Arlion has quit [Read error: Connection reset by peer]
hamburgler has joined #openvswitch
donhw has quit [Remote host closed the connection]
donhw has joined #openvswitch
froyo_ has joined #openvswitch
hamburgler has quit [Quit: Leaving]
tpires has quit [Ping timeout: 252 seconds]
donhw has quit [Remote host closed the connection]
donhw has joined #openvswitch
donhw has quit [Ping timeout: 260 seconds]
elvira1 has joined #openvswitch
elvira1 has quit [Client Quit]
vlrpl has joined #openvswitch
evplus has joined #openvswitch
GNUmoon has quit [*.net *.split]
evplus has quit [*.net *.split]
tpires has joined #openvswitch
tpires_ has joined #openvswitch
tpires has quit [Ping timeout: 248 seconds]
dceara has joined #openvswitch
froyo_ has quit [Read error: Connection reset by peer]
mj2 has joined #openvswitch
roriorden has joined #openvswitch
mj2 has quit [Quit: WeeChat 4.1.1]
mj2 has joined #openvswitch
mmichelson has joined #openvswitch
zhouhan has joined #openvswitch
<mmichelson>
Hi everyone, it's time to begin the OVN developers' meeting
<dceara>
Hi
<mj2>
hello
<mmichelson>
Tomorrow is the planned release of ovn 25.03.0. If anyone has any last-minute changes they wish to get in before release, please speak up here or on the dev list.
<mmichelson>
I'm planning to create the release patches and post them today. I won't actually merge them or perform the release until tomorrow though.
<fnordahl>
o/
<mmichelson>
As far as what I've worked on, it's been reviews, some downstream build stuff, and I posted a patch series that refactors how datapaths are synced between northbound and southbound databases.
<mmichelson>
That's all from me. Who wants to go next?
<dceara>
I see zhouhan here, i'm curious if we have results related to his hwol testing with commit-all enabled.
<zhouhan>
dceara: I can give the update later
<dceara>
zhouhan, ack
<dceara>
Aside from that I did some more reviews and applied some patches (bug fixes). I had another look at patchwork earlier and from my perspective we're good to release.
<dceara>
We did get some internal requests about detailed documentation on the BGP/dynamic-routing integration. It might be something we want to add sooner rather than later to our tree (I thing fnordahl suggested a tutorial).
<dceara>
Oh and there's a patchset up for review to make OVN build on Windows (again). I cc-ed maintainers on it to see if it's worth getting this tested in CI or not.
<imaximets>
Not much this week. Following the set of OVS releases, posted patches to update submodules in supported OVN versions. And fixed an issue with a deprecated ubuntu-20.04 image on 24.03 in CI.
<imaximets>
Thanks, dceara and mmichelson for applying those!
<imaximets>
And that's all from me.
<_lore_>
can I go next? quite fast
<mmichelson>
_lore_, go for it
<_lore_>
this week and the previous one I worked on the series to probe active mac binding entries
<_lore_>
in order to refresh them if ovn/ovs sends packets with dst ip not the default gw
<_lore_>
thx for reviewing and applying
<_lore_>
now I am working on the backport for 24.09 (done) and 24.03 (mostly done)
<_lore_>
thx, that's all from my side
<mmichelson>
_lore_, excellent, thanks for that.
<mmichelson>
Who wants to go next?
<zhouhan>
I can go next
<zhouhan>
We tested the upstream ct-commit-all change, but unfortunately it still didn't work for HW offload for the k8s nodeport LB scenario. I found the root cause, tried a fix, and then found another problem ...
<zhouhan>
I am writing test cases to cover different scenarios of DNAT + SNAT, hopefully help with a proper fix
<zhouhan>
I was also distracted by lot of downstream trouble shootings so didn't get back on this soon, sorry for that
<zhouhan>
That's it from me
<dceara>
zhouhan, having tests would be awesome! Kind of a related question: do you happen to have any news about the work to report results on patchwork entries with hwol-enabled tests?
<zhouhan>
dceara: we are still working on the security bits :D
<dceara>
zhouhan, ack, thanks
<mmichelson>
zhouhan, the root cause that you found, is it a serious issue that needs fixing ASAP?
<zhouhan>
Or we might start with testing merged patches, better than nothing, what do you think?
<dceara>
zhouhan, that's definitely good to have, we do the same with ovn-heater, we run scale tests weekly using the current upstream ovn main branch.
<zhouhan>
mmichelson: it doesn't make things worse, but just didn't help for the HW offload issue.
<mmichelson>
zhouhan, ack
<zhouhan>
mmichelson: I think it is ok to fix and backport after release
<mmichelson>
Sounds good!
<mmichelson>
Who wants to go next?
<mmichelson>
That's a full minute with no response. I guess we're done for today. Thanks everyone, and have a good day!
<mj2>
oh wait
<mmichelson>
Oh do you want to give an update mj2 ?
<mj2>
yes, sorry i was assuming other people were still to go
<mmichelson>
Sure, go right ahead.
<mj2>
so im iterating on the multinode test, done quite a bit of restructuring of it, moved the TOR switch to a netns within the ovn-gw-x container
<mj2>
currently working on the topology of the system and im getting somewhere with getting two distinct ovn-bgp setups running
<mj2>
the topology changes are going much smoother than i expected though
<mj2>
which is good
<mj2>
anyway thats all
<dceara>
"somewhere" sounds like it's in the good direction. :)
<mmichelson>
Thanks mj2
<fnordahl>
I have an update
<mj2>
i can go into further detail if you would like
<mmichelson>
mj2, only if you're comfortable talking about it.
<mj2>
i was unsure if it was needed, but ill go into detail
<mj2>
so the setup currently im aiming to have is ovn-gw-1 <-> TOR-frr (in ovn-gw-1 as a netns) and then ovn-gw-2 <-> TOR-frr (in a ovn-gw-2 netns)
<mj2>
currently I have communication between the gateway nodes and their respective TOR node in a netns
<mj2>
ovn-gw-2 does not seem to be communicating with ovn-gw-1 over bgp which is good
<fnordahl>
Having the ToR's as namespaces inside ovn-gw-x is an optimization so that we don't need additional containers on the outer testrunner, right?
<mj2>
however i expect that more problems will arise when i set up two vms (netns on ovn-gw-1 and ovn-gw-2) connected via a LR which connects to a LS vis a DGP, with the LS connected to the two ovn-bgp lrs
<mj2>
yes, I have had a number of problems getting this to work, but seemingly they are all ironed out
<mj2>
i do think the ToRs in netns is much more viable than the hacky reuing a gateway node by disconnecting eth2 to br-ex
<mj2>
*reusing
<fnordahl>
agreed
<dceara>
+1
<mj2>
anyway thats the end of what i have to say
<dceara>
mj2, if you hit any blocking issues with making the "vms" networking work I can try to help next week.
<mmichelson>
OK thanks mj2!
<mj2>
unless dceara or fnordahl have any questions
<mj2>
thanks dceara
<imaximets>
It should be possible to just run in a namespace on the host outside the containers, right?
<fnordahl>
not for now, I'll go real quick
<mmichelson>
fnordahl, go for it
<fnordahl>
Mostly downstream stuff for me this week.
<fnordahl>
dceara: Noted wrt. documentation request. mj2's multinode test will hopefully help quite a bit, and I will also set aside some time to send a high level feature or tutorial d
<fnordahl>
oc patch.
<fnordahl>
With branching tomorrow I can make an attempt at making an initial version my first thing morning task.
<fnordahl>
imaximets: doing things to the host becomes cumbersome due to it being fixed setup from ovn-fake-multinode project, it's also a bit complicated since we want multiple ToRs with distinct L2 broadcast domains.
<fnordahl>
imaximets: but it is an alternative ofc
<mmichelson>
fnordahl, a nitpick: it's not branching tomorrow. It's releasing tomorrow :)
<fnordahl>
Oh, yes.
<fnordahl>
and that's it for me
<dceara>
fnordahl, I think it's good enough if the dynamic routing tutorial is only on main.
<fnordahl>
dceara: ack
<imaximets>
fnordahl, mj2: sure. I don't have a strong opinion about namespaces / containers, so whatever works should be fine. :)
<mj2>
great
<mj2>
:+1:
<mmichelson>
Any other updates from anyone else?
<mmichelson>
I think that's everyone. Thanks everyone. Have a nice day!
<dceara>
Looking forward to OVN 25.03.0! Thanks, bye!
<imaximets>
Thanks! Bye.
<fnordahl>
\o thanks all!
<mmichelson>
bye!
mmichelson has quit [Quit: Leaving]
dceara has quit [Quit: Leaving]
zhouhan has quit [Quit: Client closed]
ChmEarl has joined #openvswitch
roriorden has quit [Ping timeout: 244 seconds]
otherwiseguy has quit [Ping timeout: 245 seconds]
otherwiseguy has joined #openvswitch
roriorden has joined #openvswitch
vlrpl has quit [Quit: Leaving]
roriorden has quit [Ping timeout: 268 seconds]
mj2 has quit [Ping timeout: 268 seconds]
roriorden has joined #openvswitch
roriorden has quit [Read error: Connection reset by peer]