rfd

adamv0025 at netconsultings.com adamv0025 at netconsultings.com
Thu Dec 27 16:33:16 UTC 2018


> Randy Bush
> Sent: Tuesday, December 18, 2018 5:40 PM
> 
> do you have rfd on?  with what parms?
> 
> randy

If I remember correctly the industry was back and forth on this several
times now.
First it was deemed good then some studies came out proving the penalty is
worse than the crime couple years later another study came out suggesting
that if correct parameters are used it should be alright, but I guess at
that time no one could have cared less already switching it on and off and
on again...

With regards to the comments made here on the number of unstable routes till
the whole system or significant parts collapse, I could easily revert that
argument and ask how many badly configured rfd till the whole system
shuts/dampens itself down... (positive vs negative feedback loop) I guess
the ideal solution is somewhere in between.

Personally I think rfd is just the aspirin, i.e. not treating the cause -but
merely helping with the headaches. 
And I suspect that Interface State Dampening would address 80% of the
route-flaps out there (it works exactly like rfd but treats the cause).
With the reminder being true protocol flaps either by misconfiguration of
max prefix limit (sessions should stay down) or BGP error handling -which
again can be solved by the enhanced BGP error handling or genuine bugs. 

adam  




More information about the NANOG mailing list