15:00:09 <danpb> #startmeeting libvirt
15:00:10 <openstack> Meeting started Tue Jul  1 15:00:09 2014 UTC and is due to finish in 60 minutes.  The chair is danpb. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:11 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:13 <openstack> The meeting name has been set to 'libvirt'
15:00:41 <ndipanov> o/ ohaio
15:00:45 <s1rp> o/
15:00:52 <nelsnelson> o/
15:02:02 <danpb> give one more min for people to arrive
15:02:29 <thomasem> o/
15:02:34 <thomasem> okay
15:03:17 <danpb> ok, that's long enough
15:03:26 <danpb> #topic LXC testing
15:03:43 <danpb> this seems to be nelsnelson + thomasem's  agenda item
15:03:52 <danpb> (from https://etherpad.openstack.org/p/nova-libvirt-meeting-agenda)
15:04:40 <thomasem> yup, so nelsnelson has started working on tempest tests for Libvirt LXC support and wanted to get some clarifications on how we can achieve promotion to upper classes of drivers via that testing, especially when some features don't seem to fit the container use case.
15:04:43 <nelsnelson> Greetings, danpb.  I recently have been tasked with working to fill out the set of gating tempest tests for the libvirt-lxc driver, and I have several questions.
15:04:54 <danpb> so on the question  "Must all of the boxes in the HypervisorSupportMatrix be supported"
15:05:07 <nelsnelson> Right.
15:05:08 <danpb> i don't think anyone in Nova has ever said that's our goal
15:05:16 <nelsnelson> By supported, I mean, "checked positively".
15:05:50 <danpb> the tier1 / 2 / 3 has only ever been defined as whether we have gating tests so far
15:05:59 <nelsnelson> Alternatively, must they simply be marked one way or another?
15:06:41 <danpb> i think this is probably something we need to raise on the dev mailing list
15:06:50 <danpb> because i doubt it has been thought about in any detail
15:07:04 <nelsnelson> Very good.  Will make a note, and do so.
15:07:10 <danpb> i expect there will be some minimum feature set people will ask for
15:07:37 <nelsnelson> In terms of implemented support?
15:07:47 <danpb> that matrix though is really crude and doesn't capture all possible permutations of supported features  in any sense
15:07:49 <nelsnelson> Or in terms of tested/verified services?
15:07:50 <danpb> yeah
15:07:58 <danpb> in terms of implemented support
15:08:04 <nelsnelson> Oh, I see.
15:08:34 <nelsnelson> One example of what you were just talking about seems to be the "resize" service for libvirt-lxc.
15:08:58 <nelsnelson> Which I imagine would only be implemented as a series of distinct operations.
15:09:20 <nelsnelson> Or perhaps a wrapping service, but that's an architectural decision.
15:10:15 <danpb> as far as moving from tier 3 to tier 2 i think there probably won't be any min supported feature set - just CI testing mandate
15:10:39 <danpb> i expect the issue of min feature set would only arise once considering tier 1
15:11:00 <nelsnelson> danpb, Do those mandates include both unit testing and functional testing?
15:11:11 <danpb> so personally I wouldn't get too fixated on min feature set yet - focus on just getting the testing system active
15:11:25 <danpb> nelsnelson: the HypevisorSupport wiki says
15:11:33 <danpb> *  unit tests that gate commits
15:11:33 <danpb> *  functional testing that gate commits
15:11:42 <danpb> (for Tier 1)
15:11:57 <danpb> Teir 2 is the same, but non-gating for functional tests
15:12:19 <nelsnelson> Understood.  I only ask because the tests which I was seeing for libvirt-lxc in the Tempest project currently, seem to be focused on API/service functionality.
15:13:20 <nelsnelson> They don't seem to be unit tests, per se.  Does that mean that those unit tests have not been written for the libvirt-lxc driver integration code?
15:14:09 <danpb> by unit tests i think it means the stuff in nova/tests/virt/libvirt
15:14:32 <danpb> and by functional tests I assume it means tempest
15:14:44 <nelsnelson> Okay, that helps to clarify.  Thank you.  :)
15:15:34 <thomasem> cool
15:16:19 <danpb> ok, so you had a few more Q's in agenda besides testing
15:16:37 <danpb> * What is the difference between "pause" and "suspend", and "unpause" and "resume"?
15:17:23 <nelsnelson> Right.  So that functionality seems to be somewhat virtual-machine specific operations.
15:17:31 <danpb> this is just a terminology thing - Nova uses  pause + unpause, while libvirt uses suspend + resume
15:17:36 <danpb> ie pause == suspend   and unpause == resume
15:17:47 <nelsnelson> So they are in fact synonymous.
15:17:52 <nelsnelson> Cool beans.
15:19:05 <danpb> * Is the iSCSI, iSCSI CHAP, and Fibre Channel features support outside of the scope of containers?
15:19:23 <danpb> can you elaborate on what you mean here ?
15:20:32 <nelsnelson> Well, is that functionality necessary/possible to support, or actually supported already, by libvirt-lxc at all?
15:21:07 <danpb> in what context - you talking about cinder volume drivers here ?
15:21:35 * danpb is trying to recall where iSCSI is involved in Nova
15:21:56 <nelsnelson> I'm frankly not familiar enough with those services to give much more detail about that context.  This particular question is actually relayed from another colleague here at Rackspace.
15:22:48 <nelsnelson> I was hoping to figure out what to do about features which don't seem to be compatible across virtual-machines/lxc contexts.
15:23:06 <nelsnelson> As far as the support matrix goes.
15:23:07 <danpb> ok, well assuming it is related to the volume driver AFAIK the iSCSI setup/usage is all done on the host
15:23:19 <danpb> so from the guest's POV it is just given a raw block device
15:23:31 <danpb> which should be possible to support with LXC too
15:23:50 <nelsnelson> Understood, noted.
15:24:19 <danpb> so its probably a matter of just testing out cinder volume usage with LXC and seeing what is working vs broken
15:24:34 <danpb> and then we'll figure out what todo from there
15:26:57 <danpb> nelsnelson: thomasem  any other Q's about LXC before we move on
15:27:10 <nelsnelson> I suppose the only other question I have...
15:27:16 <nelsnelson> which I'm not too clear about still
15:28:06 <nelsnelson> is just, what sort of documentation is there for the requirements involving the sufficiency of a collection of tempest functional service tests?
15:28:33 <nelsnelson> But I imagine that's a question the dev-list will be able to answer along with my other question regarding the support matrix substantiation.
15:28:53 <danpb> yeah, i think that's probably best for the mailing list
15:29:05 <nelsnelson> Okay, I'm good then, thank you very much, danpb.
15:29:05 <thomasem> No questions from me. =]
15:29:12 <thomasem> thanks!
15:29:56 <danpb> #topic Open discussion
15:30:11 <danpb> so anyone else with issues they want to raise today ...
15:30:54 <thomasem> I still haven't had time to scope out the network namespace stuff for ipv6, unfortunately
15:31:49 <thomasem> I did a little digging and it appeared to have a pretty extensive list. I'm not familiar enough to know what would be considered a comprehensive list for what's needed to be namespaced appropriately, regarding the /proc/sys mounted readonly
15:33:17 <thomasem> But, iirc, there was discussion to just remove that constraint anyway due to it only being a "don't hurt yourself" kind of aid, but doesn't really secure the situation.
15:33:23 <thomasem> Have you had any other thoughts on that, danpb?
15:33:44 <thomasem> atm, I'm just using a work around with post-up hooks in my lab
15:33:55 <danpb> yeah, i've not really got any other thoughts right now
15:34:28 <thomasem> Okay. I'll make a note to reply to the mailing list and see if we can get some resolution there.
15:34:39 <thomasem> Thanks
15:37:34 <danpb> so, doesn't sound like anyone else wants to raise any topics, so lets end it here
15:37:55 <thomasem> Okey doke. Thanks everyone!
15:38:07 <danpb> #endmeeting