How does load balancing operate on >1 forwarders

[Cathy Almond]

A long time ago it used to be in turn, but all current versions of BIND
sort the forwarders based on a preference value (SRTT) that's derived
from the RTT of previous query/query response interactions, with a 'time
since we last tried this server' incorporated so that servers that
aren't top of the preference list are periodically re-used.  It also
means that if a server becomes unavailable, it gets time-penalised and
therefore the others of the group will be used instead until the penalty
has decreased over time - at which point, if it's back and running once
more then it's going to be selected (or not) as before on 'nearness'.

You can see the SRTT value of nameservers in the ADB section of the
cache dump (from rndc dumpdb).  Smaller values are preferred.

What version are you using?


Jonathan Reed wrote:
> I have the forwarders statement to fwd queries to a few DNS servers on my
> LAN.
> forwarders { 10.0.0.1;
>                        10.0.0.2;
>                        10.0.0.3; }
> The bind documentation says that these fwders are queried "in turn", but
> what exactly does that mean? I understand it to mean that they are not round
> robined and if the answer is found from the first IP then it stops there and
> returns the query to the client. But assume that .1 goes unreachable. What
> is the timeout used to query the next forwarder in the list? And is this
> timeout modifiable?
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> bind-users mailing list
> bind-users at lists.isc.org
> https://lists.isc.org/mailman/listinfo/bind-users