root zone A queries

Barry Margolin barmar at alum.mit.edu
Fri Aug 19 01:37:44 UTC 2005


In article <de1um9$27rl$1 at sf1.isc.org>,
 Chris Thompson <cet1 at hermes.cam.ac.uk> wrote:

> On Aug 17 2005, Pawel Rogocz wrote:
> >> I have a number of BIND9 servers behind load balancer.
> >> The load balancer does health checks by sending queries for "." of type A.
> >> How soon, after a box looses its connection to the internet and no root
> >> servers can be reached, will these queries start to fail ?
> >
> >I do not know what I was thinking , the answer is obvious:
> >It depends on the value of max-ncache-ttl ( default 10800 secs )
> 
> Yes - capped at the value of SOA.minimum for the root zone (86400 secs).
> 
> I do wonder why this particular query is used as a "health check", though.
> With nameserver software that didn't implement RFC2308-style negative
> caching, it would go through to the real root servers. On the other hand,
> maybe that's the intention.

If all it's trying to determine is whether the server is up, it doesn't 
really matter whether the answer comes from cache or causes recursion.  
It just wants to see whether the server answers.

-- 
Barry Margolin, barmar at alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***



More information about the bind-users mailing list