Weird arp behavior

[Russell Handorf]

Hi all,

So, I've deployed Tinc in a non normal manor to which has been working 
just fine for days up until recent. Here is a description of the network.

There are two VPN host controllers, A and B. All nodes have a separate 
connection to A and B. Tinc has been configured to be in 'switch' mode 
for both the A cloud and B cloud. On the back end, I have OSPF running 
on all nodes and the host controllers which are populated the routes in 
the various clouds so that A is the primary route and B is the secondary 
route.

The OSPF routes are populating just fine, and tcpdump on all nodes shows 
this. The problem is ARP. From the host controller A, I can ping all the 
nodes and all the nodes can ping the host controller. Same as B. NONE of 
the nodes can ping each other. When I packet dump the Tinc interfaces 
across the nodes I see them sending ARP requests ("who has"). No one 
sends a response. Sometimes, and now very rarely, a response is sent. If 
I static the ARP entries, everything works.

I know Tinc keeps its own ARP tables, and I dabbled with ifconfig 
 >interfacename< -arp just to see what happens, and of course that 
doesn't help.

Does anyone have any ideas as to why ARP isnt working?

Thanks!