Difference between address-family ipv6 and address-family ipv6 labeled unicast

[mukundh86]

Hello Experts,

Can someone explain me the difference between address-family ipv6 and address-family ipv6 labeled unicast. Per my understanding, i think both of them are used to send labelled IPv6 prefix advertisements through BGP..If so, are the following configs same?

 address-family ipv6

 neighbor 192.168.0.1 activate

 neighbor 192.168.0.1 send-label

router bgp 10

neighbor 192.168.0.1

address-family ipv6 labelled unicast

Please let me know if my understanding is correct

Thanks

Mukundh

[Nagendra Kumar Nainar · Accepted solution]

Hi Mukund,

They both are same feature but in different image. "neighbor <> send-label" is the command used to enable BGP to assign local label and advertise to neighbors in IOS. "address-family ipv6 labelled unicast" is used to enable the same in XR.

-Nagendra

[mukundh86]

Thanks for the reply Nagendra...

I have another related query regarding this. We have BGP neighborship flapping between 2 routers ...One is Cisco 7204 and another is Juniper M120 I think.... The Juniper logs show that BGP is flapped due to family inet6 not configured on the Juniper end and Juniper is receiving BGP advertisements with native IPv6 as next hop from Cisco when it shouldn't be receiving that.. The following are commands on Cisco and Juniper...

##### CISCO####

outer bgp 5603

neighbor 95.176.254.10 inherit peer-session LAR  neighbor 95.176.254.10 description --- M320-LAB-LJ-CIGALETOVA  address-family ipv4

  neighbor 95.176.254.10 activate

  neighbor 95.176.254.10 inherit peer-policy LAR-ipv4  address-family ipv6

  neighbor 95.176.254.10 activate

  neighbor 95.176.254.10 send-community both

  neighbor 95.176.254.10 route-reflector-client

  neighbor 95.176.254.10 send-label

template peer-session LAR

  remote-as 5603

  update-source Loopback0

  timers 30 90

exit-peer-session

template peer-policy LAR-ipv4

  route-map LAR-ipv4-out out

  route-reflector-client

  soft-reconfiguration inbound

  send-community both

exit-peer-policy

####JUNIPER####

protocols{bgp{

group I-BGP-IPV4 {

            type internal;

            family inet {

                unicast;

            }

            family inet6 {

                labeled-unicast {

                    explicit-null;

                }

            }

            export RR-Export-All;

            neighbor 95.176.255.254 {

                description C7201-RR-IP-CIGALETOVA;

                local-address 95.176.254.10;

            }

            neighbor 95.176.255.252 {

                description C7201-RR-IP-CIGALETOVA;

                local-address 95.176.254.10;

            }

        }

By the cisco command above, shouldn't cisco be sending only labelled ipv6 prefixes or am I wrong in this. And if Cisco sends both unlabelled and labelled prefixes, is there a way to make it send only ipv6 prefixes?

Thanks

Mukundh

[Harold Ritter]

Hi Mukundh,

Since "address-family ipv6 unicast" handles both labeleld and unlabelled prefixes in IOS, you may need to have an outbound route-map with "set mpls-label" for the 6PE prefixes.

*** Update ***

Given that the Juniper side does not negotiate unlabeleldipv6 unicast, the cisco side should not have this specific AF negotiated either and should therefore not send unlabelled ipv6 prefixes across. In some IOS versions, unlabelled ipv6 unicast prefixes are sent across even though the AF has not been negotiated. This behavior is documented by CSCtf27303. "set mpls-label" is the work around for this issue.

What version of IOS do you use in your lab?

Regards

[mukundh86]

Hi Harold,

Thanks a lot for your reply..Let me know if  my understanding is correct:

1. Cisco sends both unlabelled and labelled ipv6 prefixes to juniper after bgp is established

2. Juniper drops BGP sessiona fter seeing NLRI with native IPv6

3. Cisco needs to have route-map policy which matches all ip v6 prefixes with next-hop as native ipv6 and use "set -mpls-label" command to send them out with labels so that Juniper does not drop BGP.

Also, can you provide me with a sample config of the route-map and if can match conditions beased on next hop?

I need to get back to customer for the IOS version..As of now I have control only over Juniper box.

Thanks

Mukundh

[Harold Ritter]

Hi Mukundh,

Your understanding is correct. In this particular case, you need to set mpls-label on all prefixes since the Juniper side is not even configured for unlabelled ipv6 AF. The route-map would be pretty straightforward:

route-map SendLabel permit 10

set mpls-label

Regards