18:34:55 #startmeeting networking_policy 18:34:57 Meeting started Thu Dec 7 18:34:55 2017 UTC and is due to finish in 60 minutes. The chair is SumitNaiksatam. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:34:59 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:35:01 The meeting name has been set to 'networking_policy' 18:35:26 mainly wanted to provide an update on the Pike sync work, and also discuss the outstanding patches 18:35:35 lets start with rkukura’s patch 18:35:47 ok 18:35:56 #link https://review.openstack.org/#/c/526158/ 18:36:12 rkukura: thanks for doing this, it looked good to me 18:36:22 was just waiting for zuul to report 18:36:26 I will probably have a minor update 18:36:26 will vote on it shortly 18:36:32 rkukura: ah good to know 18:37:04 this patches hopefully fixes some of the long standing issues we have had with running UTs independently 18:37:09 The back-port to newton is non-trivial, and it turned up that all_models.py was not importing some models used for testing 18:37:34 I’d like to get to where any combination of tests passes on all the active branches 18:37:39 rkukura: that later was my thought as well, i wanted to check if it indeed does import everything 18:37:44 rkukura: right 18:37:53 rkukura: whats the issue with newton? 18:38:19 one was that the apic_mapping driver didn’t exist on newer branches, and needed fixing 18:38:37 rkukura: ah, painful 18:38:43 that one is easy 18:39:18 but my attempted back-port turned up some missing DB table type errors again, so that got me tracking down what I might have missed first time 18:39:53 I’ve been the entire set of UTs, and each UT module individually, but should also be testing each directory of UTs 18:40:05 s/been/been testing/ 18:40:24 so I hope to have working patches for all three branches today 18:40:58 rkukura: nice, been on that path of testing those combinations, its painful 18:41:12 takes time :( 18:41:16 yeah 18:41:38 rkukura: actually i had a patch earlier where i was trying to eliminate the calls to create_tables 18:41:53 but it kept failing 18:42:01 i believe the reason was that i was not importing everytning 18:42:10 (after looking at your patch) 18:42:13 could be 18:42:32 i had attempted this about a year back, i obviously gave up! 18:42:37 at least the (internal) AIM stuff and the NFP stuff are separate 18:42:39 so thanks for doing this 18:42:57 no problem, but I do need to get back to the AIM validation/repair work 18:43:05 rkukura: yeah 18:43:05 +1 thanks, it will help! 18:43:19 on the pike sync 18:43:35 #link https://review.openstack.org/#/c/514864/ 18:43:41 on the above patch, i will fix the pep8 error 18:43:48 it creepd in after rebase 18:44:09 annakk: i responded to your comment, and incorporated rkukura’s comment which came later 18:44:30 so hopefully once i uploaded the pep8 fix you guys can take a quick look again 18:44:40 will do 18:44:51 thanks 18:44:53 regarding the sync patch #link https://review.openstack.org/#/c/518183/ 18:44:57 yes, you're right, since subnet pools might have added in current session, its important to keep the session 18:45:06 annakk: okay 18:45:08 i am down to 75 failures 18:45:19 so couple of things to discuss about that 18:45:28 most of the failures i am seeing now pass locally 18:45:42 there are a couple of tests which fail locally as well 18:45:58 and my current theory is that those tests are not cleaning up if they fail 18:46:08 and hence configuration gets carried over 18:46:16 and causes other tests to fail 18:46:31 so i am trying to get to the bottom of this 18:47:01 the heirarchy if UT TestCase classes is a mess! 18:47:13 the other thing is, i would like to draw your attention to this ugly (temporary) hack: https://review.openstack.org/#/c/518183/14/gbpservice/neutron/plugins/ml2plus/plugin.py 18:47:14 * tbachman nods 18:47:15 hierarchy of … 18:47:21 refer to L261 18:47:25 on the right hand side 18:48:40 i tried to debug this, and i couldnt tell where else a delete was being callled, but when it hits this point of code, some UTs complain that the “NetworkSegment” object has already been addded for deletion 18:49:04 if i just ignore this exception, the UTs pass 18:49:22 if you guys have thoughts on this please do let me know 18:49:56 a way to reproduce this would be to remove this try-except block and run the UT that i mention in the comments there 18:50:21 looks suspicious :) 18:50:36 annakk: yeah, it doesnt look good 18:51:01 I have thoughts, but they are not productive ;) 18:51:18 lol 18:51:27 i am trying to get to a point where i can get all the UTs to pass (albeit with this hack) and then we can perhaps use our collective wisdom to crack this one 18:51:31 rkukura: lol 18:51:39 rkukura: you can let us decide that 18:52:47 i dont have anything more on this, happy to discuss the patch in the time that we have (assuming you got a chance to look at it at all) 18:52:54 The neutron team took ML2’s concept of network segments and both promoted it to the core and hijacked it for the routed-networking stuff, which is entirely inconsistent with ML2’s concept 18:53:22 so I’m not surprised its causing trouble now 18:53:36 rkukura: ah okay, good data point, so perhaps might need a bit more offline discussion with you 18:53:47 so far this happens only with the NetworkSegment DB object 18:54:38 annakk: tbachman: anything you would like top bring up at your end? 18:54:46 SumitNaiksatam: nothing from me 18:55:06 tbachman: thanks, one more patch from you coming? 18:55:06 not from me 18:55:11 annakk: thanks 18:55:21 SumitNaiksatam: ACK. I have one more fix in GBP for EP updates 18:55:30 that will likely post tomorrow 18:55:31 tbachman: ok thanks, will look for it 18:55:35 SumitNaiksatam: thx! 18:55:44 rkukura: if nothing else from you, we can wrap up for today! 18:55:50 sure 18:55:54 thanks SumitNaiksatam! 18:55:55 SumitNaiksatam: thanks! 18:56:00 thanks all! 18:56:02 bye! 18:56:03 thanks, bye! 18:56:04 #endmeeting