09:32:49 <BobBall> #startmeeting XenAPI
09:32:50 <openstack> Meeting started Wed Sep 30 09:32:49 2015 UTC and is due to finish in 60 minutes.  The chair is BobBall. Information about MeetBot at http://wiki.debian.org/MeetBot.
09:32:51 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
09:32:52 <BobBall> Morning all!
09:32:54 <openstack> The meeting name has been set to 'xenapi'
09:32:58 <BobBall> johnthetubaguy: ping!
09:33:11 <huazhihao> bobball: Morning
09:33:13 <BobBall> I say morning, but, of course, for the Nanjing team it's afternoon
09:33:17 <huanxie> morning~
09:33:30 <BobBall> Reminder that the agenda is here https://wiki.openstack.org/wiki/Meetings/XenAPI
09:33:36 <jianghuaw> Good morning, guys.
09:33:44 <BobBall> If there is anything to add, then please just add it
09:33:57 <BobBall> Hmmm - one thing I noticed is that I need to update the meeting time
09:34:17 <BobBall> The agenda says that it's on Wednesdays at 15:00 UTC which is old now.
09:34:24 * johnthetubaguy nods
09:34:36 <BobBall> Anyway - no actions
09:34:38 <BobBall> Blueprints
09:34:40 <BobBall> #topic Blueprints
09:34:59 <BobBall> jianghuaw? Do we have a BP for Mitaka now?  Or is that still under internal review *don't remember the status*
09:35:36 <BobBall> What we do have is prototype code that adds vGPU to Nova
09:35:37 <jianghuaw> have not sent out yet for formal review.
09:35:47 <johnthetubaguy> so I have a ask about that...
09:35:50 <BobBall> #link https://review.openstack.org/223426 <-- prototype vGPU code
09:35:58 <jianghuaw> yes.
09:35:59 <johnthetubaguy> please don't internally review specs, please do that in gerrit
09:36:18 <BobBall> This is jianghuaw's first spec so getting some internal eyes on it is useful
09:36:30 <jianghuaw> sure, I did some change today and will send out for gerrit review soon.
09:36:35 <johnthetubaguy> but those internal eyes can be in gerrit, with a WIP on it
09:36:39 <BobBall> We weren't doing reviews of the approach - just of the format + detail contained
09:37:03 <jianghuaw> exactly.
09:37:25 <johnthetubaguy> so I would still ask that happens in gerrit please, it really helps
09:37:42 <BobBall> It will be going on to gerrit in the next few days I hope
09:37:52 <johnthetubaguy> honestly, lots of specs I reviewed last cycle has a round of updates about not meeting the template, thats totally cool
09:38:33 <jianghuaw> ok. I will send out review today.
09:38:38 <BobBall> Anyway - so there will be a spec we're proposing for Mitaka
09:38:48 <BobBall> Any chance you would be one of the core sponsors johnthetubaguy? :)
09:38:50 <jianghuaw> any comments are welcome and appreciated.
09:39:06 <johnthetubaguy> we don't really have core sponsors for blueprints
09:39:19 <johnthetubaguy> it didn't work, so we abandoned the idea really
09:39:19 <BobBall> Ah - my misunderstanding of the process then
09:39:31 <BobBall> Shows how many specs I was involved in for Liberty
09:39:48 <johnthetubaguy> I think we stopped that kilo
09:39:59 <johnthetubaguy> so there is a doc here with the details: https://wiki.openstack.org/wiki/Nova/Liberty_Release_Schedule#How_do_I_get_my_code_merged.3F
09:40:09 <johnthetubaguy> although I clearly need to move that soon!
09:40:12 <BobBall> Shows how many specs I was involved in for Kilo :)
09:40:37 <johnthetubaguy> the above doc talks about *why* we do things, which is really interesting
09:41:02 <BobBall> Thanks for the link
09:41:12 <johnthetubaguy> this one talks about internal reviews: https://wiki.openstack.org/wiki/Nova/Mentoring#Why_do_code_reviews_if_I_am_not_in_nova-core.3F
09:41:23 <johnthetubaguy> which made me realise its in the wrong place really
09:41:28 <BobBall> :)
09:41:31 <johnthetubaguy> but hey, ho, its all WIP really
09:41:44 <johnthetubaguy> ideas on structure and updates very very welcome!
09:41:48 <BobBall> Anyway - Do you know roughly when Mitaka specs open?  Is that when Liberty branches?
09:41:57 <BobBall> Or do the specs open at any time?
09:42:04 <johnthetubaguy> mitaka specs opened for review about one or two months ago
09:42:13 <johnthetubaguy> master is now mitaka
09:42:24 <BobBall> Oh blimey.  I'm behind then
09:42:29 <johnthetubaguy> we opened mitaka specs just after the midcycle
09:42:52 <johnthetubaguy> but honestly, only just starting to get core reviews on there now
09:42:58 <BobBall> OK - so jianghuaw - if you upload the spec, create the BP then we can hopefully have it discussed in a nova meeting as per the flowchart
09:43:25 <johnthetubaguy> its only spec less blueprints that need to go in the meeting
09:44:03 <BobBall> oh.  Lack of labels in the flow chart.  Clearly I didn't realise the difference between a solid and dotted line :)
09:44:17 <BobBall> I assume from the top solid is yes and dotted is no
09:44:18 <johnthetubaguy> yeah, I am lazy, the labels are only at the top
09:45:25 <BobBall> OK - well, we'd like to aim for M-1 so we'll be revising the spec quite hard when comments are received
09:45:47 <BobBall> Moving on - huanxie - update on Neutron status?
09:46:04 <johnthetubaguy> the slight issues with vGPU is that it touches flavor extra specs, but lets try not to worry about that right now
09:46:23 <johnthetubaguy> (or I assume it does)
09:46:25 <huanxie> hi, I'm working on RDO+XS+Neutron
09:46:34 <BobBall> We're very happy for the prototype code to be completely rewritten based on the approved design in the spec ;)
09:47:01 <huanxie> with single box installation, I can launch a new instance with correct IP at VLAN mode
09:47:28 <BobBall> Excellent - this is based on which OpenStack release?
09:48:17 <huanxie> it's about kilo, if I remember correct. And Guest VM is CentOS7.0
09:49:06 <BobBall> Ah great
09:49:15 <BobBall> We did find one thing that you might be interested in johnthetubaguy
09:49:22 <BobBall> We can't use vxlan with XenServer
09:49:28 <huanxie> When using all-in-one installation, we need to manually start another neutron-openvswitch-agent
09:50:01 <johnthetubaguy> why not?
09:50:09 <BobBall> vxlan requires an OVS tunnel between two endpoints.  Currently Neutron sets the tunnel up between compute <--> controller IP addresses, but because XenServer needs the OVS commands in dom0 the rule for the tunnel gets set up on the OVS in dom0, which does not have the compute's IP address
09:50:56 <johnthetubaguy> yeah, should be hypervisor IPs not compute IP
09:51:13 <BobBall> *nod* but I suspect that would be the start of the problems rather than the end
09:51:32 <johnthetubaguy> quark doesn't use neutron agents, as far as I understand
09:51:35 <BobBall> Other fun things include the OVS agent detecting IPv6 based on the Compute VM rather than dom0 :)
09:51:55 <johnthetubaguy> its an interestingly different approach (not saying I like it)
09:52:11 <BobBall> It is 'different'.  That's for sure.
09:52:25 <johnthetubaguy> what about running nova inside dom0 for the neutron support?
09:52:36 <BobBall> Possible only in Dundee
09:52:41 <BobBall> because of python versions
09:52:42 <johnthetubaguy> far from ideal, but maybe less effort
09:53:00 <BobBall> Well - vlan mode works for Neutron in our current architecture
09:53:18 <BobBall> We can have a look at proxying the retrieval of the IP address for the tunnel endpoint for vxlan or something like that
09:53:25 <johnthetubaguy> yep, just putting that out there
09:53:55 <BobBall> *nod*
09:54:07 <BobBall> Worth a look; but would have all sorts of different races etc
09:54:18 <BobBall> Just out of interest, would RAX rather run in dom0 or continue in a domu?
09:54:57 <johnthetubaguy> depends how much we could contain nova-compute I think
09:55:25 <BobBall> I see - so if it could be in a venv or similar then maybe.  Will think about that.
09:55:38 <johnthetubaguy> well we run it in a venv already
09:55:51 <johnthetubaguy> I am more thinking about cgroups and things, to restrict the process
09:55:52 <BobBall> OK - final update from huazhihao - Mirantis integration
09:55:58 <BobBall> *nod*
09:57:15 <huazhihao> Yes
09:57:42 <huazhihao> Last week we have delivered the fuel plugin
09:57:58 <huazhihao> With some docs and guides
09:58:32 <johnthetubaguy> its that specific to kilo fuel?
09:58:40 <huazhihao> Nope juno
09:58:43 <johnthetubaguy> or its it more independent of that
09:58:46 <johnthetubaguy> oh, OK
09:58:51 <BobBall> #link http://ca.downloads.xensource.com/OpenStack/Plugins
09:58:59 <huazhihao> Kilo fuel is just released. But we will catch up with this one
09:59:00 <BobBall> Fuel 6.1 which is Juno based
09:59:08 <johnthetubaguy> so juno is out of support soon, I think
09:59:10 <BobBall> Fuel 7.0 (Kilo) was just released two days ago
09:59:14 <johnthetubaguy> got it
09:59:38 <BobBall> But actually the plugin is almost entirely independent of OS version
09:59:49 <BobBall> OpenStack* version
09:59:54 <johnthetubaguy> I was hoping thats true :)
10:00:01 <huazhihao> Yes
10:00:03 <BobBall> Packages the Nova plugins, but that's all
10:00:05 <huazhihao> Mostly
10:00:21 <johnthetubaguy> first one is always the hardest, with any luck
10:00:32 <huazhihao> Yes, we have run into some problems about race conditions.
10:00:35 <BobBall> Thankfully huazhihao is now a Fuel expert
10:00:45 <huazhihao> Come on Bob
10:00:46 <BobBall> Race condition is likely to be in the Fuel tests
10:00:59 <BobBall> rather than the XenAPI code
10:01:12 <BobBall> We fixed a similar race in Tempest a year or so back
10:01:30 <BobBall> (basically around waiting for images vs serves to be 'ready' after a snapshot)
10:02:07 <BobBall> OK
10:02:10 <BobBall> Drat - we're out of time
10:02:12 <johnthetubaguy> hmm, interesting
10:02:23 <BobBall> Anything else to raise today's meeting?
10:02:39 <huazhihao> Ah that is all for me.
10:03:14 <BobBall> In which case, thanks all for the meeting
10:03:30 <BobBall> Meet again same time, same place, 2 weeks (14th October)
10:03:32 <BobBall> #endmeeting