Highly Available DNS?

[paul at anastrophe.com]

edlewis at my-deja.com wrote:
> Customer has single DNS server and wishes to take redundancy measures.
> Customer has 3000 clients all pointing to this single DNS server IP
> address and is not willing to make DNS configuration modifications on
> the client side.

> Options:

> 1) IBM eNetwork Dispatcher to direct requests to primary DNS server
> unless it fails, in which case requests directed to standby.

> 2) Clustered DNS servers in hot-standby configuration.

> 3) "Poor man's HA" scripted solution that may involve manual
> intervention.

My setup: BIG/ip HA (dual BIG/ip chassis w/failover). dual sun e4500's
behind the BIG/ip's. Both e4500a and e4500b are running bind, and are
set up as primary masters (zone updates from the 'real' master -
e4500a - are done via scp to e4500b). The BIG/ip's have two virtual IP's
set up, one for ns1, and one for ns2, *both of them* pointing to both
nameservers, thus:

VIP +------> 206.58.250.53    
    |
    +---+--> PORT 53                 
        |
        NODE 192.168.1.23:53        
        |
        NODE 192.168.1.27:53        

VIP +------> 206.58.250.54    
    |
    +---+--> PORT 53                 
        |
        NODE 192.168.1.23:53        
        |
        NODE 192.168.1.27:53        

This provides near perfect availability and reliability. Lookups are
load balanced between the nameservers, and if one of them is offline
for maintenance, the BIG/ip just pushes all the requests to the other
one - whether the request is coming in for ns1 or ns2.

This is an expensive solution if done only for nameservice. I host
most major services on the same cluster - www, ftp, smtp, pop3, etc.
which makes it very cost effective.

-- 
Paul Theodoropoulos                       Advanced TelCom Group, Inc.
Senior UNIX Systems Administrator                   Internet Services
Work: http://www.atgi.net             Play: http://www.anastrophe.com 
=Opinions are my own of course, not necessarily those of my employer=