ndc stats problem on busy DNS servers

[Alexis Oosterhoff]

I am developing an in-house system for monitoring DNS caching server within
our organisation. We're running  8.2.3-REL on a Solaris 2.6 system.

I have 9 servers which are dumping their stats to named.stats on an hourly
basis, using a call to "ndc stats". Most of the machines are dumping the
stats reliably, however, on two of our busier caching servers (approx. 400
lookups/sec), "ndc stats" works some times and not others.

Bind is running as follows:

    named -u nobody -g nobody -v /etc/named.conf

and ndc access is configured with the following controls statement:

    controls { unix "/var/named/ndc.d/ndc" perm 0774 owner 0 group
600001; };

The error message being displayed when the "ndc stats" call fails is:

    cannot open stat file, "named.stats"

The "named.stats" file is created, owned by user "nobody" but with null file
size.

If this were happening on every call to "ndc stats", I'd say it was a
configuration problem. However, as it's happening irregularly (no
discernable pattern, maybe works 30% of the time), I'm wondering if there
are issues related to system sizing (though the machine is running with dual
300MHz CPUs and 2GB RAM), or in running in a chown-ed environment.


Any thoughts would be appreciated.

Alexis Oosterhoff