BGP rib-failure - silly question :(

[kfarrington]

When BGP tries to install bestpath prefix into Routing Information Base (RIB), i.e., the IP Routing table,

RIB might reject the BGP route due to any of the following reasons:

Route with better administrative distance already present in IGP, for example, if a static route already

exists in IP Routing table.

Memory failure.

The number of routes in VPN routing/forwarding (VRF) exceeds the route-limit configured under the VRF instance.

------------------------------------------------

All, with this statement in mind, if you have a route in the routing table (say eigrp) and you redist this route

into BGP.

It goes into BGP with a weight of 32768 (next hop 0.0.0.0) and is the BGP best path, but why does it not rib the route.

As it would have a route in the router with a lower admin distance ie the EIGRP route (asuming that locally generated

and aggregate routes have an AD associated with them of iBGP AD200).

I know you would not want to RIB the route as it would defeat the object of redistributing it in, but

what stops this from happening.

I know that if you receive a route from an iBGP or eBGP peer and you have a better AD route in your routing table,

it would be ribbed. Is the BGP rib condition only applicable to BGP learned (ie ibgp and eBGP and not locally generated)

routes?

With this in mind, I assume, if you had two leaned routes Ibgp in your BGP rib, and you then received a

redistributed route from eigrp, bgp has to recalulate the best path, and then pass that descison onto the routing

table process. Would that be correct? If so, is this process flow documented?

Sorry if its a crap question.

Best regards to all,

Ken.

[Harold Ritter]

A route created from distribution can not replace the routes it was originated from.

As for your second question, BGP would indeed recalculates it's best path an most likely choose the redistributed EIGRP routes because of the weight set to 32768.

Hope this helps,

[kfarrington]

Sorry for the clarification here,

So, any routing protocol, has multiple paths to a prefix, and selects the best prefix for installtion into the IP RIB. Then the router decides if there are multiple routing protocols that have a path to the same prefix, which one to install based on AD.

As BGP Design and Implementation Zhang and Bartell state on page 15:-

"After all valid paths are installed into the BGP RIB, the BGP router process runs the path selection

and installs the best paths into the IP RIB (the IP routing table)"

hence, when a route gets redist into BGP from an IGP, there must be some indicator that this route has been redist from another protocol into bgp, hence when BGP comes to run the best path calulation on its adj-rib-in, it is not considered being forwarded from the loc-rib, or ADJ-RIB-OUT back to the IP RIB (routing table) becuase it knows it has come from the routing table.

One would assume, that if this did happen (ie BGP decided on the best path to this prefix and tried to install it in the IP RIB), becuase the route is not learned from an eBGP peer, it would (in theory) be flagged as an iBGP route and again in theory be ribbed because of AD.

This part of the BGP descion i think is important to understand.

Does it make sense?

Thx

Ken

[Harold Ritter · Accepted solution]

It does make sense indeed.

The redistributed route is neither considered has iBGP or eBGP but rather as a locally originated route. IOS knows this route is redistributed from another routing protocol (or static) and will never try to install this route in the RIB.

Hope this helps,

[kfarrington]

Thx dude.

Top man :))

Ken