[FrontPage] [TitleIndex] [WordIndex

This is a read-only archived version of wiki.centos.org

Attending

Updates

1. Xen Project

Next:

No blockers

2. Docker

Agreements:

3. Going forward / open questions

4. Process / Terminology

5. Meeting at LinuxCon in Dusseldorf?

6. Complete log

* lsm5 is here
<lars_kurth> lars_kurth just arrived for the Virt SIG meeting
<gwd> lsm5: You're Lokesh, right?  [13:58]
<lsm5> gwd: yup
*** richardm75 (~richardm7@c-24-218-110-68.hsd1.ma.comcast.net) has joined
    channel #centos-devel  [14:00]
<lars_kurth> So we have George, Lokesh, KB? Anyone I missed?  [14:02]
<gwd> Is kbsingh actually here?  [14:03]
<kbsingh> i am
<gwd> Excellent.
<lars_kurth> Cool. Can someone do me a favour and send me the log afterwards?
             My client produces badly formatted logs  [14:04]
<gwd> lars_kurth: I can do.
<lars_kurth> Thank you
<kbsingh> i have logs as well
<kbsingh> lsm5: here ?
<lsm5> kbsingh: yup
<lars_kurth> What do we need to cover? Reading through notes from 2 weeks
             ago. Seems we discussed Docker, Build System, Upstreams,
             Upstreams/Kernel and a few other items  [14:05]
<gwd> I think a quick update on the different "projects" would be helpful
<gwd> ("projects" being my tentative term for Xen, Docker, &c)
<kbsingh> sounds good
<lars_kurth> gwd: Do you want to start?
<gwd> Sure -- I got Xen 4.2.4 for C6 to build in the community build system
      yesterday.  [14:06]
<gwd> Still have to test it, but I think I'm getting the hang of it
<gwd> Next would be 4.4 for C6, then 4.4 for C7
<gwd> Then Linux 3.14 for C6, then Linux 3.14 for C7  [14:07]
<gwd> Still a lot of work to do...
<gwd> That's where Xen is.
<gwd> lsm5, any update on Docker?  [14:08]
<lars_kurth> Any blockers/dependencies ... or just time to get through
             everything?
<lsm5> docker 1.2.0  done
<lsm5> libcontainer 1.1.0 done  [14:09]
<gwd> At the moment no blockers.  kbsingh and alphacc have been pretty
      helpful.
<lsm5> these are working
<gwd> lsm5: They're building in the build system...?
<lsm5> gwd: yup, on cbs
<lsm5> docker-registry 0.8.1 built, but crashes (gotta dig in further)
<lsm5> cockpit built, but that one crashes too :(  [14:10]
<lsm5> a lot of the kubernetes deps built, but not kubernetes itself
<lsm5> that's about it
<lsm5> would be great to have more testers and debuggers for these
<lsm5> if anyone's interested  [14:11]
*** giallu__ (~giallu@87.18.185.235) has joined channel #centos-devel
<gwd> lsm5: I see docker in the "bananas7" repo -- but that's not in the
      virt7-testing yet.
<lars_kurth> Did you send a mail to the mailing list? That would be a good
             first step
<lsm5> gwd: i though I was only supposed to put it in bananas7 /cc kbsingh
                                                                        [14:12]
<lsm5> lars_kurth: ack, I'll email the list
<gwd> lsm5: Sorry, poorly worded -- I just meant to ask. :-)
<lars_kurth> gwd, lsm5: do we need these in virt7-testing for people to have a
             go? Sorry if it's a stupid question
<lsm5> gwd: aah ok :)  [14:13]
<gwd> lars_kurth: I don't think so, but bananas may have a bunch of other
      random stuff in it as well.
<lars_kurth> that's fine with me
<kbsingh> the bananas' target was mostly for people to test things etc
<lars_kurth> perfect
<kbsingh> once we have a plan for the repo layouts from the virt-sig, we can
          rebuild into other targets as needed  [14:14]
<gwd> kbsingh: For CBS users to test CBS, not for CentOS users to test
      packages, right? :-)
<kbsingh> in the mean time, we can hand grab the rpms and cherry pick what is
          needed into testing / devel repos
<kbsingh> gwd: yeah
<kbsingh> ideally, no users should be hitting the repos at cbs.centos.org -
          they should get their content elsewhere
<lsm5> so once packages are in bananas7, can I make a .repo file out of it?
       what's the correct URL for it?
<gwd> cbs.centos.org/repos/ has the tags, but that's generated from a cron job
      every hour (?)  [14:15]
<gwd> According to what I understand...
<gwd> I haven't used it as a repo yet (that's the next step)
<lsm5> hmm  [14:16]
<gwd> kbsingh: users> But if people want to test things, we can point them to
      the *-testing repos on cbs, right?
<kbsingh> so, once packages are built - something called 'mash' is run to
          collect and push into repos
<kbsingh> i dont think that is running as yet
<kbsingh> alphacc: would need to confirm  [14:17]
<kbsingh> gwd: i guess so - but then they get all the packages in the
          buildtarget
<gwd> kbsingh: alphacc told me yesterday there was a cron job running every
      hour to pull packages into cbs.centors.org/repos
<kbsingh> ah ok, that then is the mash run  [14:18]
<lars_kurth> and there are definitely RPMs in
             http://cbs.centos.org/repos/bananas7-testing/i386/os/Packages/
             (and other directories)  [14:19]
<gwd> kbsingh: well if we're worried about "collisions" I guess we'll have to
      solve that at some point.
<kbsingh> we should be good for that  [14:20]
<kbsingh> my main concern is how much b/w we have going into that machine
<kbsingh> for hundreds of users we should be ok - but if it grows beyond, we
          might have a situation
<gwd> So, should lsm5 start building his packages with the target
      virt7-testing once he's reasonably confident he knows what he's doing,
      and then point people to cbs to test it?
<kbsingh> sure  [14:21]
<kbsingh> how are we going to have this show up publicly down the road
<kbsingh> is docker stuff going to be in the virt repos ?
<kbsingh> along with xen
<kbsingh> or do we want a docker specific repo
*** pj (~pj@202-154-141-53.ubs-dynamic.connections.net.nz) has quit: Read
    error: Connection reset by peer
<kbsingh> ( eg. things like cockpit and mesos etc - where do we want those
          ending up )
<gwd> I'm easy either way.  I don't think there will be a collision between
      our packages.
<gwd> What are cockpit and mesos?  [14:22]
<kbsingh> cockpit is a webui to manage stuff on the machine.. mesos is a job
          orchestration thing  [14:23]
<gwd> I guess the risk would be that someone adds the Virt SIG repo for Xen
      and gets updated packages from the Docker project, or vice versa
<kbsingh> yea  [14:24]
<kbsingh> specially if kernels get included
<lars_kurth> BTW: I am adding key points of this conversation to
             http://wiki.centos.org/SpecialInterestGroup/Virtualization/2014-September23-minutes
             - feel free to modify afterwards
<gwd> Well can we take a timeout for a second, and ask whether this "project"
      "maintainer" terminology makes sense?
<gwd> Since we've been using it for hte last 15 minutes...  [14:25]
<lars_kurth> OK
<lars_kurth> I can't see a mention in the last set of minutes
*** bbankes (~bbankes@67-2-219-238.slkc.qwest.net) has quit: Ping timeout: 246
    seconds
<lars_kurth> Can you clarify?  [14:26]
<gwd> lars_kurth: Sorry, I sort of made it up as I was writing up the wiki:
<gwd> http://wiki.centos.org/SpecialInterestGroup/Virtualization/Roadmap
<lars_kurth> OK  [14:27]
<gwd> Basically, the idea would be that within the virt sig we have
      "projects", like Xen, Docker, oVirt, libvirt, &c
*** jitseklomp (~jitse@square.phys.uu.nl) has joined channel #centos-devel
<gwd> Each project would have an associated set of rpms with them (e.g., Xen
      would be xen, kernel, ipxe, seabios -- at least for C6)  [14:28]
<lars_kurth> That is fine. Or we could call them teams, ... I don't have a
             strong view, only that project is overloaded
<lars_kurth> And then we have a maintainer per project?  [14:29]
<gwd> And each project would have maintainer(s) who was in charge of pulling
      in / updating changes to those rpms
*** elguero (~miguel323@2001:470:1f06:12c4::2) has quit: Ping timeout: 272
    seconds
*** vu (~vu@ec2-54-72-114-143.eu-west-1.compute.amazonaws.com) has joined
    channel #centos-devel
<lars_kurth> So basically, each project needs to have at least ONE maintainer
             to be viable. Seems to be pretty standard terminology to me
                                                                        [14:30]
<lars_kurth> Do we all agree that this is sensible?
<lars_kurth> I mane gwd's proposal  [14:31]
<lars_kurth> mean
<gwd> projects> "team" is people oriented, where "project" is focused on the
      product.  I'm open to other terminology here, but I can't think of
      something off the top of my head.  [14:32]
<kbsingh> yes, but we might need to define what is a project
<kbsingh> would all the xen effort be on 1 project... would that then also
          include a management layer like xenorchestra ?
<gwd> kbsingh: I would experct xenorchestra and xapi to be different projects
                                                                        [14:33]
<kbsingh> for stuff like libvirt, if we get a newer upstream version, we
          should certainly inisit that we get someone who has a finger in that
          pie to come along
<kbsingh> but there might be some leaf nodes sort of thing that would suffer
          from this
<gwd> kbsingh: leaf nodes> what do you mean?  [14:34]
<kbsingh> maybe if we word it so its 'every component needs someone active
          around it'
<lars_kurth> gwd, kb: I suppose this comes down to the "scope" of the SIG. I
             think the practical approach would be to focus on lower level
             stuff first and  deal with the higher level stuff later (if
             needed and there is pull)
<kbsingh> gwd: leaf nodes, as in simple things that add value to what the main
          project effort is - eg. cockpit around docker
<lars_kurth> kbsingh: I like the term component better than project
<gwd> lars_kurth: See, I would call seabios a component of the Xen project
                                                                        [14:35]
<kbsingh> gwd: right, so we need someone actively looking after seabios
<lars_kurth> OK,. But you could argue that seabios is a package within the Xen
             Project component  [14:36]
<gwd> Yes; but seabios is tightly coupled with Xen, so as long as no one else
      wants to use it, the Virt SIG Xen project will just make sure it works
      for them.
<gwd> If someone else decides they need seabios for some other reason, we can
      talk about moving it somewhere else and introducing a dependency,
      depending on the situation.  [14:37]
<gwd> lars_kurth: But "package" is a technical term relating to koji, yum, &c.
                                                                        [14:38]
<kbsingh> i think we will need *bios for the qemu stuff
<gwd> Also, "component" implies something closely coupled with something else.
      Xen and Docker are pretty independent; I wouldn't call them both
      "components" of the Virt SIG.  [14:39]
<lars_kurth> OK. So to restate  : a project (or whatever we call it) is a
             collection of packages or components (chose as desired) looked
             after by one or several maintainers
<kbsingh> lars_kurth: that works
<lars_kurth> gwd, lsm5: any objections?  [14:41]
*** giallu__ (~giallu@87.18.185.235) is now known as giallu
<gwd> Do we want to say anything about them being related to accomplishing one
      logical goal?
<gwd> Or being integrated / interdependent somehow?
<kbsingh> as long as its virtsig relevant, we should be ok
<lsm5> lars_kurth: nope  [14:42]
<lars_kurth> I agree with kbsingh
<gwd> I think people will get the idea anyway. :-)
<gwd> So kbsingh asked about whether we should have different repos for Xen
      and Docker.  [14:43]
<lars_kurth> Let me add '''a project within the Virt SIG is a collection of
             packages or components (chose as desired) looked after by one or
             several maintainers that delivers the goals of the SIG''' to the
             minutes
<lars_kurth> That's slightly clearer
<gwd> Sure, but it doesn't tell you why you might group things together one
      way and not another.
<gwd> I meant, the Xen Project is one logical thing: get the hypervisor
      working.  It has several components: Xen, a dom0 kernel, and supporting
      packages (ipxe, seabios)  [14:44]
<gwd> XenOrchestra is another logical thing: get Xen Orchestra up.  That may
      involve having just one package, or other support packages.
<lars_kurth> But we don't have an argument at the moment. My pragmatic
             approach would be to revisit when we get to a case where this is
             not clear  [14:45]
<gwd> Docker is one logical thing, but apparently it involves a ton of
      supporting packages. :-)
*** holmja (~holmja@public.kdl.org) has joined channel #centos-devel
<lsm5> it does :/
*** Epic|rcl|afk (~rcl@ec2-54-229-171-156.eu-west-1.compute.amazonaws.com) is
    now known as Epic|rcl
<gwd> lsm5: Are they mostly packages that don't exist in C7, or are they
      updated versions of what you find in C7?  [14:46]
<lsm5> many are updates, some new  [14:47]
* lsm5 is logged in via phone...home wifi blew up :(
<gwd> Right.  Well it might be surprising to people to install Xen and get a
      bunch of updates to unrelated packages...  [14:48]
<kbsingh> yah, thats a good point - docker is going to need its own repo,
          since we have an older docker in centos7 as well
<lsm5> kbsingh: or should it go in atomic?
<gwd> kbsingh: Does that mean virt7-docker-* 
<kbsingh> i believe from the centos project side - the sigs are a way into the
          packaging system  [14:50]
<kbsingh> you dont need to own the resulting repo
<kbsingh> so its possible for a SIG to feed into CentOS-Extras/
<kbsingh> or CentOS-Plus/  [14:51]
<kbsingh> or a CentOS-Virt
<gwd> ...
<lars_kurth> Are we going to resolve these today? We have 8 minutes? I wanted
             to briefly discuss a possible face-2-face as per last minutes
                                                                        [14:52]
*** lalatenduM (~lalatendu@122.167.168.235) has quit: Quit: Leaving
<kbsingh> lets do the repo convo on the lists
*** lalatenduM (~lalatendu@122.167.168.235) has joined channel #centos-devel
                                                                        [14:53]
<lars_kurth> OK. O made a note under open questions
<lars_kurth> I
<gwd> OK, but from a practical perspective: lsm5 can start building in
      virt7-testing when he's ready, and point people to those repos for
      testing?
<lars_kurth> The other open question is whether we should try to meet at
             LinuxCon
<lsm5> gwd: sure thing
<kbsingh> gwd: basically, we can workout what the scope is for the rpms, and
          then work through whatever is needed, eg: a single centos-virt repo
          can also contain sub repos and people can either get the sub repo or
          the main one etc - and we can even have deps between repos etc (
          much rope to hang with )
<kbsingh> gwd: yes  [14:54]
<kbsingh> lsm5: that would be good, i am going to be there for a few days
<kbsingh> erm lars_kurth that was for you - meeting would be good
*** kbsing: No such nick/channel
<lars_kurth> OK. Just to recap, who is there?  [14:55]
<lars_kurth> kbsingh, lars_kurth, lsm5
<kbsingh> I am there
<lars_kurth> gwd: did you make a decision?
* lsm5 is here  [14:56]
<lsm5> ahh wait
<lsm5> not at linuxcon :(
<lars_kurth> Ah, OK. How about I send a mail to the list and see how much
             traction there is
<kbsingh> ok  [14:57]
<kbsingh> the other thing is that since we are all UK ( the rest of us ) -
          maybe we can try and organise something locally
<lars_kurth> lsm5: are you based in the UK?
<kbsingh> lets thrash this out on the list as well  [14:58]
<kbsingh> although we'd need to decide soon'ish
<lsm5> lars_kurth: usa
<lars_kurth> What I thought
<lars_kurth> Let's trash this out at the list then
<kbsingh> yeah  [14:59]
*** mattgriffin (~textual@99-181-54-51.lightspeed.rcsntx.sbcglobal.net) has
    joined channel #centos-devel
<lars_kurth> Have we got anything else
<kbsingh> i think ovirt is going to be represented there, and the kvm forum
          might mean some of the qemu folks are around as well - and i suspect
          libvirt as well
<kbsingh> so we might still be able to run adhoc conversations - or atleast do
          some ice breaking with those groups and send them to the centos-virt
          list
<DV> lot of the libvirt team will be around KVM Forum next month, indeed
                                                                        [15:00]
<lars_kurth> I am tempted to do a dinner or beer thing, if there are more than
             a few people
<kbsingh> sounds good! I'm all for beer and dinner
<lars_kurth> Or have KB pull sdome people together adhioc
<lars_kurth> oops; bad spelling
<gwd> lsm5: Do you want to try to fill out some of the Docker bits of the Virt
      SIG roadmap?
<lars_kurth> lsm5: that would be great  [15:01]
<gwd> lars_kurth: Are you going to LinuxCon?
<lars_kurth> Yes. I am speaking
<lsm5> gwd: lars_kurth will do
<lsm5> on the wiki right?
<lars_kurth> But I was not going to stay for KVM Forum
<gwd> lsm5: Yep, see the link above.
<lsm5> ack
<lars_kurth> Yes,
             http://wiki.centos.org/SpecialInterestGroup/Virtualization/Roadmap
<vu> [13:58] < lars_kurth> lsm5: are you based in the UK?  [15:02]
* DV will pass next month, but should be back in Europe for FOSDEM
<lars_kurth> OK. In this case let's close (have not had lunch and am hungry?
             FOSDEM definitely sounds like something we should plan for
<lars_kurth> Feel free to update
             http://wiki.centos.org/SpecialInterestGroup/Virtualization/2014-September09-minutes
                                                                        [15:03]
<lars_kurth> And gwd: please sendf me the log or paste it into the minutes
<kbsingh> vu: hello
<vu> kbsingh: :-)  [15:04]
<kbsingh> lars_kurth: sounds good. thanks
<kbsingh> we have a few convo's to pickup on the list from here, lets do that
<lars_kurth> We do  [15:05]

2023-09-11 07:23