No delay to client when the primary DNS server goes down?
kcd at daimlerchrysler.com
Mon Feb 20 23:29:40 UTC 2006
nick.seccombe at gmail.com wrote:
>Is it possible to get a DNS client to automatically start using its
>secondary nameserver when its primary nameserver goes down - the way a
>caching-only DNS server dynamically starts using a different forwarder
>to avoid repeated timeouts/delays to queries.
>All I can think of at present is by reducing the timeout to 1 second in
>a client's resolve.conf to reduce the impact, or clustering the DNS
>But a client side configuration would be nice.
Trying to make stub resolvers this smart is like trying to teach a pig
to sing. Stub resolvers will always have severely limited knowledge of
how to resolve DNS, severely limited options for
failover/retry/fail-forward, etc. There's only so much you can do within
the stub-resolver framework.
If one really cares about redundancy, availability, etc. then put
full-service iterative resolvers on one's clients (e.g. the Wintel port
of BIND) and be done with it. If *necessary*, a given iterative-resolver
instance could be configured in a "forwarding" mode, which would dumb
down its behavior to more-or-less mimic that of a stub resolver, but
with integrated caching capability, and a better forwarder-selection
algorithm, much as you described. (In all fairness, I should mention
that I have never actually set up and run the Wintel port of BIND, but
that's more a reflection of my unfamiliarity/hostility with respect to
Wintel platforms, than the putative benefits of running "real" DNS
software whenever and wherever possible).
More information about the bind-users