17:18:51 <mmichelson> #startmeeting ovn_community_development_discussion
17:18:52 <openstack> Meeting started Thu Jul 16 17:18:51 2020 UTC and is due to finish in 60 minutes.  The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:18:53 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:18:55 <openstack> The meeting name has been set to 'ovn_community_development_discussion'
17:19:09 <mmichelson> I can get things started.
17:19:53 <mmichelson> I'm finishing up testing a patch that will allow dnat_force_snat and lb_force_snat options to work with dual-stack routers.
17:20:06 <mmichelson> Planning on submitting the patch to the list as soon as this meeting is done.
17:20:26 <mmichelson> Also still working on my patch for forcing symmetric replies to incoming traffic on ECMP routes.
17:20:44 <mmichelson> I've hit a snag there where the ct_commit() command is very restrictive with what you can set ct_label and ct_mark to
17:20:55 <mmichelson> I can't just do something like ct_commit(ct_label=eth.src)
17:21:10 <_lore_> hi all
17:21:28 <mmichelson> Because "eth.src" isn't an integer. So I need to expand the command to be able to accept tokens instead of just literals.
17:21:59 <mmichelson> In doing so, I'll also likely expand it to allow for specific chunks of ct_label and ct_mark to be set (i.e. ct_label[0..47]=eth.src)
17:22:44 <mmichelson> Only other bit of news is that the ovn.org "Releases" section now actually has content on it instead of placeholder text
17:22:58 <mmichelson> #link https://www.ovn.org/en/releases/
17:23:04 <mmichelson> And that's all from me.
17:23:50 <numans> Hi
17:23:56 <numans> I can go real quick.
17:24:38 <numans> Nothing much this week. dceara found an issue yesterday with monitor-all set to true
17:24:47 <numans> It was the side effect of the I-P patches.
17:25:07 <numans> thanks to dceara for finding 3 bugs so far in the I-P patches
17:25:18 <numans> I submitted the fix and it is merged to master and branch-20.06
17:25:39 <numans> I have been working on optimizing the conntrack stuff in ovn.
17:25:56 <numans> But nothing concrete yet. although I've a WIP RFC patch which is working
17:26:35 <numans> If some one wants to take a look
17:26:36 <numans> #link https://github.com/numansiddique/ovn/commit/7fe68a080c1db09c0fcdcad3a1557f9f1ba23534
17:26:51 <numans> I couldn't spend much time in reviews this week,
17:26:55 <numans> That's it from me.
17:27:23 <mmichelson> Thanks. I should also schedule some review time this week :(
17:28:03 <dceara> numans: thanks for the quick fix for the I-P bug!
17:28:21 <numans> dceara, thanks for the review :)
17:29:56 <mmichelson> Who would like to go next?
17:31:15 <dceara> I have a quick note this week: I played a bit with imaximets's stream record/replay series and it's great! And I really hope we can get it to the state where it can be used as a debugging tool in production.
17:32:09 <imaximets> dceara, thanks!
17:33:27 <dceara> I don't know how relevant it is for this meeting but just in case, ovn-heater supports now running the OVN scale tests on CentOS/RHEL platforms too. Thanks to Jianlin Shi!
17:33:49 <mmichelson> That'll probably be useful for us RH folks :)
17:33:54 <numans> dceara, may be we can consider moving it under ovn-org/ovn-heater if everyone is fine
17:34:32 <dceara> numans: sure, if people think it fits there, I'm all for it.
17:34:36 <mmichelson> +1 from me
17:35:31 <numans> +1 from me too. may be we can check with zhouhan and if he is fine
17:35:41 <dceara> numans: mmichelson: I'll bring it up on the mailing list then.
17:35:49 <numans> sounds good
17:36:14 <_lore_> can I go next?
17:36:18 <dceara> That's it on my side, thanks!
17:36:47 <_lore_> dceara: sorry :(
17:37:07 <dceara> _lore_: no pb :)
17:37:36 <_lore_> so this week I started working on this bz: https://bugzilla.redhat.com/show_bug.cgi?id=1847570
17:37:36 <openstack> bugzilla.redhat.com bug 1847570 in OVN "[RFE] Add support for BFD from OVN routers to other directly connected L3 devices" [Urgent,New] - Assigned to lorenzo.bianconi
17:37:53 <_lore_> the capability to monitor ecmp next hop for HA
17:38:12 <_lore_> I found the discussion upstream between tom and zhouhan
17:38:27 <_lore_> and I would like to have their opnionion about next steps
17:38:38 <_lore_> *opinions
17:39:28 <_lore_> as suggested by numans, we can implement bfd monitoring in ovn-controller, using lib/bfd from ovs
17:40:25 <_lore_> another approach would be to rely on ovs implementation, but this will trigger the issue about ovs interface to use for monitoring
17:41:02 <numans> since ovn can't be sure which interface to use, I personally prefer implementing in ovn.
17:41:06 <_lore_> and last approch would be to extend svc_monitorning for that, but this would be non rfc compliant
17:41:13 <mmichelson> _lore_, do you have a link to the discussion between tom and zhouhan?
17:41:18 <dceara> _lore_: we do something similar for IGMP/MLD, we reuse the OVS library and we implement the protocol in ovn-conroller.
17:41:24 <_lore_> uhm, let me look for it
17:41:42 <numans> mmichelson, _lore_ its actually trozet :)
17:42:21 <mmichelson> Ah, Tim, not Tom :)
17:42:45 <_lore_> mmichelson: sorry :)
17:42:48 <_lore_> https://www.mail-archive.com/ovs-discuss@openvswitch.org/msg07197.html
17:42:59 <mmichelson> _lore_, Thanks!
17:43:47 <_lore_> so 2 for first option :)
17:45:35 <_lore_> no other opinions??
17:46:42 <mmichelson> _lore_, my initial reaction was to use the ovs BFD library in ovn-controller. zhouhan points out a good point that determining the ovn-controller that is responsible for performing the BFD is important
17:46:56 <_lore_> yes
17:47:28 <_lore_> for the moment I was thinking about running it on the gw_router or node where gw router port is scheduled
17:49:06 <_lore_> ok, I will exmplore the first option
17:49:13 <_lore_> that's all from my side
17:49:17 <_lore_> thx
17:50:00 <mmichelson> OK, do we have anyone else?
17:50:40 <imaximets> I don't have anything OVN specific. Mostly just reviewing things for OVS 2.14 branching.
17:52:05 <mmichelson> OK. I didn't notice anybody else in here talking. Is it safe to assume we're done?
17:53:00 <mmichelson> OK, I'll take the silence to assume we're done. By everyone!
17:53:06 <mmichelson> *Bye
17:53:11 <mmichelson> #endmeeting