Information Technology Reference
In-Depth Information
in Step 2. What happens next is the same as what happened in Steps 2
through 6. This cycle continues indefinitely.
Table 7-7
New Paths on R1
Path
BGP Next Hop
AS_PATH
MED
RID
1
R4
300 400
6
172.16.67.6
2
R2
200 400
10
192.168.45.4
3
*
R3
300 400
5
192.168.45.5
The problem of persistent convergence loops can be identified by observing the following
two events:
•
Ever-increasing BGP table version. Whenever there is a best-path change, the BGP
table version is incremented.
•
Constant next-hop change in the IP RIB for the destination. When there is a BGP best-
path change, the IP RIB is updated with a new next hop.
Solutions to Incomparable Inter-AS Metrics
Because this problem is inherent in the architecture and the protocol, solutions have to
come from design workarounds. Several options can be implemented individually or
combined:
•
Use a full iBGP mesh
•
Enable
always-compare-med
•
Enable a
deterministic-MED
comparison
•
Reset MEDs to 0s
•
Use communities
These five solutions are explored in the following sections.
Using Full iBGP Mesh
When full mesh is used, all iBGP routers have the complete
routing information, and a convergence loop is not formed. However, this option might be
unacceptable if RRs were selected in the first place to increase scalability.
Enabling
always-compare-MED
When MED is compared among all neighboring
autonomous systems, the path with the lowest MED always wins. In Figure 7-17, for
example, AS 100 will always prefer the path via R3. However, this
always-compare-MED
option has a couple issues:
•
MEDs from different autonomous systems might not always be comparable. Making
a comparison requires close coordination among all peering autonomous systems to
associate the MED with consistent and meaningful metrics.