16:00:42 <dcavalca> #startmeeting Hyperscale SIG
16:00:42 <centbot> Meeting started Wed Apr 13 16:00:42 2022 UTC.  The chair is dcavalca. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:42 <centbot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:00:51 <dcavalca> #chair dcavalca jvreeland
16:00:51 <centbot> Current chairs: dcavalca jvreeland
16:00:54 <dcavalca> morning folks
16:00:56 <jvreeland> morning
16:00:59 <aekoroglu> morning
16:01:12 <kcwells> Mornin'!
16:01:14 <davdunc> hello.
16:01:23 <oidoming> hi
16:01:28 <Eighth_Doctor> hey folks!
16:02:30 <dcavalca> let's get started
16:02:32 <dcavalca> #topic Followups
16:02:41 <dcavalca> two things on my end
16:02:46 <dcavalca> first of all, the SIG report is out
16:02:53 <dcavalca> thanks everyone for the help writing this up
16:03:16 <dcavalca> #info hyperscale SIG report is out: https://blog.centos.org/2022/04/centos-hyperscale-sig-quarterly-report-for-2022q1/
16:03:28 <dcavalca> this got some decent press coverage as well, which is always nice
16:04:06 <dcavalca> the other thing I had is a heads up that you might see some fallout from the rpmdb path change in our Hyperscale build of the rpm stack
16:04:12 <davdunc> Great work on the report dcavalca
16:04:25 <dcavalca> or at least, we did on our end, so I've asked Eighth_Doctor to add a compat symlink like Fedora does
16:05:05 <dcavalca> anybody else has followups?
16:06:00 <Eighth_Doctor> not from me
16:07:00 <dcavalca> alright, let's move to
16:07:01 <dcavalca> #topic Announcements
16:07:19 <dcavalca> I'm in the process of updating our kpatch backport
16:07:34 <dcavalca> just tagged it for c8s, and will do c9s afterwards
16:07:54 <dcavalca> I also built compsize yesterday, but as mentioned in the channel there's still some issues with dist-git layouts for c8s
16:08:13 <dcavalca> bstinson: I think we're missing the right version of centpkg-minimal in our buildroot there
16:08:30 <dcavalca> the c9s build is fine and I've already tagged it
16:08:32 <DaanDeMeyer[m]> Hi everyone, sorry for being late
16:08:43 <dcavalca> hey DaanDeMeyer[m]
16:08:50 <dcavalca> we're doing announcements
16:09:03 <Eighth_Doctor> dcavalca: well, it's because we have an override package that I want to remove once bstinson has figured it all out with the baseline centpkg
16:09:13 <Eighth_Doctor> *centpkg-minimal
16:09:19 <dcavalca> Eighth_Doctor: did we have an infra ticket for this already?
16:09:25 <Eighth_Doctor> yes
16:09:36 <Eighth_Doctor> I believe Daan De Meyer filed it
16:09:49 <dcavalca> https://pagure.io/centos-infra/issue/678
16:10:02 <dcavalca> I'll bump that, thanks
16:11:05 <dcavalca> anybody else has announcements?
16:11:36 <Eighth_Doctor> I'm planning to do CentOS Hyperscale 9 images as GPT-by-default for BIOS
16:11:47 <Eighth_Doctor> as a step toward upstream deprecation of BIOS
16:12:08 <osandov> I updated btrfs-progs to v5.16.2 (plus some Facebook-specific patches). That's already tagged for c8 and c9
16:12:16 <Eighth_Doctor> our cloud images will be hybrid boot, but our install media and live media will prefer GPT even on BIOS systems
16:12:57 <DaanDeMeyer[m]> systemd 250.3-6.5 with hopefully all the package conflicts with EPEL stuff fixed
16:12:57 <DaanDeMeyer[m]> *is released
16:13:40 <Eighth_Doctor> we're going to have a meeting with the RHEL kernel folks in the office hours later today to talk about the stuff we plan to contribute and maintain
16:14:39 <dcavalca> lots of great work here, lemme tag some of these for the minutes
16:15:01 <dcavalca> #info systemd 250.3-6.5 released with fixes for EPEL package conflicts
16:15:12 <dcavalca> #info btrfs-progs updated to v5.16.2
16:15:33 <dcavalca> #info CentOS Hyperscale 9 images will be GPT-by-default for BIOS, Cloud images will be Hybrid boot
16:16:13 <dcavalca> #info kpatch-0.9.6-2.9.hs.el8 release, c9s build coming up
16:16:34 <dcavalca> #info compsize is now build for c9s in Hyperscale, c8s build pending on https://pagure.io/centos-infra/issue/678
16:16:44 <dcavalca> alright, anything else?
16:17:04 <Eighth_Doctor> I don't have anything
16:18:54 <dcavalca> cool, time for
16:18:56 <dcavalca> #topic Tickets
16:19:27 <dcavalca> I've filed a few documentation tickets
16:19:36 <dcavalca> which reminds me, I've forgot an announcement
16:20:02 <dcavalca> we now have an ELN Extras workflow for hyperscale: https://tiny.distro.builders/workload-overview--eln_extras_hyperscale--repository-fedora-eln.html
16:20:41 <dcavalca> the idea is to track here packages from EPEL that we care about, so that we can ensure they stay in good shape and will eventually become available in epel10
16:21:06 <dcavalca> source is at https://github.com/minimization/content-resolver-input/blob/master/configs/eln_extras_hyperscale.yaml
16:21:19 <dcavalca> feel free to add more, I'm going to get this documented properly
16:21:38 <dcavalca> #info ELN Extras workflow for Hyperscale: https://tiny.distro.builders/workload-overview--eln_extras_hyperscale--repository-fedora-eln.html
16:22:00 <dcavalca> back to tickets, let's see
16:22:11 <dcavalca> I'm taking care of #96 for compsize
16:22:35 <dcavalca> jvreeland: is there anything left to do for aarch64 in #71?
16:22:54 <chantra_> dcavalca, may need to merge https://pagure.io/centos-sig-hyperscale/sig/pull-request/108
16:23:04 <chantra_> I fixed the typo
16:23:21 <jvreeland> dcavalca: just rebuilding the -release packages I think. I've pinged a few people but don't think it's been rebuilt yet.
16:23:24 <dcavalca> chantra: oh thanks, I don't think pagure sends emails on updates
16:23:29 <dcavalca> I'll merge that now
16:23:50 <jvreeland> tbh not sure who the right person is anymore
16:23:56 <dcavalca> jvreeland: I _think_ we can do that ourselves now?
16:24:08 <jvreeland> I though that was only for C9?
16:24:20 <jvreeland> *thought
16:24:40 <dcavalca> https://sigs.centos.org/guide/delivery/#centos-release-package seems to imply it's for both
16:24:53 <dcavalca> Eighth_Doctor iirc you were the last one to mess with this
16:25:42 <Eighth_Doctor> the release packages?
16:25:43 <Eighth_Doctor> in what way?
16:25:51 <Eighth_Doctor> I've done the c9s process, but not the c8s ones
16:25:55 <dcavalca> I meant building a release package
16:25:58 <dcavalca> oh alright
16:26:07 <dcavalca> jvreeland: I'd say give it a try, and file an infra ticket if it doesn't work
16:26:10 <jvreeland> i'll try it out and check permissions
16:26:38 <dcavalca> jvreeland: while you're here, any kernel updates of note?
16:26:44 <Eighth_Doctor> I believe only dcavalca and I have permissions on the release packages
16:26:54 <Eighth_Doctor> that's what bstinson originally setup, iirc
16:27:02 <jvreeland> dcavalca: no updates
16:27:25 <Eighth_Doctor> but yeah the process for c9s looks right
16:27:27 <Eighth_Doctor> no idea about c8s
16:27:37 <Eighth_Doctor> since that was all done by carlwgeorge and dcavalca
16:27:46 <dcavalca> Eighth_Doctor: yeah you're right, https://accounts.centos.org/group/sig-extras/ is the group
16:27:54 <dcavalca> jvreeland: so, file a ticket to infra to ask to get added to that
16:28:19 <jvreeland> will do
16:28:34 <dcavalca> anybody else has ticket updates?
16:28:53 <Eighth_Doctor> still working through the images for c9s
16:29:08 <Eighth_Doctor> trying to get the live media working without crashing or breaking
16:29:30 <Eighth_Doctor> made updates to livesys-scripts to try to improve things: https://bodhi.fedoraproject.org/updates/FEDORA-EPEL-2022-00f288cc10
16:29:39 <kcwells> Not for a ticket in our pagure -- I'm just still waiting on feedback from bstinson on Infra ticket #694
16:29:51 <kcwells> I just sent another follow-up email response to his
16:31:03 <dcavalca> speaking of infra, good news, sounds like the pagure upgrade work will start next month: https://pagure.io/centos-infra/issue/456#comment-792235
16:31:12 <dcavalca> which will hopefully get us working PRs and stuff for git.c.o
16:31:58 <osandov> Nice
16:32:01 <Eighth_Doctor> yay
16:32:23 <DaanDeMeyer[m]> That'll be useful
16:33:46 * chantra_ heading off ttyl
16:34:22 <dcavalca> if there's nothing else, let's do
16:34:26 <dcavalca> #topic Membership
16:34:32 <dcavalca> I don't think we have anything here this week?
16:35:17 <dcavalca> #topic Misc
16:35:21 <Eighth_Doctor> nope
16:35:38 <dcavalca> one last thing I wanted to bubble up, https://bugzilla.redhat.com/show_bug.cgi?id=2068423 hasn't been fun to deal with on my end
16:35:40 <Eighth_Doctor> I realized you forgot to announce our c9s container :)
16:35:46 <dcavalca> so if you're deploying c9s, you may wanna pay attention to that
16:35:53 <dcavalca> Eighth_Doctor: that was in the SIG report already :)
16:36:11 <Eighth_Doctor> ahh right blech
16:36:14 <Eighth_Doctor> I didn't read it :)
16:37:42 <Eighth_Doctor> (well, not today anyway)
16:38:40 <Eighth_Doctor> my merge request for adding libva-devel to all arches was merged today
16:39:06 <Eighth_Doctor> https://gitlab.com/redhat/centos-stream/release-engineering/comps/-/merge_requests/220
16:39:23 <Eighth_Doctor> this helps for adding ffmpeg to EPEL 9 in the future
16:39:44 <Eighth_Doctor> and may be useful for folks doing hw-accelerated encoding and decoding with Hyperscale machines
16:41:20 <aekoroglu> We would like to add glibc, zlib, R and openblas packages into hyperscale experimental repo, intel tagged packages might take some time as it seems. How we gonna do that, which procedures we should follow
16:42:45 <dcavalca> aekoroglu: where/how would the experimental builds for these differ?
16:42:46 <aekoroglu> I was testing our optimization patches for those packages rpminspect shows no abidiff too. So we would like to run at least in exp
16:42:52 <dcavalca> glibc in particular seems particularly scary
16:43:50 <dcavalca> if this is specifically optimizations, a dedicated tag might make more sense, so users know what they're getting
16:44:16 <Eighth_Doctor> yeah, if you're going to do stuff like that, it might make sense to be dedicated tags+branches
16:44:17 <aekoroglu> yes they are
16:44:24 <dcavalca> also, I don't see R in centos at all, so you'll likely have to backport that from Fedora
16:44:40 <dcavalca> and having done some of that internally a long time ago, I expect it won't be particularly fun, so heads up
16:44:54 <aekoroglu> yeap we already have R in our repo
16:45:34 <Eighth_Doctor> it might be worth also asking for copr support for hyperscale buildroots
16:45:47 <dcavalca> Eighth_Doctor: I'd asked before
16:45:54 <Eighth_Doctor> some of the stuff we're trying to do might not make sense as something we push into CBS
16:46:07 <dcavalca> https://pagure.io/copr/copr/issue/2057
16:46:08 <Eighth_Doctor> dcavalca: oh?
16:46:13 <dcavalca> well, not me but someone had
16:46:35 <dcavalca> aekoroglu: are these optimizations intel-specific, or would they apply to all platforms?
16:46:56 <aekoroglu> intel specific
16:47:05 <dcavalca> ok, in that case you definitely want a dedicated tag for these
16:47:27 <aekoroglu> %50 efficiency increase zlib for example
16:47:32 <Eighth_Doctor> yeah, and we'll probably want it to have its own disttag (.hs+intel)
16:47:49 <Eighth_Doctor> wtf
16:47:49 <Eighth_Doctor> that's pretty high
16:48:00 <Eighth_Doctor> is this with zlib or zlib-ng?
16:48:06 <aekoroglu> zlib
16:48:16 <Eighth_Doctor> O.o
16:48:22 <aekoroglu> backported patch set from intel-zlib version
16:48:39 <aekoroglu> zlib-ng has our changes too
16:48:43 <dcavalca> we can file a ticket for infra to request a tag
16:49:15 <dcavalca> e.g. hyperscaleXs-packages-intel-{candidate,release,testing} and `.hs+intel.elX` as the disttag
16:49:20 <dcavalca> does that work?
16:49:26 <aekoroglu> would be perfect
16:49:47 <kcwells> aekoroglu -- to clarify, yes these are intel-specific optimizations, but the patches build successfully on other platforms
16:50:10 <kcwells> Right?
16:50:25 <aekoroglu> sure Im also testing for armv7, aarch64, s490 and ppc too :)
16:50:42 <aekoroglu> 390 :P
16:51:21 <dcavalca> if these don't introduce regressions for !intel platforms, IMO you should aim to get them upstream in the relevant projects
16:51:43 <dcavalca> but in the meantime, I think building these in a dedicated tag here is fine, and will give users the option to use them if they'd like
16:51:48 <Eighth_Doctor> yes
16:51:59 <Eighth_Doctor> and we can also discuss getting this upstreamed in Fedora and backported to RHEL
16:53:39 <kcwells> Yeah, for each of the packages aekoroglu mentioned, the patches have been submitted upstream one way or another. With zlib in particular, the maintainer gave it a cursory "looks good, but I'm too busy to work on zlib right now" response
16:57:56 <dcavalca> https://pagure.io/centos-infra/issue/748 -- please double check that I got this right
16:58:39 <Eighth_Doctor> dcavalca: `hs+intel`, not `hs.intel`
16:58:55 <Eighth_Doctor> also it needs to be layered on EPEL
16:59:00 <Eighth_Doctor> they don't do that by default
16:59:15 <dcavalca> fixed
16:59:33 <kcwells> Yeah sounds good with those changes
16:59:40 <aekoroglu> yeap thanks
16:59:47 <Eighth_Doctor> do you not want aarch64 too?
17:00:04 <Eighth_Doctor> not sure if you care about that or not (might matter from an upstreaming perspective)
17:00:07 <dcavalca> I don't think it makes a ton of sense, given that these are intel-specific optimizations
17:00:23 <Eighth_Doctor> 👍️
17:00:25 <dcavalca> from a user standpoint I mean
17:00:30 <dcavalca> alright, we're out of time
17:00:38 <Eighth_Doctor> yup
17:00:41 <dcavalca> thanks everyone, and see you at office hours in a moment
17:00:45 <dcavalca> #endmeeting