Can anyone throw any light on this.
I have a small network connecting 2 x sites together. The LAN's are
192.168.143.X/24 and 10.10.X.X /16 respectively. All the routers WAN IP's are 10.250.X.X/16.Between RTR1 and RTR3 I have a point-to-point GRE Tunnel 172.10.1.1 &
172.16.1.2 /24. Between RTR2 and RTR4 I run a DMVPN with RTR5 (the NHRP server). The DMVPN Tunnel IP's are 172.16.0.1, .2 & .3 /24 repectively). I use a route-map on RTR1 and RTR3 that prioritise important IP traffic to go over the RTR1 to RTR3 Tunnel. Any traffic not matched is forwarded normally over the DMVPN. NB RTR1 & RTR3 are HSRP active routers for their LANs.On the RTR1 - RTR3 connection the bandwidth is 32k and delay 5000. On the RTR2 - RTR4 link the bandwidth is 256k and delay 1000. This means all traffic should prefer the RTR2 - RTR4 link due to the better metric assuming no route-map match.
I advertise my 10.250.X.X WAN in RIP (including no validate update source + passive interface on he LAN). I advertise my Tunnels and LAN in EIGRP.
My problem - RTR1 & RTR3 (on their respective LAN's) think the best path to the remote LAN is via RTR2 & RTR4. However, RTR2 & RTR4 think the best path is via RTR1 & RTR3.
Is it possible that the switch simulating the WAN may be throwing the routing metrics out when RIP and EIGRP routing updates are being learned by the routers.
I don't redistribute between routing protocols.
Diagram:
192.168.143.0 /24 | | Rtr3 Rtr4 (DMVPN) | | (WAN - Catalyst Switch) 10.250.X.X ----Rtr5 (DMVPN) for Rtr 2 & 4 | | Rtr1 Rtr2 (DMVPN) | | 10.10.0.0 /16