16:01:11 <Sukhdev> #startmeeting networking_ml2
16:01:12 <openstack> Meeting started Wed Mar  2 16:01:11 2016 UTC and is due to finish in 60 minutes.  The chair is Sukhdev. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:01:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:01:16 <openstack> The meeting name has been set to 'networking_ml2'
16:01:28 <Sukhdev> #topic: Agenda
16:01:36 <Sukhdev> #link: https://wiki.openstack.org/wiki/Meetings/ML2#Meeting_March_2.2C_2016
16:01:58 <Sukhdev> It is a very brief agenda today - so, probably be a short meeting
16:02:04 <rkukura> +1
16:02:20 <Sukhdev> #topic: Announcements
16:02:30 <Sukhdev> M3 is later this week
16:03:01 <Sukhdev> Anybody has anything else to announce?
16:03:45 <Sukhdev> #topic: Modular L2 agent update
16:04:36 <Sukhdev> most of the code is already in - I reviewed the doc patch
16:04:45 <Sukhdev> https://review.openstack.org/#/c/282349
16:05:11 <Sukhdev> Andreas does not seem to be here foter any other upda
16:05:48 <Sukhdev> opps hit the wrong key - meant to say he is not here for any other update
16:06:29 <Sukhdev> have a look at his patch and provide any feedback
16:06:55 <Sukhdev> #topic: Routed networks
16:07:24 <Sukhdev> Looks like this was discussed during the mid-cycle - which I missed
16:07:40 <Sukhdev> it has moved into implementation phase
16:07:52 <Sukhdev> I do not have any further update -
16:08:01 <rkukura> spec hasn’t merged, so I think feedback is still possible
16:08:16 <Sukhdev> I was watching it to make sure that  there is not much impact on ML2
16:08:24 <rkukura> carl_baldwin uploaded a new patch on 3/1, I’m reviewing that one now
16:08:45 <Sukhdev> rkukura : yes - in fact, there is lot of feedback coming -
16:08:53 <rkukura> I’m still kind of uncomfortable changing the semantics of a neutron network to no longer be an L2 broadcast domain
16:08:59 <Sukhdev> so, feel free to provide your feedback
16:09:36 <rkukura> I can live with it being optionally an L2 broadcast domain, as long as that is clearly visible from the API
16:10:35 <Sukhdev> rkukura : you should review the patch and post the comments
16:10:44 <rkukura> Sukhdev: I am and will
16:11:06 <rkukura> Curious if others here have thought about the implications of this
16:11:45 <Sukhdev> there have been a lot of debate about it
16:12:17 <Sukhdev> in the previous mid-cycle, when it was introduced, there was quite a bit of dicussion
16:13:32 <Sukhdev> Anything else on this?
16:14:25 <Sukhdev> #topic: Open Discussion
16:14:48 <Sukhdev> I did not have anything else on the agenda
16:15:10 <Sukhdev> Any body wants to discuss anything today?
16:15:45 * Sukhdev waiting
16:16:02 <rkukura> Has anyone else seen any need for ML2 to support extension drivers and maybe mech driver calls on subnet pools and address scopes?
16:17:34 <Sukhdev> I guess silence means either no or no opinion :-)
16:17:38 <rkukura> guess so
16:18:10 <Sukhdev> or we do not have critical mass to answer your question :-)
16:18:48 <rkukura> my team is working on a driver that may need to track the association between address scopes and subnets, so we may be looking at adding support for this to ML2
16:19:31 <Sukhdev> what do you mean by address scope?
16:20:00 <rkukura> its a neutron core resource, added in liberty, I think
16:20:35 <yamamoto> rkukura: why do you need to track them?
16:20:52 <rkukura> by knowing that subnets are associated with the the same address scope, you know they don’t overlap and are potentially routable
16:21:00 <Sukhdev> does it have to do with carving out of the IP address space across sub-nets?
16:21:11 <rkukura> yes
16:21:29 <rkukura> subnets can belong to subnet pools which can belong to address scopes
16:22:15 <rkukura> anyone, just wanted to see if anyone else had thought about ML2 support for these
16:23:25 <rkukura> s/anyone/anyway/
16:23:33 <rkukura> that’s it from me
16:23:54 <Sukhdev> rkukura : so, at the time of create of sub-net, you do not specify the cidr, instead specify the address scope?
16:24:27 <rkukura> actually, the subnet pool
16:24:45 <rkukura> and the pool can be associated with an address scope
16:25:20 <Sukhdev> hmm...so these new APIs are already merged and available?
16:25:48 <rkukura> yes
16:25:52 <Sukhdev> so, create address scope, then sub-net pools, and then subnet
16:25:57 <rkukura> right
16:26:10 <rkukura> but the address scope attribute of subnet pool is mutable
16:26:30 <Sukhdev> I see
16:26:54 <yamamoto> rkukura: i guess a callback for SUBNETPOOL_ADDRESS_SCOPE is enough for many cases
16:27:30 <rkukura> right, although my preference would be to treat subnet pools and address scopes just like the other resources in ML2
16:29:18 <yamamoto> rkukura: i tend to agree but many folks seem to like callbacks recently.
16:29:37 <rkukura> the callbacks are not really usable without significant work
16:29:56 <rkukura> there are no callbacks for most resources, and no way to really see what is changed on updates
16:30:20 <rkukura> right now, they are just special cased when needed, rather than uniformily implemented
16:31:09 <Sukhdev> I guess the idea is to get the state of resource by query the DB
16:32:23 <Sukhdev> however, the question is who updates the DB, if multiple entities are trigged on callbak
16:33:07 <rkukura> lets wrap this up for today
16:33:22 <Sukhdev> +1
16:33:47 <yamamoto> sure
16:33:56 <Sukhdev> so, if nothing else on this, I good to call it a day
16:34:06 <rkukura> unless there are other topics anyone wants to discuss
16:34:15 <yamamoto> nothing from me.
16:34:26 <Sukhdev> OK - cool - we are done then
16:34:29 <Sukhdev> Thanks
16:34:31 <yamamoto> thank you!
16:34:33 <rkukura> thanks Sukhdev!
16:34:35 <Sukhdev> bye
16:34:39 <Sukhdev> #endmeeting