17:15:31 <mmichelson> #startmeeting ovn_community_development_discussion
17:15:32 <openstack> Meeting started Thu May 20 17:15:31 2021 UTC and is due to finish in 60 minutes.  The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:15:33 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:15:35 <openstack> The meeting name has been set to 'ovn_community_development_discussion'
17:15:41 <mmichelson> Hi everyone!
17:15:53 <imaximets> Hi.
17:16:07 <ihrachys> o/
17:16:24 <mmichelson> Tomorrow is supposed to be our date to branch 21.06. However, looking at patchwork, there's still a good number of patches up that were put up before the soft freeze and that need review.
17:16:39 <numans> Hello
17:16:48 <mmichelson> I'm thinking that delaying branching by a week might be a good idea. Does anyone object?
17:16:54 <numans> I'm fine with that.
17:17:09 <numans> I'm hoping I'll contribute to the reviews next week.
17:17:41 <mmichelson> Yep, I'm planning to spend basically the rest of today doing reviews, and probably some amount of time tomorrow, too.
17:18:15 <mmichelson> Also, those of you that have privileges on patchwork, please be sure to mark patches as accepted, rejected, etc. as the state changes. That way we can have a more accurate view of things.
17:19:05 <mmichelson> As for what I've been working on, I've more-or-less rewritten my ARP FIP fix from scratch. The new approach seeks to program logical flows directly in ovn-northd instead of creating MAC Bindings in ovn-controller.
17:19:43 <mmichelson> I'm planning to create an option that will allow for routing flows to be created automatically for NAT and LB addresses on routers one hop away.
17:20:00 <blp> Hi!
17:20:02 <mmichelson> This can prevent having to create static routes in routers for typical E-W traffic
17:20:03 <blp> Sorry I'm late.
17:20:04 <mmichelson> blp, hi!
17:20:26 <mmichelson> Current status is that I have the code written in ovn-northd.c but need to tackle ddlog still.
17:20:40 <mmichelson> ETA is some time next week, I think.
17:20:47 <mmichelson> And that's all from me for the moment.
17:21:12 <mmichelson> Actually before I give up the floor, blp and zhouhan have joined since I started, so I'll repeat what I led with
17:21:20 <mmichelson> Tomorrow is supposed to be our date to branch 21.06. However, looking at patchwork, there's still a good number of patches up that were put up before the soft freeze and that need review.
17:21:32 <mmichelson> zhouhan, blp, do you have strong opinions about delaying branching for a week?
17:21:50 <zhouhan> I am ok with it
17:22:04 <blp> I'm OK with it. I hope it will give some time for folks to look at my series.
17:22:21 <mmichelson> blp, yep, I'm planning to look at it and _lore_'s CoPP series today
17:22:31 <_lore_> mmichelson: thx
17:22:52 <mmichelson> blp, and once that series is merged, I think we'll do a round of scale testing with DDLog again. Hopefully we'll be in a better position than before.
17:23:08 <mmichelson> And now I'm really done :)
17:23:19 <blp> mmichelson: Sounds good!
17:23:37 <blp> I haven't rebased my series in a while. I wouldn't be surprised if there was a conflict (probably in NEWS, at least).
17:23:54 <mmichelson> Probably. But it should still be reviewable
17:24:40 <numans> blp, thanks to ovsrobot, it can be easily rebased - https://github.com/ovsrobot/ovn/tree/series_242583
17:24:50 <blp> numans: Great!
17:27:51 <blp> OK, it seems quiet, so I'll bring up the choice of IRC server.
17:28:12 <blp> freenode is undergoing some turmoil. The staff have almost all quit.
17:28:21 <blp> The successor is libera.chat. I've started an #openvswitch channel there.
17:28:27 <blp> I hope that everyone here can create a nick on libera.chat
17:28:34 <numans> +1 to move over there.
17:28:36 <blp> and that starting next week we can have the meeting there instead.
17:29:09 <numans> I'd support that.
17:29:32 <mmichelson> Yep, that's fine. I set myself up over there this morning
17:29:53 <blp> I'm going to try to be in both places for a few weeks to make sure that we get everyone.
17:30:00 <blp> I don't know whether the meeting bot is available there yet.
17:30:07 <imaximets> sounds good to me.  another option is oftc, but I don't know much about it and therefore have no preferences.
17:30:08 <mmichelson> Sounds good. Also update the topic to point to the new location
17:30:31 <mmichelson> blp, I think we can ask some openstack people about entering the bot there. Hopefully it's not a pain to do.
17:30:52 <blp> I've tried "/invite openstack" but no response currently. I *think* that's how we got the bot in here.
17:31:49 <numans> we may need to reach out to dalvarez for some help
17:32:03 <zhouhan> mmichelson: seems the bot didn't work for last meeting on 5/13?
17:32:13 <blp> It's possible that the bot just hasn't been moved yet.
17:32:23 <zhouhan> I didn't see meeting record here: http://eavesdrop.openstack.org/meetings/ovn_community_development_meeting/2021/
17:32:31 <ihrachys> I can dig on how to get it on libra and what's the status of probably openstack migration
17:32:53 <blp> The log is here: http://eavesdrop.openstack.org/meetings/ovn_community_development_discussion/2021/ovn_community_development_discussion.2021-05-13-17.15.log.html
17:33:00 <mmichelson> zhouhan, oh, that's because I changed it back to the name in the topic
17:33:01 <blp> (from 5/13)
17:33:04 <mmichelson> s/meeting/discussion/
17:33:23 <zhouhan> ok :)
17:35:25 <mmichelson> It's quiet again. Anyone who wants to share,  please go ahead.
17:35:33 <numans> I can go real quick
17:35:41 <numans> Mark Gray and I've been looking into addressing issues in offloading datapath flows when the pkt hits conntrack of router pipeline.
17:35:51 <numans> There is some asymmetry there. I'm hoping to submit the patch to fix these issues tomorrow.
17:36:07 <numans> I couldn't do much reviews this week and hoping to catch up next week. That's it from me.
17:36:51 <_lore_> can I go next?
17:37:04 <mmichelson> _lore_, go for it
17:37:27 <_lore_> this week I worked to add support for check_packet_larger for ingress traffic
17:37:49 <_lore_> for traffic entering the cluster through a gw router
17:38:35 <_lore_> ddlog part is missing
17:38:57 <_lore_> moreover I had some pending patch for reivews
17:39:41 <blp> _lore_: Do you need help on the ddlog part? Or just haven't had time for it?
17:39:48 <_lore_> - CoPP series http://patchwork.ozlabs.org/project/ovn/list/?series=241400
17:40:23 <_lore_> - localport - localnetport issue: http://patchwork.ozlabs.org/project/ovn/list/?series=241400
17:40:42 <_lore_> sorry wrong link: http://patchwork.ozlabs.org/project/ovn/patch/e2df86ba8357477e7c54cf4d99f4659edd3633ce.1621265763.git.lorenzo.bianconi@redhat.com/
17:41:19 <_lore_> this is the link for ingress check_packet_larger series: http://patchwork.ozlabs.org/project/ovn/list/?series=244729
17:41:23 <zhouhan> _lore_: the localport fix has been merged 2-3 days ago
17:41:38 <_lore_> zhouhan: it is a different patch
17:41:55 <_lore_> it is for dropping traffic from localport to a localnet one
17:42:10 <zhouhan> _lore_: at least the link you provided is merged :)
17:42:18 <_lore_> zhouhan: sorry, I put the wrong link
17:42:20 <_lore_> one sec
17:43:00 <_lore_> http://patchwork.ozlabs.org/project/ovn/patch/8008fa9867d210cf18ad31f912535f2c14e85c43.1620151078.git.lorenzo.bianconi@redhat.com/
17:43:11 <_lore_> sorry for the noise
17:43:48 <_lore_> blp: I have not time to look at it yet, thanks for the help..I will get back to you :)
17:44:16 <_lore_> zhouhan: maybe you missed the link for the patch I was referring to
17:44:30 <_lore_> http://patchwork.ozlabs.org/project/ovn/patch/8008fa9867d210cf18ad31f912535f2c14e85c43.1620151078.git.lorenzo.bianconi@redhat.com/
17:44:37 <zhouhan> yes, sorry I was disconnect
17:44:50 <_lore_> that's all from me, sorry for the confusion :(
17:44:51 <zhouhan> now I see it, thanks!
17:45:07 <ihrachys> I can do next?
17:45:20 <mmichelson> sure hting ihrachys
17:45:41 <ihrachys> spent time fixing some embarrasing bugs with the new allow-stateless, here's the result: https://patchwork.ozlabs.org/project/ovn/list/?series=244288
17:45:50 <ihrachys> hoping we can squeeze them into 21.06
17:46:17 <zhouhan> ihrachys: thanks. I am reviewing it
17:46:21 <ihrachys> it's mostly around proper handling priority/direction of other acls
17:46:27 <ihrachys> zhouhan: thanks!
17:46:57 <ihrachys> also, thanks blp for fixing that ddlog relation bug, it was over the top of my abilities :)
17:47:19 <ihrachys> that's it for me
17:47:20 <mmichelson> Yeah I would never have figured that one out
17:48:58 <zhouhan> I have a quick one: mmichelson: please take a look at v2 of the ssl patch series: http://patchwork.ozlabs.org/project/ovn/list/?series=244813
17:49:11 <mmichelson> zhouhan, can and will do
17:49:14 <blp> ihrachys: That was a wild bug.
17:49:25 <mmichelson> call me Review Man
17:49:33 <mmichelson> Because that's all I'm going to do for a bit.
17:49:34 <zhouhan> thanks mmichelson!
17:51:54 <zhouhan> blp: just curious do you expect more such kind of issues that require debugging in ddlog itself?
17:52:25 <mmichelson> zhouhan, I think that's an impossible question to answer :)
17:52:49 <blp> zhouhan: I don't, but... even GCC has bugs.
17:53:21 <zhouhan> Or, maybe better to ask: can we easily tell if it is bug in the northd-ddlog code or in ddlog itself?
17:54:37 <zhouhan> I guess the answer is that if we are familiar enough with ddlog then it is easier to tell :D
17:55:15 <blp> Easily? It's a matter of debugging until there's a contradiction. That's what I ended up with.
17:55:48 <blp> I don't normally suspect my tools either.
17:56:08 <zhouhan> That's fair :)
17:56:24 <ihrachys> in this case it was probably slightly easier since we knew the two commits that interacted badly; in the future we may not have that privilege of scoping the issue to particular relations / definitions. correct?
17:56:53 <blp> Some kinds of bugs will probably never end up blamed on ddlog itself. Since it's in Rust, I wouldn't expect to ever see a memory error, or if there is one it's probalby due to the C code.
17:57:29 <blp> ihrachys: That's correct, but you can usually reduce it to one or two commits, which does tend to focus the search.
17:58:42 <ihrachys> that's if it doesn't happen in the wild on a customer machine :)
17:59:04 <ihrachys> "my northdb is empty, please help; what did you do? I dunno"
18:00:17 <imaximets> if you have a prefect tool, you, probably, didn't test it enough. :)
18:01:19 <imaximets> *perfect
18:01:21 <mmichelson> is there anyone left that hasn't gone yet?
18:01:39 <imaximets> I'm here, but I don't really have much to say.
18:02:22 <mmichelson> imaximets, :)
18:03:45 <mmichelson> OK, I guess that's everyone then. Thanks!
18:03:49 <mmichelson> #endmeeting