbes> Eighth_Doctor: A lot of that has gotten better now, everything happens in os-build now
13:32:29 <ttomecek> #undo
13:32:29 <zodbot> Removing item from minutes: INFO by ttomecek at 13:31:50 : Complex merge workflows of kernel-ark make it hard for external contributors to perform correct contributions.
13:32:38 <Eighth_Doctor> yeah, but you are still merging things into that branch basically between upstream and downstream
13:32:48 <Eighth_Doctor> so I can't easily derive patch series deltas
13:32:53 <jforbes> Order is pretty clear, but even if you don't want to dig through git, there is a Patchlist.changelog which lists the downstream patches, in order, with git hash
13:33:23 <ttomecek> #info Complex merge workflows of kernel-ark made it hard for external contributors to perform correct contributions in the past. This is now better that things happen in os-build now.
13:33:57 <ttomecek> anyway, this is a clear indication, that whatever we build, we need to make sure it is well documented and understandable
13:34:01 <jforbes> Yes, though that is supposed to get rebased every upstream release.  Problem has been too many open MRs, but we got that from over 100 down to 35, so maybe we can start the rebase series for 5.13. Of course the fedora-5.11 branch was a rebase, fedora-5.12 will too
13:34:01 <Eighth_Doctor> I guess I'll have to look at it again, but we originally dismissed ark for Hyperscale because it was too complicated to figure out
13:34:31 <jforbes> Cloning still takes a while, that is gitlab's issue, not sure what we can do about it
13:34:32 <Eighth_Doctor> the idea would be that we'd work from a kernel branch from ark and contribute backports that we cared about
13:34:49 <Eighth_Doctor> and use that for the hyperscale kernel
13:35:02 <Eighth_Doctor> but none of us could figure out how to reasonably implement that workflow
13:35:49 <ttomecek> jforbes: flachman is actually working on a caching solution right now so we can reuse git objects from a local clone during the cloning process in CI (and potentially for contributors locally as well)
13:35:52 <Eighth_Doctor> e.g. assuming we went with 5.12, we'd work from that branched one and contribute backports as needed for things we cared about, like btrfs
13:35:53 <jforbes> Eighth_Doctor: I can help walk you through some of it.  The docs need to be better, but it isn't too difficult
13:36:05 <Eighth_Doctor> that'd be great :)
13:36:19 <jforbes> I am going on a motorcycle trip in a few minutes, will be back Monday
13:36:24 <Eighth_Doctor> by jforbes :)
13:36:28 <Eighth_Doctor> bye even :)
13:36:37 <jforbes> Well, around some in the AM, but from a hotel, so not sure how reliable
13:36:49 <ttomecek> jforbes, thanks for joining and enjoy your time off!
13:37:13 <Eighth_Doctor> but yeah, to make it clear, a big part of the goal for Hyperscale is to bring our expertise to the wider Enterprise Linux community and help support features and technologies we care about there
13:37:47 <ttomecek> my hope is that the workflow we build here should be usable by kernel-ark and userspace packages at the same time
13:38:14 <Eighth_Doctor> so in my case, virtualization and package management are my core concerns
13:38:21 <Eighth_Doctor> in malmond's case, package management
13:38:23 <ttomecek> Eighth_Doctor, nice, do you write some reports of work you are doing? would love to read about it
13:38:30 <Eighth_Doctor> yup
13:38:33 <Eighth_Doctor> one sec
13:38:51 <Eighth_Doctor> #link https://wiki.centos.org/SpecialInterestGroup/Hyperscale
13:39:13 <Eighth_Doctor> #link https://blog.centos.org/2021/04/centos-hyperscale-sig-quarterly-report/
13:39:43 <Eighth_Doctor> the first link covers the Hyperscale SIG effort, and the second covers what we've done so far
13:39:45 <ttomecek> Eighth_Doctor, wow, that's really impressive, thank you
13:40:01 <Eighth_Doctor> #link https://pagure.io/centos-sig-hyperscale/sig
13:40:21 <Eighth_Doctor> And this is our project tracking repository where we do all our planning and work tracking
13:40:22 <ttomecek> since we only have 20 minutes left, I'd love to discuss organization stuff before we part our ways
13:40:42 <ttomecek> #topic how do we communicate and track work?
13:41:03 <ttomecek> and I'd really appreciate some of your expertise here, Eighth_Doctor
13:41:19 <ttomecek> what worked for you in terms of running SIGs, collaboration and communication
13:41:45 <Eighth_Doctor> I've been driving a couple of SIGs that I chair using Pagure projects, video calls, and Matrix rooms for realtime chat
13:42:07 <Eighth_Doctor> Workstation WG and KDE SIG work this way and it's been phenomenal
13:42:30 <Eighth_Doctor> Hyperscale SIG is not quite up and going to do all that yet, but we're slowly getting everything set up
13:43:24 <ttomecek> Eighth_Doctor, why did you choose Matrix over IRC?
13:43:53 <ttomecek> I can see that a common practice is to create a pagure.io group and track work there
13:43:55 <Eighth_Doctor> more accessible for people and accounts for the increasing usage of multiple devices
13:44:10 <Eighth_Doctor> Fedora as a whole is migrating to Matrix very soon too
13:44:12 <ttomecek> what would others prefer?
13:44:27 <Eighth_Doctor> (IRC bridges are expected to be maintained, but Matrix is going to be the primary realtime chat)
13:44:39 <Eighth_Doctor> #link https://discussion.fedoraproject.org/t/the-future-of-real-time-chat-discussion-for-the-fedora-council/24628
13:44:40 * ttomecek didn't know about Fedora using Matrix officially
13:45:03 <Eighth_Doctor> we're waiting on Red Hat to pay for the server so we can start the process...
13:46:19 <ttomecek> Eighth_Doctor, thanks for the update!
13:46:27 <Eighth_Doctor> ttomecek: for pagure based project management, I'm particularly proud of the fedora-btrfs project: https://pagure.io/fedora-btrfs/project
13:46:47 <Eighth_Doctor> you can see the issues and boards and how we've organized things
13:46:53 <ttomecek> so would it work for everybody if we had a dedicated pagure.io group? (pretty much mirror how the Hyperscale SIG is organized)
13:47:00 <ttomecek> and a dedicated IRC channel, for now
13:47:24 <ttomecek> #link https://pagure.io/fedora-btrfs/project
13:47:29 <Eighth_Doctor> I'm cool with that
13:47:41 <flachman> +1
13:47:45 <Eighth_Doctor> if you need help with setup, feel free to hit me up and I can help :)
13:47:48 <malmond> +1
13:47:51 <Eighth_Doctor> +1
13:47:52 <ttomecek> #info Neal likes how the fedora-btrfs project is organized
13:48:37 <ttomecek> oki, no minuses, so let's do that :) I'll set that up (unless anyone else wants to do that :D)
13:50:18 <ttomecek> one problem with project management for us is that we already have somewhat-internal jira board, github project and a dozen of repositories with issues, internal gitlab repos so for our team, it would be _another_ source of work - I'll spend some time thinking how we to integrate that into our team's current development workflow
13:51:02 <ttomecek> #action ttomecek to set up https://pagure.io/group/fedora-sig-source-git and #fedora-source-git @ freenode
13:51:38 <ttomecek> Eighth_Doctor, thank you!
13:52:02 <Eighth_Doctor> ttomecek: you should ask nb for help to plumb it through to Matrix
13:52:34 <ttomecek> Eighth_Doctor, I'd need to poke Matrix first, I have never seen it :D
13:52:41 <Eighth_Doctor> 😀
13:53:08 <ttomecek> #action ttomecek to check out Matrix and plug the IRC channel into it; ask nb for help
13:53:24 <ttomecek> and finally: meetings
13:53:30 <ttomecek> any preference?
13:53:46 <Eighth_Doctor> I prefer video meetings, but text meetings are fine too
13:54:32 <ttomecek> Eighth_Doctor, what platform you like the most?
13:54:47 <ttomecek> we're used to google meet in our team
13:54:52 <csomh> yep, video would be better indeed, with some notes that could be shared afterwards
13:54:55 <Eighth_Doctor> Bluejeans or Google Meet are fine
13:55:11 <Eighth_Doctor> Bluejeans is often easier for people who don't necessarily have google accounts
13:55:26 <Eighth_Doctor> (which may be rare among us, anyway)
13:55:56 <ttomecek> I can see that Hyperscale meets biweekly, which I'd prefer personally
13:56:04 <malmond> +1 on the meetings sentiments
13:56:14 <Eighth_Doctor> yeah, I don't know if source-git stuff would change so much to need weekly meetings
13:56:25 <ttomecek> it also seems that a video meeting would be way more efficient if I look at the amount of information in here
13:57:17 <ttomecek> awesome! another agreement, I hope that the workflow definition and implementation will go so smoothly as well
13:58:29 <ttomecek> would it be better to start one hour later than we started this one? (meaning in 2 minutes)
13:58:47 <malmond> +1 to that :D
13:59:56 <ttomecek> I'll do another when-is-good for everyone who joined the meeting today so we can pick time in 2 weeks and do it via google meet
14:00:00 <Eighth_Doctor> ideally not on Thursday then
14:00:13 <Eighth_Doctor> since Fedora Social Hour every other Thursday is at this time
14:00:31 <ttomecek> in the meantime, we can start discussions on irc and on the pagure issue tracker[s]
14:00:37 <Eighth_Doctor> yup
14:01:14 <ttomecek> we're one minute over, thank you all for joining!
14:01:31 <ttomecek> I'm gonna end this, let me know if you have anything else to discuss
14:01:36 <ttomecek> #endmeeting