[c-nsp] unknown unicast flooding - particularly regarding fhrp's

[Aaron]

What do y'all know about the effects of implementing fhrp's (glbp, hsrp,
vrrp) WITH route diversity from the distribution (fhrp router) to the
internet. (which I'd imagine is a pretty typical scenario in HA nets)

 

I mean as packets arrive from the internet to the non-active fhrp router,
then this router probably won't have arp entries (perhaps at 4 hour timeouts
it will) but it more than likely won't have bridge table entries, nor will
the L2 distribution / access devices have bridge table entries (at 300 secs
aging probably not)

 

How does constant unknown unicast flooding affect networks?  Better yet, how
to design in mitigation ?  is it all about lower arp timeouts below 300 secs
so to artificially prop-up bridge tables and keep them fresh?  My goodness
that's making arp very busy.

 

This is also being asked since I'm suspecting this behavior on my asr9k's
via their bvi's (hsrp'd) since they have separate internet uplinks and I'm
suspecting unknown unicast flooding from the non-active hsrp asr9k over the
vpls domain towards customers.  (but ugh, my dual 7609's over my legacy net
have been running like this forever!)

 

Aaron