How to advertise learned iBGP routes to eBGP peer?

[Laurent Ost]

Hi,

I'm a student and I'm currently building a full IPv6 lab with a provider backbone in AS 65000 with XR routers and a customer AS 64800 with a CE router and want to redistribute the learned iBGP routes from the backbone to the customer CE peer. I have configured IS-IS for router connectivity in the core over each router's loopback 0. For the XR3 to XE4 I've used a static route.

Afterwards, iBGP was configured to advertise the loopback 1 from XR2 (simulating a network segment) to all iBGP peers which worked. XR1 was used as route reflector to omit the full mesh requirement.

XR3 and XE4 have successfully established eBGP peering but I have currently two problems:

- CE does not receive the BGP routes of XR2 Lo1 through XR3 and XR3 does not receive the Loopback 1 created on XE4.
- I cannot ping loopback 0 (XR3 and XE4) of each other even though BGP peering is successfully established.

Following BGP configuration on XR3:

router static
  address-family ipv6 unicast
    2014::1/128 GigabitEthernet0/0/0/1
    2014::1/128 2001:db8:34::3
  !
!

router bgp 65000
  bgp router-id 3.3.3.3
  address-family ipv6 unicast
    network 2013::1/128
    network 2001:db8:34::/64
    redistribute connected
    redistribute isis 97
  !
  neighbor 2011::1
    remote-as 65000
    update-source Loopback0
    address-family ipv6 unicast
      next-hop-self
     !
   !
  neighbor 2014::1
    remote-as 64800
    ebgp-multihop 2
    update-source Loopback0
    address-family ipv6 unicast
      next-hop-self
     !
   !
!

-------------------------------

On XE4

router bgp 64800
  bgp router-id 4.4.4.4
  bgp log-neighbor-changes
  no bgp default ipv4-unicast
  neighbor 2013::1 remote-as 65000
  neighbor 2013::1 ebgp-multihop 2
  neighbor 2013::1 update-source Loopback0
  !
  address-family ipv4
  exit-address-family
  !
  address-family ipv6
    network 2001:DB8:34::/64
    network 2014::1/128
    network 2114::1/128
    neighbor 2013::1 activate
  exit-address-family
!

ipv6 route 2013::1/128 2001:DB8:34::3

------ 

XR3 has XR2's route.

show route ipv6

....

B 2112::1/128
[200/0] via 2012::1, 2w6d

Do I have to configure some kind of redistribution for eBGP or is it automatically done and I haven't configured something correctly?

Thanks in advance.

[MHM Cisco World]

XR3

router bgp 65000
  neighbor 2014::1
    remote-as 64800
    ebgp-multihop 2 <<-
    update-source Loopback0 <<-
    address-family ipv6 unicast
      next-hop-self

between XR3 and XR4 you use loopback as update source but are you config any static or routing protocol for these Loopback??

try use direct connect instead of update-source in all router and check, I think you missing config IGP reach all LO inside AS and between AS.

[Laurent Ost]

Thanks for your response.

IS-IS is used as IGP inside the core AS and static routing between the AS. I removed update-source loopback0 and multihop 2 on XR3 and XE4 but it broke the BGP peering. IGP is working fine inside the core, but maybe the static route was not a good idea.

In another lab I configured a similar setup with IPv4, OSPF as IGP and static route between AS (only XE routers) which also worked.

I'll try to configure a L1/L2 IS-IS link between XR3 and XE4. If that doesn't work I'm will configure them all as directly connected instead of going over loopbacks. It's interesting that the peering is established.

[Harold Ritter]

Hi @Laurent Ost ,

The static routes on XR3 are incorrect. Try the following:

router static

  address-family ipv6 unicast

    no 2014::1/128 GigabitEthernet0/0/0/1

    no 2014::1/128 2001:db8:34::3

    2014::1/128 2001:db8:34::4 

Regards,

[Laurent Ost]

Hi @Harold Ritter,

Thanks, I realized 

2001:db8:34::3 

was wrong but the removal of 

2014::1/128 GigabitEthernet0/0/0/1

made the ping  finally go through from Loopback0 to Loopback0. Thanks a lot.

However, I still receive no prefixes over the eBGP peer when running:

show bgp ipv6 unicast neighbors 2013::1 routes

Is there some redistribution needed? I would like the advertise the iBGP received route 2112::1/128 of XR2 from XR3 to XE4. 

Regards,

Laurent

[Harold Ritter · Accepted solution]

Hi @Laurent Ost ,

With IOS-XR, eBGP drops all inbound and outbound routes by default. You need to configure an inbound and outbound

route-policy

permitting the routes or configure

bgp unsafe-ebgp-policy

to disable the default behaviour.

route-policy pass

pass

end-policy

router bgp 65000

neighbor 2014::1

address-family ipv6 unicast

route-policy pass in

route-policy pass out

or

router bgp 65000

bgp unsafe-ebgp-policy

Regards,

[Laurent Ost]

Hi @Harold Ritter,

Awesome, it works! I didn't know about this default behaviour.

Thank you very much.

Regards,

Laurent