15:01:01 #startmeeting neutron_qos 15:01:02 Meeting started Tue Feb 11 15:01:01 2020 UTC and is due to finish in 60 minutes. The chair is ralonsoh. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:03 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:04 hi 15:01:06 The meeting name has been set to 'neutron_qos' 15:01:07 o/ 15:01:25 I don't think slawek is going to join us today 15:01:28 kk 15:01:37 let's wait 30 secs more 15:02:53 #topic RFEs 15:03:15 #link https://bugs.launchpad.net/neutron/+bug/1476527 15:03:16 Launchpad bug 1476527 in neutron "[RFE] Add common classifier resource" [Wishlist,Triaged] - Assigned to Igor D.C. (igordcard) 15:03:20 davidsha, ^ 15:04:23 Ya, I'm wrapping up the unit tests on the service plugin atm, however my team is being redeployed onto other tasks so I won't be able to continue development of the feature going forward. 15:05:18 davidsha, so are you going to stop the development of the feature? 15:05:56 ralonsoh: yesll post an update onto the patches so everyone is aware. 15:06:52 ok, I'll pack all the open patches in the bug 15:06:58 just to have them in a single place 15:07:16 thanks. 15:07:26 but I no one is going to continue the development, I don't think we should merge any of them 15:07:37 including the first one, the n-lib one 15:08:22 ack, I'll put a workflow -1 on the patches for the time being. 15:09:24 anyway, thank you very much for working in such a difficult feature 15:09:48 and if you are not going to work (for now) in OpenStack, thank you very much for your time 15:10:08 it was a pleasure to work with you in the same team 15:10:21 thanks, and thank you for all the reviews and recommendations, I'll need to message Lajos aswell and thank them for the API test work 15:10:38 Thank you, the pleasure was all mine :) 15:11:17 ok, let's move then to the next RFE we have 15:11:23 #link https://bugs.launchpad.net/neutron/+bug/1858610 15:11:24 Launchpad bug 1858610 in neutron "[RFE] Qos policy not supporting sharing bandwidth between several nics of the same vm." [Undecided,New] 15:11:53 we are still debating the in launchpad about the feasibility of this feature 15:12:11 we have 3 debating points 15:12:14 You would need to aggreagate all the ports is it? 15:12:31 that's one question 15:12:37 lest' start by 15:12:40 1) the db model 15:12:53 this is the easiest one 15:13:08 to create a new qos rule, but this rule is exclusive with maxBW 15:13:25 a qos policy can have maxBW or maxBW shared 15:13:27 not both 15:13:53 (this must be defined with more detail in a spec) 15:13:58 2) the neutron server 15:14:30 are we going to allow ports from different VMs with the same qos rule? 15:14:52 a rule can define how a port is scheduled? 15:15:33 IMO, we should NOT block a VM to be scheduled if a port has a shared maxBW rule and there are others ports, in different VMs, with this rule 15:15:45 in this case, the aggreagation is per VM level 15:16:09 and the shared BW will apply only for ports in the same VM with the same maxBW shared qos rule 15:16:34 (but this is just my opinion, no spec is defined yet) 15:16:38 davidsha, any comment? 15:17:11 ralonsoh, Just so I'm clear, is it a 1:1 mapping of Rule to VM to say all these ports with these rules are in the same VM? 15:17:54 no 15:17:58 with *this* rule are in the same VM* 15:18:00 kk 15:18:07 there is no 1:1 mapping between rule and VM 15:18:12 that's what I want to avoid 15:18:29 ack 15:18:41 if we do this, we need to create a placement resource provider to limit the port scheduling 15:18:53 and.... this is too much for this RFE 15:19:00 (just my opinion) 15:19:09 This nearly sounds like it should be a policy applied to the instance rather than the port. 15:19:30 you can have VMs with ports with the same rule 15:19:36 1 maxBW shared rule 15:19:42 applied to 100 ports, 10 per VM 15:19:56 the 10 ports assigned to a VM will share the defined BW 15:20:10 of course, this must be enforced in the backend 15:20:16 yup 15:20:26 and this is the third point 15:20:29 3) backend 15:20:58 3.1) LB: similar to the proposal Jie is doing, but I would like to see more detail on this 15:22:06 3.2) OVS: IMO, we **cannnot** introduce a LB in the middle of br-int and a device port 15:22:28 this won't work with DPDK or offload HW 15:22:56 with DPDK, probably not. 15:23:56 and that's why I would like Jie to propose a spec, to define those points 15:23:59 For OvS, it's QoS objects do let you route traffic into queues, I think it's only for egress traffic though 15:24:25 davidsha, you can make it work like in hybrid OVS 15:24:34 ack 15:24:39 with a LB in the middle of br-int and the VM 15:24:45 kk 15:24:48 and using the same strategy as in LB 15:24:52 but the performance... 15:25:04 and as I said, some compatibility issues 15:25:11 It's possible, just not practicle. 15:25:17 exactly 15:26:04 ok, let's move to next section 15:26:07 #topic Bugs 15:26:18 #link https://bugs.launchpad.net/neutron/+bug/1853171 15:26:20 Launchpad bug 1853171 in neutron "Deprecate and remove any "ofctl" code in Neutron and related projects " [Medium,In progress] - Assigned to David Shaughnessy (david-shaughnessy) 15:26:33 as commented, this will take time 15:27:03 I'm looking into the firewall file at the moment, I'll try to push some initial work for it. 15:27:08 thanks! 15:27:17 no rush 15:27:43 ack 15:28:25 and the next one we have is 15:28:28 #link https://bugs.launchpad.net/neutron/+bug/1648525 15:28:29 Launchpad bug 1648525 in neutron "[OVN] QoS doesn't work with DVR, vlan tenant networks or provider networks." [Undecided,New] 15:28:37 (filled in 2016) 15:28:42 :O 15:29:00 I'll talk to OVN guys to know more about this bug 15:29:12 and how to reproduce it 15:29:35 so far, I don't have more information about it 15:30:14 and that's all I have 15:30:20 #topic Open Discussion 15:30:32 do you have something for this section, davidsha ? 15:31:13 Just to thank everyone again, I enjoyed working with you all and hope to get the chance again! 15:31:20 for sure 15:31:28 thank you very much! 15:31:36 and see you around 15:31:48 See you! 15:32:09 bye 15:32:10 #endmeeting