15:00:24 <haleyb> #startmeeting neutron_dvr
15:00:25 <openstack> Meeting started Wed Jul  6 15:00:24 2016 UTC and is due to finish in 60 minutes.  The chair is haleyb. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:26 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:29 <openstack> The meeting name has been set to 'neutron_dvr'
15:00:33 <haleyb> #chair Swami
15:00:37 <openstack> Current chairs: Swami haleyb
15:01:38 <haleyb> #topic Announcements
15:02:09 <haleyb> there's finally a quorum - welcome back Swami
15:02:17 <Swami> sure
15:02:58 <haleyb> let's get right to bugs
15:03:01 <haleyb> #topic Bugs
15:03:54 <Swami> yes, there are couple of new bugs
15:04:18 <Swami> #link https://bugs.launchpad.net/neutron/+bug/1597561
15:04:19 <openstack> Launchpad bug 1597561 in neutron "L3 agent allows multiple gateway ports in fip namespace" [High,In progress] - Assigned to Brian Haley (brian-haley)
15:04:52 <haleyb> Carl actually did the work, I just edited the patch to fix a nit
15:05:07 <haleyb> https://review.openstack.org/#/c/335755/
15:05:17 <Swami> I did see it.
15:05:50 <Swami> I was not sure, how this occurred. But the patch at least checks for the stale and removes it. Looks good.
15:05:59 <haleyb> and oleg just gave it a +A
15:06:36 <Swami> The next in the list is.
15:06:40 <Swami> #link https://bugs.launchpad.net/neutron/+bug/1599089
15:06:40 <openstack> Launchpad bug 1599089 in neutron "DVR: floating ip stops working after reassignment" [High,In progress] - Assigned to Oleg Bondarev (obondarev)
15:06:53 <Swami> There is also a patch for this bug pushed in by Oleg.
15:07:15 <haleyb> https://review.openstack.org/#/c/337591/
15:07:27 <Swami> Thanks for posting the patch.
15:08:03 <Swami> I did see that you have already reviewed it. I will take a look at it again.
15:08:12 <Swami> It should be good to go.
15:08:33 <haleyb> yes, that one looks good
15:08:42 <Swami> The next in the list is
15:08:46 <Swami> #link https://bugs.launchpad.net/neutron/+bug/1593354
15:08:46 <openstack> Launchpad bug 1593354 in neutron "SNAT HA failed because of missing nat rule in snat namespace iptable" [Undecided,New]
15:09:11 <Swami> This seems to be like a duplicate, but will check with adolfo, before we tag the priority on this bug.
15:11:06 <Swami> assaf has posted a question in there to check if it is a duplicate of #link https://bugs.launchpad.net/neutron/+bug/1571113
15:11:06 <openstack> Launchpad bug 1571113 in neutron "SNAT interface not created for dvrha in some scenarios" [Medium,Fix released] - Assigned to Adolfo Duarte (adolfo-duarte)
15:11:46 <haleyb> ok
15:11:50 <Swami> adolfo-duarte has already addressed that bug, but will re-check with adolfo again and will triage the bug.
15:12:30 <Swami> The next one is performance related to DVR and IPv6.
15:12:34 <Swami> #link https://bugs.launchpad.net/neutron/+bug/1596473
15:12:34 <openstack> Launchpad bug 1596473 in neutron "Packet loss with DVR and IPv6" [Undecided,New]
15:13:19 <Swami> This need to be triaged. I am not sure we have seen such a degradation in performance.
15:14:07 <haleyb> right, looks like packets are going multiple places.  i had asked about reproducing in a newer release but had no answer
15:14:25 <Swami> haleyb: yes I saw your message.
15:15:03 <Swami> So let us wait and see or check on the master branch and prioritize based on the findings.
15:15:20 <haleyb> IPv6 with DVR should only route externally via the CVR
15:15:57 <Swami> haleyb: Is that what that was reported as performance degradation.
15:17:19 <Swami> The next in the list is
15:17:21 <haleyb> i can't tell, but it seems like if they force a route via the CVR everything works
15:18:29 <Swami__> got disconnected.
15:18:38 <Swami__> The next bug is
15:18:54 <Swami__> #link https://bugs.launchpad.net/neutron/+bug/1599287
15:18:54 <openstack> Launchpad bug 1599287 in neutron "Cleanup snat redirect rules when agent restarts after stale snat namespace is cleaned." [Undecided,In progress] - Assigned to Swaminathan Vasudevan (swaminathan-vasudevan)
15:20:10 <haleyb> https://review.openstack.org/337855
15:20:52 <Swami__> haleyb: I got some push back from assaf and oleg on this patch since it is too much to cache and use it later.
15:21:24 <Swami__> Let me see if there is a way to delete all the ip rule and default gateway without the snat_idx.
15:22:01 <haleyb> Swami__: yes, if you can somehow just cleanup all the rules it would be better, will be more work though
15:23:20 <Swami__> haleyb: I thought caching would help us, to re-instantiate if someone deletes the namespace and if we wanted to recreate those rules, but may be we can do a full-sync at that time to recreate the rules.
15:23:54 <Swami__> haleyb: This is again similar to the fip and its priority retained in the localfilesystem to handle reboots.
15:24:25 <Swami__> haleyb: But that might have a different use cases.
15:24:34 <Swami__> s/use cases/use case
15:25:07 <haleyb> yes, and i have a feeling if we proposed the fip priority code today it might get push-back.  Best to try and sync if possible
15:25:27 <Swami__> haleyb: ok will try to see what can be done to it.
15:25:36 <Swami__> haleyb: one the same lines, #link https://review.openstack.org/#/c/326729/
15:26:22 <Swami__> After removing the check for the namespace exists check in create, the jenkins is passing on this patch. What do you think. Is there any other work on this patch or it can merge.
15:27:05 <haleyb> Swami__: yes, i saw the update.  I did just have a question on self.snat_ports usage since it's initialized differently
15:27:43 <Swami__> haleyb: I know I saw your comments, about the empty list and none used at different places.
15:27:58 <Swami__> haleyb: I think you lean towards the empty list rather than none.
15:28:17 <Swami__> So, should we make that fix in a different patch or add it to this patch.
15:28:39 <haleyb> Swami__: well, i think they should be the same, right?  then maybe one of the changes isn't necessary?
15:29:23 <Swami__> haleyb: what do you mean one of the changes isn't necessary
15:29:56 <Swami__> So what you are saying is if we define the variable in dvr_local_router, we don't need to refine it in the dvr_edge_router
15:30:04 <haleyb> https://review.openstack.org/#/c/326729/18/neutron/agent/l3/dvr_local_router.py - you check self.snat_ports, but don't think you need to
15:30:06 <Swami__> s/refine/redefine
15:30:50 <haleyb> get_snat_port_for_internal_port() can take None or [] as an argument
15:31:49 <haleyb> I didn't know if it was related to one place being None and the other []
15:32:00 <Swami__> so the best option is to change it to None.
15:32:25 <Swami__> I meant change it to None in dvr_local_router as well.
15:33:02 <haleyb> if you make it [] then we might not need another change in the dvr ha edge router code
15:33:24 <haleyb> because it does a 'for port in self.snat_ports' which will blow-up with None
15:33:48 <Swami__> haleyb: ok let me check that out and see what makes sense.
15:34:02 <haleyb> I know it's been a while since you made that change, so I was trying to figure out what error you saw
15:35:33 <Swami__> haleyb: Yes basically the only issue that I saw was unable to iterate on self.snat_ports, since it would be None.
15:35:59 <Swami__> That's all I had for bugs right now.
15:36:18 <haleyb> So perhaps it should be [], then there's just a question of should it be in the parent class, but we can discuss that in the review
15:36:33 <Swami__> haleyb: Ok, will address it in the review.
15:36:58 <Swami__> The other thing that is pending for a long time is #link https://review.openstack.org/#/c/275073/
15:37:39 <haleyb> Swami__: did the tempest patch merge?
15:37:46 <Swami__> This is the nova patch and probably we should be pushing it before the nova mid cycle, I don't want this to caught up.
15:37:53 <Swami__> haleyb: no it did not merge.
15:38:40 <Swami__> haleyb: can you ping tempest core to take a look at that patch.
15:39:45 <haleyb> Swami__: i can once i find the patch
15:40:14 <Swami__> #link https://review.openstack.org/#/c/286855/
15:40:19 <Swami__> tempest patch link above
15:40:38 <haleyb> thanks
15:41:56 <haleyb> Swami__: i saw you update the WIP patch for "fast exit", https://review.openstack.org/#/c/297468/ - is that ready for review?
15:42:05 <Swami__> haleyb: Once I finish of this patch, I might be focussing on the fast path exit.
15:42:15 <haleyb> great, thanks
15:42:21 <Swami__> Yes, one of the patch is ready, but after the rebase it is failing jenkins, I will fix it today.
15:42:39 <Swami__> The next one I will be working on is the static routes on the fip namespace.
15:42:55 <Swami__> I have a WIP patch, but need to revisit, since it was done long back.
15:43:48 <Swami__> Thats all I had on bugs to discuss
15:43:53 <haleyb> ok
15:44:15 <Swami__> haleyb: I had another question
15:44:27 <haleyb> sure
15:44:32 <Swami__> I had a couple of patches for addressing the floatingips for unbound ports.
15:44:59 <haleyb> that was an RFE, but yes, I remember
15:45:54 <Swami__> armax had a -2 on this because I was using a config option. I think I got rid of the config option.
15:46:00 <haleyb> at least i think it was an RFE
15:46:12 <Swami__> can you take a look at it. Meanwhile I will ping armax and see if he is ok to remove the -2 on it.
15:46:34 <Swami__> #link https://review.openstack.org/#/c/320669/
15:46:39 <haleyb> yes, i'll look
15:46:51 <Swami__> #link https://review.openstack.org/#/c/323618/
15:47:02 <Swami__> haleyb: yes that was an RFE.
15:47:33 <Swami__> Yes that's all I had for bugs.
15:48:32 <haleyb> ok, thanks
15:49:00 <haleyb> #topic Stable backports
15:49:52 <haleyb> Ihar sent a reminder to the list to not forget about pro-active backports
15:50:18 <haleyb> https://goo.gl/sx0KL5 has a link to the bugs, I've been working through them
15:50:24 <Swami__> haleyb: ok thanks
15:50:29 <haleyb> that's in addition to https://etherpad.openstack.org/p/stable-bug-candidates-from-master
15:51:03 <haleyb> we haven't merged a lot of DVR fixes, but there were some that needed a push along
15:51:20 <Swami__> haleyb: ok
15:51:32 <haleyb> #topic Open discussion
15:51:46 <haleyb> Any other topics to discuss?
15:51:54 <Swami__> I have not looked at the gate failures lately.
15:51:57 <Swami__> Is it stable.
15:53:08 <haleyb> it is much better.  the multinode dvr job is still 2x in the check queue
15:53:37 <haleyb> http://grafana.openstack.org/dashboard/db/neutron-failure-rate
15:53:57 <Swami__> Yes we should shoot for the multinode stability and voting pretty quick
15:54:48 <Swami__> That's all I had for today.
15:55:24 <haleyb> any lurkers have comments? :)
15:56:29 <haleyb> thanks Swami__
15:56:32 <haleyb> #endmeeting