15:03:41 <carl_baldwin> #startmeeting neutron_routed_networks
15:03:42 <openstack> Meeting started Tue Jun 21 15:03:41 2016 UTC and is due to finish in 60 minutes.  The chair is carl_baldwin. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:03:43 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:03:45 <blogan> hi
15:03:45 <openstack> The meeting name has been set to 'neutron_routed_networks'
15:03:57 <carl_baldwin> #topic Open Agenda
15:04:03 <carl_baldwin> We'll just start with this.
15:04:12 <john-davidge> I'd just like to draw attention to #link https://review.openstack.org/#/c/300207/ - we need it to merge ASAP
15:04:21 <carl_baldwin> First, look over the etherpad and gerrit topic for reviews.
15:04:31 <carl_baldwin> #link https://etherpad.openstack.org/p/routed-provider-networks-notes
15:04:54 <carl_baldwin> #link https://review.openstack.org/#/q/topic:bp/routed-networks+status:open
15:06:08 <carl_baldwin> john-davidge: I'll ping kevinbenton again to have another pass.
15:06:27 <john-davidge> carl_baldwin: Thanks :)
15:06:54 <carl_baldwin> I'm still working on Nova side.  I will commit to having some kind of Nova patch up this week to allow deferred IP alloc.
15:07:25 <carl_baldwin> blogan: how is DHCP going?
15:07:44 <blogan> had the week off last week, going to push up a new patch today
15:07:54 <blogan> though one question
15:08:33 <carl_baldwin> blogan: What's that?
15:08:54 <blogan> kevinbenton suggested just scheduling all segments in a network, instead of just the one segment affected that the code is currently doing
15:09:37 <blogan> does anyone feel strongly for or against that method?
15:10:14 <xiaohhui> I think scheduling to one segment one time is fine.
15:10:32 <xiaohhui> After all, we are adding one segment-subnet one time
15:11:03 <blogan> xiaohhui: true but it wouldn't hurt to attempt to do them all, and really it'll probably just be one that gets scheduled 99% of the time anyway if thats always done
15:11:27 <carl_baldwin> blogan: I don't feel strongly.
15:11:43 <john-davidge> blogan: Did kevin highlight a particualr advantage to doing it that way?
15:12:22 <blogan> i believe the advantage would be that we wouldn't have to pass specific segment information down to the scheduler from the notification code
15:12:54 <carl_baldwin> blogan: That could simplify things a bit.
15:13:02 <blogan> currently its being packed into the network dictionary so that the scheduler knows what segment is actually needing scheduling
15:13:38 <xiaohhui> I am adding candidates as parameter in schedule method in this patch for refactoring l3-agent-scheduler.
15:13:49 <xiaohhui> https://review.openstack.org/#/c/320347/
15:13:49 <john-davidge> I'm not sure I can think of any obvious downsides of scheduling them all at once
15:14:03 <john-davidge> So it seems fine to me
15:14:28 <blogan> xiaohhui: could the same thing be done in your case?
15:15:09 <xiaohhui> I am thinking it is the same thing to pass candidates as parameter for segment scheduling.
15:16:10 <xiaohhui> So, it that patch is merged, I think we can avoid pullute network dict
15:16:15 <xiaohhui> it -> if
15:16:35 <blogan> i thought there was a reason we didn't want to change the base scheduler's arguments like that, it breaking the interface
15:17:13 <xiaohhui> May be we should discuss in the review of that patch...
15:17:22 <blogan> xiaohhui: +1
15:17:48 <carl_baldwin> blogan: Let us know when you get your update posted.
15:17:59 <blogan> carl_baldwin: will do
15:18:39 <carl_baldwin> Anything else to bring up?
15:20:33 <carl_baldwin> Please review and update the etherpad regularly.
15:20:53 <carl_baldwin> Going once...
15:21:39 <carl_baldwin> Going twice...
15:21:59 <carl_baldwin> Thanks!
15:22:06 <carl_baldwin> #endmeeting