tto / docs / azure / bgp

Feature overview

Uses public or private AS numbers. Note: [reserved-asn]

Range	Description
1-64495	Public
64496-65511	Reserved for use in Documentation
65512-65534	Reserved for private use (like RFC 1918 IP addresses)

Uses tcp/179 for reliability

Used in three major configuration options: default routes, partial updates, full updates.

Used to provide outbound and inbound high availability for internet connectivity.

Defaults to use AS_Path as metric, so the route with less AS_Paths (Hops) wins.

iBGP vs eBGP

iBGP is used internally (same ASN)

eBGP is used with external partners, providers (different ASN)

Requirements for forming eBGP neighbor ships

• A local router’s ASN (router bgp NR) must match the neighboring router’s reference (neighbor remote-as NR)
• BGP Router ID of the two routers must NOT be the same
• MD5 authentication must pass, if configured (neighbor ipaddr password pass)
• TCP connection must be established between each router. Source address used for TCP must match the IP that the other router configures as BGP neighbor
• Neighbors must be directly connected, or the ebgp-multihop router configuration command can be used to set the TTL

BGP Router ID	32 Bit Identifier, dotted-decimal notation Configured: bgp router-id RID Highest Loopback: Use highest numeric IP address of up/up loopback interface Highest other interface: Use highest numeric IP address of up/up interface
MD5 Authentication	Must be configured on none or both routers, passwords must match
TCP	Routers create a TCP connection before exchanging BGP messages. IOS, by default, picks the IP address of the outgoing interface as a source address for the TCP connection. If there are multiple paths to reach a neighbor (2x T1 to same ISP) BGP could be configured twice, using each interface’s IP Address or one could configure BGP to use a loopback address as source address for any given neighbor.

BGP Neighbor States

1. Router tries to establish a TCP connection on destination port: tcp/179
2. When three-way TCP connection completes, the router sends firs BGP Open massage. Generally performs same function as EIGRP/OSPF Hello messages. Open messages contain several BGP parameters that must match for the routers to become neighbors.
3. After Open messages have been sent and received and the parameters match the neighbor relationship is formed and the neighbors reach “established” state.

State	Reason
Idle	BGP process is either administratively down or awaiting the next retry attempt
Connect	Waiting for the TCP connection to be completed
Active	TCP connection established, no BGP messages sent
Opensent	BGP Open Message has been sent, but no matching Open Massage has been received yet
Openconfirm	Open Massages sent and received. Waiting for BGP Keepalive (all Neighbor parameters match) or BGP Notification (to learn there is some mismatch in neighbor parameters)
Established	All neighbor parameters match, relationship works, peers can now exchange BGP Update messages

Cisco details

Network

Unlike other routing protocols the network command is not used to enable bgp on a specified interface, the network command is only used to inject routes into the bgp process.

Another possibility to get networks into bgp is to used redistribution, even though it might not be very useful to fully redistribute a IGP into BGP. Redistribute connected can be used instead.

Auto Summary

Summarizes network to class full boundaries.

Auto-Summary turned off by default from 12.2(8)T and newer

Turn it off for earlier versions: (config-router)#no auto-summary

BGP Synchronization

BGP does not use or advertise a route learned from iBGP until that route has also been learned from an internal routing protocol (try to prevent network blackholes)

Is turned off by default from 12.2(8)T and newer

Turn it off in earlier versions: (config-router)#no synchronization

BGP Next Hop processing

For eBGP peers: Change the next hop address

For iBGP peers: Don’t change the next hop address

Update next hop addresses for routes advertised to this neighbor: (config-router)#neighbor x.x.x.x next-hop-self

BGP Attributes

There are five Types of BGP Attributes

Well Known	Supported by all vendors
Mandatory	Must be in every update
Discretionary	Optional
Transitive	Advertised to eBGP peers, so ‘travels’ between AS
Non-Transitive	Stay “local” to Autonomous Systems

Mandatory, Well Known Attributes

AS-Path

Next Hop Address

Origin

BGP Path decision

BGP has to find one (and only one) best route for every destination, so if routes tie the following Attributes are used to break the tie (ordered)

N	Next Hop reachable?
W	Prefer higher Weight (Cisco, Default 0)
L	Prefer highest Local Preference (Default 100)
L	Prefer locally originated via network command
A	Prefer shortest AS_PATH length
O	Origin prefer I > E > ?
M	Prefer lowest metric/Multi Exit Discriminator
N	Neighbor type, prefer eBGP over iBGP
I	Prefer route with smaller IGP metric to next Hop

Weight

Is local to the ROUTER!

!Set all routes received from this neighbor to 30
(config-router)#neighbor x.x.x.x weight 30

!Use a route map to set weight for specific routes
route-map set weight

Local Preference

Is local to the Autonomous System, if set on one router it is advertised into the local AS

!Set the default local_pref for this router to 200
(config-router)#bgp default local-preference 200

!Use a route map to set local_pref for specific routes
route-map set local-preference