[c-nsp] bgp holdtime adjustment

[John Elliot]

> 
> The mechanism you want to investigate is BFD to check if the BGP neighbor 
> is reachable, not lowering BGP hold timers. BFD might give you anything 
> from 100 ms to 3 seconds of detection time depending on platform, changing 
> BGP hold timers to less than ~10 seconds is usually unreliable.
> 

Thanks for the info - Is it necessary for the Upstream to also configure bfd on their side?
 
I would have something like this under L3 Interface to Upstream?:
 
bfd interval 100 min_rx 100 multiplier 3
 
And under bgp:
 
neighbor UPSTREAM fall-over bfd
 
Cheers.