Hi,
Returning back to this issue after a while as I'm migrating ml2/ovs to
ml2/ovn.
Seems like the same issue from 2025 still persists.
refs:
[0]
https://mail.openvswitch.org/pipermail/ovs-discuss/2025-February/053456.html
[1]
https://mail.openvswitch.org/pipermail/ovs-discuss/2025-March/053484.html
case:
Big L2 domain with border device learning IPs by flooding ARP. For some
reason in case L2 device (nic with VLANs attached) is attached to the
br-ex bridge, then after a while OVN stops sending DHCP packets
(OFFER/ACK/etc).
Anybody else observed the same issue? The only way to stabilize region
is to switch to L3 networking using ovn-bgp-agent (eth0 is detached from
br-ex, so no more ARPs delivered to ovn-controller), but there is
monstrous overhead using kernel routing: IRQ higher up to x5-6 times,
like 10-12% while L2 networking is just 2% which is fine.
Meanwhile, no errors, warns, resubmit logs in logs.
--
Regards,
Ilia Baikov
[email protected]
_______________________________________________
discuss mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-discuss
