BGP uses TCP as its transport protocol. Two BGP routers form a TCP connection between one another (peer routers) and exchange messages to open and confirm the connection parameters.

BGP routers exchange network reachability information. This information is mainly an indication of the full paths (BGP autonomous system numbers) that a route should take to reach the destination network. This information helps construct a graph that shows which autonomous systems are loop free and where routing policies can be applied to enforce restrictions on routing behavior.

Any two routers forming a TCP connection to exchange BGP routing information are called peers or neighbors. BGP peers initially exchange their full BGP routing tables. After this exchange, incremental updates are sent as the routing table changes. BGP keeps a version number of the BGP table, which is the same for all of its BGP peers. The version number changes whenever BGP updates the table due to routing information changes. Keepalive packets are sent to ensure that the connection is alive between the BGP peers and notification packets are sent in response to error or special conditions.

Note	Other than enabling RTC (route target constraint) with address-family ipv4 rtfilter command, there is no separate configuration needed to enable RTC for BGP EVPN.

Note

For information on configuring BGP to distribute Multiprotocol Label Switching (MPLS) Layer 3 virtual private network (VPN) information, see the Cisco IOS XR Multiprotocol Label Switching Configuration Guide for the Cisco CRS-1 Router. For information on BGP support for Bidirectional Forwarding Detection (BFD), see the Cisco IOS XR Interface and Hardware Configuration Guide for the Cisco CRS-1 Router and the Cisco IOS XR Interface and Hardware Command Reference for the Cisco CRS-1 Router.

For BGP sessions between neighbors to be established, BGP must be assigned a router ID. The router ID is sent to BGP peers in the OPEN message when a BGP session is established.

BGP attempts to obtain a router ID in the following ways (in order of preference):

• By means of the address configured using the bgp router-id command in router configuration mode.

• By using the highest IPv4 address on a loopback interface in the system if the router is booted with saved loopback address configuration.

• By using the primary IPv4 address of the first loopback address that gets configured if there are not any in the saved configuration.

If none of these methods for obtaining a router ID succeeds, BGP does not have a router ID and cannot establish any peering sessions with BGP neighbors. In such an instance, an error message is entered in the system log, and the show bgp summary command displays a router ID of 0.0.0.0.

After BGP has obtained a router ID, it continues to use it even if a better router ID becomes available. This usage avoids unnecessary flapping for all BGP sessions. However, if the router ID currently in use becomes invalid (because the interface goes down or its configuration is changed), BGP selects a new router ID (using the rules described) and all established peering sessions are reset.

Note	We strongly recommend that the bgp router-id command is configured to prevent unnecessary changes to the router ID (and consequent flapping of BGP sessions).

IOS XR BGP maximum-prefix feature imposes a maximum limit on the number of prefixes that are received from a neighbor for a given address family. Whenever the number of prefixes received exceeds the maximum number configured, the BGP session is terminated, which is the default behavior, after sending a cease notification to the neighbor. The session is down until a manual clear is performed by the user. The session can be resumed by using the clear bgp command. It is possible to configure a period after which the session can be automatically brought up by using the maximum-prefix command with the restart keyword. The maximum prefix limit can be configured by the user. Default limits are used if the user does not configure the maximum number of prefixes for the address family. For default limits, refer to BGP Default Limits.

Discard Extra Paths

An option to discard extra paths is added to the maximum-prefix configuration. Configuring the discard extra paths option drops all excess prefixes received from the neighbor when the prefixes exceed the configured maximum value. This drop does not, however, result in session flap.

The benefits of discard extra paths option are:

• Limits the memory footstamp of BGP.

• Stops the flapping of the peer if the paths exceed the set limit.

When the discard extra paths configuration is removed, BGP sends a route-refresh message to the neighbor if it supports the refresh capability; otherwise the session is flapped.

On the same lines, the following describes the actions when the maximum prefix value is changed:

• If the maximum value alone is changed, a route-refresh message is sourced, if applicable.

• If the new maximum value is greater than the current prefix count state, the new prefix states are saved.

• If the new maximum value is less than the current prefix count state, then some existing prefixes are deleted to match the new configured state value.

There is currently no way to control which prefixes are deleted.

For detailed configuration steps, see Configuring Discard Extra Paths.

Cisco IOS XR BGP imposes maximum limits on the number of neighbors that can be configured on the router and on the maximum number of prefixes that are accepted from a peer for a given address family. This limitation safeguards the router from resource depletion caused by misconfiguration, either locally or on the remote neighbor. The following limits apply to BGP configurations:

• The default maximum number of peers that can be configured is 4000. The default can be changed using the bgp maximum neighbor command. The limit range is 1 to 15000. Any attempt to configure additional peers beyond the maximum limit or set the maximum limit to a number that is less than the number of peers currently configured will fail.

• To prevent a peer from flooding BGP with advertisements, a limit is placed on the number of prefixes that are accepted from a peer for each supported address family. The default limits can be overridden through configuration of the maximum-prefix limit command for the peer for the appropriate address family. The following default limits are used if the user does not configure the maximum number of prefixes for the address family:

  • IPv4 Unicast: 1048576

  • IPv4 Labeled-unicast: 131072

  • IPv4 Tunnel: 1048576

  • IPv6 Unicast: 524288

  • IPv6 Labeled-unicast: 131072

  • IPv4 Multicast: 131072

  • IPv6 Multicast: 131072

  • IPv4 MVPN: 2097152

  • VPNv4 Unicast: 2097152

  • IPv4 MDT: 131072

  • VPNv6 Unicast: 1048576

  • L2VPN EVPN: 2097152

  A cease notification message is sent to the neighbor and the peering with the neighbor is terminated when the number of prefixes received from the peer for a given address family exceeds the maximum limit (either set by default or configured by the user) for that address family.

  It is possible that the maximum number of prefixes for a neighbor for a given address family has been configured after the peering with the neighbor has been established and a certain number of prefixes have already been received from the neighbor for that address family. A cease notification message is sent to the neighbor and peering with the neighbor is terminated immediately after the configuration if the configured maximum number of prefixes is fewer than the number of prefixes that have already been received from the neighbor for the address family.

BGP receives notifications from the Routing Information Base (RIB) when next-hop information changes (event-driven notifications). BGP obtains next-hop information from the RIB to:

• Determine whether a next hop is reachable.

• Find the fully recursed IGP metric to the next hop (used in the best-path calculation).

• Validate the received next hops.

• Calculate the outgoing next hops.

• Verify the reachability and connectedness of neighbors.

BGP is notified when any of the following events occurs:

• Next hop becomes unreachable

• Next hop becomes reachable

• Fully recursed IGP metric to the next hop changes

• First hop IP address or first hop interface change

• Next hop becomes connected

• Next hop becomes unconnected

• Next hop becomes a local address

• Next hop becomes a nonlocal address

Note	Reachability and recursed metric events trigger a best-path recalculation.

Event notifications from the RIB are classified as critical and noncritical. Notifications for critical and noncritical events are sent in separate batches. However, a noncritical event is sent along with the critical events if the noncritical event is pending and there is a request to read the critical events.

• Critical events are related to the reachability (reachable and unreachable), connectivity (connected and unconnected), and locality (local and nonlocal) of the next hops. Notifications for these events are not delayed.

• Noncritical events include only the IGP metric changes. These events are sent at an interval of 3 seconds. A metric change event is batched and sent 3 seconds after the last one was sent.

The next-hop trigger delay for critical and noncritical events can be configured to specify a minimum batching interval for critical and noncritical events using the nexthop trigger-delay command. The trigger delay is address family dependent.

The BGP next-hop tracking feature allows you to specify that BGP routes are resolved using only next hops whose routes have the following characteristics:

• To avoid the aggregate routes, the prefix length must be greater than a specified value.

• The source protocol must be from a selected list, ensuring that BGP routes are not used to resolve next hops that could lead to oscillation.

This route policy filtering is possible because RIB identifies the source protocol of route that resolved a next hop as well as the mask length associated with the route. The nexthop route-policy command is used to specify the route-policy.

For information on route policy filtering for next hops using the next-hop attach point, see the Implementing Routing Policy Language on Cisco IOS XR Software module of Cisco IOS XR Routing Configuration Guide (this publication).

Autonomous system numbers (ASNs) are globally unique identifiers used to identify autonomous systems (ASs) and enable ASs to exchange exterior routing information between neighboring ASs. A unique ASN is allocated to each AS for use in BGP routing. ASNs are encoded as 2-byte numbers and 4-byte numbers in BGP.

RP/0/RP0/CPU0:router(config)# as-format [asdot | asplain]
RP/0/RP0/CPU0:router(config)# as-format asdot

BGP in Cisco IOS XR software follows a neighbor-based configuration model that requires that all configurations for a particular neighbor be grouped in one place under the neighbor configuration. Peer groups are not supported for either sharing configuration between neighbors or for sharing update messages. The concept of peer group has been replaced by a set of configuration groups to be used as templates in BGP configuration and automatically generated update groups to share update messages between neighbors.

BGP does not support the concept of a default address family. An address family must be explicitly configured under the BGP router configuration for the address family to be activated in BGP. Similarly, an address family must be explicitly configured under a neighbor for the BGP session to be activated under that address family. It is not required to have any address family configured under the BGP router configuration level for a neighbor to be configured. However, it is a requirement to have an address family configured at the BGP router configuration level for the address family to be configured under a neighbor.

For default VRF, starting from Cisco IOS XR Software Release 6.2.x, both IPv4 Unicast and IPv4 Labeled-unicast address families are supported under the same neighbor.

For non-default VRF, both IPv4 Unicast and IPv4 Labeled-unicast address families are not supported under the same neighbor. However, the configuration is accepted on the Router with the following error:

bgp[1051]: %ROUTING-BGP-4-INCOMPATIBLE_AFI : IPv4 Unicast and IPv4 Labeled-unicast Address families together are not supported under the same neighbor.

When one BGP session has both IPv4 unicast and IPv4 labeled-unicast AFI/SAF, then the routing behavior is nondeterministic. Therefore, the prefixes may not be correctly advertised. Incorrect prefix advertisement results in reachability issues. In order to avoid such reachability issues, you must explicitly configure a route policy to advertise prefixes either through IPv4 unicast or through IPv4 labeled-unicast address families.

External BGP (eBGP) neighbors must have an inbound and outbound policy configured. If no policy is configured, no routes are accepted from the neighbor, nor are any routes advertised to it. This added security measure ensures that routes cannot accidentally be accepted or advertised in the case of a configuration omission error.

Note	This enforcement affects only eBGP neighbors (neighbors in a different autonomous system than this router). For internal BGP (iBGP) neighbors (neighbors in the same autonomous system), all routes are accepted or advertised if there is no policy.

In the following example, for an eBGP neighbor, if all routes should be accepted and advertised with no modifications, a simple pass-all policy is configured:

  RP/0/RP0/CPU0:router(config)# route-policy pass-all
  RP/0/RP0/CPU0:router(config-rpl)# pass
  RP/0/RP0/CPU0:router(config-rpl)# end-policy
  RP/0/RP0/CPU0:router(config)# commit
  

Use the route-policy (BGP) command in the neighbor address-family configuration mode to apply the pass-all policy to a neighbor. The following example shows how to allow all IPv4 unicast routes to be received from neighbor 192.168.40.42 and advertise all IPv4 unicast routes back to it:

  RP/0/RP0/CPU0:router(config)# router bgp 1
  RP/0/RP0/CPU0:router(config-bgp)# neighbor 192.168.40.24
  RP/0/RP0/CPU0:router(config-bgp-nbr)# remote-as 21
  RP/0/RP0/CPU0:router(config-bgp-nbr)# address-family ipv4 unicast
  RP/0/RP0/CPU0:router(config-bgp-nbr-af)# route-policy pass-all in
  RP/0/RP0/CPU0:router(config-bgp-nbr-af)# route-policy pass-all out
  RP/0/RP0/CPU0:router(config-bgp-nbr-af)# commit
 

Use the show bgp summary command to display eBGP neighbors that do not have both an inbound and outbound policy for every active address family. In the following example, such eBGP neighbors are indicated in the output with an exclamation (!) mark:

  RP/0/RP0/CPU0:router# show bgp all all summary
  
  Address Family: IPv4 Unicast
  ============================
  
  BGP router identifier 10.0.0.1, local AS number 1
  BGP generic scan interval 60 secs
  BGP main routing table version 41
  BGP scan interval 60 secs
  BGP is operating in STANDALONE mode.
  
  Process         RecvTblVer    bRIB/RIB  SendTblVer
  Speaker                 41          41          41
  
  Neighbor        Spk    AS MsgRcvd MsgSent   TblVer  InQ OutQ Up/Down  St/PfxRcd
  10.0.101.1        0     1     919     925       41    0    0 15:15:08       10
  10.0.101.2        0     2       0       0        0    0    0 00:00:00 Idle
  

  Address Family: IPv4 Multicast
  ==============================
  
  BGP router identifier 10.0.0.1, local AS number 1
  BGP generic scan interval 60 secs
  BGP main routing table version 1
  BGP scan interval 60 secs
  BGP is operating in STANDALONE mode.
  
  Process         RecvTblVer    bRIB/RIB  SendTblVer
  Speaker                  1           1           1
  
  Some configured eBGP neighbors do not have both inbound and
  outbound policies configured for IPv4 Multicast address family.
  These neighbors will default to sending and/or receiving no
  routes and are marked with ’!’ in the output below. Use the
  ’show bgp neighbor <nbr_address>’ command for details.
  
  Neighbor        Spk    AS MsgRcvd MsgSent   TblVer  InQ OutQ Up/Down  St/PfxRcd
  10.0.101.2        0     2       0       0        0    0    0 00:00:00 Idle!
  
  
  Address Family: IPv6 Unicast
  ============================
  
  BGP router identifier 10.0.0.1, local AS number 1
  BGP generic scan interval 60 secs
  BGP main routing table version 2
  BGP scan interval 60 secs
  BGP is operating in STANDALONE mode.
  
  Process         RecvTblVer    bRIB/RIB  SendTblVer
  Speaker                  2           2           2
  
  Neighbor        Spk    AS MsgRcvd MsgSent   TblVer  InQ OutQ Up/Down  St/PfxRcd
  2222::2           0     2     920     918        2    0    0 15:15:11        1
  2222::4           0     3       0       0        0    0    0 00:00:00 Idle
  
  
  Address Family: IPv6 Multicast
  ==============================
  
  BGP router identifier 10.0.0.1, local AS number 1
  BGP generic scan interval 60 secs
  BGP main routing table version 1
  BGP scan interval 60 secs
  BGP is operating in STANDALONE mode.
  
  Process         RecvTblVer    bRIB/RIB  SendTblVer
  Speaker                  1           1           1
  
  Some configured eBGP neighbors do not have both inbound and
  outbound policies configured for IPv6 Multicast address family.
  These neighbors will default to sending and/or receiving no
  routes and are marked with ’!’ in the output below. Use the
  ’show bgp neighbor <nbr_address>’ command for details.
  
  Neighbor        Spk    AS MsgRcvd MsgSent   TblVer  InQ OutQ Up/Down  St/PfxRcd
  2222::2           0     2     920     918        0    0    0 15:15:11        0
  2222::4           0     3       0       0        0    0    0 00:00:00 Idle!  

  

The table policy feature in BGP allows you to configure traffic index values on routes as they are installed in the global routing table. This feature is enabled using the table-policy command and supports the BGP policy accounting feature.

BGP policy accounting uses traffic indices that are set on BGP routes to track various counters. See the Implementing Routing Policy on Cisco IOS XR Software module in the Routing Configuration Guide for Cisco CRS Routers for details on table policy use. See the Cisco Express Forwarding Commands on Cisco IOS XR Software module in the IP Addresses and Services Command Reference for Cisco CRS Routers for details on BGP policy accounting.

Table policy also provides the ability to drop routes from the RIB based on match criteria. This feature can be useful in certain applications and should be used with caution as it can easily create a routing ‘black hole’ where BGP advertises routes to neighbors that BGP does not install in its global routing table and forwarding table.

The BGP Update Groups feature contains an algorithm that dynamically calculates and optimizes update groups of neighbors that share outbound policies and can share the update messages. The BGP Update Groups feature separates update group replication from peer group configuration, improving convergence time and flexibility of neighbor configuration.

To use this feature, you must understand the following concepts:

The BGP cost community is a nontransitive extended community attribute that is passed to internal BGP (iBGP) and confederation peers but not to external BGP (eBGP) peers. The cost community feature allows you to customize the local route preference and influence the best-path selection process by assigning cost values to specific routes. The extended community format defines generic points of insertion (POI) that influence the best-path decision at different points in the best-path algorithm.

The cost community attribute is applied to internal routes by configuring the set extcommunity cost command in a route policy. See the Routing Policy Language Commands on Cisco IOS XR Software module of Cisco IOS XR Routing Command Reference for information on the set extcommunity cost command. The cost community set clause is configured with a cost community ID number (0–255) and cost community number (0–4294967295). The cost community number determines the preference for the path. The path with the lowest cost community number is preferred. Paths that are not specifically configured with the cost community number are assigned a default cost community number of 2147483647 (the midpoint between 0 and 4294967295) and evaluated by the best-path selection process accordingly. When two paths have been configured with the same cost community number, the path selection process prefers the path with the lowest cost community ID. The cost-extended community attribute is propagated to iBGP peers when extended community exchange is enabled.

The following commands include the route-policy keyword, which you can use to apply a route policy that is configured with the cost community set clause:

• aggregate-address

• redistribute

• network

BGP routers typically receive multiple paths to the same destination. The BGP best-path algorithm determines the best path to install in the IP routing table and to use for forwarding traffic. This section describes the Cisco IOS XR software implementation of BGP best-path algorithm, as specified in Section 9.1 of the Internet Engineering Task Force (IETF) Network Working Group draft-ietf-idr-bgp4-24.txt document.

The BGP best-path algorithm implementation is in three parts:

• Part 1—Compares two paths to determine which is better.

• Part 2—Iterates over all paths and determines which order to compare the paths to select the overall best path.

• Part 3—Determines whether the old and new best paths differ enough so that the new best path should be used.

Note

The order of comparison determined by Part 2 is important because the comparison operation is not transitive; that is, if three paths, A, B, and C exist, such that when A and B are compared, A is better, and when B and C are compared, B is better, it is not necessarily the case that when A and C are compared, A is better. This nontransitivity arises because the multi exit discriminator (MED) is compared only among paths from the same neighboring autonomous system (AS) and not among all paths.

An administrative distance is a rating of the trustworthiness of a routing information source. In general, the higher the value, the lower the trust rating. For information on specifying the administrative distance for BGP, see the BGP Commands module of the Routing Command Reference for Cisco CRS Routers

Normally, a route can be learned through more than one protocol. Administrative distance is used to discriminate between routes learned from more than one protocol. The route with the lowest administrative distance is installed in the IP routing table. By default, BGP uses the administrative distances shown in Table 1.

Note	Distance does not influence the BGP path selection algorithm, but it does influence whether BGP-learned routes are installed in the IP routing table.

In most cases, when a route is learned through eBGP, it is installed in the IP routing table because of its distance (20). Sometimes, however, two ASs have an IGP-learned back-door route and an eBGP-learned route. Their policy might be to use the IGP-learned path as the preferred path and to use the eBGP-learned path when the IGP path is down. See Figure 1.

In Figure 1, Routers A and C and Routers B and C are running eBGP. Routers A and B are running an IGP (such as Routing Information Protocol [RIP], Interior Gateway Routing Protocol [IGRP], Enhanced IGRP, or Open Shortest Path First [OSPF]). The default distances for RIP, IGRP, Enhanced IGRP, and OSPF are 120, 100, 90, and 110, respectively. All these distances are higher than the default distance of eBGP, which is 20. Usually, the route with the lowest distance is preferred.

Router A receives updates about 160.10.0.0 from two routing protocols: eBGP and IGP. Because the default distance for eBGP is lower than the default distance of the IGP, Router A chooses the eBGP-learned route from Router C. If you want Router A to learn about 160.10.0.0 from Router B (IGP), establish a BGP back door. See .

In the following example, a network back-door is configured:

RP/0/RP0/CPU0:router(config)# router bgp 100
RP/0/RP0/CPU0:router(config-bgp)# address-family ipv4 unicast
RP/0/RP0/CPU0:router(config-bgp-af)# network 160.10.0.0/16 backdoor

Router A treats the eBGP-learned route as local and installs it in the IP routing table with a distance of 200. The network is also learned through Enhanced IGRP (with a distance of 90), so the Enhanced IGRP route is successfully installed in the IP routing table and is used to forward traffic. If the Enhanced IGRP-learned route goes down, the eBGP-learned route is installed in the IP routing table and is used to forward traffic.

Although BGP treats network 160.10.0.0 as a local entry, it does not advertise network 160.10.0.0 as it normally would advertise a local entry.

Multiprotocol BGP is an enhanced BGP that carries routing information for multiple network layer protocols and IP multicast routes. BGP carries two sets of routes, one set for unicast routing and one set for multicast routing. The routes associated with multicast routing are used by the Protocol Independent Multicast (PIM) feature to build data distribution trees.

Multiprotocol BGP is useful when you want a link dedicated to multicast traffic, perhaps to limit which resources are used for which traffic. Multiprotocol BGP allows you to have a unicast routing topology different from a multicast routing topology providing more control over your network and resources.

In BGP, the only way to perform interdomain multicast routing was to use the BGP infrastructure that was in place for unicast routing. Perhaps you want all multicast traffic exchanged at one network access point (NAP). If those routers were not multicast capable, or there were differing policies for which you wanted multicast traffic to flow, multicast routing could not be supported without multiprotocol BGP.

Note	It is possible to configure BGP peers that exchange both unicast and multicast network layer reachability information (NLRI), but you cannot connect multiprotocol BGP clouds with a BGP cloud. That is, you cannot redistribute multiprotocol BGP routes into BGP.

Figure 1 illustrates simple unicast and multicast topologies that are incongruent, and therefore are not possible without multiprotocol BGP.

Autonomous systems 100, 200, and 300 are each connected to two NAPs that are FDDI rings. One is used for unicast peering (and therefore the exchange of unicast traffic). The Multicast Friendly Interconnect (MFI) ring is used for multicast peering (and therefore the exchange of multicast traffic). Each router is unicast and multicast capable.

Figure 2 is a topology of unicast-only routers and multicast-only routers. The two routers on the left are unicast-only routers (that is, they do not support or are not configured to perform multicast routing). The two routers on the right are multicast-only routers. Routers A and B support both unicast and multicast routing. The unicast-only and multicast-only routers are connected to a single NAP.

In Figure 2, only unicast traffic can travel from Router A to the unicast routers to Router B and back. Multicast traffic could not flow on that path, so another routing table is required. Multicast traffic uses the path from Router A to the multicast routers to Router B and back.

Figure 2 illustrates a multiprotocol BGP environment with a separate unicast route and multicast route from Router A to Router B. Multiprotocol BGP allows these routes to be incongruent. Both of the autonomous systems must be configured for internal multiprotocol BGP (IMBGP) in the figure.

A multicast routing protocol, such as PIM, uses the multicast BGP database to perform Reverse Path Forwarding (RPF) lookups for multicast-capable sources. Thus, packets can be sent and accepted on the multicast topology but not on the unicast topology.

Route dampening is a BGP feature that minimizes the propagation of flapping routes across an internetwork. A route is considered to be flapping when it is repeatedly available, then unavailable, then available, then unavailable, and so on.

For example, consider a network with three BGP autonomous systems: autonomous system 1, autonomous system 2, and autonomous system 3. Suppose the route to network A in autonomous system 1 flaps (it becomes unavailable). Under circumstances without route dampening, the eBGP neighbor of autonomous system 1 to autonomous system 2 sends a withdraw message to autonomous system 2. The border router in autonomous system 2, in turn, propagates the withdrawal message to autonomous system 3. When the route to network A reappears, autonomous system 1 sends an advertisement message to autonomous system 2, which sends it to autonomous system 3. If the route to network A repeatedly becomes unavailable, then available, many withdrawal and advertisement messages are sent. Route flapping is a problem in an internetwork connected to the Internet, because a route flap in the Internet backbone usually involves many routes.

One way to reduce the iBGP mesh is to divide an autonomous system into multiple subautonomous systems and group them into a single confederation. To the outside world, the confederation looks like a single autonomous system. Each autonomous system is fully meshed within itself and has a few connections to other autonomous systems in the same confederation. Although the peers in different autonomous systems have eBGP sessions, they exchange routing information as if they were iBGP peers. Specifically, the next hop, MED, and local preference information is preserved. This feature allows you to retain a single IGP for all of the autonomous systems.

BGP requires that all iBGP speakers be fully meshed. However, this requirement does not scale well when there are many iBGP speakers. Instead of configuring a confederation, you can reduce the iBGP mesh by using a route reflector configuration.

Figure 1 illustrates a simple iBGP configuration with three iBGP speakers (routers A, B, and C). Without route reflectors, when Router A receives a route from an external neighbor, it must advertise it to both routers B and C. Routers B and C do not readvertise the iBGP learned route to other iBGP speakers because the routers do not pass on routes learned from internal neighbors to other internal neighbors, thus preventing a routing information loop.

With route reflectors, all iBGP speakers need not be fully meshed because there is a method to pass learned routes to neighbors. In this model, an iBGP peer is configured to be a route reflector responsible for passing iBGP learned routes to a set of iBGP neighbors. In Figure 2 , Router B is configured as a route reflector. When the route reflector receives routes advertised from Router A, it advertises them to Router C, and vice versa. This scheme eliminates the need for the iBGP session between routers A and C.

The internal peers of the route reflector are divided into two groups: client peers and all other routers in the autonomous system (nonclient peers). A route reflector reflects routes between these two groups. The route reflector and its client peers form a cluster. The nonclient peers must be fully meshed with each other, but the client peers need not be fully meshed. The clients in the cluster do not communicate with iBGP speakers outside their cluster.

Figure 3 illustrates a more complex route reflector scheme. Router A is the route reflector in a cluster with routers B, C, and D. Routers E, F, and G are fully meshed, nonclient routers.

When the route reflector receives an advertised route, depending on the neighbor, it takes the following actions:

• A route from an external BGP speaker is advertised to all clients and nonclient peers.

• A route from a nonclient peer is advertised to all clients.

• A route from a client is advertised to all clients and nonclient peers. Hence, the clients need not be fully meshed.

Along with route reflector-aware BGP speakers, it is possible to have BGP speakers that do not understand the concept of route reflectors. They can be members of either client or nonclient groups, allowing an easy and gradual migration from the old BGP model to the route reflector model. Initially, you could create a single cluster with a route reflector and a few clients. All other iBGP speakers could be nonclient peers to the route reflector and then more clusters could be created gradually.

An autonomous system can have multiple route reflectors. A route reflector treats other route reflectors just like other iBGP speakers. A route reflector can be configured to have other route reflectors in a client group or nonclient group. In a simple configuration, the backbone could be divided into many clusters. Each route reflector would be configured with other route reflectors as nonclient peers (thus, all route reflectors are fully meshed). The clients are configured to maintain iBGP sessions with only the route reflector in their cluster.

Usually, a cluster of clients has a single route reflector. In that case, the cluster is identified by the router ID of the route reflector. To increase redundancy and avoid a single point of failure, a cluster might have more than one route reflector. In this case, all route reflectors in the cluster must be configured with the cluster ID so that a route reflector can recognize updates from route reflectors in the same cluster. All route reflectors serving a cluster should be fully meshed and all of them should have identical sets of client and nonclient peers.

By default, the clients of a route reflector are not required to be fully meshed and the routes from a client are reflected to other clients. However, if the clients are fully meshed, the route reflector need not reflect routes to clients.

As the iBGP learned routes are reflected, routing information may loop. The route reflector model has the following mechanisms to avoid routing loops:

• Originator ID is an optional, nontransitive BGP attribute. It is a 4-byte attributed created by a route reflector. The attribute carries the router ID of the originator of the route in the local autonomous system. Therefore, if a misconfiguration causes routing information to come back to the originator, the information is ignored.

• Cluster-list is an optional, nontransitive BGP attribute. It is a sequence of cluster IDs that the route has passed. When a route reflector reflects a route from its clients to nonclient peers, and vice versa, it appends the local cluster ID to the cluster-list. If the cluster-list is empty, a new cluster-list is created. Using this attribute, a route reflector can identify if routing information is looped back to the same cluster due to misconfiguration. If the local cluster ID is found in the cluster-list, the advertisement is ignored.

Border Gateway Protocol (BGP) routers receive multiple paths to the same destination. As a standard, by default the BGP best path algorithm decides the best path to install in IP routing table. This is used for traffic forwarding.

BGP assigns the first valid path as the current best path. It then compares the best path with the next path in the list. This process continues, until BGP reaches the end of the list of valid paths. This contains all rules used to determine the best path. When there are multiple paths for a given address prefix, BGP:

• Selects one of the paths as the best path as per the best-path selection rules.

• Installs the best path in its forwarding table. Each BGP speaker advertises only the best-path to its peers.

Note	The advertisement rule of sending only the best path does not convey the full routing state of a destination, present on a BGP speaker to its peers.

After the BGP speaker receives a path from one of its peers; the path is used by the peer for forwarding packets. All other peers receive the same path from this peer. This leads to a consistent routing in a BGP network. To improve the link bandwidth utilization, most BGP implementations choose additional paths satisfy certain conditions, as multi-path, and install them in the forwarding table. Incoming packets for such are load-balanced across the best-path and the multi-path(s). You can install the paths in the forwarding table that are not advertised to the peers. The RR route reflector finds out the best-path and multi-path. This way the route reflector uses different communities for best-path and multi-path. This feature allows BGP to signal the local decision done by RR or Border Router. With this new feature, selected by RR using community-string (if is-best-path then community 100:100). The controller checks which best path is sent to all R's. Border Gateway Protocol routers receive multiple paths to the same destination. While carrying out best path computation there will be one best path, sometimes equal and few non-equal paths. Thus, the requirement for abest-path and is-equal-best-path.

The BGP best path algorithm decides the best path in the IP routing table and used for forwarding traffic. This enhancement within the RPL allows creating policy to take decisions. Adding community-string for local selection of best path. With introduction of BGP Additional Path (Add Path), BGP now signals more than the best Path. BGP can signal the best path and the entire path equivalent to the best path. This is in accordance to the BGP multi-path rules and all backup paths.

Remotely triggered black hole (RTBH) filtering is a technique that provides the ability to drop undesirable traffic before it enters a protected network. RTBH filtering provides a method for quickly dropping undesirable traffic at the edge of the network, based on either source addresses or destination addresses by forwarding it to a null0 interface. RTBH filtering based on a destination address is commonly known as Destination-based RTBH filtering. Whereas, RTBH filtering based on a source address is known as Source-based RTBH filtering.

RTBH filtering is one of the many techniques in the security toolkit that can be used together to enhance network security in the following ways:

• Effectively mitigate DDoS and worm attacks

• Quarantine all traffic destined for the target under attack

• Enforce blocklist filtering

Note	RTBH is not supported in cases such as L3VPN iBGP route over NULL0.

RTBH is implemented by defining a route policy (RPL) to discard undesirable traffic at next-hop using set next-hop discard command.

RTBH filtering sets the next-hop of the victim's prefix to the null interface. The traffic destined to the victim is dropped at the ingress.

The set next-hop discard configuration is used in the neighbor inbound policy. When this config is applied to a path, though the primary next-hop is associated with the actual path but the RIB is updated with next-hop set to Null0. Even if the primary received next-hop is unreachable, the RTBH path is considered reachable and will be a candidate in the bestpath selection process. The RTBH path is readvertised to other peers with either the received next-hop or nexthop-self based on normal BGP advertisement rules.

A typical deployment scenario for RTBH filtering would require running internal Border Gateway Protocol (iBGP) at the access and aggregation points and configuring a separate device in the network operations center (NOC) to act as a trigger. The triggering device sends iBGP updates to the edge, that cause undesirable traffic to be forwarded to a null0 interface and dropped.

Consider below topology, where a rogue router is sending traffic to a border router.

route-policy RTBH-trigger
  if tag is 777 then
    set community (1234:4321, no-export) additive
    pass
  else
    pass
  endif
  end-policy

router bgp 65001
 address-family ipv4 unicast
  redistribute static route-policy RTBH-trigger
 !
 neighbor 192.168.102.1 
  remote-as 65001
  address-family ipv4 unicast
   route-policy bgp_all in
   route-policy bgp_all out

router static
 address-family ipv4 unicast
 10.7.7.7/32 Null0 tag 777

route-policy RTBH
  if community matches-any (1234:4321) then
    set next-hop discard
  else
    pass
  endif
end-policy

router bgp 65001
 address-family ipv4 unicast
 !
 neighbor 192.168.102.2 
  remote-as 65001
  address-family ipv4 unicast
   route-policy RTBH in
   route-policy bgp_all out

Most of the show commands provide address family (AFI) and subaddress family (SAFI) arguments (see RFC 1700 and RFC 2858 for information on AFI and SAFI). The Cisco IOS XR software parser provides the ability to set the afi and safi so that it is not necessary to specify them while running a show command. The parser commands are:

• set default-afi { ipv4 | ipv6 | all }

• set default-safi { unicast | multicast | all }

The parser automatically sets the default afi value to ipv4 and default safi value to unicast . It is necessary to use only the parser commands to change the default afi value from ipv4 or default safi value from unicast . Any afi or safi keyword specified in a show command overrides the values set using the parser commands. Use the following show default-afi-safi-vrf command to check the currently set value of the afi and safi.

Maximum Segment Size (MSS) is the largest amount of data that a computer or a communication device can receive in a single, unfragmented TCP segment. All TCP sessions are bounded by a limit on the number of bytes that can be transported in a single packet; this limit is MSS. TCP breaks up packets into chunks in a transmit queue before passing packets down to the IP layer.

The TCP MSS value is dependent on the maximum transmission unit (MTU) of an interface, which is the maximum length of data that can be transmitted by a protocol at one instance. The maximum TCP packet length is determined by both the MTU of the outbound interface on the source device and the MSS announced by the destination device during the TCP setup process. The closer the MSS is to the MTU, the more efficient is the transfer of BGP messages. Each direction of data flow can use a different MSS value.

Carrier supporting carrier (CSC) is a term used to describe a situation in which one service provider allows another service provider to use a segment of its backbone network. The service provider that provides the segment of the backbone network to the other provider is called the backbone carrier. The service provider that uses the segment of the backbone network is called the customer carrier.

A backbone carrier offers Border Gateway Protocol and Multiprotocol Label Switching (BGP/MPLS) VPN services. The customer carrier can be either:

• An Internet service provider (ISP) (By definition, an ISP does not provide VPN service.)

• A BGP/MPLS VPN service provider

You can configure a CSC network to enable BGP to transport routes and MPLS labels between the backbone carrier provider edge (PE) routers and the customer carrier customer edge (CE) routers using multiple paths. The benefits of using BGP to distribute IPv4 routes and MPLS label routes are:

• BGP takes the place of an Interior Gateway Protocol (IGP) and Label Distribution Protocol (LDP) in a VPN routing and forwarding (VRF) table. You can use BGP to distribute routes and MPLS labels. Using a single protocol instead of two simplifies the configuration and troubleshooting.

• BGP is the preferred routing protocol for connecting two ISPs, mainly because of its routing policies and ability to scale. ISPs commonly use BGP between two providers. This feature enables those ISPs to use BGP.

For detailed information on configuring MPLS VPN CSC with BGP, see the Implementing MPLS Layer 3 VPNs on Cisco IOS XR Software module of the MPLS Configuration Guide for the Cisco CRS Routers.

BGP keychains enable keychain authentication between two BGP peers. The BGP endpoints must both comply with draft-bonica-tcp-auth-05.txt and a keychain on one endpoint and a password on the other endpoint does not work.

See the System Security Configuration Guide for Cisco CRS Routers for information on keychain management.

BGP is able to use the keychain to implement hitless key rollover for authentication. Key rollover specification is time based, and in the event of clock skew between the peers, the rollover process is impacted. The configurable tolerance specification allows for the accept window to be extended (before and after) by that margin. This accept window facilitates a hitless key rollover for applications (for example, routing and management protocols).

The key rollover does not impact the BGP session, unless there is a keychain configuration mismatch at the endpoints resulting in no common keys for the session traffic (send or accept).

The BGP Multicast VPN feature uses the IPv4 multicast distribution tree (MDT) subaddress family identifier (SAFI) in Border Gateway Protocol (BGP).

Multicast VPN (MVPN) extends the VPN architecture to provide multicast services over a shared service provider backbone using native multicast technology. This is achieved using virtual connections between provider edge (PE) routers in each VPN and using native multicast forwarding inside the provider network. An MDT may span across multiple customer sites and the provider network, allowing traffic to flow freely from one source to multiple receivers.

MVPN is supported on VPN networks based on MPLS and on networks based on IP Layer 2 Tunnel Protocol version 3 (L2TPv3).

PE routers are the only routers that must be MVPN-aware and that must be able to signal to remote PEs information regarding the MVPN. Therefore, all PE routers must have a BGP relationship with each other—either directly or using a route reflector (RR).

Generally the source address of the default MDT is the same address used to source the internal BGP (iBGP) sessions with the remote PE routers that belong to the same VPN and multicast VPN routing and forwarding (MVRF) instance. When Protocol Independent Multicast–Source Specific Multicast (PIM–SSM) is used for transport inside the provider core, it is through the BGP relationship that the PEs indicate that they are MVPN-capable and provide for source discovery. This capability is indicated using the updated BGP message.

Note	The source address can also be configured uniquely per VRF instance under multicast-routing configuration. See Multicast Configuration Guide for Cisco CRS Routers.

When a PE receives a BGP update, which includes the rendezvous point (RP) and the group information, it joins the root of that tree, thereby joining the MDT.

Figure 1 shows Multiprotocol iBGP updates for MVPN. On PE1, PE2 is configured as its iBGP peer. This BGP peer configuration within a VRF triggers the MP-iBGP updates that send PE1 local VPN routes to PE2. BGP process on PE2 receives the VPN updates and installs VPN routes in the Routing Information Base (RIB) VRF table. When PIM looks up a VRF source or rendezvous point address that is reachable through the provider core, it receives an MP-iBGP route from the RIB.

When an MVPN-specific default MDT group is configured on PE1, PIM creates a virtual MDT tunnel interface with the tunnel source address the same as the BGP local peering address. This MDT interface is used by PIM to send VPN packets to the provider network and to receive VPN packets from the provider network. PIM also exchanges control messages over this MDT interface.

Each time a default MDT group is configured for a specific VRF, BGP builds an MDT SAFI update, with network layer reachability information (NLRI) containing the local PE BGP peering address and the newly configured MDT group address (The NLRI format is 8-byte-RD:IPv4-address followed by the MDT group address). This update is sent to all the BGP peers including PE2. The BGP process on PE2 receives this MDT update and notifies PIM. If the group is a PIM–SSM group, PIM on PE2 begins sending SSM joins to the BGP peering address on PE1 to establish an SSM tree in the core. This SSM tree is used to carry PIM control traffic and multicast data traffic in the corresponding VRF.

In summary, PIM requires the following from BGP:

• A new BGP MDT SAFI, which carries the VRF RD and BGP local peering address and default MDT group in its NLRI.

• A notification mechanism from BGP to PIM about the availability of the MDT SAFI update.

• A notification mechanism from PIM to BGP about the default MDT group address and source address.

See Internet Engineering Task Force (IETF) draft-nalawade-idr-mdt-safi-03 for detailed information on MDT SAFI.

The Border Gateway Protocol (BGP) Nonstop Routing (NSR) with Stateful Switchover (SSO) feature enables all bgp peerings to maintain the BGP state and ensure continuous packet forwarding during events that could interrupt service. Under NSR, events that might potentially interrupt service are not visible to peer routers. Protocol sessions are not interrupted and routing states are maintained across process restarts and switchovers.

BGP NSR provides nonstop routing during the following events:

• Route processor switchover

• Process crash or process failure of BGP or TCP

  Note

  In case of process crash or process failure, NSR will be maintained only if nsr process-failures switchover command is configured. In the event of process failures of active instances, the nsr process-failures switchover configures failover as a recovery action and switches over to a standby route processor (RP) or a standby distributed route processor (DRP) thereby maintaining NSR. An example of the configuration command is RP/0/RSP0/CPU0:router(config) # nsr process-failures switchover The nsr process-failures switchover command maintains both the NSR and BGP sessions in the event of a BGP or TCP process crash. Without this configuration, BGP neighbor sessions flap in case of a BGP or TCP process crash. This configuration does not help if the BGP or TCP process is restarted in which case the BGP neighbors are expected to flap.

• In-Service System Upgrade (ISSU)

• Minimum Disruption Restart (MDR)

During route processor switchover and In-Service System Upgrade (ISSU), NSR is achieved by stateful switchover (SSO) of both TCP and BGP.

NSR does not force any software upgrades on other routers in the network, and peer routers are not required to support NSR.

When a route processor switchover occurs due to a fault, the TCP connections and the BGP sessions are migrated transparently to the standby route processor, and the standby route processor becomes active. The existing protocol state is maintained on the standby route processor when it becomes active, and the protocol state does not need to be refreshed by peers.

Events such as soft reconfiguration and policy modifications can trigger the BGP internal state to change. To ensure state consistency between active and standby BGP processes during such events, the concept of post-it is introduced that act as synchronization points.

BGP NSR provides the following features:

• NSR-related alarms and notifications

• Configured and operational NSR states are tracked separately

• NSR statistics collection

• NSR statistics display using show commands

• XML schema support

• Auditing mechanisms to verify state synchronization between active and standby instances

• CLI commands to enable and disable NSR

NSR can be provisioned on a multishelf router. The following guidelines should be observed when provisioning NSR on a multishelf router:

• When provisioning NSR for line cards installed on a single rack, provision the active and standby applications on the distributed route processor (DRP) of that rack. If a rack failure occurs, sessions are dropped, because all line cards go down.

• When provisioning NSR for line cards installed on different racks, use one of the following three options:

  • Provision the active and standby applications on a distributed route processor (DRP) redundant pair, where there is a separate route processor in each rack. This configuration uses up two revenue-producing line-card slots on each rack, but is the most secure configuration.

  • Provision the active and standby applications on a distributed route processor (DRP) pair that spans two racks. In this configuration, the active/standby role of the line cards is not dependent on the active/standby role of the DRPs. This is called flexible process redundancy and provides for rack loss and efficient use of LC slots. Use of distributed BGP is not required with this solution.

  Note

  Sessions on line cards in a lost rack are not protected with any of the above options, because there is no line-card redundancy. These options ensure only that sessions on other racks are not affected by a lost rack. However, lost sessions from a lost rack may cause some traffic loss on other racks, because destinations learned through those lost sessions may no longer have alternate routes. Also, rack loss may cause the CPUs on route processors of active racks to slow as they attempt to define new paths for some routes.

The Border Gateway Protocol (BGP) best–external path functionality supports advertisement of the best–external path to the iBGP and Route Reflector peers when a locally selected bestpath is from an internal peer.

BGP selects one best path and one backup path to every destination. By default, selects one best path . Additionally, BGP selects another bestpath from among the remaining external paths for a prefix. Only a single path is chosen as the best–external path and is sent to other PEs as the backup path.

BGP calculates the best–external path only when the best path is an iBGP path. If the best path is an eBGP path, then best–external path calculation is not required.

The procedure to determine the best–external path is as follows:

1. Determine the best path from the entire set of paths available for a prefix.

2. Eliminate the current best path.

3. Eliminate all the internal paths for the prefix.

4. From the remaining paths, eliminate all the paths that have the same next hop as that of the current best path.

5. Rerun the best path algorithm on the remaining set of paths to determine the best–external path.

BGP considers the external and confederations BGP paths for a prefix to calculate the best–external path.

BGP advertises the best path and the best–external path as follows:

• On the primary PE—advertises the best path for a prefix to both its internal and external peers

• On the backup PE—advertises the best path selected for a prefix to the external peers and advertises the best–external path selected for that prefix to the internal peers

The advertise best-external command enables the advertisement of the best–external path in global address family configuration mode and VRF address family configuration mode.

When a primary PE-CE link fails, BGP withdraws the route corresponding to the primary path along with its local label and programs the backup path in the Routing Information Base (RIB) and the Forwarding Information Base (FIB), by default.

However, until all the internal peers of the primary PE reconverge to use the backup path as the new bestpath, the traffic continues to be forwarded to the primary PE with the local label that was allocated for the primary path. Hence the previously allocated local label for the primary path must be retained on the primary PE for some configurable time after the reconvergence. BGP Local Label Retention feature enables the retention of the local label for a specified period. If no time is specified, the local lable is retained for a default value of five minutes.

The retain local-label command enables the retention of the local label until the network is converged.

Cisco IOS XR software provides the capability to run Border Gateway Protocol (BGP) over Generic Routing Encapsulation (GRE) tunnel interfaces.

GRE protocol transports packets of one protocol over another protocol by means of encapsulation. Service Providers can provide IP, MPLS VPN or L2VPN services between their networks that are connected together by a public network using GRE encapsulation to carry data securely over the public network.

The packet that needs to be transported is first encapsulated in a GRE header, which is further encapsulated in another protocol like IPv4 or IPv6 and then forwarded to the destination.

The Cisco IOS XR software GRE implementation is compliant with GRE encapsulation defined in RFC 2784. Key and Sequence numbering as defined in RFC 2890 is not supported in Cisco IOS XR software GRE implementation. To be backward compliant with RFC 1701, Cisco IOS XR software transmits GRE packets with Reserved0 field set to zero. A receiver that is compliant with RFC 1701 treats key present, sequence number, and strict source route as zero and do not expect key and sequence number. The Cisco IOS XR software discards a GRE packet with any of the bits in Reserved0 field set.

From Cisco IOS XR Release 3.9.0 onwards, the Border Gateway Protocol (BGP) commands use disable keyword to disable a feature. The keyword inheritance-disable disables the inheritance of the feature properties from the parent level.

The Border Gateway Protocol (BGP) Additional Paths feature modifies the BGP protocol machinery for a BGP speaker to be able to send multiple paths for a prefix. This gives 'path diversity' in the network. The add path enables BGP prefix independent convergence (PIC) at the edge routers.

Note	BGP Additional Path feature is not supported under vrf.

BGP add path enables add path advertisement in an iBGP network and advertises the following types of paths for a prefix:

• Backup paths—to enable fast convergence and connectivity restoration.

• Group-best paths—to resolve route oscillation.

• All paths—to emulate an iBGP full-mesh.

  Note	Add path is not be supported with MDT, tunnel, and L2VPN address families and eBGP peerings.

When a Border Gateway Protocol (BGP) speaking router that has no local policy configured, receives multiple network layer reachability information (NLRI) from the internal BGP (iBGP) for the same destination, the router will choose one iBGP path as the best path. The best path is then installed in the IP routing table of the router.

The iBGP Multipath Load Sharing feature enables the BGP speaking router to select multiple iBGP paths as the best paths to a destination. The best paths or multipaths are then installed in the IP routing table of the router.

When there are multiple border BGP routers having reachability information heard over eBGP, if no local policy is applied, the border routers will choose their eBGP paths as best. They advertise that bestpath inside the ISP network. For a core router, there can be multiple paths to the same destination, but it will select only one path as best and use that path for forwarding. iBGP multipath load sharing adds the ability to enable load sharing among multiple equi-distant paths.

Configuring multiple iBGP best paths enables a router to evenly share the traffic destined for a particular site.

The iBGP Multipath Load Sharing feature functions similarly in a Multiprotocol Label Switching (MPLS) Virtual Private Network (VPN) with a service provider backbone.

For multiple paths to the same destination to be considered as multipaths, the following criteria must be met:

• All attributes must be the same. The attributes include weight, local preference, autonomous system path (entire attribute and not just length), origin code, Multi Exit Discriminator (MED), and Interior Gateway Protocol (iGP) distance.

• The next hop router for each multipath must be different.

Even if the criteria are met and multiple paths are considered multipaths, the BGP speaking router will still designate one of the multipaths as the best path and advertise this best path to its neighbors.

Per-vrf label mode is not supported for Carrier Supporting Carrier (CSC) network with internal and external BGP multipath setup

Per VRF label mode cannot be used for BGP PIC edge with eiBGP multipath as that might cause loops. Only per prefix label supports per VRF label mode.

Traditional BGP multipath feature allows a router receiving parallel paths to the same destination to install the multiple paths in the routing table. By default, this multipath feature is applied to all configured peers. BGP selective multipath allows application of the multipath feature only to selected peers.

The BGP router receiving multiple paths is configured with the maximum-paths ... selective option. The iBGP/eBGP neighbors sharing multiple paths are configured with the multipath option, while being added as neighbors on the BGP router.

The following behavior is to be noted while using BGP selective multipath:

• BGP selective multipath does not impact best path calculations. A best path is always included in the set of multipaths.

• For VPN prefixes, the PE paths are always eligible to be multipaths.

For information on the maximum-paths and multipath commands, see the Cisco ASR 9000 Series Aggregation Services Router Routing Command Reference.

Topology

A sample topology to illustrate the configuration used in this section is shown in the following figure.

Router R4 receives parallel paths from Routers R1, R2 and R3 to the same destination. If Routers R1 and R2 are configured as selective multipath neighbors on Router R4, only the parallel paths from these routers are installed in the routing table of Router R4.

Configuration

Note	Configure your network topology with iBGP/eBGP running on your routers, before configuring this feature.

To configure BGP selective multipath on Router R4, use the following steps.

1. Configure Router R4 to accept selective multiple paths in your topology.

   
   /* To configure selective multipath for iBGP/eBGP
   RP/0/RP0/CPU0:router(config)# router bgp 1
   RP/0/RP0/CPU0:router(config-bgp)# address-family ipv4 unicast
   RP/0/RP0/CPU0:router(config-bgp-af)# maximum-paths ibgp 4 selective 
   RP/0/RP0/CPU0:router(config-bgp-af)# maximum-paths ebgp 5 selective 
   RP/0/RP0/CPU0:router(config-bgp-af)# commit
   
   /* To configure selective multipath for eiBGP
   RP/0/RP0/CPU0:router(config)# router bgp 1
   RP/0/RP0/CPU0:router(config-bgp)# address-family ipv4 unicast
   RP/0/RP0/CPU0:router(config-bgp-af)# maximum-paths eibgp 6 selective
   RP/0/RP0/CPU0:router(config-bgp-af)# commit
   

2. Configure neighbors for Router R4.

   Routers R1 (1.1.1.1) and R2 (2.2.2.2) are configured as neighbors with the multipath option.

   Router R3 (3.3.3.3) is configured as a neighbor without the multipath option, and hence the routes from this router are not eligible to be chosen as multipaths.

   
   RP/0/RP0/CPU0:router(config-bgp)# neighbor 1.1.1.1
   RP/0/RP0/CPU0:router(config-bgp-nbr)# address-family ipv4 unicast
   RP/0/RP0/CPU0:router(config-bgp-nbr-af)# multipath     
   RP/0/RP0/CPU0:router(config-bgp-nbr-af)# commit
   
   RP/0/RP0/CPU0:router(config-bgp-nbr)# neighbor 2.2.2.2
   RP/0/RP0/CPU0:router(config-bgp-nbr)# address-family ipv4 unicast
   RP/0/RP0/CPU0:router(config-bgp-nbr-af)# multipath
   RP/0/RP0/CPU0:router(config-bgp-nbr-af)# commit
   
   RP/0/RP0/CPU0:router(config-bgp-nbr)# neighbor 3.3.3.3
   RP/0/RP0/CPU0:router(config-bgp-nbr)# address-family ipv4 unicast
   RP/0/RP0/CPU0:router(config-bgp-nbr-af)# commit
   
   

You have successfully configured the BGP selective multipath feature.

The Accumulated Interior Gateway Protocol (AiGP)Attribute is an optional non-transitive BGP Path Attribute. The attribute type code for the AiGP Attribute is to be assigned by IANA. The value field of the AiGP Attribute is defined as a set of Type/Length/Value elements (TLVs). The AiGP TLV contains the Accumulated IGP Metric.

The AiGP feature is required in the 3107 network to simulate the current OSPF behavior of computing the distance associated with a path. OSPF/LDP carries the prefix/label information only in the local area. Then, BGP carries the prefix/lable to all the remote areas by redistributing the routes into BGP at area boundaries. The routes/labels are then advertised using LSPs. The next hop for the route is changed at each ABR to local router which removes the need to leak OSPF routes across area boundaries. The bandwidth available on each of the core links is mapped to OSPF cost, hence it is imperative that BGP carries this cost correctly between each of the PEs. This functionality is achieved by using the AiGP.

The per VRF and per CE label for IPv6 feature makes it possible to save label space by allocating labels per default VRF or per CE nexthop.

All IPv6 Provider Edge (6PE) labels are allocated per prefix by default. Each prefix that belongs to a VRF instance is advertised with a single label, causing an additional lookup to be performed in the VRF forwarding table to determine the customer edge (CE) next hop for the packet.

However, use the label mode command with the per-ce keyword or the per-vrf keyword to avoid the additional lookup on the PE router and conserve label space.

Use per-ce keyword to specify that the same label be used for all the routes advertised from a unique customer edge (CE) peer router. Use the per-vrf keyword to specify that the same label is to be used for all the routes advertised from a unique VRF. In 6PE, the label is IPV6 explicit null label.

Constrained Route Distribution is a feature that service providers use in Multiprotocol Label Switching (MPLS) Layer 3 Virtual Private Networks (L3VPNs) to reduce the number of unnecessary routing updates that route reflectors (RR) send to provider edge (PE) routers. The reduction in routing updates saves resources. RRs, autonomous system boundary routers (ASBRs), and PEs will have fewer routes to carry. Route targets are used to constrain routing updates.

Some service providers have a very large number of routing updates being sent from RRs to PEs, using considerable resources. A PE does not need routing updates for VRFs that are not on the PE; therefore, the PE determines that many routing updates it receives are "unwanted." The PE filters out these unwanted updates.

Now consider a scenario where there are two RRs with another set of PEs. Not only are there unwanted routing updates from RR to PE, there are also unwanted routing updates between the RRs. As a result, a large number of unwanted routes might be advertised among RRs and PEs. The Constrained Route Distribution feature addresses this problem by filtering unwanted routing updates. When the Constrained Route Distribution is in place, the RR filters the updates.

Selective VRF Download (SVD) feature enables the downloading of only those prefixes and labels to a line card that are actively required to forward traffic through the line card.

To meet the demand for a consolidated edge MSE platform, the number of VRFs, VRF interfaces, and the prefix capacity increase. Convergence timings differ in different line card engines. One of the major factors that determine convergence timing is the time taken to process and program a prefix and its associated data structures. A lesser number of prefixes and labels ensure better convergence timing. By enabling selective download of VRF routes to both Engine-3 (E3) and Engine-5 (E5) line cards, SVD reduces scalability and convergence problems in Layer 3 VPNs (L3VPNs)..

By default, SVD is enabled on the line cards. Use the selective-vrf-download disable command to disable SVD. Use the show svd role and show svd state commands to display role and state information of SVD on line cards.

The BGP Accept Own feature enables handling of self-originated VPN routes, which a BGP speaker receives from a route-reflector (RR). A "self-originated" route is one which was originally advertized by the speaker itself. As per BGP protocol [RFC4271], a BGP speaker rejects advertisements that were originated by the speaker itself. However, the BGP Accept Own mechanism enables a router to accept the prefixes it has advertised, when reflected from a route-reflector that modifies certain attributes of the prefix. A special community called ACCEPT-OWN is attached to the prefix by the route-reflector, which is a signal to the receiving router to bypass the ORIGINATOR_ID and NEXTHOP/MP_REACH_NLRI check. Generally, the BGP speaker detects prefixes that are self-originated through the self-origination check (ORIGINATOR_ID, NEXTHOP/MP_REACH_NLRI) and drops the received updates. However, with the Accept Own community present in the update, the BGP speaker handles the route.

One of the applications of BGP Accept Own is auto-configuration of extranets within MPLS VPN networks. In an extranet configuration, routes present in one VRF is imported into another VRF on the same PE. Normally, the extranet mechanism requires that either the import-rt or the import policy of the extranet VRFs be modified to control import of the prefixes from another VRF. However, with Accept Own feature, the route-reflector can assert that control without the need for any configuration change on the PE. This way, the Accept Own feature provides a centralized mechanism for administering control of route imports between different VRFs.

BGP Accept Own is supported only for VPNv4 and VPNv6 address families in neighbor configuration mode.

Accept Own Configuration Example

In this configuration example:

• PE11 is configured with Customer VRF and Service VRF.

• OSPF is used as the IGP.

• VPNv4 unicast and VPNv6 unicast address families are enabled between the PE and RR neighbors and IPv4 and IPv6 are enabled between PE and CE neighbors.

The Accept Own configuration works as follows:

1. CE1 originates prefix X.

2. Prefix X is installed in customer VRF as (RD1:X).

3. Prefix X is advertised to IntraAS-RR11 as (RD1:X, RT1).

4. IntraAS-RR11 advertises X to InterAS-RR1 as (RD1:X, RT1).

5. InterAS-RR1 attaches RT2 to prefix X on the inbound and ACCEPT_OWN community on the outbound and advertises prefix X to IntraAS-RR31.

6. IntraAS-RR31 advertises X to PE11.

7. PE11 installs X in Service VRF as (RD2:X,RT1, RT2, ACCEPT_OWN).

Border Gateway Protocol demilitarized zone (BGP DMZ) Link Bandwidth for Unequal Cost Recursive Load Balancing provides support for unequal cost load balancing for recursive prefixes on local node using BGP DMZ Link Bandwidth. The unequal load balance is achieved by using the dmz-link-bandwidth command in BGP Neighbor configuration mode and the bandwidth command in Interface configuration mode.

Bi-directional Forwarding Detection Multihop (BFD-MH) support is enabled for BGP. BFD Multihop establishes a BFD session between two addresses that may span multiple network hops. Cisco IOS XR Software BFD Multihop is based on RFC 5883. For more information on BFD Multihop, refer Interface and Hardware Component Configuration Guide for Cisco CRS Routers and Interface and Hardware Component Command Reference for Cisco CRS Routers.

Multiple BGP instances are supported on the router corresponding to a Autonomous System (AS). Each BGP instance is a separate process running on the same or on a different RP/DRP node. The BGP instances do not share any prefix table between them. No need for a common adj-rib-in (bRIB) as is the case with distributed BGP. The BGP instances do not communicate with each other and do not set up peering with each other. Each individual instance can set up peering with another router independently.

Multi-AS BGP enables configuring each instance of a multi-instance BGP with a different AS number.

Multi-Instance and Multi-AS BGP provides these capabilities:

• Mechanism to consolidate the services provided by multiple routers using a common routing infrastructure into a single IOS-XR router.

• Mechanism to achieve AF isolation by configuring the different AFs in different BGP instances.

• Means to achieve higher session scale by distributing the overall peering sessions between multiple instances.

• Mechanism to achieve higher prefix scale (especially on a RR) by having different instances carrying different BGP tables.

• Improved BGP convergence under certain scenarios.

• Cisco IOS XR CRS Multi-chassis systems can be used optimally by placing the different BGP instances on different RP/DRPs.

• All BGP functionalities including NSR are supported for all the instances.

• The load and commit router-level operations can be performed on previously verified or applied configurations.

A BGP route associates an address prefix with a set of autonomous systems (AS) that identify the interdomain path the prefix has traversed in the form of BGP announcements. This set is represented as the AS_PATH attribute in BGP and starts with the AS that originated the prefix.

To help reduce well-known threats against BGP including prefix mis-announcing and monkey-in-the-middle attacks, one of the security requirements is the ability to validate the origination AS of BGP routes. The AS number claiming to originate an address prefix (as derived from the AS_PATH attribute of the BGP route) needs to be verified and authorized by the prefix holder.

The Resource Public Key Infrastructure (RPKI) is an approach to build a formally verifiable database of IP addresses and AS numbers as resources. The RPKI is a globally distributed database containing, among other things, information mapping BGP (internet) prefixes to their authorized origin-AS numbers. Routers running BGP can connect to the RPKI to validate the origin-AS of BGP paths.

The BGP RPKI Bind Source feature allows you to specify the source IP address and interface used for the RPKI server connection. This feature enables you to have RPKI session that source from loopback interface, for example.

BGP origin-as validation is enabled by default.

The BGP 3107 PIC Updates for Global Prefixes feature supports Prefix Independent Convergence (PIC) updates for global IPv4 and IPv6 prefixes in an MPLS VPN provider network. This feature is based on RFC 3107 that describes using BGP to distribute MPLS labels for global IPv4 or IPv6 prefixes. This enables IGP to scale better and also provides PIC updates for fast convergence.

BGP 3107 PIC is supported on CRS-1 and CRS-3 line cards.

RFC 3107 enables routes and labels to be carried in BGP. When BGP is used to distribute a particular route, it can also be used to distribute an MPLS label that is mapped to that route. The label mapping information for a particular route is piggybacked in the same BGP Update message that is used to distribute the route itself. RFC 3107 allows filtering of Next-Hop Loops from OSPF and reduces labels advertised by LDP. This implementation significantly reduces OSPF and LDP database.

The 3107 PIC implementation supports the following address-families with additional-path configuration.

• address-family ipv4 unicast

• address-family ipv6 unicast

• address-family vpnv4 unicast

• address-family vpnv6 unicast

Note	The address-family l2vpn vpls-vpws does not support additional-path. Hence, the l2vpn service that uses address-family l2vpn vpls-vpws does not guarantee PIC convergence time.

The 3107 PIC implementation supports these Cisco IOS XR features:

• PIC Edge for 3107

• Traffic Engineering Fast-reroute (TE FRR)—Traffic convergence for core link failure is guaranteed within 50 milliseconds using verbatim tunnel.

• L2VPN Service (VPWS)

• L3VPN VPNv4 Service

• 6 PE Service

• 6 VPE Service

• VPLS Service

BGP 3107 PIC Updates for Global Prefixes implementation uses a shared recursive Load Info (RLDI) forwarding object in place of a Light-Weight recursive (LW-RLDI) object. The RLDI is shared between multiple leaves, while the LW-RLDI is instantiated per leaf. Sharing helps in handling PIC updates since it will be prefix independent.

BGP PIC for RIB and FIB adds support for static recursive as PE-CE and faster backup activation by using fast re-route trigger.

The BGP PIC for RIB and FIB feature supports:

• FRR-like trigger for faster PE-CE link down detection, to further reduce the convergence time (Fast PIC-edge activation).

• PIC-edge for static recursive routes.

• BFD single-hop trigger for PIC-Edge without any explicit /32 static route configuration.

• Recursive PIC activation at third level and beyond, on failure trigger at the first (IGP) level.

• BGP path recursion constraints in FIB to ensure that FIB is in sync with BGP with respect to BGP next-hop resolution.

When BGP PIC Edge is configured, configuring the neighbor shutdown command does not trigger CEF to switch to the backup path. Instead, BGP starts to feed CEF again one by one from the top prefix of the routing table to the end thus causing a time delay.

Caution

The time delay causes a black hole in the network. As a workaround, you must route the traffic to the backup path manually before configuring the neighbor shutdown command.

The BGP UPDATE message error handling changes BGP behavior in handling error UPDATE messages to avoid session reset. Based on the approach described in IETF IDR I-D:draft-ietf-idr-error-handling, the Cisco IOS XR BGP UPDATE Message Error handling implementation classifies BGP update errors into various categories based on factors such as, severity, likelihood of occurrence of UPDATE errors, or type of attributes. Errors encountered in each category are handled according to the draft. Session reset will be avoided as much as possible during the error handling process. Error handling for some of the categories are controlled by configuration commands to enable or disable the default behavior.

According to the base BGP specification, a BGP speaker that receives an UPDATE message containing a malformed attribute is required to reset the session over which the offending attribute was received. This behavior is undesirable as a session reset would impact not only routes with the offending attribute, but also other valid routes exchanged over the session.

The BGP Attribute Filter feature checks integrity of BGP updates in BGP update messages and optimizes reaction when detecting invalid attributes. BGP Update message contains a list of mandatory and optional attributes. These attributes in the update message include MED, LOCAL_PREF, COMMUNITY etc. In some cases, if the attributes are malformed, there is a need to filter these attributes at the receiving end of the router. The BGP Attribute Filter functionality filters the attributes received in the incoming update message. The attribute filter can also be used to filter any attributes that may potentially cause undesirable behavior on the receiving router.

Some of the BGP updates are malformed due to wrong formatting of attributes such as the network layer reachability information (NLRI) or other fields in the update message. These malformed updates, when received, causes undesirable behavior on the receiving routers. Such undesirable behavior may be encountered during update message parsing or during re-advertisement of received NLRIs. In such scenarios, its better to filter these corrupted attributes at the receiving end.

When a router receives a malformed update packet, an ios_msg of type ROUTING-BGP-3-MALFORM_UPDATE is printed on the console. This is rate limited to 1 message per minute across all neighbors. For malformed packets that result in actions "Discard Attribute" (A5) or "Local Repair" (A6), the ios_msg is printed only once per neighbor per action. This is irrespective of the number of malformed updates received since the neighbor last reached an "Established" state.

%ROUTING-BGP-3-MALFORM_UPDATE : Malformed UPDATE message received from neighbor 13.0.3.50 - message length 90 bytes,
 error flags 0x00000840, action taken "TreatAsWithdraw". 
Error details: "Error 0x00000800, Field "Attr-missing", Attribute 1 (Flags 0x00, Length 0), Data []"

[4843.46]RP/0/0/CPU0:Aug 21 17:06:17.919 : bgp[1037]: %ROUTING-BGP-5-UPDATE_FILTERED : 
One or more attributes were filtered from UPDATE message received from neighbor 40.0.101.1 - message length 173 bytes,
 action taken "DiscardAttr".
 Filtering details: "Attribute 16 (Flags 0xc0): Action "DiscardAttr"". NLRIs: [IPv4 Unicast] 88.2.0.0/17

[391.01]RP/0/0/CPU0:Aug 20 19:41:29.243 : bgp[1037]: %ROUTING-BGP-5-UPDATE_FILTERED :
 One or more attributes were filtered from UPDATE message received from neighbor 40.0.101.1 - message length 166 bytes,
 action taken "TreatAsWdr".
 Filtering details: "Attribute 4 (Flags 0xc0): Action "TreatAsWdr"". NLRIs: [IPv4 Unicast] 88.2.0.0/17

BGP Link-State (LS) is an Address Family Identifier (AFI) and Sub-address Family Identifier (SAFI) defined to carry interior gateway protocol (IGP) link-state database through BGP. BGP LS delivers network topology information to topology servers and Application Layer Traffic Optimization (ALTO) servers. BGP LS allows policy-based control to aggregation, information-hiding, and abstraction. BGP LS supports IS-IS and OSPFv2.

Note	IGPs do not use BGP LS data from remote peers. BGP does not download the received BGP LS data to any other component on the router.

The Border Gateway Protocol (BGP) dynamic route leaking feature provides the ability to import routes between the default-vrf (Global VRF) and any other non-default VRF, to provide connectivity between a global and a VPN host. The import process installs the Internet route in a VRF table or a VRF route in the Internet table, providing connectivity.

Note	Directly connected routes cannot be leaked using BGP VRF Dynamic Route Leaking from default VRF to non-default VRF

A route-policy is mandatory to filter the imported routes. This reduces the risk of unintended import of routes between the Internet table and the VRF tables and the corresponding security issues.

There is no hard limit on the number of prefixes that can be imported. The import creates a new prefix in the destination VRF, which increases the total number of prefixes and paths. However, each VRF importing global routes adds workload equivalent to a neighbor receiving the global table. This is true even if the user filters out all but a few prefixes. Hence, importing five to ten VRFs is ideal.

The Resilient Per-CE Label is an extension of the Per-CE label mode to support Prefix Independent Convergence (PIC) and load balancing.

In the Resilient Per-CE label scheme, BGP installs a unique rewrite label in LSD for every unique set of CE paths or next hops. There may be one or more prefixes in BGP table that points to this label. BGP also installs the CE paths (primary) and optionally a backup PE path into RIB. FIB learns about the label rewrite information from LSD and the IP paths from RIB.

In steady state, labeled traffic destined to the resilient per-CE label is load balanced across all the CE next hops. When all the CE paths fail, any traffic destined to that label will result in an IP lookup and will be forwarded towards the backup PE path, if available. This action is performed on the label independently of the number of prefixes that may point to the label, resulting in the PIC behavior during primary paths failure.

• Overwriting of next-hop calculation for multipath prefixes is not allowed. The next-hop-unchanged multipath command disables overwriting of next-hop calculation for multipath prefixes.

• The ability to ignore as-path onwards while computing multipath is added. The bgp multipath as-path ignore onwards command ignores as-path onwards while computing multipath.

When multiple connected routers start ignoring as-path onwards while computing multipath, it causes routing loops. Therefore, you should not configure the bgp multipath as-path ignore onwards command on routers that can form a loop.

Consider three routers R1, R2 and R3 in different autonomous systems (AS-1, AS-2, and AS-3). The routers are connected with each other. R1 announces a prefix to R2 and R3. Both R2 and R3 are configured with multipath and also with bgp multipath as-path ignore onwards command. Since R3 is configured as multipath, R2 will send part of its traffic to R3. Similarly, R3 will send part of its traffic to R2. This creates a forwarding loop between R3 and R2. Therefore, to avoid such forwarding loops you should not configure the bgp multipath as-path ignore onwards command on connected routers.

BGP supports Subsequent Address Family Identifier (SAFI)-2 and SAFI-129 for multicast VPNs (MVPNs).

SAFI-129 provides the capability to support multicast routing in the core IPv4 network. SAFI-129 supports BGP-based MVPNs. The addition of SAFI-129 allows multicast to select an upstream multicast hop that may be independent of the unicast topology. Multicast routes learned from the customer edge (CE) router or multicast VPN routes learned from remote provider edge (PE) routers are installed into the multicast Routing Information Base (MuRIB). This MuRIB will be populated with routes that are specific to multicast, and are not used by unicast forwarding. The PE-CE BGP prefixes are advertised using SAFI-2, the PE-PE routes are advertised using SAFI-129.

By default, if a link goes down, all BGP sessions of any directly adjacent external peers are immediately reset. Use the bgp fast-external-fallover disable command to disable automatic resetting. Turn the automatic reset back on using the no bgp fast-external-fallover disable command.

RP/0/RP0/CPU0:router#configure 
RP/0/RP0/CPU0:router(config)#lpts pifib hardware police location 0/2/CPU0
RP/0/RP0/CPU0:router(config-pifib-policer-per-node)#flow bgp configured rate 4000
RP/0/RP0/CPU0:router(config-pifib-policer-per-node)#flow bgp known rate 4000
RP/0/RP0/CPU0:router(config-pifib-policer-per-node)#flow bgp default rate 4000
RP/0/RP0/CPU0:router(config-pifib-policer-per-node)#commit

Logging neighbor changes is enabled by default. Use the log neighbor changes disable command to turn off logging. The no log neighbor changes disable command can also be used to turn logging back on if it has been disabled.