15:02:43 <danpb> #startmeeting libvirt
15:02:44 <openstack> Meeting started Tue Sep  9 15:02:43 2014 UTC and is due to finish in 60 minutes.  The chair is danpb. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:45 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:02:48 <openstack> The meeting name has been set to 'libvirt'
15:03:54 <danpb> hi, anyone here for a libvirt meeting ?
15:03:56 <thomasem> o/
15:03:59 <nelsnelson> o/
15:04:00 <sew> o/
15:04:12 <mtesauro> o/
15:04:22 <thomasem> us again. :P
15:04:40 <apmelton> o/
15:05:07 <danpb> add any agenda items to https://etherpad.openstack.org/p/nova-libvirt-meeting-agenda
15:06:36 <thomasem> I think the only issue we've been itching about is this shutdown issue, but that sounds to me now like a distribution issue with Wheezy.
15:06:53 <thomasem> The only other concern is that Libvirt seems to assume /dev/initctl
15:06:55 <danpb> #topic LXC shutdown issue
15:07:08 <thomasem> with the shutdown command
15:07:31 <thomasem> but, I think rcleere or sew could speak better to that than I
15:07:35 <danpb> #topic LXC shutdown issue
15:07:36 <thomasem> since they've done the bulk of that work
15:07:52 <danpb> hmm, seems the bot isn't playing
15:07:56 <thomasem> =[
15:08:14 <rcleere> like you said thomasem, I think its just a broad assumption made by libvirt that initctl will be in /dev.
15:08:21 <rcleere> and debian wheezy puts it in /run
15:08:28 <danpb> so IIUC, the first problem is that the setns() syscall isn't available
15:08:43 <rcleere> Oh yeah.. that too
15:08:51 <danpb> the patch referenced only makes sense if glibc is missing it, but the nuderlying kernel has it
15:08:52 <rcleere> THere is a patch to stub that out
15:09:01 <danpb> otherwise you're just moving the error slightly later
15:09:06 <danpb> to a syscall not existing
15:10:00 <sew> we used the patch here to get ns working on wheezy: http://www.redhat.com/archives/libvir-list/2014-March/msg00082.html
15:10:22 <thomasem> And we had to set up a /dev/initctl link in our images to get the shutdown command to work.
15:10:23 <danpb> sew: so you confirmed that does in fact make it work fully ?
15:10:40 <thomasem> With that patch and no /dev/initctl, it just noops, essentially
15:10:51 <thomasem> iirc
15:10:51 <sew> yes, our testing with libvirt 1.2.3 and that patch makes shutdown work cleanly for us
15:11:01 <sew> from a libvirt perspective
15:11:12 <sew> but then /dev/initctl comes into play
15:11:16 <danpb> sew: ok, can you reply to that msg on the upstream libvirt list
15:11:27 <danpb> or just point it out as something to be merged upstream
15:11:28 <sew> yep, can do
15:12:06 <danpb> for  /dev/initctl, I see no harm in trying /dev/initctl and then falling back to try /run/initctl
15:12:53 <danpb> i'm surprised that /dev/initctl isn't simply a symlink to /run/initctl on those distros using /run, but that's besides the point i guess
15:13:18 <danpb> so if someone wants to write a patch to fallback to /run/initctl that'd be nice
15:14:17 <sew> danpb: we can take a look at writing that patch
15:15:12 <danpb> #topic LXC testing
15:15:31 <danpb> it has been a while since we talked about testing, so i wondered if there was any update on CI tests for LXC ?
15:16:16 <thomasem> nelsnelson has been working on the tempest tests for LXC
15:16:36 <thomasem> We've been having issues with devstack leaking NBD devices and causing tempest tests to fail
15:16:48 <thomasem> Not sure what other issues have been run into
15:17:10 <apmelton> thomasem: pretty sure I've identified that leak
15:17:14 <thomasem> sweet
15:17:19 <apmelton> thomasem: they leak when terminating a shutdown container
15:17:24 <apmelton> the cleanup code never gets called
15:17:25 <thomasem> interesting
15:17:32 <apmelton> we just need to move it outside an iff
15:17:43 <apmelton> I'm confirming it today, will open a bug and get a fix up for review soon
15:17:52 <thomasem> =]
15:18:32 <danpb> great, so its basically work in progress still
15:19:08 <apmelton> danpb: pretty much, I think we're almost at the point to where most of the tests should pass
15:19:20 <apmelton> nelsnelson: which tests were you skipping?
15:19:27 <nelsnelson> all rescue
15:19:29 <nelsnelson> all suspend
15:19:31 <nelsnelson> for now
15:19:40 <nelsnelson> also skipping xml_api2 tests
15:19:46 <thomasem> I know the infra guys wanted to see our skips
15:19:52 <thomasem> with a reasoning as to why
15:19:54 <thomasem> Just as a heads up
15:20:08 <nelsnelson> we're only skipping them for now to isolate issues
15:20:12 <thomasem> gotcha
15:20:45 <nelsnelson> Actually, I think s1rp Rick Harris indicated that rescue and suspend might not be supported yet?
15:20:56 <danpb> right, suspend makes sense to skip since lxc can't spport that
15:21:14 <danpb> rescue ought to be doable if it doesn't work yet
15:21:26 <danpb> since that's merely booting with a slightly different disk config
15:21:29 <s1rp> yeah, should be possible just hasn't been done
15:26:27 <thomasem> added notes
15:39:51 <thomasem> Anything else we want to chat about, or shall we call it?
15:40:01 <danpb> seems, like we're done
15:40:08 <thomasem> cool, thanks everyone!!
15:40:13 <danpb> see ya all next week
15:40:19 <apmelton> o/
15:40:20 <danpb> #endmeeting