[CentOS-virt] Virt SIG Meeting minutes for 16 Dec 2014
dunlapg at umich.edu
Wed Dec 17 14:47:23 UTC 2014
Attending: imcleod, lars_kurth, gwd, kbsingh, dcaro, jonludlam
Meetings over Christmas:
- Skip the one on 30 Dec; do IRC meeting Jan 13th.
- 4.4.1, libvirt pushed
- Notes: xend disabled by default
- Plan: Sign, release soon,
- RHEL 7.1 will already rebuild qemu with snapshotting on
- But we may want to do the qemu rebuild for 6 as well.
- kbsingh: I think oVirt is fine in the sig
- sandro to be maintainer, dcaro to also have access / co-maintainer
- Probably a hackathon / dojo thing the Friday before
- Further discussion to happen on centos-promo
Docker update (from Lokesh)
- 1.4.0 into c7-testing
- working on rocket
-------------- next part --------------
<imcleod> So, is the virt SIG meeting here? [14:01]
<gwd> Virt SIG meeting!
<gwd> imcleod: Yes.
<lars_kurth> Hi all [14:02]
<gwd> I think David Caro said he was coming.
<gwd> kbsingh: You around?
<lars_kurth> Who is here for the virt SIG meeting? I see gwd, imcleod,
<kbsingh> gwd here
<kbsingh> was david going to make it grom ovirt?
<lars_kurth> Alright, what do we have to discuss? (I have to drop off early)
<kbsingh> lsm5 too [14:04]
<gwd> On my agenda:
<gwd> - oVirt inclusion in SIG
<gwd> - Update on Xen
<gwd> - Meetings over Christmas
<gwd> - Update on CBS (particularly a way for a SIG get signed packages)
<gwd> I assume there are updates on docker too if lsm5 is around.
<lars_kurth> Should we do the things we need kbsingh for first? On meetings I
would suggest to re-start in Jan [14:06]
<gwd> While kb is here we should probably touch base about an event around
<lars_kurth> That too
*** dcaro (~dcaro at 130.Red-79-145-201.dynamicIP.rima-tde.net) has joined
<gwd> meetings> Yes, I think we can just skip the next meeting, and meet again
on IRC in 4 weeks' time
<dcaro> hi everyone, sorry I'm late, ^agree
<kbsingh> +1 to skip next meeting
<gwd> That way nobody has to mess with their calendars except to remove the
one meeting. :-)
<lars_kurth> +1, so this means we re-start on Jan 18th [14:07]
<gwd> You mean Jan 13th?
<kbsingh> 13th on my cal as well [14:08]
<gwd> OK, that sounds like we're all set then. [14:09]
<gwd> Just a quick update on Xen:
<gwd> After getting a bit of automated testing up and finding some bugs, I
pulled some stuff into 4.4.1, and got libvirt building. [14:10]
<gwd> I did some smoke testing of libvirt (virt-install), which seems to work
fine for libxl.
<gwd> Haven't done testing of libvirt with xend yet.
<lars_kurth> kbsingh: yes, 13th
*** agrimm (agrimm at nat/redhat/x-ohvbnjookashdser) has joined channel
<gwd> Also -- I added a patch to disable xend by default.
<gwd> My hope was that it would prompt people using xend to start trying to
switch to xl / libxl, while they still have the fallback available.
<gwd> xend has been entirely removed in 4.5, so once we update past then
they'll be stuck without a fallback option. [14:12]
<kbsingh> how would that impact people with existing installs though
<kbsingh> ( i mean in the 4.4.1 build, people who get this from the existing
4.2 installs )
*** DV (~veillard at 2001:41d0:a:f29f::1) has quit: Ping timeout: 272 seconds
<lars_kurth> gwd: agreed. Also given that we agreed to use even Xen numbers
for CentOS, so any issues would need to be raised and fixed in
the 4.6 release cycle which is startting in Jan
*** jonludlam (jonludlam at nat/citrix/x-vnynbphzofovbwpi) has joined channel
<gwd> Yes -- I'm not sure; in at least one case it seemed like doing an update
didn't disable xend. [14:13]
<gwd> But I haven't tested it thoroughly, that may have been a quirk of
something I was doing.
<kbsingh> ok [14:14]
<gwd> I'm not sure what else we can do. In 4.4.1 the xm command already
prints a warning message every time you use it. :-)
<gwd> I think I might also change the error message xm prints if xend is not
running -- to 1) suggest switching to xl, and 2) to tell them how to
re-enable xend if they need it. [14:15]
*** dominic_ (dgeevarg at nat/redhat/x-dxboixqkmynhyiae) has quit: Quit: Leaving
<kbsingh> that should be more than enough [14:16]
<kbsingh> so, with yur 4.4.1 builds, these are all in the virt6-testing repo
on cbs.centos.org/repos ?
<kbsingh> looks like yesterday was a busy day :) [14:17]
<kbsingh> given that package set - should we sign and put into a testing repo
then do some announcements around it ?
*** DV (~veillard at 2001:41d0:a:f29f::1) has joined channel #centos-devel
<kbsingh> let people play with the package set and report issues over the
holiday season, with the aim of push to stable-release into
xen4centos repos early Jan ?
<gwd> Yes, I think that would be helpful; with one caveat: this is my last
work day this year, so I won't be able to answer any bugs until January.
<gwd> If you don't think 3 weeks of radio silence will be a problem, then
yeah, we can do that.
<kbsingh> ok, i guess we can set expectations around that - are there any
other notes ? [14:19]
*** mcclurmc (~mcclurmc at cust2438-mfw1.atlanticmetro.net) has joined channel
<gwd> kbsingh: But I thought you wanted to sign it with the SIG signing key?
Is that already available (to you guys anyway)?
*** Bahhumbug_ (jrd at gerdesas.gerdesas.com) has joined channel #centos-devel
*** Bahhumbug_ (jrd at gerdesas.gerdesas.com) has quit: Changing host
*** Bahhumbug_ (jrd at serentos/admin/jrd) has joined channel #centos-devel
<gwd> mcclurmc: Hey
<mcclurmc> hey gwd, what's up?
*** chorrell_ (~chorrell at unaffiliated/chorrell) has joined channel
<gwd> You here for the Virt SIG meeting?
<lars_kurth> kbsingh: yes announcements would be good, We can also do the xen
blog and once we are ready a press release (assuming it is less
painful than last time) [14:21]
*** telex (teletype at freeshell.de) has quit: Remote host closed the connection
<lars_kurth> gwd: just let people know in the mail that you won't be able to
get back to them until Jan [14:22]
<gwd> let people know> check.
*** kkeithley1 (~kaleb at c-73-16-206-47.hsd1.ma.comcast.net) has joined channel
*** telex (teletype at freeshell.de) has joined channel #centos-devel
*** crossbuilder (~crossbuil at p4FC14B09.dip0.t-ipconnect.de) has joined channel
<mikem> kbsingh, if you know the dep order, you can use koji chain-build
<gwd> OK, so shall we move on to oVirt? [14:23]
<lars_kurth> +1 [14:24]
<gwd> I wasn't what the procedure should be here.
<gwd> I think it makes total sense to have the qemu rebuild in the Virt SIG.
*** saltsa (joonas at dsl-hkibrasgw1-58c01a-36.dhcp.inet.fi) has joined channel
*** der_mode1 (~der_moder at 81-89-98-215.blue.kundencontroller.de) has joined
<dcaro> +1 (I'm from oVirt, os not sure it counts xd) [14:25]
<gwd> I don't have any objection in principle to having oVirt RPMS or the
different images proposed in the SIG as well, but I didn't know if there
might be a better place for them.
<lars_kurth> kbsingh: we wanted to consult with you, as you had raised
concerns re whether the VIRT SIG is the best place for oVirt. I
don't have any objections, but don't have the big picture on what
is going on in other SIGs
<gwd> kbsingh: Did you get a chance to read the oVirt proposal? [14:26]
<lars_kurth> seems we lost kbsingh [14:27]
*** kbsingh_ (~kbsingh at n12.lon1.karan.org) has joined channel #centos-devel
*** ChanServ (ChanServ at services.) has changed mode for #centos-devel to +o
<kbsingh_> thanks freenode [14:28]
<gwd> Indeed we did lose kbsingh_
*** talori (~timothy.l at n219077078209.netvigator.com) has joined channel
<gwd> dcaro: FYI a lot of the infrastructure for the SIGs isn't quite in yet;
so even if we said "yes you're in", it may be some time before you can
actually build something... [14:29]
* kbsingh_ back - freenode decided to dump my connection
<lars_kurth> OK, let me re-post
<kbsingh_> did we close off the xen 4.4.1 for testing
<dcaro> gwd: what is needed there? We might be able to help (I'm from the
infra team on oVirt)
*** kbsingh_ (~kbsingh at n12.lon1.karan.org) is now known as kbsingh [14:30]
*** rtnpro (shashisp at gateway/shell/waartaa/session) has joined channel
<lars_kurth> kbsingh_: I think we did "‹mikem› [14:23:05]kbsingh, if you know
the dep order, you can use koji chain-build "
*** Bahhumbug_ (jrd at serentos/admin/jrd) is now known as Bahhumbug [14:31]
<kbsingh> ok, i will check the meeting log once were done here. and then go do
whatever is needed w.r.t sign and release
<gwd> kbsingh: Yes, I think the plan is to go ahead with signing and
announcing "betas", as long as I make it clear I'll be away until
<gwd> kbsingh: So we moved on to oVirt:
<gwd> <gwd> I don't have any objection in principle to having oVirt RPMS or
<gwd> different images proposed in the SIG as well, but I didn't know if
<gwd> might be a better place for them.
<gwd> <lars_kurth> kbsingh: we wanted to consult with you, as you had raised
<gwd> concerns re whether the VIRT SIG is the best place for oVirt. I
<gwd> don't have any objections, but don't have the big picture on what
<gwd> is going on in other SIGs
<kbsingh> so, w.r.t Ovirt - it looked that rather than ovirt itself, the
desire is to just have a qemu with the one option turned on ?
<gwd> kbsingh: The proposal sent to the virt list had 4 components.
<gwd> One was a qemu re-build; I think that makes sense to have in the Virt
*** srirama (~srirama at 220.127.116.11) has quit: Ping timeout: 265 seconds
<kbsingh> the qeu story took a bit of a sudden turn, with the rhel7.1
announcement stating that future rhel qemu builds will have these
features turned on anyway ( for ceph ) [14:33]
<gwd> Then there was "oVirt Live", a live CD with oVirt
<kbsingh> would that then remove the need to have a qemu here as well ?
<gwd> And some "oVirt Node", a minimal image for running oVirt, presumably.
*** mattgriffin (~textual at 99-181-54-51.lightspeed.rcsntx.sbcglobal.net) has
joined channel #centos-devel
<gwd> And finally oVirt itself.
<jonludlam> what's the story for CentOS 6 for these SIGs?
<kbsingh> gwd: ok, i guess if ovirt itself was going to come in, then livecd /
node etc can follow eventually.
<kbsingh> lars_kurth: i think ovirt is ok in the virt sig
<kbsingh> jonludlam: sig's can target 5/6/7 as they want, its left upto the
SIG to workout what story it wants to deliver for what release
<lars_kurth> kbsingh: so it seems that's a +1 then assuming that there is
component leadership from oVirt to make this happen [14:36]
<jonludlam> so is it still desirable to repackage qemu for 6?
<lars_kurth> what I meant was someone maintaining those repos in the SIG
<kbsingh> lars_kurth: yeah
<dcaro> sandro and I will take care of that (mainly sandro)
<gwd> jonludlam: I think at this point it makes sense for each "maintainer" to
set their own level of commitment. [14:37]
<kbsingh> jonludlam: my understanding is that the same qemu spec will make it
into 6 as well
<kbsingh> jonludlam: but we can check
<lars_kurth> jonludlam: If we include oVirt, then a different version of qemu
for 6 is a minor detail
<gwd> So Xen is going to try to maintain c6 for some time; hopefully with the
same basic package structure as c7.
<gwd> Docker ATM is c7-only.
<gwd> I'd leave the qemu rebuild / oVirt stuff up to the oVirt maintainer.
<kbsingh> ok ( re: leave to ovirt maintainer)
<dcaro> lars_kurth: kbsingh so what is needde on our side to get the packages
<lars_kurth> dcaro, gwd: so I suppose we should make an announcement on the
<gwd> I think to begin with we need to designate one person as the oVirt
maintainer for the sig [14:40]
<lars_kurth> dcaro: there some infrastructure access which needs to be sorted
such as wiki, etc. - kbsingh needs to do this (there is no self
<lars_kurth> gwd: agreed
<gwd> Then that person (at least) needs access to the CBS and permission to
own the relevant packages.
<dcaro> gwd: can it be more than one person?
<kbsingh> yeah, there is a procss for the acces, via bugs.centos.org - but
that will be once you nominate a person to own the packages / do the
buids, and it will need either gwd or lars_kurth to ack the request
<kbsingh> dcaro: you can make it as many people as you like, but its always
good to channel via a small number of people [14:41]
<dcaro> kbsingh: okok, I said because it's sandro and me for now hte ones that
will be pushing this
<lars_kurth> I think notionally it is best to have one person as maintainer,
access to systems can be several [14:42]
<dcaro> (and sandro is on pto right now)
<dcaro> ok, that should be done through the list?
*** DV (~veillard at 2001:41d0:a:f29f::1) has quit: Ping timeout: 265 seconds
<dcaro> (if not I propose sandro as maintainer and me as 'access'
<gwd> Yes, I'll send mail this afternoon with a proposal; if nobody objects in
a week we can call it a deal. [14:43]
<kbsingh> lars_kurth: gwd: it might be good to then also edit the wiki page
for the SIG and add those people details there
<dcaro> gwd: okok good :)
<gwd> kbsingh: Yes.
<lars_kurth> dcaro: can you raise a ticket on bugs. for sandro and yourself
and CC me (lars.kurth.xen at gmail.com) and gwd onto it and we can
then pick this up in the new year when sandro is back from pto
<dcaro> lars_kurth: sure
<kbsingh> http://wiki.centos.org/HowTos/CommunityBuildSystem has some details
on the bootstrap process as well
<lars_kurth> dcaro: drop me a private mail with names & ITRC handles and I
will add to the page [14:44]
<kbsingh> well, it has all the details for the bootstrap, anything missing
should be added in there
<dcaro> lars_kurth: ok
<lars_kurth> I have to drop off soon. [14:45]
<lars_kurth> I have 5 more mins
*** Zequal (~polaroid at dsl-207-112-88-18.static.tor.primus.ca) has joined
<gwd> kbsingh: Is anything going on around FOSDEM?
<lars_kurth> Can we cover FOSDEM
<gwd> kbsingh: I personally probably won't be able to make it, but Roger
should be able to come.
<kbsingh> ttp://wiki.centos.org/Events/Dojo/Brussels2015 [14:46]
<Evolution> we have a dojo pre-fosdem... yeah thzt
<Evolution> er, that even.
<kbsingh> that is the day before Fosdem, and then during Fosdem we will have
the booth and talks etc
<Zequal> When doing an installation of CentOS to, for example, a HDD.. If you
were to plug that into a different machine, it wouldn't recognize the
hardware on boot. However, the LiveCD seems to handle this so much
better than an actual install. What is the LiveCD doing differently?
<kbsingh> for the Dojo - am trying to make it mostly demo/tutorial/hack
session sort of content, we have 2 rooms, 45 people capacity per
*** DV (~veillard at 2001:41d0:a:f29f::1) has joined channel #centos-devel
<kbsingh> was thinking of making one track be cloudy / virt specific - the
other track be more traditional domains ( like install / patch
management / lifecycle / config management etc ) [14:48]
<gwd> kbsingh: OK, keep us updated and we'll see what we can get going there.
<kbsingh> the conversation should all be on centos-promo from here on, it was
private as I was trying to confirm venue [14:49]
<lars_kurth> kbsingh: please keep us in the loop & FOSDEM is soon. After XMas
there are only 2-3 weeks to turn this round [14:50]
<gwd> Last thing is -- jonludlam, have you managed to get access to the CBS
<gwd> You've submitted a bug report as per the wiki page several weeks ago,
<jonludlam> that's right
<kbsingh> lars_kurth: so, watch the centos-promo list
<kbsingh> jonludlam: do you have a bugs number ?
<lars_kurth> kbsingh: nope [14:52]
<jonludlam> just looking it up [14:53]
<lars_kurth> Just signed up now
<lars_kurth> have to go [14:54]
<jonludlam> I've been using copr for the prototype of the OCaml SCL
<kbsingh> jonludlam: so this is for the scl SIG right ? [14:55]
<Evolution> kbsingh: yes, but would be required to build the xen bits. [14:56]
<Evolution> well, not required but useful
<kbsingh> in that case, hhorak will need to ack the request first. then
alphacc can action it. I suspect alphacc is waiting on a SIG owner
to ack the request
<jonludlam> Originally I was only interested in the virt sig, but since the
suggestion was to make an SCL for the OCaml stuff I'm now
insterested in both
<gwd> jonludlam: Are you looking to get xapi into a SIG at some point as well?
<kbsingh> Evolution: plot thickens... so this is a scl via the scl-sig needed
for the virt-sig
<jonludlam> quite possibly
<Evolution> kbsingh: not needed. but could be useful to. [14:57]
<kbsingh> if xapi is going in, its going to be needed
<Evolution> as I understand it right now the ocaml packages used *replace* the
<gwd> kbsingh: AFAIK the core xen packages only need ocaml for oxenstored
<kbsingh> gwd: and at the moment were using the cxenstored ?
<kbsingh> i guess the real question is - are we accepting SCL's into sig's
other than the scl-sig [14:58]
* gwd doesn't know
<jonludlam> cxenstored is not default in xen these days is it?
<gwd> jonludlam: It will use cxenstored if the ocaml dev libraries are not
<jonludlam> but will default to oxenstored if they are
<gwd> kbsingh: From what I understand, the resulting oxenstored binary is
statically linked, so there's no dependency for the runtime; there might
be one for build time if we wanted updated ocaml libs.
<kbsingh> ok [15:00]
<jonludlam> same for xapi
<gwd> jonludlam: default to oxenstored> That's my understanding last time I
<kbsingh> MerlinTHP: alphacc: mikem: can we consume scl's in mock roots in
koji for this kinda stuff ?
* gwd goes to check...
<jonludlam> xapi wont need the SCL at runtime
<kbsingh> if so, then this is a clear scl-sig thing, and we can handle it +
work it there
<MerlinTHP> We can build SCLs in koji, and add SCL packages to buildroots,
<mikem> kbsingh, I assume so.
<MerlinTHP> I've not read much scrollback
<kbsingh> okay. I guess its just a case of that build tag having a reference
to the scl repo [15:02]
<gwd> BTW, we definitely are building w/ ocaml at the moment...
<kbsingh> jonludlam: so you need hhorak to ack the req, and can get the access
in via the scl sig.
<kbsingh> gwd: i suspect its the distro ocaml though
* hhorak reading the backlog
<jonludlam> aha [15:03]
<kbsingh> jonludlam: untested, since we've not had this in the past - but
access via sig usually maps to targets you can build against.
*** DV (~veillard at 2001:41d0:a:f29f::1) has quit: Ping timeout: 244 seconds
<kbsingh> so you can just also ask lars_kurth or gwd to ack the request, and
alphacc can get you setup to build into the virt targets as well
<jonludlam> kbsingh, fair enough.
<gwd> Hmm -- I seem to see /usr/sbin/xenstored but no /usr/sbin/oxenstored...
<alphacc> yes I will catch up with the pending ticket today (irl meeting feel
free to assign action to me)
<jonludlam> I think the spec might rename it [15:05]
<kbsingh> alphacc: wildo
<hhorak> I'm a bit desoriented between abbrevs I don't know but if the ack
should be for including ocaml402 into the scl sig, then I'm fine with
<jonludlam> hhorak, good to know :-) I think that's the plan [15:07]
<gwd> Oh weird -- it builds it but then puts it in xen-ocaml, which won't get
installed unless you ask for it.
<jonludlam> oh nice
<gwd> That should get sorted...
<jcpunk> alphacc, MerlinTHP, kbsingh, Arrfab : I'd love any assistance getting
the chainbuild srpm patch into koji. I've a simple python script for
ordering a pile of srpms I can also share if you want [15:09]
<Evolution> hhorak || jonludlam if you guys are fine with that name, I'll
create the repo for it and add to the mix.
<jonludlam> jcpunk, we've got a similar script that dumps stuff into a
makefile to get make to do it [15:10]
<Evolution> will create an ocaml and ocaml402
<jonludlam> Evolution, the name is good for me
*** DV (~veillard at 2001:41d0:a:f29f::1) has joined channel #centos-devel
<jcpunk> jonludlam: awesome :) great minds and thinking
*** puzzled (~patrick at puzzled.xs4all.nl) has quit: Quit: Leaving [15:13]
<jonludlam> jcpunk, github.com/xenserver/planex
<jonludlam> readme is out of date - must fix that! [15:14]
<kbsingh> ok, so howse the virt-sig meeting going ten [15:15]
<kbsingh> lsm5: anything from you ?
<kbsingh> gwd: are we good to close off then ?
<gwd> kbsingh: Yes, I think we're done. [15:16]
<gwd> Thanks all
More information about the CentOS-virt