Meeting started at 2025-10-22 15:07:17 UTC

The Meeting name is 'CentOS Hyperscale SIG'

!group members centos-hyperscale-sig

Sorry, but group 'centos-hyperscale-sig' does not exist

!group members centos-sig-hyperscale

Sorry, but group 'centos-sig-hyperscale' does not exist

huh we don't have a Pagure group? or I'm forgetting what it's called

anyway

!hi

Neal Gompa (ngompa) - he / him / his

!group members sig-hyperscale

Members of sig-hyperscale: abulimov, Ali Erdinc Koroglu, alphacc, Anita Zhang, bkircher, chantra, Chris Murphy, Daan De Meyer, David Duncan, dbrandonjohnson, Davide Cavalca, djohansen, dminer, filbranden, Gabriele Mambrini, hyperscalebot, Igor Raits, ikruglov, Jack Aboutboul, Jonathan Wright, junwang, jvreeland, kcwells, leoleovich, malmond, Quentin Deslandes, neil (@neil:fedora.im, @nhanlon:beeper.com), ngompa (@conan_kudo:matrix.org, @ngompa:fedora.im, @pharaoh_atem:opensuse.org, @ngompa:kde.org, @ngompa:almalinux.im), Anton Tiurin, oidoming, osandov, Raymond Colebaugh, Roberto Campesato, richardphibel, Ryan Wilson, salimma (@michel-slm:matrix.org, @salimma:fedora.im), steelcowboy, tdmackey, Matteo Croce, tunger10, vfedorenko, vishalmishra434, Yaping Li

!hi

Gabriele Mambrini (gmambro) - he / him / his

!hi

Roberto Campesato (render) - he / him / his

!hi

ah I was just logging into accounts to find out :)

Raymond Colebaugh (rcolebaugh) - he / him / his

we really need to prune it

!hi

Michel Lind (salimma) - he / him / his

that's why I don't really use the group sig-hyperscale thing right now

it's too big

!hi

Davide Cavalca (dcavalca) - he / him / his

there is a gitlab equivalent, I wonder if it has less people

we can try it later in a place that is not logged so it's not spammy

gitlab will have the same people as long as they logged into gitlab at least once

gitlab-centos-sig-hyperscale

ah so it might be quieter as people who left before gitlab is a thing won't show up

anyway, enough time for roll call I guess

rcolebaughhad two gitlab MRs for openssh we have been remiss in reviewing. I just landed and built the c10s-hs+fb one

for those not at Meta, this does not have anything interesting outside of our use cases so except for one time when it's an emergency CVE fix we only built it in the facebook variant - it's just some pesky extra logging we have been trying to upstream

the c9s one is interesting. we have had an out of sync patch for FIPS mode at least since I cut that emergency CVE release last year - I noticed when reviewing rcolebaugh's update

not sure how that happens, but we should just check out the c9s version of that patch from the revision we're merging and get back in sync. even though we don't really care about FIPS

Thanks Michel! I'll take a look into that and see what happened there. I seen to remember some patches being superseded upstream but will need to dig into the history to verify this

that file was slightly out of sync when I did a merge, which I suspect makes your git merge ignore it somehow (not sure what different git merge strategies do for files that are considered modified)

or we both manually resolved that file wrong? anyway...

I have a related thing for open floor later, remind me to bring it up if I forget

related two things i guess

anyone else has a follow-up?

None from me

nothing from me, I've been sick for the past couple of weeks

oops, might want to see a doctor Conan Kudo 😷🤒🤧

just in case you end up with a secondary bacterial infection and need antibiotics

I did, I got medications for it

sinus infection triggered bronchitis

oops

hope you can rest more after F43 is out of the door next week :)

anyone has anything? I guess the AlmaLinux Btrfs announcement is tangentially related

Connect cfp is open

https://connect.centos.org/

Assuming I can put myself back together this week, I'd like to get started on rebasing to 6.17 for hs10

ah speaking of things that need rebasing, perf's standalone tarball for 6.16 is also fubar on aarch64 - acme did not fix the missing files issue

do you plan to poke him about it for 6.17?

so I'm just going to give up on it. I think there's an additional problem with some BPF headers being too old

I followed up on the mailing list earlier this week

but this should not block us, I'll just do the fallback of using the full kernel tree source but a separate SRPM so we can keep things focused

that way we can keep the HS kernel going for longer, if like right now perf won't build out of the box because of outdated dependencies :)

hoping if this works well enough we can persuade jforbes to start yanking userspace tools out of the kernel tree

kernel srpm, sorry. but anyway, that is it from me on this topic

there are issues with internal kernel bpf code with cs9, which have been blocking kernel rebases on cs9 for a while too

it might be the same issue, or similar

I guess we should just focus on c10 first then

should we move on to tickets?

I don't see any new ticket

did we finish processing James' membership after the last meeting?

don't think so

I'm starting to experiment with Niri. Better hurry. Just kidding. Get well soon Neal! 🙏

Davide is back from PTO so it might be easier if he does this

yeah sure

so that takes care of memberships too, since I don't see any new ticket related to that :)

james: what's your FAS?

so ... documentation. since James is joining and all

Davide Cavalcaissue 197

jreilly1821 should be it?

yeah looks it it matches

cool thanks

our internal docs still mentions git.centos.org and c8s, I think we should replace them with gitlab and c10s/c9s

yeah

and Conan Kudo 😷🤒🤧has been itching to replace the meeting template to not say Miscellaneous anymore

yes

we also need to finish moving to gitlab

so we should just do it at some point :)

yeah the issue tracker is still on pagure

remind me tomorrow if this does not happen :)

the meeting template... we probably want to also add group members ping to it like FESCo does, but that means we need to clean up our membership roster first :). I still owe an activity script for it

the other thing I want to bring up is package testing... I have not heard from arrfab but ideally we want to have signed testing packages, right?

yes

just like Bodhi - so apart from unit tests in %check and integration tests in tmt, you want users to test things in -testing, right now we often just yolo tag as release because we need signed packages to install on production hosts

anyone knows if it's actually possible already? iirc test packages are not signed

I don't mind if they are not mirrored, since they won't really get downloaded much

it is currently not

I think we are the first to ask for it

so let's file an infra ticket and see what arrfab says?

sounds good

alrighty, anyone has another topic?

otherwise we can give Neal time to rest before our next common meeting in... oh 12 mins

I got nothin'

related, Europe goes off daylight saving this weekend :)

so for a brief 2 weeks US-centric meeting times will be less painful :P

and likewise EU-centric meeting times I guess, it will be less early for Americans

and on that bombshell... thanks all for coming! you all get 10 mins back