17:21:40 <mmichelson> #startmeeting ovn_community_development_discussion
17:21:41 <openstack> Meeting started Thu Oct 15 17:21:40 2020 UTC and is due to finish in 60 minutes.  The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:21:42 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:21:44 <openstack> The meeting name has been set to 'ovn_community_development_discussion'
17:22:11 <mmichelson> I can start things off, but there's not much for me to report.
17:22:11 <dceara> Hi
17:22:37 <mmichelson> Biggest time sink of mine this week has been reviews. I still have more in my queue to look at
17:23:06 <mmichelson> I'm also writing up a post that will hopefully go on the OVN website that gives an overview of changes in 20.09
17:23:26 <mmichelson> Nothing too deep, but more than just what's in the NEWS file.
17:23:40 <mmichelson> And that's all for me to report
17:26:43 <imaximets> I have a very little update.
17:27:19 <imaximets> This week I tried to run OVN tests with clang MemorySanitizer and found on issue with ic.
17:27:51 <imaximets> It hashes struct v46_ip with uninitialized memory inside.
17:27:52 <mmichelson> I think I saw a patch from Han with a fix for that
17:28:22 <imaximets> Yep, I sent a bug report to maillist and zhouhan sent a patch for this.
17:28:53 <imaximets> That's it from my side for OVN this week.
17:29:05 <_lore_> hi all
17:29:28 <_lore_> can I go next?
17:29:32 <_lore_> very quick
17:29:36 <mmichelson> go for it
17:29:39 <_lore_> this week I posted:
17:29:47 <_lore_> 1- iPXE support for ovn
17:29:57 <_lore_> tested by OSP folks
17:30:11 <_lore_> 2- Renew/Rebind support for IPv6 PD
17:30:45 <_lore_> 3- now I stared working on a BFD implementation in OVN to enable link failure detection for emcp routing
17:30:53 <_lore_> that's all from my side
17:31:44 <dceara> I have a quick update too if I may go next.
17:32:22 <mmichelson> yeah I like these quick updates :)
17:32:46 <dceara> Thanks to zhouhan and Mark Gray for reviews on the ofctrl predictable conflict resolution patches! I sent a v6 and today a v7: http://patchwork.ozlabs.org/project/ovn/patch/1602753150-31918-1-git-send-email-dceara@redhat.com/
17:33:11 <zhouhan> dceara: thank you! I am reviewing now.
17:33:30 <dceara> zhouhan: Sorry, I think we had some confusion at some point, but I hope it's all sorted out now.
17:34:17 <dceara> Except for that I did some reviews this week and I wanted to bring up to attention numans' question on patch http://patchwork.ozlabs.org/project/ovn/patch/20201014091525.3391981-1-numans@ovn.org/
17:35:01 <dceara> That's it from my side, thanks!
17:35:10 <numans> dceara, thanks for bringing it up.
17:35:25 <dceara> numans: I didn't know if you'd join :)
17:35:26 <numans> I can go real quick too.
17:35:32 * numans just joined
17:36:09 <numans> It would be great if mmichelson zhouhan and others can provide their comments to the link dceara shared.
17:36:32 <numans> based on that, I'll spin up next version of reject patch series.
17:36:50 <zhouhan> numans: is the question about whether we should swap IPs and ETHs together?
17:36:54 <numans> other than that I worked for reducing the hairpin flows.
17:37:14 <numans> zhouhan, yes
17:37:32 <numans> zhouhan, or versus whether to go another ovn field ip.src <-> ip.dst
17:37:46 <numans> which results in another packet-in.
17:38:14 <mmichelson> numans, I'll have a look after this meeting.
17:38:21 <numans> mmichelson, thanks.
17:38:24 <zhouhan> another packet-in sounds bad :(
17:38:52 <mmichelson> yeah the issue is that the ip.src <-> ip.dst results in a packet in, and then the reject {} does a second packet in.
17:38:57 <numans> zhouhan, which means we need to swap eth src and eth dst and ip src and ip dst
17:39:16 <numans> mmichelson, zhouhan FYI we do the same for icmp4.frag_mtu too
17:40:08 <zhouhan> numans: you mean two packet-ins for frag_mtu?
17:40:17 <numans> I'm fine to implicitly swap these fields in reject pkt-in, but normally in OVN,  we specify the inner actions to any OVN action
17:40:45 <numans> zhouhan, the usage of it is - icmp4 { eth.src <-> eth.dst ,...., icmp4.frag_mtu = 1500 ...}
17:40:53 <numans> zhouhan, this results in packetin for icmp4 action
17:41:11 <dceara> numans: In my opinion "reject" is explicit enough to imply that it will generate a packet with addresses swapped.
17:41:13 <numans> and then again to set icmp4.frag_mtu (with pause flag set for this one)
17:41:33 <numans> zhouhan, that was suggested by Ben when I worked on it while back
17:42:10 <numans> I'm fine with both the approaches. I'll respin the patches based on the comments :)
17:42:29 <numans> That's it from my side.
17:42:31 <zhouhan> numans: thanks for the information. I will check it offline.
17:42:43 <numans> zhouhan, welcome
17:43:04 <zhouhan> May I go next?
17:43:27 <zhouhan> I want to briefly discuss the _is_new problem, if Mark Gray is here?
17:44:26 <mmichelson> Hm, if he's in here he's using a nick I'm not familiar with
17:44:38 <zhouhan> dceara: thanks for sharing the code that relies on _is_new() for a row inserted in current transaction
17:45:09 <zhouhan> dceara: I think there is some confusion for those interfaces _is_new(), _is_modified() ...
17:45:44 <dceara> zhouhan: I agree. But we do use them for locally inserted records too. We could make them work in that case too. Which is what Mark's patch does, IIUC.
17:45:52 <zhouhan> dceara: for my understanding these are for tracked changes only. But it seems it is used for checking a row added in current transaction (not committed yet)
17:46:22 <zhouhan> dceara: yes, it is definitely possible to change it.
17:47:03 <dceara> zhouhan: I think the patch will not break the tracked changes functionality so I think it would be good to have.
17:47:35 <zhouhan> dceara: but if we change, we should make sure it is changed for modify and delete, not only just for insert. (I am not 100% sure if there is any potential problems, need to study more on this)
17:47:40 <dceara> But I'm not extremely familiar with that code so I might be wrong.
17:48:18 <dceara> zhouhan: OK, I see.
17:48:35 <zhouhan> dceara: let's check more and discuss with Mark Gray.
17:48:45 <dceara> zhouhan: Ack.
17:49:00 <zhouhan> I don't have any other updates except the bug fix and some reviews
17:49:05 <zhouhan> thanks
17:50:51 <mmichelson> Okie-dokie. Anyone else have updates?
17:52:25 <mmichelson> I guess that's it then. Thanks everyone!
17:52:28 <mmichelson> #endmeeting