So the questions are: 1) is it possibile to announce/inject the same external subnets on two different egress nodes? If yes, which one will be preferred? Is there a concept of metric or cost or backup-routes or load-balancing of any sort? 2) is there a similar case study, or someone willing to share her experience on this? 3) is there relevant documentation or code comments or whatever on the route selection mechanisms? It should be something happening at the wireguard level (cryptokey routing) but nodes send/receive configs via netclient and then configure themeselves using that config so it's a kind of coop.

Currently it is not possible to have an egress with an overlapping range on the same network

Hey dlrow_olleh, thanks for the info. Does "currently" mean that it's planned? If so, is it a long-term goal or something closer?

not planned at this time.

ok, I think I can figure out a way to make it happen anyway

one way to accomplish it is to have multiple networks... and nodes connect/disconnect from networks to reach each egress as required.

that's why I asked the q above 😉

lets say you have egressA and egressB both with the same egress range. They are on networks A and B respectively. If nodeC needs to reach egressA it joins networkA. If it subsequently needs to reach egressB it disconnects (not leave) netA and joins netB. Then it then needs to reach A again, disconnect from B and connectA. So nodeC is on both networks but only connected to one network at a time. In the case where nodeC is connected to both networks at the same time, the egress range that would be reachable is the one from the first network it connected to.

I was thinking of a different scenario: 3 networks (a,b,c), 100 clients, 4 egress nodes. Clients can use one network and only one. The egress nodes are connected to all the networks and route traffic towards endpoints on different VPCs. Since the networks are separated, there's no overlapping but this works only if I can connect to all the networks from the egress nodes. Wireguard is fine with this, what about netmaker?

as long as the egress ranges do not overlap all egress gateways can be on all networks

the egress nodes will be announcing the same ranges to the three (netmaker) networks. So, there's no overlapping on the single netmaker network but there's multiple range announcements for the same networks on the egress nodes. I wonder what the egress node does in this scenario: ideally, it should NOT modify its routing table since that is already done by the routing protocols. I just need a way to "attract" (netmaker) traffic from the egress clients and then route it through the rest of the network. For the return traffic, I'd "attract" traffic destined TO the netmaker clients by announcing the netmaker subnets in the routing protocol and that's it. Does it sound doable to you?

the egress node(s) have to forward traffic between the netmaker network and the egress range. netmaker nodes will get an update to their routing table to send data to the egress range via the netmaker network. To route traffic from the egress range, either the egress gateway will do NAT (default) or routing on egress range machines needs to be updated (manual process)

ok, if it's manual and the egress nodes do no try to mess up with their own routing table (and thus the routing protocol) I think it's doable. I just need to find a way to automate the range announcements start/stop in case of (egress) node failure. If you, future reader, have found a way, let my future self know, please 🙂