[c-nsp] bgp filtering to save memory question

[Bruce Pinsky]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Scott Granados wrote:
> I have an issue and wonder if anyone has any pointers.
> 
>  
> 
> I have a 7507 taking two full views.  I presently filter for routes on
> the /24 level but do to increasing table size and so on I wish to filter
> out /24's and say accept only /23's or /22's.
> 
>  
> 
> So, I've done the following, I have my learned routes match against a
> prefix list with a route map.  The list looks like
> 
>  
> 
> Ip prefix-list specific seq 5 0.0.0.0/0 le 24
> 
> I'll just change this to /22
> 
> Then add default routes and reset the inbound sessions with a soft in.
> 
>  
> 
> My question is this, when I do this I seem to not gain any memory
> savings.  Also, my number of network entries stays the same, about
> 188700 or so yet my installed routes match the filters.  Do I have to
> disable soft inbound to make this work?  What step have I forgotten?
> 
>  

If you have soft inbound configured, then you are keeping a copy of all
routes learned from your neighbors in the RIB-IN as well as the filtered
ones in the BGP-RIB.  By limiting to /22, you are only eliminating longer
mask prefixes from the BGP-RIB so you may not see a significant reduction.

Assuming your neighbors support route refresh capability, you don't need
soft inbound unless you have a need to see what routes you are receiving
from your neighbors prefiltered.

I assume, of course, you did a soft clear of the peering session once you
changed your filters.

- --
=========
bep

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.4 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFEu9u3E1XcgMgrtyYRAut5AKDS4DtAO06tlVJYgq4rOOTm0+6ZCQCdG0FX
NV+QUEWdQ/juF3vLn1BlMR0=
=sTrj
-----END PGP SIGNATURE-----