dynmamic DNS failover

[Kevin Darcy]

If you don't mind constant spill-over to your backup webserver, you could
associate *both* A records with the name, and configure your master and all of
your slaves with a "fixed" rrset-order. You'll still get spill-over because
when other caching nameservers give out the name, they'll generally
"round-robin" the answers. You can alleviate that somewhat with a short TTL.
The real beauty of this approach is that most clients these days should be
smart enough to *automatically* failover to the backup webserver.

The long-term solution though is for clients to use SRV records. See
http://bugzilla.mozilla.org/show_bug.cgi?id=14328 for an attempt to get
SRV support included in Mozilla.


- Kevin

Evgeni Dobrev wrote:

> Hello,
>
> I am trying to do something which I never tried before.
>
> I wan to establish a backup machine which is offsite.  The backupp machine
> is running a watchdog process which checks whether the HTTP server we host
> is available. if it is not it updates the DNS servers for the domain and a
> new A record is created for the HTTP server. and all http requests go to an
> alternative machine.
>
> My question is : How can I bring the down time to minimum? when I update
> the DNS it will take some time for the changes to propagate. I have set the
> TTL in our zone files to 3 mins but I am not sure whether this is all the
> downtime I will get?
> Are there better solution to what I want to do?
>
> Anyone please, help me out with this one. Any help is highly appreciated
>
> evgeni