15:13:27 <tmorin> #startmeeting bgpvpn
15:13:28 <openstack> Meeting started Tue Nov 15 15:13:27 2016 UTC and is due to finish in 60 minutes.  The chair is tmorin. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:13:30 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:13:32 <openstack> The meeting name has been set to 'bgpvpn'
15:13:35 <tmorin> hi everyone
15:13:39 <bobmel> tmorin: hi
15:13:46 <tmorin> hi bobmel
15:13:55 <bobmel> tmorin: thx for your comments
15:14:12 <bobmel> tmorin: review comment
15:14:13 <tmorin> hi doude, eon`, lewo, matrohon, pcarver, timirnich
15:14:21 <tmorin> you're welcome :)
15:14:28 <tmorin> thanks for working on this !
15:14:52 <bobmel> tmorin: I have some follow up questions but perhaps we can take those here after this meeting
15:15:41 <tmorin> we sure can, during of after, depending what else we  cover
15:16:56 <bobmel> tmorin: ok, great
15:18:05 <tmorin> hmm
15:18:13 <tmorin> we aren't many :)
15:18:36 <tmorin> some people may still be on DST...
15:19:11 <tmorin> bobmel: let's discuss the point you have about the tempest scenario testing
15:19:29 <tmorin> we can still cover other topics when/if more people come up
15:19:43 <bobmel> Ok, yes we might as well do that since nobody else seems to be around
15:20:13 <bobmel> So I have my single node devstack setup with bpgvpn up (bagpipe driveR)
15:20:58 <bobmel> Now, in the scenario test outlines with the two networks/subnets attached to the vpn
15:21:42 <bobmel> Should that vpn instance forward packets between the vms on those two networks/subnets
15:21:44 <bobmel> ?
15:22:47 <tmorin> in a single node setup, the one OVS will forward the packets between the VMs on those two networks
15:23:28 <tmorin> the traffic path will be br-int -> br-tun -> br-mpls -> br-tun -> br-int
15:23:50 <tmorin> with the internal vlan id changed by br-mpls
15:24:05 <tmorin> does it answer your question ?
15:24:37 <bobmel> Yes think so.
15:25:19 <bobmel> Is it the bagpipe agent driver that will create the ovs rules to make that vlan change?
15:26:10 <tmorin> bagpipe_bgpvpn extension inside the agent will receive BGPVPN information in RPCs and pass them to bagpipe-bgp via its local REST API
15:26:16 <bobmel> At the moment i have not been able to ping between the two VMs
15:26:30 <tmorin> and its bagpipe-bgp that will (a) exchange BGP VPN routes and (b) populate br-mpls
15:26:58 <bobmel> Aha, ok, I see.
15:27:01 <tmorin> (b) above include rule that change the vlan id
15:27:07 <tmorin> rules
15:27:28 <tmorin> traffic is processed in a given VRF based on its local vlan id
15:28:15 <tmorin> if the prefix matched in the VRF is on the same compute, the packet is resubmitted to appear as incoming MPLS traffic, and matched in the destination VRF
15:28:47 <tmorin> this destination VRF pops the MPLS label, sets the destination local vlan id, and passes the packet to br-tun
15:29:14 <tmorin> these diagrams may help: http://docs.openstack.org/developer/networking-bgpvpn/bagpipe/index.html
15:29:29 <tmorin> (to some degree only, as they don't have all details)
15:31:12 <bobmel> Yes I have had a look at that one.
15:31:59 <bobmel> So what you wrote above describes what should happen in the single node setup?
15:33:20 <tmorin> yes
15:34:48 <tmorin> in a multinode setup, the path would be :  node A [ br-int -> br-tun -> br-mpls ] --> -- (phy or GRE) --> node B [ br-mpls -> br-tun -> br-int]
15:37:45 <bobmel> I wonder if it is maybe the route targets config that prevents the ping to work
15:38:31 <tmorin> can you do a "bagpipe-looking-glass -r" and put the output on  paste.openstack.org ?
15:40:14 <bobmel> on sec
15:42:22 <bobmel> http://paste.openstack.org/show/589301/
15:44:09 <bobmel> The devstack host IP is 172.16.6.20
15:45:14 <tmorin> vpns / instances is empty
15:45:32 <bobmel> Lots of those OpenWaitTimeout messages in hte log
15:45:38 <tmorin> it means that the bagpipe_bgpvpn agent extension passed no BGP VPN attachement information to bagpipe-bgp
15:46:31 <tmorin> you can ignore these OpenWaitTimeout message: bagpipe-bgp is trying to connect its default BGP peer, but we don't need any in a single node setup  (I think we should actually get rid of this default BGP peer setup)
15:46:54 <bobmel> Ok, got it.
15:46:57 <tmorin> so its likely that the bagpipe_bgpvpn agent extension has no BGPVPN attachement information
15:47:05 <tmorin> do you see warnings in q-agt.log ?
15:47:12 <bobmel> Let me check
15:49:37 <bobmel> No warnings
15:49:55 <bobmel> I see lines like this:
15:49:56 <bobmel> bagpipe-bgp client request: GET ping [None
15:50:07 <bobmel> bagpipe-bgp returns [200:96867148709979766103418085043297248131]
15:50:25 <bobmel> bagpipe-bgp PING response received with sequence number 96867148709979766103418085043297248131 f
15:50:31 <tmorin> it simply means that bagpipe_bgpvpn extension can properly talk to bagpipe-bgp
15:50:52 <bobmel> Ok
15:50:53 <tmorin> the sequence number is used to let the agent extension know when/if bagpipe-bgp restarts
15:51:22 <tmorin> if you grep for 'bagpipe' in q-agt.log, you should see RPCs about BGP VPNs
15:51:32 <tmorin> assuming that Neutron is actually sending thos
15:52:08 <tmorin> those
15:52:27 <tmorin> but you could first confirm that (checking about warnings/errors in q-svc.log, grepping for bagpipe)
15:55:02 <bobmel> Here's some output from the q-agt log
15:55:03 <bobmel> http://paste.openstack.org/show/589303/
15:55:50 <tmorin> based on these logs you should have vpn attachment defined in bagpipe-bgp
15:56:42 <tmorin> "bagpipe-looking-glass summary vpn_instances_count" should give you something different from 0
15:56:47 <bobmel> so it should be visible in the looking glass?
15:56:54 <tmorin> but you had zero in the output you sent earlier
15:56:55 <tmorin> yes
15:57:30 <bobmel> it return 0
15:57:32 <bobmel> odd
15:59:15 <bobmel> RPC to the q-agent for bagpipe driver, is that triggered when a network is attached to the vpn?
15:59:25 <tmorin> yes
15:59:59 <bobmel> Is there a way to restart bagpipe itself?
15:59:59 <tmorin> or when a Port goes active
16:00:11 <bobmel> Ok, got it.
16:00:17 <tmorin> in screen, b-bgp tab, stop restart
16:00:18 <tmorin> yes
16:00:51 <tmorin> is the q-agt.log that you sent complete, I'm surprised to see nothing after the last line
16:01:08 <bobmel> No, it was a subset of it
16:01:26 <adrian_otto> time to wrap up?
16:01:40 <tmorin> #endmeeting bgpvpn