fedora-coreos-meeting
LOGS
<@jbtrystram:matrix.org>
16:30:36
!startmeeting fedora_coreos_meeting
<@meetbot:fedora.im>
16:30:38
Meeting started at 2024-03-20 16:30:36 UTC
<@meetbot:fedora.im>
16:30:39
The Meeting name is 'fedora_coreos_meeting'
<@jbtrystram:matrix.org>
16:30:42
!topic roll call
<@ravanelli:matrix.org>
16:31:19
.hi
<@marmijo:fedora.im>
16:31:26
!hi
<@zodbot:fedora.im>
16:31:27
Michael Armijo (marmijo)
<@mnguyen:fedora.im>
16:31:27
.hi
<@dustymabe:matrix.org>
16:31:32
!hi
<@zodbot:fedora.im>
16:31:34
Dusty Mabe (dustymabe) - he / him / his
<@jlebon:fedora.im>
16:31:52
!hi
<@jbtrystram:matrix.org>
16:31:52
!hi jbtrystram@matrix.org
<@zodbot:fedora.im>
16:31:53
None (jlebon)
<@zodbot:fedora.im>
16:31:53
Sorry, I can only look up one username at a time
<@mnguyen:fedora.im>
16:31:55
!hi
<@zodbot:fedora.im>
16:31:57
Michael Nguyen (mnguyen)
<@jbtrystram:matrix.org>
16:32:16
!hi jbtrystram@matrix.org
<@zodbot:fedora.im>
16:32:18
Sorry, I can only look up one username at a time
<@aaradhak:matrix.org>
16:32:29
!hi aaradhak
<@zodbot:fedora.im>
16:32:31
Aashish Radhakrishnan (aaradhak)
<@spresti:fedora.im>
16:33:03
!hi
<@zodbot:fedora.im>
16:33:04
Steven Presti (spresti)
<@jbtrystram:matrix.org>
16:33:33
!hi jbtrystram
<@zodbot:fedora.im>
16:33:41
Jean-Baptiste Trystram (jbtrystram) - he / him / his
<@spresti:fedora.im>
16:34:00
jbtrystram: you dont need to Hi if you are hosting
<@spresti:fedora.im>
16:34:11
or atleast I have never done that
<@jbtrystram:matrix.org>
16:34:18
okay :)
<@jbtrystram:matrix.org>
16:35:14
Ok I assume everyone is here let's start
<@jbtrystram:matrix.org>
16:35:29
!topic Action items from last meeting
<@jbtrystram:matrix.org>
16:36:13
<@jbtrystram:matrix.org>
16:36:28
fifofonix to bring up a 1.28 cluster with zswap
<@jbtrystram:matrix.org>
16:36:55
it does not look like fifofonix is around today
<@jbtrystram:matrix.org>
16:37:24
let's re-action this one for next time
<@jbtrystram:matrix.org>
16:37:45
I feel it's been re-actionned 5 or 6 times already :)
<@jbtrystram:matrix.org>
16:37:56
!action fifofonix to bring up a 1.28 cluster with zswap
<@jbtrystram:matrix.org>
16:38:30
there were no other actions items in the last meeting
<@jbtrystram:matrix.org>
16:38:39
let's start with the topics
<@ydesouza:fedora.im>
16:38:43
!hi
<@zodbot:fedora.im>
16:38:44
Yasmin Valim de Souza (ydesouza)
<@jbtrystram:matrix.org>
16:39:06
!topic New Package Request: zip
<@jbtrystram:matrix.org>
16:39:18
<@jlebon:fedora.im>
16:40:13
actually, let's skip this
<@jlebon:fedora.im>
16:40:34
the reporter wanted to join the meeting but I failed to get back to them in time to give that info
<@jbtrystram:matrix.org>
16:40:35
we forgot to tell PhrozenByte when to join the community meeting
<@jbtrystram:matrix.org>
16:40:37
https://github.com/coreos/fedora-coreos-tracker/issues/1690#issuecomment-1995593202
<@jlebon:fedora.im>
16:40:44
let's see if they can join next week
<@jbtrystram:matrix.org>
16:41:28
let's make an action item
<@jlebon:fedora.im>
16:41:44
https://github.com/coreos/fedora-coreos-tracker/issues/1690#issuecomment-2010024306
<@jbtrystram:matrix.org>
16:41:51
ah, you already update the issue. Thanks Jonathan Lebon
<@jbtrystram:matrix.org>
16:42:03
let's skip.
<@jbtrystram:matrix.org>
16:42:36
!info we will ask @ PhrozenByte to join the community meeting to discuss the request next week
<@jbtrystram:matrix.org>
16:42:49
!topic Ship dnf in FCOS and RHCOS
<@jbtrystram:matrix.org>
16:42:58
<@jlebon:fedora.im>
16:43:29
i can take this one
<@jlebon:fedora.im>
16:43:51
so this is related to the cliwrap discussion we had recently-ish
<@jlebon:fedora.im>
16:44:28
the motivation is the same: to provide a consistent experience when writing Containerfiles to derive from FCOS
<@jbtrystram:matrix.org>
16:45:22
so we would ship dnf but it would only work when deriving images ?
<@jlebon:fedora.im>
16:45:26
the cliwrap approach was intended as a stopgap until we had better dnf support for the client-side, but there were UX concerns with it that made it less appealing
<@ravanelli:matrix.org>
16:45:33
:q
<@jbtrystram:matrix.org>
16:45:46
what would happen if i do `dnf` on a running fcos ?
<@jbtrystram:matrix.org>
16:46:01
what would happen if i do `dnf install` on a running fcos ? It layers the packages ?
<@jlebon:fedora.im>
16:46:12
jbtrystram: today, it would be useful in two situations: when layering on the container side, and client-side in an unlocked system (e.g. `ostree admin unlock`)
<@jbtrystram:matrix.org>
16:46:52
TIL about `unlock` ! does it break updates ?
<@jlebon:fedora.im>
16:46:59
today it'll error out with an error about the filesystem being read-only. but there's rough consensus to have it give a more useful error error soon that points to different options (one of them being `rpm-ostree install`)
<@dustymabe:matrix.org>
16:47:04
I think the "client-side in an unlocked system" isn't compelling - that need has always been there and we've never suggested to install dnf to solve that problem before
<@jlebon:fedora.im>
16:48:14
dustymabe: there have been *many* times when i've directly `rpm -i` packages in an unlocked system to debug things
<@jlebon:fedora.im>
16:48:38
it's doable today sure. having actual dnf makes it easier :)
<@dustymabe:matrix.org>
16:48:48
Jonathan Lebon: yes, but only now (many years later) have you suggested we should install `dnf` to solve that problem
<@jlebon:fedora.im>
16:49:08
dustymabe: it's not a primary motivation. i'm just laying out where the gaps are and what works today
<@dustymabe:matrix.org>
16:49:47
agreed. all i'm saying is that if that were the only reason then we wouldn't do this
<@jlebon:fedora.im>
16:49:56
agreed
<@jbtrystram:matrix.org>
16:50:25
reading the linked issue a question comes to mind : if I were to do `admin unlock` then `rpm -i` , the installed stuff would be gone after a reboot ?
<@dustymabe:matrix.org>
16:50:34
so really the reason we should focus on is "make the container installing experience easier"
<@jlebon:fedora.im>
16:50:44
jbtrystram: yup, exactly :)
<@mnguyen:fedora.im>
16:51:00
There's a hotfix option which will make it persist
<@dustymabe:matrix.org>
16:51:00
jbtrystram: yes, unless you added an option to make it persistent (I think `--hotfix` does that)
<@jbtrystram:matrix.org>
16:51:14
Jonathan Lebon: that's a really cool trick to know !!
<@dustymabe:matrix.org>
16:51:41
IMO you really only want to do this on systems you are debugging or developing on and you can easily throw them away after
<@jbtrystram:matrix.org>
16:52:24
what's the difference between `hotfix` and layering with `rpm-ostree install` ?
<@dustymabe:matrix.org>
16:53:03
that may be a good question for Jonathan Lebon or jmarrero after the meeting :)
<@jlebon:fedora.im>
16:53:41
jbtrystram: hotfix permanently modifies a specific deployment
<@jbtrystram:matrix.org>
16:53:48
yeah sorry to disrupt
<@jlebon:fedora.im>
16:53:51
`rpm-ostree install` carries it across updates
<@dustymabe:matrix.org>
16:53:55
Jonathan Lebon: are there any other reasons to install dnf? other than the two you listed?
<@jlebon:fedora.im>
16:54:27
dustymabe: alignment with the future of bootc, dnf, and rpm-ostree
<@dustymabe:matrix.org>
16:54:54
i'm still not sure clear on what that future is TBH
<@dustymabe:matrix.org>
16:55:01
i'm still not super clear on what that future is TBH
<@dustymabe:matrix.org>
16:55:44
kind of like a soup where things are moving around willy nilly
<@jlebon:fedora.im>
16:55:47
it's not fully fleshed out, but essentially: client-side logic in rpm-ostree will drain into dnf
<@jbtrystram:matrix.org>
16:55:48
dustymabe: making the system more fedora-like as well ? If we want atomic workstations to become the default (something that travier have been working on IIRC) then `dnf` should do something otherwise people will be confused
<@dustymabe:matrix.org>
16:56:28
jbtrystram: but including DNF here won't do that (other than the case where you are building a derived container)
<@jlebon:fedora.im>
16:57:05
also of note that dnf is shipping in the base bootable images. once we rebase on top of that, we'd have to consciously remove it (and possibly differ from other variants)
<@dustymabe:matrix.org>
16:58:06
I guess that would need to play into our decision to move over to that? which IIUC we haven't discussed anywhere
<@dustymabe:matrix.org>
16:58:20
this seems a little "cart before the horse"ey
<@jlebon:fedora.im>
16:58:37
for FCOS at least this should at least wait until dnf5 is the default. in RHCOS... well this is going to ship in other variants very soon...
<@dustymabe:matrix.org>
16:59:32
yeah, that's fine. hopefully they don't push off dnf5 to after F41
<@dustymabe:matrix.org>
16:59:41
they've already pushed it quite a bit already
<@jlebon:fedora.im>
16:59:43
so, i'm trying to keep them to not differ too much, but of course we could also do it in one and hold in the other
<@dustymabe:matrix.org>
17:01:35
ok what's next?
<@jlebon:fedora.im>
17:01:51
I know there's a lot of uncertainty surrounding this. I think we'll get more clarity in the mid-term at least, but that's where things happening around us are pointing currently.
<@jlebon:fedora.im>
17:03:22
how about we re-discuss this once the f41 dnf5 change has been approved? i think we need to discuss the RHCOS side too, but we can do that separately
<@jbtrystram:matrix.org>
17:03:50
> also of note that dnf is shipping in the base bootable images sre there base bootable image from fedora that we could rebase on ?
<@jbtrystram:matrix.org>
17:03:57
> also of note that dnf is shipping in the base bootable images are there base bootable image from fedora that we could rebase on ?
<@jbtrystram:matrix.org>
17:04:02
> also of note that dnf is shipping in the base bootable images are there base bootable image from fedora that we could rebase on ?
<@dustymabe:matrix.org>
17:04:03
I'm not the biggest fan of waiting
<@dustymabe:matrix.org>
17:04:20
could we for now install microdnf and symlink `/usr/bin/dnf` to it?
<@jlebon:fedora.im>
17:05:00
dustymabe: yeah, was thinking about that... possibly? i don't know enough about the gap between the two. but one thing i tested was that even the fedora-minimal image doesn't use microdnf
<@jlebon:fedora.im>
17:05:37
so i'm not sure where it's in active use
<@dustymabe:matrix.org>
17:05:48
k - so probably has issues, or at least if there are issues it won't get fixed
<@dustymabe:matrix.org>
17:06:23
```[core@cosa-devsh ~]$ sudo microdnf install htop Package Repository Size Installing: htop-3.3.0-3.fc40.x86_64 rawhide 200.2 kB hwloc-libs-2.10.0-3.fc40.x86_64 rawhide 2.2 MB Transaction Summary: Installing: 2 packages Reinstalling: 0 packages Upgrading: 0 packages Obsoleting: 0 packages Removing: 0 packages Downgrading: 0 packages Is this ok [y/N]: y Downloading packages... Running transaction test... error: Error -1 running transaction ```
<@jlebon:fedora.im>
17:06:25
it does ship dnf5 though. so possibly dnf5 renders it obsolete?
<@jlebon:fedora.im>
17:06:52
(e.g. i mean `fedora-minimal:39` ships dnf5 today)
<@dustymabe:matrix.org>
17:07:02
interesting.. didn't know that
<@dustymabe:matrix.org>
17:07:19
could we just pick up and use dnf5 then (since it's obviously built for f39/f40)?
<@jbtrystram:matrix.org>
17:07:20
I think it's used in redhat `ubi` images
<@jlebon:fedora.im>
17:08:52
dustymabe: we could... i do like though the signal we get from it being the default in Fedora
<@jlebon:fedora.im>
17:11:42
i would really like also to have `dnf install` output a more useful error message before we add it to FCOS
<@jlebon:fedora.im>
17:15:37
my initial goal with this discussion is more about having rough alignment of where we're going. there isn't anything actionable right now. if we agree to ship it in f41, great! but I'm also OK re-discussing that when there's more clarity.
<@jlebon:fedora.im>
17:16:30
well, i guess we could add it to the rawhide stream now
<@dustymabe:matrix.org>
17:16:40
I imagine we could actually ship this in rawhide already and make it conditional on F41?
<@dustymabe:matrix.org>
17:16:41
yeah
<@jlebon:fedora.im>
17:16:47
we could say "we'll add it in f41 and rediscuss at branching"
<@dustymabe:matrix.org>
17:17:01
I assume rawhide is dnf5 already?
<@jlebon:fedora.im>
17:17:14
i'm not sure. i'll check
<@jlebon:fedora.im>
17:17:44
not in `fedora:41` at least
<@dustymabe:matrix.org>
17:18:48
ahh, good to know
<@dustymabe:matrix.org>
17:19:02
ok, wrap up this topic?
<@jlebon:fedora.im>
17:19:20
ok here's my proposal and we'll see what comes out: proposed: we will add dnf5 to the rawhide stream to test it out and help shake out issues. we will re-evaluate at f41 branching.
<@jlebon:fedora.im>
17:19:48
ok here's my proposal and we'll see what comes out: proposed: we will add dnf5 to the rawhide stream to test it out and help shake out issues. we will re-evaluate at f41 branching whether to continue with it or hold off.
<@dustymabe:matrix.org>
17:20:26
that seems fine to me, though I'm not sure about "re-evaluate at f41 branching whether to continue with it or hold off" -> we need a mechanism to make us re-evaluate if we really want to re-evaluate remembering to re-evaluate, probably won't work well
<@jlebon:fedora.im>
17:20:57
we can add it to the f41 checklist (which we can file now)
<@jlebon:fedora.im>
17:22:36
ack, nack? the faster we vote, the faster we move on :)
<@aaradhak:matrix.org>
17:23:11
+1
<@jbtrystram:matrix.org>
17:23:20
+1
<@dustymabe:matrix.org>
17:23:25
+1
<@dustymabe:matrix.org>
17:23:56
might be worth including our reasoning for including DNF in the agreed
<@dustymabe:matrix.org>
17:24:14
or maybe the conversation in the ticket is sufficient
<@jbtrystram:matrix.org>
17:24:22
so 1 action to add dnf to rawhide, and 1 action to add to f41 checklist to rediscuss it
<@jlebon:fedora.im>
17:25:00
jbtrystram: you can assign those to me. i can also take care of posting the agreement in the ticket after the meeting, where i'll add more details about some of the things we discussed
<@jbtrystram:matrix.org>
17:26:27
!agreed: @jlebon will add dnf5 in the rawhide stream to test it out. We will add to the f41 branching checklist to reconsider this and evaluate at this point in time
<@jbtrystram:matrix.org>
17:27:30
the semicolon prevent the bot to pick it up. Before I confirm this, no objections ?
<@dustymabe:matrix.org>
17:28:14
yeah probably need to drop the colon
<@jbtrystram:matrix.org>
17:28:22
!agreed @jlebon will add dnf5 in the rawhide stream to test it out. We will add to the f41 branching checklist to reconsider this and evaluate at this point in time
<@jbtrystram:matrix.org>
17:29:08
we are running out of time. Should we carry on or push the remainings topics to the next time ?
<@jlebon:fedora.im>
17:30:25
i think let's just do an open floor
<@jbtrystram:matrix.org>
17:30:53
the reporter of the next issue is not here I think
<@jbtrystram:matrix.org>
17:31:12
!topic Open Floor
<@jbtrystram:matrix.org>
17:31:27
Any other topics to discuss ?
<@dustymabe:matrix.org>
17:31:44
we do have a test day coming up (assuming we get a Beta GO this week)
<@jlebon:fedora.im>
17:33:06
ash, jbtrystram how's that coming along?
<@aaradhak:matrix.org>
17:33:28
The Fedora test week is now rescheduled to the first week of April
<@aaradhak:matrix.org>
17:37:11
We are at the stage of shipping next as part of the tasks under Fedora 40 beta
<@aaradhak:matrix.org>
17:39:27
We are at the stage of shipping rebased next as part of the tasks under Fedora 40 beta
<@jbtrystram:matrix.org>
17:41:13
ok, if no one else as anything I'll end the meeting shortly
<@jbtrystram:matrix.org>
17:43:27
!endmeeting