Advertising Routes

Each BGP speaker advertises to its peers the routes to prefixes that it can reach. These routes include:

• Routes to prefixes originating within the speaker’s AS
• Routes redistributed from another protocol, including static routes

By default, BGP does not advertise any route unless the router’s IP routing table also contains the route.

Prefixes Originating in an AS

Use the network command to configure a router with the prefixes that originate within its AS. Thereafter the router advertises these configured prefixes with the origin attribute set to IGP. See Understanding the Origin Attribute for more information about origins. Figure 13 shows a network structure of three autonomous systems, each with a router that originates certain prefixes.

Figure 13: Prefixes Originating in an AS

The following commands configure router NY:

The following commands configure router Boston:

Notice that a mask was not specified for the prefix originating with router Boston. The natural mask is assumed for networks without a mask.

The following commands configure router LA:

Advertising Best Routes

By default, BGP selects from its routing table one best route to each destination. If BGP learned that best route from an internal peer, then the BGP speaker does not advertise a route to that destination to the speaker’s internal peers.

In earlier software releases, the default behavior was for BGP to select two best routes to any destination. The best route learned from external (including confederation) peers was advertised to the speaker’s internal peers. The best route learned from all sources was advertised to the speaker’s external peers.

You can issue the bgp advertise-external-to-internal command to cause BGP to revert to advertising two potentially different routes to its peers. See Selecting the Best Path for information about the process BGP uses to determine best routes.

Redistributing Routes into BGP

BGP can learn about routes from sources other than BGP updates from peers. Routes known to other protocols can be redistributed into BGP. Similarly, routes manually configured on a router—static routes—can be redistributed into BGP. After the routes are redistributed, BGP advertises the routes. When you redistribute routes, BGP sets the origin attribute for the route to Incomplete. See Understanding the Origin Attribute for more information about origins.

The following commands configure three static routes on router Boston and configure router Boston to redistribute the static routes and routes from OSPF into BGP for the network structure shown in Figure 14:

Figure 14: Redistributing Routes into BGP

Redistributing Routes from BGP

If you have redistributed routes from BGP into an IGP, by default only EBGP routes are redistributed. You can issue the bgp redistribute-internal command followed by clearing all BGP sessions to permit the redistribution of IBGP routes in addition to EBGP routes.

Note: This default behavior does not apply to VPN routes. Redistribution of IBGP routes (routes received from an internal BGP peer) in a VRF is always enabled. You do not have to issue this command to enable redistribution of internal BGP routes in a VRF.

Configuring a Default Route

Default routes can provide backup routes if primary connections fail or if the route information for a destination is unknown. A router uses the default route in its IP forwarding table to route traffic toward a destination for which no routing entry exists. The accepted BGP convention is to represent a default route by the network prefix 0.0.0.0/0.

Advertising Default Routes

If you want a router to serve as a default destination for traffic from other routers that do not know where to forward traffic, you can configure the router to advertise a default route. Use the neighbor default-originate command to specify the neighbors to which this router will advertise the default route. Said another way, these neighbors will dynamically learn the default route from the router you configure.

If you issue the neighbor default-originate command, BGP sends the default route to that neighbor regardless of whether the default route exists in the IP forwarding table.

In Figure 15, router NY originates the default route 0.0.0.0/0 to router Albany only. Router Chicago does not receive the default route.

Figure 15: Advertising a Default Route

To configure router NY:

You can also specify a route map to modify the attributes of the default route. If the default route does not match the route map, then the default route is not advertised.

Redistributing Default Routes

By default, the redistribute command does not permit a default route to be redistributed into BGP. You can use the default-information originate command to override this behavior and permit the redistribution of default routes into BGP.

Setting a Static Default Route

You might not want your routers to rely on dynamically learned default routes. Instead, you might prefer to specify a static default route that your routers use to forward traffic when they do not have a routing entry for a destination. Use the ip route command to configure a default route on a router. The static route can point to a network number, an IP address, or a physical interface. You can add a distance value to give preference to a specific static route when multiple entries exist for the same route.

Suppose that in Figure 16, router KC has been configured to advertise a default route to router Chicago:

You prefer that router Chicago send traffic with unknown destinations to router StLouis, so you configure a static default route on router Chicago:

Router StLouis is configured to advertise network 172.25.122.0/23 to router Chicago:

Figure 16: Setting a Static Default Route

Setting the Minimum Interval Between Routing Updates

You can use the neighbor advertisement-interval command to set the minimum interval between the sending of BGP updates. Lower values for the advertisement interval cause route changes to be reported more quickly, but may cause routers to use more bandwidth and processor time.

In the following example, the minimum time between sending BGP routing updates is set to 5 seconds:

Aggregating Routes

Aggregation applies only to routes that are present in the BGP routing table. BGP advertises an aggregate route only if the routing table contains at least one prefix that is more specific than the aggregate. You aggregate IPv4 routes by specifying the aggregate IP address, and IPv6 routes by specifying the aggregate IPv6 prefix.

Figure 17 illustrates an IPv4 network structure where you might use aggregation. The following commands configure router LA and router SanJose so that router SanJose advertises an IPv4 aggregate route, 172.24.0.0/16, for the more specific prefixes 172.24.1.0/24, 172.24.2.0/24, and 172.24.24.0/21.

Figure 17: Configuring Aggregate Addresses

To configure router LA:

To configure router SanJose:

As configured above, router SanJose advertises the more specific routes as well as the aggregate route to router Boston. Alternatively, you can use the summary-only option to configure router SanJose to suppress the more specific routes and advertise only the aggregate route:

Each of these configurations sets the atomic-aggregate attribute in the aggregate route. This attribute informs recipients that the route is an aggregate and must not be deaggregated into more specific routes.

Aggregate routes discard the path information carried in the original routes. To preserve the paths, you must use the as-set option. This option creates an AS-Set that consists of all the AS numbers traversed by the summarized paths. The AS-Set is enclosed within curly brackets; for example, {3, 2}. Each AS number appears only once, even if it appears in more than one of the original paths. If you use the as-set option, the atomic-aggregate attribute is not set for the aggregated route. The following commands configure router SanJose to aggregate the routes while preserving the path information:

If you do not want to aggregate all more specific routes, you can use a route map to limit aggregation. Consider Figure 17 again. Suppose you do not want router SanJose to aggregate prefix 172.24.48.0/20. The following commands show how you can configure a route map on router SanJose to match this prefix, and how to invoke the route map with the advertise-map option:

You can use the attribute-map option to configure attributes for the aggregated route. In Figure 17, suppose that router LA has been configured to set the community attribute for route 172.24.160.0/19 to no-export. This attribute is passed along to router SanJose and preserved when the aggregate route is created. As a result, the aggregate route is not advertised outside the AS. The following commands demonstrate how to configure router SanJose to prevent the aggregate from not being advertised:

Advertising Inactive Routes

Under normal circumstances, routes that are not being used to forward traffic—inactive routes—are not advertised to peers unless synchronization is enabled. For example, suppose a BGP speaker receives a route to a particular prefix, determines that it is the best route to the prefix, and stores the route in the IP routing table (sometimes known as the forwarding information base, or FIB). This route might not be used for forwarding to that prefix; for example, if you have configured a static route to the same destination prefix. Because static routes have better administrative distances than BGP received routes, IP will use the static route rather than the BGP received route for forwarding traffic to that prefix. The BGP received route is inactive and is not advertised to peers. You can use the bgp advertise-inactive command to enable the advertisement of inactive received routes.

Verifying an AS Path

You can use the bgp enforce-first-as command to cause BGP to compare the first AS in the AS-path of a received route with the configured remote AS number of that EBGP peer. If the check fails, BGP returns a notification message to the peer.

Advertising IPv4 Routes Between IPv6 BGP Peers

When an IPv6 network connects two separate IPv4 networks, you can use IPv6 to advertise the IPv4 routes over the BGP session, using TCP IPv6 as the transport mechanism. Similarly, you can advertise IPv6 routes between two IPv4 peers over their BGP session.

Configure the peers by using IPv6 addresses within the IPv4 unicast address family. You can set the IPv4 next hop with a static route or by configuring an inbound or outbound route map. This action overrides the IPv4 next hop that is advertised to the peer for IPv4 routes over BGP IPv6 peers.

If you do not use the route map, then the advertised IPv4 next hop is set to the BGP router ID. That value generally makes the next hop unreachable by the other BGP IPv6 peer.

Advertising Routes Conditionally

By default, a BGP speaker advertises the best routes in its routing table to its peers. However, in some circumstances, you might prefer that some routes be advertised to a peer or peer group only when another route is in the BGP routing table, or only when that route is not in the routing table. BGP conditional advertisement enables you to control route advertisement without having to rely on only the best routes.

For example, in a multi-homed network, you might want to advertise certain prefixes to one of the providers when a failure occurs in the peering session with a different provider, or when there is only partial reachability to that peer.

In other cases, the advertisement to a peer of certain routes might be useful only in the event that some other routes are present in the BGP routing table.

You can use the neighbor advertise-map command with route maps to configure conditional advertisement of BGP routes to a peer or peer group within an address family. BGP conditional advertisement does not create routes. The routes specified by the route map in the neighbor advertise-map command must already be present in the BGP routing table.

BGP conditional advertisement is supported in only the following address families:

• Unicast IPv4
• Unicast IPv6
• Multicast IPv4
• Multicast IPv6
• VPNv4 unicast
• VPNv6 unicast

  Note: For VPNv4 unicast and VPNv6 unicast address families, we recommend that you include a match extcommunity clause to match a route with a specific route target. However, conditional advertisement in these address families can sometimes result in unintended behaviors: advertisement of or based on an incorrect VPN route or a non-VPN route.

BGP conditional advertisement is not supported in the following address families:

• L2VPN
• Route-target
• VPLS
• VPWS

Use the exist-map keyword when you want a route advertised only when another route is present. The determining route must match the specified route map. If the route map you specify with the exist-map keyword references multiple routes, only one of those routes needs to be in the routing table to trigger the conditional advertisement.

Use the non-exist-map keyword when you want a route advertised only when another route is absent. The determining route must match the specified route map. If the route map you specify with the non-exist-map keyword references multiple routes, all of those routes must be absent to trigger the conditional advertisement.

You can optionally specify a sequence number for the advertise route map that matches the determining route. The sequence number specifies the order in which the advertise route maps are processed. It indicates the position the specified advertise route map has in the list of all advertise route maps that are configured for a particular neighbor within the same address family.

If you do not specify a sequence number, the position of the advertise route map is considered to be the sum of the current largest sequence number plus five. An advertise route map with a lower sequence number has a higher priority and is processed before one with a higher sequence number.

If the route matches more than one advertise route map, only the first matching advertise route map, based on the sequence, controls the advertisement of a BGP route.

You can configure a maximum of 50 advertise maps for a given peer or peer group in an address family. However, the name and sequence number for the advertise route map must be unique for each entry. BGP applies any policy specified by the advertise map to the conditionally advertised routes before outbound policy specified for the neighbor is applied.

The route maps referenced by the neighbor advertise-map command must include a match ip-address clause. You can also include additional match clauses. All match commands supported by existing outbound policies are supported. The additional clauses are useful when you want to match only on a specific route with a specific set of attributes. Only the permit keyword is acted on in a match clause. The deny keyword is ignored. Only exact matching of a prefix referenced by exist maps or non-exist maps is supported. Consequently a range specified by the ge or le keyword in the prefix list referenced by these route maps is ignored.

Clauses in a route map that include set commands or the match-set summary prefix-tree command are ignored. To change the attributes of conditionally advertised routes, you must use outbound routing policy.

If the contents of a referenced route map are changed, the new route map takes effect automatically.

Advertising a Route Only When Another Route is Present

You can use the exist-map keyword with the neighbor advertise-map command to advertise a route only when the routing table contains some other particular route.

In the network shown in Figure 18, router 2 (R2) has established BGP sessions with both router 1 (R1) and router 3 (R3). The plan is for router 2 to send router 1 an advertisement for the route to prefix 10.10.20.0/24 only if router 2 has received a route to prefix 172.24.19.0/24 from router 3.

Alternatively, if the route to prefix 172.24.20.0 has been installed in the BGP routing table on router 2, then router 2 advertises to router 1 the route to prefix 10.10.30.0. In this case, the route does not have to be learned from router 3.

Figure 18: Advertising a Route When Another Route is Present

The following commands represent a partial configuration of router R2:

The match as-path clause in the route map referenced by the exist-map keyword ensures that router 2 sends router 1 the route to prefix 10.10.20.0 only if a route to 172.24.19.0/24 with an AS path of 300 is present in the BGP routing table. Similarly, you can impose additional restraints by including any other match clause that is supported by an existing outbound policy.

In this configuration, the condition1 route map has a sequence number of ten. Advertise route maps configured for this peer within the same address family and a lower sequence number are processed before the condition1 route map. The condition2 route map has no sequence number configured, thus giving the route map a sequence number of 15 and ensuring that condition2 is processed after the condition1 route map.

Advertising a Route Only When Another Route is Absent

You can use the non-exist-map keyword with the neighbor advertise-map command to advertise a route only when the BGP routing table does not contain some other particular route.

In the network shown in Figure 19, router R2 has established BGP sessions with both router R1 and router R3. The plan is for router R2 to send peergroup1 an advertisement for the route to prefix 10.10.30.0/24 only if the route to prefix 172.24.20.0/24 is not present in the BGP routing table. Alternatively, if router R2 has not received a route to prefix 172.21.30.0 from router R3, then router R2 advertises to peergroup1 the route to prefix 10.10.20.0. In this sample network, router R3 advertises neither of the routes to router R2. Consequently, router R2 advertises both 10.10.20.0/24 and 10.10.30.0/24 to peergroup1.

Figure 19: Advertising a Route When Another Route is Absent

The following commands configure router R2:

In this configuration, the condition1 route map has a sequence number of five, placing it high in the list of all configured advertise route maps for this peer group within the same address family. The condition2 route map has no sequence number configured, thus placing it at the bottom of the route map list.

In this configuration, the condition1 route map has a sequence number of ten. Route maps configured for this peer group within the same address family and a lower sequence number are processed before the condition1 route map. The condition2 route map has no sequence number configured, thus giving the route map the sequence number of ten and ensuring that condition2 is processed after the condition1 route map.

Advertising a Default Route Only When Another Route Is Present

In some circumstances, you might want to control the advertisement of a default route based on the reachability of an IGP prefix. Because conditional advertisement tracks the BGP routing table rather than the IP routing table, the prefixes that govern the advertisement (the conditional prefixes) must be present in the BGP routing table. In order to use the IGP prefix as a condition, you must import the IGP prefixes into the BGP routing table. You must also configure the origination of the default route.

In the network shown in Figure 20, router R2 has an EBGP session with router R3 and has an IGP session with router R1. Suppose you want to advertise the default route to router R3 based on the reachability of an IGP prefix, 172.55.55.0/24, on router R2.

On router R2, configure a conditional advertisement entry for the neighbor R3. The advertise map must match the default route and the route map referenced by the exist-map keyword must match the imported IGP prefix.

In case router R3 must not learn about the IGP prefix 172.55.55.0/24, you must configure an additional outbound route map to deny this prefix so that it is not advertised to router R3.

With this configuration, the default route is advertised to router R3 only when the IGP prefix 172.55.55.0/24 is reachable on router R2. The default route is withdrawn if this prefix becomes unreachable.

Figure 20: Advertising a Default Route When Another Route is Present

The following commands configure router R2: