> This is good stuff. Most of these solutions have ben employed up to this
> point. It is good to see that most are using BGP for this. I have some
> tricky requirements to get around, but wanted to see what everyone else is
> doing before I went ahead down a path that may lead to a re-design.
Alternatively, if you can't employ associated "network" statements
(presumably as a result of some non-intuitive address allocation) that
encompass one or more of the static routes, redistributing is likely
the most feasible approach.
As for associating a community with the prefix to suppress propagation
to external BGP peers, I prefer a slightly differ model. The model
I prefer actually only permits external advertisement of explicitly
permitted communities and denies all others. This provides a number
of benefits over implicit announce models. Also, associating some
community value with POP aggregates or other regional aggregates to
alleviate some of the brokennes of BGP MEDs (w/aggregation) may be of
benefit down the road (though this opens another entire can of worms).
The problem with using "aggregate-address" stuff is that if things are
badlt allocated already, you may inadvertently break things by auto-
creating aggregates. I've seen this happen a few times.
HTH,
-danny
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:17 EDT