[47955] in North American Network Operators' Group
Re: route statistics
daemon@ATHENA.MIT.EDU (Stephane Bortzmeyer)
Sun May 19 08:02:57 2002
Date: Sun, 19 May 2002 13:00:04 +0100
From: Stephane Bortzmeyer <bortzmeyer@gitoyen.net>
To: Ralph Doncaster <ralph@istop.com>
Cc: nanog@merit.edu
Message-ID: <20020519120004.GA2355@laperouse.sources.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Pine.LNX.4.21.0205181858530.11751-100000@cpu1693.adsl.bellglobal.com>
Errors-To: owner-nanog-outgoing@merit.edu
On Sat, May 18, 2002 at 07:02:58PM -0400,
Ralph Doncaster <ralph@istop.com> wrote
a message of 10 lines which said:
> I'm trying to collect statistics on how many routes match certain
> patterns. So far I've been using zebra, set term len 0, and then sh ip
> bgp regexp, and wait for the total prefixes count at the end of the list.
> I figure there must be a better way than this, but so far haven't found
> one. Any ideas?
Compile zebra with --enable-snmp (the Debian binary package just
switched on this option) and snmpwalk the BGP table (1.3.6.1.2.1.15 ==
mib-2.15, see RFC 1657) ?
I didn't benchmark the two solutions against each other. If the BGP
machine is an actual forwarding router, not just a dedicated looking
glass, be sure to look at its load, not just at the wall-clock
response time.
Another solution is to dump the routing table
<URL:http://manticore.2y.net/doc/zebra/bgpd.html#Dump BGP packet and
table> in MRT format and to use MRT tools to analyze it (I tried that
and at least the Python version of these tools is hopelessly broken).
