fab23 has quit [Quit: macOS is user friendly. It is just picky about who his friends are.]
fab23 has joined #openvswitch
jlibosva has joined #openvswitch
dceara has joined #openvswitch
ChmEarl has joined #openvswitch
erlon has joined #openvswitch
mmichelson has joined #openvswitch
mkalcok has joined #openvswitch
<mmichelson>
Hi everyone, it's time to begin the weekly OVN developers' meeting.
<fnordahl>
o/
<mkalcok>
o/
zhouhan has joined #openvswitch
<mmichelson>
I can get thing started this week.
imaximets__ is now known as imaximets
imaximets has quit [Changing host]
imaximets has joined #openvswitch
<mmichelson>
We branched 25.09 last week. I'm in the midst of reviewing numan's ovn-bridge-controller patch series.
<erlon>
\o
<mmichelson>
I think that's probably it for me. Who would like to go next?
<mkalcok>
I can go next
moloings has joined #openvswitch
<mkalcok>
I proposed a potential solution for unnecessary conntrack entries for DNAT_AND_SNAT traffic. I'll wait for amusil's opinion before spinning it into a real change proposal
<mmichelson>
mkalcok, ack, amusil is out on PTO this week. I believe he'll be back next week though.
<mmichelson>
Who's next?
<erlon>
\o
<imaximets>
May I?
<mmichelson>
imaximets, go right ahead
<imaximets>
Not a lot of updates, did some reviews and officially released OVS 3.6.
<imaximets>
A reminder for everyone: Submit your talks for OVS/OVN conference if you didn't already.
<imaximets>
That's it from me.
<mmichelson>
thanks imaximets
<mmichelson>
Who wants to go next?
<erlon>
So me
<mmichelson>
erlon, sure thing, go right ahead
* dceara
joins the party late
<erlon>
We are proposing a fix for the UDP connection tracking. We already got to feed it back from the version one of the patch, but why you working on the second version we are hitting some errors related to conjunction rules.
<erlon>
I pointed this specific line because, it seems that adding that or condition fixes a problem with the reverse flows from the first version, but then it breaks when ofctrl creates conjuntion rules
<erlon>
So I'm floating the idea, just to see if people have an the best way to fix that
<mmichelson>
erlon, interesting, I don't recall what OFPBMC_BAD_PREREQ means. Does it have to do with conjunction flows being formatted incorrectly or something?
<erlon>
@fnordahl removing the usages of it did not solve the issue, it passes if you completely remove translation, but as soon as you add it even if it's in a single place, it starts breaking again
<imaximets>
It means one of the fields doesn't have prerequisites matched.
<erlon>
It seems It's related to the rules not matching some prerequisites
<imaximets>
I'd suggest looking at the actions of the main flow to see what it changes and check that all conjunction flows have prerequisites for that.
<imaximets>
e.g. the conjunctive flow matches on packet to be ip, so any updates of l4 fields will fail.
<imaximets>
* will fail prerequisite checks.
<erlon>
right
zhouhan52 has joined #openvswitch
<imaximets>
It may mean that we have to generate different flows for the first and later fragments.
<erlon>
right, I was also considering generating one for each one of the (ct.new || ct.est || ct.rpl) conditions, since adding those flags seems to be necessary to allows reverse traffic
zhouhan has quit [Ping timeout: 250 seconds]
<mmichelson>
OK, I'm not sure that anyone has any immediate help for this issue here. I guess we can continue this discussion on the mailing list instead.
<mmichelson>
In the meantime, who would like to give the next update?
<zhouhan52>
May I
<mmichelson>
zhouhan52, yes you may
<zhouhan52>
I am working on support flow-based tunnels (instead of port-based), so that in large scale env we can avoid huge number of tunnel port in OVS
<zhouhan52>
The only thing I am not sure is if the BFD is still going to work
<imaximets>
ipsec will not work.
<zhouhan52>
imaximets: hmm ...
<zhouhan52>
So, probably add it as an optional feature
<zhouhan52>
imaximets: BFD won't work either, right?
<imaximets>
Yes, BFD will not work as well, IIRC.
<imaximets>
Why the number of tunnels is a probelm?
<zhouhan52>
We have seen OVS main process performance issue in the past when there are thousands of ports
<zhouhan52>
is that not a problem any more?
<zhouhan52>
I will have a test first
<imaximets>
I haven't seen such issues for a long time.
<zhouhan52>
imaximets: thanks for the feedback
<imaximets>
Would be good to see some test numbers / find where the issue is coming from. Maybe we can just fix that in OVS instead.
<imaximets>
np
<zhouhan52>
imaximets: that would be good. It seems not much a problem with thousands of tunnel ports, except that it doesn't look clean when doing ovs-vsctl show, but shouldn't be a big deal :)
<zhouhan52>
imaximets: I mean if the performance is ok or can be fixed in OVS
<zhouhan52>
That's it from me
<imaximets>
You can use the --filter now with ovs-vsctl!
<fnordahl>
mmichelson: for once its not an ovn flaky test!
<fnordahl>
:)
<dceara>
yay :)
<mmichelson>
"for once" lol
<mmichelson>
Do we have any more updates today?
<moloings>
Oneliner, I am working through some legacy transit switch tests that I have broken while doing transit router development.
<mmichelson>
moloings, thanks for the update!
<mmichelson>
Anyone else?
<dceara>
I also have a oneliner (two lines in one I guess): reviewed patches before branching, reviewing some more now including the Network Function Insertion v6 for which an exception request was raised for late acceptance in 25.09.
<mmichelson>
I replied to that exception request on-list, btw.
<dceara>
Ack, I saw mmichelson, I agree, let's see how that develops in the coming days.
<dceara>
Thanks!
<mmichelson>
Anybody else to give an update today?
<mmichelson>
All right, thanks everyone. Have a good day!
<mkalcok>
\o
<fnordahl>
\o thanks, cheers!
<dceara>
Thanks, bye!
<mmichelson>
bye!
<imaximets>
Thanks! Bye.
<moloings>
Thanks, Bye!
mmichelson has quit [Quit: Leaving]
mkalcok has quit [Quit: leaving]
moloings has quit [Quit: WeeChat 4.6.3]
<zhouhan52>
thank, bye
dceara has quit [Quit: Leaving]
fnordahl has quit [Remote host closed the connection]
zhouhan52 has quit [Ping timeout: 250 seconds]
jlibosva has quit [Quit: WeeChat 4.5.1]
elvira has quit [Ping timeout: 260 seconds]
kuraudo has quit [Remote host closed the connection]
kuraudo has joined #openvswitch
kuraudo has quit [Remote host closed the connection]