BGP basic configuration

The neighboring nodes must be accessible to each other at the network layer.

Follow these steps to create a BGP connection:

	CAUTION: Since a router can reside in only one AS, the router can run only one BGP process.

BGP uses TCP as the transport layer protocol. By default, BGP uses the output interface of the optimal router to a peer as the source interface for establishing TCP connections to the peer. If a BGP router has multiple links to a peer, when the source interface fails, BGP has to reestablish TCP connections, causing network oscillation. Therefore, it is recommended to use a loopback interface as the source interface to enhance stability of BGP connections.

Follow these steps to specify the source interface of TCP connections:

BGP requires TCP as the transport protocol. To enhance security, you can configure BGP to perform MD5 authentication when establishing a TCP connection. The two parties must have the same password configured to establish TCP connections. BGP MD5 authentication is not for BGP packets, but for TCP connections. If the authentication fails, no TCP connection can be established.

In general, direct physical links should be available between eBGP peers. If not, you can use the neighbor ip-address ebgp-multihop command to establish a TCP connection over multiple hops between two peers.

Follow these steps to allow establishment of eBGP connection to a non-directly connected peer.

Before configuring this task, you should have completed the BGP basic configuration.

You can redistribute IGP routes into BGP. During route redistribution, BGP can filter routing information from specific routing protocols.

	NOTE: The ORIGIN attribute of routes redistributed using the import-route command is Incomplete. The ORIGIN attribute of networks advertised into the BGP routing table with the network command is IGP. These networks must exist in the local IP routing table. Using a routing policy makes route control more flexible.

Follow these steps to configure BGP route reception filtering policies:

	CAUTION: Only routes permitted by the specified filtering policies can be installed into the local BGP routing table.

Before configuring this task, you should have configured BGP basic functions.

You can configure BGP route attributes to influence BGP route selection.

Follow these steps to configure BGP route attributes.

	CAUTION: Using a routing policy can set preferences for routes matching it. Routes not matching it use the default preferences. If other conditions are identical, the route with the smallest MED value is selected as the best external route. Using the neighbor next-hop-self command can specify the router as the next hop for routes sent to a peer. If BGP load balancing is configured, the router specifies itself as the next hop for routes sent to a peer regardless of whether the neighbor next-hop-self command is configured. In a “third party next hop” network, that is, a BGP router has two eBGP peers in a common broadcast subnet, the BGP router does not specify itself as the next hop for routes sent to such an eBGP peer, unless the neighbor next-hop-self command is configured. BGP checks if the AS_PATH attribute of a route from a peer contains the local AS number. If so, it discards the route to avoid routing loops. You can specify a fake AS number to hide the real one. The fake AS number applies to routes sent to eBGP peers only, that is, eBGP peers in other ASs can only find the fake AS number. The neighbor as-override command is used only in specific networking environments. Inappropriate use of the command may cause routing loops.

BGP connections have been created.

After establishing a BGP connection, two routers send keepalive messages periodically to each other to keep the connection. If a router receives no keepalive or update message from the peer within the holdtime, it breaks the connection.

If two parties have the same timer assigned with different values, the smaller one is used.

Follow these steps to configure BGP keepalive interval and holdtime.

	CAUTION: The maximum keepalive interval should be one third of the holdtime and no less than 1 second. The holdtime is no less than 3 seconds unless it is set to 0. Intervals set with the neighbor timers command are preferred to those set with the timers command. If the router has established a neighbor relationship with a peer, you need to reset the BGP connection to validate the new set timers.

A BGP community must be configured. Follow these steps.

	CAUTION: When configuring the BGP community, you must configure a routing policy to define the community attribute, and then apply the routing policy to the route advertisement.

Follow these steps to configure a BGP route reflector:

	CAUTION: It is not required to make clients of a route reflector fully meshed. The route reflector forwards routing information between clients. If clients are fully meshed, you can disable route reflection between clients to reduce routing costs. A cluster has only one route reflector, and the router ID is used to identify the cluster. You can configure multiple route reflectors to improve network stability. In this case, you must specify the same cluster ID for these route reflectors to avoid routing loops.