BGP withdrawing the best path

[lmessenger]

Hi.

I have a problem where the best path to a particular destination is being withdrawn by BGP.

To explain.

Site A has 2 links to site B. 1 via an eBGP peer over MPLS, 2 via an iBGP peer over a VPN backup.

I have configured the ebgp peer with a higher weight so that it is preferred.

The problem is this.

When the link to the ebgp peer goes down, the link via the ibgp peer is preferred.

When the link via the ebgp peer comes back up, the route via the ebgp peer does not go back into the BGP table (in fact it goes back for a second and is then withdrawn).

Could someone help me on this one.

Thanks

Lee

[Harold Ritter]

Can you tell us more about the topology of site A and B. Is the iBGP and eBGP session from the same two CEs? A sample topo would be nice.

Thanks,

[lmessenger]

Hi,

thanks for the reply, I have attached a diagram of the topology and the BGP configs of all 3 routers is below.

One other thing, the only way I can get the preffered route to appear back in the BGP table once the MPLS link is back is to do a hard reset of the peer connected via the backup link

Thanks in advance

Lee

RTR A Config

router bgp 65500

no synchronization

bgp log-neighbor-changes

bgp redistribute-internal

network 10.3.64.0 mask 255.255.224.0

redistribute connected metric 10

redistribute static metric 10

redistribute ospf 10 metric 10 match internal external 1 external 2

neighbor RTR B remote-as 65500

neighbor RTR B next-hop-self

neighbor RTR B route-map NEXTHOP in

neighbor MPLS remote-as 65000

neighbor MPLS soft-reconfiguration inbound

neighbor MPLS weight 200

no auto-summary

RTR B Config

router bgp 65500

no synchronization

bgp log-neighbor-changes

network 10.3.64.0 mask 255.255.224.0

redistribute connected metric 10

neighbor VPN_BACKUP_MESH peer-group

neighbor VPN_BACKUP_MESH remote-as 65500

neighbor VPN_BACKUP_MESH route-reflector-client

neighbor RTR C peer-group VPN_BACKUP_MESH

neighbor RTR A peer-group VPN_BACKUP_MESH

neighbor 10.3.77.33 weight 200

no auto-summary

RTR C Config

router bgp 65500

no synchronization

bgp log-neighbor-changes

network 10.1.144.0 mask 255.255.240.0

redistribute connected metric 10

neighbor RTR B remote-as 65500

neighbor RTR B next-hop-self

neighbor RTR B soft-reconfiguration inbound

neighbor RTR B prefix-list SYDNEY_ROUTES out

neighbor RTR B route-map NEXTHOP in

neighbor MPLS remote-as 65000

neighbor MPLS weight 200

neighbor MPLS soft-reconfiguration inbound

no auto-summary

[Harold Ritter]

The issue is that the PE router prefers the route received via the CE (iBGP) rather than the one received via the other PE (iBGP).

On the site A CE, try setting the MED for site B routes sent to the MPLS cloud to 2 and 1 for local routes. The same should also be done on the site B CE. This will ensure that when the eBGP session comes back up, the PEs will prefer the route from the directly connected PE.

Hope this helps,

[lmessenger]

Hi hritter

Thanks again for your help. I have tried to do what you suggested however I am not quite sure I understand what you are asking me to do.

Is there any chance you coudl find the time to write down the config changes I would need to make on each of my routers ?

Thanks

Lee

[Harold Ritter · Accepted solution]

It would go something like this:

RtrA:

router bgp 65500

neighbor MPLS route-map setMed out

route-map setMed permit 10

match ip address 1

set med 2

route-map setMed permit 20

set med 1

access-list 1 permit

RtrC:

router bgp 65500

neighbor MPLS route-map setMed out

route-map setMed permit 10

match ip address 1

set med 2

route-map setMed permit 20

set med 1

access-list 1 permit

Let me know if you have any issues,

[lmessenger]

Hritter,

That works perfectly, thanks very much for your help.

Lee