Hi
having a few problems with our EIGRP WAN over a L2 VPLS. We have primary and backup fibre or ADSL lines at almost all sites (55 in total) with 2 core sites that have 4506s (Supervisor 7E) and a non-core site with a 4503. The non-core sites either have a 100mbit primary and 10mbit backup or 10mbit primary and ADSL backup. The 2 core sites are 1gbit primary and backup. The fibre is sourced from different providers, but both of them are run through separate NTUs from the primary provider, primary subnet is a xx.xx.40.0/24 and backup is xx.xx.20.0/24. The ADSL lines are sourced by the backup provider and there is a peering router in the same subnet as the backup lines that links into our EIGRP process and then into a core site. Every site has a default-network set to their nearest core site's loopback interface and delay 1000 is configured on the backup interface as well as the correct bandwidth setting. the ADSL only sites have a default route pointing at the dialer and the service provider does some BGP then into EIGRP on the peering router they have set up for us.
The issue we have is that our 4506 core switches connected to the 2 NTUs from the provider are flapping constantly. They are seeing the same MAC address bouncing between both of them and our adjacencys are going up and down every minutes or so - the ports were set up as access ports with a vlan, and then their 40.x/24 and 20.x/24 were on these VLANS. All the VLANs on the 4500 have the same MAC address. When you change the interface to a Routed port and set an IP address, it still has the same MAC address as all the VLANS. We can't change the MAC address on the VLAN or interface for some reason here, no mac-address command from the interface config. Is there any other way of doing this?
Our solution at one of the core sites has been to put a 3925 between the core router and the provider NTU which seems to be mostly working at the moment (with some manipulating of the delays but i'd rather not do this - it's getting out of hand tracking delays across this mean routers), but its not always picking it as the preferred route. Sometimes it would prefer to go through a slower 10mbit .20.x interface to hop onto the 40.x network on its way to the core site. Nothing should be using the 20.0/24 interfaces if the primary lines are up, but the ADSL connections have to use them all the time to link into the network, and we're getting a whole bunch of packet loss across the core site with the flapping MAC/route.
I can put another 3925 or 2911 between the 4500 and the NTU but i'd much prefer a cleaner option in fixing the 4500s if its available but i can't seem to find anything on google about it? Or some other way of influencing the EIGRP to use the proper routes?
Thanks
