16:00:12 <dcavalca> #startmeeting Hyperscale SIG
16:00:12 <centbot> Meeting started Wed Sep  1 16:00:12 2021 UTC.  The chair is dcavalca. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:12 <centbot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:00:22 <dcavalca> #chair dcavalca jvreeland
16:00:22 <centbot> Current chairs: dcavalca jvreeland
16:00:24 <dcavalca> morning folks
16:00:44 <jvreeland> Morning
16:00:49 <anitazha> morning
16:01:02 <Eighth_Doctor> hey all
16:01:14 <themayor> hello hello
16:01:32 <Eighth_Doctor> ooh
16:01:36 <Eighth_Doctor> hey themayor :)
16:02:01 <dcavalca> let's get started
16:02:03 <dcavalca> #topic Followups
16:02:13 <dcavalca> anybody has followups from the last meeting?
16:03:22 <dcavalca> looks like not
16:03:29 <dcavalca> #topic Announcements
16:03:49 <dcavalca> c9s has updated to the 5.14 kernel
16:03:58 <dcavalca> so that's something we'll wanna keep an eye on
16:04:38 <Eighth_Doctor> they've forked the source tree from ARK
16:04:45 <Eighth_Doctor> so now we need to start thinking about things to stack on top of 5.14 from mainline
16:06:03 <jvreeland> #info we're enabling builds for aarch64
16:06:08 <Eighth_Doctor> we should branch from the kernel-5.14.0-1.el9 tag into our repo and start stacking changes to submit
16:06:40 <dcavalca> Eighth_Doctor: should we keep doing that on pagure, or should we make a repo on gitlab to ease collaboration there?
16:06:55 <dcavalca> I don't think we can do PRs from pagure to gitlab
16:07:09 <dcavalca> but I'm also not sure what the actual contribution flow is / will be, so it might not matter much in practice
16:07:58 <jvreeland> I think for 9 we need to work in gitlab.
16:08:24 <Eighth_Doctor> dcavalca: we will probably want to have a fork on GitLab.com for changes we send there.
16:08:47 <dcavalca> we already have https://gitlab.com/centos-sig-hyperscale so we can probably keep a fork up there
16:09:02 <Eighth_Doctor> but for the tree we actually ship (being the collection of patches that I can trivially `git format-patch` to export), pagure is easier
16:09:19 <Eighth_Doctor> the way that pagure represents commit histories makes it considerably easier to track deltas
16:09:29 <dcavalca> works for me
16:10:09 <Eighth_Doctor> for example: https://pagure.io/centos-sig-hyperscale/anaconda/commits/c8s-hs-v33.16.5.3
16:10:35 <Eighth_Doctor> you can see how branches and tags are stacked to create a unified history
16:10:42 <jvreeland> sounds good
16:11:11 <Eighth_Doctor> do we have any simple things we could start with to test a workflow to contribute?
16:11:58 <jvreeland> not that i'm aware of
16:12:31 <Eighth_Doctor> we should come up with something so that we can concretely work out the workflow
16:12:42 <Eighth_Doctor> maybe stuff with btrfs or something?
16:13:16 <dcavalca> yeah, we'll come up with something
16:14:01 <Eighth_Doctor> if ntfs3 was in mainline already, I'd do that
16:14:06 <Eighth_Doctor> because that's self-contained and totally new
16:15:53 <dcavalca> oh, the shadow inode thing, probably
16:16:00 <dcavalca> but it's not in mainline yet
16:16:18 <dcavalca> https://lore.kernel.org/linux-fsdevel/20210614211904.14420-4-hannes@cmpxchg.org/
16:17:14 <Eighth_Doctor> we should probably create a ticket for things we want to backport from mainline to the c9s kernel
16:17:17 <dcavalca> that's definitely not self container, so if we have something simpler that's probably better
16:17:22 <dcavalca> yeah, good call
16:17:40 <Eighth_Doctor> dcavalca: can you do that? you have a good idea of the stuff we'd want to do here
16:18:10 <dcavalca> #action dcavalca file ticket to track patches for 5.14 ark
16:18:18 <dcavalca> alright, let's move on
16:18:35 <dcavalca> the only other thing I had in announcements was a reminder for various conferences
16:18:43 <dcavalca> devconf.us starts tomorrow
16:18:53 <dcavalca> the cfp for dojo in october ends this week
16:19:01 <dcavalca> and the cfp for devconf.cz in jan just opened
16:20:11 <Eighth_Doctor> I suppose I can submit a talk about the Hyperscale Workstation for the october dojo
16:20:16 <Eighth_Doctor> that'll make rbowen happy
16:20:17 <dcavalca> I've submitted the usual sig update to Dojo
16:20:21 <dcavalca> Eighth_Doctor: please do
16:20:43 <dcavalca> I'll probably do a "c9 at FB" at devconf.cz, depending on how much progress we actually make
16:20:54 <dcavalca> someone else should submit a hyperscale talk there too
16:21:21 <dcavalca> (unclear whether devconf.cz is in person or not, but I'll likely attend remotely either way)
16:22:17 <dcavalca> I think that's all for announcements, so
16:22:21 <dcavalca> #topic Tickets
16:22:40 <dcavalca> bkircher: any update on #67 for the qemu package?
16:23:50 <rbowen> We do already have a hyperscale sig update on the (tentative) dojo schedule. Let me know if I should swap the speaker.
16:24:18 <dcavalca> rbowen: yeah I think that's the one I submitted
16:24:25 <rbowen> Yes
16:24:33 <dcavalca> Eighth_Doctor: was thinking of doing a separate one on the Workstation spin
16:24:45 <rbowen> Ah. Got it.
16:24:50 <dcavalca> but we could also just merge them into one talk if you already have enough submissions to fill the schedule
16:25:04 <Eighth_Doctor> yeah, I don't mind being pulled into the sig update talk
16:25:23 <rbowen> We currently have 7 submissions, so technically I have space for 2 more
16:25:26 <dcavalca> I'm fine either way, I'll probably improv this as usual anyway
16:26:17 <Eighth_Doctor> improv is the way to go :P
16:27:13 <dcavalca> did we have anything else for tickets?
16:27:19 <dcavalca> I think we already covered the kernel stuff before
16:27:55 <Eighth_Doctor> well, the cloud image stuff
16:28:13 <dcavalca> rbowen: btw, I'm trying to get two more people to submit talks for Dojo, hopefully they'll get them out soon
16:28:14 <Eighth_Doctor> as part of the prep for that, I've moved livemedia-kickstarts to spin-kickstarts
16:28:27 <Eighth_Doctor> and I'm working on the cloud image kickstarts for trying out in koji
16:28:54 <dcavalca> Eighth_Doctor: that's a different flow from the live media one?
16:29:23 <Eighth_Doctor> yes
16:29:40 <Eighth_Doctor> it uses ImageFactory and Oz
16:29:57 <dcavalca> yay more image creation tools
16:30:39 <Eighth_Doctor> so we'll be up to livecd-creator, imgfac, and appliance-creator
16:31:13 <Eighth_Doctor> arrfab wants me to switch to livemedia-creator at some point, but that requires beating on anaconda some more to make it do the right things
16:31:18 * dcavalca sights
16:31:25 <dcavalca> yeah if we can dedupe this it'd be nice
16:31:35 <Eighth_Doctor> oh there's no deduping here
16:31:39 <Eighth_Doctor> it would be a net-add
16:31:45 <dcavalca> lol awesome
16:31:46 <Eighth_Doctor> that's what makes this insane :(
16:31:59 <dcavalca> we also still need to finish the btrfs stuff in osbuild
16:32:03 <Eighth_Doctor> yeah
16:32:12 <Eighth_Doctor> that's going to become much more critical with 9
16:32:15 <dcavalca> I have zero bandwidth for that right now though
16:32:30 <Eighth_Doctor> I know, and I don't really have any bandwidth either
16:32:59 <ssmoogen> switch to osbuild :)
16:33:35 <ssmoogen> or are you using that already and need to switch to livemedia?
16:33:40 <dcavalca> ssmoogen: osbuild needs some non-trivial work to understand btrfs subvolumes properly
16:34:07 <ssmoogen> everything needs non-trivial work to understand btrfs subvolumes
16:34:11 <Eighth_Doctor> and osbuild still doesn't support most of the artifacts we want to produce
16:34:11 <dcavalca> we hacked together basic btrfs support as part of the f33 change but doing it properly is tricky
16:34:30 <dcavalca> yes, that too
16:34:38 <ssmoogen> [read through the lwn.net articles on getting nfs to understand it]
16:34:53 <ssmoogen> what artifacts are you wanting to produce?
16:35:22 <ssmoogen> i can't help on the btrfs side but I might be able to see what other artifacts
16:35:46 <Eighth_Doctor> disk images, install ISOs, live media ISOs, cloud images for AWS, Azure, and GCP, netboot images
16:36:30 <Eighth_Doctor> today, osbuild only supports disk images and cloud images
16:37:05 <ssmoogen> ah ok. thanks
16:37:55 <Eighth_Doctor> essentially, we want CentOS Hyperscale to exist as a full distribution of artifacts in the same way that CentOS Stream does
16:37:59 <Eighth_Doctor> or Fedora does
16:40:29 <Eighth_Doctor> dcavalca: I'd use kiwi for this, except there's no koji integration for it
16:40:52 <Eighth_Doctor> the koji devs weren't opposed to having support for kiwi, but it's not currently there
16:42:05 <dcavalca> yeah I suspect that's be a somewhat major project to add
16:44:05 <dcavalca> I think this is all for tickets
16:44:10 <dcavalca> #topic Membership
16:44:32 <dcavalca> don't think we have anything here this week?
16:45:54 <dcavalca> #topic Misc
16:46:30 <rbowen> Is there any chance that any of y'all would be at Supercomputing '21, and, if so, if you'd be willing to do a presentation there in the Red Hat minitheater about the Hyperscale effort.
16:46:45 <rbowen> Not an exact fit, but I think it might be of interest to the audience.
16:46:48 <dcavalca> rbowen: when / where is that?
16:47:08 <rbowen> Nov 15-18, St Louis
16:47:09 <rbowen> https://sc21.supercomputing.org/
16:47:25 <Eighth_Doctor> rbowen: in person or virtual?
16:47:32 <rbowen> It's in person, as of today.
16:47:44 <rbowen> SOrry I missed the end-of-meeting flag. I'm in that meeting right now.
16:48:18 <Eighth_Doctor> looks like it's a hybrid event
16:48:23 <dcavalca> I'm unlikely to be able to attend in-person conferences this year
16:48:24 <Eighth_Doctor> sure, we could probably do something for that
16:48:53 <dcavalca> but yeah, if they take remote speakers we can probably figure something out
16:49:03 <rbowen> Ok, I'll email some info once I'm out of this meeting.
16:49:47 <Eighth_Doctor> likewise
16:50:50 <dcavalca> there's also packagingcon in Nov which probably overlaps in interests with some of this crowd
16:50:59 <Eighth_Doctor> yeah
16:51:11 <Eighth_Doctor> is malmond going to submit a talk for that?
16:51:59 <dcavalca> Eighth_Doctor: I was just harassing him about that earlier :)
16:52:08 <rbowen> CFP closed yesterday
16:52:39 <dcavalca> iirc it got extended? at least that's what the email I got said
16:53:10 <dcavalca> yeah, it closes EOW
16:53:12 <Eighth_Doctor> yeah it was extended
16:53:33 <rbowen> Oh, the website doesn't reflect that. Cool.
16:57:58 <dcavalca> alright, I think that's is for today
16:58:04 <dcavalca> have a good one folks!
16:58:15 <dcavalca> #endmeeting