IPv6 ND not seemingly being updated for default-gw/next-hop

[f0o]

Hi,

let me start with that I'm relatively unsure if I traced the issue correctly or not but hear me out;

I've got a standard OpenStack setup with OVN and 2 dedicated Gateway Nodes that host ovsdb-server, ovs-vswitchd, ovn-controller and neutron-ovn-metadata-agent as per refarch,

However after some hardware failure of rt1 the setup runs on the 2ndary host (rt2) and while IPv4 flows fine, all IPv6 traffic is being dumped into the ovs-bridge with the destination of rt1's MAC leading to a full IPv6 outage.

If I ip link set address 11... on rt2 with rt1's MAC then IPv6 resumes.

So my assumption is that somewhere in the flow/s or router setup the Neighbor Discovery of the gateway (anycast' ::) is stale and it keeps the old rt1 MAC instead of refreshing it to obtain the new rt2 MAC.

Is my assumption correct?

If so - how do I "flush" the ND cache?

I checked with ovs-ofctl dump-flows br-int | grep 11... but no results; likewise with ovn-sbctl list logical_flow | grep 11... :/

Are there any other things I can try to debug this?

Thanks in advance!

//EDIT:

I also checked the Logical_Router_Static_Route but there's nothing notable either, just the ip_prefix ::/0 and the anycast-router next-hop. How do I see what OVN believes the ND/MAC is for a given next-hop ?

//EDIT2:

Issuing ndptool -v -U -t na -i vlan1234 -T dead:babe:: send fixed it. So my suspicion of stale NDP is correct; why didnt it re-resolve the NDP? It's been hours since