All nameservers unresponsive when master is down

[Simon Waters]

Sam Pointer wrote:
> 
> Simon, sorry I do have a $TTL statement - I just omitted it in this email
> exerpt.
> 
> I want a very quick failover, would you suggest 5 as a minimum TTL?

This value is just how long remote servers can cache the data
safely, it doesn't affect how quickly remote servers will retry
other name servers if one is down.

In most cases a value of 3600 for nameservers will be adequate,
beyond that your balancing the problems caused by loss of
connectivity/DNS service, by the problems of switching IP
addresses at short notice.

I usually use 86400 on DNS records I'm not expecting to have to
change at short notice.

Values below about 15 seconds mean that data can expire from
cache whilst it is being used to look up other data. Could be
your problem, what data is actually being looked up as the
domain seems largely to consist of name servers...