15:00:55 <apuimedo> #startmeeting kuryr
15:00:56 <openstack> Meeting started Mon Nov  9 15:00:55 2015 UTC and is due to finish in 60 minutes.  The chair is apuimedo. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:57 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:00 <openstack> The meeting name has been set to 'kuryr'
15:01:19 <apuimedo> Hello everybody and welcome to the first Mitaka cycle Kuryr meeting!
15:01:37 <apuimedo> who's up for a quick meeting?
15:01:43 <tfukushima> o/
15:01:53 <banix> hi
15:02:27 <banix> sounds good :)
15:03:03 <gsagie> Sorry :)
15:03:08 <gsagie> thought its in an hour as clock changed
15:03:16 <gsagie> hi everyone :)
15:03:38 <apuimedo> #info only cores present today. tfukushima, banix, gsagie and apuimedo
15:04:30 <apuimedo> I'm glad you could make it to the meeting
15:04:57 <apuimedo> and I want to start by saying that the Friday morning meetup was very enjoyable. Thanks ;-)
15:05:16 <gsagie> yep, was great to meet everyone
15:05:40 <apuimedo> I also want to apologize for not having posted an agenda
15:05:52 <apuimedo> basically I'd like to keep following up on the work items
15:06:13 <apuimedo> gsagie is working on having devstack working
15:06:46 <gsagie> yeah, basically it will need to also install a default k/v store, as Docker engine installation doesnt come with a default one and its not working otherwise
15:06:52 <apuimedo> #topic devstack
15:07:18 <gsagie> i hope to send something out tommorow but also will need the OVS binding part, as the example local.conf is going to be with the reference implementation
15:07:38 <apuimedo> #info devstack needs to set un a default k/v store and set up Docker to use it.
15:07:50 <apuimedo> gsagie: I guess you will do it with consul, right?
15:08:07 <gsagie> apuimedo: actually etcd is simpler for me
15:08:13 <apuimedo> fine
15:08:17 <gsagie> but we can later add other options
15:08:32 <apuimedo> #info: gsagie will set the default k/v store to be etcd
15:08:49 <apuimedo> any objection from the tfukushima or banix on this point?
15:09:09 <gsagie> and of course to test it fully i will need the OVS binding part merged
15:09:22 <gsagie> is diga here?
15:09:30 <banix> no not really; haven’t used it with Kuryr but I am sure it will be fine
15:09:36 <tfukushima> apuimedo: The key-value store should be transparent and anything can be fine for me.
15:09:45 <apuimedo> perfect, etcd it is then
15:09:55 <apuimedo> #topic bindings
15:09:58 <gsagie> tfukushima: yeah the plan is that it can be changed by configuration in the local.conf
15:10:13 <gsagie> so we can add consul quite easily
15:10:41 <apuimedo> banix: I assume you have an ovs binding since you demoed it :-)
15:10:49 <banix> :)
15:11:04 <banix> yes I do; I offered to help diga as he has a patch on it
15:11:17 <apuimedo> banix: that's what I wanted to ask
15:11:25 <banix> I haven’t heard back though
15:11:33 <apuimedo> could you please help him get it this week
15:11:44 <banix> Yes I will reach out to him directly
15:11:52 <apuimedo> if he does not respond I guess we'll have to move on
15:12:06 <banix> yes sure
15:12:16 <apuimedo> and use either feiskyer abandoned patch or the work you used for the demo
15:12:23 <apuimedo> nice
15:12:32 <tfukushima> Would it be better to have the deadline?
15:12:50 <apuimedo> tfukushima: I'm not sure where that'd be posted
15:12:51 <gsagie> i think we really only need the script and maybe pass the MAC (and optionally bridge)
15:12:53 <gsagie> in the binding
15:13:06 <gsagie> is there anything else?
15:13:08 <apuimedo> gsagie: I'm wondering when I'll see a Dragonflow binding
15:13:19 <gsagie> apuimedo: the OVS one will fit that too :)
15:13:25 <apuimedo> ah, cool
15:14:00 <apuimedo> about the extra parameters
15:14:05 <apuimedo> that ovs binding needs
15:14:09 <banix> gsagie: I think we want to have the option of having the hybrid OVS as well, for security groups. or may be that is the only thing we need?
15:14:38 <gsagie> not only, because OVN for example doesnt need it
15:14:43 <gsagie> but yeah thats a good point
15:15:05 <apuimedo> banix: gsagie: I think that we don't need the hybrid ovs
15:15:11 <apuimedo> if
15:15:13 <apuimedo> of course
15:15:30 <gsagie> apuimedo: its needed for the reference implementation, if i understood correctly what you mean by "hybrid ovs"
15:15:32 <apuimedo> neutron really manages to move ovs to use conntrack for security groups
15:15:53 <gsagie> apuimedo: yeah, but not sure when this will be merged, dont think its planned for Mitaka
15:15:59 <banix> apuimedo: well, is that going to happen soon?
15:16:02 <apuimedo> oh. I thought it was
15:16:10 <apuimedo> well. Let's figure it out
15:16:14 <apuimedo> and if it isn't planned
15:16:18 <banix> yup
15:16:29 <apuimedo> we have to do the funky linux bridge binding, of course :-)
15:17:13 <banix> apuimedo: I agree
15:17:18 <gsagie> you can add an action on me to find out when its planned
15:17:22 <apuimedo> #action gsagie to check conntrack for security groups in the mitaka cycle
15:17:29 <apuimedo> thanks gsagie
15:17:56 <apuimedo> #action banix to work with diga on the ovs binding (with the hybrid as follow up patches if necessary)
15:18:39 <apuimedo> vikasc_: hi! Joining the meeting?
15:18:47 <vikasc_> apuimedo, hi
15:18:56 <vikasc_> sorry for being late
15:19:06 <apuimedo> no problem, glad you made it
15:19:15 <apuimedo> #topic multi-node
15:19:45 <apuimedo> #info vikasc_, banix and tfukushima have worked on this
15:19:50 <apuimedo> what's the current state?
15:20:42 <tfukushima> I confirmed I could ping from a host to another host with my setting where I have Docker 1.9.0 release.
15:21:29 <apuimedo> tfukushima: master or some other patch necessary?
15:21:55 <tfukushima> We need Vikas's validation patches.
15:22:20 <tfukushima> Then, it's basically the same as master.
15:22:21 <gsagie> can you describe how the multi node is working?
15:23:04 <tfukushima> Actually I tried with MidoNet and it's specific to it but I just launched two Docker instances in the different hosts.
15:23:25 <apuimedo> #action apuimedo to review validation vikasc patches
15:23:52 <apuimedo> banix did try it with ovs though, right?
15:23:55 <gsagie> what do you mean specific to Midonet?
15:23:57 <apuimedo> or was it ovn?
15:24:05 <gsagie> it needs to communicate with the same Neutron right?
15:24:07 <banix> apuimedo: ovs
15:24:10 <tfukushima> And I just pinged. Before that, I needed to configure them to be in the same tunnel zone, which is the concept specific to MidoNet.
15:24:13 <apuimedo> do you mean the tunneling between hosts, tfukushima ?
15:24:38 <tfukushima> Yes, GRE andVXLAN.
15:25:13 <apuimedo> oh well... I would appreciate if we put some docs in the repo for each binding/vendor that we support
15:25:19 <apuimedo> explaining how to set it up
15:25:20 <gsagie> I am more interested how we do it in Kuryr side, given that there is a Neutron solution deployed at both hosts what is needed in Kuryr?
15:25:55 <apuimedo> gsagie: kuryr had to be configured in global mode, which vikasc rightly spotted
15:25:56 <gsagie> apuimedo: isnt the tunneling just a Neutron specific implementation?
15:26:06 <apuimedo> and docker had to point to the kv in both machines
15:26:12 <gsagie> it needs to work regardless if we have containers or not
15:26:13 <banix> gsagie: with OVS, nothing beyond the binding is needed on the Kuryr side
15:26:21 <apuimedo> gsagie: it is, that's why tfukushima said it was a midonet thing that he had to address
15:26:57 <gsagie> ok, and just to understand we running one kuryr server in this case?
15:27:17 <apuimedo> gsagie: no. Always one per host
15:27:24 <vikasc_> apuimedo, true
15:28:08 <gsagie> ok, so you added code that knows to recognize the same network for example and not re-create it in Neutron and so on in order to implement this
15:28:22 <apuimedo> gsagie: that's handled by docker
15:28:31 <apuimedo> recently
15:28:44 <apuimedo> before Taku had to make a patch to prevent the re-creation
15:28:53 <apuimedo> but they fixed the propagation
15:29:29 <tfukushima> Instead, the networks are shared through the distributed datastore.
15:29:30 <gsagie> ahh ok, so basically its mostly testing ?
15:29:37 <gsagie> in Kuryr part
15:30:21 <vikasc_> tfukushima, currently we running independent neutron on each node?
15:30:36 <gsagie> it has to be the same neutron
15:30:39 <banix> vikasc_: no one neutron service
15:30:45 <apuimedo> #topic open discussion vikasc_ you could, but for most deployments a single neutron server would be the norm
15:30:52 <apuimedo> (you can put more behind ha)
15:31:00 <apuimedo> oops
15:31:05 <gsagie> :)
15:31:06 <apuimedo> wrong command
15:31:06 <tfukushima> vikasc_: No. Every Kuryr agent points to the same Neutron API in my setting.
15:31:21 <apuimedo> #topic still multi-node
15:31:23 <vikasc_> tfukushima, got it
15:31:55 <apuimedo> ok. let's move on
15:32:01 <gsagie> k
15:32:16 <apuimedo> tfukushima: did you give any look at connecting to existing nets?
15:32:29 <apuimedo> (existing neutron nets)
15:32:36 <tfukushima> apuimedo: No, not yet.
15:32:43 <apuimedo> ok ;-)
15:32:59 <tfukushima> Before that, we need to solve the IPAM and the API change in 1.9.0.
15:33:06 <apuimedo> true
15:33:21 <tfukushima> That's a huge gap for now.
15:33:22 <apuimedo> #topic stabilization
15:33:27 <vikasc_> i am working on null ipam driver , will be pushing for review soon
15:33:32 <apuimedo> tfukushima: please, list the missing parts
15:33:46 <apuimedo> #info vikasc_ is working on a null ipam driver
15:33:51 <apuimedo> thanks vikasc_ !
15:34:15 <tfukushima> 1. IPv4Data and IPv6Data in the request of /NetworkDriver.CreateNetwork
15:34:40 <banix> vikasc_: does that make it possible to use kuryr as is?
15:34:48 <tfukushima> 2. Doing something for IPAM driver (Vikas is already working probably)
15:34:49 <vikasc_> yes banix
15:35:00 <banix> vikasc_: great
15:35:05 <vikasc_> yes
15:35:24 <vikasc_> our existing changes will make sense
15:35:26 <tfukushima> 3. /NetworkDriver.DiscoverNew and /NetworkDriver.DiscoverDelete (banix is looking at them?)
15:35:41 <banix> tfukushima: yes
15:36:03 <tfukushima> I'm working on 1.  now.  It's taking some time because it's a big internal change.
15:36:40 <vikasc_> working on 2
15:37:11 <tfukushima> So I'm modifying Kuryr to allocate subnets on creating the Docker network instead of on creating endpoints.
15:37:18 <apuimedo> #info tfukushima working on IPv4Data and IPv6Data in /NetworkDriver.CreateNetwork requests
15:37:30 <apuimedo> tfukushima that's a welcome change :-)
15:38:01 <tfukushima> Yes indeed. But they should have that in the first place.
15:38:08 <apuimedo> yeah
15:38:39 <vikasc_> we have to dance to their(docker) tunes :)
15:38:53 <apuimedo> #info banix working on /NetworkDriver.DiscoverNew and /NetworkDriver.DiscoverDelete
15:39:02 <apuimedo> vikasc_ indeed
15:39:29 <apuimedo> #topic open discussion
15:39:41 <apuimedo> does anybody have another topic to bring up?
15:40:05 <apuimedo> remember that next week we run the meeting in the alternate time
15:40:10 <vikasc_> external connectivity
15:40:12 <apuimedo> I guess banix will chair it
15:40:45 <apuimedo> vikasc_: can you give more context (for logging purposes)?
15:41:01 <tfukushima> #link external network connectivity blue print https://blueprints.launchpad.net/kuryr/+spec/external-network-connectivity
15:41:04 <banix> apuimedo: sure
15:41:05 <vikasc_> https://blueprints.launchpad.net/kuryr/+spec/external-network-connectivity
15:41:23 <apuimedo> thanks tfukushima, vikasc_
15:41:45 <apuimedo> #action apuimedo to review the external network connectivity blueprint
15:41:51 <gsagie> same
15:41:55 <apuimedo> sorry vikasc_. I didn't get around to it yet
15:42:00 <gsagie> its about creating the virtual router right?
15:42:07 <apuimedo> let's discuss it online during the week
15:42:17 <vikasc_> sure
15:42:25 <vikasc_> gsagie, yes
15:42:30 <apuimedo> (and of course we can discuss a bit now if more people have the right context)
15:42:35 <gsagie> vikasc_ : i think we need to imitate what magnum are doing
15:42:48 <gsagie> or at least be close to their implementation
15:42:48 <vikasc_> gsagie, yes Gal
15:43:09 <gsagie> You already have the design ready in the blue print?
15:43:30 <vikasc_> High level choices i have mentioned there on bp
15:43:36 <gsagie> i will make sure to go over it and give you feedback
15:43:43 <tfukushima> gsagie: Could you give us the link of what Magnum is doing?
15:44:01 <gsagie> tfukushima: dont have it, i need to search it up and look
15:44:12 <vikasc_> gsagie, thanks
15:44:15 <apuimedo> about blueprints. (and sorry for going a bit off-topic) We should define what deserves a spec and what a blueprint
15:44:28 <tfukushima> Ok. I'll search it.
15:45:22 <gsagie> going back to the multinode, does this mean that we have full integration with Swarm, or something else needs to be done?
15:45:24 <vikasc_> gsagie, tfukushima  through heat templates they create router and their and plumb external networking
15:45:38 <vikasc_> i have little idea baout that
15:45:41 <vikasc_> *about
15:46:07 <gsagie> vikasc_ : will take a look and we can talk about it online
15:46:16 <vikasc_> gsagie, thanks gal
15:46:26 <gsagie> you can always also talk with danyeon from the Magnum team about it
15:46:53 <tfukushima> gsagie: It's basically the same as the default overlay driver. So probably yes although I've never tried anything with Swarm.
15:47:14 <tfukushima> But again, Kuryr is lacking the compatibility with the latest libnetwork.
15:47:14 <apuimedo> I'll be trying swarm with kuryr this week
15:48:06 <gsagie> yeah assuming we stabilize it with latest libnetwork
15:48:07 <gsagie> ok thanks
15:48:11 <apuimedo> yes.
15:48:17 <apuimedo> Any other topic?
15:49:37 <apuimedo> alright. Thank you all for the meeting and let's see if the EMEA folks are able to be on the next meeting :-)
15:49:41 <apuimedo> #endmeeting