bind forward to rbldns server intermittently fails

[David Nolan]

--On Friday, November 17, 2006 15:26:31 +0530 Ramprasad <ram at netcore.co.in> 
wrote:

<snip>
>  forward first;
<snip>
> Everything seems to work perfectly fine .. but sometimes all of sudden
> the named server seems to have failed When I "dig" for a listed ip I get
> a negative response. I restart bind and then for the same ip I get
> positive response
>

I believe you want 'forward only' not 'forward first'.  'forward first' 
means "if forwarded query fails, perform normal recursion", 'forward only' 
means "never recurse, if the forwarded query fails return failure"

I've seen the same behavior from a similar setup here.  What happens is 
that at some point BIND gets a timeout from rbldns, and falls back to 
performing normal recursion, which gets a negative response from some 
server, and that response is then subject to standard negative caching.  So 
the next query bumps into the negative entry in the cache and is aborted 
without being forwarded to your rbldns server.

-David Nolan
 Network Software Designer
 Computing Services
 Carnegie Mellon University