by Nate Campi

I work for an internet portal, and once I was tasked with reducing the load on an overwhelmed BIND8 nameserver that was both an authoritative server for our public domains and a cache for internal hosts. I wanted to know the rate at which queries came into the nameserver over a period of time, as well as what types of queries they received.

I took some in-house scripts we developed for graphing memory/CPU/disk stats, and modified them to dump BIND stats, parse them and insert them into a local RRD. I also wrote a simple CGI to display the graphs. Before long I decided I wanted to graph the stats on all nameserver hosts, but to collect from a central point. I modified the procedure to collect the stats over SNMP, and enhanced the CGI to show short and long-term stats, and wrote client-side scripts for DNS servers running BIND 9, dnscache and tinydns.

For those interested, the scripts and usage instructions are here