BIND 8.4.5 as cache server, long delay to resolve some domains...

[rene mathis]

I have the same problem with BIND 9.3.0.

Before I was using BIND 9.2.1 (and 9.2.3) and there the Name Server was
sometimes waiting for 2 seconds before sending a recursive query to another
nameserver on the internet (not only root servers).

Since I am using BIND 9.3.0 I have a lot of "Lame Servers" logged (maybe one
per 500 queries). 


On Wed, 26 Jan 2005 14:53:02 +0100, Jacques-André Petit wrote
> Thank You, Ronan, for your response
> 
> Ronan Flood wrote:
> 
> >
> >
> >That's exactly what it does: it drops the request and hopes to have
> >the information ready by the time the client asks again.
> >
> >  
> >
> >>Once resolved (ns.another_domain.com), BIND stops and never resolves 
> >>www.my_domain.net
> >>
> >>Then, after about 5 seconds, there is a new request from the client 
> >>(timeout) and BIND can resolve further as it then knows how to resolve 
> >>ns.another_domain.com.
> >>
> >>A realworld example would be www.isfc.org  (dig response pasted below)
> >>
> >>We suppose that this situation is not RFC compliant (where NS pointers 
> >>must belong to the same root as the domain they manage), and that this 
> >>    
> >>
> >
> >Which RFC?
> >  
> >
> Actually, in Belgium, to request a domain name, your name server 
> MUST have a .be  A Record. This avoid such kind of trouble. So I 
> suppose there is a internet directive to act that way.
> 
> >>Is there a way to fix that behavior?
> >>    
> >>
> >
> >Install BIND9 ...
> >  
> >
> Does BIND9 solve this problem? We tried BIND9 on different servers 
> and had performance problems (there are currently 35.000 users on 
> the network) and were suggested to downgrade to BIND8 to get better results.
> As long as cache fills up, the problem remains but after a while, it 
> disappear nearly, with some customers still having delays. It is 
> these delays we try to put out of the scene.


--
Rene Mathis
rene at solosaina.ch