internal networks unreachable if iBGP neighbor wan goes down

[Hasrat Raja]

Hello,

I have 2 Nexus 5000's as iBGP neighbors, both have their own wan access, if wan goes down on one of the nexus, I loose access to internal networks even from neighbor router with active wan, I can only ping physical interfaces on the nexus.

I have attache BGP config for reference.

thanks

[de1denta]

Hi,

Can you post the output of 'show ip bgp' when both WAN interfaces are up and when one of the WAN interfaces is down?

Thanks,

[Hasrat Raja]

Thank you for your response, I don't have sh ip bgp when circuit is down, I have attached current with both circuits up.

regards

[de1denta]

Hi,

Thanks for these.

I can see in the show ip bgp outputs that the ibgp path to all destinations is using the next-hop address of the opposite N5ks WAN connection. For example dist-1 is using next-hop address 10.50.132.18 which is the WAN connection for dist-2. For BGP to consider a path valid, the router must be able to route to the advertised next-hop addess.

Can you check if dist-1 has a route for 10.50.132.18 this address in its local RIB?

The normal way to resolve this issue is to set 'next-hop-self' for the iBGP connection on both N5ks or advertise the WAN next hop addresses into IGP.

I can see in the BGP configuation that an iBGP peer-policy has been configured with next-hop-self but it doesnt appear to be applied.

[Hasrat Raja]

I am new to Nexus, how to apply ? I thought adding in configuration will do it .

thanks

[de1denta]

Hi,

I would just configure next-hop-self for each neighbor as follows

dist-1

neighbor 10.25.38.12 remote-as 64415

    description << iBGP Peering to DIST2 >>

    update-source loopback0

    address-family ipv4 unicast

      next-hop-self

dist-2

neighbor 10.25.38.11 remote-as 64415

    description << iBGP Peering to DIST1 >>

    update-source loopback0

    address-family ipv4 unicast

      next-hop-self

[Hasrat Raja]

This is how it's configured on dist1, I thought I have to do a committ, will schedule a windiow & test if internal routes are reachable.

Thanks again for all your help, will update with result.

[Hasrat Raja]

Hi Will,

I applied next-hop-self to both ibgp neighbors, still same issue. In sh ip bgp default route from dist2 is advertised to dist1 as you mentioned, no change.

thanks