15:01:15 <carl_baldwin> #startmeeting neutron_routed_networks
15:01:16 <openstack> Meeting started Tue Apr 12 15:01:15 2016 UTC and is due to finish in 60 minutes.  The chair is carl_baldwin. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:17 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:20 <openstack> The meeting name has been set to 'neutron_routed_networks'
15:01:30 <carl_baldwin> #topic Announcements
15:01:37 <carl_baldwin> #link https://wiki.openstack.org/wiki/Meetings/Neutron-Routed-Networks
15:01:47 <tonytan4ever> o/
15:01:52 <carl_baldwin> I don't have any announcements
15:01:56 <carl_baldwin> Anyone else?
15:02:07 <reedip__> nope
15:02:11 <tonytan4ever> Well I guess I can use this chance to introduce myself
15:02:21 <mlavalle> o/
15:02:31 <rtheis> o/
15:02:49 <mlavalle> carl_baldwin: just remind the team the summit is 2 weeks aways
15:03:12 <carl_baldwin> mlavalle: Yes!  It is coming fast.
15:03:18 <reedip__> not coming .... Visa not yet approved :(
15:03:30 <reedip__> stuck in Administrative processing....
15:03:44 <reedip__> will wait for the notes :)
15:03:49 <vikram_> reedip__: Hope you get it soon
15:03:52 <carl_baldwin> reedip__: Very sorry to hear that.
15:04:35 <carl_baldwin> Okay, moving on.
15:04:45 <carl_baldwin> #topic Progress
15:05:10 <carl_baldwin> Progress slowed down this week a bit for me.  But, there is still some.
15:05:21 <carl_baldwin> #topic Basic CRUD
15:05:47 <carl_baldwin> This patch has stalled.  I got feedback from armax that I need to address which will result in some rework.
15:06:04 <carl_baldwin> I decided to finish up my action items from last week before I go back to address that.
15:06:06 <blogan> carl_baldwin: is the rework just making it its own plugin and not a mixin to ml2?
15:06:40 <carl_baldwin> blogan: I think you're thinking down the right lines but the terminology might be a little off.
15:06:54 <carl_baldwin> The feedback is to use the extension manager instead of a mixin.
15:07:07 <carl_baldwin> The mixin was never really tied to ml2.
15:07:44 <carl_baldwin> Anyway, I'm hoping to sneak in the changes before summit.  But, with summit approaching, it is beginning to take my time away.
15:08:01 <blogan> carl_baldwin: i guess i dont understand how the extension manager comes into play, but I can wait to see
15:09:08 <carl_baldwin> blogan: I don't fully understand it either but mlavalle has been through a similar change for DNS.  I have some memories of reviewing that.
15:09:37 <carl_baldwin> As I recall, mlavalle didn't think it was as bad as it first looked.  Is that right mlavalle ?
15:09:58 <mlavalle> carl_baldwin: it is really easy
15:10:09 <mlavalle> carl_baldwin: but maybe blogan has a point
15:10:22 <blogan> mlavalle: i probably dont, i'm more ignorant right now
15:10:37 <carl_baldwin> mlavalle: Could you elaborate?
15:10:45 <blogan> i was thinking it'd be more like a service plugin
15:11:02 <mlavalle> the way I used th extension manager (and the way I understand it) the extension manager allows you to process additional attributes for existing top resources
15:11:23 <reedip__> Thats what my understanding is as well ... ^^
15:11:32 <blogan> mlavalle: it allows to extend existing resources yes, but something still has to handle it
15:12:00 <mlavalle> carl_baldwin: are you adding create_segment, update_segment, etc methods to ml2?
15:12:30 <carl_baldwin> mlavalle: Later on, those will come.
15:12:49 <mlavalle> in that case the extension manager will be useful
15:13:11 <xiaohhui> so, the extension manager is for ml2, right?
15:13:22 <mlavalle> xiaohhui: yes
15:13:24 <blogan> ohhh ml2 extensions
15:13:58 <mlavalle> carl_baldwin: and the extension manager will allow you to add segment related functionality to existing top resouces
15:14:01 * blogan finally put 1 and 1 together
15:14:31 <reedip__> carl_baldwin: while developing l2  adjacency, I also was thinking how to make all the extensions work in a single file
15:14:34 <mlavalle> xiaohhui, blogan: neutron/plugins/ml2/extensions
15:14:46 <blogan> mlavalle: yeah i was thinking of api extensions
15:15:51 <xiaohhui> but the segment now is as a first class object, we will create/delete it in ml2 directly, will the extension manager still apply?
15:15:51 <mlavalle> in any case, is not difficult at all. not worst than a mixin
15:16:13 <carl_baldwin> mlavalle: Interesting.  I might have to pick your brain on this once I get the IPAM changes up for review.
15:16:22 <mlavalle> xiaohhui: yeah that is why I am making the distinction
15:16:45 <carl_baldwin> reedip__: I might be able to provide some assistance once I get over the learning curve myself.
15:17:18 <mlavalle> xiaohhui: we are going to add create, update, etc methods to ml2. Just for that, we might not need an extension
15:17:18 <carl_baldwin> This discussion has given me a good starting point.  I hope to get to it tomorrow.
15:17:47 <reedip__> carl_baldwin: okay, I think review would be the best way to move forward
15:17:48 <blogan> i look forward to learning about it too
15:18:11 <carl_baldwin> OK, let's move on and plan to get to that in review later this week.
15:18:22 <carl_baldwin> #topic Associate Subnets to Segments and IPAM
15:18:52 <carl_baldwin> I'm still working this.  I have it basically working in my sandbox but I have a bunch of UTs to fix.
15:19:19 <tonytan4ever> Well I am interested in picking this task up
15:19:57 <carl_baldwin> Basically, what I did was I added a way to pass the IPAM driver a list of subnet ids instead of just one.  That way the driver can optimize.  But, identifying the candidate subnets will still be in Neutron.
15:20:39 <carl_baldwin> tonytan4ever: I will welcome your feedback on the patch when I post it for review.  From there, maybe we can work on a way for you to help it continue to move forward.
15:21:12 <tonytan4ever> sounds like a good plan. I will look at the PR once it is ready
15:21:32 <carl_baldwin> I will reset my goal to get this posted by the end of the day today.
15:22:27 <tonytan4ever> cool.
15:22:29 <carl_baldwin> tonytan4ever: Thanks, I'll set the topic to "bp/routed-networks" so it should be easy to find once it is up.  Watch for it in about 24 hours.
15:22:48 <tonytan4ever> Cool. I will keep an eye on it.
15:22:57 <carl_baldwin> That's all I have on this.
15:23:05 <carl_baldwin> #topic Nova Scheduler
15:23:17 <carl_baldwin> mlavalle: Hi.  I think we've made some progress on this.
15:23:30 <mlavalle> yes, we made great progress on this
15:24:13 <mlavalle> esentially we wanted to get guidance / agreement from jaypipes on how to handle a couple of use cases between Nova resource pools and Neutron
15:24:16 <carl_baldwin> I ran our thoughts about Neutron remembering at what point a port got its IP address (deferred or at port create) by armax.
15:24:36 <carl_baldwin> armax pointed out some synergy with that idea and this blueprint:
15:24:37 <carl_baldwin> https://blueprints.launchpad.net/neutron/+spec/vm-without-l3-address
15:25:49 <mlavalle> great, will take a look and compare to what we discussed yesterday with Jay
15:25:56 <carl_baldwin> In that blueprint, someone suggested a tri-state flag on the port to tell if the IP address will be 1) assigned right away, 2) deferred, or 3) never assigned.
15:26:23 <mlavalle> that starts sounding to something we might use
15:26:28 <carl_baldwin> mlavalle: Given this data on the port, we can derive what we need to know about the port when Nova goes to unbind the port.
15:26:47 <mlavalle> yeah, I'll dig in :-)
15:26:48 <carl_baldwin> It was a very astute observation by armax, our fearless leader.
15:27:10 <mlavalle> he is always astute... especially in Gerrit
15:28:18 <carl_baldwin> In a way, this gives us a head start in learning what obstacles we would run in to trying to get deferred IP allocation working with Nova.
15:28:59 <carl_baldwin> From a different more pessimistic perspective, it does look like a little bit more work.
15:29:25 <mlavalle> yes
15:29:36 <carl_baldwin> mlavalle: Anyway, let's make sure that we can make that blueprint successful and I think we'll get what we need out of it.
15:29:43 <carl_baldwin> I've assigned it to myself.
15:29:47 <mlavalle> the good news is that we are uncovering this hurdles early in the game
15:29:53 <carl_baldwin> mlavalle: ++
15:30:22 <reedip__> I think the logs of the discussion would be very valuable ....
15:30:31 <carl_baldwin> Anything else related to Nova?
15:31:01 <carl_baldwin> mlavalle: ^
15:31:07 <mlavalle> carl_baldwin: did you have a chance to talk to armax about discussing routed networks in the jont session with Nova in Austin?
15:31:24 <carl_baldwin> mlavalle: Ah, yes we did discuss that a bit.
15:31:52 <carl_baldwin> He was aware of the session and the time slot and this is on the top of his list for that session.
15:32:01 <mlavalle> ++
15:32:31 <mlavalle> that's what we need, especially since this sesion goes right after the Nova team discusses the scheduler
15:33:01 <carl_baldwin> I'm actually feeling pretty good about the decision to keep track of those use cases in Neutron and adjusting the resource pools in Nova.
15:33:10 <mlavalle> it was setup this way by mriedem to lead to the routed networks conversation
15:33:51 <carl_baldwin> We're going to have to have a pretty good idea about how we're going to manage host aggregates and resource pools in Nova.
15:33:52 <blogan> mlavalle, carl_baldwin: was that on the schedule armax posted yesterday?
15:33:56 <blogan> the joint nova sessions?
15:34:16 <mlavalle> blogan: no, it is in the nova schedule
15:34:25 <blogan> mlavalle: okay good to know
15:35:01 <mlavalle> carl_baldwin: the other thing I want to mention to the team is that the Newton spec for generic pools resources is up for review here:
15:35:09 <carl_baldwin> blogan: Looks like our sessions start later that day.
15:35:10 <mlavalle> #link https://review.openstack.org/#/c/300176/
15:35:37 <carl_baldwin> mlavalle: Thanks, I'll read through it.
15:35:50 <mlavalle> we need this spec to get off the ground on the Nova side for our work on routed networks to move ahead
15:36:16 <carl_baldwin> mlavalle: ++
15:36:38 <mlavalle> so we can use all the help we can get
15:37:01 <carl_baldwin> Yes.
15:37:13 <carl_baldwin> mlavalle: Anything else to discuss on this topic?
15:37:25 <mlavalle> that's it for this week. we can move on
15:37:36 <carl_baldwin> #topic Host / Segment Mapping
15:37:56 <mlavalle> hi again
15:37:59 <carl_baldwin> mlavalle: You're a busy guy
15:38:19 <mlavalle> so I've been reviewing https://review.openstack.org/#/c/205631/33 since carl_baldwin pointed me to it
15:38:42 <mlavalle> I definitely learned a bit from it
15:39:44 <mlavalle> we can re use check_segment_for_agent in SimpleAgentMechanismDriverBase
15:40:22 <mlavalle> to decide on host segment mapping on data coming from the agents
15:41:11 <mlavalle> and then we might refactor the work in https://review.openstack.org/#/c/205631/33 to help that woprk to scale up better
15:41:22 <blogan> this looks very similar to the dhcp agent scheduling for a segment too
15:41:24 <carl_baldwin> mlavalle: That sounds pretty good.
15:41:43 <mlavalle> the only de-tour we will have to take is with the SR-IOV mechanism agent
15:42:05 <carl_baldwin> blogan: Yes, it is very related.  In fact, it might be a good place to jump in and help things along.
15:42:21 <blogan> carl_baldwin: definitely
15:42:23 <mlavalle> it wasn't originally immplemented inheriting from SimpleAgentMechanismDriverBase
15:42:50 <mlavalle> so it doesn't inherit check_segment_for_agent
15:43:16 <mlavalle> the reason is that the SR-IOV driver originbally had the option to run without agent
15:43:27 <mlavalle> but that option was eliminated in Mitaka
15:43:37 <carl_baldwin> mlavalle: interesting
15:44:26 <mlavalle> I checked with irenab this morning (the SR-IOV implementer) and she doesn't see any problem in having this mechanism driver inherit from SimpleAgentMechanismDriverBase
15:44:56 <mlavalle> I will start a patch to do just that
15:45:11 <carl_baldwin> mlavalle: Great, I'm happy to be a reviewer.  blogan you too?
15:45:40 <mlavalle> with that all the in tree mechanism drivers will have a uniform interface that we will leverage for host segment mapping and then improve the dhcp agent work
15:45:55 <blogan> carl_baldwin: yes!
15:46:18 <mlavalle> carl_baldwin: I thnk this is good progress report for this week :-)
15:46:45 <carl_baldwin> mlavalle: great!
15:46:59 <carl_baldwin> #topic Create / delete segment in ML2
15:47:21 <carl_baldwin> xiaohhui: hi, I'm not sure you can do much with this until I figure out what to do with the basic CRUD patch.
15:47:35 <xiaohhui> yeah,
15:48:11 <xiaohhui> I added a new patch, the code looks good from DB's view
15:48:34 <xiaohhui> But I am afraid it will need change based on your change on basic CRUD
15:49:03 <xiaohhui> One thing to make sure, we will not support update for segment, right?
15:49:34 <rtheis_> if name and description are added then update will be supported
15:49:48 <rtheis_> I think reedip was working on that
15:49:51 <carl_baldwin> xiaohhui: I don't think so for the existing attributes.  We might support it for the name attribute.  But, that shouldn't be much of a problem I don't think.
15:50:19 <carl_baldwin> rtheis_: Right, and description .
15:50:25 <reedip__> my patch is STUCK
15:50:51 <carl_baldwin> reedip__: Mine too.  ;)
15:51:07 <xiaohhui> OK, I can continue to look into to see if there is any issue with the real functionality, now I only tried with DB data change.
15:51:36 <xiaohhui> I think that will be useful even if the code might change for the extension manager.
15:51:43 <reedip__> dont know why ...aLL: care to give some help ? https://review.openstack.org/293305 ?
15:52:30 <carl_baldwin> reedip__: I've got to prioritize the preceding patch adding the basic CRUD.  It will need some rework.
15:52:39 <carl_baldwin> xiaohhui: ack
15:52:48 <carl_baldwin> xiaohhui: I think you're right.
15:53:02 <reedip__> carl_baldwin :syn ack...
15:53:10 <xiaohhui> I think that is all for this work
15:53:20 <carl_baldwin> reedip__: Thanks.
15:53:27 <blogan> reedip__: i think that should be a new extension
15:53:41 <blogan> api extension
15:53:46 <carl_baldwin> blogan: Why a new extension?
15:54:23 <reedip__> blogan; I dont think so, but please let me know ur reasons for  the same?
15:54:24 <blogan> oh god nvm dont mind me
15:54:45 <reedip__> I think u r still in the extension Universe ....
15:54:47 <blogan> i was thinking it was an older extension but realized it was the one just added
15:54:56 <carl_baldwin> blogan: :)
15:54:58 <blogan> too early for me, carry on
15:55:16 <carl_baldwin> blogan: After I asked I realized that's probably what you were thinking.
15:55:29 <carl_baldwin> I think we're safe to keep this extension open until the release.
15:55:43 <carl_baldwin> Let's just skip right to Open Agenda.
15:55:44 <blogan> carl_baldwin: yep!
15:55:47 <carl_baldwin> #topic Open Agenda
15:56:27 <carl_baldwin> Anything on the client, DHCP, or anything else?
15:56:34 <carl_baldwin> 3.5 minutes remaining.
15:56:36 <blogan> so i've been ramping up on the dhcp agent stuff, just learing at this point
15:57:19 <reedip__> nothing on neutronclient , WIP for l2 publisjhed, the name/desc patch is stuck but needs API tests rewritten I guess
15:57:20 <rtheis_> osc client patches were updated to align with API changes
15:57:36 <carl_baldwin> reedip__: Thanks.
15:57:39 <blogan> who is someone who knows a lot about the dhcp agent parts that I could ask questions to whenever I need to?
15:57:39 <rtheis_> still a WIP but good progress
15:57:44 <carl_baldwin> rtheis_: Thanks!
15:57:52 <rtheis_> yw
15:58:02 <reedip__> check the Lieutenant
15:58:05 <xiaohhui> there is a patch in networking-ovn to read the data that routed networks need
15:58:06 <reedip__> blogan :
15:58:17 <carl_baldwin> blogan: ZZelle might be a good one.  mlavalle too by now.
15:58:31 <blogan> id rather nto talk to mlavalle anymore
15:58:40 <blogan> did i say that out loud?
15:58:45 <mlavalle> blogan: i'd buy you luncj ;-)
15:58:47 * blogan loves mlavalle
15:58:53 <carl_baldwin> blogan: I'm always looking for opportunities to talk to mlavalle
15:58:54 <mlavalle> lunch^^^
15:59:15 <blogan> carl_baldwin: you're a masochist i see
15:59:35 <reedip__> who said there is no such thing as a free lunch :P
15:59:59 <carl_baldwin> Well, we've got seconds left ...
16:00:03 <carl_baldwin> #endmeeting