09:00:13 <oanson> #startmeeting Dragonflow
09:00:14 <openstack> Meeting started Mon Jul  4 09:00:13 2016 UTC and is due to finish in 60 minutes.  The chair is oanson. Information about MeetBot at http://wiki.debian.org/MeetBot.
09:00:15 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
09:00:17 <openstack> The meeting name has been set to 'dragonflow'
09:00:26 <oanson> Hi.
09:00:29 <oshidoshi> hello
09:00:32 <oanson> Who's here for the meeting?
09:00:48 <oshidoshi> o/
09:01:02 <oanson> We'll maybe wait a minute for people to start rolling in
09:01:12 <yuval> o/
09:01:23 <oanson> hujie: hi
09:01:34 <hujie> oanson: Hi :)
09:01:48 <oanson> yuli_s, you here?
09:02:10 <oanson> #info oshidoshi yuval hujie oanson are in the meeting
09:02:20 <yuli_s> Hello !
09:02:32 <oanson> #info yuli_s is also in the meeting
09:02:37 <yamamoto_> hi
09:02:50 <oanson> #info yamamoto_ is also in the meeting
09:02:51 <Shlomo_N> hi
09:02:56 <oanson> yamamoto_, Hi.
09:03:00 <hujie> Hi
09:03:03 <oanson> #info Shlomo_N is also in the meeting
09:03:10 <oanson> #topic Roadmap
09:03:30 <oanson> Reminder, our roadmap document is here: https://etherpad.openstack.org/p/dragonflow-newton
09:03:35 <oanson> #link https://etherpad.openstack.org/p/dragonflow-newton
09:03:55 <oanson> hujie, do you have anything to share about the DB synchronization?
09:04:09 <hujie> yes, I have committed a patch
09:04:19 <hujie> https://review.openstack.org/#/c/336377/7..8
09:04:41 <hujie> https://review.openstack.org/#/c/336377/
09:04:55 <oanson> hujie, do you want us to start reviewing it?
09:05:20 <oshidoshi> i can only see a typo fix in that patch
09:05:28 <hujie> it is only WIP
09:05:32 <oanson> oshidoshi, the patch is marked WIP
09:05:48 <oshidoshi> okay, sure
09:05:51 <hujie> I will delete WIP when the UT is ok
09:06:01 <oanson> hujie, sure.
09:06:25 <oanson> Everyone, please review the patch once the WIP tag is removed.
09:06:50 <hujie> thx
09:06:54 <oanson> Is wangyongben here?
09:07:10 <hujie> not yet, I'll try to find him
09:07:23 <oanson> All right. We'll move on to packaging in the meantime.
09:07:24 <DuanKebo> He is in hangzhou
09:07:49 <oanson> #info DuanKebo is in the meeting.
09:07:56 <hujie> wangyongben will come soon
09:08:06 <oanson> All right.
09:08:30 <oanson> I started playing with openstack-ansible.
09:08:48 <oanson> but I haven't made a lot of progress with it. I hope that in two weeks I will have some interesting results.
09:09:16 <oanson> The IGMP application is marked as a nice to have.
09:09:29 <hujie> what do you want to do by using ansible for dragonflow?
09:09:37 <oanson> Sorry, as a MUST.
09:09:58 <oanson> hujie, the plan is to use ansible to deploy compute nodes and controller nodes in the cloud
09:10:35 <oanson> We want to add dragonflow support to openstack-ansible, which shouldn't be difficult since dragonflow is a python library that can be installed via pip
09:10:44 <oanson> or from the git repository.
09:10:49 <DuanKebo> We are also developing the deploy scripts
09:11:11 <oanson> DuanKebo, what are you working with?
09:11:12 <DuanKebo> for Huawei's cloud service
09:11:26 <oshidoshi> DuanKebo: are you developing these scripts based on open source, or some internal platform?
09:11:30 <DuanKebo> we have our own platform
09:11:39 <hujie> yes, UVP
09:11:54 <DuanKebo> we have to develop based on the platform
09:11:54 <oanson> DuanKebo, the work I'm doing is targeted for open source. For external deployers.
09:12:11 <DuanKebo> Yes, I see
09:12:32 <DuanKebo> We can discuss it to see  if we can share something in common
09:12:33 <oshidoshi> right now, we only have devstack for dragonflow, and that's intended for dev, not for prod
09:12:46 <oanson> DuanKebo, definitely. I will contact you off-line.
09:13:15 <oanson> oshidoshi, yes. With e.g. ansible, the same deployment can be used for both - only difference is in version and configuration.
09:13:42 <oshidoshi> oanson: re IGMP, is it really a MUST, or should we move it to NTH, and move on with VLAN?
09:13:55 <oanson> There are other solutions - e.g. puppet and kolla, but I am taking the ansible direction at the moment.
09:14:09 <oanson> oshidoshi, in the document it is marked as MUST. I have no objections to bumping it down.
09:14:23 <oanson> Does it have an owner?
09:14:35 <oshidoshi> it does not, but I'm guessing... you?
09:14:47 <oanson> In that case, it has been demoted to 'Nice to have'.
09:14:56 <oshidoshi> yep, I just did
09:15:11 <oanson> VLAN networking - is liuhaixia here?
09:15:41 <DuanKebo> Yes
09:15:54 <DuanKebo> she is not far from me
09:16:08 <hujie> she's coming
09:16:16 <oanson> Hi liuhaixia
09:16:22 <oanson> #info liuhaixia is also in meeting
09:16:28 <liuhaixia> hi
09:16:35 <liuhaixia> Here.
09:16:37 <oanson> liuhaixia, do you have anything to update about the VLAN features?
09:17:05 <liuhaixia> yes
09:17:31 <liuhaixia> I give a new patch last week.
09:17:46 <oshidoshi> which of the VLAN tasks are we discussing, the TRUNK PORT or the NETWORKS
09:17:54 <wangyongben> hi oanson, I am coming
09:18:06 <oanson> Hi, wangyongben.
09:18:26 <hujie> oshidoshi: vlan network
09:18:27 <oanson> #info wangyongben is also in meeting.
09:18:29 <liuhaixia> My patch only support access port.
09:18:53 <oanson> liuhaixia, do you plan to take port trunking as well?
09:18:55 <DuanKebo> It's about vlan networks, trunk port isn't supported yet.
09:19:25 <oanson> #link VLAN networking spec https://review.openstack.org/#/c/288270/
09:19:41 <oanson> Everyone, please review this spec so that implementation can get under way.
09:19:43 <liuhaixia> I  have no plan by now :)
09:20:02 <oanson> Fair enough. It needs an owner. Any volunteers?
09:20:43 <oanson> I guess we'll find someone later :)
09:21:06 <oanson> wangyongben, any update on the ML2/L3 service plugin?
09:21:27 <wangyongben> I have commit a patch for qos
09:22:06 <wangyongben> it  change the function for update_port ， create_netwrok and update_network
09:22:34 <liuhaixia> This patch is the implementation. It inludes 'change tunnel key from port key to segmentation', and supports vlan and flat networks(access port).https://review.openstack.org/#/c/334798/
09:22:44 <wangyongben> https://review.openstack.org/#/c/331932/
09:23:44 <oanson> liuhaixia, this is a very long patch. I understand that it handles several issues. Is that correct?
09:24:48 <liuhaixia> It implements l2 communication as what l2_app does, what's more it 'change tunnel key from port key to segmentation', and supports vlan and flat networks.
09:25:39 <liuhaixia> I give a new app, because if i change l2-app, it always conflict with others. I  have change it a lot of times.
09:25:49 <oanson> liuhaixia, would it be possible to split it to a few smaller patches? e.g. 1 for l2 communication, 1 for tunnel key -> segmentation id, 1 for vlan, and 1 for flat networks?
09:26:30 <oanson> About QoS patch, this is the link to the QoS API: http://developer.openstack.org/api-ref-networking-v2-ext.html#qos-ext
09:27:33 <liuhaixia> oanson,  if i only split it, for example, 1 for tunnel key. It will be also a big patch.
09:27:33 <oanson> liuhaixia?
09:27:45 <oanson> but it will be smaller, right?
09:27:50 <liuhaixia> With vlan and flat, it just a small branch.
09:28:15 <liuhaixia> It will be a little smaller, just a little :)
09:28:38 <oanson> Reviewing a patch with 1000+ lines of code is very difficult. I understand there is a lot of content, but anything that can be done in a separate patch, should.
09:28:39 <hujie> liuhaixia mean for tunnel key 900 loc, for vlan 100 loc :)
09:29:04 <oanson> Otherwise, it will take a lot of time to review, and will be very difficult to merge.
09:29:12 <oanson> and 900 lines is better than a 1000 :)
09:29:31 <liuhaixia> That's why i add a new app:)
09:29:33 <oanson> and it means that the small parts, vlan, flat, segmentation id, will be merged faster.
09:30:00 <liuhaixia> I tried to split it, but it does not take effect obviously.
09:30:13 <oanson> liuhaixia, please do what you can, and we will review it.
09:30:16 <hujie> oanson: what about my remote port patch?
09:30:38 <DuanKebo> we need to merge the vxlan patch first.
09:30:51 <DuanKebo> the other modifications depend on that patch.
09:30:57 <oanson> hujie, this one? https://review.openstack.org/#/c/329806/
09:31:10 <oanson> DuanKebo, which patch? Please provide a link or number.
09:31:15 <liuhaixia> oanson ,I have tried, it didn't change obvirously.
09:31:31 <hujie> no, this one : https://review.openstack.org/#/c/328706/
09:32:31 <DuanKebo> https://review.openstack.org/#/c/334798/
09:32:32 <oanson> I'll review it again. I'll also consult Yamamoto, and see if it can be merged.
09:32:42 <DuanKebo> this patch
09:32:42 <oanson> liuhaixia, I'm sorry, I don't understand. Can you please explain?
09:32:48 <wangyongben> Sorry, I am off-line just now. I commit a new patch "Dragonflow Support QoS", this patch will update ml2 mech driver, it add a parameter "qos_policy"  when update port, create network and update network.
09:32:50 <liuhaixia> Because it's a full implementation of l2 commutication.
09:33:40 <oanson> wangyongben, we'll review it, but the QoS is marked nice to have.
09:33:43 <hujie> oanson: ok, please tell me if you have a conclusion :)
09:33:57 <oanson> What about the ML2/L3 plugin? I think it has higher priority?
09:33:59 <oanson> hujie, sure.
09:34:19 <oanson> #action oanson consult Yamamoto about https://review.openstack.org/#/c/328706/
09:34:49 <oanson> liuhaixia, but you can still move vlan, flat network, and segmentation out, right?
09:34:51 <hujie> https://review.openstack.org/#/c/329806/  what about this patch, could it be merged?
09:34:51 <yamamoto_> i'll revisit
09:34:57 <oanson> this way we can merge these changes.
09:35:02 <DuanKebo> Omer, Liuhaixia can split the patch into serveral patches.
09:35:20 <DuanKebo> but the vxlan patch is still very big.
09:35:35 <oanson> Yes, but 900 lines is still better than 1000
09:35:35 <hujie> yamamoto_
09:35:35 <hujie> : thx
09:35:38 <oanson> yamamoto_, thanks.
09:35:44 <liuhaixia> If i split it to three patch ,it will reduce about 100 lines of code.
09:36:01 <DuanKebo> maybe we need to do this painstaking work
09:36:05 <oanson> liuhaixia, still good. And the other two patches can be reviewed faster and be merged.
09:36:12 <DuanKebo> get it merged asap.
09:36:21 <oanson> Sure.
09:36:43 <oanson> Everyone, please help with the reviews. It will help a lot in moving things forwards.
09:36:48 <oanson> All right, let's move on.
09:36:58 <oanson> #topic Barcelona Summit
09:37:24 <oanson> Our etherpad (https://etherpad.openstack.org/p/dragonflow_barcelona_summit_talks) is still very empty.
09:37:35 <liuhaixia> Becuase, now  l2_ml2_app based on l2_app, previous got some changes, it need to change the later either.
09:37:59 <oanson> We have some talk ideas that we will post, if anyone would want to take or collaborate.
09:38:35 <oanson> liuhaixia, I think we should continue off-line. I would like to move on to other items.
09:38:48 <liuhaixia> OK:)
09:38:59 <oanson> If anyone has any talks they would like to share or have reviewed, please post in the etherpad.
09:39:04 <oanson> #link https://etherpad.openstack.org/p/dragonflow_barcelona_summit_talks
09:39:29 <oanson> Anyone has something else to add in this topic?
09:41:10 <oanson> #topic Performance Testing
09:41:18 <oanson> Shlomo_N, anything to update here?
09:41:55 <Shlomo_N> sure
09:42:23 <Shlomo_N> Basically the performance testing framework is working end-to-end
09:43:08 <Shlomo_N> 1. setting the testing environment, starts the tests, collects the results from all VMs and finally generate a report
09:43:50 <Shlomo_N> You can review the patch
09:44:04 <Shlomo_N> https://review.openstack.org/#/c/304470/
09:44:28 <oanson> Shlomo_N, great!
09:44:41 <oanson> Is it still Work-in-progress?
09:44:45 <Shlomo_N> yes
09:45:00 <oanson> All right.
09:45:24 <oanson> Anyone else has something to add?
09:45:48 <oanson> #topic Bugs
09:45:50 <DuanKebo> Has anyone considering vm qos
09:46:22 <oanson> DuanKebo, can you please explain?
09:46:32 <DuanKebo> current qos is applied to port
09:46:57 <DuanKebo> sometimes we need to limit the speed of the whole vM
09:47:15 <hujie> what about this bug: https://review.openstack.org/#/c/329806/
09:47:18 <DuanKebo> this feature has been supported by AWS
09:47:42 <oanson> DuanKebo, We are bound by Neutron's API.
09:47:50 <oanson> I am looking at it now.
09:48:09 <DuanKebo> yes
09:48:39 <DuanKebo> Neutron api need to be extended.
09:49:09 <DuanKebo> but this feature is interested by some customers.
09:49:32 <oanson> hujie, I workflow+1 the patch. Please review the cosmetic issues I've raised.
09:50:03 <oanson> DuanKebo, then yes, the QoS API has to be extended. This has to be proposed to them.
09:51:15 <DuanKebo> We are working on it now. If you are interested, we can discuss when you come to China.
09:51:33 <oanson> Sure. I'd be happy.
09:51:51 <oanson> yuli_s, Any updates on our bugs?
09:53:56 <yuli_s> yes
09:54:02 <oanson> Go for it
09:54:39 <yuli_s> ops, sorry
09:54:50 <yuli_s> i had only old news regarding bugs
09:55:09 <oanson> Any objections that we move on, then?
09:55:11 <yuli_s> i fixed the dual handling of messages
09:55:16 <oanson> Great.
09:55:25 <DuanKebo> Great!
09:55:28 <yuli_s> i am working on control test with the dockers
09:56:10 <yuli_s> so, I am doing it is part of control plane stress test
09:56:25 <yuli_s> it looks better now,
09:56:44 <yuli_s> hope to be able to run the tests today on separate servers
09:58:19 <oanson> All right. Good luck.
09:58:29 <oanson> Looks like we won't have time for open discussion and open reviews.
09:58:58 <oanson> Please everyone, review the patches. Especially the long ones, since they need the most effort and take most of the time.
09:59:05 <oanson> Anything else before we finish?
09:59:27 <oanson> All right. Thanks everyone.
09:59:37 <Shlomo_N> thank you oanson
09:59:39 <yuli_s> thanks !
09:59:40 <oanson> #endmeeting