cache problems

[Kevin Darcy]

joe wrote:

> Hello,
>
>  I run Bind 9.2.1 on RedHat 7.2. I've noticed that recently I will
> do an nslookup for "www.food.com" and I get a reply. On occasion
> I will try that same domain 12 sometimes 24 hours later and get
> a timeout. I looked at the SOA for food.com and noted the refresh
> was set for 1 hour and the min TTL was set for 5 minutes. Is this
> the reason why I get timeouts when performing a lookup?

It doesn't help that the zone sets (apparently) the TTL for all of its
records to 5 minutes.

Even worse, one of the nameservers for the food.com zone
(ns2.an.food.com) appears to be timing out queries.

The domain needs some work...

> My co-worker states the our DNS is broken on account that
> lookups from time to time are failing. I read some posts on ISC
> regarding this issue and I've explained to him that the windows
> resolver is set for a 2 second timeout with one retry. Would I
> be out of line telling him to set the value in nslookup as follows:
>
> set timeout=5
> set retry=4
>
> I also tell him to use the FQDN as well. Is this a valid response
> or am I really having problems?

I think food.com itself is having problems. The steps you've mentioned
are just workarounds for that.

> Also am I correct in that I have no control over non-authoratative
> data that is stored in my cache, that whatever the TTL is set to
> controls how long that information stays in my cache? Am I correct
> in assuming that I cannot alter the duration an object stays in my
> cache?

In BIND 9, there is a max-cache-ttl option which puts a limit on how
long cache entries persist.

But, there is nothing that makes cache entries stick around beyond their
expiration. To do so would go against a fundamental principle upon which
DNS is based, i.e. that the domain owner/maintainer is the one that
controls maximum cache persistence.

I'd suggest contacting the domain admins of food.com.


- Kevin