17:13:35 <mmichelson> #startmeeting ovn-community-development-discussion
17:13:35 <openstack> Meeting started Thu Mar 26 17:13:35 2020 UTC and is due to finish in 60 minutes.  The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:13:36 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:13:38 <openstack> The meeting name has been set to 'ovn_community_development_discussion'
17:15:10 <mmichelson> Hi everyone, it's about time we got started with this meeting.
17:15:36 <mmichelson> I've been working on SCTP-related things lately. You've probably seen my patch for SCTP load balancing, as well as the discussion about conntrack states for SCTP
17:15:43 <numans> Hello
17:16:12 <mmichelson> The end result of that SCTP discussion is that OVN is trying to do an invalid snat operation in the case where there is a load balancer configured and an SNAT for the same external IP.
17:16:35 <mmichelson> numans wrote a patch that I've acked. Once that's put in place, I'll rebase my SCTP load balancer patch and it should be good to review.
17:16:58 <mmichelson> The other major thing I've been doing lately is looking at pw-ci (the repo of scripts that 0-day robot uses)
17:17:10 <_lore_> hi
17:17:11 <blp> I'm in a program committee meeting today but I hope that whoever is running the OVN meeting uses the new meeting bot.
17:17:26 <mmichelson> I want to introduce more CI into the OVN process. I have an idea of how that would work, and I'm going to send an e-mail about it probably tomorrow
17:17:37 <mmichelson> blp, yep, I started the meeting using the bot :)
17:17:48 <numans> FYI - logs wil be here - http://eavesdrop.openstack.org/meetings/ovn_community_development_discussion/
17:17:55 <mmichelson> And now I'm done.
17:18:11 <mmichelson> BTW, I have a baby in the room with me and he's actively throwing toys around the room, so I may not be as responsive as normal during this meeting
17:18:12 <flaviof> hi all
17:18:18 <numans> flaviof, hi
17:18:20 <mmichelson> So who'd like to go next?
17:19:08 <numans> I can go real quick
17:19:22 <numans> As mmichelson mentioned I submitted the patch to address the nat issue
17:19:31 <numans> I did few reviews.
17:19:41 <numans> And nothing much. That's it.
17:19:54 <numans> zhouhan, Thanks for the comments for the I-P patches
17:20:02 <numans> I'll get back with some comments.
17:20:03 <zhouhan> numans: np
17:20:35 <numans> zhouhan, We can discuss more in the ML, but I had the correctness in my mind
17:20:46 <numans> I mean the flow correctness.
17:20:58 <zhouhan> numans: ok, let's discuss there
17:21:11 <numans> In my testing though I didn't find any inconsistencies though
17:21:31 <numans> zhouhan, If you have any comments which you want to share, please do
17:21:40 <numans> Sometimes its easier over the chat.
17:22:01 <numans> That's it from me.
17:22:13 <mmichelson> thanks numans
17:23:30 <flaviof> I'll jump in real quick :)
17:23:46 <flaviof> I'll be putting notes about this meeting in the site
17:23:49 <flaviof> #link https://github.com/ovn-org/ovn-website/pull/9
17:24:01 <mmichelson> cool!
17:24:08 <flaviof> If any of you have ideas on what you want to see there, don't be shy to make a pr
17:24:34 <flaviof> that is all from me :)
17:25:02 <mmichelson> ok I think it's going to take some getting used to all the features that the meeting bot has to offer.
17:25:16 <mmichelson> But for now having logs and being able to easily link them is a huge thing
17:27:14 <zhouhan> I can go next
17:27:45 <zhouhan> I did some review on ovn, will pickup the ovn-scale-test reviews.
17:28:13 <zhouhan> numans: For your I-P patches, my major concern is the noop change-handler. If an input is used in full-compute, but it is not handled in the change-handler, it seems something is missing.
17:29:07 <zhouhan> I haven't looked at the OVSDB update lost problem. It is still in my list
17:29:12 <zhouhan> That's it from me
17:29:36 <numans> zhouhan, I understand. But running full (lflow_run) recompute for every runtime data change is definitely costing lot of time
17:29:53 <numans> mainly when there are lots of lflows
17:30:24 <numans> but ofcourse correctness is important :)
17:30:43 <zhouhan> numans: yes, so what's in my mind for incrementally processing "runtime data" change is to split "runtime data" into different nodes, and handle some of them incrementally.
17:31:49 <numans> zhouhan, that's gonna be tricky.
17:32:00 <avishnoi> numans: zhouhan what is the sceario of update lost ? sb db sends the update but ovn-controller is disconnected?
17:32:03 <zhouhan> currently the "runtime data" is just a bundle of many things that I didn't have time to think through one-by-one. If we want to handle it incrementally, I think we will have to break it down
17:32:14 <numans> ok.
17:32:26 <numans> I see your point
17:32:53 <avishnoi> and when it connects back, sb db thinks that updated is already sent ?
17:32:58 <zhouhan> avishnoi: the update lost is something reported in last meeting by dceara
17:33:21 <numans> zhouhan, Let me explore more on your suggestion
17:33:23 <numans> thanks
17:34:11 <zhouhan> avishnoi: he noticed for some reason the IDL didn't got some of the update, and it will never get it until restart
17:34:23 <dceara> avishnoi, this is the issue: https://bugzilla.redhat.com/show_bug.cgi?id=1808580#c11
17:41:23 <openstack> bugzilla.redhat.com bug 1808580 in ovn2.12 "[OVN SCALE] many flows missing from new node after scaling up 179 nodes" [Unspecified,Assigned] - Assigned to dceara
17:41:23 <zhouhan> numans: sure, thank you
17:41:23 <avishnoi> dceara: thanks.
17:41:23 <dceara> I have a quick update as well if I may jump in.
17:41:23 <numans> dceara, jump in
17:41:23 <dceara> zhouhan, numans, imaximets, thanks for the ovsdb-idl review of https://patchwork.ozlabs.org/patch/1261637/
17:41:23 <zhouhan> dceara: np
17:41:23 <flaviof> #link  https://bugzilla.redhat.com/show_bug.cgi?id=1808580#c11  issue reported in last meeting by dceara
17:49:37 <dceara> If everyone is ok with it, it would be great to have this merged so we can use ovn-monitor-all safely
17:49:37 <mmichelson> dceara, With the reviews you have I think it's probably fine to merge.
17:49:37 <dceara> regarding the missing SB DB update, I didn't find the root cause yet so I'll be at least sending a patch force a reconnect with monitor_cond_since last_id 0. Even if we fix the root cause it would still be good if ovn-controller would recover.
17:59:50 <ihrachys> numans: btw tried as you said - bring up fake-multinode, then set vxlan through external-ids for all nodes, then ping from port to ovnfake-ext and vice versa, all works, vxlan tunnels replaced geneve ones
18:00:31 <numans> ihrachys, ok. cool. probably something wrong in my setup
18:00:35 <numans> thanks for confirming
18:15:13 <numans> test
18:22:33 <flaviof> numans: I think we are missing an #endmitting ;)
18:22:55 <flaviof> # e n d m e e t i n g :)
18:23:20 <flaviof> mmichelson: ^^
18:23:59 <flaviof> #link http://eavesdrop.openstack.org/meetings/ovn_community_development_discussion/2020/ovn_community_development_discussion.2020-03-26-17.13.log.txt
20:36:03 <flaviof> #endmeeting