16:03:14 <rkukura> #startmeeting networking_ml2
16:03:14 <openstack> Meeting started Wed Dec  7 16:03:14 2016 UTC and is due to finish in 60 minutes.  The chair is rkukura. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:03:15 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:03:18 <openstack> The meeting name has been set to 'networking_ml2'
16:03:29 <rkukura> #topic Agenda
16:03:42 <rkukura> #link https://wiki.openstack.org/wiki/Meetings/ML2#Meeting_December_7.2C_2016
16:04:03 <rkukura> same agenda as we’ve been using for a while
16:04:16 <rkukura> Anything to add or remove?
16:04:44 <yamamoto> no updates from me
16:05:05 <rkukura> #topic Announcements
16:05:54 <rkukura> Reminder that PTG (Project Teams Gathering) registration is open - around 300 slots remaining
16:06:02 <rkukura> #link http://www.openstack.org/ptg
16:06:27 <Sukhdev_> I already registered
16:06:48 <Sukhdev_> I will be attending
16:06:50 <rkukura> I believe Sukhdev_ said he is going, and I’m likely to go, so its an opportunity to discuss ML2 work
16:07:07 <yamamoto> i'm going
16:07:14 <rkukura> great!
16:07:43 <rkukura> Any other announcements?
16:08:18 <rkukura> ok…
16:08:29 <rkukura> #topic Bugs/RFEs
16:09:04 <rkukura> Does andreas_s or brianstajkowski have any update on the nova live migration work?
16:09:19 <brianstajkowski> yes, so the spec is close
16:09:31 <brianstajkowski> #link https://review.openstack.org/#/c/309416
16:10:07 <brianstajkowski> waiting on our testing in wip patch #link https://review.openstack.org/#/c/404293/
16:10:33 <brianstajkowski> to finalize what we are doing with portbinding and distributedportbinding tables
16:10:46 <Sukhdev_> I reviewed the spec - it looks good
16:10:47 <brianstajkowski> in that same patch we completed the sub resource work, so that's complete
16:11:15 <rkukura> brianstajkowski: Do you mean exposing the bindings as sub resources of the port in the API?
16:11:25 <brianstajkowski> yes
16:11:39 <rkukura> great!
16:11:53 <rkukura> Any issues or concerns to discuss?
16:12:03 <brianstajkowski> in the spec line 638 I made some assumptions on how the activate would work, need some feedback there
16:12:44 <brianstajkowski> mainly trying to figure out a way to make the switchover a bit seemless although there will be slight interruptions in traffic
16:13:12 <rkukura> brianstajkowski: Is this active RPC from the L2 agent, or from nova?
16:13:17 <brianstajkowski> the only way to get around this is for the source migration host to send out a spoofed GARP, while the target migration host is plugging the vif maybe?
16:13:22 <brianstajkowski> yes
16:13:34 <brianstajkowski> nova, this is the activate call
16:13:42 <brianstajkowski> which will trigger an rpc to plug the vif
16:13:55 <brianstajkowski> on the target host and unplug on the source host
16:14:24 <rkukura> Is this a AMQP RPC, or a REST call?
16:14:52 <brianstajkowski> so REST from nova to activate, then this might trigger an AMQP RPC to the source and target to plug unplug their vifs
16:15:20 <brianstajkowski> maybe i'm off base here but just looking for feedback in the spec
16:15:21 <rkukura> OK. I usually think AMQP when I see “RPC”
16:15:31 <brianstajkowski> so do i lol
16:16:30 <rkukura> Originally, the nova/neutron interactions were entirely REST, but we did start using notifications, so I was wondering if AMQP RPCs were now on the table?
16:16:43 <brianstajkowski> not from nova
16:16:54 <rkukura> OK
16:17:00 <brianstajkowski> rest to neutron api, rpc from neutron api to agents etc
16:17:07 <rkukura> OK
16:17:27 <rkukura> Any concerns with this model supporting agentless ML2 mechanism drivers?
16:17:52 <brianstajkowski> in that case possibly, i'll go back on that one
16:19:04 <Sukhdev_> I think this should be OK for agent less drivers
16:19:31 <rkukura> OK, I clearly need to review the current spec before asking too many more questions
16:19:34 <Sukhdev_> as long as these are new calls and old ones are unchanged - this should be fine
16:20:12 <brianstajkowski> ok, good to know, i'll go through it as well, but yes line 638 is the rough area that needs some feedback
16:20:32 <rkukura> anything else on this?
16:20:36 <brianstajkowski> no thank you
16:20:44 <rkukura> thanks brianstajkowski!
16:21:09 <rkukura> Are there any other bugs/RFEs anyone would like to discuss?
16:21:50 <rkukura> #topic Driver API Feature Requests
16:22:38 <rkukura> I don’t see any new developments on https://bugs.launchpad.net/neutron/+bug/1583096
16:22:38 <openstack> Launchpad bug 1583096 in neutron "ml2 supported extensions list is inaccurate" [Medium,Confirmed]
16:23:31 <yamamoto> i guess we should start from something simple.
16:25:38 <yamamoto> eg. report a set of extensions which all configured drivers support
16:26:20 <rkukura> yamamoto: I think we need to address two categories of extensions - those built into the ML2 core plugin, and those implemented as ML2 extension drivers
16:27:33 <rkukura> Are you suggesting that if an extension built into the ML2 core plugin is not supported by all registered MDs, it should not be enabled in the API (and advertized in the extensions list)?
16:27:49 <yamamoto> yes
16:28:11 <rkukura> I’m not sure I would agree
16:28:22 <yamamoto> do we have any extension implemented purely in extension driver?
16:29:02 <rkukura> I see the API as advertizing everything that can be done, but port binding determining what actually gets done
16:29:14 <rkukura> Isn’t QoS in an ED?
16:29:57 <yamamoto> qos needs notification drivers, which is usually 1:1 with mech drivers
16:30:00 <rkukura> And I know there are 3rd party MDs that have corresponding EDs that implement extensions
16:31:27 <yamamoto> for example?
16:33:11 <rkukura> yamamoto: New drivers being developed for Cisco APIC expose the APIC resources to which Neutron resources are mapped via an extension (normally admin-only)
16:34:23 <yamamoto> a problem of port-binding time decision is we don't have a good way to report a failure to api users.
16:35:04 <rkukura> And that driver can be used via HPB with other drivers that know nothing about APIC or its extension
16:35:34 <yamamoto> interesting. thank you
16:35:47 <rkukura> I suggest we try to have an in-depth discussion of how extensions are advertized and enforced at the PTG
16:36:12 <rkukura> And come up with a plan if we are going to change anything
16:36:20 <Sukhdev_> +1
16:36:36 <rkukura> I think there are several valid but conflicting views on this
16:37:04 <rkukura> lets move on…
16:37:10 <yamamoto> sure
16:37:40 <rkukura> There is a patch in review for https://bugs.launchpad.net/neutron/+bug/1587401
16:37:40 <openstack> Launchpad bug 1587401 in neutron "Helper method to change status of port to abnormal state is needed in ml2." [Wishlist,In progress] - Assigned to Trevor McCasland (twm2016)
16:37:52 <rkukura> Its https://review.openstack.org/#/c/336068/
16:38:27 <rkukura> Looks like no updates in the past month or so
16:39:13 <yamamoto> i think the patch itself is fine.  multiple MD concern is still valid though.
16:40:07 <rkukura> I see I had one minor comment drafted, but never completed my review
16:41:18 <rkukura> Any other dicussion of driver api feature requests?
16:41:40 <rkukura> if not…
16:41:46 <rkukura> #topic Open Discussion
16:42:05 <rkukura> Any other items to discuss today?
16:42:08 <Sukhdev_> radhikam : are you here?
16:43:01 <Sukhdev_> I guess we are done
16:43:36 <Sukhdev_> rkukura : nothing from me
16:43:38 <rkukura> Sukhdev_: I assume radhikam is the person you said pinged you?
16:43:45 <Sukhdev_> yup
16:43:49 <radhikam> I am here
16:43:57 <rkukura> hi radhikam!
16:44:25 <radhikam> hi rkukura
16:44:28 <Sukhdev_> radhikam : this is the forum to ask any questions you may still have
16:44:43 <andreas_s> brianstajkowski, FYI: just commented on your spec
16:45:02 <Sukhdev_> andreas_s : hey stranger!!!
16:45:36 <andreas_s> Sukhdev_, hehe
16:46:05 <Sukhdev_> andreas_s : are you going to PTG in Atlanta?
16:46:31 <radhikam> Thanks Sukhdev! I am going over the information you gave me yesterday. Will bring up any questions I have in this meeting next time.
16:46:43 <andreas_s> Sukhdev_, no,  I'm not planning to attend...
16:46:53 <Sukhdev_> radhikam : sounds good -
16:47:27 <rkukura> Anything else today, or can we wrap up a bit early?
16:47:28 <Sukhdev_> andreas_s : rkukura, yamamoto, and I plan on going - you should come along - we can have detailed ML2 discussions
16:48:30 <rkukura> OK, thanks everyone!
16:48:44 <rkukura> #endmeeting