15:01:49 #startmeeting neutron_l3 15:01:50 Meeting started Thu Nov 30 15:01:49 2017 UTC and is due to finish in 60 minutes. The chair is mlavalle. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:51 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:53 The meeting name has been set to 'neutron_l3' 15:02:09 hi 15:02:13 hi 15:02:15 o/ 15:02:28 sorry for the delay, the Sahara guys were wrapping up their meeting 15:02:52 #topic Announcements 15:03:35 First announcement is that the Q-2 milestone will be next week 15:04:04 after that, we are on the home stretch of the Queens cycle 15:04:44 Second annoucement is that the Rocky PTG is open to buy tickets and make hotel reservations 15:05:31 #link https://www.openstack.org/ptg/ 15:05:59 the venue is a soccer stadium that also has meeting rooms to host conferences 15:06:24 nice 15:06:39 s/soccer/football right? :) 15:07:17 I made my reservation at the event hotel yesterday, to make sure I get the discounted price of 155 euros and comply with my employer policies 15:07:59 haleyb: lol, you made doubt whether my subconcious had betrayed me.... 15:08:10 i just got approved today so will be doing the same. i will only be there wed-fri though, the ptg always conflicts with my vacation 15:08:29 Sorry I'm late, just got out of a meeting @.@ 15:08:32 that rate includes breakfasts and taxes 15:08:47 so that is the net net rate 15:09:38 haleyb: nice, we can have go to the Irish pubs with davidsha to discuss kick ass dvr 15:09:47 :D 15:09:55 any other announcements? 15:10:33 ok, moving on 15:10:40 #topic Bugs 15:10:49 Swami: please fire away 15:10:57 mlavalle: thanks 15:11:24 #link https://bugs.launchpad.net/neutron/+bug/1717302 15:11:24 Launchpad bug 1717302 in neutron "Tempest floatingip scenario tests failing on DVR Multinode setup with HA" [High,Confirmed] 15:12:05 mlavalle, haleyb: I was able to debug this issue further and found some issues in the floatingip translation from the router namespace to the fip namespace. 15:12:21 Swami: great! 15:12:26 I have captured all my findings in the bug report. 15:13:03 The rules are configured in the router namespace and the packet is getting out of the router namespace with a private ip and so it is not getting in sync. 15:13:19 I see both the nodes reponding to the ping, but they are not syncing up. 15:14:19 haleyb: Do you know why this would happen or have you seen similar issues. 15:14:21 i will have to read through all the notes 15:15:02 This is very hard to reproduce, because out of the four tests either one fails randomly. The only test that does not fail is the 'without FIP' condition for both SRC and DEST. 15:15:30 sorry I meant hard to debug, but easy to reproduce, you will get a failure at least once in two runs. 15:16:29 Swami: thanks for debugging further 15:16:33 haleyb: I think the culprit is the floatingip-chain in the router-namespace, but let me look further and update. In the meanwhile if you see any thing let me know or update the bug report. 15:17:09 So I will continue my debugging on this. 15:17:15 Thanks! 15:17:45 #link https://bugs.launchpad.net/neutron/+bug/1733987 15:17:45 Launchpad bug 1733987 in neutron "name resolution error with DVR+HA routers" [Medium,In progress] - Assigned to Armando Migliaccio (armando-migliaccio) 15:17:55 There is a patch up for review for this bug. 15:18:45 #link https://review.openstack.org/#/c/522362/ 15:19:05 i see armando updated it, will take another look 15:19:19 Nothing else to discuss on this patch, since we already have some review comments from anil and haleyb 15:19:44 haleyb: ok, thanks. 15:20:11 #link https://bugs.launchpad.net/neutron/+bug/1733852 15:20:11 Launchpad bug 1733852 in neutron "Incorrect ARP entries in new DVR routers for Octavia VRRP addresses" [Medium,In progress] - Assigned to Daniel Russell (danielr-2) 15:20:47 Sorry I just noticed this issue. I am backlogged from my vacation in Sydney. 15:20:58 haleyb: thanks for triaging this. 15:21:12 np, i just haven't looked at the review yet 15:21:38 I need to take a further look at this and see what is triggering this ARP entry. May be we might have some code that needs cleanup on the server side. 15:21:55 But I will update the bug report after doing my testing. 15:22:07 haleyb: I did see that the patch was abandoned. 15:22:27 there's a new one, don't know why the first was abandoned 15:22:28 there is another one: 15:22:52 #link https://review.openstack.org/#/c/524037/ 15:23:06 yeah, that seems to be the good one 15:23:15 mlavalle: yes this patch is still active against the bug. I will take a look at it. The other one has been abandoned. 15:23:59 #link .launchpad.net/neutron/+bug/1731857 15:24:26 #link https://bugs.launchpad.net/neutron/+bug/1731857 15:24:26 Launchpad bug 1731857 in neutron "DVR scenario tests fail in default deployment" [Low,New] 15:24:51 The bug says there is a mismatch in the default settings with DVR and legacy routers. 15:25:23 I need to check this out. I might not have tested with legacy routers. 15:26:36 for the migration tests it would seem legacy was not a good choice, so maybe it is just a quick fix? 15:27:00 haleyb: yes 15:27:31 #link https://bugs.launchpad.net/neutron/+bug/1728665 15:27:31 Launchpad bug 1728665 in neutron "Removing gateway ip for tenant network (DVR) causes traceback in neutron-openvswitch-agent" [Medium,In progress] - Assigned to Brian Haley (brian-haley) 15:28:11 #link https://review.openstack.org/#/c/521199/ 15:28:18 There is a patch up for review. 15:28:21 Swami: i had forgoteen about that one, will look at your comments 15:28:31 haleyb: no problem. 15:29:17 #link https://bugs.launchpad.net/neutron/+bug/1716401 15:29:17 Launchpad bug 1716401 in neutron "FWaaS: Ip tables rules do not get updated in case of distributed virtual routers (DVR)" [Undecided,New] 15:30:03 I haven't got any information on this from the FWaaS team. I tried to sync up with Sridhar in Sydney but he was busy and mentioned that he will through this issue with me later. 15:30:10 So this is still pending. 15:30:24 mlavalle: may be you can bring this up in the FWaaS meeting. 15:30:34 That's all I had for today's meeting. 15:30:36 Swami: ok, I will 15:30:46 mlavalle: back to you 15:31:28 on my side we have https://bugs.launchpad.net/neutron/+bug/1706750 15:31:28 Launchpad bug 1706750 in neutron "FlushError for automatic IPAllocations when subnet create is retried" [High,Confirmed] - Assigned to Ihar Hrachyshka (ihar-hrachyshka) 15:31:47 ihrachys is wroking on it 15:31:55 he has two WIP patchsets 15:32:13 #link https://review.openstack.org/518988 15:32:29 #link https://review.openstack.org/518989 15:32:40 so I guess keep an eye on them 15:33:20 and that's all I got 15:33:27 any other bucgs from the team? 15:33:42 i have one 15:33:57 https://bugs.launchpad.net/neutron/+bug/1731595 15:33:57 Launchpad bug 1731595 in neutron "L3 HA: multiple agents are active at the same time" [High,In progress] - Assigned to venkata anil (anil-venkata) 15:34:17 anil has a patch that addresses the problem, just needs an update with a test 15:34:55 is that the one where he has patgchset for master and stable branches? 15:35:18 mlavalle: yes, he proposed stable ones just to test, and make it easier to pick downstream for testing 15:35:38 ok, cool, they are in my radar screen then 15:36:30 that's it from me 15:36:47 ok thanks for the updates 15:36:58 #topic Kick-ass DVR 15:37:28 Thanks, I'm working on breaking out the roter_info_base file into a seperate patch atm 15:38:30 There are a few calls outside of the routers I'm looking into absorbing into the base router classes 15:38:51 An example of some are these calls: https://github.com/openstack/neutron/blob/master/neutron/agent/metadata/driver.py#L283-L324 15:39:24 mlavalle: haleyb: I need to drop off my son at school, so will signogg 15:39:36 s/signogg/signoff 15:39:39 Swami: go ahead, thanks for the updates :-) 15:39:43 Just as a general question, are these functions ok to move into router info base? 15:39:48 Swami: ok, thanks for the updates - jinx 15:39:51 Swami: cya! 15:40:55 By moving them into router_info, I'd hope that we could abstract out the use of iptables from the main l3-agent 15:41:41 it seems worth exploring to me 15:42:04 they are not even part of the meta driver per se 15:42:30 yeah, they are self-contained 15:44:42 Kk, I'll try to have this patch up for review by next Friday. Are both of you ok with the idea of trying to abstract iptables calls into the routers and out of the Agent? 15:45:36 well it seems to me that is an effort to start re-factoring code in prepration for adopting the openflow approach, while we fix our DVR mess 15:45:56 under that assumption, yes 15:45:57 yes, it is fine with me 15:47:02 cool, thanks. thats all from me unless there are any questions? 15:47:58 davidsha: none from me. Thanks for continuing with this. We are aware your priorities have been changing lately and highly appreciate the effort 15:49:02 let's move on 15:49:28 #topic QoS for floating ips 15:49:55 The patchset for the server side merged yesterday. Thanks haleyb for the reviews :-) 15:51:12 The agent side is here: 15:51:18 #link https://review.openstack.org/#/c/521079/ 15:52:04 and and seems ripe for reviews 15:52:34 I am thinking we also need some docs, either added to this last patchset or in another one 15:53:36 and the python openstackclient patch is here: 15:53:49 #link https://review.openstack.org/#/c/517845/ 15:55:25 ok, moving on 15:55:32 #topic Open agenda 15:55:49 Does anyone have any toher topics to discuss today? 15:56:16 I have one 15:56:38 haleyb: I took a look at this review: https://review.openstack.org/#/c/516055/ 15:57:08 it seems good to me. The question is, won't we break someone who is using the extension as it is today? 15:58:18 mlavalle: yes, i thought about that, but it seems they are maybe broken as-is, so perhaps noone is using it much? 15:58:46 that was my thinking also but I wanted to check anyway 15:59:04 I also think we need some sort of doc 15:59:09 Just before we run out of time! 15:59:23 for the extension, because we have none now 15:59:24 I have one also, Someone has been assigned to take over the DVR patch, I'll ask him to try and join this meeting next week, but he is in China I believe, so it is late for them. 15:59:46 davidsha: ok, thanks for the heads up 15:59:47 (feel free to run over a bit, we api-sig people in no rush) 16:00:05 cdent: I think we are done 16:00:10 mlavalle: doc the metering extension? sure 16:00:16 * cdent WAVES AT MLAVALLE 16:00:16 #endmeeing 16:00:31 * mlavalle waves back at cdent 16:00:37 * edleafe waves at mlavalle too 16:00:40 #endmeeting