Why does the Nexus 7k Adjacency Manager insert ARP-learned routes into the IPv4 routing table with an Administrative Distance/Preference of 250?
I'm not sure why is that but apparently Cisco sets it by this way on purpose:
Nexus#show run all | i "ip adjacency"
ip adjacency route distance 250
Meaning that all routes installed by Adjacency Manager have AD set to 250. In my opinion this could be very dangerous because if by any chance there is same /32 route installed by any routing protocol (e.g. OSPF or BGP) it would have lower AD and would be preferred over directly attached route installed by Adjacency Manager. Actually I have seen this on one of our Nexus switches where vPC keepalive peer link was broken because of this.
You can set AD of routes installed by Adjacency Manager to lower value by command:
ip adjacency route distance <value>
I'm not sure but I would say that older NX-OS versions had AD value for Adjacency Manager routes set to 2. Not sure why Cisco changed it to 250 in recent NX-OS versions.
I know im 2 years late to the party but im running into this very issue. Whats happening is im trying to force path control using host routes, but OSPF is propagating the host route and overriding the AM host route in neighboring Nexus Switches because OSPF has a lower AD (110) than the AM (250). This makes zero sense to me as to why a directly connected AM route learned from an ARP would be less attractive than one learned from literally ANY routing protocol. Just thought Id share my plight and say I found this thread very helpful.
Do you recommend setting the AD of the AM learned routes to 2? Any adverse affects from this? Thanks.
This change is to support host mobility. In an overlay extended L2 network design e.g. VxLAN. If a host is moved from one site to another. It's preferred to take the remotely newly learned /32 host route by BGP rather than the no longer existed but not yet expired local AM host route entry. HTH