I've been playing around with Azure's vnet offerings and I can't seem to get a simple network topology working. I am using the Resource Manager model. It seems that Azure's capabilities for even the simplest set ups is incredibly limited. The scenario that I want to model is one where I have one local network trying to communicate with Azure using a PolicyBased (IKEv1) VPN, and various clients that need to connect to the network independently (P2S). PolicyBased VNetGateways do not support P2S, so I need to have two gateways, which seems to be the root of the problem.
My setup then becomes this:
_____ _______ ______
|LAN | |VNet1 | |VNet2 |
| M0 |--S2S (PolicyBased)--| |--Peering--| VM |--P2S--(Client1)
|_____| |_______| |______|
And what I need to accomplish is:
- Client1 must be able to communicate with VM
- VM must be able to communicate with M0
Issue 2 is the one that I can't get to work. A machine in VNet2 is able to communicate with a machine in VNet1, and a machine in VNet1 can see M0. However, VM cannot see M0. I can get it to work if I remove the vnet gateway in VNet2 required for P2S and having VNet1 act as a gateway for VNet2, but this obviously fails requirement 1.
Any suggestions? In the documentation for VNet peering they seem to hint that RouteTables (UDR) could be a solution. That is, routing requests from a subnet to a Network Virtual Appliance (NVA). However, I can't find any lightweight NVA's for simple routing, and I'm not sure if this would even work.