Meeting started at 2025-04-09 16:30:04 UTC

The Meeting name is 'fedora_coreos_meeting'

!hi

Yasmin Valim de Souza (ydesouza)

!hi

Hristo Marinov (hricky) - he / him / his

!hi

!hi @

Timothée Ravier (siosm) - he / him / his

Sorry, but Fedora Accounts user '@' does not exist

!hi :)

!hi

Sorry, I can only look up one username at a time

Michael Armijo (marmijo)

hello

!hi

Tiago Bueno (tlbueno)

!hi

Jean-Baptiste Trystram (jbtrystram) - he / him / his

!hi

Dusty Mabe (dustymabe) - he / him / his

!hi

No Fedora Accounts users have the @nemric:relativit.fr Matrix Account defined

!hi mnguyen

Michael Nguyen (mnguyen)

Make sure to add the meeting tags to topics that you want to discuss. We only have one right now

!hi aaradhak

Sorry, I can only look up one username at a time

!hi ravanelli

Renata Ravanelli (ravanelli)

I don't see any actions in the last meeting log

We are in final freeze. Does anyone has anything in mind that we need to do?

We probably need a representative to join the Go/No Go meetings on Thursdays

Ah yes, indeed

From https://github.com/coreos/fedora-coreos-tracker/issues/1851, it looks like we did not yet set a barrier release for the last N-1 next release

Hypothetically F42 could GA next Tuesday

hmm travier I did that, but maybe I forgot to update something

https://github.com/coreos/fedora-coreos-streams/pull/1039 > looks like we did, ticking the bix

https://github.com/coreos/fedora-coreos-streams/pull/1039 > looks like we did, ticking the box

https://github.com/coreos/fedora-coreos-streams/pull/1039

I can attend the go/no-go meetings

One thing we haven't done this cycle (all the way at the top of the rebase tracker ticket) is do an evaluation of the package additions/removals

I will be doing the next FCOS release. Since its tha F42 GA, there is any concerns or things that I should know about it?

while that's not strictly necessary, it is a nice sanity check

Yasmin Valim de Souza: it *may* be F42 GA, we don't know yet

if it is, we'll partner with you heavily

Nice! Thank you!

Is anyone up to doing an update from F41 to F42 to get a list of packages to investigate?

I've created https://github.com/coreos/fedora-coreos-tracker/issues/1921

I can get the additions/removals list pretty easily. The harder part is going through them to evaluate

marmijo: may have the list handy too :)

I can help with the evaluation :)

I'm sure I can get that fairly easily. I can help with this!

dustymabe has already given cookies to marmijo during the F41 timeframe

dustymabe has already given cookies to siosm during the F41 timeframe

jbtrystram gave a cookie to dustymabe. They now have 190 cookies, 2 of which were obtained in the Fedora 41 release cycle

Foo. Zodbot, cookies should be an infinite resource

ravanelli gave a cookie to siosm. They now have 33 cookies, 4 of which were obtained in the Fedora 41 release cycle

ravanelli gave a cookie to dustymabe. They now have 191 cookies, 3 of which were obtained in the Fedora 41 release cycle

ravanelli gave a cookie to marmijo. They now have 10 cookies, 4 of which were obtained in the Fedora 41 release cycle

why not do that instead of the community meeting next week ?

:)

jbtrystram gave a cookie to marmijo. They now have 11 cookies, 5 of which were obtained in the Fedora 41 release cycle

jbtrystram gave a cookie to siosm. They now have 34 cookies, 5 of which were obtained in the Fedora 41 release cycle

marmijo gave a cookie to dustymabe. They now have 192 cookies, 4 of which were obtained in the Fedora 41 release cycle

marmijo gave a cookie to siosm. They now have 35 cookies, 6 of which were obtained in the Fedora 41 release cycle

we need to do that before next week in case F42 is GA next week

wow, Dusty got 2 cookies this cycles, for 190 cookies total. That means you have been there since fedora -52 !

Ideally this is a "now" thing as the no/no-go is tomorrow

(on average)

ydesouza has already given cookies to marmijo during the F41 timeframe

ydesouza gave a cookie to siosm. They now have 36 cookies, 7 of which were obtained in the Fedora 41 release cycle

ok, let's move to the next topic

👋 nirik

nirik is here if we have questions :)

hello

We can probably move "as soon as" the new cluster is up?

I think basically the choices are a) move in late may, or b) move in mid june.

yeah...

but we need access to the builders

How do you access builders currently?

let's dig into this just a little more

in the new DC I think we might have enough hardware to get a dedicated ppc64le machine?

if that's the case we could have it set up and use when we do the migration

if that's the case we could have it set up and use when we do the migration (could be May)

no. In the new datacenter we are hoping to have power10's... but there will be only 2 of them for a bunch of uses. We can (and I was planning on) giving you a lpar on one...

but you can't have an enitre machine. ;)

hmm i'm still pretty ignorant about the architecture I guess. I thought lpar was a s390x thing

some of this is handwavy, because I don't have access to the new machines yet. ;) But this is my planning at least

I think its a ibm thing. ;)

anyway - the way we connect to different things is via SSH so hypothetically our ppc64le builder could remain in place where it is and our pipeline could move to the new DC

as long as SSH still worked

we could possibly make that work. A new power10 lpar would be nicer of course (I would hope it would be a faster for one thing)

are your builds cpu or i/o bound? or both like other builds?

1. coreos-ostree-importer

2. fedora-ostree-pruner

the only other question I have is related to our few "agents" that run and do access the OSTree repo

nirik: on ppc - i'm not actually sure

the tests are probably more I/O bound than CPU bound (we test installs which involves a lot of disk I/O)

yeah, so storage. If we moved you, we would move storage to the new place too... I should perhaps look at that and see if it has any entanglements with non moved stuff.

Does this p10 support kvm?

it's a unified OSTree repo

"move storage"

but we aren't the only users of that storage ?

it's complicated. ;) You mean for nested virt?

Yeah! I don't know how the new p10 are supporting it, if it is a matter of firmware our p10 model

yeah, thinking about it, storage might mess this early plan up. ;( Since we still need it in the old dc for atomic desktops, etc

nirik: the early plan can still happen, we just need to leave those two services where they currently are until storage moves

everything else can move

nested virt should work. The layers are a but confusing with power stuff... there are layers

dustymabe: but you need to write your composes? or ?

or the importer does that work?

nirik: we shove everything in s3 first and then send a fedmsg that the coreos-ostree-importer picks up and processes.. it pulls the commit from s3 and imports it into the ostree repo

ah ha. great. then yes...

so our pipeline can actually live in a different place than the importer

ok, cool. So from my perspective, I would like to move anything thats independent as soon as I can/before the switch... that will mean less stuff in the move week to worry about.

https://fedoraproject.org/wiki/Changes/CoreOSStopPublishingOSTree

i'll note that in F43+ we'll no longer be pushing to the OSTree repo

nirik: at what point do you think the new openshift cluster will be up and ready for workloads?

Yeah, nested virt is a new thing for P10s, before it did not support it at all, but now it does. How? I don't know, my guess is only a matter of firmware

I am hoping early mayish... but it's subject to my getting access to the hardware, getting enough infra setup to install, etc...

nested virt should be working for you with power9 as well... you just have to tweak the xml to allow it...

+1 - keep us in the loop

we're happy to start moving stuff over

definitely can do.

will circle back when I know more.

should I update that issue/keep it?

I just posed an update to the issue. Ping us back there when you have updates

if you get access to the p10 we can probably go ahead and start playing with it too

I just posted an update to the issue. Ping us back there when you have updates

ok. Note there is a ppc64le-test02.fedorainfracloud.org that maintainers have access to if you want to look at anything power10 beforehand.

ehh. I'm more interested in getting FCOS installed on top of it

yeah, fair

which is the OS we use for our builders

Happy to answer any other questions as they occur to you all.

dustymabe: I'm also curious to see a benchmark with a P9 build

like comparing performance between p9 and p10 ?

Yeah

at the very least our power9's have slow spinning disk and the power10's should be all ssd.

Seems this nested virt is some kind of emulation on top of the power hypervisor

Just curious to see, we can also ask Mick helps, he was working I guess to test it as well vs P9

yeah, there's a aix like hypervisor -> lpars -> etc...

but it's all a bit murky without the actual hardware to look at.

OK, looks like there is plenty to do but no immediate blockers

anyhow thats all I had. :) thanks for hearing me out

Coordinating in the issue tracker is best

Thanks for keeping the infra working for all of us!

Coordinating the next steps in the issue tracker is best

siosm gave a cookie to kevin. They now have 724 cookies, 32 of which were obtained in the Fedora 41 release cycle

ravanelli gave a cookie to kevin. They now have 725 cookies, 33 of which were obtained in the Fedora 41 release cycle

jbtrystram gave a cookie to kevin. They now have 726 cookies, 34 of which were obtained in the Fedora 41 release cycle

tlbueno gave a cookie to kevin. They now have 727 cookies, 35 of which were obtained in the Fedora 41 release cycle

I had a good discussion with a person from Akamai Cloud (formerly Linode) so we may get some credits and could start making image for this cloud as well. We'll see how this goes.

Alright, if there is nothing else, I'll close this one