15:00:37 <gsagie_> #startmeeting Kuryr 15:00:37 <openstack> Meeting started Mon Oct 19 15:00:37 2015 UTC and is due to finish in 60 minutes. The chair is gsagie_. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:38 <tfukushima> Toni is on vacation, BTW. 15:00:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:41 <openstack> The meeting name has been set to 'kuryr' 15:00:49 <gsagie_> Yeah, Toni is in Tokyo :) 15:00:59 <gsagie_> setting the grounds for us 15:01:13 <tfukushima> He's visiting Kyoto actually. 15:01:22 <gsagie_> Who is in for the meeting, i see banix feisty and tfukushima 15:01:27 <gsagie_> feisky 15:02:00 <gsagie_> i think we will have a short meeting this week because everyone are probably busy with the summit.. 15:02:12 <gsagie_> #topic libnetwork updates 15:02:32 <gsagie_> tfukushima: anything to update us on the progress? most patches are merged 15:02:44 <tfukushima> So my patches for catching up with the libnetwork was merged. 15:03:00 <tfukushima> Vikas is working on updading the devref. 15:03:14 <salv-orlando> aloha 15:03:21 <gsagie_> ok cool once this is done i will update our Kuryr spec at the neutron repository 15:03:24 <gsagie_> Hello salv-orlando 15:03:31 <tfukushima> Vikas is also working on the validation issue for /NetworkDriver.CreateEndpoint . 15:03:55 <gsagie_> #action gassy update Kuryr neutron spec with lib network new workflow 15:04:08 <gsagie_> #action gsagie update Kuryr neutron spec with lib network new workflow 15:04:47 <gsagie_> tfukushima: so we are right now synced with the service removal and it should work if i rebase on your Join patch? 15:05:43 <tfukushima> Actually Vikas found the behaviour of change of creating endpoints as I put above. 15:06:02 <tfukushima> Except for that I confirmed it works against the specific version of libnetwork. 15:06:12 <tfukushima> In the latest one, I found some issues. 15:06:33 <tfukushima> vagrant@devstack:~/kuryr$ sudo docker run -it --net=foo --rm ubuntu /bin/bash 15:06:34 <tfukushima> Error deleting container: Error response from daemon: Driver aufs failed to remove root filesystem f7f5134b9043da4d5d44e6849ec4a609f1e41df0942ae9581714208134080332: rename /var/lib/docker/0.0/aufs/mnt/f7f5134b9043da4d5d44e6849ec4a609f1e41df0942ae9581714208134080332 /var/lib/docker/0.0/aufs/mnt/f7f5134b9043da4d5d44e6849ec4a609f1e41df0942ae9581714208134080332-removing: device or resource busy 15:06:34 <tfukushima> Error response from daemon: Cannot start container f7f5134b9043da4d5d44e6849ec4a609f1e41df0942ae9581714208134080332: failed to create endpoint admiring_pike on network foo: driver modified interface address: endpoint interface IP present (172.19.0.2/16). Cannot be modified with (172.19.0.2/16).; rolled back 15:07:57 <tfukushima> I tested against the following version and everything works with that. 15:07:58 <tfukushima> https://apt.dockerproject.org/repo/pool/experimental/d/docker-engine/docker-engine_1.9.0~dev~git20150924.164351.0.1e514de-0~jessie_amd64.deb` 15:08:25 <gsagie_> ok good work, we will probably demo from that version anyway and not the latest 15:08:34 <vikasc__> o/ 15:08:36 <tfukushima> So we'd use that version, yes. 15:08:39 <gsagie_> and nice work from both you and vikas 15:08:43 <vikasc__> sorry for being late 15:08:47 <banix> so 1.9.0-rc1 is the one we nee dto work it but I think this late in our cycle staying with a couple of commits earlier we are ok 15:08:48 <gsagie_> hi Vikas! 15:08:57 <vikasc__> Hi Gal 15:09:25 <vikasc__> yes , IMO alo we should work on 1.9.0_rc1 15:09:47 <gsagie_> okie, vikas is that the version you changed the vagrant installation to point to? 15:09:58 <vikasc__> yes Gal 15:10:27 <tfukushima> Yeah, I tested with the env created by Vagrant. Thanks for that Vikas. 15:10:40 <gsagie_> okie great, after the summit we can probably investigate the things tfukushima noticed with latest builds 15:11:06 <gsagie_> #topic demo in openstack summit 15:11:14 <gsagie_> mestery: here? 15:11:20 <mestery> gsagie_: howdy! :) 15:11:33 <gsagie_> Anything is missing on your end for the demo? 15:11:40 <gsagie_> Hi daneyon ! :) 15:12:38 <gsagie_> tfukushima, banix: i think we agreed not to demo with Docker Swarm, or has that changed? 15:13:02 <banix> gsagie_: no that’s my understanding 15:13:15 <gsagie_> okie 15:13:51 <gsagie_> anything new to update on that Integration band? 15:13:53 <gsagie_> banix 15:14:39 <banix> gsagie_: are you referring to swarm? 15:14:50 <gsagie_> yes 15:15:34 <banix> gsagie_: we have been doing some work on that area but nothing interesting to report yet; simply catching up with new changes 15:15:44 <gsagie_> okie cool 15:16:17 <gsagie_> tfukushima: is there any help you need for the demo ? i think you helping Toni on that, please let me know if there is anything you need from me 15:16:43 <vikasc__> i am also available for any help Taku 15:17:12 <banix> gsagie_: tfukushima it’s good to know what exactly the plan is for the demo 15:17:16 <tfukushima> I don't have issues for now. So I heard no multi node connectivity demo. Am I understanding correctly? 15:17:51 <tfukushima> Basically it'd be the same as the last demo I put but with Horizon. 15:18:11 <tfukushima> banix: I heard you worked on some multi node connectivity stuff, which might be the next topic. 15:18:33 <banix> tfukushima: thanks; regarding horizon just to confirm my understanding, the conatiners show up only if we use compute:nova as owner? 15:19:06 <gsagie_> banix: i believe so, we need to add support there feisky is working on it i believe 15:19:16 <banix> tfukushima: yes, I have a multi node system but as independent docker hosts without swarm fronting them 15:19:18 <tfukushima> banix: Yes. We just show we set the device_owner field at this point. 15:19:28 <banix> ok thanks 15:19:32 <gsagie_> we want to be able to use kuryr:container as the owner 15:19:40 <gsagie_> (similar to what Ironic is doing) 15:20:14 <gsagie_> #topic trello board 15:20:54 <gsagie_> Ok, i believe everyone know but we have a Trello board for the project and after the summit i think we should start clearing it and transferring things to bugs/blue prints and track everything in launchpad 15:21:06 <gsagie_> #link https://trello.com/b/cbIAXrQ2/project-kuryr 15:21:35 <gsagie_> so we should probably clean everything that is already done, i will work on it after the summit 15:21:40 <tfukushima> I finished some tasks, I believe, i.g., "Docker 1.9 new API adjustments". 15:21:46 <tfukushima> Oh, Ok. 15:22:05 <gsagie_> tfukushima: Feel free to delete it, and anything else you finished 15:22:18 <gsagie_> i think all the members should have edit premissions 15:22:20 <tfukushima> gsagie_: Got it. 15:22:31 <gsagie_> #topic Neutron design summit 15:23:04 <gsagie_> We will have part of a session to speak about Kuryr, i am still not sure how much time exactly but wanted to make sure the time it self fits everyone 15:23:24 <gsagie_> http://mitakadesignsummit.sched.org/event/9be2ec1db45ea3267f811b8d35816e1c#.ViUK4kKfx-M 15:23:28 <gsagie_> #link http://mitakadesignsummit.sched.org/event/9be2ec1db45ea3267f811b8d35816e1c#.ViUK4kKfx-M 15:23:42 <gsagie_> If anyone has any conflict, please let me know and i will see what we can do 15:25:14 <gsagie_> And if anyone from the team has something that he/she thinks should be on the items, please let me know as we are working on the exact schedule 15:26:13 <banix> gsagie_: sounds good; it would be mice if we could find an hour or two for Kuryr to talk among ourselves. Saying it now as I am sure scheduling it will be a challenge 15:26:43 <gsagie_> banix: yeah, its actually an action item on my self, are you staying friday as well? 15:26:47 <banix> no mice, nice :) 15:26:58 <banix> yes 15:27:38 <gsagie_> so worst thing if we won't be able to find anything sooner we can do it on Friday, i will try to find something else and update everyone, would also like the Magnum team to join us 15:27:56 <gsagie_> or at least the ones that are involved with the networking part 15:28:51 <banix> sounds good 15:29:04 <gsagie_> #topic open discussion 15:29:24 <gsagie_> Anything else from anyone? 15:29:34 <tfukushima> If we don't have any other topic, I'd like to talk about the multi node configuration a little bit. 15:29:43 <gsagie_> tfukushima: sure 15:30:25 <tfukushima> So I tried --cluster-store and --cluster-adverties options of Docker 1.9.0. 15:30:32 <vikasc__> Can we discuss remote ipam driver support after that? 15:31:04 <tfukushima> vikasc__: Yeah, mine would be short. 15:31:16 <vikasc__> thanks tfukushima 15:31:30 <gsagie_> hi irenab 15:31:43 <tfukushima> Actually I tried ZooKeeper as the backend because that's what I have right now. 15:31:54 <tfukushima> From the log I can see they're communicating each other. 15:32:22 <tfukushima> However, when I create a network on a host, it's not synced on another host. 15:33:05 <tfukushima> With overlay networking driver, it'd be synched among multiple hosts. 15:33:07 <banix> tfukushima: when you say not sync, what do you mean? the network doesnt show on netowrk ls? 15:33:21 <tfukushima> banix: Yes. 15:34:00 <tfukushima> banix: How about your system? Have you experienced that? If so how did you add the capability to syncing multiple hosts? 15:34:14 <banix> tfukushima: hmmm. have tried it with consul and that works fine 15:34:18 <tfukushima> It's possible that I'm configuring it wrong. 15:35:01 <tfukushima> Ok, I'd appreciate if you could share the configuration. I must be doing wrong. 15:35:22 <banix> tfukushima: --cluster-store=consul://localhost:8500 --cluster-advertise=<your IP>:2375 15:35:31 <tfukushima> DOCKER_OPTS="-D -H unix:///var/run/docker.sock -H tcp://0.0.0.0:2376 --cluster-advertise=192.168.11.14:2376 --cluster-store=zk://192.168.11.14:2181” 15:36:13 <tfukushima> banix: Thanks I'll try that. And I'm glad it worked on your side. 15:36:49 <banix> tfukushima: this is the command line omn one node: docker -dD -H :2375 --cluster-store=consul://localhost:8500 --cluster-advertise=192.168.122.185:2375 15:37:19 <banix> tfukushima: sure; ping on IRC if you think i can be of any help 15:37:37 <tfukushima> banix: Ok. Thank you very much. 15:38:00 <tfukushima> vikasc__: I think I finished. Please go ahead. 15:38:50 <vikasc__> how are we going to support remote ipam driver apis? 15:39:26 <vikasc__> in the same process or a seperate server for ipam driver also 15:39:45 <vikasc__> thoughts? 15:39:59 <banix> vikasc__: seperate sounds better 15:40:15 <banix> as it allows possibly having different versins of it? if need be 15:40:32 <vikasc__> banix, make sense 15:40:43 <gsagie_> Vikas: can you explain more what do you mean 15:41:23 <salv-orl_> vikasc__: yeah curious about that too. I thought we'd leverage ipam via neutron only. 15:41:57 <gsagie_> yeah same, and neutron has a pluggable IPAM already, so we will just need to connect to it 15:42:07 <vikasc__> gsagie_, user has two options now.. either use built-in ipam of libnetwork or configure driver through docker cli 15:42:45 <vikasc__> we will not need default subnets now i guess 15:42:49 <gsagie_> vikasc__ : can you share a #link where its described maybe ? 15:43:08 <vikasc__> one moment please 15:43:33 <vikasc__> https://github.com/docker/libnetwork/blob/master/docs/ipam.md 15:43:36 <salv-orl_> vikasc__: I think I see what you mean. A user can specify a 3rd party IPAM driver in addition to the libnetwork driver? 15:43:38 <gsagie_> and are those lib network IPAM calls (that are done using docker cli) not being directed to the remote driver *Kuryr) ? 15:43:57 <gsagie_> #link https://github.com/docker/libnetwork/blob/master/docs/ipam.md 15:44:07 <vikasc__> gsagie_, no 15:44:17 <vikasc__> gsagie_, first they will go to ipam driver 15:44:36 <vikasc__> if user has not mentioned .. libnetwork has a built it 15:44:51 <banix> well the remore drivers and the remote ipam drivers are essentially different entities; but they can of course be handled by a single kuryr driver 15:45:03 <salv-orl_> vikasc__: it is interesting that this interface seems very close to neutron's ipam driver interface 15:45:12 <banix> by remore driver i mean the dcker betwork plugin talking to the remote driver 15:45:23 <vikasc__> banix, agreed 15:45:43 <salv-orl_> vikasc__: as docker has a default driver, does this mean that neutron's IPAM is going to be pretty much bypassed by docker's now? 15:45:58 <salv-orl_> unless a kuryr IPAM driver is provided as well? 15:46:01 <gsagie_> banix: just wondering why do you think that running those separately is better? 15:46:02 <banix> salv-orl_: right now yes 15:46:12 <vikasc__> salv-orl_, yes 15:46:20 <tfukushima> vikasc__: I'm wondering what is the interface from the perspective of Kuryr or other remote drivers? 15:46:41 <tfukushima> #link new create network API https://github.com/docker/libnetwork/blob/master/docs/remote.md#create-network 15:46:47 <banix> right now there is no option of having a “null” libnetwork ipam driver so you either have that or you bring your own that implements the api 15:46:57 <salv-orl_> banix, vikasc__: do you think an IPAM driver could just be implemented using Neutron's APIs or do you reckon we might need direct access to Neutron IPAM, which is not exposed via the API atm? 15:46:59 <vikasc__> banix, exactly 15:47:04 <banix> gsagie_: just because these are logically different components/modules 15:47:54 <tfukushima> I thought it would be in /NetworkDriver.CreateNetwork. 15:47:55 <banix> salv-orl_: i think using Neutron API would be sufficient but need to look more closely 15:48:28 <vikasc__> salv-orl_, abstraction layer need to be introduced for defining interface 15:49:39 <banix> tfukushima: there are access through /IpamDriver.RequestPool etc 15:49:47 <banix> they are accesses 15:49:51 <banix> accessed 15:49:55 <vikasc__> salv-orl_, we will have to define ipam apis. Then we can see how much we can reuse from neutron 15:50:32 <gsagie_> banix: i am afraid that splitting these two in Kuryr, might lead to all strange race conditions, but i can't really back that up until i read more and understand the lib network IPAM better 15:51:04 <vikasc__> gsagie_, thats why i brought this topic so that we can discuss more in next meeting 15:51:11 <banix> gsagie_: let’s look into it more closely and discuss further 15:51:20 <vikasc__> we will have more thoughts by then 15:51:22 <gsagie_> okie, 15:51:39 <gsagie_> #action everyone to examine the new lib network IPAM APIs 15:51:52 <gsagie_> #link https://github.com/docker/libnetwork/blob/master/docs/ipam.md 15:52:03 <salv-orl_> vikasc__: why? bike shedding about things you don't actually know if fun ;) 15:52:25 * salv-orl_ is fun 15:52:41 <vikasc__> salv-orl_, :D 15:52:46 <tfukushima> BTW, do we have the meeting next week in the same timeline? 15:52:56 <banix> is there a session on bike shedding at the summit? 15:53:11 <vikasc__> here is the blueprint i drafted for same.https://blueprints.launchpad.net/kuryr/+spec/remote-ipam-driver 15:53:12 <gsagie_> tfukushima: i don't think so 15:53:14 <banix> tfukushima: i would say let’s cancel the next week meeting 15:53:54 <tfukushima> Ok, there'd be tons of meetings next week. :-p 15:54:01 <vikasc__> :) 15:54:22 <gsagie_> yeah, we could all meet in person, no need to do it on IRC :) 15:54:50 <feisky> I'm coming back. Has the discussion finished? 15:55:29 <gsagie_> feisky: yes, but you can read the log in a minute 15:55:35 <feisky> I have two questions 15:55:53 <gsagie_> sure, but we have 2 minutes before the other meeting start 15:56:03 <salv-orl_> banix: the summit is actually all about bike shedding, as you surely know 15:56:25 <feisky> first about device_owner, we must have an agreement before take other actions to horizon 15:56:28 <banix> :) 15:57:31 <vikasc__> :) 15:57:51 <feisky> We can't expect expect horizon to accept a new device_owner since we are still be debating 15:57:56 <gsagie_> feisky: lets take this offline, we discussed about it 15:58:19 <gsagie_> feisky: we are going to use a patch with nova:compute for the demo and then proceed with what was mentioned in the patch 15:58:34 <gsagie_> i think irena was commenting 15:58:41 <feisky> OK 15:59:02 <gsagie_> ok everyone are time is up, thanks for coming 15:59:06 <gsagie_> our 15:59:13 <gsagie_> and see you all in the summit 15:59:23 <tfukushima> Have safe trips, guys! 15:59:23 <gsagie_> #endmeeting