OSPF route redistribution across separate OSPF processes

← RouterOS basic OSPF router configuration | posts | IOS-XR BGP network advertisement →

In short I have been doing a networking lab with route leaking using OSPF. To keep things simple the idea has been to use only BGP and OSPF as dynamical routing protocols. The OSPF redistribution router has 2 OSPF processes and can do mutual redistribution between these two running OSPF processes. The redistribution router has following tasks:

OSPF process 1 has to pick up the default route from the global routing table and redistribute it into the OSPF process 10, the local routing table
OSPF process 10 has to pick up OSPF prefixes from the local routing table and redistribute to the global routing table OSPF process 1

Almost all things have worked with mutual redistribution, but the default route redistribution did not work as expected . Redistributing 0.0.0.0/0 did not work with following:

distribute lists
prefix-lists
route-maps
access-lists

After reading further documentation I have found Ivan's Papelnjak's blog post about this process, and have discovered a additional documentation about OSPF Redistribution among different OSPF processes. At this point the RFC1925 comes to mind paragraph (8) It is more complicated than you think.

A few days a more simpler solution came to my mind. Each routing process advertises the default network route instead of redistributing it. The simpler solution has been to advertise a default route to the local OSPF process with higher OSPF metrics. The default route in the global routing tables has lower OSPF metrics. Interesting has been, which I have not been aware of, IOS tagged routes have their tag still attached after redistribution. The redistribution router has following configuration:

configure terminal
!
!global routing table
!
router ospf 1
 router-id 10.255.255.200
 auto-cost reference-bandwidth 1000
 redistribute ospf 10 subnets
!
!local routing table
!
router ospf 10
 router-id 10.0.3.0
 auto-cost reference-bandwidth 1000
 default-information originate always metric 10000
!
end

The redistribution is done in the global router ospf 1 process. Importing IP OSPF prefixes from router ospf 10 process, using following command:

redistribute ospf 10 subnets

The local ospf process router ospf 10, advertises the default route with the command:

default-information originate always metric 10000

The default route metric in the global routing table is set to 1000 which is a smaller value than metric of 10000. This is what has been configured on the router that is doing route leaking . Leaking routes from the local routing table to the global routing table. This router has 2 separate OSPF processes, and is lacking MPLS/BGP/VRF configuration.

The router that has MPLS/BGP/VRF configuration is directly attache to the mutual redistribution router, uses following router configuration:

configure terminal
!
!local routing table
!
router ospf 10 vrf vpn10
 router-id 10.0.3.2
 auto-cost reference-bandwidth 1000
 redistribute bgp 65001 subnets tag 200
!
!global routing table
!
router ospf 1
 router-id 10.255.255.1
 auto-cost reference-bandwidth 1000
!
!bgp routing table
!
router bgp 65001
 bgp router-id 10.255.255.1
 bgp log-neighbor-changes
 neighbor 10.255.255.60 remote-as 65001
 neighbor 10.255.255.60 update-source Loopback0
 neighbor 10.255.255.61 remote-as 65001
 neighbor 10.255.255.61 update-source Loopback0
 !
 address-family ipv4
  neighbor 10.255.255.60 activate
  neighbor 10.255.255.61 activate
  maximum-paths ibgp 4
 exit-address-family
 !
 address-family vpnv4
  bgp nexthop trigger delay 0
  neighbor 10.255.255.60 activate
  neighbor 10.255.255.60 send-community both
  neighbor 10.255.255.61 activate
  neighbor 10.255.255.61 send-community both
 exit-address-family
 !
 address-family ipv4 vrf vpn10
  network 0.0.0.0
 exit-address-family
!
end

The default route again is not redistributed from local routing table router ospf 10. The BGP process announces the default route with the command:

address-family ipv4 vrf vpn10
 network 0.0.0.0
exit address-family

This way the default route is send to the BGP route reflectors. The BGP route reflectors announce the default route to the BGP route reflector clients.