18:17:13 #startmeeting ovn_community_development_discussion 18:17:13 Meeting started Thu Dec 10 18:17:13 2020 UTC and is due to finish in 60 minutes. The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:17:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:17:16 The meeting name has been set to 'ovn_community_development_discussion' 18:17:56 Last Friday I branched OVN 20.12, but I completely forgot that there were other commits required in addition to actually creating the branch (whoops!) 18:18:09 Luckily Numan swept in and got those created. 18:18:10 <_lore_> hi all 18:18:35 This week I've worked on getting SCTP up to feature parity with TCP and UDP when it comes to reject ACL actions. I have a patch on patchwork. 18:19:15 I will need to update it once _lore_'s patch that rejects packets on empty LB VIPs gets merged, since I will need to include SCTP support there as well 18:19:25 (that may just require updating documentation. I'm not 100% sure) 18:19:46 In other news, the OVS conference was held virtually over the past two days. 18:19:59 Thanks to blp for being such a good host. 18:20:16 And that's all from me! 18:26:22 Do we have anyone else that wants to share? 18:27:55 <_lore_> I can go next, very quick 18:28:13 <_lore_> this week I worked on rejects packets on empty LB VIPs, posted v3 18:28:44 <_lore_> then I worked on addressing numans's comments on bfd series, I am now integrating bfd on static routes 18:28:55 <_lore_> I think will post v2 by the end of the week 18:29:06 <_lore_> that's all from my side 18:33:58 OK, last chance if anyone else wants to share 18:34:08 This may be a record for shortest OVN meeting :) 18:35:23 mmichelson, sorry, there was a meeting collision. 18:35:41 imaximets__, ah ok 18:36:12 I have a quick update 18:36:51 First is that dceara found a perfromance issue with the datapath groups patch-set. 18:37:26 It's the swapping of logical datapath filed in logical flows in case where datapath groups disable. 18:37:35 *disabled 18:37:50 I have a fix for it and will send a bit later today. 18:38:17 And the second one is that I worked on optimization of ovsdb-server. 18:38:48 To store in database and raft logs only diff of the row, not the whole new value. 18:39:15 This should significantly improve cases where we have a lot of UUIDs in a set. 18:39:47 Each database file transaction and raft log entry will have only what is modified. 18:41:02 In my testing it allows to significantly reduce size of the on-disk database and the proces memory consumption. 18:41:59 In a simple test that creates 4000 acls in a single port group the size of the on-disk db reduced from 369MB to 2.7MB 18:42:38 Memory consumption of the ovsdb-server (in a raft cluster) process reduced from 2.7 GB to 580 MB. 18:43:23 Currently I'm working on database upgrade scenario for this patch. Will send on a mailing list somewhere soon. 18:43:33 That's it from me. 18:43:50 imaximets, just for clarification, the ovsdb optimization is an OVS change, right? Not OVN? 18:44:07 mmichelson, yep. pure OVS patch. 18:44:16 imaximets, OK, thanks 18:44:25 Anyone else care to share? 18:46:22 OK, my toddler son has woken up from his nap, so this seems like as good a time as any to wrap things up. 18:46:28 #endmeeting