14:01:54 <rkukura> #startmeeting networking_ml2 14:01:55 <openstack> Meeting started Wed Aug 28 14:01:54 2013 UTC and is due to finish in 60 minutes. The chair is rkukura. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:56 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:59 <openstack> The meeting name has been set to 'networking_ml2' 14:02:18 <rkukura> #link https://wiki.openstack.org/wiki/Meetings/ML2 Agenda 14:02:53 <rkukura> mestery is on vacation, so I'll do my best to run the meeting 14:03:06 <rkukura> who do we have? 14:03:13 <rcurran> rcurran 14:03:16 <apech> apech 14:03:32 <sukhdev> sukhdev 14:04:13 <rkukura> #topic Blueprints 14:04:54 <rkukura> BP implementation proposal deadline was last Friday 14:05:06 <rkukura> BP implementation merge deadline is end-of-day next Tuesday, August 3rd 14:05:33 <rkukura> There were a couple of recent ML2 merges: 14:05:47 <rkukura> https://bugs.launchpad.net/neutron/+bug/1196963 14:05:49 <uvirtbot> Launchpad bug 1196963 in neutron "Update the OVS agent code to program tunnels using ports instead of tunnel IDs" [Wishlist,Fix committed] 14:06:02 <rkukura> https://review.openstack.org/#/c/43678/ 14:06:34 <rkukura> the above is: https://blueprints.launchpad.net/neutron/+spec/ml2-mechanism-driver-subnet-calls 14:07:48 <rkukura> So we now have the subnet-related calls on mechanism drivers, thanks to Zng MingJie 14:08:03 <apech> great! 14:08:35 <sukhdev> it seems it changes the context in the calls to mech drivers 14:08:45 <rkukura> This merge also added @property to the context class property calls 14:09:13 <rkukura> sukhdev: Right. Hopefully this isn't too much of a problem for the drivers currently in review. 14:09:47 <rkukura> It does require rebasing them, and removing the "()" from usage of the properties. 14:10:05 <sukhdev> sounds good 14:10:29 <rkukura> Anything else related to recent merges? 14:11:36 <rkukura> So we now have 5 BPs in review targeting H-3, and a few other things we should track 14:12:44 <rkukura> https://blueprints.launchpad.net/quantum/+spec/ml2-portbinding is in review at https://review.openstack.org/#/c/43129/ 14:13:45 <rkukura> I've gotten some good feedback from reviews, and will post an updated patch very soon today rebasing and addressing the nits 14:14:05 <rkukura> I'll need a bit more time to resolve a couple more significant items: 14:14:37 <rkukura> DB migration and duplication among the agent-based mech drivers were both identified in the reviews 14:15:24 <sukhdev> I created a bug to integrate with this after the merge 14:15:31 <rkukura> I also discovered at the last minute that the hyperv-agent does not yet implement the agents-db functionality, so it seems we should add that, or not claim to support hyperv 14:16:13 <rkukura> any thoughts on hyperv? 14:16:26 <rcurran> no 14:16:31 <sukhdev> no 14:17:05 <rkukura> I'm going to try to get one of its developers to test it with ml2. 14:17:46 <rkukura> Other main portbinding issue is how to get the host-id for ports of L3, DHCP, etc. agents? 14:18:27 <rkukura> Seems cleanest to have those agents set binding:host_id when they create/update their ports 14:18:35 <sukhdev> yes, we need host-id for dhcp 14:19:04 <sukhdev> presently, there is no host-id when a dhcp call is invoked 14:19:07 <rkukura> the alternative is for the RPC code to take the agent-id and look it up in the agents-db to find the agent's host 14:19:30 <rkukura> any preferences between these approaches, or other ideas? 14:19:41 <sukhdev> cleaner will be if it provided via port context to mech drivers 14:20:55 <rkukura> sukhdev: in either approach it would be in the port dictionary as binding:host_id 14:21:30 <rkukura> anything else on ml2-portbinding? 14:21:32 <sukhdev> ah OK - in that case I OK with either approach 14:22:06 <rkukura> OK 14:23:13 <rkukura> mestery posted implementation of https://blueprints.launchpad.net/neutron/+spec/ml2-multi-segment-api 14:23:27 <rkukura> https://review.openstack.org/#/c/43569/ 14:23:46 <rkukura> has anyone had a chance to look at this in detail yet? 14:24:25 <apech> not yet 14:24:49 <rcurran> no 14:25:12 <sukhdev> no 14:25:48 <rkukura> Lets try to get him some feedback before he's back online, which I think will be sometime later this week 14:26:12 <sukhdev> he is back on 9/3 14:26:12 <rkukura> Anything else on multi-segment-api? 14:26:29 <apech> general question - what's the cutoff date for all changes? 14:26:38 <rkukura> sukhdev: I got the impression he may be available to update this before then, but I'm not sure 14:26:40 <apech> we have a few significant blueprints coming in that will change our mechanism drivers 14:26:45 <apech> trying to understand our timeline 14:27:34 <rkukura> I believe the BP implementations need to be merged by end-of-day 9/3 - next Tuesday. 14:28:16 <rcurran> lot of balls in the air between now and then 14:28:32 <rkukura> There will still be opportunity for bug fixes, so if drivers needs additional updates to support multiple segments, that could be done after 9/3 14:28:40 <rkukura> rcurran: agreed 14:29:13 <apech> ok got it 14:29:35 <rkukura> but I don't think 5 merges is unreasonable with 3 of them being drivers that shouldn't effect other merges 14:29:55 <rcurran> glad to read to state that :-) 14:30:43 <rkukura> So lets discuss https://blueprints.launchpad.net/neutron/+spec/arista-ml2-mechanism-driver 14:30:58 <rkukura> lhttps://review.openstack.org/#/c/41213/ 14:31:11 <rkukura> link: https://review.openstack.org/#/c/41213/ 14:31:13 <sukhdev> I sent out an email yesterday in response to your comments 14:31:30 <rkukura> sukhdev: and I tried to summarize that in the agenda 14:31:34 <sukhdev> does that sound reasonable? 14:31:43 <rkukura> I think so 14:32:05 <rkukura> We need to document that single physical network limitation in the README and wiki 14:32:40 <sukhdev> Did you look at the wiki, I mentioned this in the overview section 14:32:51 <sukhdev> will mention in the readme as well 14:33:04 <rkukura> I took a quick look - must have missed it - sorry 14:33:09 <rkukura> great 14:34:02 <sukhdev> So, I will upload a new patch which will rebase based upon the latest merge and new readme file - but, no other changes - are you OK with that? 14:34:09 <rkukura> I think the redundant DB state is justified for now, given that there is no reasonable way to sync state without it, but we should look for a good general solution for this in icehouse 14:34:34 <sukhdev> I agree 14:34:50 <rkukura> sukhdev: I'm OK with that, as long as all review issues have been addressed 14:35:14 <sukhdev> I have addressed all other issues 14:35:31 <sukhdev> unless - a new one pops up :-) 14:35:40 <rkukura> sukhdev: It woudn't hurt to include some of the email text in the review so others can see 14:36:02 <rkukura> Seems the Arista driver is on track to merge - anything else on this? 14:36:21 <rkukura> Next is the Cisco Nexus driver 14:36:33 <rkukura> link: https://blueprints.launchpad.net/quantum/+spec/ml2-md-cisco-nexus 14:36:41 <rcurran> 2nd patch went up friday 14:36:43 <rkukura> link: https://review.openstack.org/#/c/43196/ 14:36:51 <rcurran> still had an issue w/ one of the UTs 14:37:20 <rcurran> i think this has been fixed (rkukura: has dane leblanc reached out to you?) 14:37:40 <rkukura> rcurran: not that I recall 14:38:06 <rcurran> ok, that may be coming ... he was having an issue that may require an update to core ml2 code 14:38:27 <rcurran> regarding db duplication 14:38:30 <rkukura> rcurran: is that related to the UT failure? 14:38:40 <rcurran> yes 14:38:52 <rcurran> dane is working on ml2 cisco nexus ut 14:39:24 <rkukura> OK, so hopefully we can work through this quickly and get an updated patch set 14:39:31 <rcurran> yes 14:39:34 <rcurran> db duplication 14:40:11 <rcurran> you'll see this in the ml2 cisco nexus driver also. this is because the code is mostly a port from the current plugins/cisco 14:40:37 <rcurran> i think we decided a while ago that plugins/ml2 drivers shouldn't import modules from another plugin 14:41:12 <rkukura> rcurran: OK, so this is code duplication with the cisco plugin? 14:41:19 <rcurran> but because cisco doesn't want to supprot duplicate code we'll thinking in the icehouse timeframe to have the cisco plugin import the ml2 code 14:41:54 <rcurran> eventurally (if all goes as planned) the plugin/cisco code gets deprecated and we only have ml2 left 14:42:17 <rcurran> we can then clean up the duplicated db stuff 14:42:37 <rkukura> sounds good to me - a README on the cisco driver would be helpful, especially setting expectations relative to the cisco plugin 14:42:43 <rkukura> README/wiki 14:42:50 <rcurran> yeah, i saw that comment coming :-) 14:43:17 <rkukura> I'll definitely review the upcoming patch 14:43:22 <rkukura> anything else on this one? 14:43:57 <rkukura> So Tail-f is next 14:44:14 <rkukura> link: https://blueprints.launchpad.net/quantum/+spec/tailf-ncs 14:44:31 <rkukura> link: https://review.openstack.org/#/c/37647/ 14:46:12 <rkukura> Is Luke here? 14:48:16 <rkukura> This needs to be rebased and a couple nits fixed 14:48:57 <rkukura> I haven't looked this too closely, but will. It seems to be on track 14:49:06 <rkukura> anything else on tail-f? 14:50:02 <rkukura> has anyone been tracking the PXE work? 14:50:20 <rkukura> link: https://blueprints.launchpad.net/neutron/+spec/pxeboot-ports 14:50:40 <rkukura> link: https://review.openstack.org/#/c/30441/ 14:51:15 <rkukura> This adds an extension to openvswitch, and I had commented a while back that it should address ml2 as well 14:51:30 <rkukura> dkehn: Are you online? 14:51:39 <dkehn> yes 14:52:05 <rkukura> Are you planning to add ml2 support to the PXE boot patch? 14:52:35 <dkehn> yes, agreed on the ml2, what markmcclain and others have agreed upon is after the patch lands we will add ml2 support in 14:53:05 <dkehn> https://review.openstack.org/#/c/30441/, see lifeless comments 14:53:11 <rkukura> OK, I guess we can treat that as a bug fix and git it in H-3 14:53:58 <rkukura> Looks like the plan is to do that in icehouse 14:54:13 <dkehn> correct 14:56:29 <rkukura> I guess that will do, but we've been trying to keep parity with openvswitch so it can be deprecated 14:56:52 <rkukura> dkehn: I may followup with you offline 14:57:09 <dkehn> rkukura, sounds good 14:57:34 <rkukura> Last BP items are devstack and documentation. Any updates on these? 14:59:01 <apech> I had volunteered to help, mestery was going to reach but I didn't hear anything 14:59:06 <apech> reach out that is 14:59:34 <rkukura> Lets focus on those, plus any bugs, next week then 15:00:15 <rkukura> feleouet doesn't seem to be online - anyone have any update on l2-population? 15:00:39 <safchain> Hi 15:00:51 <safchain> on my side I'll submit a new patch for l2population, with an implementation as Mechanism Driver 15:01:23 <rkukura> safchain: I assume this will be targeting icehouse, right? 15:01:57 <rkukura> Out time is up. Thanks everyone! 15:02:14 <safchain> probably, but I'll do my best for H5 15:02:16 <safchain> H3 15:02:17 <sukhdev> thanks 15:02:51 <rkukura> OK, lets get these reviews and merges done 15:02:55 <rkukura> #endmeeting