(no subject)

[Kevin Darcy]

aaron.g at hushmail.com wrote:

> We recently experienced an Internet outage for an extended amount of time
> due to problems with our ISP. About 3 hours after the outage began, and
> before it was restored, performance of the name servers became degraded.
> The named process was using lots more cpu than normal and resolving names
> was very slow for the domains the servers are authoritative for.
>
> I assume that their were lots of requests that went out and no responses
> came back during this outage, and that was causing the degradation in performance.
>
> In such a case, is it advisable to implement your own root cache (create
> your own root name server) , with only your domains delegated until the
> Internet connection is restored?

I think I've suggested something like that in the past.

If you're using MX records to route mail, however, you may wish to add some wildcard
MX records to your "disconnected-mode" root zone, so that mail won't bounce with
"host unknown" errors.

In fact, you may want to use wildcard A records too, so that clients for all
protocols will get a "host unreachable", temporary kind of error, instead of a fatal
"host unknown" error. For protocols like HTTP, you could even direct browsers to a
"service temporarily unavailable" web page. Watch out for proxy caching, though: once
service is restored, you don't want the "unavailable" web page still being served out
of the proxy's cache, so you may want to purge it.


- Kevin