#meeting-1:fedoraproject.org log

<@aaradhak:matrix.org>

15:30:05

!startmeeting fedora_coreos_meeting

<@meetbot:fedora.im>

15:30:07

Meeting started at 2025-08-13 15:30:05 UTC

<@meetbot:fedora.im>

15:30:07

The Meeting name is 'fedora_coreos_meeting'

<@aaradhak:matrix.org>

15:30:14

!topic roll call

<@aaradhak:matrix.org>

15:30:19

!hi aaradhak

<@zodbot:fedora.im>

15:30:20

Aashish Radhakrishnan (aaradhak)

<@aaradhak:matrix.org>

15:30:38

Hello Everyone!

<@hricky:fedora.im>

15:30:44

!hi

<@dustymabe:matrix.org>

15:30:45

!hi

<@zodbot:fedora.im>

15:30:46

Hristo Marinov (hricky) - he / him / his

<@zodbot:fedora.im>

15:30:48

Dusty Mabe (dustymabe) - he / him / his

<@ydesouza:matrix.org>

15:31:43

!hi

<@zodbot:fedora.im>

15:31:46

No Fedora Accounts users have the @ydesouza:matrix.org Matrix Account defined

<@ydesouza:fedora.im>

15:32:44

hi!

<@marmijo:fedora.im>

15:33:06

!hi

<@zodbot:fedora.im>

15:33:07

Michael Armijo (marmijo)

<@dustymabe:matrix.org>

15:33:17

Yasmin, we are going to have to let your two accounts fight to the death and allow one of them to emerge victorious :)

<@ydesouza:fedora.im>

15:33:51

I guess is the only way! haha

<@jlebon:fedora.im>

15:34:35

!hi

<@zodbot:fedora.im>

15:34:36

None (jlebon)

<@aaradhak:matrix.org>

15:35:02

Alright may be we can get started now..

<@aaradhak:matrix.org>

15:35:16

<@aaradhak:matrix.org>

15:35:16

<@aaradhak:matrix.org>

15:35:16

!topic Action items from last meeting

<@aaradhak:matrix.org>

15:36:00

Looks like there is no Action item from the last meeting

<@aaradhak:matrix.org>

15:36:17

!topic Review Fedora 43 Release Schedule

<@aaradhak:matrix.org>

15:36:23

!link https://fedorapeople.org/groups/schedule/f-43/f-43-key-tasks.html

<@aaradhak:matrix.org>

15:37:33

The deadline for Branch Fedora Linux 43 from Rawhide was yesterday. I believe it is already done

<@aaradhak:matrix.org>

15:38:55

https://docs.fedoraproject.org/en-US/fesco/Fails_to_build_from_source_Fails_to_install/

<@aaradhak:matrix.org>

15:38:55

also Retire FTI packages in NEW state by 08-19 before our next community meeting.

<@aaradhak:matrix.org>

15:38:55

Next up is Post-branch Freeze that needs to be done by 08-15 and

<@dustymabe:matrix.org>

15:39:28

I know marmijo opened a few PRs to get us rolling

<@marmijo:fedora.im>

15:40:05

Yes, branching happened yesterday. There are still a few PRs to merge, but CI is failing one one and I'm looking into today.

<@aaradhak:matrix.org>

15:41:51

ok thanks for the update!

<@aaradhak:matrix.org>

15:43:33

If there is nothing else to add , may be we can move on to the next topic

<@aaradhak:matrix.org>

15:43:56

!topic Allow passing in Butane configs directly to instances

<@aaradhak:matrix.org>

15:43:56

!link https://github.com/coreos/fedora-coreos-tracker/issues/200

<@dustymabe:matrix.org>

15:44:23

👋

<@dustymabe:matrix.org>

15:44:44

I opened this issue upstream.

<@dustymabe:matrix.org>

15:45:21

The basic question at a high level is:

<@dustymabe:matrix.org>

15:45:21

<@dustymabe:matrix.org>

15:45:21

Should we allow our instances to be able to parse butane configs as well as Ignition configs on boot?

<@dustymabe:matrix.org>

15:46:01

I've had this idea for a long time, but was under the impression it would be a lot of work to pull off.

<@dustymabe:matrix.org>

15:46:36

and while it may actually be a lot of work in the end, at least a PoC was pretty easy to get up and running (actually working way better than I expected)

<@dustymabe:matrix.org>

15:47:40

I know we've all talked to users (or even ourselves) who have had some pain related to having a "transpile" step

<@dustymabe:matrix.org>

15:48:12

What are the perceived benefits people can see? What are the negatives of such an approach?

<@dustymabe:matrix.org>

15:48:12

<@dustymabe:matrix.org>

15:48:12

Should we consider trying to solve that pain point?

<@dustymabe:matrix.org>

15:48:33

Also very interested in the perspective of the flatcar folks on this? cc tormath1

<@hricky:fedora.im>

15:49:13

I think the correct link to the issue in question is https://github.com/coreos/fedora-coreos-tracker/issues/2006

<@dustymabe:matrix.org>

15:49:35

!link https://github.com/coreos/fedora-coreos-tracker/issues/2006

<@aaradhak:matrix.org>

15:50:02

Yes thats right!

<@tormath1:matrix.org>

15:50:29

<@tormath1:matrix.org>

15:50:29

Folks understand and they get use to it.

<@tormath1:matrix.org>

15:50:29

* Provide an extra check *before* starting the instance

<@tormath1:matrix.org>

15:50:29

Most of the time, when you explain why:

<@tormath1:matrix.org>

15:50:29

* Provide some additional sugar (e.g it generates the mount units)

<@tormath1:matrix.org>

15:50:29

Yeah, thanks for the PoC first. This is something we've been asked quite some times in Flatcar workshop: why do we need two steps?

<@tormath1:matrix.org>

15:50:39

Folks understand and they get used to it.

<@tormath1:matrix.org>

15:50:39

<@tormath1:matrix.org>

15:50:39

- Provide an extra check _before_ starting the instance

<@tormath1:matrix.org>

15:50:39

- Provide some additional sugar (e.g it generates the mount units)

<@tormath1:matrix.org>

15:50:39

<@tormath1:matrix.org>

15:50:39

Most of the time, when you explain why:

<@tormath1:matrix.org>

15:50:39

Yeah, thanks for the PoC first. This is something we've been asked quite some times in Flatcar workshop: why do we need two steps?

<@jlebon:fedora.im>

15:51:25

yeah, the extra check part is underappreciated. the cost of spinning up a VM only to find out e.g. you wrote JSON wrong is annoying

<@jlebon:fedora.im>

15:51:38

that said, writing YAML is easier at least to write

<@tormath1:matrix.org>

15:52:21

What I've proposed on Mastodon: why should we bother to maintain both JSON and YAML spec?

<@jlebon:fedora.im>

15:52:26

but of course there's lots of other checks it does. and obviously, people can still run those. but it becomes easier not to if the machine just accepts your file as is

<@tormath1:matrix.org>

15:52:59

*i.e* no more JSON, the Ignition specs becomes the Butane spec.

<@tormath1:matrix.org>

15:53:07

_i.e_ no more JSON, the Ignition spec becomes the Butane spec.

<@jlebon:fedora.im>

15:53:53

tormath1: that gets into the other main difference, which is distro-specific sugar vs ignition being generic

<@dustymabe:matrix.org>

15:53:53

I guess one reason would be backwards compat. We'd be throwing away existing workflows people have set up if we started to barf when someone handed us JSON.

<@dustymabe:matrix.org>

15:53:53

tormath1: that's definitely a larger change, for sure..

<@dustymabe:matrix.org>

15:53:53

<@tormath1:matrix.org>

15:55:38

For the backward compatibility, we can still keep the JSON spec around (like we are currently doing on Flatcar for Ignition v2 vs /v3).

<@tormath1:matrix.org>

15:55:38

Bonus: we get closer from cloud-init spec (YAML), so it's easier for folks coming from Ubuntu & Cie to understand.

<@dustymabe:matrix.org>

15:56:29

tormath1: so your argument is more.. going forward we'd do spec bumps and such only in butane yaml - so we'd only focus on one area rather than two.

<@tormath1:matrix.org>

15:57:53

That would drop the Butane / Ignition matrix compatibility which to me brings useless complexity and more time to propagate new features (and bugs as well).

<@jlebon:fedora.im>

15:58:21

tormath1: the butane spec is already 90% the ignition spec. i think it comes back to that 10% distro-specific sugar

<@dustymabe:matrix.org>

15:58:46

I guess there's not really a decision we'd need to make here other than we just don't bump the ignition spec in the future unless we decide we need to

<@dustymabe:matrix.org>

15:59:23

if we decided there was a need to bump the ignition spec; then I guess we'd understand it would have to be worth the effort?

<@dustymabe:matrix.org>

16:00:20

Should we list out potential downsides of something like this?

<@tormath1:matrix.org>

16:00:22

I just think your PoC translates the issue here: what's the point of having two specs. if you can use one or the other with no effect on the end result.

<@dustymabe:matrix.org>

16:01:17

right, one is a superset?? - but yeah we can de-emphasize the "less featurefull" one in the future?

<@tormath1:matrix.org>

16:01:56

if tomorrow your PoC becomes stable, no one will use explicitly Butane

<@tormath1:matrix.org>

16:02:14

and YAML will become the *de-facto* Ignition configuration syntax.

<@dustymabe:matrix.org>

16:02:48

If that's true, then it speaks pretty strongly to what we think users actually want?

<@jlebon:fedora.im>

16:03:24

again, to be clear, this would be equivalent to making Ignition no longer distro-agnostic. but i guess we could add a "default" or "raw" variant?

<@tormath1:matrix.org>

16:04:04

I guess. I recall we actually got this conversation / idea for Flatcar a few years ago: what if we could auto-translate butane on the fly?

<@dustymabe:matrix.org>

16:04:06

<@dustymabe:matrix.org>

16:04:06

> Ignition no longer distro-agnostic

<@dustymabe:matrix.org>

16:04:06

Is that necessarily true?

<@jlebon:fedora.im>

16:04:45

all butane specs are tied to a distro

<@dustymabe:matrix.org>

16:04:52

<@dustymabe:matrix.org>

16:04:52

> what if we could auto-translate butane on the fly?

<@dustymabe:matrix.org>

16:04:52

That's basically what the PoC does.. Ignition itself doesn't understane the butane spec. It's just using the butane code as a library.

<@dustymabe:matrix.org>

16:06:00

Or is Igntion just leveraging another library that does support non-distro-agnostic configs?

<@dustymabe:matrix.org>

16:06:00

<@dustymabe:matrix.org>

16:06:00

Does Ignition understand the non-distro-agnostic bits?

<@dustymabe:matrix.org>

16:06:00

<@dustymabe:matrix.org>

16:06:00

Jonathan Lebon: right. I guess the distinction in my head is subtle

<@dustymabe:matrix.org>

16:06:29

i.e. the ignition binary is different than the ignition spec

<@dustymabe:matrix.org>

16:06:29

Maybe it's terminology.

<@dustymabe:matrix.org>

16:06:29

<@dustymabe:matrix.org>

16:06:45

the ignition spec (which the ignition binary interprets) is still distro agnostic

<@dustymabe:matrix.org>

16:07:13

the butane spec (which the ignition binary can translate into the ignition spec using a library) is not distro agnostic

<@jlebon:fedora.im>

16:08:18

right i'm speaking to tormath1's suggestion of folding the Butane specs into Ignition

<@jlebon:fedora.im>

16:10:19

there's definitely a bunch of core Butane sugars that aren't distro-specific (e.g. mount units) which could in theory be folded into the Ignition spec

<@jlebon:fedora.im>

16:10:51

some of the Butane sugar interestingly isn't really sugar per se but more... processing (e.g. the file embedding). that still requires running a binary before starting the system

<@dustymabe:matrix.org>

16:11:38

So I think we're mostly at implementation details, which are definitely important, but backing up a bit..

<@dustymabe:matrix.org>

16:11:38

<@dustymabe:matrix.org>

16:11:38

Does anyone think this is a bad idea?

<@dustymabe:matrix.org>

16:11:38

<@dustymabe:matrix.org>

16:12:07

So I think we're mostly at implementation details, which are definitely important, but backing up a bit..

<@dustymabe:matrix.org>

16:12:07

<@dustymabe:matrix.org>

16:12:07

<@dustymabe:matrix.org>

16:12:07

Does anyone think at a high level this is a bad idea?

<@jlebon:fedora.im>

16:12:27

i feel like i need to think about it more :)

<@dustymabe:matrix.org>

16:12:43

There could definitely be technical reasons not to do this (and we'll probably find some we have to work through), but just in general - would providing users with a UX like this be beneficial?

<@jlebon:fedora.im>

16:12:45

i definitely empathize with the UX issues of the current flow

<@tormath1:matrix.org>

16:13:00

From a maintainer perspective, it removes one spec to maintain.

<@tormath1:matrix.org>

16:13:00

From a user perspective, it's one less thing to understand and one less thing than could break.

<@dustymabe:matrix.org>

16:13:03

I think that's fair. I'm also really interested in spresti take on this

<@tormath1:matrix.org>

16:13:12

I need to sleep on it as well :D

<@dustymabe:matrix.org>

16:13:28

Maybe let's table it and bring it up again next week?

<@aaradhak:matrix.org>

16:13:54

spresti is out today.

<@tormath1:matrix.org>

16:14:16

dustymabe in the meantime you can maybe push your PoC in a wip PR or on a fork to make it visible

<@jlebon:fedora.im>

16:14:23

I think an easier discussion that might still address some UX friction is: "make Ignition support YAML, fold some sugars into Ignition"

<@dustymabe:matrix.org>

16:15:07

tormath1: definitely.. I'll try to make it clear that it's by no means a complete PR.

<@jlebon:fedora.im>

16:15:36

("support YAML" here basically means: still the Ignition spec, just written as YAML)

<@dustymabe:matrix.org>

16:15:44

that seems harder IMO, but might be better to maintain longer term if we decide to drop butane

<@jlebon:fedora.im>

16:15:47

which again is like 90% of Butane

<@dustymabe:matrix.org>

16:16:18

aaradhak: we can probably move to the next topic

<@tormath1:matrix.org>

16:16:31

(I'll ask Flatcar users / maintainers inputs as well)

<@dustymabe:matrix.org>

16:16:37

Thank you tormath1 for joining!

<@jlebon:fedora.im>

16:16:46

indeed, thanks for chiming in!

<@aaradhak:matrix.org>

16:16:52

Alright, lets table this discussion to the next meeting.

<@aaradhak:matrix.org>

16:17:24

!link https://github.com/coreos/fedora-coreos-tracker/issues/1999

<@aaradhak:matrix.org>

16:17:24

!topic Introducing the fcos-assistant Bot

<@aaradhak:matrix.org>

16:17:24

Moving on to the next topic..

<@aaradhak:matrix.org>

16:17:24

<@ydesouza:fedora.im>

16:18:49

So, I wanted to know what are the community inputs about having a AI bot in our channel or in or server as a separate room to answer questions.

<@ydesouza:fedora.im>

16:19:45

The idea is not handle complex question, and also will not use discussions, meetings and other channel messages as data. It will be used only our own documentation and web search.

<@dustymabe:matrix.org>

16:20:11

I'm a bit skeptical that we should throw AI at everything, but I think in targeted use cases it could be useful.

<@dustymabe:matrix.org>

16:20:45

> The idea is not handle complex question

<@dustymabe:matrix.org>

16:20:45

<@dustymabe:matrix.org>

16:20:45

well. We could throw complex questions at it at times to see if it's useful :)

<@hricky:fedora.im>

16:21:00

I think it should start in a separate room.

<@dustymabe:matrix.org>

16:22:28

Hristo Marinov: that's fair, but I'm interested to know what your concerns are (I have a few for sure, but am interested in what yours are)?

<@jlebon:fedora.im>

16:22:46

it might be a good exercise to do backtesting on real questions that came up in the channel and see how it performs. or did you do that already?

<@ydesouza:fedora.im>

16:23:52

I did not tried with our channel questions, but it's a good idea. For now the bot is down, still working on get a key and find a good way to host it.

<@hricky:fedora.im>

16:24:00

I think we can ask people to test it there. I also think we need some form of consent from the people who use it.

<@dustymabe:matrix.org>

16:24:57

<@dustymabe:matrix.org>

16:24:57

Hristo Marinov: the ticket states one option as:

<@dustymabe:matrix.org>

16:24:57

> Main channel: The bot would respond only when explicitly tagged and mark their answers with a robot emoji (e.g., @fcos-assistant 🤖).

<@dustymabe:matrix.org>

16:25:41

in that scenario AI actually wouldn't be looking at every message in the channel. Only ones were a human added the 🤖 emoji as a reaction to a message.

<@dustymabe:matrix.org>

16:26:17

I think that helps alleviate some of the concerns I had. because 1. AI isn't looking at everything. 2. some human had to request AI engage.

<@jlebon:fedora.im>

16:26:33

that seems reasonable. we would probably want to mention it in the topic though otherwise it's not very discoverable

<@dustymabe:matrix.org>

16:26:44

and then it would only respond in a thread so at least AI isn't polluting the main channel scrollback with a bunch of stuff

<@dustymabe:matrix.org>

16:26:58

in that scenario AI actually wouldn't be looking at every message in the channel. Only ones where a human added the 🤖 emoji as a reaction to a message.

<@jlebon:fedora.im>

16:27:32

or I wonder if we could customize a "welcome" message when people join a room like in Slack

<@hricky:fedora.im>

16:28:14

Can we test it more before announcing it?

<@dustymabe:matrix.org>

16:28:29

Hristo Marinov: if implemented as described in https://matrix.to/#/!YWqcsiUQiCaqimYdQT:fedoraproject.org/$TChkBqK2RLnFkacfsnLaHYTUBXrWadt7jPC6QvWs4Jk?via=fedoraproject.org&via=fedora.im&via=matrix.org help with your concerns?

<@dustymabe:matrix.org>

16:28:48

absolutely. we could invite you to a channel for testing

<@jlebon:fedora.im>

16:29:31

yeah, i think backtesting it would be good. if we judge its answers were more helpful than not, that seems sufficient?

<@hricky:fedora.im>

16:30:00

I think I'm already in the channel...

<@jlebon:fedora.im>

16:30:09

(including in the "not helpful" the baseline overhead of having to read an AI message at all)

<@dustymabe:matrix.org>

16:30:39

if it's not helpful no one would ever use it and we'd eventually drop it probably

<@dustymabe:matrix.org>

16:30:53

if it's not helpful no one would ever use it and we'd eventually drop it probably (or try to improve it to become helpful)

<@ydesouza:fedora.im>

16:31:03

I added some people in the channel already but I will invite anyone who is interested in it as soon I update the bot with a new key.

<@dustymabe:matrix.org>

16:31:12

the important trait here is that it's only invoked if someone asks it to be involved. So it's not just noise in the channel

<@jlebon:fedora.im>

16:31:39

dustymabe: right, i'm saying that should be a criteria for initial introduction. but yes, we should continually review it

<@dustymabe:matrix.org>

16:31:43

the important trait here is that it's only invoked if someone asks it to be involved. So it's automatically noise in the channel

<@dustymabe:matrix.org>

16:31:51

the important trait here is that it's only invoked if someone asks it to be involved. So it's not just automatically noise in the channel

<@aaradhak:matrix.org>

16:32:00

Just a quick time check, we are past our meeting schedule.

<@aaradhak:matrix.org>

16:32:48

Should we continue this discussion in the next meeting?

<@ydesouza:fedora.im>

16:32:50

Thanks for all your inputs about this topic, feel free to take a look in the repo and comment at this issue if you have any ideas or sugestions.

<@aaradhak:matrix.org>

16:33:36

<@aaradhak:matrix.org>

16:33:36

Alright, lets move to open floor.

<@aaradhak:matrix.org>

16:33:37

!topic Open Floor

<@aaradhak:matrix.org>

16:33:37

<@jlebon:fedora.im>

16:33:46

have to drop, but thanks all!

<@dustymabe:matrix.org>

16:34:16

related to the AI discussion

<@dustymabe:matrix.org>

16:34:55

I know of at least one other project we have that wants to use AI, but for analyzing pipeline failure logs instead (in a dedicated channel)

<@dustymabe:matrix.org>

16:35:22

I think we should create a common coreos bot matrix account that can be used for these various purposes. Does that sound reasonable?

<@ydesouza:fedora.im>

16:35:44

Yes, I agree with that. Makes sense!

<@aaradhak:matrix.org>

16:35:59

yes, +1 for creating a coreos bot matrix account for the pipeline failures

<@dustymabe:matrix.org>

16:36:24

I guess we already to have one matrix thing triggered by a bot over in https://matrix.to/#/#jenkins-coreos:fedoraproject.org but not sure what account that is using or if we could re-use it.

<@aaradhak:matrix.org>

16:37:00

May be we can create a new one ?

<@dustymabe:matrix.org>

16:37:06

Does anyone want to look into creating that "fedora coreos" matrix account ?

<@aaradhak:matrix.org>

16:37:27

I would like to look into it

<@dustymabe:matrix.org>

16:38:00

We should maybe consult fedora infra/releng to see if there is any established mechanism for bots already in Fedora or if it's just freeform.

<@ydesouza:fedora.im>

16:38:33

Nice, the account that I create to test the bot is linked in my e-mail so I guess there is a better way to do it.

<@aaradhak:matrix.org>

16:40:00

> We should maybe consult fedora infra/releng to see if there is any established mechanism for bots already in Fedora or if it's just freeform.

<@aaradhak:matrix.org>

16:40:00

<@aaradhak:matrix.org>

16:40:00

<@aaradhak:matrix.org>

16:40:00

I will raise this question in the fedora infra channel and get back .

<@ydesouza:fedora.im>

16:40:50

Thank you aaradhak and thanks for running the meeting :)

<@dustymabe:matrix.org>

16:41:38

!action aaradhak will try to determine the best path forward for creating a common coreos bot matrix account to use for various automations/helpers we want to create/use.

<@aaradhak:matrix.org>

16:42:01

Thanks for adding that dustymabe

<@hricky:fedora.im>

16:42:13

Yasmin Valim de Souza: Is the bot testing room the same one you invited me to earlier?

<@aaradhak:matrix.org>

16:42:25

If there is no other topic for open floor, we shall wrap up this meeting

<@aaradhak:matrix.org>

16:42:41

!endmeeting