15:00:16 <tidwellr> #startmeeting neutron_l3
15:00:17 <openstack> Meeting started Thu Feb 11 15:00:16 2016 UTC and is due to finish in 60 minutes.  The chair is tidwellr. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:18 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:21 <openstack> The meeting name has been set to 'neutron_l3'
15:00:31 * tidwellr lets people file in
15:00:55 <vikram__> hi
15:01:22 <tidwellr> #topic Announcements
15:01:49 <tidwellr> Mitaka-3 is approaching quickly
15:02:17 <carl_baldwin> Hi
15:02:29 <tidwellr> Any other announcements?
15:03:01 <mlavalle> the Neutron mid-cycle in Rochester MN
15:03:23 * neiljerram says hi
15:03:31 <mlavalle> https://etherpad.openstack.org/p/neutron-mitaka-midcycle
15:03:50 <tidwellr> yes, lots of good topics being discussed
15:04:06 <tidwellr> unfortunately I will not be able to make it :(
15:04:43 <tidwellr> any other announcements?
15:05:26 <tidwellr> #topic Bugs
15:05:56 <tidwellr> #link https://bugs.launchpad.net/neutron/+bug/1478100
15:05:57 <openstack> Launchpad bug 1478100 in neutron "DHCP agent scheduler can schedule dnsmasq to an agent without reachability to the network its supposed to serve" [High,In progress] - Assigned to Cedric Brandily (cbrandily)
15:07:14 <tidwellr> any updates or items to discuss on this one?
15:07:34 <carl_baldwin> Has been somewhat starve for review time.
15:08:07 <mlavalle> JUst that the patchset https://review.openstack.org/#/c/205631/ got a -1 from Garyk. I think CEdric will respond soon
15:08:43 <mlavalle> actually he already responded
15:09:59 <tidwellr> ok, anything we need to discuss in this setting?
15:10:43 <mlavalle> not from me
15:11:10 <tidwellr> mlavalle: thanks
15:11:13 <tidwellr> next is https://bugs.launchpad.net/neutron/+bug/1543094
15:11:15 <openstack> Launchpad bug 1543094 in neutron "[Pluggable IPAM] DB exceeded retry limit (RetryRequest) on create_router call" [High,Confirmed] - Assigned to Salvatore Orlando (salvatore-orlando)
15:12:55 <carl_baldwin> Looks like we just wait for salv-orlando to post a fix.
15:13:01 <mlavalle> yeap
15:13:03 <tidwellr> ok
15:13:41 <tidwellr> alright, moving on
15:14:43 <tidwellr> there a couple of IPAM RFE's #link https://bugs.launchpad.net/neutron/+bug/1541895 and #link https://bugs.launchpad.net/neutron/+bug/1540512
15:14:45 <openstack> Launchpad bug 1541895 in neutron "[RFE] [IPAM] Make IPAM driver a per-subnet pool option" [Wishlist,Confirmed] - Assigned to John Belamaric (jbelamaric)
15:14:46 <openstack> Launchpad bug 1540512 in neutron "Host dependent IPAM" [Wishlist,Confirmed]
15:15:15 <neiljerram> I can provide context for the latter, if helpful
15:15:28 <mlavalle> I added them yesterday to the agenda. carl_baldwin mentioned a few days ago that we want to track the rfe's now
15:15:35 <carl_baldwin> :)
15:15:45 <tidwellr> neiljerram: please go ahead :)
15:16:00 <neiljerram> OK, interesting situation here.
15:16:27 <neiljerram> Basic idea is to be able to allocate VM IP addresses in a clustered way - per-host, per-rack or per-pod.
15:17:03 <neiljerram> Main reason for that is to allow fabric routers to aggregate VM routes - if you're using a routed network implementation.
15:17:34 <neiljerram> I suggested this as an Outreachy project, and Petra Sargent stepped up there, and hence is now my mentee on this.
15:18:13 <neiljerram> Interestingly, though, it came to my notice yesterday that the Romana project have already implemented something that looks very like this.
15:18:34 <neiljerram> But anyway, questions for this forum....
15:19:33 <carl_baldwin> neiljerram: Do you have any links to Romana?
15:19:36 <neiljerram> 1. whether it's interesting to do a pluggable IPAM driver for this class of use cases; Carl suggested that the routed networks spec would provide an alternative approach based on the proposed Segments
15:19:57 <neiljerram> https://github.com/romana/networking-romana/blob/master/networking_romana/driver/ipam_romana.py
15:20:25 <carl_baldwin> neiljerram: I think there is that one significant different between your requirements and those of the routed networks spec.
15:20:41 <neiljerram> 2. if yes, whether there's any part of this that is worth commonizing/upstreaming, and hence for possible inclusion in core Neutron
15:20:42 <carl_baldwin> neiljerram: namely that your boundaries are soft.
15:20:53 <neiljerram> carl_baldwin, yes indeed
15:21:14 <carl_baldwin> neiljerram: Looking at the spec now, I see that both use cases have deferred port binding in common.
15:21:41 <neiljerram> carl_baldwin, yes, and that's a part of the problem that I haven't really tackled at all yet.
15:21:47 <carl_baldwin> ... because you need the host binding information to make the IP allocation.
15:21:57 <tidwellr> neiljerram: From a BGP perspective I like the idea of being able to aggregate DVR host routes where possible, this sounds like a way we could enable that
15:21:59 <carl_baldwin> neiljerram: We'll need to figure that one out together.
15:22:37 <carl_baldwin> tidwellr: Good point, when we work floating IPs in to routed networks, this could indeed be useful in that scenario.
15:23:03 <neiljerram> Another angle here is our experience of ramping up on the pluggable IPAM driver API...
15:23:36 <neiljerram> First that it's been hard, and secondly that a relatively simple enhancement like this seems to require subclassing lots of classes.
15:23:53 <carl_baldwin> neiljerram: I notice that your third point in the rfe description is a prerequisite for the first point.
15:24:34 * carl_baldwin doesn't see johnbelemaric around.
15:24:54 <neiljerram> Strictly no, I think
15:25:20 <neiljerram> I managed to find a sequence of existing CLI commands that would cause Neutron to allocate a new IP after the host was known
15:26:38 <neiljerram> Stepping back... the reason this RFE exists is because the neutron-specs process requires it.
15:27:11 <neiljerram> And the reason I asked Petra to write a neutron-spec was to be a forum where we could start discussing design with John, Pavel etc.
15:27:59 <carl_baldwin> neiljerram: Is there a spec up for review?
15:28:01 <neiljerram> But as it's not clear that any upstreaming is needed, perhaps I should make that a networking-calico spec instead, and cancel this as a Neutron RFE
15:28:31 <neiljerram> carl_baldwin, yes: https://review.openstack.org/#/c/273868/
15:28:43 <neiljerram> But currently it doesn't have much more content than the RFE
15:28:49 <carl_baldwin> neiljerram: Thanks.
15:29:55 <carl_baldwin> neiljerram: You make some good points.  It sounds like you can make some progress without any upstreaming.  But, I think to do deferred port binding correctly in all cases will require some work in Neutron and Nova.
15:30:33 <carl_baldwin> Could you document your cli commands that cause Neutron to allocate after the host is known?  Is it just letting Nova create the port?
15:31:01 <neiljerram> Yes, will do.
15:31:17 <neiljerram> Unfortunately I don't recall the detail right now - it was tricky!
15:31:37 <carl_baldwin> neiljerram: Thanks.
15:31:55 <tidwellr> neiljerram: very interesting topic, I have thought a lot about host-aware IPAM, thanks doing more than just thinking!
15:32:14 <carl_baldwin> neiljerram: I'm going to move the RFE to Triaged to discuss it in an upcoming drivers meeting.
15:32:33 <neiljerram> carl_baldwin, OK, thanks.
15:32:36 <tidwellr> neiljerram: anything else you'd like to discuss?
15:32:48 <neiljerram> No, thank you.
15:32:58 <tidwellr> alright, moving on
15:33:07 <tidwellr> next RFE is #link https://bugs.launchpad.net/neutron/+bug/1370033
15:33:08 <openstack> Launchpad bug 1370033 in neutron "Admin should be able to manually select the active instance of a HA router" [Wishlist,Triaged] - Assigned to Hong Hui Xiao (xiaohhui)
15:33:39 <mlavalle> tidwellr: I just wanted to draw attention to this one. Seems to have been languishing for some time
15:34:32 <tidwellr> mlavalle: thanks, attention drawn :)
15:34:43 <carl_baldwin> More or less, yes.  We may need the rest of the time for other topics though.
15:34:54 <tidwellr> ok, we'll move on then
15:35:06 <tidwellr> #topic Routed Networks
15:35:28 <tidwellr> carl_baldwin: you have the floor
15:35:56 <carl_baldwin> I justed wanted to let people know that I'm going to run another meeting to dive in to the details of routed networks.
15:36:19 <carl_baldwin> #link https://review.openstack.org/#/c/278114/
15:36:42 <carl_baldwin> We'll make the decision on the time(s) soon and then I'll worry about why Jenkins is failing and get it merged.
15:36:51 <carl_baldwin> It is time for the rubber to hit the road.
15:36:56 <carl_baldwin> Also, please review the specs
15:37:06 <carl_baldwin> #link https://review.openstack.org/#/c/225384/
15:37:17 <tidwellr> ok, so we'll start having deeper discussions in that forum
15:37:18 <carl_baldwin> #link #link https://review.openstack.org/#/c/263898/
15:37:27 <carl_baldwin> tidwellr: Back to you.
15:37:46 <tidwellr> carl_baldwin: thanks
15:37:53 <tidwellr> #topic BGP
15:38:13 <tidwellr> Vikram__: ping
15:38:35 <carl_baldwin> tidwellr: I just rebased your patch due to conflict.
15:38:41 <carl_baldwin> The conflict resolution was trivial.
15:39:08 <tidwellr> carl_baldwin: thanks
15:39:16 <carl_baldwin> tidwellr: I'd be happy to continue rebasing the chain if you want to focus on other stuff.
15:39:50 <tidwellr> carl_baldwin: thanks for the help, that would be great
15:40:08 <tidwellr> I think we have a solution to some of the gate issues for now
15:40:18 <carl_baldwin> tidwellr: vikram__ : Be advised that I'll be rebasing the chain.
15:41:18 <carl_baldwin> Actually, looks like vikram__ has already been working on rebasing.
15:41:24 <carl_baldwin> I'll pick up where he left off.
15:41:36 <tidwellr> I'm going to spend some time today testing with the patch sets we have
15:41:56 <carl_baldwin> tidwellr: Great, I'll be reviewing these patches today as my top priority.
15:42:17 <tidwellr> I want to be sure we can start merging the code that handles centralized routers, the DVR stuff can wait for a minute while we get that moving
15:43:04 <tidwellr> I didn't have anything else, Vikram__?
15:43:46 <carl_baldwin> ... or vikram__ ?
15:44:17 <carl_baldwin> Let's just keep merging!
15:44:30 <tidwellr> +1
15:44:38 <tidwellr> #topic DNS
15:44:51 <mlavalle> hi again
15:44:56 <tidwellr> mlavalle: anything on your mind?
15:45:05 <mlavalle> quick update
15:45:22 <mlavalle> the Nova side patch is waiting a final +2 https://review.openstack.org/#/c/271578
15:45:54 <mlavalle> I also added unit tests and a realease note here: https://review.openstack.org/#/c/277302
15:46:26 <mlavalle> and finally over the next few days I will be pushing the Networking Guide chapter that covers this topic
15:46:52 <mlavalle> so, good progress overall
15:46:57 <mlavalle> any questions?
15:47:58 <carl_baldwin> mlavalle: Just reading Michael's comment about validation.
15:48:16 <mlavalle> carl_baldwin: I responded to him
15:48:18 <carl_baldwin> mlavalle: Are they really validating against the same RFCs?
15:48:47 <mlavalle> carl_baldwin: I let myself to be guided by the comment... you don't think so?
15:48:50 <carl_baldwin> Maybe I forgot the outcome of the discussion but I still had it in mind that their validation was *slightly* different.
15:49:06 <carl_baldwin> Though, I agree that it shouldn't be seen as a problem either way.
15:49:16 <mlavalle> yeah, I agree
15:49:47 <carl_baldwin> But, now that the discussion is coming back to me.  Maybe my concern is no longer valid.
15:50:06 <carl_baldwin> I had thought they would allow a name to begin with a digit and we would not.
15:50:56 <tidwellr> mlavalle: thanks for the update and the great work here
15:51:05 <mlavalle> :-)
15:51:08 <tidwellr> anything else you'd like to discuss?
15:51:15 <mlavalle> I'm done
15:51:33 <tidwellr> #topic Address Scopes
15:51:53 <tidwellr> carl_baldwin: we have address scopes!!!!
15:52:07 <carl_baldwin> Yeah!
15:52:37 <carl_baldwin> I need to document now.  I've started but that is a whole other learning curve to climb.
15:52:46 <carl_baldwin> I didn't make progress as quickly as I wanted to.
15:52:59 <mlavalle> ++
15:53:19 <carl_baldwin> But, I blogged with some good technical content that I plan to work in to the docs
15:53:21 <carl_baldwin> #link http://blog.episodicgenius.com/post/neutron-address-scopes/
15:54:02 <tidwellr> blogs are how we doc things these days, right? ;)
15:55:16 <carl_baldwin> Don't let Sam-I-Am hear you say that!
15:55:37 * tidwellr ducks under the table
15:55:52 <mlavalle> yeah, tidwellr you are getting yourself in deep trouble :-)
15:56:20 <carl_baldwin> It did help me.  There is something about writing a blog post that makes it easier for me to get through.
15:56:26 <carl_baldwin> Anyway, back to you tidwellr
15:57:03 <tidwellr> thanks, looks like features are getting buttoned up nicely for Mitaka
15:57:17 <tidwellr> #topic Open Discussion
15:57:54 * tidwellr begins counting down
15:58:17 <mlavalle> LOL
15:58:43 <tidwellr> alright, thanks everyone!
15:58:51 <mlavalle> thanks!
15:59:01 <tidwellr> #endmeeting