BGP runs over the tunnel it's just that both routers are using the same AS so the routes learned over the tunnel are iBGP and are preferred over the eBGP routes learned from the MPLS network.

My issue with EIGRP is that I'm already using EIGRP locally at each site with different AS's. If I was to use the same AS in both sites I believe I would run into the same problem, EIGRP routes learned via the tunnel being preferred to the routes being redistributed from BGP into the local EIGRP AS.

I am still experimenting with different options.

BGP runs over the tunnel it's just that both routers are using the same AS so the routes learned over the tunnel are iBGP and are preferred over the eBGP routes learned from the MPLS network.

I suggest going with BGP as you initially stated. You can alter the route preference by using the correct BGP attribute.

In your case, the iBGP route is chosen over the eBGP because the length of the AS_PATH.

If you review the BGP Best Path Selection, WEIGHT and Local_Pref go before the AS_PATH on the route selection so if you manipulate those attributes, eBGP/MPLS route will be preferred, please refer to the documentation:

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a0080094431.shtml

HTH,

__

Edison.

Please rate helpful posts

I definitely considered that as an option. In fact in the lab I was able to get it to work using the "local-as" command which allowed me to change the AS on one of the routers (at the central site for example) while spoofing the original AS that the provider is expecting.

router bgp 65002

neighbor 192.168.1.2 remote-as xxxxx

neighbor 192.168.1.2 local-as 65001 no-prepend replace-as

Trying to manipulate the iBGP routes the way I want is proving to be rather challenging and seems to require not only weighting of the preferred eBGP routes, but extensive route filtering as well to keep routes from looping all over the place. Not to mention the challenges of redistributing iBGP routes into an IGP.

hi

what do you think of floating static routes?

Cheers

Floating static routes were not something I was familiar with until you mentioned them.  Certainly seems simple enough.  I suppose I would have to redistribute both BGP and the floating statics into the IGP at each site to achieve the desired results.

The only downside I can see would if later down the road new networks were added to BGP without the corresponding floating statics being added as well.

Thanks for the suggestion.

Hi

this will depend on your overall network design.

We use as primary network a dmvpn network with most sites connected with a simple rip setup and for backup we use isdn with floating statics. the sites have only one route for LAN and WAN, so the setup is simple. for sites with higher demands we have a second dmvpn network with eigrp running and all sites have two wan routers and redundant LAN routers

so we have load sharing and failover.

the management of the floating statics are not so complicated because from the remote site only one route is needed an and in the datacenter also. management is done by tools like the rme from cisco or with scripts like rancid.

for your situation it may be a solution to use mpls as transport for some kind off vpn and to ignore the provider bgp. or use bgp over the vpn.

it highly depends on what your requirements are.