Using dynamic DNS and TTL as a poor man's redundancy

[Rick Kunkel]

Hello all,

I have a potential customer that has several machines hosting
multimedia in a fairly well situated NOC.  However, due to routine
maintenance or unforeseen events, his servers occasionally become
unreachable.  He want to put some older, junkier duplicates of these
servers in our NOC to fail over to in the even that his main site goes
down.  He doesn't want these servers hit during under normal operating
circumstances.  They're for failover only.

Anyhow, one of the guys here suggested a Dynamic DNS solution similar
to the way that no-ip.com does their stuff.  In essence, we'd be
running a dynamic DNS server with a very low TTL on the zone that
would receive heartbeat information from their main servers, and as
long as that hearbeat remained present, the server would point people
over to his main NOC.  However, in the event that the heartbeat
stopped, the dyanmic DNS server would call back to a default of the
servers hosted at our location.

My job is to pick this thing apart and try to to break it.  Does
anyone see problems with this as a solution?  Simply put, I feel that
it's got to have holes in it, or else people would be doing this all
over the place.  Heck, maybe they do, and I'm just out of it.

Potential problems I've already read about mainly revolve around DNS
servers overriding TTL values with their own, or browsers caching IPs.
(There seems to have been strong suspicions that AOL in particular has
overridden the TTL value on their servers.)  However, in most cases,
these complaints seem to be old.

So, is it realistically possible, and perhaps even advisable, to use
DNS in this fashion?  Somehow it just doesn't seem right to me, but I
don't necessarily trust myself, and need to know why, if at all..

Thanks,

Rick Kunkel