18:16:12 <mmichelson> #startmeeting ovn_community_development_meeting
18:16:12 <openstack> Meeting started Thu Jan 21 18:16:12 2021 UTC and is due to finish in 60 minutes.  The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:16:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
18:16:15 <openstack> The meeting name has been set to 'ovn_community_development_meeting'
18:16:31 <mmichelson> I can lead things off
18:17:01 <mmichelson> The biggest thing I have to report is my change I put up for review yesterday to make OVS a submodule in OVN. I got good feedback from dceara and imaximets and am working on v2 now.
18:17:44 <mmichelson> And that's all I have to report.
18:17:46 <blp> This does seem like an appropriate use of submodules.
18:17:54 <blp> I have a quick report.
18:18:28 <blp> I see that my idl restructuring change was reviewed.
18:18:44 <blp> I expect to push it today, as long as the fixes that people requested are simple.
18:19:07 <blp> And then I'll need to rebase, update, and retest the ddlog changes for ovn.
18:19:19 <blp> As long as they don't break the non-ddlog case, I guess it'll be ready then.
18:19:41 <blp> There were a few people who reported problems linking the ddlog changes; I found the problem and fixed that in the last version.
18:21:57 <mmichelson> anything else blp, or were you finished?
18:22:02 <blp> That's all.
18:22:51 <imaximets> I have a quick update.
18:23:34 <imaximets> patch to use column diffs in file transactions is merged now, will be part of 2.15 release.
18:23:58 <mmichelson> nice
18:24:00 <blp> imaximets: that is a remarkably effective change. well done
18:24:08 <imaximets> thanks
18:24:44 <imaximets> I'm also working an a small change for ovn-ctl to allow disabling this feature for the ovsdb-server upgrade case.
18:24:56 <imaximets> will send a patch somewhere soon.
18:25:04 <imaximets> that's mostly it from me.
18:26:12 <mmichelson> Okie-dokie. Anyone else?
18:26:49 <imaximets> mmichelson, do we need to have submodules for older OVN branches?
18:28:09 <mmichelson> imaximets, hmm
18:28:50 <mmichelson> imaximets, my initial thought is not to put the submodule on older branches, but I can be convinced otherwise
18:29:59 <imaximets> mmichelson, it might be good if we have some random OVS commits to build with, i.e. if some branch cannot be built with existing OVS releases.  But I'm not sure.
18:33:13 <dceara> I have a quick update too, I sent a couple of bug fix patches and I started looking at some scale related issues reported by OpenShift in ovn-controller: configurable lflow cache size/limit and reducing number of load balancer openflows.
18:33:14 <mmichelson> imaximets, right that's my problem. I'm not 100% sure either
18:33:26 <dceara> Oops, sorry mmichelson :)
18:33:34 <mmichelson> dceara, no, go ahead :)
18:33:45 <dceara> That's it, one liner, thanks :)
18:34:51 <mmichelson> ah ok
18:36:34 <mmichelson> Anyone else?
18:38:08 <dceara> blp: Regarding the IDL patches, I mentioned two memleaks but they should be straight forward to fix.  Otherwise the rest looked good to me so I think it's safe to push them if imaximets agrees too.
18:38:34 <blp> That was my impression from a first look at the reviews. imaximets?
18:38:55 <dceara> blp: imaximets: Related we consider including them in the already created 2.15 OVS branch?
18:39:00 <blp> (I've been doing almost nothing but paper writing recently.)
18:39:05 <dceara> s/we/should we
18:39:16 <blp> I'll defer to imaximets on that.
18:40:08 <imaximets> With memory leaks fixed and some small comment update it should be good to go.  If we really want this update on 2.15, I think, it's OK to apply there.
18:40:19 <dceara> I'm thinking that we wouldn't have to build ovn20.06 with a non-release branch if the IDL changes go to 2.15.
18:40:32 <dceara> ovn21.03
18:40:42 * dceara is still in 2020 apparently :(
18:42:33 <imaximets> blp, dceara: I'm OK with applying to 2.15.  Not sure if building with released OVS is a big problem, though.  Especially if we will have a submodule.
18:43:18 <blp> I don't know. Is it an issue for Fedora packaging? It could be an issue for Debian.
18:43:32 <blp> I recently stepped back from Debian, so it's not *my* problem anymore though.
18:44:55 <mmichelson> numans maintains the fedora OVN packages. I don't think building against non-release OVS has been a problem, but I'd defer to him to be certain. I don't know if he's present though.
18:45:03 <dceara> blp: numans usually does the Fedora packaging but, afaik, based on how we build OVN packages for RHEL there shouldn't be a problem.
18:46:46 <imaximets> blp, Debian might be problematic.  But package versions are frozen there already, IIUC.  So it will be 20.12 at most that should be fine with ovs 2.15.
18:47:55 * imaximets doesn't really know what is happening in Debian.  Don't listen to him. :)
18:49:06 <imaximets> blp, let's just apply idl split to branch-2.15 and forget about it. :)
18:49:26 <mmichelson> Remember, commits can always be undone.
18:49:47 * mmichelson imagines yall are about to give me counterexamples now
18:50:40 <blp> imaximets: OK. I'll make it happen today. Thank you.
18:51:08 <imaximets> ok
18:51:15 <blp> I actually think the idl split makes the idl easier to understand, so it should be helpful to maintainers.
18:51:32 <mmichelson> that's a win
18:55:00 <mmichelson> All right, I guess we're done if no one else wants to share.
18:55:13 <mmichelson> bye everyone
18:55:28 <mmichelson> #endmeeting