[dns-operations] Reverse Lookup speedup

Edward Lewis Ed.Lewis at neustar.biz
Thu Aug 2 12:15:55 UTC 2007

At 10:50 +0200 8/2/07, Michael Monnerie wrote:

>This is for sure valid for a running bind, but after a restart all
>caches are emptied, right? We're currently in a big reconfiguration, so
>bind restarts happen more often than I would want. Having
>the .in-addr.arpa zone locally saves at least the 1st level lookups,
>thus helping after a restart, or isn't that worth the simple zone copy?

I don't mean to question neither your requirements nor the solution 
being engineered, but before thinking about slaving the reverse map 
zones and the ensuing management overhead needed to do that my 
recommendation would be to examine why there are "restarts."

Name servers should not have to suffer from restarts.  Notice I said 
"should not" - I don't mean to say that that is the way it has to be. 
But most of the considered operation experience and opinion you will 
run into on a list like this will assume the name servers operate 
consistently and continually.  (And not just operating BIND, name 
service in general.)

The argument that "why slave since you cache" is a potent one but 
assumes that you have a name server operating in an environment in 
which it is not restarting and losing it's cache memory.

Perhaps you need to restart, so your reply (if you choose to) could 
be so and that would lend credence to wanting to slave locally. 
However, from my experience I would recommend that you first engineer 
a way around the need to restart as that is probably causing a lot 
more instability than latency resolving queries against the reverse 
map zones.
Edward Lewis                                                +1-571-434-5468

Think glocally.  Act confused.

More information about the dns-operations mailing list