Multiple networks in s2s

GoodCloud, AstroWarp and AstroRelay
1

i want to try having multiple s2s networks configured to avoid high latency between spoke to spoke traffic. for example.

i have 3 routers in total, like R1,R2 & R3.

s2s network one: This network will have R1 as hub, while R2 & R3 are spokes
s2s network Two: This network will have R2 as hub, while R1 & R3 are spokes

please advise.

2

Pretty sure this won't work because routes should be unique.

Maybe you should look into Tailscale or ZeroTier instead?

3

for sure routes will be unique. moreover, s2s works with masquerade between the nodes so i think that should be fine.

4

Why do you want to use masq between the nodes?

5

Im not using it. Its already used by design

6

@admon, any thoughts on it ?

7

tl;dr -> static routes are needed.

I think the terminology is wrong here.

Three sites, three routers.

In general all communication will be send to one point and router there. Let's say one HQ and two branches.
But now we want to make all tree sites equal. Be aware you can have only one default route! A common issue in this kind of construct.

So set up each of the three servers.
Site A: 192.168.10.0/24
Site B: 192.168.11.0/24
Site C: 192.168.12.0/24

And set up a WireGuard Server on each router. Set up two client connections, each.

If site B connects to site A, it should be able to access from client 192.168.11.2 the system 192.168.10.2.
If you activate the second tunnel from site B to site C, in worst case this connection will overwrite this. You could access 192.168.12.2, but not 193.168.10.2 anymore. But as long as the tunnel to site A is active, you could add a static route, so the route on site B knows that 192.168.10.0/24 is reachable behind [IP of WireGuard tunnel within your network] ... I forgot to bring the tunnel networks in. Just make sure they are unique.

8

So basically masq happing on the tunnel networks is what i would take advantage off.

So lets say following are the tunnel networks.

  1. 10.1.1.0/24, for which R1 is main Node and R2 & R3 clients.
  2. 10.1.2.0/24, for which R2 is main Node and R1 & R3 clients.
  3. 10.1.3.0/24, for which R3 is main Node and R1 & R2 clients.

So, basically each networks main node will announce its lan subnet and relevant network subnets behind its main node. This way static routes can be avoided and i can avoid routing loops and network lags between any two nodes.