I have an interesting situation which I have not seen before. Maybe someone can help. We have a Cisco 2921 with IOS 15.1 connected to the internet on Gi0/0. The 2921 is doing NAT as that is the border router.
Gi0/1 of the 2921 is connected to a Cisco ASA 5505 on Eth0/0 (outside). Eth0/1 of ASA is connected to a Layer 3 switch (3750).
This is a very common deployment scenario. Now the problem:
We are using a radius server behind the ASA on the inside interface to authenticate our VPN users. VPN termination is NOT done on the ASA instead on the 2921. ASA is running in routed mode.
The interesting thing I am noticing is when Windows Server 2008 receives an authentication request from the router, the source address in the packet capture is the ASA's inside address. So obviously the server responds back to that address with success however that request never reaches the router. The 2921 keeps waiting for a response from Radius and then times out.
However if i ping from the 2921 to this Radius server and check the packet capture using Wireshark, the source ip on the packet is correct (2921's gi0/1). Since the ASA is not doing NAT (no nat-control) looks like somewhere on the ASA there is a problem.
We don't have any outbound ACL's on the ASA and for testing I opened up everything using permit ip any any and still the same result. Since ping's can reach from the 2921 to the radius server, i did rule out ACL.