Facts to remember about BGP - Part 2

Recall that the next-hop value is only updated by default when prefixes are advertised to true EBGP peers, not iBGP peers or confederation EBGP peers.


BGP Next-Hop Processing - Next-Hop-Self
                                                   R1<---(iBGP)---->R2<----(eBGP)----->R3

• R2 learns a new prefix from R3 which becomes the next-hop for this prefix
• When R2 advertise this prefix to R1, the next-hop value for R1 is still R3 address
• R1 will drop packets sent to the prefix  cause next-hop (R3) is unreachable  for R1
• The solution would be if R1 has R2 as next-hop to reach this new prefix
• next-hop-self statement will change this in the router which receives the eBGP prefix (R2)

router bgp 100

 network 150.1.7.7 mask 255.255.255.255
 neighbor 155.1.146.1 remote-as 100
 neighbor 155.1.146.1 next-hop-self
 neighbor 155.1.79.9 remote-as 54

BGP Next-Hop Processing - Manual Modification

Two possible scenarios where the next-hop can be fixed. On R1(RR) or in R7 (RR client)

R1:
route-map SET_NEXT_HOP_FROM_R8 permit 10
 set ip next-hop 155.1.58.8
!
router bgp 100
 network 150.1.1.1 mask 255.255.255.255
 neighbor 155.1.58.8 remote-as 100
 neighbor 155.1.58.8 route-map SET_NEXT_HOP_FROM_R8 in
 neighbor 155.1.67.7 remote-as 100
 neighbor 155.1.58.8 route-reflector-client
 neighbor 155.1.67.7 route-reflector-client

R7:
route-map SET_NEXT_HOP_TO_R1 permit 10
 set ip next-hop 155.1.67.7
!
router bgp 100
 network 150.1.7.7 mask 255.255.255.255
 neighbor 155.1.146.1 remote-as 100
 neighbor 155.1.146.1 route-map SET_NEXT_HOP_TO_R1 out
 neighbor 155.1.79.9 remote-as 54




iBGP Synchronization
If your AS passes traffic from another AS to a third AS, BGP should not advertise a route before all the routers in your AS have learned about the route via IGP. BGP waits until IGP has propagated the route within the AS. Then, BGP advertises the route to external peers.

BGP synchronization is disabled by default cause none IGP can scale to the size of the current Internet BGP table. When BGP is redistribute into IGP there will be a "rib-failure" (show ip bgp) that means the BGP route will not be installed in the routing table but the IGP route cause its lower administrative distance than iBGP (200).
Route map filtering should always be used when redistribute BGP into IGP to avoid CPU and memory exhaustion like the example

R8:
ip as-path access-list 1 permit ^54_
!
route-map BGP_TO_IGP permit 10
 match as-path 1
!
router eigrp 100
 redistribute bgp 100 metric 100000 1000 255 1 1500 route-map BGP_TO_IGP
 network 150.1.8.8 0.0.0.0
!
router bgp 100
 synchronization

BGP Redistribute Internal
Redistribute iBGP into IGP is disabled by default cause potential loops like the example

• R3 and R4 know that the next-hop to 1.1.1.1 is through R2 (iBGP advertise)
• R3 will always reach R2 through R4 cause its lowest IGP metric (20)
• If R3 distributes 1.1.1.1 into OSPF, R2 will have a new route to reach 1.1.1.1 with R3 as next-hop, this route will be prefered over the iBGP learned route cause a better AD (110 vs 200)
• When R3 try to go 1.1.1.1 through R2, R2 will try to go through R3 and the loop will begin
• To enable iBGP redistribution into IGP use the statement bgp redistribute-internal under the router bgp process 

BGP Peer Groups

R1:
router bgp 100
 neighbor IBGP_PEERS peer-group
 neighbor IBGP_PEERS remote-as 100
 neighbor IBGP_PEERS update-source Loopback0
 neighbor IBGP_PEERS route-reflector-client
 neighbor 150.1.2.2 peer-group IBGP_PEERS
 neighbor 150.1.3.3 peer-group IBGP_PEERS
 network 150.1.1.1 mask 255.255.255.255

R2:
router bgp 100
 neighbor 150.1.1.1 remote-as 100
 neighbor 150.1.1.1 update-source Loopback0
 network 150.1.2.2 mask 255.255.255.255

R3:
router bgp 100
 neighbor 150.1.1.1 remote-as 100
 neighbor 150.1.1.1 update-source Loopback0
 network 150.1.3.3 mask 255.255.255.255

BGP Network Statement
Another way to deal with external next-hop besides next-hop-self is to propagate the external link subnet to all the iBGP peers so they will install it in their RIBs and do a recursive look up till they  find the next hop

R7:
router bgp 100
 no neighbor 150.1.1.1 next-hop-self
 network 155.1.79.0 mask 255.255.255.0







BGP Auto-Summary

Works with the network statement (classfull network subnet in the IGP table must exist) and when we do redistribution

R7:
route-map CONNECTED_TO_BGP
 match interface Loopback0
 match interface GigabitEthernet1.79
!
router bgp 100
 no network 150.1.7.7 mask 255.255.255.255
 no network 155.1.79.0 mask 255.255.255.0
 auto-summary
 redistribute connected route-map CONNECTED_TO_BGP


R8:
router bgp 100
 no network 150.1.8.8 mask 255.255.255.255
 no network 155.1.108.0 mask 255.255.255.0
 auto-summary
 network 150.1.0.0
 network 155.1.0.0

BGP best path selection process

1. Ignore invalid paths (no valid next-hop, not synchronized, looped).
2. Prefer path with the highest locally assigned weight value.
3. Prefer path with the highest Local Preference attribute value.
4. Prefer locally originated prefixes (originated via the network, aggregate-address, or redistribution commands).
5. Prefer path with the shortest AS_PATH attribute length.
6. Prefer path with the lowest Origin Type (value for the Origin code), where IGP < EGP < Incomplete.
7. Prefer path with the lowest MED attribute value (provided that the first AS in the list is the same).
8. Prefer external BGP paths over internal BGP paths.
9. Prefer path with the smallest IGP metric to reach the NEXT_HOP IP address.
10. Prefer path originated from the router with the lowest BGP Router ID.

Before the procedure runs, the bestpath process excludes some prefixes based on the following criteria:

No valid next-hop. This is the most common cause for the prefix being ignored by the selection process. BGP prefixes carry their next-hop as a separate attribute (NEXT_HOP attribute). If the next-hop address is NOT reachable via IGP, the prefix is marked as invalid and is not considered. This usually happens with eBGP-learned prefixes when you forget to enter the command next-hop-self or advertise the link subnet into IGP/BGP.
BGP Synchronization enabled and the prefix is not in the IGP table. The bestpath process will ignore this prefix. This is a legacy restriction, but you may occasionally run into it.
AS_PATH Loop. Prefixes from the neighbor that has the local AS number in the AS_PATH attribute are dropped. This is the well-known BGP loop detection mechanism.