understanding ip bgp summary state prefix

[williammanurung]

Hi expert,

can you explain what exactly of the value 146 state/pfxrcd from show ip bgp summary?

I know there is number of receiving prefix from neighbor 10.251.1.62, but its not same value of total number of prefixes .

And if you said the value 146 is the number network on routing table, its not true.

Please explain to me what is the value 146 and how i confirm that value on routing table?

Thanks

[Georg Pauwen]

Both values should match. Which platform is this on, and which IOS are you using ?

[williammanurung]

Cisco Router 3945 , Version IOS 15.4(3)M1.

[Georg Pauwen]

It actually looks more like a bug such as the one below. Version 15.4 and 15.5 are affected, try and upgrade to a 15.6 release...

Incorrect prefix count upon clearing bgp peering
CSCuy03504
Description
Symptom:
The number of prefixes received from a peer displayed in "show ip bgp summary" is more than the actual number of prefixes received from that peer.

Conditions:
Graceful restart is enabled.
An event on the router causes the session to go down.
Peer sends router a TCP FIN before the router sends out a TCP FIN.

Workaround:
No workaround

Further Problem Description:
The "event" seen in this DDTS is issuing a "clear ip bgp "
This behavior is more likely to be seen with a higher number of prefixes on the router.

[Julio E. Moisa]

Hi

As Georg mentioned, it usually is the same value, have you executed: clear ip bgp * soft, previously?

[williammanurung]

My collogues said :

146 is the total number of prefixes received from the neighbor "after" the inbound policy is applied. 214 is the total number of prefixes "before" the inbound policy is applied. This means that this router has rejected 68 prefixes coming from 10.251.1.62 from being installed in its BGP table. This also means that you have soft-reconfiguration inbound configured which is why your router is able to store a copy of "unfiltered" routes from 10.251.1.62.



146 is the total number of prefixes from 10.251.1.62 installed in the BGP table. However, what is in the BGP table doesn't always get installed in the RIB. There are a couple of reasons for this to happen such as a route with a better AD (IGP/static/connected) is already in the RIB, or your router can't recurse to the next-hop for the prefix, or memory failure if your router has exceeded the total number of routes that it can handle.

[Georg Pauwen]

Hello,

actually, I tested that exact same thing, but before and after applying an inbound route map filtering, both values still match, so i don't think that is the reason...

[Julio E. Moisa]

Hi

Are you using any kind of filtering method (prefix list, route-map, etc)? It will show you that behavior,

The show ip bgp summary will show you the allowed Prefixes to be received only.

The show ip bgp neighbor x.x.x.x received-routes will show you the total of the received prefixes, but you will see the symbol (>) for the allowed networks in your filter method.

To avoid this behavior and extra processing, you could configure BGP ORF if you are filtering with prefix list. 

Hope it is useful

:-)