Multiple default routes / backup internet site.

[Bill19795_2]

I have 15 sites connected via MPLS. Currently all of the sites go through the main location for internet access. I am doing this by using the default-information originate command within BGP out to the MPLS provider. I am also using EIGRP internally and redistributing the routes learned from BGP into EIGRP. I would like to add a redundant site for outbound internet connectivity. I only want to use this site internet when the main location internet is down. The main site may or may not be up but the internet would be down. I will use IP SLA to remove the route if the main internet connection is down.

My question is how can I influence routing so I only use the backup site internet when the main internet is not available? I tried putting a static route in with a higher AD and that works when I take the main internet route out but it does not go back to normal when I put the main internet site as the primary.

In the diagram attached the HQ site is the main site, RS1 is the backup site and RS2 is a normal remote site.

[Bill19795_2]

So I think my problem is that when HQR1 receives the redistributed route from HQR3 it has an AD of 1

70. This is fine for every other external EIGRP route but I want to set just the default route to a higher AD of BGP 20.

I tried to use example in the following document and it does not work. Any other suggestions?

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a00800c2d96.shtml

Change the Administrative Distance at R2

It is also possible to alter the path selection process by changing the administrative distance, on R2, of the route learned from R4. This method is less ideal than the others. It can increase the potential for routing loops for the reasons listed here:

• Administrative distance is generally used to determine the method by which a route was learned. If set incorrectly, the individual router cannot choose a redistributed route instead of the actual best path.

• Administrative distance is not propagated to other routers. Routing protocols rely on the fact that all routers choose the same path given the same set of parameters. Altering parameters on a single router can lead to routing loops.

1. Change R2's configuration so that when a routing update from R4 (10.1.2.4) for subnet 10.1.3.0/24 is heard, the administrative distance is increased to 91.

   91 is chosen because it is 1 greater than the default EIGRP administrative distance for internals (which is 90). The default administrative distance for EIGRP externals (routes redistributed into EIGRP) is 170. Refer to the Default Distance Value Table in the What Is Administrative Distance? document for the default values of all routing protocols.

   R2# configure terminalEnter configuration commands, one per line.  End with CNTL/Z.
   
   R2(config)# access-list 99 permit 10.1.3.0 0.0.0.255R2(config)# router eigrp 1R2(config-router)# distance 91 10.1.2.4 0.0.0.0 99R2(config-router)# end
   
   R2#

[rtjensen4]

Do you want to advertise the other default route via BGP to the rest of the network? Here's what I do for an almost identical setup. I have MPLS WAN with 18 locations, a primary datacenter and a DR datacenter. BGP on the WAN, OSPF at / DR locations, inernet at HQ location, DR internet only to be used if primary is offline:

At HQ:

router bgp 64519
...

network 0.0.0.0
....

no auto-summary
!

track 190 rtr 190

!
ip route 0.0.0.0 0.0.0.0 track 190

ip route 255.255.255.255

at DR:

router bgp 64519
....
bgp log-neighbor-changes
network 0.0.0.0
...
neighbor remote-as xxxx
neighbor route-map AS-PREPEND out
no auto-summary
!
ip route 0.0.0.0 0.0.0.0
!
ip prefix-list default-route seq 5 permit 0.0.0.0/0

!

route-map AS-PREPEND permit 10
match ip address prefix-list default-route
set as-path prepend 64519 64519
!
route-map AS-PREPEND permit 20
!

At HQ, I created an IP SLA to ping the next-hop IP of our internet connection and tie it to a static route via track 190. General idea is if the router can't ping "the internet", then it's down.whether the circuit is actually down or the firewall is down or whatever.

Both HQ and DR advertise a default route via BGP to the MPLS cloud via network 0.0.0.0. The default out of DR location has a longer as-path, and is not seen as primary.

When the internet stops replying, default route out of HQ is no longer advertised (removed via track command), this allows the default route from DR to take over. The other default route gets installed at all branch locations as well as HQ. When Internet at HQ is back online, the route is re-installed via BGP and since it has a better AS-Path, it's used as primary.

Note that I have to have a host-level route for the next-hop IP of the internet at HQ. This is because once the router will keep trying to ping the next-hop IP via the DR internet connection. In my case, the next-hop IP isn't pingable via my DR internet connection, so it would never fail back to primary. I put in the host-level route to ensure it goes through the HQ firewalls to get to it.

HTH