15:00:09 <mlavalle> #startmeeting neutron_l3
15:00:10 <openstack> Meeting started Thu Jun  8 15:00:09 2017 UTC and is due to finish in 60 minutes.  The chair is mlavalle. 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 'neutron_l3'
15:00:22 <davidsha> Hi
15:00:35 <haleyb> hi
15:00:53 <mlavalle> #chair haleyb
15:00:53 <openstack> Current chairs: haleyb mlavalle
15:01:27 <mlavalle> hi haleyb davidsha . Nice to see you!
15:01:38 * haleyb waves
15:01:48 <hjensas> hello
15:02:00 <mlavalle> hjensas: welcome
15:02:18 <hjensas> mlavalle: thanks.
15:02:26 <mlavalle> #topic Announcements
15:03:17 <mlavalle> This is the week of the Pike-2 milestone:
15:03:24 <mlavalle> #link https://releases.openstack.org/pike/schedule.html
15:03:48 <mlavalle> Before we know it, Pike will be over
15:04:07 <mlavalle> any other announcements from the team?
15:05:03 <mlavalle> ok, let's move on....
15:05:10 <mlavalle> #topic Bugs
15:05:47 <mlavalle> I don't see swami around. haleyb do you want to bring any dvr bugs up?
15:05:51 <haleyb> sure
15:06:00 <haleyb> bug 1695101 is top of list
15:06:01 <openstack> bug 1695101 in neutron "DVR Router ports and gateway ports are not bound to any host and no snat namespace created" [Undecided,In progress] https://launchpad.net/bugs/1695101 - Assigned to Swaminathan Vasudevan (swaminathan-vasudevan)
15:06:17 <haleyb> https://review.openstack.org/#/c/470063/ is out, swami updated last night
15:06:24 <haleyb> swami is moving this week btw
15:06:32 * mlavalle now remembers that swami is off this week
15:06:45 <haleyb> i will update that today as this is a regression
15:07:37 <haleyb> bug 1695140 was filed the other day, I have not triaged
15:07:39 <openstack> bug 1695140 in neutron "SRC without FIP,DEST with FIP case is broken for dvr-multinode" [Undecided,New] https://launchpad.net/bugs/1695140
15:08:45 <haleyb> although this could be related since it would require snat namespace, which the other bug is fixing
15:09:54 <haleyb> i'll add a note to the latter
15:12:22 <haleyb> that's all for DVR that's important for now
15:12:31 <mlavalle> thanks haleyb
15:13:09 <mlavalle> on the plain vanilla dhcp-l3 bugs, first one up is https://bugs.launchpad.net/neutron/+bug/1655567
15:13:10 <openstack> Launchpad bug 1655567 in neutron "DHCPv6 failures with "address already allocated in subnet" error" [High,In progress] - Assigned to Miguel Lavalle (minsel)
15:13:34 <mlavalle> I have a fix for it marked WIP:
15:13:42 <mlavalle> #link https://review.openstack.org/#/c/469669/
15:14:20 <mlavalle> First, it got hit by the linuxbridge bug that has been around hitting us all
15:14:35 <mlavalle> so I spent some time triaging this latter bug
15:15:21 <mlavalle> Also ihrachys ponted out that we might break agentless plugins with the chosen approach
15:16:04 <mlavalle> combining that with a better understanding on my part of how this bugs comes about, earlier today I updated the fix in my local system
15:16:18 <mlavalle> I will push it again after this meeting
15:17:26 <mlavalle> Next one is https://bugs.launchpad.net/neutron/+bug/1683227
15:17:27 <openstack> Launchpad bug 1683227 in neutron "test_connection_from_same_address_scope failed with: Cannot find device "qg-74372988-7f"" [High,Confirmed] - Assigned to Miguel Lavalle (minsel)
15:17:46 <mlavalle> I've started working on this one, having made not much progress yet
15:18:33 <mlavalle> and finally we have https://bugs.launchpad.net/neutron/+bug/1694764
15:18:34 <openstack> Launchpad bug 1694764 in neutron "test_metadata_proxy_respawned failed to spawn metadata proxy" [High,Confirmed]
15:18:43 <mlavalle> which has no owner yet
15:19:24 <mlavalle> any comments or other bugs from the team?
15:20:16 <mlavalle> ok, moving on
15:20:20 <mlavalle> #topic DNS
15:20:44 <mlavalle> I've also been busy adding dns_domain attribute to ports
15:21:34 <mlavalle> specifically, adding unit tests. combining this new attribute with the existing port and network dns related attribute, gave a lot of tests cases
15:21:59 <mlavalle> I have on last to implement today and then add a release note and it will be good for reviews
15:22:45 <haleyb> nice, i will keep an eye out
15:22:48 <mlavalle> ok, moving on
15:23:02 <mlavalle> #topic Open Agenda
15:23:36 <mlavalle> hjensas: do you have a topic to bring up?
15:23:38 <hjensas> #link https://bugs.launchpad.net/neutron/+bug/1695740 - I wanted to bring up this.
15:23:40 <openstack> Launchpad bug 1695740 in neutron "Segments, routed-networks, IPAM only usecase regression" [Undecided,In progress] - Assigned to Harald Jensås (harald-jensas)
15:24:31 <hjensas> working on tripleo to deploy on routed networks, we need the ipam only usecase.
15:25:41 <hjensas> #link https://review.openstack.org/#/c/470788/ <- fix proposed here.
15:26:02 <haleyb> mlavalle: i am ok with this change, but know you have been more involved with routed networks
15:26:36 <mlavalle> hjensas: what is the ipam only use case?
15:27:26 <hjensas> mlavalle: if we create a port, and pass fixed_ips: [subnet_id|ip_address] IP allocation should not be deferred.
15:28:16 <hjensas> mlavalle: basically in heat templates several ports are created, and attributes of the port resource are used in a software-config that drive os-net-config.
15:29:31 <mlavalle> hjensas: are we going to create the port and then pass it to nova for the instance creation?
15:29:44 <hjensas> mlavalle: no, we do not pass it to nova.
15:30:50 <mlavalle> hjensas: who receives it and makes sure the subnet / segment is accesible?
15:31:49 <hjensas> mlavalle: we pass the attributes of the port, the fixed_ips: ip_address specifically, to a software-config that is executed on the target instance.
15:32:41 <haleyb> ironic, correct?
15:33:09 <hjensas> no nova or ironic never see these.
15:34:27 <mlavalle> hjensas: reading the bug, you quote use case 2 of the nova neutron-routed-networks spec. is this the case we are talking about?
15:35:08 <hjensas> mlavalle: yes, we do that too when the user is allowed to pre-select ips.
15:36:12 <mlavalle> hjensas: so the port will be passed to nova in the request to create the instance. At least that was wht the spec was talking about. are we in agreement?
15:36:13 <hjensas> but for the most, we just request a IP from the pool, and call the API to read it later and use it via configuration management on the target node.
15:36:40 <hjensas> mlavalle: this is true only for the provisioning network in triplo.
15:37:09 <mlavalle> hjensas: ok, I start to understand
15:37:41 <mlavalle> and that is what I have found confusing about it
15:37:50 <hjensas> mlavalle: but yes if the port is later bound to a host the scheduler would ideally read the attribute and sync with neutron to schedule the instance onto the good host. As we would do with instance migrations?
15:39:06 <mlavalle> hjensas: ok, in principle I am also ok with this
15:40:04 <mlavalle> I will read this bug carefully and the patchset and will ask questions if I need to
15:40:10 <mlavalle> hjensas: is that fair?
15:40:32 <hjensas> mlavalle: the way I see it, the user asked for "subnetX". So there is no need to defer, the user gave us enough info to decide. If the user ask for networkX, and does not provide a host then we need to defer.
15:41:31 <hjensas> mlavalle: yes, that is fair.
15:42:10 <mlavalle> cool, thanks
15:42:26 <mlavalle> anything else from you, hjensas
15:42:32 <mlavalle> ?
15:42:35 <hjensas> mlavalle: not today.
15:42:41 <hjensas> thanks
15:43:03 <davidsha> I have the PoC up if anyone is interested: https://review.openstack.org/#/c/472289
15:43:07 <mlavalle> hjensas: if I remeber correctly, you also filed one or more bugs related to this and related to ironic?
15:43:15 <davidsha> sorry
15:43:22 <mlavalle> davidsha: not a problem
15:43:30 <mlavalle> will be glad to look at it
15:43:44 <hjensas> mlavalle: related to heat, this bug explain the issue better.
15:44:14 <mlavalle> hjensas: can we disregard the other ones?
15:44:16 <hjensas> mlavalle: I can cross post bug references on the two bugs.
15:44:33 <hjensas> mlavalle: the neutron part of the one also related to heat, yes.
15:44:48 <mlavalle> hjensas: thanks!
15:45:06 <mlavalle> davidsha: is there anything you want to comment about the PoC?
15:45:21 <hjensas> mlavalle: # link https://bugs.launchpad.net/heat/+bug/1692567 <- this one is invalid for neutron. :)
15:45:22 <openstack> Launchpad bug 1692567 in heat "can't create neutron port fixed_ip if subnet associated with segment" [Undecided,In progress] - Assigned to Harald Jensås (harald-jensas)
15:45:38 <mlavalle> hjensas: thank you :-)
15:46:00 <davidsha> mlavalle: Ya, it currently supports east-west and north-south, but it doesn't support floating ip yet.
15:47:30 <davidsha> At the moment it makes a new agent mode and works from there. I did have to make some modifications to the agent to stop it making ip table calls, but let me know if they shouldn't be touched.
15:48:38 <mlavalle> it is fine with me
15:50:58 <mlavalle> davidsha: anything else you want us to know?
15:50:59 <davidsha> cool, let me know what you all think and I'll try to respond quickly to comments
15:51:33 <davidsha> mlavalle: one more thing
15:52:16 <davidsha> If your deploying this, the L3 agent will need to use ovs-vsctl not native
15:52:29 <mlavalle> ok
15:52:50 <mlavalle> in any case, I'll ping you if I have problems, davidsha :-)
15:53:04 <mlavalle> I know where you live
15:53:13 <davidsha> Thanks, I'll respond as quick as I can!
15:54:38 <mlavalle> haleyb, hjensas, davidsha thanks for attending the meeting :-)
15:54:48 <davidsha> Thanks, cya!
15:54:50 <mlavalle> #endmeeting