<@conan_kudo:matrix.org>
15:04:58
!startmeeting CentOS Hyperscale SIG
<@meetbot:fedora.im>
15:05:00
Meeting started at 2025-05-07 15:04:58 UTC
<@meetbot:fedora.im>
15:05:00
The Meeting name is 'CentOS Hyperscale SIG'
<@conan_kudo:matrix.org>
15:05:15
!topic Roll call
<@conan_kudo:matrix.org>
15:05:29
!hi
<@zodbot:fedora.im>
15:05:31
Neal Gompa (ngompa) - he / him / his
<@rcolebaugh:matrix.org>
15:06:12
!hi
<@zodbot:fedora.im>
15:06:14
Raymond Colebaugh (rcolebaugh) - he / him / his
<@zodbot:fedora.im>
15:06:49
jflory7 has already given cookies to amoloney during the F42 timeframe
<@rrbc:matrix.org>
15:08:09
!hi
<@salimma:fedora.im>
15:08:10
!hi
<@zodbot:fedora.im>
15:08:11
Roberto Campesato (render) - he / him / his
<@zodbot:fedora.im>
15:08:11
Michel Lind (salimma) - he / him / his
<@davide:cavalca.name>
15:08:27
!hi
<@zodbot:fedora.im>
15:08:29
Davide Cavalca (dcavalca) - he / him / his
<@jonathanspw:fedora.im>
15:12:04
!hi
<@zodbot:fedora.im>
15:12:05
Sorry, could not get info from FASJSON (code 500)
<@conan_kudo:matrix.org>
15:12:10
O.o
<@conan_kudo:matrix.org>
15:12:17
that's a new one
<@conan_kudo:matrix.org>
15:12:36
try again?
<@salimma:fedora.im>
15:13:42
FAS is overloaded I guess
<@jonathanspw:fedora.im>
15:13:50
!hi
<@zodbot:fedora.im>
15:13:51
Sorry, could not get info from FASJSON (code 500)
<@jonathanspw:fedora.im>
15:13:55
:o
<@jonathanspw:fedora.im>
15:14:00
!hi jonathanspw
<@zodbot:fedora.im>
15:14:01
Sorry, could not get info from FASJSON (code 500)
<@conan_kudo:matrix.org>
15:14:03
wow you cause server errors
<@jonathanspw:fedora.im>
15:14:07
oof
<@jonathanspw:fedora.im>
15:14:10
I'm bad news for infra
<@conan_kudo:matrix.org>
15:14:48
!hi jonathanspw
<@zodbot:fedora.im>
15:14:50
Sorry, could not get info from FASJSON (code 500)
<@conan_kudo:matrix.org>
15:14:57
welp
<@jonathanspw:fedora.im>
15:15:03
Will try again in a bit
<@conan_kudo:matrix.org>
15:15:11
well, we know jonathanspw is Jonathan Wright anyway
<@conan_kudo:matrix.org>
15:15:19
let's move on to...
<@jonathanspw:fedora.im>
15:15:23
I tried to go with a pretty obvious fas name lol
<@conan_kudo:matrix.org>
15:15:27
!topic Followups
<@conan_kudo:matrix.org>
15:15:45
Anyone have stuff to follow up from last meeting?
<@salimma:fedora.im>
15:16:16
I have not had the chance to work on the script to identify inactive packagers
<@davide:cavalca.name>
15:16:21
I've moved some more stuff to GitLab
<@conan_kudo:matrix.org>
15:16:49
I've been working on moving our container builds to kiwi in koji
<@davide:cavalca.name>
15:16:53
And I'm talking to Ivan to see if we can generalize the systemd CI thing to use it for other packages
<@conan_kudo:matrix.org>
15:16:54
but that's been... a bit rough
<@salimma:fedora.im>
15:17:06
same - I rebased perf to 6.14 and also moved it to gitlab
<@conan_kudo:matrix.org>
15:17:14
it might be that we'll stopgap this with gitlab-ci using kiwi for the short term or something
<@conan_kudo:matrix.org>
15:17:21
at least we'll be halfway there
<@salimma:fedora.im>
15:17:27
(the hs+fb.el9 build still needs to be pinned, annoyingly, it now pins against llvm 19)
<@conan_kudo:matrix.org>
15:17:41
I don't love it but it's better than nothing
<@gwmngilfen:fedora.im>
15:18:53
happy to do so!
<@conan_kudo:matrix.org>
15:19:44
Davide Cavalca: do you think it's possible for us to port our container-releng thing to gitlab-ci?
<@davide:cavalca.name>
15:19:57
Daan is going to follow up with the testing farm folks and see if there's something we can do to improve reliability there
<@davide:cavalca.name>
15:20:07
It needs a full rewrite I think
<@davide:cavalca.name>
15:20:20
I haven't looked at the code in a while tbh
<@conan_kudo:matrix.org>
15:20:32
the main thing is we need a way to push to quay
<@conan_kudo:matrix.org>
15:20:42
everything else can be thrown away
<@conan_kudo:matrix.org>
15:21:11
we can invoke kiwi in gitlab-ci to build OCI images
<@conan_kudo:matrix.org>
15:21:25
so it turns into just two steps: kiwi build then skopeo push
<@salimma:fedora.im>
15:22:45
see, Kiwi is container native
<@conan_kudo:matrix.org>
15:22:56
😆
<@salimma:fedora.im>
15:23:16
there's a bunch of other things I want to port that looks like we want to make them run in gitlab ci too anyway
<@conan_kudo:matrix.org>
15:23:27
would you be willing to look into this for me?
<@salimma:fedora.im>
15:23:40
installability check for Hyperscale - this one is new - and new versino check
<@salimma:fedora.im>
15:23:47
look into which part?
<@conan_kudo:matrix.org>
15:23:58
container build + quay push in gitlab ci
<@salimma:fedora.im>
15:24:22
what's the scope - just for systemd, or for something else?
<@salimma:fedora.im>
15:24:38
(I probably can't do anything new until H2, juggling too many balls already)
<@rrbc:matrix.org>
15:24:50
I started testing builds of qemu on c10s and pulling some updates from fedora/rawhide, but had to switch to something else, so will go back to it probably tomorrow
<@conan_kudo:matrix.org>
15:25:00
it would just be a repo that has a script to check out kiwi descriptions in the right branch, then run the kiwi build, then use skopeo to push the archive into quay.io/centoshyperscale/centos
<@conan_kudo:matrix.org>
15:25:10
if you can't, I'll do it and get to it
<@conan_kudo:matrix.org>
15:25:20
but if you're looking at other things, it might harmonize with that
<@conan_kudo:matrix.org>
15:25:45
ehh, I guess I'll do it
<@davide:cavalca.name>
15:25:52
We have qemu and libvirt in gitlab now
<@conan_kudo:matrix.org>
15:25:55
I think I just talked myself into it
<@conan_kudo:matrix.org>
15:26:05
don't worry about it then :)
<@conan_kudo:matrix.org>
15:27:03
!action Neal to make new GitLab CI based container build and publish pipeline using kiwi and skopeo
<@salimma:fedora.im>
15:27:39
yeah, just want to give you a realistic timeline
<@salimma:fedora.im>
15:28:00
I think we should document any finding about gitlab CI containers anyway since I'm sure we'd reuse it for different purposes
<@conan_kudo:matrix.org>
15:28:05
yeah
<@conan_kudo:matrix.org>
15:28:28
I might even refresh some of our demo content stuff with GitLab CI pipelines if I get some spare time
<@salimma:fedora.im>
15:28:32
speaking of which do we have a "how to use GitLab" guideline?
<@conan_kudo:matrix.org>
15:28:58
I've got a document to push into our docs
<@conan_kudo:matrix.org>
15:29:07
so I'm going to try to finish that today
<@salimma:fedora.im>
15:29:07
I think just a rough draft is fine, doesn't have to be very polished. but we should keep an eye on what part can be enforced/set via automation and what cannot (and ideally then file issues to ask gitlab to support it)
<@salimma:fedora.im>
15:29:16
if anything we can point it to the forgejo people as a shopping list
<@salimma:fedora.im>
15:29:21
nice
<@conan_kudo:matrix.org>
15:29:21
yeah
<@conan_kudo:matrix.org>
15:29:36
it's mostly that I need to write down all the individual options for setting up a gitlab repo
<@conan_kudo:matrix.org>
15:29:43
there's a lot :(
<@jonathanspw:fedora.im>
15:31:22
!hi
<@zodbot:fedora.im>
15:31:23
Sorry, could not get info from FASJSON (code 500)
<@jonathanspw:fedora.im>
15:31:25
I give up
<@salimma:fedora.im>
15:33:32
!hello jonathanspw
<@zodbot:fedora.im>
15:33:34
Sorry, could not get info from FASJSON (code 500)
<@salimma:fedora.im>
15:33:39
it's really dead
<@conan_kudo:matrix.org>
15:35:39
anyway, any other followups?
<@salimma:fedora.im>
15:36:25
trying to get some initial website for Proposed Updates this week (though I am off to run errands on Friday so hopefully today or tomorrow)
<@conan_kudo:matrix.org>
15:36:48
oh jeez I forgot the week is halfway over
<@salimma:fedora.im>
15:36:58
I wonder if we should borrow one of the weeks Hyperscale is off to hold the meeting on the same slot, but maybe once a month
<@conan_kudo:matrix.org>
15:37:24
that would probably be reasonable
<@salimma:fedora.im>
15:37:34
on that front - I think some of the issues can be closed now, but I have not had the time to update our tracking spreadsheet yet. since there's a spreadsheet and we don't have to rely on the metadata from the issues, it's not a big deal
<@conan_kudo:matrix.org>
15:37:34
I expect that a monthly frequency is probably reasonable
<@salimma:fedora.im>
15:37:50
yeah, I expect it'd fit within 30 mins
<@conan_kudo:matrix.org>
15:40:31
anyway, I think that's it for followups
<@conan_kudo:matrix.org>
15:40:32
so...
<@conan_kudo:matrix.org>
15:40:46
let's move on to...
<@conan_kudo:matrix.org>
15:40:47
!topic Announcements
<@conan_kudo:matrix.org>
15:41:29
So, I have a pretty awesome announcement
<@conan_kudo:matrix.org>
15:41:54
!info The tooling for CentOS Hyperscale composes (Kurchu) has been open sourced and released!
<@conan_kudo:matrix.org>
15:42:02
!link https://blog.centos.org/2025/05/introducing-kurchu-for-centos-sig-content-collections/
<@salimma:fedora.im>
15:42:12
it's alive!
<@conan_kudo:matrix.org>
15:42:19
Yep, finally!
<@conan_kudo:matrix.org>
15:42:40
I'm going to re-remove the container builds from our definitions and then we can start setting up regular automatic composes.
<@rcolebaugh:matrix.org>
15:42:51
Nice!
<@conan_kudo:matrix.org>
15:44:17
but yeah we're now in a place where we can have periodic, regular snapshots of CentOS Hyperscale
<@conan_kudo:matrix.org>
15:44:24
with installable/bootable/usable images
<@conan_kudo:matrix.org>
15:44:27
including even WSL ones
<@conan_kudo:matrix.org>
15:45:26
the only thing we don't have is a netinstall ISO, and that's only because I haven't reverse engineered how they are made
<@conan_kudo:matrix.org>
15:45:33
to reimplement them in kiwi
<@salimma:fedora.im>
15:45:36
it's quite curious how this was not possible for so long
<@salimma:fedora.im>
15:45:56
I guess Fedora is mostly self contained (though with kurchu you can compose *with* COPR too huh, like for Asahi)
<@conan_kudo:matrix.org>
15:46:02
yeah
<@salimma:fedora.im>
15:46:06
and CentOS is ... a total mess with private build repos and such
<@conan_kudo:matrix.org>
15:46:21
CentOS and RHEL use custom tooling not used in Fedora too
<@conan_kudo:matrix.org>
15:46:43
their ODCS does it differently from Pungi proper
<@salimma:fedora.im>
15:47:07
but... but it's an open source first company
<@conan_kudo:matrix.org>
15:47:16
they're all open source
<@conan_kudo:matrix.org>
15:47:24
it's just that they aren't well documented or well understood
<@conan_kudo:matrix.org>
15:47:36
a big part of the design effort in kurchu was me digging into how those things worked
<@conan_kudo:matrix.org>
15:48:43
there are really only two features we don't have that pungi has: gathering signed packages from koji to generate flat repos, and using those gathered repos directly as inputs for image builds
<@conan_kudo:matrix.org>
15:48:56
I've managed to implement pretty much everything else that anyone would want
<@conan_kudo:matrix.org>
15:49:18
the first first is easy if you don't need to segment packages... that's when it gets messy
<@conan_kudo:matrix.org>
15:49:26
the first feature is easy if you don't need to segment packages... that's when it gets messy
<@conan_kudo:matrix.org>
15:49:54
a lot of pungi's complexity here is supporting the ability to take a big pile of packages and split up how they are published (or if they are published at all)
<@salimma:fedora.im>
15:50:01
the signed packages thing is surprisingly... opaque
<@salimma:fedora.im>
15:50:13
I was surprised when even Adam did not realize if you just download from Bodhi you don't get them
<@conan_kudo:matrix.org>
15:50:19
yeah though actually that part is easy now that I know how it works
<@salimma:fedora.im>
15:50:30
(pretty much I think anyone apart from you and nirik did not know when I was asking around)
<@salimma:fedora.im>
15:50:42
oh yeah the weird grobbisplitter thing
<@conan_kudo:matrix.org>
15:50:47
yup
<@conan_kudo:matrix.org>
15:51:08
some of that I consider "specialty" functionality
<@conan_kudo:matrix.org>
15:51:26
but some of it is generally applicable
<@conan_kudo:matrix.org>
15:51:44
for example, being able to auto-split packages that are `rust-*-devel` or `golang-*-devel` into a separate repo is a feature pungi has (though it is difficult to use)
<@conan_kudo:matrix.org>
15:52:33
my general philosophy is that I don't think it's necessarily a good idea to split up packages from a single pile into multiple piles
<@conan_kudo:matrix.org>
15:52:40
if you want them split, they should be split at the source
<@conan_kudo:matrix.org>
15:52:51
that way dependency relationships are clearer
<@conan_kudo:matrix.org>
15:53:59
anyway, the other bit is the ability to use gathered repos as inputs for image builds... this one is more challenging because for hyperscale we basically can't do it
<@conan_kudo:matrix.org>
15:54:42
since our setup involves only publishing the repo to s3, the gathered content on the system cannot be made available to Koji to consume before we finish everything
<@conan_kudo:matrix.org>
15:55:06
not to mention that Hyperscale 9 takes over 9 hours to gather all the input content anyway
<@salimma:fedora.im>
15:55:22
we definitely want to split rust-*-devel and golang-*-devel out - I wonder where that conversation ended up after mattdm is no longer FPL
<@salimma:fedora.im>
15:55:39
at this rate I think `golang-*-devel` is a dead end anyway so it's only an issue for Rust
<@conan_kudo:matrix.org>
15:56:24
I've been iffy about it for a while simply because it opens a pathway that would be hard to shut later
<@conan_kudo:matrix.org>
15:56:34
the split repos in RHEL are pretty aggravating
<@conan_kudo:matrix.org>
15:57:12
anyway, the announcement includes links to initial composes for both 9 and 10 and we should start having more created regularly soon enough
<@conan_kudo:matrix.org>
15:57:57
we will need to come up with a practice and support cycle for running composes though, I don't think we want to keep running 9 composes for that much longer, for example
<@salimma:fedora.im>
15:58:13
yeah... I think if we do this it has to be well thought out
<@salimma:fedora.im>
15:58:37
on lifecycle.. interesting question
<@salimma:fedora.im>
15:59:30
release date + four years? that gives one year to transition to the new version. or just X+1 + 1 year to account for the new version being slightly delayed
<@salimma:fedora.im>
15:59:44
kind of like Fedora stopping support for the N-2 release a month after N
<@conan_kudo:matrix.org>
15:59:54
release date + six months maybe?
<@conan_kudo:matrix.org>
16:00:09
anyway, we can discuss that later
<@conan_kudo:matrix.org>
16:00:15
need to roll through the rest
<@conan_kudo:matrix.org>
16:00:30
!topic Tickets
<@conan_kudo:matrix.org>
16:01:19
We have a couple of tickets related to kernel features, we should ask Red Hat to see if they're okay with enabling in RHEL first before we make a diff in our kernel config
<@conan_kudo:matrix.org>
16:01:40
because I like being able to just incorporate from their config in our kernel
<@conan_kudo:matrix.org>
16:02:07
!topic Membership
<@salimma:fedora.im>
16:02:37
I'm suggested the bluefin person who seems very enthusiastic about GNOME on CentOS to join
<@conan_kudo:matrix.org>
16:02:52
tulip: right?
<@salimma:fedora.im>
16:02:56
yes
<@conan_kudo:matrix.org>
16:03:08
tulip: could you introduce yourself?
<@salimma:fedora.im>
16:03:39
long term idk if we want to split off the GNOME SIG anyway - or does it need to remain GNOME focused? maybe Desktop SIG. are there things in KDE that can't be upgraded / packaged in EPEL?
<@conan_kudo:matrix.org>
16:03:49
yes
<@salimma:fedora.im>
16:04:05
(If tulip is not here right now, no rush, we literally told him about the meeting after it already started - though I did point him to Hyperscale yesterday)
<@conan_kudo:matrix.org>
16:04:21
let me see there's a ticket about this
<@salimma:fedora.im>
16:04:39
any example? (also KDE and GNOME shares certain components anyway and those component might end up being stale in EPEL, like malcontent)
<@conan_kudo:matrix.org>
16:05:00
https://pagure.io/fedora-kde/SIG/issue/487
<@conan_kudo:matrix.org>
16:05:04
!link https://pagure.io/fedora-kde/SIG/issue/487
<@salimma:fedora.im>
16:05:05
to be clear I don't foresee doing all the paperwork for yet another SIG until maybe H1 next year, one a year is enough :)
<@salimma:fedora.im>
16:05:22
but it's good to plan ahead
<@salimma:fedora.im>
16:05:36
anyway, I don't want to hijack the membership talk - anyone else already in the pipeline?
<@conan_kudo:matrix.org>
16:05:40
I was originally just planning on adding this in hyperscale-main
<@conan_kudo:matrix.org>
16:05:58
yeah there's one other person... I'd like to propose replacing themayor with Jonathan Wright
<@jonathanspw:fedora.im>
16:06:38
Replacing themayor where?
<@conan_kudo:matrix.org>
16:06:41
themayor joined us originally to work on Alma Hyperscale but hasn't been present in almost two years
<@conan_kudo:matrix.org>
16:06:53
Jonathan Wright: as a member of the Hyperscale SIG
<@jonathanspw:fedora.im>
16:07:00
I'm already a member :p
<@conan_kudo:matrix.org>
16:07:11
you're not in the membership roll I think?
<@conan_kudo:matrix.org>
16:07:17
I didn't see you pinged in #centos-hyperscale:fedoraproject.org
<@conan_kudo:matrix.org>
16:07:44
oh I'm dumb
<@conan_kudo:matrix.org>
16:07:54
it's there but fas is failing at your MXID thing
<@salimma:fedora.im>
16:08:01
who is themayor?
<@salimma:fedora.im>
16:08:18
FAS sounds vaguely familiar but I normally try to remember people's real names
<@conan_kudo:matrix.org>
16:08:22
Jack Aboutboul
<@salimma:fedora.im>
16:08:26
ahh
<@conan_kudo:matrix.org>
16:08:42
we can ask him if he wants to still participate, he might if he's interested from the Azure front
<@conan_kudo:matrix.org>
16:08:51
same reason davdunc is a member for AWS
<@jonathanspw:fedora.im>
16:08:56
I bet he'd love to engage on the azure front
<@salimma:fedora.im>
16:09:30
when we do our membership purge I think we can retain people who we know do useful work but the work is sporadic
<@conan_kudo:matrix.org>
16:09:36
yeah
<@salimma:fedora.im>
16:09:46
just purge people we know have either definitely moved on to other things, *or* we have no idea what they're doing :)
<@conan_kudo:matrix.org>
16:09:55
we need to do it soon, it's getting hard to figure out who is really here
<@salimma:fedora.im>
16:10:00
need a better word than purge too :P
<@salimma:fedora.im>
16:10:04
"graduate"
<@salimma:fedora.im>
16:10:05
yeah
<@conan_kudo:matrix.org>
16:10:21
emeritus :)
<@salimma:fedora.im>
16:10:37
if we make a library to check meeting activity in meetlog I think it will generally be useful for other SIGs too
<@salimma:fedora.im>
16:10:41
yeah but what's the verb for emeritus
<@salimma:fedora.im>
16:10:52
"promote upstairs"
<@salimma:fedora.im>
16:10:53
"retire"
<@conan_kudo:matrix.org>
16:11:08
retired
<@conan_kudo:matrix.org>
16:11:13
so we can do that
<@conan_kudo:matrix.org>
16:11:17
retire to the emeritus
<@conan_kudo:matrix.org>
16:11:21
anyway, all that's left is...
<@conan_kudo:matrix.org>
16:11:30
!topic Miscellaneous
<@conan_kudo:matrix.org>
16:11:41
okay, we should rename this to open floor, typing Miscellaneous is hard
<@salimma:fedora.im>
16:11:58
yeah
<@jonathanspw:fedora.im>
16:12:13
What about users/people that joined but never did anything...like they went through the motions to join with the implication that they'd so *something* or at least show up to meetings after they joined, but didn't?
<@salimma:fedora.im>
16:12:20
I like spelling long words - I grew up in a wrong country, I'd have loved spelling bees - but I am annoyed by typing that too
<@jonathanspw:fedora.im>
16:12:22
Seems unfair to give them recognition at all....
<@salimma:fedora.im>
16:12:31
good question
<@conan_kudo:matrix.org>
16:12:35
if there's zero, we'll just drop them
<@salimma:fedora.im>
16:12:51
I think if they join meetings but they have no package activity ever (either directly in HS or HS adjacent) we should drop them
<@conan_kudo:matrix.org>
16:12:53
emeritus is for recognized work but no longer participating
<@conan_kudo:matrix.org>
16:13:02
like all the Twitter folks
<@jonathanspw:fedora.im>
16:13:18
I have no package activity in HS directly, just adjacent.
<@conan_kudo:matrix.org>
16:13:33
and the people who worked on rpm-cow before Matteo
<@salimma:fedora.im>
16:13:35
right, that's why I intentionally include 'adjacent'
<@jonathanspw:fedora.im>
16:14:00
:)
<@rcolebaugh:matrix.org>
16:16:30
We have a new patch on the hs build for openssh: QoS bits were not being set properly by the server during handshake. The patch author mentioned looking into upstreaming it too
<@conan_kudo:matrix.org>
16:18:27
cool
<@salimma:fedora.im>
16:19:08
oh speaking of patched builds - we retired bash from hs+fb
<@salimma:fedora.im>
16:19:21
probably not worth publicly announcing since it was only hs+fb ever
<@conan_kudo:matrix.org>
16:19:25
awesome
<@salimma:fedora.im>
16:19:42
will probably retire fish too at least for hs+fb.el10
<@conan_kudo:matrix.org>
16:19:43
I'm going to try to get the 6.14 kernel rebase going for Hyperscale this week
<@jonathanspw:fedora.im>
16:20:13
!hi
<@zodbot:fedora.im>
16:20:14
Jonathan Wright (jonathanspw)
<@salimma:fedora.im>
16:20:28
welcome back zodbot
<@salimma:fedora.im>
16:20:42
wait are we overtime already
<@salimma:fedora.im>
16:20:48
isn't this meeting only 1 hour
<@conan_kudo:matrix.org>
16:21:40
yeah
<@conan_kudo:matrix.org>
16:21:56
I let it go over because of the length of roll call
<@conan_kudo:matrix.org>
16:22:11
but I think we're done here and we can continue conversations in #centos-hyperscale:fedoraproject.org
<@conan_kudo:matrix.org>
16:22:17
See you all next time!
<@conan_kudo:matrix.org>
16:22:24
!endmeeting