DNS Redundancy After a Disaster

[Tim Wilde]

On Mon, 23 Jan 2006, Martin McCormick wrote:

> 	One of the beautiful things about the DNS protocol is that one
> should have one or more slaves listed along with the master.  Other
> than not being able to dynamically update one's zones, how bad does
> service get if the master is dead and the only working DNS is an
> off-site slave?

Most DNS resolvers (particularly BIND) have algorithms designed to 
determine the closest servers, network-wise, and also ignore dead servers 
for some time period.  There would be some initial latency the first 
time(s) a given resolver is trying to resolve something in your domain, 
but it shouldn't be a huge impact.  You can lessen the impact by having 
multiple off-site slaves; many companies (including my own) offer 
secondary DNS services that will provide you with multiple off-site 
servers for low (and in some cases, no) cost.

Google for [secondary dns] or [secondary dns service], you'll find lots of 
different options out there.

> 	Do large companies do anything special to insure that if a
> master DNS fails, there will always be something at the IP address of
> the master?

Many people will use anycast or other technological solutions to ensure 
that their master is always available.  This only really impacts the 
slaves, though, as any random resolver out on the Internet has no way of 
knowing your master from your slaves when they're just doing a normal 
query.  They just see a list of nameservers, and pick one using their 
internal selection algorithms.

An outside secondary DNS service is a great way to ensure your domain 
remains up and running if your primary network connection is down, or your 
primary DNS server(s) fail.

Regards,
Tim Wilde

-- 
Tim Wilde
twilde at dyndns.com
Systems Administrator
Dynamic Network Services, Inc.
http://www.dyndns.com/