00:27 <varesa> well, picking those two commits was a bit misguided, since a ton of work in commits that don't touch the version number

00:29 <varesa> it seems that a simple standalone setup of two ovsdbs and a single northd instance can reproduce the issue on branch-24.03 but not on some older versions

00:31 <varesa> got a `git bisect run` looking for the change that triggers it. Just accidentally picked some really old commit (still being fixated on OVN_INTERNAL_MINOR_VER at that point) which in 2022 or something like that, so there are a few unnecessary steps when I could have started at 24.03-something. Well, it can run a few more iterations in the background

00:52 <varesa> https://github.com/ovn-org/ovn/commit/a9e8d8931c12d194efe1a15ac96536b100fcfa40 is what breaks half of my LSPs

16:17 <dceara> Hi all! I don't think Mark is here (yet). I guess I can start the OVN developers' meeting and hand over when/if he joins.

16:17 <fnordahl> o/

16:17 <zhouhan> hi

16:17 <imaximets> o/

16:18 <dceara> Who would like to start with their updates?

16:18 <imaximets> May I?

16:18 <dceara> imaximets, sure, please go ahead!

16:19 <imaximets> Don't really have much. Been working mostly on OVS stuff, declared soft freeze yesterday. Before that worked on adding support for checking server-side types

16:19 <imaximets> in C IDL.

16:19 <imaximets> This is merged now, will work on a change for OVN to stop parsing the schema.

16:19 <imaximets> That's all from me.

16:19 <dceara> imaximets, thanks, it would be nice to see the parsing code removed from OVN!

16:20 <dceara> Who would like to go next?

16:20 <dceara> I guess I can if nobody else volunteers. :D

16:21 <fnordahl> me

16:21 <dceara> fnordahl, please go ahead!

16:21 <fnordahl> Pretty swamped in downstream issues, so did unfortunately not get around to upstream reviews (sorry dceara, I will get to it.)

16:21 <fnordahl> One of our customers have hit an issue similar to what was fixed in 20a96b9d2c3d ("northd: Allow fragmented traffic through LB."), but it affects regular vm - vm traffic and not strictly realted to LBs.

16:21 <fnordahl> Have not looked in detail yet, but have had them file a bug here: https://bugs.launchpad.net/ubuntu/+source/ovn/+bug/2115795

16:22 <fnordahl> That's about it for me

16:22 <dceara> Thanks fnordahl!

16:22 <dceara> mmichelson, hi! We started without you, sorry.

16:22 <mmichelson> Sorry for the late arrival.

16:23 <dceara> mmichelson, would you like to share an update?

16:23 <mmichelson> Sure, I can do that.

16:23 <mmichelson> I've been addressing reviews on the datapath and port binding refactor series for the most part.

16:23 <mmichelson> dceara pointed out a performance regression due to the lack of incremental processing in the new engine nodes introduced in the series.

16:24 <mmichelson> _lore_ and I are working to add incremental processing to the datapath sync refactor portion so we can at least try to get that merged before soft freeze.

16:24 <mmichelson> Patch 4 (port binding refactor) may end up waiting until 26.03 instead, depending on how quickly the datapath incremental processing goes.

16:24 <mmichelson> That's all from me.

16:25 <dceara> Thanks, mmichelson! And sorry for running the scale tests so late in the development cycle. It would be nice to get the changes in 25.09 though.

16:25 <dceara> I guess I can go next.

16:26 <dceara> Similar to previous weeks I've been: reviewing patches (a notable one is a new feature to advertise load balancer IPs across OVN IC zones; another notable one is mmichelson's datapath/port binding refactor which I really liked)

16:27 <dceara> and I've been working together with Ales on adding the EVPN support to OVN. Slow progress on my side - focusing on writing e2e tests now (in parallel with Ales doing the actual implementation).

16:27 <dceara> That's it on my side.

16:28 <zhouhan> May I go next

16:28 <dceara> I see zhouhan also said "hi" in the beginning. Would you like to share an update, Han?

16:28 <dceara> zhouhan, yes, definitely! :)

16:28 <zhouhan> I am still fixing the HW offload issue introduced by 20a96b9d2c3d ("northd: Allow fragmented traffic through LB."). Worked out a patch of adjusting flow priorities, which helped for some scenarios but not others. The problem was that the LS and LR pipelines are sharing same tables, so it is hard to make the priority right. So I am working on a new

16:28 <zhouhan> approach that introduces new stages that reads the ct_xxx info from separate tables

16:28 <zhouhan> I hope that would work

16:28 <zhouhan> That's it from me

16:29 <dceara> zhouhan, if I'm not wrong, I think together with imaximets we discussed a similar approach of loading ct info in side tables (to avoid overlapping pipelines) and IIRC that might work indeed.

16:29 <dceara> zhouhan, looking forward to the patches!

16:29 <zhouhan> cool!

16:29 <dceara> Is there anyone else who'd like to share today?

16:31 <dceara> One more thing on my side: soft freeze for 25.09 is approaching (scheduled for July 18th), please post your feature patches! :)

16:31 <dceara> Thanks everyone for joining, have a nice day!

16:31 <fnordahl> \o thanks, cheers!

16:31 <mmichelson> Thanks dceara for taking over for me today!

16:31 <mmichelson> Bye everyone!

16:31 <dceara> mmichelson, no problem! :)

16:31 <imaximets> Thanks! Bye.

16:31 <zhouhan> bye

16:31 <dceara> Bye!