15:01:29 #startmeeting neutron_l3 15:01:29 o/ 15:01:29 Meeting started Thu Feb 26 15:01:29 2015 UTC and is due to finish in 60 minutes. The chair is carl_baldwin. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:30 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:31 hi 15:01:33 The meeting name has been set to 'neutron_l3' 15:01:38 #topic Announcements 15:01:44 #link https://wiki.openstack.org/wiki/Meetings/Neutron-L3-Subteam 15:01:59 March 19th is the big day. Kilo-3 15:02:05 #link https://wiki.openstack.org/wiki/Kilo_Release_Schedule 15:02:11 Any other announcements? 15:02:34 #topic Bugs 15:02:43 Anything super-important here? 15:03:23 If not, we’ll move on and spend a little extra time on bugs next week. 15:03:40 #topic L3 Agent Restructuring 15:04:37 My last patch here is https://review.openstack.org/#/c/158495/. I need to turn my attention mostly away from this so I’ll just be care-and-feeding the patches that I currently have up. 15:04:55 I know there are loose ends and I encourage you to continue working on those loose ends. 15:05:02 mlavalle: Your patch is next in line. 15:05:14 mlavalle: I liked it overall when I reviewed it the other day. 15:05:18 mlavalle: Anything to discuss? 15:05:30 carl_baldwin: I saw your feedbacb and will work on it today. That's all 15:05:36 for today 15:05:38 #link https://review.openstack.org/#/c/147744/26 15:05:59 mlavalle: Thanks. I didn’t see anything major in there. Just minor cleanup. 15:06:05 :-) 15:06:59 Anything else? 15:07:30 #topic neutron-ipam 15:07:43 johnbelamaric: tidwellr: salv-orlando: ping 15:07:50 carl_baldwin: pong 15:07:56 pong 15:08:14 Did we reach a conclusion on the discussion from yesterday? 15:08:31 I believe so - it's in the review comments now. 15:08:38 I think so 15:09:10 any status on the reference driver? salv-orlando here? 15:09:20 johnbelamaric: Do you have a link handy? I haven’t seen the review since the comments have been added. 15:09:26 one sec 15:09:57 #link https://review.openstack.org/#/c/153236/8/neutron/db/db_base_plugin_v2.py that one? 15:10:00 johnbelamaric: I thought we made some progress in salv-orlando ’s ML thread from the week. 15:10:57 pavel_bondar: yes, that's it 15:11:17 pavel_bondar: Thanks, I think johnbelamaric ’s comment there is correct. 15:11:23 carl_baldwin: yes, I think progress is being made i just haven't seen a new patch since then. maybe I missed it 15:11:59 johnbelamaric: To me, the progress was more around narrowing down the scope for Kilo so that we don’t have to worry about all of the possibilities. 15:12:17 carl_baldwin: ah yes, I see your point. 15:12:24 I haven’t seen a new patch either but I expect one will come soon. 15:12:36 pavel_bondar: how is the dbbase plugin coming - are you blocked waiting on reference driver or you can proceed? 15:13:10 I still have work to do, but as soon as refence driver appears I could start testing 15:14:09 biggest part that left to do is rollback on exception, since ipam call is done in separate transaction 15:14:22 pavel_bondar: ok 15:14:35 aloha people. sorry for being late. 15:14:48 salv-orlando: hi 15:15:02 salv-orlando: hi 15:15:32 salv-orlando: hi 15:16:30 salv-orlando: Are there still open issues from the ML thread you started this week? 15:17:03 carl_baldwin: that thread has been inconclusive so far, which means 15:17:16 that for kilo we'll just adopt the same algorithm that we have, based on av ranges 15:17:24 adding only retry on deadlock 15:17:27 easy and simple 15:17:38 also reliable. Maybe not efficient but reliable 15:17:50 we can have fun in the next release and experiment with different things 15:18:20 salv-orlando: I agree. I think that is one mini-conclusion that came out of the thread. 15:18:34 salv-orlando: You’re right though, the meat of the discussion was inconclusive. 15:19:54 carl_baldwin: it's not easy to reach any sort of agreement on this matter. 15:20:19 salv-orlando: indedd 15:20:19 salv-orlando: with pluggable, we don't need to - we can have multiple drivers and let experience show which algo is best :) 15:20:25 s/indedd/indeed/ 15:21:31 johnbelamaric: yeah pretty much. But we're also free to change the internal behaviour of the driver without worrying about affecting anything at the mgmt layer 15:21:55 salv-orlando: yes, good point too 15:22:06 so I'm not worried at all about just sticking with what we have - and defer optimization to the next release 15:22:56 salv-orlando: I think we’re all in agreement. I’m glad to have a direction that could get us to Kilo. 15:23:11 Anything more to discuss? 15:23:28 one question about passing subnetpool_id, since it should be property of Pool, it should present in Pools init(), right? 15:24:16 So should we update Pool interface and include it into __init__? 15:24:45 I can add it to interface review page, and we could discuss it there 15:25:22 pavel_bondar: Yes, please do. 15:25:47 ok 15:26:28 Anything else to discuss here? 15:27:06 #topic neutron-ovs-dvr 15:27:14 carl_baldwin: see #link https://review.openstack.org/#/c/147479/14/neutron/manager.py 15:27:15 mrsmith: Swami: Rajeev: ping 15:27:25 Pong 15:27:27 for a discussion of subnetpool_id, driver laoding, etc 15:27:39 #undo 15:27:40 Removing item from minutes: 15:27:59 * carl_baldwin looking... 15:28:05 carl_baldwin: doesn't have to be now - can be addressed later 15:28:25 carl_baldwin: just wanted to bring it to your attention in relation to pavel_bondar's question 15:28:35 johnbelamaric: Thanks for pointing it out. It is still early for me here, I hadn’t seen it. 15:29:02 #topic neutron-ovs-dvr 15:29:11 mrsmith: hi, anything to discuss here? 15:29:42 Working on the non-voting test diffs between centralized 15:29:43 Rajeev got me looking at a bug that might be due to the weak reference tracking of the fip namespace. I don’t have any insight yet. 15:29:48 Slow progress 15:29:54 Ok 15:30:21 #link https://bugs.launchpad.net/neutron/+bug/1425639 15:30:23 Launchpad bug 1425639 in neutron "A functional floating ip stops working sometimes if another floating ip is created and deleted." [Undecided,In progress] - Assigned to Rajeev Grover (rajeev-grover) 15:30:23 There has been some improvement from swamis patches 15:30:34 there's those 3 DVR bugs I found, I don't have time to add tests to those patches. 15:31:02 amuller: Could you link them here for posterity? 15:31:16 coming up 15:31:29 I was swamped last week and didn’t have time to look at them at all. Sorry about that. 15:31:49 #link https://bugs.launchpad.net/neutron/+bug/1423777 15:31:50 Launchpad bug 1423777 in neutron "TRACE when removing a floating IP from a DVR router that has no floating IPs" [Undecided,In progress] - Assigned to Assaf Muller (amuller) 15:31:59 #link https://bugs.launchpad.net/neutron/+bug/1423775 15:32:00 Launchpad bug 1423775 in neutron "TRACE when restarting openvswitch if using OVS DVR agent" [Undecided,In progress] - Assigned to Assaf Muller (amuller) 15:32:05 #link https://bugs.launchpad.net/neutron/+bug/1423774 15:32:06 Launchpad bug 1423774 in neutron "Fix TRACE when removing a DVR router port from a VLAN network" [Undecided,In progress] - Assigned to Assaf Muller (amuller) 15:32:15 amuller: Thanks. 15:32:48 all 3 have patchs up, marun -1 because they have no tests to mitigate regressions. I agree with him but I don't have time to deal with that. 15:33:00 amuller: ack 15:33:02 I was hoping someone could take those off my hands 15:33:17 I can try to find some cycles 15:33:29 mrsmith: thanks 15:33:30 I'll take a look at least 15:34:04 Functional or unit? 15:34:24 mrsmith: I’ll keep looking in to bug 1425639. I suspect we’ll need to get some more logging added for when we can catch the bug in the gate and check queues. 15:34:25 bug 1425639 in neutron "A functional floating ip stops working sometimes if another floating ip is created and deleted." [Undecided,In progress] https://launchpad.net/bugs/1425639 - Assigned to Rajeev Grover (rajeev-grover) 15:35:36 amuller: functional or unit tests? 15:35:49 mrsmith: depends on the case 15:35:56 K 15:36:13 whatever is more appropriate. If you find yourself mocking out the entire world maybe a functional test would be better 15:37:48 Anything else on DVR to discuss here? 15:38:52 #topic Open Discussion 15:39:55 carl_baldwin: pc_m just +2'd https://review.openstack.org/#/q/topic:bug/1425759,n,z 15:39:56 on the vpnaas repo 15:40:17 I want to ask to merge these ASAP so we can start moving from though gigantic chain before K3... 15:40:27 moving through* 15:40:28 amuller: I think we can get those in quickly. 15:41:04 carl_baldwin: Sounds good :) 15:41:27 I think we'll be OK, although am a little concerned as the impression is that we'll have seperate processes, for the AdvancedService instance, if > 1 agent for a service, however the test failure seems to imply instance was shared by two processes. 15:42:15 a singleton cannot be shared across different processes 15:42:22 Not an issue currently, as we have only one agent per service. Something we need to keep an eye on though. 15:42:58 amuller: Right, so I'm having a hard time understanding how the functional test failed. 15:43:32 Me too :) The only thing I can think of is that somehow, two tests were running at the same time in the same process (Which shouldn't happen according to my understanding) 15:43:48 or there was some shared point in time where one tests did not finish and the other already started 15:43:54 that would screw up the first test 15:44:21 though that would imply two agents in same process... 15:44:22 one test* 15:44:29 pc_m: each test has its own agent 15:44:34 initialized in the test setUp 15:44:48 pc_m: I think the test runner may run multiple tests in the same process. I don’t know for sure how it works but it doesn’t surprise me. 15:45:13 mind you we dont actually start the L3 agent like you do through the CLI, but the test setUp does create a L3 agent instance, which initializes its own self.conf 15:45:14 carl_baldwin: must be. Only way I could account for the issue. 15:45:21 so each L3 agent instance has its own conf object 15:49:16 carl_baldwin: who can I contact for FWaaS repo merge? 15:49:41 amuller: dougwig, SumitNaiksatam 15:50:22 amuller: https://review.openstack.org/#/admin/groups/500,members 15:50:53 amuller: also any neutron drivers 15:51:07 salv-orlando: Like yourself? :) 15:51:17 as long as I'm there, yes 15:55:32 I think we’ll call it a meeting. 15:55:49 Thanks, everyone for all your work. 15:56:08 #endmeeting