BGP route leaking eigrp redistribution

[Dallas Brown]

Hello,

I have an issue with redistributing eigrp into bgp. I have a router dual connected to other routers, lets call it a routerleaker. Its sole purpose is to leak routes between vrf's. It is is peered eigrp with the vrf routers and has an instance of bgp running on it to do the route leaking. I am find that when the routerleaker router has more than one equal cost route from the upstream routers as show below the redistribution breaks and bgp shows the eigrp routes as learned from default route(which does not exist, there are no default routes anywhere). If I shut one of the links to the upstream routers leaving only one usable eigrp route to redistribute, all is well.

routeleaker#show ip route vrf vrf120

D        10.120.0.0/18
           [90/25600] via 172.20.1.81, 00:32:17, GigabitEthernet0/1.120       
           [90/25600] via 172.20.1.77, 00:32:17, GigabitEthernet0/2.120
D        10.120.120.0/21
           [90/25600] via 172.20.1.81, 00:32:17, GigabitEthernet0/1.120
           [90/25600] via 172.20.1.77, 00:32:17, GigabitEthernet0/2.120

routeleaker#show ip route vrf SHARED

B        10.120.0.0/18
           [20/25600] via 0.0.0.0, 00:00:55, GigabitEthernet0/1.120
B        10.120.120.0/21
           [20/25600] via 0.0.0.0, 00:00:55, GigabitEthernet0/1.120

routeleaker#show ip bgp vpnv4 vrf vrf120 10.120.0.0
BGP routing table entry for 65535:120:10.120.0.0/18, version 1048
Paths: (1 available, best #1, table vrf120)
  Not advertised to any peer
  Refresh Epoch 1
  Local
    0.0.0.0 (via default) from 0.0.0.0 (1.1.1.1)
      Origin incomplete, metric 25600, localpref 100, weight 32768, valid, sourced, best
      Extended Community: RT:65535:120 Cost:pre-bestpath:128:25600
        0x8800:32768:0 0x8801:120:1024 0x8802:65283:2560 0x8803:65281:1500
        0x8806:0:0
      rx pathid: 0, tx pathid: 0x0

Here is when I shut one link, all is well

routeleaker#show ip route vrf vrf120

D        10.120.0.0/18
           [90/25600] via 172.20.1.77, 00:00:16, GigabitEthernet0/2.120
D        10.120.120.0/21
           [90/25600] via 172.20.1.77, 00:00:16, GigabitEthernet0/2.120

routeleaker#show ip route vrf SHARED

B        10.120.0.0/18
           [20/25600] via 172.20.1.77 (vrf120), 00:00:46, GigabitEthernet0/2.120
B        10.120.120.0/21
           [20/25600] via 172.20.1.77 (vrf120), 00:00:46, GigabitEthernet0/2.120

routeleaker#show ip bgp vpnv4 vrf vrf120 10.120.0.0
BGP routing table entry for 65535:120:10.120.0.0/18, version 1115
Paths: (1 available, best #1, table vrf120)
  Not advertised to any peer
  Refresh Epoch 1
  Local
    172.20.1.77 (via vrf vrf120) from 0.0.0.0 (1.1.1.1)
      Origin incomplete, metric 25600, localpref 100, weight 32768, valid, sourced, best
      Extended Community: RT:65535:120 Cost:pre-bestpath:128:25600
        0x8800:32768:0 0x8801:120:1024 0x8802:65283:2560 0x8803:65281:1500
        0x8806:0:0
      rx pathid: 0, tx pathid: 0x0

vrf config

ip vrf SHARED
 rd 65535:1
 route-target export 65535:1
 route-target import 65535:120
 route-target import 65535:121
!
ip vrf vrf120
 rd 65535:120
 route-target export 65535:120
 route-target import 65535:1

bgp config

router bgp 65535
 bgp router-id 1.1.1.1
 bgp log-neighbor-changes
 !
 address-family ipv4 vrf SHARED
  redistribute connected
  redistribute static
 exit-address-family
 !
 address-family ipv4 vrf vrf120
  redistribute connected
  redistribute eigrp 120
 exit-address-family

[eduardopozo56 · Accepted solution]

By default BGP only selects 1 route as the best (EIGRP supports multiple routes by default), have you tried enabling multipath support on your BGP config?

The following document might help you with the configuration:
https://learningnetwork.cisco.com/docs/DOC-11233

Update: the multipath command wont work on your case, as is only used for MP-BGP received routes not for route redistribution within IGP and BGP. The issue is that EIGRP has this special feature of allowing multipathing which other protocols don't

In order to allow you scenario (import all eigrp load-balanced routes from one VRF to another) you can not use BGP for the leaking between the routes, you have to use other feature as the "route-replicate" command under vrf definition for EVN services:

http://www.cisco.com/c/en/us/products/collateral/ios-nx-os-software/layer-3-vpns-l3vpn/whitepaper_c11-638769.html

I just did this on a lab environment, and adding the command:
"route-replicate from vrf vrf120 unicast all" into the address family for vrf SHARED, successfully imported all the routes to the VRF without the need for BGP (I removed all bgp config from the test router). You could have more granularity using route-maps for the replication

[Dallas Brown]

Thank you for the response. The first issue I had was actually a code problem. I downgraded the code and the problem of having the two links up and the route changing to a directly connected went away. Then I wan into exactly what you are explaining here. So thank you for that. I ran into the wall of trying to get two routes into MP-BGP and could not make it work, so I am glad you posted this update.