[quagga-dev 5254] BGP memory utilization

Scher, Dave dscher at mitre.org
Tue Feb 5 19:14:53 GMT 2008


Hi all,
    I am running bgpd on a PPC board with a linux kernel 2.6.15.  bgpd
is configured to peer with two external neighbors in different ASs,
each of which advertise 50K unique best path routes.  Upon convergence,
I have the two external peers (actually two simulated peers using
Spirent TestCenter TeraMetric Routing suite) withdraw all routes.
During this entire trial run, memory utilization linearly increases
from approximately 0 at start to x% percent utilization when bgpd
completes the route exchange with its two peers.  However, utilization
remains at x% for the remainder of the trial where the output of show
ip bgp does not display any routes.  Is this a protocol feature
(perhaps flap dampening) that will keep the routes as stale in the
adj-rib-in data structs but remove them from the global RIB?  Also I am
trying to verify that bgpd does NOT support Graceful Restart is this
true?
 
v/r
Dave
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quagga.net/pipermail/quagga-dev/attachments/20080205/b206e13f/attachment-0001.html>


More information about the Quagga-dev mailing list