no BGP default ipv4-unicast

[Ibrahim Jamil]

Hi Guys

when we deal in MPLS MP-BGP, why we need to disable the "BGP default ipv4-unicast" behavior .with the command

"no BGP default ipv4-unicast"

what if we kept it on

thanks 

[Giuseppe Larosa · Accepted solution]

Hello Ibrahim,

the need for the command

"no bgp default ipv4-unicast"

was sure years ago.

The meaning is that you are disabling the default behaviuor of BGPv4 to advertise only IPv4 unicast routes.

The command is an enabler for Multi protocol BGP mode where multiple address families can be negotiated during the BGP session setup when the two peers exchange the respective capabilities.

Related to this command is the need to activate the neighbor under each address family to say to the router " I want you to talk also this family with this peer"

The need for this command

"no bgp default ipv4-unicast"

may have been removed in recent IOS images by reverting the default BGP behaviuor to be Multi protocol.

I think I have seen configuration examples of this here in the forums.

However, this can be platform specific so it may be better to use the command for safety if you want to run BGP in multi protocol mode (BGP4+).

Hope to help

Giuseppe

[Peter Paluch · Accepted solution]

Ibrahim, Giuseppe,

To my best knowledge, the purpose of the "no bgp default ipv4-unicast" command is slightly different.

By default, whenever a new neighbor statement is added to BGP configuration pointing to the IPv4 address of the neighbor, the router automatically and implicitly assumes that we want to exchange IPv4 routes with this neighbor. This might be true, but with today's use of BGP to advertise so many diverse kinds of control information (VPNv4 routes, VPNv6 routes, autodiscovery for VPLS endpoints, signalling for multicast VPNs, FlowSpec information, ...), it would be a premature conclusion to assume that just because a neighbor is defined using its IPv4 address, we want to exchange IPv4 routes with it.

This is where "no bgp default ipv4-unicast" command comes in. When entered, BGP will no longer assume anything about the neighbor, and instead, it will expect us to tell it exactly what kind of content should be exchanged with this neighbor. This is done by using the "address-family" sections inside BGP configuration where this neighbor needs to be explicitly activated. In fact, without this command, every IPv4 neighbor is automatically added to the "address-family ipv4 unicast" section; with the "no bgp default ipv4-unicast" command entered, we have to do this manually.

Please note that either the presence or impresence of the "no bgp default ipv4-unicast" command has no impact on whether we can exchange different address families with the neighbor; in other words, it does not activate or deactivate the multi-protocol BGP capability. It only affects the default behavior of adding IPv4 neighbors to the IPv4-unicast address family automatically.

Years ago, I have answered another thread that was very relevant to this topic - you might be interested in reading it, too:

https://supportforums.cisco.com/discussion/11476526/when-use-bgp-address-family

Best regards,
Peter

[Ibrahim Jamil]

thanks all

[Philip D'Ath · Accepted solution]

This is a bit outside my area, but I think if you don't do this BGP defaults to the single address family of "ipv4".  When you use the above command it allows the exchange of additional address families.

You use this when configuring IPv6 BGP as well.