15:10:33 <tmorin> #startmeeting bgpvpn
15:10:34 <openstack> Meeting started Tue Dec  6 15:10:33 2016 UTC and is due to finish in 60 minutes.  The chair is tmorin. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:10:36 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:10:38 <tmorin> hi everyone
15:10:39 <openstack> The meeting name has been set to 'bgpvpn'
15:11:18 <tmorin> hi bfernando, bobmel, doude, eon`, lewo, matrohon, pcarver
15:11:25 <pcarver> hi
15:11:39 <bfernando> hello
15:12:43 <tmorin> let's see who may attend today...
15:13:09 <lewo> hi
15:13:22 <tmorin> hi :)
15:14:48 <matrohon> hi
15:15:41 <tmorin> let's see what we can discuss today
15:15:46 <tmorin> what I have in mind:
15:15:55 <tmorin> * status on tempest testing
15:16:26 <tmorin> * stadium things
15:16:46 <tmorin> * driver status
15:16:50 <tmorin> anything else ?
15:17:49 <tmorin> ok let's start
15:17:53 <tmorin> #topic tempest testing
15:18:10 <tmorin> the new tempest scenario base test has just merged
15:18:18 <tmorin> very nice work by bobmel
15:18:22 <tmorin> many thanks to him
15:18:42 <tmorin> it's really a good thing to have scenario coverage in our check/gate pipeline
15:18:47 <bobmel> Thanks! And hi all!
15:18:51 <bobmel> Sorry for being late
15:18:56 <tmorin> hi bobmel :)
15:18:58 <tmorin> np
15:19:30 <tmorin> what the test does is setup two VMs in two networks associated to the same BGPVPN, and confirm that one can ping the other
15:19:56 <tmorin> this exercises a large part of the neutron-agent-bagpipe-ovs stack
15:20:25 <tmorin> exercising more would involved setting up a multinode test, or doing things the fullstack way
15:20:44 <tmorin> so we now have a very good starting point to enrich the scenario test
15:21:19 <tmorin> it should be pretty easy to add extra tests (eg. Router associations,   doing the associatio before or after the VMs are up, etc.)
15:21:37 <tmorin> comments ?
15:21:40 <bobmel> the things you pointed out seems like good next steps to gradually move towards test-wise
15:22:18 <tmorin> yes, adding more tests will be welcome
15:22:45 <tmorin> and we can think about how to do a multinode test ; that will involve deploying a BGP route reflector in a devstack node
15:22:55 <tmorin> nothing that can't be done, but not trivial either
15:24:40 <tmorin> #topic stadium things:  API ref and APi definition + moving OSC client
15:24:52 <tmorin> we have a few things to do
15:25:00 <matrohon> can't we use "full mesh" with two bagpipe bgp speakers?
15:25:19 <matrohon> to avoid deploying a RR
15:25:46 <tmorin> #undo
15:25:47 <openstack> Removing item from minutes: <ircmeeting.items.Topic object at 0x7f63d9725090>
15:25:48 <tmorin> matrohon: we could have bagpipe with two nodes with the fake/lightweight/limited RR bundled with bagpipe
15:25:57 <tmorin> but not more than two
15:26:17 <matrohon> well, that would be nice enough to me :)
15:26:19 <bobmel> is that a limitation of bagpipe?
15:26:20 <tmorin> bagpipe's BGP implementation is "connect-only", can't listen on BGP socket at this point
15:26:26 <tmorin> bobmel: yes
15:26:29 <bobmel> Ok
15:27:34 <tmorin> well, neutron has scripts to compile/start a fresh OVS from git,  so running quagga or gobgp is certainly in the range of feasible things
15:28:08 <tmorin> but in the meantime, we could have a two node multinode test, one controller+compute node and one compute-only node
15:29:09 <tmorin> next topic ?
15:29:16 <matrohon> yep thanks
15:29:24 <tmorin> #topic stadium things:  API ref and APi definition + moving OSC client
15:29:37 <tmorin> pcarver, you have been pursuing the API work
15:29:47 <tmorin> https://review.openstack.org/#/c/390196/
15:29:58 <tmorin> anything you want to comment on pcarver ?
15:30:03 <tmorin> I'd say the work is mostly done
15:30:29 <pcarver> I will push another update that adds get_sub_attribute_map but I also posted a comment to Armando's "example" review to see what he says.
15:30:54 <pcarver> https://review.openstack.org/#/c/407308/2/neutron_lib/api/definitions/_dummy.py@62
15:30:56 <tmorin> link ?
15:30:58 <tmorin> thanks
15:31:27 <pcarver> If he decides to give an example that's different, I'll look at what he proposes
15:32:01 <tmorin> ok
15:32:21 <tmorin> in your API change, I think there are still places with "uuid" instead of the base "string" JSON type
15:32:37 <pcarver> I thought I got them all, but I'll look again
15:33:20 <tmorin> I just commented inline
15:34:00 <tmorin> I think you can add the subresource definition and accessor, and remove the action map, without waiting for armando's comment on the dummy/example definition
15:34:10 <pcarver> ok
15:35:11 <tmorin> then there is the move of OSC client from networking-bgpvpn to python-neutronclient
15:35:17 <tmorin> doude, any update on this ?
15:36:15 <tmorin> ah, maybe doude isn't here today
15:37:32 <tmorin> change https://review.openstack.org/#/c/407271/ in under review and describes how neutron stadium project should push osc client extensions in python-neutronclient
15:38:25 <tmorin> ok, something to follow up with doude next meeting ...
15:38:30 <tmorin> next topic ?...
15:38:34 <tmorin> #topic driver status
15:39:24 <tmorin> a brief update on bagpipe driver
15:40:00 <tmorin> our CI jobs and tempest stuff is now in good order to test bgpvpn with bagpipe driver in a much better way
15:40:48 <tmorin> last week we run a few tests more in-depth than usual that lead us to find two bugs in bagpipe driver and one in OVS MPLS implementation
15:41:07 <tmorin> these are good example of test we can add to the test suite
15:42:15 <tmorin> https://bugs.launchpad.net/bgpvpn/+bug/1646526  => will require work on neutron/agent extension framework side, but not easy to test
15:42:15 <openstack> Launchpad bug 1646526 in networking-bgpvpn "bgpvpn functionality lost on openvswitch restart" [Medium,New]
15:42:50 <tmorin> https://bugs.launchpad.net/networking-bagpipe/+bug/1644151 => this one is typically something that a scenario test can cover
15:42:50 <openstack> Launchpad bug 1644151 in BaGPipe "removal of a VM disconnects other VMs on same Network" [Undecided,Fix released]
15:44:04 <bobmel> For the current vendor drivers, anyone done anything on 3rd party CI?
15:44:22 <tmorin> link to the discussion on the ovs bug: https://mail.openvswitch.org/pipermail/ovs-dev/2016-December/325567.html
15:44:46 <tmorin> bobmel: we have briefly discussed this in Barcelona, but without coming to a conclusion
15:44:55 <bobmel> ok
15:44:59 <tmorin> I don't know which vendor would be interested to add this
15:45:06 <tmorin> timirnich, doude ?
15:45:40 <tmorin> the tempest suite will be run as part of the OPNFV tests (against ODL, an later against bagpipe)
15:45:57 <tmorin> but this is not third-party CI integrated in Openstack Gerrit/Zuul tools
15:47:09 <bobmel> If we implement a driver for some of our platforms we might have an interest in the 3rd party ci aspects. Not an immediate issue. Was just curious if anyone had looked into that.
15:47:31 <tmorin> we haven't but this would be welcome
15:47:49 <bobmel> ok, thanks.
15:48:00 <tmorin> when it becomes a real goal, I'll be happy to help and see how this would be configured
15:48:29 <tmorin> last thing on bagpipe, is that ythomas plans to push fullstack tests (framework plus some base tests) to networking-bagpipe very soon
15:48:59 <tmorin> this will allow testing various bagpipe specific parameter/driver combinations
15:49:49 <tmorin> anything else ?
15:50:11 <tmorin> #topic open discussion
15:51:38 <tmorin> ok ...
15:51:51 <tmorin> it seems we're done for today
15:51:55 <tmorin> thanks everyone !
15:52:00 <bobmel> One things,
15:52:06 <tmorin> yep :)
15:52:32 <bobmel> were you planning anything for the PTG?
15:52:42 <tmorin> good question
15:53:09 <tmorin> I'm still confirming that I can go
15:53:28 <tmorin> anything we plan would be part of what is planned for Neutron during the PTG
15:54:17 <tmorin> has anything be announced for neutron ?
15:55:25 <bobmel> Not sure
15:55:37 <tmorin> I didn't see anyhting on the list
15:55:44 <bobmel> I have missed the two last neutron ircs
15:56:22 <tmorin> definitly a good thing to discuss during next meetings
15:56:57 <tmorin> the PTG looks like a good time to discuss BGPVPN API evolution, a topic on which we didn't advance much during H2 2016
15:58:44 <tmorin> time to close
15:58:49 <tmorin> thanks everyone !
15:58:55 <tmorin> #endmeeting