[c-nsp] Setting "weight 255" as default for customer BGP with uRPF strict

[Pete Templin]

Brian Feeny wrote:
> 
> Your right, what I meant to say, was they advertised the same routes to 
> both isp's, but just
> prepended to one ISP more than the other, so that say, ISP A preffered 
> the route thru ISP B, rather than thru its own direct link with the 
> customer, since the customer may have prepended say 5 times to ISP A.  
> In that case, weight will fix strict uRPF.

Very true.  One of the notes in my original reply was a question about 
your selection of weight value.  255 looked to me like you were trying 
to set it to max, rather than a non-zero value.  Max is 65535, and it's 
good to note that locally-originated prefixes earn an automatic weight 
of 32768.

Probably not applicable in your situation, but possibly useful for 
others: I use BGP to carry customer prefixes, and therefore redistribute 
static into BGP (don't worry, I mark them all no-export).  Some of my 
customers are multi-homed with other providers.  We've assigned them 
some space from our aggregate, and since it's "our" space I statically 
route it to them with a high AD.  I cut two of these customers over 
Saturday morning, and realized that the redistributed routes came up 
first, and therefore the redistributed static route had a weight of 
32768.  The equivalent route learned from the customer had a weight of 
0, and lost.  Oops...but easy to fix.

pt