18:00:22 <dcavalca> #startmeeting Hyperscale SIG
18:00:22 <centbot> Meeting started Wed Mar  3 18:00:22 2021 UTC.  The chair is dcavalca. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:00:22 <centbot> Useful Commands: #action #agreed #help #info #idea #link #topic.
18:00:26 <dcavalca> morning folks
18:00:34 <dcavalca> #chair dcavalca jvreeland
18:00:34 <centbot> Current chairs: dcavalca jvreeland
18:00:42 <smooge> hello
18:00:48 <dcavalca> let's wait a few mins for folks to join
18:01:01 <anitazha> morning
18:02:16 <jvreeland> morning
18:02:38 <dcavalca> #topic Followups
18:02:48 <dcavalca> ok, let's review actions from the last meeting
18:03:18 <dcavalca> I said I'd followup with bstinson on secure boot, and I totally forgot, so I'll take that again :)
18:03:43 <dcavalca> #action dcavalca follow up with bstinson wrt secure boot and kernel builds
18:03:59 <dcavalca> jvreeland: I saw you created the linux repo on pagure, any other updates there?
18:05:00 <jvreeland> I pushed the kernel source tree to pagure. I have a local version of the rpms/kernel take from fedora but it's failing to finish the build at towards the end when working with the vmlinuz and i haven't gotten to debugging it. I'm moving next week so probably won't have much time before the next meeting either.
18:05:30 <dcavalca> thanks
18:05:40 <dcavalca> #info jvreeland pushed kernel source tree to pagure
18:06:11 <dcavalca> jvreeland: you also had an action to follow up on pbisaacs membership request
18:06:13 <jvreeland> Also got a response from Paul Isaacs, not sure if that should wait till the memeber ship part or now.
18:06:19 <dcavalca> ah perfect
18:06:51 <dcavalca> eh, I figured we'd do all followups at the top so we don't forget
18:08:11 <jvreeland> He said that him and his team is experienced with "Apache Bigtop, Tensorflow, Pytorch, OpenStack, OpenHPC all on Arm" and has available " ThunderX2, Kunpeng920 and A64FX systems." for their testing resources and that the hyperscale seems like a good fit.
18:08:43 <dcavalca> works for me, though we should get them to join IRC and come to these meetings
18:08:57 <dcavalca> also, right now we only have x86_64 in koji
18:09:31 <dcavalca> if we want to add aarch64 it's doable, but it'll require some infra work, and obviously someone to commit shepherding that the build actually work
18:09:31 <jvreeland> Agreed, I reminded them about the meeting and was hoping they'd show up. Should we make a pagure ticket to talk in a more open place about how they might join to help?
18:09:59 <dcavalca> jvreeland: yeah that works for me
18:10:20 <dcavalca> #action jvreeland create ticket to continue discussion with pbisaacs in public
18:10:31 <dcavalca> let's see, what's next
18:10:50 <dcavalca> filbranden had an action to followup with the Cloud SIG but he's not around today, so I'll put this back in the queue
18:11:01 <dcavalca> #action filbranden followup with Cloud SIG on how to build Hyperscale cloud images
18:11:20 <dcavalca> King_InuYasha: had an action to followup with pgreco on kernel builds for altarch, any update there?
18:13:57 <dcavalca> uh, not sure if King_InuYasha is around, we can pick that back up later in case
18:14:21 <dcavalca> let's move on for now to
18:14:28 <dcavalca> #topic Announcements
18:15:10 <dcavalca> I gave a talk at DevConf.cz which went well, covering the SIG work among other things
18:15:29 <dcavalca> I also submitted something to LISA to rbowen suggestion, we'll see if that gets picked up
18:16:19 <dcavalca> #info DevConf.cz talk, video tbd
18:16:33 <dcavalca> #info submitted talk to LISA CfP, tbd
18:16:48 <dcavalca> the only other announcement I have is that I pushed a new grep package to the SIG
18:17:11 <dcavalca> which has major perf improvements (like, 30% according to upstream)
18:17:21 <dcavalca> so if you happen to use grep a lot, you may wanna check it out
18:17:28 <jvreeland> always nice to have a faster grep
18:17:40 <dcavalca> #info pushed grep package to the sig with major perf improvements
18:17:52 <dcavalca> anybody has any other announcements?
18:18:17 <jvreeland> none here
18:19:01 <rbowen> I just announced a Dojo too! Hope to see your submissions there, too!
18:19:21 <dcavalca> oh sweet
18:19:39 <dcavalca> #action dcavalca check out CfP for Dojo
18:19:53 <dcavalca> thanks rbowen
18:20:08 <dcavalca> alright, let's move on to
18:20:11 <dcavalca> #topic Tickets
18:20:30 <dcavalca> first one is #27 setup systemd staging repo
18:20:41 <dcavalca> so that's mine, and I mostly wanted to give a little context
18:20:51 <dcavalca> FB has an internal repo for systemd that we use to stage patches and stuff
18:21:18 <dcavalca> where the main branch tracks systemd upstream, and then we have feature branches for the various releases we build (e.g. fb-246, fb-247, etc)
18:21:41 <dcavalca> we found this workflow works fairly well, but there's not really any reason for this to be FB specific, so we wanted to move it to a public repo
18:22:01 <dcavalca> I've created https://pagure.io/centos-sig-hyperscale/systemd for this purpose
18:22:12 <dcavalca> but need to figure out a sane way to keep the main branch in sync with upstream
18:22:26 <dcavalca> pingou on the ticket suggested using https://pagure.io/fedora-infra/toddlers
18:22:36 <dcavalca> which I'm not familiar with, but looks like it might work
18:23:04 <dcavalca> if anybody has other ideas or suggestions I'm definitely open to that
18:23:41 <dcavalca> an obviously, the idea is that we'd be able to use this to collaborate on systemd in general, it's not a FB specific thing
18:24:05 <jvreeland> no new suggestions, just want to make sure i'm understanding, the flow right. When you say upstream do you mean systemd or systemd-stable. iirc the facebook systemd rpm comes from stable?
18:24:34 <dcavalca> jvreeland: good question, I believe we've been tracking systemd.git so far for this
18:24:38 <dcavalca> anitazha: ^^ can you confirm?
18:25:08 <anitazha> the rpms are built on systemd-stable. the FB git repo has branches for both systemd.git and systemd-stable.git
18:25:10 <dcavalca> jvreeland: in general, we try to upstream patches as much as possible, so this repo mostly ends up carrying either backports, or stuff that's currently being tested
18:25:32 <anitazha> so i guess we'd want to stage on systemd-stable. i use the systemd.git branch for bleeding edge testing
18:25:32 <dcavalca> cool, we should replicate the same setup here then I think
18:25:58 <dcavalca> yeah, we can have both, I don't think there's any downside here
18:25:58 <jvreeland> awesome, thanks
18:26:18 <dcavalca> I'll followup on the toddlers thing and see if that works for doing the automated sync
18:26:53 <dcavalca> #action dcavalca followup on toddlers for syncing systemd repo
18:27:00 <dcavalca> anything else on this one?
18:27:43 <dcavalca> ok, so the next one I have tagged is
18:27:45 <dcavalca> #23 systemd 247 fails to boot if selinux is enabled
18:27:49 <dcavalca> so, this is "fun"
18:28:06 <dcavalca> turns out we'd never actually tested our systemd backport against the upstream selinux policy
18:28:19 <dcavalca> because FB doesn't really use selinux (and where we do, it's with a custom policy)
18:28:46 <dcavalca> I don't know anything about selinux, but I managed to hack together a patch that's good enough to make the system boot
18:29:32 <dcavalca> still, I'm not comfortable releasing this as-is, as a bunch of daemon are outright broken
18:30:03 <dcavalca> does anybody here have selinux experience and can have a look?
18:30:18 <dcavalca> I think this is the only actual blocker to tagging systemd 247 for -release
18:30:49 <dcavalca> the alternative option, I guess, is that we could say we just don't support selinux
18:31:06 <dcavalca> but that seems kind of crappy, and even in that case we shouldn't release packages that will brick a default install with it enabled
18:31:30 <jvreeland> yeah that doesn't seem like a good option.
18:31:37 <anitazha> is this new to 247 or do prior releases also brick
18:31:51 <dcavalca> it likely affects previous releases as well, we just never tested it
18:32:01 <dcavalca> https://git.centos.org/rpms/selinux-policy/c/7cee3bf1e3294a5a5b9733d488e683d38c9792fb?branch=c8s-sig-hyperscale is the policy patch fwiw
18:34:33 <jvreeland> Any chance we could ping the maintainers of of the current policy on cent/fedora for took take a quick look, that looks fine to me but i also don't know too much about selinux.
18:35:01 <dcavalca> jvreeland: yeah, that was going to be my next step, but wanted to see first if we had someone here with domain specific expertise
18:37:20 <dcavalca> looks like not, so I'll move forward with that and put updates in the ticket
18:37:35 <dcavalca> the other two tickets we have are the same as the last meeting
18:37:51 <dcavalca> I think we already covered the kernel stuff earlier with jvreeland's update
18:38:00 <dcavalca> and I don't think we have any news for cloud
18:38:05 <dcavalca> does anybody have anything else here?
18:38:49 <dcavalca> #action dcavalca followup with selinux-policy maintainer on systemd 247 issue
18:39:42 <dcavalca> ok, let's move on to
18:39:46 <dcavalca> #topic Membership
18:39:58 <dcavalca> does anybody have anything here?
18:40:27 <dcavalca> the queue only has pbisaacs, which we'd already discussed before
18:41:09 <dcavalca> ok, let's close with
18:41:12 <dcavalca> #topic Misc
18:41:29 <dcavalca> smooge: mentioned this meeting conflicts with the Fedora Server one
18:41:35 <dcavalca> so we should probably reschedule it
18:42:25 <jvreeland> seems reasonable
18:42:36 <dcavalca> I'll see if I can find another hole in the centos calendar and follow up
18:42:49 <dcavalca> ideally I'd keep it around this time as I think it works well for everybody
18:43:07 <dcavalca> #action dcavalca reschedule to avoid conflict with Fedora Server meeting
18:43:37 <dcavalca> the only other thing I wanted to mention, is that we should put together some kind of announcement/update once we have something that people can actually use
18:44:03 <dcavalca> we do already have a centos-release-hyperscale package and a repo
18:44:17 <dcavalca> I'd say probably once systemd is sorted out this would be a good opportunity
18:44:43 <dcavalca> I was hoping to do this last week but then the selinux thing came up and kinda derailed things
18:45:47 <dcavalca> anybody has anything else they'd like to discuss?
18:48:31 <dcavalca> ok, I think that's it for this one
18:48:35 <dcavalca> thanks everybody
18:48:45 <dcavalca> #endmeeting