18:01:03 #startmeeting networking_policy 18:01:04 Meeting started Thu Sep 15 18:01:03 2016 UTC and is due to finish in 60 minutes. The chair is SumitNaiksatam. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:01:06 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:01:08 The meeting name has been set to 'networking_policy' 18:01:30 #info agenda https://wiki.openstack.org/wiki/Meetings/GroupBasedPolicy#Sept_15th.2C_8th_2016 18:01:41 just a follow up of last week’s discussion about the QoS patch 18:01:48 #topic Quality of Service support via NSPs 18:01:53 #link https://review.openstack.org/#/c/301701 18:01:55 igordcard: hi 18:02:11 igordcard: i think the only outstanding comment is the one from Bob about the devstack setup, right? 18:02:32 bob -> rkukura :-) 18:03:06 rkukura: in the comment you have already provided the suggestion on what to fix, right? 18:03:21 rkukura: he will need the to have the qos extension 18:03:23 actually 2 comments 18:03:34 rkukura: but perhaps not the port_security 18:04:09 If the normal devstack enables port_security, I think we should too 18:04:22 rkukura: okay, i dont think it does, but i am not sure 18:05:29 my bad, it does: #link https://github.com/openstack-dev/devstack/blob/fac8adbe1b2bdd2bae105228eb7acf2dd3f30da7/lib/neutron#L169 18:05:36 SumitNaiksatam: I know the ENABLE_APIC_AIM setup does 18:06:08 rkukura: yeah, and i was going to check if the GBP devstack also does it, but seems like you are confirming it does 18:06:29 rkukura: so all you are saying is that it need not be specified explicitly in the GBP conf? 18:06:37 GBP -> GBP devstack 18:07:18 When GBP overrides something from regular devstack, it should add what it needs, and not remove thinkgs unless there is good reason 18:07:28 rkukura: right 18:07:53 rkukura: what is the second comment? 18:08:02 My point with my comment is just that he should use the override-defaults file to override extension_drivers 18:08:15 rkukura: okay, got it 18:08:43 and also that it needs to take into account the fact that ENABLE_APIC_AIM already overrides extension_drivers 18:09:07 rkukura: i will try to reach out to igordcard offline to settle this 18:09:20 The 2nd comment is that adding this config stuff to local.conf seems redundant 18:09:29 rkukura: okay 18:09:30 Isn’t it handled by the GBP plugin? 18:09:50 rkukura: i did not check if there was an overlap 18:10:05 We shouldn’t need to set extension_drivers both via override-defaults and via local.conf 18:10:12 rkukura: agreed 18:10:51 There may be some stuff he sets in local.conf that he doesn’t set via some other mechanism, such as the agent extension 18:11:45 rkukura: agreed, i havent been paying much attention until this point if there is any redundant config in local.conf, i have just been going with the mindset that it doesnt hurt 18:11:56 rkukura: but yeah, over time this can lead to issues 18:12:16 We should minimize what needs to be added in local.conf for GBP to work 18:12:24 rkukura: agreed 18:12:38 rkukura: other than that you are okay with the patch? 18:12:51 I think so 18:12:56 rkukura: ok good 18:13:19 so if no one else has any objections, we can merge this patch once the devstack issues are sorted out 18:13:25 rkukura: thanks for the review 18:13:29 I had forgotten to submit my comments, and didn’t realize when I finally did that igordcard had already uploaded patch set 22 18:13:38 rkukura: ah okay, np 18:13:49 looks like igordcard is not around 18:13:52 So maybe I should re-post my comments on the latest patch set 18:13:56 so moving on 18:14:00 rkukura: sure 18:14:18 #topic Tags 18:14:44 a couple of weeks back we had the discussion about extending the Neutron tags framework to GBP resources 18:15:36 there was some back and forth about whether using that framework was the best choice for the use case that we immediately wanted to solve, which is providing a micro-segmentation tag on the PT 18:16:20 while i think we should still extend the neutron tags framework, to specifically sove the above use case, i am leaning towards just creating a driver specific extension for the apic_mapping driver 18:16:56 we can use this as a experiment to gain experience in terms of how want to progress in the future with other resources 18:17:34 rkukura: sound okay? 18:18:03 SumitNaiksatam: What’s your view on the argument that tags are intended for orchestration layers above, and should not influence back-end behaviour? 18:18:33 rkukura: so in the driver extension we will not call them tags :-) 18:18:52 tomato, tomAHtoe 18:18:55 That’s one way of avoiding that argument ;) 18:18:58 rkukura: i think makes sense not to override established convention 18:19:15 tbachman: :-) 18:19:57 tbachman: to give a brief summary, we want to be able to use something akin to what people seem to call “labels” 18:20:03 SumitNaiksatam: Wasn’t there already some sort of tag/label proposed for the GBP model, but not implemented? 18:20:16 rkukura: that is right, it was not implemented 18:20:27 rkukura: we were calling those policy tags 18:20:32 Is that equivalent to what you are proposing to add now? 18:20:50 rkukura: but that terminology pre-dates the “tag” usage in openstack 18:21:24 rkukura: that proposal was a full blown resource, with notion of hierarchy, etc 18:21:42 I figured we used “tag” instead of “label” in the same way we used “policy target” instead of “endpoint” ;) 18:21:52 rkukura: for now i just mean to add a string driver extension attribute to the PT 18:22:03 SumitNaiksatam: understood 18:22:22 rkukura: ha, could be (my memory fails me) 18:22:31 I don’t see a problem with that, but would want to make sure the semantics were cleanly defined in a way that could eventually be required of all policy drivers 18:22:58 rkukura: okay 18:24:25 anyone else have thoughts on this? 18:25:46 okay moving on 18:25:54 #topic Open Discussion 18:26:08 is songole or hemanthravi here? 18:26:35 i dont have anything else, just wanted to check about the pending NFP patches 18:26:47 oh btw, we do have a new “aim” gate job 18:27:11 and currently the aim and the nfp devstack gate jobs are failing because they are running on xenial 18:27:23 i tried to fix that in this patch: 18:27:43 #link https://review.openstack.org/#/c/369009/ 18:28:02 i fixed the nfp job with that but aim is failing for other reasons, so we need to fix that 18:28:19 hi 18:28:25 hi 18:28:31 if nothing else i think i should make this fix only for the nfp patch 18:28:35 and get it to work 18:28:38 igordcard: hi, welcome 18:28:40 sorry I had this scrolled up 18:28:47 igordcard: we discussed your patch 18:29:01 igordcard: thanks for posting the new patchset 18:29:03 keep going, we can go perhaps go back to qos at the end? 18:29:24 igordcard: are you fine with the devstack related comments that rkukura has provided? 18:29:46 SumitNaiksatam: yes I just replied... the local.conf is simply because of the functional tests 18:30:13 igordcard: ah okay, so its not the devstack plugin local.conf, its the one for the gate job? 18:30:19 SumitNaiksatam: yeah 18:30:28 igordcard: then its not a problem 18:30:36 gbpservice/tests/contrib/devstack/local.conf 18:30:42 igordcard: okay cool 18:31:01 igordcard: You should probably also use override-defaults for setting agent:extensions 18:31:04 igordcard: of course at some point we need to move the gate job to run the devstack plugin 18:31:19 igordcard: the above comment was not directed at you 18:31:38 meant that as a general comment 18:31:45 Is the gate job not using the plugin? 18:33:40 rkukura: no 18:34:10 SumitNaiksatam: OK, didn’t realize that 18:34:14 rkukura: the gate job setup pre-dated the plugin, and we never got around to changing the gate job once the plugin was there 18:34:31 rkukura: but, the configuration is very similar (if not identical) 18:35:40 SumitNaiksatam: Lets use the plugin for the new AIM gate job and get that working, then switch the regular gate job over. 18:35:54 rkukura: okay 18:36:47 anyone else wants to discuss anything else today? 18:37:38 alrighty, thanks everyone for joining 18:37:40 bye! 18:37:44 SumitNaiksatam: thanks! 18:37:55 #endmeeting