Interfaces on the mesh have an IP in the range 10.0.0.0/8. Devices you connect to the LAN behind the node will have IP addresses in the 10.x.x.x/29 or /30 range. These ranges overlap. Mesh interfaces will see all LAN addresses as part of their local link, but LAN device interfaces will see the wider mesh as a separate subnet. If two LAN subnets have the same network prefix, each subnet will be unable to access the other because they will see the foreign addresses as part of their own subnet.
I'm sure there's a reason for this? Is it because it's too expensive to do duplicate address detection? Why not use IPv6? That would pretty much eliminate the risk of address conflicts.
I'm sure there's a reason for this? Is it because it's too expensive to do duplicate address detection? Why not use IPv6? That would pretty much eliminate the risk of address conflicts.
OK, this was a very simple explanation. Someone else may expand on it.
+1 K6CCC
NE5IE:
Not just an overlap, but the local 'LAN address' network is completely inside the 'Primary address' network.
In the AREDN firmware there is logic to assigning a 'LAN address' network that attempts to avoid a conflict.
Should you find a conflict, you may manually set your 'LAN address' network.
The largest allowed network is /27 and that allows for 262,144 subnets.
The default (/29) allows 1,048,576 subnets.
So, as an AREDN network approaches one million nodes, there is an ever increasing chance of a conflict.
" 'mesh' is a topology, not a protocol."
73, Chuck