16:29:50 <dustymabe> #startmeeting fedora_coreos_meeting
16:29:50 <zodbot> Meeting started Wed Mar 18 16:29:50 2020 UTC.
16:29:50 <zodbot> This meeting is logged and archived in a public location.
16:29:50 <zodbot> The chair is dustymabe. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:29:50 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
16:29:50 <zodbot> The meeting name has been set to 'fedora_coreos_meeting'
16:29:51 <jlebon> .hello2
16:29:52 <zodbot> jlebon: jlebon 'None' <jonathan@jlebon.com>
16:29:55 <dustymabe> #topic roll call
16:32:13 <jlebon> hmm, i guess i have to do this again?
16:32:15 <jlebon> .hello2
16:32:16 <zodbot> jlebon: jlebon 'None' <jonathan@jlebon.com>
16:32:19 <dustymabe> #chair jlebon
16:32:19 <zodbot> Current chairs: dustymabe jlebon
16:32:40 <dustymabe> skunkerk: I forget what was the right invocation for you: .hello <fasusername>
16:32:40 <davdunc> .hello2
16:32:41 <zodbot> davdunc: davdunc 'David Duncan' <davdunc@amazon.com>
16:32:44 <bgilbert> .hello2
16:32:45 <zodbot> bgilbert: bgilbert 'Benjamin Gilbert' <bgilbert@backtick.net>
16:32:48 <dustymabe> well high there davdunc
16:32:53 * bgilbert waves at davdunc
16:32:55 <davdunc> hi dustymabe
16:33:06 <davdunc> hey everybody.
16:33:11 <ksinny> .hello sinnykumari
16:33:12 <zodbot> ksinny: sinnykumari 'Sinny Kumari' <ksinny@gmail.com>
16:33:21 <cyberpear> .hello2
16:33:22 <zodbot> cyberpear: cyberpear 'James Cassell' <fedoraproject@cyberpear.com>
16:33:34 <skunkerk> .hello sohank2602
16:33:35 <zodbot> skunkerk: sohank2602 'Sohan Kunkerkar' <skunkerk@redhat.com>
16:33:46 <dustymabe> #chair skunkerk davdunc ksinny cyberpear bgilbert
16:33:46 <zodbot> Current chairs: bgilbert cyberpear davdunc dustymabe jlebon ksinny skunkerk
16:33:47 <mnguyen_> .hello mnguyen
16:33:48 <zodbot> mnguyen_: mnguyen 'Michael Nguyen' <mnguyen@redhat.com>
16:33:50 <dustymabe> #chair mnguyen_
16:33:50 <zodbot> Current chairs: bgilbert cyberpear davdunc dustymabe jlebon ksinny mnguyen_ skunkerk
16:34:18 <dustymabe> thanks all for coming!
16:34:22 <dustymabe> i'll get started now
16:34:45 <dustymabe> #topic Fedora CoreOS content for Community Central for RH Virtual Summit
16:34:51 <dustymabe> #link https://github.com/coreos/fedora-coreos-tracker/issues/416
16:35:18 <dustymabe> so RH summit is a virtual event this year
16:35:53 <dustymabe> and one of our community leaders for Fedora is asking if we have any ideas for ways to show off FCOS in a demo
16:36:37 <dustymabe> one of the features of FCOS (I think) is that it can be underwhelming depending on who you are, so this might be tough
16:36:50 <dustymabe> but we do a lot of inovative things with the OS
16:36:54 <jlebon> maybe at least one demo showing automatic updates?
16:37:15 <red_beard> .hello redbeard
16:37:16 <zodbot> red_beard: redbeard 'Brian 'redbeard' Harrington' <bharring@redhat.com>
16:37:17 <dustymabe> jlebon: +1
16:37:27 <jlebon> going into it too with the cincinnati graph could be interesting
16:37:53 <dustymabe> jlebon: i.e. some of the mechanics behind the updates ?
16:38:07 <jlebon> assuming people aren't familiar with ostree, just showing `rpm-ostree status` and explaining the output is always nice
16:38:28 <jlebon> dustymabe: right, exactly
16:38:39 <jlebon> just a bit to show it's not just a dumb polling thing
16:38:40 <red_beard> honestly before that means anything to folks i think there's more brick laying to be done on going over why the OS is important.
16:39:06 <dustymabe> jlebon: did we get readonly sysroot working? we could demo `rm -rf /` not breaking the system
16:39:16 <red_beard> there's more than enough non-working demoware that gets cooked up every single year for summit.
16:39:35 <cyberpear> lol, you'll still lose all user data
16:39:45 <jlebon> dustymabe: yup good idea
16:39:56 <dustymabe> cyberpear: who needs data :)
16:40:27 <dustymabe> red_beard: any concrete ideas there? unfortunately we don't have a lot of time to do much more than short demos
16:40:36 <dustymabe> but would love to do that
16:40:45 <miabbott> .hello miabbott
16:40:46 <zodbot> miabbott: miabbott 'Micah Abbott' <miabbott@redhat.com>
16:41:43 <dustymabe> does anybody think a demo/overview of ignition would be useful?
16:42:54 <dustymabe> #chair red_beard miabbott
16:42:54 <zodbot> Current chairs: bgilbert cyberpear davdunc dustymabe jlebon ksinny miabbott mnguyen_ red_beard skunkerk
16:42:58 <ksinny> I think so, perhaps showcasing ignition together with FCOS installation?
16:43:05 <jlebon> dustymabe: yeah, i think so
16:43:33 <red_beard> dustymabe, i guess my point was lost.  a demo just for the sake of having a demo isn't very interesting, especially when the majority of the audience doesn't understand why what you're building matters.
16:43:48 <dustymabe> TBH I think the value of ignition is really illustrated when you do an openshift install - it's crazy how much information is transmitted and applied through Ignition
16:44:15 <red_beard> if it's just to satisfy the community leader's itch, then they should already have an idea of what will resonate.  if they don't have an idea, than there is no cohesive narritive that's being conveyed.
16:44:32 <dustymabe> red_beard: so you think it'd be worth having a full talk about FCOS, what the benefits are, and how we got here ?
16:45:07 <red_beard> i do.  imagine that you _don't_ have that context... what is a demo actually showing off?
16:45:14 <miabbott> to red_beard's point, we should demonstrate the fault tolerance of the update mechanism of the OS with container workloads in place.  in theory, we should be able to pull the plug on the system in the middle of an update and it will boot back up normally, with the workloads resumed
16:45:18 <red_beard> it's no different than yum update in a cron job
16:45:45 <red_beard> or... show off an intentionally broken update and show the system recovering
16:46:06 <jlebon> right yeah, i think for the full effect, a presenter would need to cover the overall message while interspersing demos
16:46:46 <dustymabe> jlebon: yeah that's probably best
16:47:18 <jlebon> that might be more work than the person who opened that issue initially imagined though
16:47:38 <dustymabe> yeah, i mean usually the community booth has talks at summit
16:47:46 <dustymabe> they can range from 30-45 minutes
16:47:58 <dustymabe> and they also do short demos at the booth itself for smaller audiences
16:48:08 <dustymabe> so this could probably follow either of those forms
16:48:11 <jlebon> ohh gotcha. ok yup, that's much better
16:48:31 <dustymabe> we could have 1 "talk" that covers the overview along with interspersed demos
16:48:44 <dustymabe> and we could also cut the demos out into short 5-10 minute clips to share separately
16:48:48 <dustymabe> and link back to the talk
16:49:27 <dustymabe> i'll enumerate some of these items.. in the meantime, any volunteers?
16:50:06 <dustymabe> - show off ignition (simple example) + maybe show how much it's leveraged during an openshift/OKD install
16:50:24 <dustymabe> - show off automatic updates
16:50:34 <dustymabe> - show off a failed update and being able to recover
16:51:00 <dustymabe> - show off immutability (possibly rm / protection)
16:51:13 <jlebon> re. "- show off a failed update and being able to recover" - that... doesn't work yet.  unless we're talking about manually rolling back
16:51:25 <dustymabe> i think manually rolling back would suffice for now
16:51:32 <dustymabe> i know it's not where we want to be
16:51:35 <jlebon> ack
16:52:23 <dustymabe> - show off rpm-ostree status output (i.e. rich information about the OS)
16:52:49 <dustymabe> anything else ?
16:53:56 <davdunc> a bare-metal install?
16:54:12 <dustymabe> ok i'll put some of these in the ticket - i'll also ask riecatnor what the deadline is for something like this
16:54:31 <dustymabe> davdunc: interesting suggestion from the AWS guy :)
16:54:50 <davdunc> :D hey! hybrid is real.
16:55:08 <dustymabe> davdunc: i'm not sure people would care much - probably assume bare-metal is table stakes
16:55:52 <dustymabe> but I do suppose using Ignition with bare metal is a fairly new concept (i.e. something like cloud-init for non-cloud)
16:56:08 <dustymabe> well, not brand new, but a lot of people probably don't think about it
16:56:27 <davdunc> but it is very different from where they may have learned to use  coreos and with the impending support eol, they may want to know all the options.
16:56:58 <dustymabe> +1
16:57:43 <dustymabe> #info dustymabe will update #416 with suggestions for talk/demo for the Red Hat Summit virtual event and ask about deadline for such content
16:57:55 <dustymabe> anything else before the next topic ?
16:58:57 <dustymabe> #topic talks for upcoming conferences
16:59:04 <dustymabe> #link https://github.com/coreos/fedora-coreos-tracker/issues/420
16:59:11 <dustymabe> #link https://github.com/coreos/fedora-coreos-tracker/issues/421
16:59:58 <dustymabe> assuming pandemic stuff calms down, we have a few conferences later this year where we can show off FCOS
17:00:17 <dustymabe> any ideas for talks we could give to help raise awareness?
17:01:21 <jlebon> where a "state of FCOS" talk for flock?
17:01:29 <jlebon> s/where/maybe/  not sure what happened there :)
17:01:40 <dustymabe> :)
17:02:00 <dustymabe> that's a good idea - obstacles overcome, current state, where we're headed, etc..
17:02:33 <bgilbert> +1
17:02:34 <jlebon> hmm, a more technical talk on the osmet stuff could be interesting
17:02:38 <dustymabe> for devconf.us maybe it's more "Intro to FCOS" since there's a decent audience that probably still doens't know about it
17:03:16 <dustymabe> i really liked luca's talk at devconf.cz about the update service
17:03:41 <dustymabe> he *could* maybe run that again for a different audience if he was willing to travel
17:03:44 <jlebon> oh thanks for the reminder, i've been meaning to see that still!
17:04:39 <dustymabe> a talk on live PXE could be interesting
17:05:03 <dustymabe> i feel like there's a good number of people in our industry who've never used it and might like the benefits
17:05:37 <jlebon> i think walters did a talk on that at devconf too
17:05:57 <dustymabe> yeah he did I think.. and talked about persisting data and stuff
17:06:02 <jlebon> but yeah, still worth it for a different audience
17:06:23 <davdunc> A deep-dive on ignition would be great.
17:06:45 <dustymabe> davdunc: yeah I think andrew had prepared some slides for a talk on that
17:06:58 <dustymabe> jlebon: did you ever get access to his talk material ^^ ?
17:07:15 <jlebon> dustymabe: i don't think so, but i'll double check
17:07:27 <jlebon> walters might've
17:07:41 <dustymabe> cool. ok these are all good ideas. please add them to the tickets and we'll try to get some talks submitted!
17:08:03 <dustymabe> #topic Integrate FCOS better into the existing Fedora release process
17:08:11 <dustymabe> #link https://github.com/coreos/fedora-coreos-tracker/issues/426
17:08:20 <dustymabe> jlebon: want to add the context here?
17:08:46 <jlebon> sure
17:09:24 <cyberpear> would also be good to distribute on the normal mirrors so folks can browse what's there and easily sync locally
17:09:38 <jlebon> there were discussions yesterday on the devel fedora list re. what we can do to improve integration into FCOS
17:09:48 <jlebon> and integration here doesn't really refer to the technical stuff
17:10:13 <jlebon> but overall messaging, vision, marketing as a Fedora edition
17:10:38 <bgilbert> cyberpear: browsing is intentionally disabled.  we want users to access artifacts via links in the stream metadata so they're always running the recommended version.
17:10:58 <jlebon> i'd wager the majority of Fedora contributors don't really have any idea what we're doing
17:10:59 <cyberpear> bgilbert: it's user- hostile, IMO
17:11:42 <jlebon> cyberpear: the bucket layout can (and has) change(d)
17:11:48 <bgilbert> also, in general there's not much reason to download all the artifacts
17:11:49 <dustymabe> jlebon: you're right we are kind of silo'd right now
17:12:02 <cyberpear> bgilbert: and now requires a special parser to sync locally, vs rsync or wget
17:12:14 <red_beard> cyberpear, i'm a bit "new" to the os-tree side of things, but in my experience browsing that content (os-tree updates) didn't seem valuable.  what was i missing when looking at it?
17:12:32 <bgilbert> `coreos-installer download` can handle fetches for you
17:12:45 <red_beard> whoops, nm.  didn't auto-scroll.
17:12:52 <dustymabe> cyberpear: i think he's talking more about our image artifacts (bare metal, qemu, aws, etc)
17:12:56 <cyberpear> red_beard: it's more about being able to sync the thing locally without smart tools
17:13:03 <dustymabe> oops, meant red_beard
17:13:32 <jlebon> not sure if there's any super actionable thing for us to take right now other than continue the discussion with Fedora leads
17:13:46 <cyberpear> but you're right, the browsing isn't good, but I'll let it go back on topic
17:14:26 <dustymabe> jlebon: +1 - i responded to the email thread.. TL;DR - integrating with the rest of Fedora is something that we (our SIG/working group) will get better at as the project matures
17:15:30 <dustymabe> cyberpear: re: browsing, I'm not personally oposed to making the directories more browsable, it is a decision we made early on, but if there's real benefit to opening them up maybe we can reconsider
17:15:31 <cyberpear> for actionable on topic, it would be nice to have the assembly process happen nightly with the compose, to be integrated...
17:16:22 <dustymabe> cyberpear: regarding the "compose", unfortunately I don't think we'll ever be part of Fedora's nightly compose
17:16:47 <dustymabe> we just operate on a different cadense with the `stable`, `testing`, `next` streams
17:16:57 <dustymabe> ignore spelling mistakes :)
17:17:08 <cyberpear> why not?  I understand that cosa has to come closer to the end, but...
17:17:13 <jlebon> somewhat related: https://pagure.io/fedora-infrastructure/issue/8370
17:17:26 <cyberpear> how about a nightly stream?
17:17:29 <jlebon> if we start running on fedora infra, i think visibility into the process will naturally increase
17:17:40 <jlebon> cyberpear: we have that (more than just nightly in fact :)  )
17:17:55 <jlebon> https://builds.coreos.fedoraproject.org/browser
17:17:59 * dustymabe finds link to the design doc
17:18:08 <cyberpear> jlebon: so make it part of compose?
17:18:17 <dustymabe> https://github.com/coreos/fedora-coreos-tracker/blob/master/stream-tooling.md#introduction
17:18:54 <bgilbert> cyberpear: what's the value?
17:19:06 <jlebon> cyberpear: the root difference here is that fcos is not created by pungi. trying to merge those artifacts would be pretty artificial
17:19:06 <dustymabe> cyberpear: more or less, we can't be a part of Fedora's regular compose because our inputs are different for most of our streams
17:19:27 <cyberpear> bgilbert: better integration with rest of Fedora
17:19:30 <dustymabe> so being a part of Fedora's nightly for one stream wouldn't make much sense when all of our other ones run a different way
17:19:35 <bgilbert> cyberpear: what does that mean, though?
17:19:54 <bgilbert> in large part, FCOS exists to experiment with doing things differently
17:22:10 <dustymabe> cyberpear: i'm happy to hop on a call (higher bandwidth) to try to justify the split too
17:24:48 <cyberpear> I've been around Fedora a lot longer that CoreOS, but I'm probably not the most qualified to make these arguments
17:25:22 <dustymabe> yeah. part of the problem we were faced with (with FCOS) is merging two communities together
17:25:29 <dustymabe> which is a good and a bad thing
17:25:48 <dustymabe> Fedora Atomic Host, was a bit more like Fedora and could follow the Fedora model better
17:26:03 <dustymabe> CoreOS Container Linux obviously was not part of Fedora
17:26:26 <dustymabe> we're trying to take the best of both worlds and creat a new community out of it
17:26:36 <dustymabe> one of those is the "stream" model along with automatic updates
17:27:14 <dustymabe> we deemed early on that there's just not a reliable way to take bodhi updates and testing and get the `alpha`, `beta`, `stable` model that CL had
17:27:34 <dustymabe> so we had to modify our approach slightly and this is what we've come up with for now
17:28:10 * dustymabe notes the time
17:28:22 <dustymabe> anything else for this topic before we move to open floor ?
17:29:18 <dustymabe> #topic open floor
17:29:36 <dustymabe> davdunc: any chance you've got anything for us :)
17:30:09 <davdunc> I am just working on getting started again.
17:30:42 <dustymabe> +1
17:30:45 <dustymabe> glad to have you here
17:30:56 <dustymabe> anyone with any topics for open floor?
17:30:57 <davdunc> dustymabe: I'd like to make sure that you have access to the aarch64 instance types we have though.
17:31:12 <dustymabe> davdunc: new ones ?
17:31:19 <davdunc> yes.
17:31:45 <davdunc> I can whitelist your eng account.
17:31:48 <dustymabe> davdunc: the cart might be before the horse on this one a bit :) - we don't yet have aarch64 content from the FCOS pipeline
17:31:56 <davdunc> right.
17:31:57 <dustymabe> we do for Fedora Cloud base
17:32:05 <dustymabe> but I can give you account numbers if you want to whitelist things
17:32:17 <davdunc> I do. you can email them.
17:32:23 <dustymabe> sounds good
17:32:41 * dustymabe will close meeting in 60 seconds unless a new topic comes in
17:33:39 <dustymabe> #endmeeting