18:01:13 #startmeeting networking_policy 18:01:13 Meeting started Thu Oct 13 18:01:13 2016 UTC and is due to finish in 60 minutes. The chair is SumitNaiksatam. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:01:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:01:16 The meeting name has been set to 'networking_policy' 18:01:40 # #info agenda https://wiki.openstack.org/wiki/Meetings/GroupBasedPolicy#Oct_13th_2016 18:02:10 there were a few bugs on the server side, client and UI which were fixed over the last few days 18:02:17 i dont have anything critical at my end 18:02:21 #topic Bugs 18:02:38 ivar-lazzaro: this one #link https://review.openstack.org/#/c/382036/ 18:02:46 “fix proxy_group migration default” 18:02:59 ivar-lazzaro: are you planning on updating that patch? 18:03:27 SumitNaiksatam: looking 18:03:35 songole: hi 18:03:41 hi SumitNaiksatam 18:03:42 ivar-lazzaro: the db migtation scheme 18:03:51 yeah I'll update 18:03:54 scheme -> schema script 18:03:59 ivar-lazzaro: ok great, thanks 18:04:36 songole: this one needs another +2 #link https://review.openstack.org/#/q/topic:bug/1630903 18:04:55 hi 18:05:06 SumitNaiksatam: ok, will do 18:05:46 songole: the NFP patches are in the queue for a long time 18:06:07 songole: also the NFP gate test keeps failing intermittently, but is more often ow 18:06:39 songole: is there anything we need to discuss on that front here? 18:06:54 Jagadish is on vacation this week. They will get to it early next week. 18:07:13 songole: okay thanks 18:07:34 i noticed some spurious posts a couple of weeks back, e.g. #link https://review.openstack.org/#/c/380661/ 18:07:53 there was another patch which was posted as a revert of my patch by the same author ^^^ 18:08:01 not sure what is going on 18:08:12 if anyone else has seen or heard of this before 18:08:57 #topic Queue DHCP and Nova notifications 18:09:03 the email address on that is weird 18:09:06 #link https://review.openstack.org/#/c/384282/ 18:09:15 tbachman: yeah, i tried to dig up 18:09:18 xaxadmin 18:09:22 tbachman: but no leads 18:09:34 SumitNaiksatam: did you talk to any Infra folks? 18:09:43 tbachman: interesting thing is that it was against the GBP projects 18:09:47 tbachman: no i didnt 18:09:57 tbachman: should have 18:09:58 * tbachman blames it on russian hackers ;) 18:10:11 lol 18:10:39 tbachman: the openstack wiki system was heavily under attack 18:11:04 yeah -- I recalll that 18:11:09 not sure what that is 18:11:15 but certainly doesn't seem legitimate 18:11:19 tbachman: yeah 18:11:21 and not sure how/why it's happening 18:11:29 so regarding the above patch 18:11:36 i got comments from rkukura and tbachman 18:12:05 the queueing in this patch is currently triggered only for aim_mapping driver 18:12:27 but the goal is to eventually move the resource_mapping driver also to a async style like the aim_mapping driver 18:12:38 and we would use this queueing over there as well 18:13:15 regarding rkukura’s comment about extending this to other notifications which neutron makes, but original goal was to only address the notifications we send from local_api 18:13:39 and my understanding was that it would fix the DHCP issue we were facing by doing this 18:14:08 queueing up all notifications going out from the neutron server is more challenging 18:14:16 SumitNaiksatam: I agree we can change this later if we need to queue other notifications, such as those made by ML2 18:14:22 rkukura: okay 18:14:30 rkukura: i have been thinking about how we can do that 18:14:41 rkukura: but will require some more prototyping 18:15:12 for now i am hoping that we can make progress with the framework i have proposed in the patch, and validate tif it actually fixes the issue 18:15:14 SumitNaiksatam: Quick question: do notifications only get queued if there is a transaction? I.e. do they get sent immediately if there is no transaction? 18:15:59 rkukura: good question, currently everything that gets called from that local_api module is within a transaction 18:16:10 rkukura: at least to the extent that i observed 18:16:54 So I’m a bit concerns something could get queued but not sent until the next transaction 18:16:57 rkukura: so if a notification is not being pushed from the local_api, it will not get queued 18:17:10 rkukura: no, that wont happen 18:17:19 s/concerns/concerned/ 18:17:34 SumitNaiksatam: dumb question - do you listen for these notifications on rabbit and fwd them to APIC? 18:17:45 rkukura: i can check that the notification_queue is empty at the end of the test 18:18:21 songole: no, we just queue up the notifications before sending them to rabbit 18:18:55 songole: there is kind of an interceptor which queues up the calls that were intended to go the message queue 18:19:32 rkukura: in my debugging i had checked that all notifications were always getting sent out, as there were no notifications that remained in the queue forever 18:19:43 SumitNaiksatam: I kind of think the add_notification_to_queue should just send immediately if transaction_key is None, rather than returning 18:19:45 SumitNaiksatam: Ah ok. thanks 18:20:09 rkukura: okay 18:21:12 songole: i can explain offline if you have more questions 18:21:47 thank you 18:21:48 in the UTs i am trying to compare the order and content of the notifications between those sent when there is no queuing versus those when there is queuing 18:22:00 in case it was not apparent from the UTs 18:22:27 the patching of the uuid and mac generation in the UTs to aid in comparison 18:22:52 so that i get the same resource identities in successive runs, and i can just compare them directly 18:23:39 any more thoughts/concerns on this patch? 18:24:23 #topic Open Discussion 18:24:32 anything else we need to discuss today? 18:24:46 i wont be travelling to Barcelona 18:25:39 i believe Ivar will be there, so he will have to keep the GBP flag flying :-) 18:25:44 Any GBP sessions planned during the Summit? 18:26:00 songole: i havent planned anything at my end since i wont be travelling 18:26:15 songole: if you are planning to attend, please feel free to schedule 18:26:30 my qos patch failed to merge, I haven't yet checked the logs but will do soon 18:26:43 songole: i think MikeC might have some talks in the vendor forum 18:26:53 SumitNaiksatam: I am not traveling either. Prasad is going 18:26:53 igordcard: great thanks 18:27:13 igordcard: let me know if you need me to look at anything, sorry we should have merged this patch much earlier 18:27:19 songole: okay 18:27:51 SumitNaiksatam: okay, thanks 18:27:58 igordcard: i have feeling this has something to do with the requirements or upper-constraints 18:28:06 alright thanks all 18:28:09 see you next time! 18:28:10 bye 18:28:17 SumitNaiksatam: bye! 18:28:25 bye 18:28:25 bye 18:28:26 #endmeeting