#meeting:fedoraproject.org log

<@davide:cavalca.name>

15:03:48

!startmeeting CentOS Hyperscale SIG

<@meetbot:fedora.im>

15:03:50

Meeting started at 2025-04-09 15:03:48 UTC

<@meetbot:fedora.im>

15:03:50

The Meeting name is 'CentOS Hyperscale SIG'

<@davide:cavalca.name>

15:03:55

morning folks

<@davide:cavalca.name>

15:03:59

!topic Roll call

<@salimma:fedora.im>

15:04:04

!hi

<@zodbot:fedora.im>

15:04:05

Michel Lind (salimma) - he / him / his

<@ngompa:fedora.im>

15:04:15

!hi

<@zodbot:fedora.im>

15:04:16

Neal Gompa (ngompa) - he / him / his

<@gmoro:matrix.org>

15:04:31

!hi

<@ngompa:fedora.im>

15:04:31

(my main account is being a bit sluggish today in the meeting rooms)

<@zodbot:fedora.im>

15:04:33

Guilherme Moro (guilhermemoro)

<@salimma:fedora.im>

15:05:18

Matrix federation is the worst

<@davide:cavalca.name>

15:06:12

let's get started

<@davide:cavalca.name>

15:06:13

!topic Followups

<@davide:cavalca.name>

15:06:20

any followups to share from the last meeting?

<@ngompa:fedora.im>

15:06:38

I'm nearly able to make Hyperscale 10 images now

<@ngompa:fedora.im>

15:06:48

There are only two remaining issues, one I can work around and one I can't

<@rcolebaugh:matrix.org>

15:07:03

!hi

<@zodbot:fedora.im>

15:07:06

Raymond Colebaugh (rcolebaugh) - he / him / his

<@ngompa:fedora.im>

15:07:51

The first issue is that zstd is not enabled in erofs and that caused the images to crash when booting. I worked around this temporarily by using lzma. But I'm not happy about the reduced performance and higher memory used for lzma, so I'm going to try to get this fixed.

<@ngompa:fedora.im>

15:08:27

The second issue is that the Mesa vulkan drivers are missing on non-x86_64, which caused the KDE image build to fail due to broken dependencies.

<@ngompa:fedora.im>

15:09:00

(There's also a third issue related to CBS Koji being too old preventing me from setting some flags for the image builds, but I worked around that by just not setting them for now.)

<@salimma:fedora.im>

15:09:51

we really should talk to the ISA SIG about issue nr 2

<@ngompa:fedora.im>

15:09:57

Yes.

<@ngompa:fedora.im>

15:09:58

I've filed a Jira and a merge request for the Mesa issue.

<@salimma:fedora.im>

15:10:17

if they're gung ho on onboarding architectures like riscv they probably care about making sure... the primary secondary architecture people normally have is not broken

<@ngompa:fedora.im>

15:10:29

!link https://issues.redhat.com/browse/RHEL-86174

<@salimma:fedora.im>

15:10:30

should we enable proposed updates in Hyperscale

<@ngompa:fedora.im>

15:10:40

!link https://gitlab.com/redhat/centos-stream/release-engineering/pungi-centos/-/merge_requests/970

<@salimma:fedora.im>

15:10:43

because we can build that *today* and unblock you

<@ngompa:fedora.im>

15:10:48

yes please

<@salimma:fedora.im>

15:10:58

want to file the infra ticket? I'll do the build

<@ngompa:fedora.im>

15:11:07

for sure

<@davide:cavalca.name>

15:11:09

oh yeah we should do that, good call

<@salimma:fedora.im>

15:11:15

(ask Fabian to plumb in proposed updates repos)

<@salimma:fedora.im>

15:11:29

!action Neal will ask Fabian to enable proposed updates repos for Hyperscale builds

<@davide:cavalca.name>

15:11:31

do we need to add a dependency between the release packages?

<@ngompa:fedora.im>

15:11:39

Yes.

<@salimma:fedora.im>

15:11:45

oh, yeah, that too

<@davide:cavalca.name>

15:11:46

we don't depend on epel explicitly though

<@davide:cavalca.name>

15:11:49

maybe we need that too

<@ngompa:fedora.im>

15:11:52

Yes we do.

<@ngompa:fedora.im>

15:11:59

the release packages will not install without epel-release

<@salimma:fedora.im>

15:12:00

maybe we should - yeah enable them both at once

<@davide:cavalca.name>

15:12:04

and update the mock configs as well

<@salimma:fedora.im>

15:12:08

Davide, want to do that?

<@davide:cavalca.name>

15:12:21

actually Michel Lind UTC-6 we should add proposed to the centos sig mock configs too

<@salimma:fedora.im>

15:12:30

I'll have my day full bringing up some packages for work today, so I can do the vulkan build but probably nothing else

<@ngompa:fedora.im>

15:12:41

!action Neal will file a ticket asking for CBS to be upgraded to support new kiwi plugin features

<@davide:cavalca.name>

15:12:42

any objections to moving the release package to gitlab while we're at it?

<@salimma:fedora.im>

15:12:46

Davide Cavalca: yeah, enable but off by default right?

<@ngompa:fedora.im>

15:12:51

I already moved the package to gitlab

<@salimma:fedora.im>

15:12:53

so people can turn it on if they want

<@salimma:fedora.im>

15:13:01

no objection, I move everything I touch to gitlab now

<@ngompa:fedora.im>

15:13:22

that's what I've been doing too

<@davide:cavalca.name>

15:13:34

on that note, Shaun and I met with gitlab yesterday, and we're good to go with the migration and to use it officially

<@salimma:fedora.im>

15:13:39

!action Davide will plumb proposed updates and EPEL to the hyperscale release package as dependencies and work on adjusting mok configs

<@ngompa:fedora.im>

15:13:50

https://gitlab.com/CentOS/Hyperscale/rpms/centos-release-hyperscale

<@salimma:fedora.im>

15:13:50

I am on my Dell laptop today, apologies for the bad typing

<@ngompa:fedora.im>

15:14:08

no worries, today I'm on the bad MBP so same issue :)

<@ngompa:fedora.im>

15:14:44

(the butterfly keyboard one)

<@salimma:fedora.im>

15:15:12

yeah, I would use the framework but now it needs a jump start each time it's off

<@salimma:fedora.im>

15:16:10

next topic I guess?

<@davide:cavalca.name>

15:16:32

!topic Announcements

<@salimma:fedora.im>

15:16:59

does a package update count as an announcement?

<@ngompa:fedora.im>

15:17:05

sure :)

<@salimma:fedora.im>

15:17:34

!info ethtool is updated to 6.14 in hs.el9 - I'll build it in hs.el10 too today now that it's in Fedora rawhide

<@salimma:fedora.im>

15:17:46

hmm I probably should not have used pronouns there oh well

<@salimma:fedora.im>

15:18:04

it's also moved to gitlab so the branches are c9s-hs and (soon) c10s-hs

<@salimma:fedora.im>

15:18:39

upstream was apparently awol for a while, so the version jumps from 6.11 to 6.14 (we never had 6.11 in HS, we had 6.10 plus a backported patch)

<@ngompa:fedora.im>

15:18:49

neato

<@ngompa:fedora.im>

15:19:11

is that it for you?

<@salimma:fedora.im>

15:20:31

yes

<@salimma:fedora.im>

15:20:37

sorry, babysitting

<@ngompa:fedora.im>

15:20:47

cool

<@ngompa:fedora.im>

15:21:32

!info Hyperscale kernel will rebase to 6.14 next week, pending the results of the Fedora Kernel Test Week.

<@ngompa:fedora.im>

15:21:44

!link https://fedoraproject.org/wiki/Test_Day:2025-04-06_Kernel_6.14_Test_Week

<@ngompa:fedora.im>

15:22:44

I held off from 6.13 for so long because the test week during that rebase bubbled up a lot of serious issues. Hopefully things will be smoother this time. Anyone who is able to participate in the test week in Fedora, please do so, it helps us too!

<@salimma:fedora.im>

15:23:01

yeah, 6.13 seems horrible, I wish we skipped it in Fedora

<@salimma:fedora.im>

15:23:13

there's a lot of noise at work from people who don't realize they can just keep booting old kernels

<@ngompa:fedora.im>

15:23:37

Empirically, it seems the longterm and first stable after longterm releases seem to be worse than the surrounding releases

<@ngompa:fedora.im>

15:23:51

after a couple of years of doing this, I've noticed that trend

<@salimma:fedora.im>

15:24:37

the longterm is worse? lol

<@ngompa:fedora.im>

15:24:58

yup

<@salimma:fedora.im>

15:25:02

I hope the point release for longterm is better then

<@ngompa:fedora.im>

15:25:05

I've observed _that_ issue for years

<@salimma:fedora.im>

15:25:07

otherwise... this is really bad engineering

<@ngompa:fedora.im>

15:25:19

going all the way back to almost 10 years ago

<@salimma:fedora.im>

15:25:27

grumble grumble year of desktop linux but desktop vendors never test their own hardware

<@ngompa:fedora.im>

15:25:37

people come out of the woodwork for longterm releases that don't show up otherwise

<@ngompa:fedora.im>

15:26:00

so a pile of new code that basically lived in someone's vendor tree and wasn't broadly tested gets integrated _right then_

<@salimma:fedora.im>

15:26:38

maybe anything new should be marked as experimental and only after they've been around for at least one kernel release can you get them into longterm

<@salimma:fedora.im>

15:26:39

anyway...

<@ngompa:fedora.im>

15:26:49

I don't disagree

<@ngompa:fedora.im>

15:26:57

but they got rid of that stuff 10+ years ago :(

<@salimma:fedora.im>

15:27:09

not sure you can do that with the current kernel model. anyway, moving on

<@ngompa:fedora.im>

15:27:14

indeed

<@ngompa:fedora.im>

15:27:26

just would like folks to participate in the test week if they can

<@ngompa:fedora.im>

15:27:30

it helps me out :)

<@salimma:fedora.im>

15:27:41

I'll post it at work

<@salimma:fedora.im>

15:27:54

at least then if things blow up afterwards I can point out they did not test so...

<@ngompa:fedora.im>

15:27:59

yup

<@ngompa:fedora.im>

15:28:28

I don't have anything else, so we can move on

<@davide:cavalca.name>

15:28:40

!topic Tickets

<@davide:cavalca.name>

15:28:52

I don't think we have anything here?

<@davide:cavalca.name>

15:30:20

!topic Membership

<@salimma:fedora.im>

15:30:34

so I think we dropped the ball again here

<@salimma:fedora.im>

15:30:54

if it's a Meta coworker let's send them a calendar invite next time

<@davide:cavalca.name>

15:31:09

yeah, fair

<@davide:cavalca.name>

15:31:15

!topic Miscellaneous

<@salimma:fedora.im>

15:31:21

but Neal wants to talk about purging membership

<@davide:cavalca.name>

15:31:29

oh right, sorry

<@salimma:fedora.im>

15:31:30

so ... probably the previous topic still applies? :)

<@davide:cavalca.name>

15:31:34

yeah we should do that

<@salimma:fedora.im>

15:31:45

!topic Membership

<@salimma:fedora.im>

15:31:59

should we adopt the same policy Fedora use to purge provenpackagers?

<@salimma:fedora.im>

15:32:11

is it 6 months or 12 months of complete inactivity?

<@salimma:fedora.im>

15:32:21

we can make that "either package builds or showing up in the meeting log"

<@ngompa:fedora.im>

15:32:36

I think 6 months is reasonable here

<@salimma:fedora.im>

15:32:39

and probably script it otherwise this becomes hellish

<@salimma:fedora.im>

15:33:17

do we make an exception if they used to be more involved? like our Twitter friends

<@ngompa:fedora.im>

15:33:19

we have weekly meetings, so you have to show up at least 1 in 24 meetings

<@salimma:fedora.im>

15:33:59

biweekly :)

<@ngompa:fedora.im>

15:34:02

I'm fine with giving more grace to specifically those who helped us create it and were heavily involved long enough

<@ngompa:fedora.im>

15:34:15

blech, so 1 in 12 then

<@ngompa:fedora.im>

15:34:23

it's still plenty :)

<@salimma:fedora.im>

15:34:37

so - 6 months inactivity, put up for review, then we get to vote on who to waive at most once

<@ngompa:fedora.im>

15:34:44

and there's also the social hangout, which adds 6 more chances

<@salimma:fedora.im>

15:34:57

ugh those are hard to track though

<@ngompa:fedora.im>

15:35:03

fair

<@ngompa:fedora.im>

15:35:14

but if someone has made an impression with us, we'll remember :)

<@salimma:fedora.im>

15:35:16

I guess we can do that during the review, if anyone remembers seeing them at a hangout we can say they are active and move on

<@salimma:fedora.im>

15:35:20

if nobody remembers then too ba

<@salimma:fedora.im>

15:35:23

too bad

<@ngompa:fedora.im>

15:35:26

yeah

<@salimma:fedora.im>

15:35:30

I hate this keyboard so much, it's worse than Apple ones

<@ngompa:fedora.im>

15:35:44

even the crappy butterfly ones?

<@salimma:fedora.im>

15:35:49

yes

<@ngompa:fedora.im>

15:35:56

that one can't keep up with my typing speed and my fingers _hurt_

<@salimma:fedora.im>

15:36:06

Dell UX at its worse is basically aping Apple just like GNOME

<@ngompa:fedora.im>

15:36:12

that's fair :)

<@salimma:fedora.im>

15:36:16

yeah, I get both too. it can't keep up and my fingers hurt

<@salimma:fedora.im>

15:36:36

and I got burned by framework too so probably no framework for a year until the sting passes

<@salimma:fedora.im>

15:36:54

anyway... ok, I guess someone should write up the membership process

<@salimma:fedora.im>

15:36:58

I can do that if we're not in a rush

<@salimma:fedora.im>

15:37:50

we should also probably revisit the meeting time, I suspect most of our potential members is on the west coast so 8am is a bit brutal

<@ngompa:fedora.im>

15:37:55

no rush but we really do need to clean this up

<@ngompa:fedora.im>

15:38:12

we probably can't move it because we're surrounded by other meetings that we attend

<@salimma:fedora.im>

15:38:14

yeah, I think I can have a PR with this change text this week

<@salimma:fedora.im>

15:38:23

yeah, meeting bankrupty

<@salimma:fedora.im>

15:38:29

bankruptcy ffs

<@salimma:fedora.im>

15:39:01

also if anyone knows who can remove stale meetings from the Fedora calendar that'd be great

<@salimma:fedora.im>

15:39:09

the Rust SIG meeting has not happened for years :P

<@salimma:fedora.im>

15:39:36

!action Michel to send a PR with proposed membership cleanup process

<@salimma:fedora.im>

15:39:42

!topic Miscellaneous

<@ngompa:fedora.im>

15:40:07

wow we went through this quick

<@ngompa:fedora.im>

15:42:38

I want to highlight that because of our work and the work we've done with Troy Dawson on CentOS Alternative Images, it is very likely that official CentOS images will move to kiwi.

<@salimma:fedora.im>

15:43:18

oh nice

<@zodbot:fedora.im>

15:43:29

salimma has already given cookies to ngompa during the F41 timeframe

<@salimma:fedora.im>

15:43:45

now we just need to declare it Container Native

<@zodbot:fedora.im>

15:44:11

rcolebaugh has already given cookies to ngompa during the F41 timeframe

<@ngompa:fedora.im>

15:44:14

Yes! 😂

<@ngompa:fedora.im>

15:45:23

I'm helping this effort by writing the initial descriptions (based on my work in Hyperscale, Alt Images, and ELN): https://gitlab.com/Conan_Kudo/centos-stream-kiwi-descriptions

<@ngompa:fedora.im>

15:45:41

at the very least I will be syncing back in the container definitions into the hyperscale descriptions

<@ngompa:fedora.im>

15:45:55

so we can have Hyperscale 10 container images

<@salimma:fedora.im>

15:46:44

oh that will be great for quick testing

<@ngompa:fedora.im>

15:46:50

and yes, this means we can also have toolbox images if we want

<@salimma:fedora.im>

15:47:20

isn't it spelled toolbx now? or did they revert bak

<@salimma:fedora.im>

15:47:25

I really hate this keyboard

<@ngompa:fedora.im>

15:47:35

project is toolbx, thing is still toolbox, and the images are still called toolbox

<@salimma:fedora.im>

15:47:48

so on one hand I agree toolbox is a better design but distrobox is so much more practical

<@ngompa:fedora.im>

15:48:01

both use the same formatted images

<@salimma:fedora.im>

15:48:23

not quite true, distrobox let you use *any* image. it's a superset

<@ngompa:fedora.im>

15:48:31

when we make hyperscale atomic with btrfs transactional updates, we could ship distrobox if you'd like

<@salimma:fedora.im>

15:48:34

unless toolbox has improved on that front?

<@salimma:fedora.im>

15:48:49

you just point distrobox to the registry url, that's all

<@ngompa:fedora.im>

15:49:01

I don't pay too much attention to the weirdness of the toolbox ecosystem personally

<@salimma:fedora.im>

15:49:03

so it can use toolbox images but also just any Docker image

<@salimma:fedora.im>

15:49:44

and yes distrobox is a pile of bash scripts so... scary

<@ngompa:fedora.im>

15:50:41

but with Jonathan Wright making progress on dnf5 for EPEL 10, we should prioritize porting things to DNF 5 sooner rather than later

<@salimma:fedora.im>

15:51:21

yeah testing dnf5 in containers would definitely help, lower the barrier of entry, before people try it on bare metal

<@jonathanspw:fedora.im>

15:51:28

Oh hi

<@salimma:fedora.im>

15:51:39

we have summoned Jonathan

<@salimma:fedora.im>

15:51:46

"a wild Jonathan appeared"

<@jonathanspw:fedora.im>

15:54:00

I hope to have dnf5 into epel-testing this week or next

<@jonathanspw:fedora.im>

15:54:31

only blocker that I know of now is lining up the logistics of some merges and builds into fedora and subsequently branching into epel10

<@zodbot:fedora.im>

15:55:01

salimma gave a cookie to jonathanspw. They now have 19 cookies, 1 of which were obtained in the Fedora 41 release cycle

<@zodbot:fedora.im>

15:56:42

rcolebaugh gave a cookie to jonathanspw. They now have 20 cookies, 2 of which were obtained in the Fedora 41 release cycle

<@davide:cavalca.name>

15:56:57

I think that's it for today

<@davide:cavalca.name>

15:57:16

thanks everyone, and see you soon

<@rcolebaugh:matrix.org>

15:57:24

Thank you!

<@zodbot:fedora.im>

15:57:25

ngompa gave a cookie to jonathanspw. They now have 21 cookies, 3 of which were obtained in the Fedora 41 release cycle

<@davide:cavalca.name>

15:57:26

!endmeeting