high availability configuration error?
kcd at daimlerchrysler.com
Tue Feb 27 22:05:22 UTC 2007
Marco Strullato wrote:
> Hi all!
> what do you think about the following configuration for high
> availability of web servers?
> In the domain zone the filed www has "nslp" as name server and nslp
> has 2 A field
> www NS nslp
> nslp A ip1
> A ip2
> so if a client tries to resolve www and the ip1 is down, the the
> client should use the next ip2.
> Unluckly this configuration does not work because the client uses only
> ip1 and never ip2.
What kind of client are you talking about? A *DNS* client, i.e. a
resolver of some kind? Typically it will try ip1 and/or ip2 in a
particular sequence depending on how fast each has responded to DNS
queries in the past. But a DNS client only resolves DNS queries. This
has only an indirect relationship to what a *web* client will do once
the query is resolved. The address(es) that are seen by the *web* client
for the "www" name would be the A records defined at the apex of the
"www" zone, the contents of which you have _not_ shown us. If you've
only defined ip1 at the apex, then that's all the web client will ever
see, and this is totally expected behavior.
> Do you have suggestions?
Yeah, don't try to use DNS alone as a cheap way to do high-availability.
It was never built for that. Try Googling for "website high
availability" and I think you'll get a number of useful hits to put you
in the right direction.
More information about the bind-users