16:03:00 #startmeeting networking_ml2 16:03:01 Meeting started Wed May 4 16:03:00 2016 UTC and is due to finish in 60 minutes. The chair is rkukura. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:03:02 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:03:05 The meeting name has been set to 'networking_ml2' 16:03:31 Hope everyone enjoyed the summit and Austin! 16:03:44 #topic Agenda 16:03:53 #link https://wiki.openstack.org/wiki/Meetings/ML2#Meeting_May_4.2C_2016 16:04:36 relatively light agenda today, kind of open ended 16:04:47 anyone have anything specific they’d like to add? 16:05:17 #topic Announcements 16:05:19 kind of short week - we were in a meeting Friday :-) 16:05:34 I don’t have any announcements - does anyone else? 16:06:06 Friday session was very useful - we got few conclusions 16:06:39 Sukhdev: “few” as in “a few”, or as in “very few”? 16:07:06 ha ha - you are getting technical now :-) 16:07:38 I will agree it was useful 16:07:55 #topic Newton ML2 Charter 16:08:47 So Sukhdev and I would both like the ML2 subteam to figure out what we’d like to accomplish during Newton, so we can make progress, and justify the time we spend at these meetings ;) 16:09:14 Sukhdev: Did you want to add anything motivational here? 16:09:32 rkukura: we are already motivated :) 16:10:04 well, I do not have any motivational speech :-) 16:10:35 I am interested in the vlan trunking feature that was discussed during the joint session between neutron and ironic 16:10:40 perhaps we can brainstorm as to what we believe will be useful to cover during Newton cycyle 16:10:58 sadasu: That’s definitely something I’m interested in as well 16:11:12 not sure if this would be considered ml2 specific, but would be useful for ml2 drivers 16:11:33 sadasu : in that case you should join us on Mondays at the same time slot as this meeting 16:11:41 I’ve got my usual wish list for ML2 16:12:10 Sukhdev: ok will do 16:12:27 sadasu : we run weekly meeting meeting on Monday's to cover Ironic/Neutorn integrarion and we have been driving that initiative there 16:12:40 it takes place on #openstack-meeting-4 16:12:48 I’d like to see this team focus on trunk ports in general, covering L2 agent solutions, ironic, network service appliances, etc. 16:12:53 Sukhdev: thanks! 16:13:48 For example, ToR MDs will need to understand trunk ports even if the L2 agent is trunking them together, right? 16:14:08 rkukura : I think that is a good list - add to it Security, and Manila integration as welll 16:14:27 rkukura : +1 16:15:01 rkukura: The SR-IOV based MDs would want that too. the UCSM MD included 16:15:16 sadasu: good point 16:16:41 So I think we have cases where the VLAN tags can be set by the client (L2 agent), and cases where the VLAN tags need to be those used in the infrastructure (ironic, appliances, maybe SR-IOV). Is that accurate? 16:16:53 By 16:16:56 By 16:17:14 rkukura: +1 16:17:15 By “set by the client”, I mean specified using the API vs getting whatever you get 16:17:43 right 16:18:24 So it sounds like trunk ports should be part of our charter for Newton 16:18:35 rkukura : actually, you mean tag translation 16:19:04 Sukhdev: Yes, the L2 agent can translate (or apply) tags, but in other cases there may be no means to translate them. 16:19:26 rkukura : So, spec for this on the neutron side is already approved and merged 16:19:54 rossella_s is driving that charter 16:20:24 the spec on the ironic side needs to be updated based upon our discussion on Friday 16:20:51 Sukhdev: Is rossella_s covering the API in cases where there is no L2 agent that can translate tags? 16:20:53 Sam mentioned he will update and post it 16:21:21 yes - that is what we covered on Friday 16:21:29 ok will look for Sam's updated specs 16:21:37 we did not use L2 agent case, but, used the TOR case - 16:21:51 but, in general, yes 16:22:15 OK, what else do we want to accomplish in ML2 during Newton? 16:22:51 rkukura : API/visibility into the segments of a multi-segmented networks 16:23:22 rkukura : I filed an RFE earlier, but, based upon Friday's discussion, I will update it 16:23:35 will do it either later today or tomorrow - 16:23:45 Sukhdev: I agree, and I think the routed networks spec is already covering most of this, right? What else is needed beyond what they are doing? 16:24:30 rkukura : for manila use-case, they need to know to which segment a specific port is bound to 16:25:17 That’s different - that is the portbinding result. I do agree we want to make that visible, but Manila will still need some way to make port binding happen. 16:25:21 so, we need some generic API/mechanism in neutron to provide that information 16:25:39 Some sort of MD that calls set_binding() for manila ports 16:26:19 rkukura: that should be done by vendor's MDs, e.g. Cisco, Arista, etc.. 16:26:59 And what if there is no vendor ToR switch involved? Shouldn’t something bind for a plain VLAN-based network? 16:28:24 rkukura : no - that will not be valid deployment scenario (I think) as the storage device has to hang off from a TOR 16:29:08 For a simple pure open source PoC (or CI setup), an unmanaged switch might be reasonable. 16:29:48 Perhaps 16:30:29 rkukura : I am assuming they would have something like that already 16:30:32 Not something we need to solve right here - the Manila team will need to decide what’s important for their specific use case. 16:30:38 similar to Ironic 16:31:57 So I think the Newton charter could include both helping get Manila integrated with more complex network deployments, and making port binding results visible via the API, with possibly the same actual solution applying to both 16:32:51 The RFE that I will be submitting will cover the visibility part for now 16:33:04 Sukhdev: Sounds good! 16:33:12 There is also the live migration improvement that scheuran has been investigating. Should we include anything on that in the charter? 16:35:37 rkukura: is there a spec out for that already? 16:36:20 sadasu: I don’t thihk so. He was looking into whether the generalized distributed port stuff I had been working on might be applicable. 16:37:09 So lets wait and see if he comes up with something that can be accomplished in Newton. 16:38:47 There are a coule items on my wishlist that we might tackle: enforcing extension semantics, consistent driver access to all ML2-implemented resources, and coelescing status from multiple MDs 16:38:58 s/coule/couple/ 16:42:15 Anything else anyone would like to accomplish in Newton for ML2? 16:43:12 If not, lets move onto the next agenda item 16:43:28 #topic Bugs 16:44:03 So there is one regarding missing options in config reference 16:44:19 #link https://bugs.launchpad.net/openstack-manuals/+bug/1567792 16:44:21 Launchpad bug 1567792 in openstack-manuals "Mitaka Configuration Reference Nework section is missing ML2 parameters" [Undecided,Confirmed] 16:45:30 Is this only 3rd party (i.e. vendor) specific config options? 16:47:14 schueran filed this, so we can cover it next time he’s here if its not resolved, assigned, or invalidated 16:47:24 Any other bugs to discuss? 16:47:55 #topic Open Discussion 16:48:16 Anything else to discuss, or should we wrap up for today? 16:49:04 Seems we are done 16:49:08 Thanks everyone! 16:49:10 nothing from me 16:49:20 Thanks rkukura 16:49:26 #endmeeting