Operational BGP Communities

Prefix redistribution control

Route server peers are able to manipulate outbound routing policies via an in-band mechanism using BGP communities,
Currently, we offer the following options:

BGP Community	BGP Extended Community	BGP Large Community	Description
0:peer-as	RT:0:peer-as	63529:0:peer-as	Do not announce a prefix to a certain peer
63529:peer-as	RT:63529:peer-as	63529:1:peer-as	Announce a prefix to a certain peer
0:63529	RT:0:63529	63529:0:0	Do not announce a prefix to any peer
63529:63529	RT:63529:63529	63529:1:0	Announce a prefix to all peers
65003:peer-as	RT:65003:peer-as	63529:103:peer-as	Prepend 3 time to peer-as
65002:peer-as	RT:65002:peer-as	63529:102:peer-as	Prepend 2 time to peer-as
65001:peer-as	RT:65001:peer-as	63529:101:peer-as	Prepend 1 time to peer-as

For destination peers employing a 32-bit ASN, please consider using BGP Extended Communities or BGP Large Communities.

With BGP Large Communities Attribute [RFC8092], Network operators can continue to build upon well-established practises in the tradition of [RFC1997].

The Use of BGP Large Communities is also described in [RFC8195].

Example

{63529:63529} or no community

This route will be advertised to anyone (Default)

{63529:63529 0:63529} or {0:63529}

This route will not be advertised to anyone

{0:65001}

This route will be advertised to everyone except AS 65001

{0:65001 0:65002}

This route will be advertised to everyone except AS 65001 and AS 65002

{0:65003 63529:65003}

This route will be advertised to everyone except AS 65003

{0:63529 63529:65001}

This route will be advertised toward AS 65001 only

{0:63529 63529:65001 63529:65002}

This route will be advertised toward AS 65001 and AS 65002 only

{0:63529 63529:65001 63529:65002 63529:65003 0:65002}

This route will be advertised toward AS 65001 and AS 65003 only

Config guide

ip classless
ip bgp-community new-format
!
router bgp (your AS)
.
.
.
address-family ipv4
  neighbor 203.159.68.68 activate
  neighbor 203.159.68.69 activate
  neighbor BKNIX-RS-v4 send-community
  neighbor BKNIX-RS-v4 next-hop-self
  neighbor BKNIX-RS-v4 route-map BKNIX-out out
  neighbor BKNIX-RS-v4 route-map BKNIX-in in
!
route-map BKNIX-in permit 10				#Accept all prefixes from RS
!
route-map BKNIX-out permit 10				#Do not announce this prefixes to everyone
   match ip address prefix-list filter-out
   set community 0:63529 additive
!
route-map BKNIX-out permit 20
   match community .....				#Announce all route matched these community to everyone
   set community 63529:63529

route-map BKNIX-out permit 30
   match community .....				#Deny route matched these community to ASN131071 (32 bits ASN)
   set extcommunity rt 0:131071

!
ip prefix-list filter-out seq 10 permit x.x.x.x/y
!

Graceful BGP Session Shutdown

BKNIX Route servers accept this well-known BGP Community, RFC 8326 (Graceful BGP Session Shutdown), which indicate our route servers to recalculate and redistribute an alternate path (if available) for your prefixes announcement.
*It is recommended that all RS clients apply GRACEFUL_SHUTDOWN at the inbound direction towards your RS BGP sessions. NLNOG has a good example here.

Example : Operations at Maintenance Time

• Member announces all prefixes with BGP Community GRACEFUL_SHUTDOWN (65535:0) to route servers.
• Route servers will set BGP local_preference of these prefixes to 0.
• Route servers will recalculate the alternative path of these prefixes.
• if prefixes have an alternative path, new advertised prefixes will be redistributed to clients.
• if prefixes do not have an alternative path, the same prefixes remain advertised with the BGP Community GRACEFUL_SHUTDOWN.
• Other RS clients have learned alternate paths and forward traffic through the new paths. At this point, a member can safely shutdown their BGP session.

---