Problem with dnsservers

Hi!

I'm having problems with a new squid server i'm installing in a RedHat
6.0 box.

When I have the cache running for a significant ammount of time (more
than just a couple of minutes), the avg service time in the "dnsserver
statistics" page of Cachemgr.cgi gives me about 29000 msec (incredible
high value).

I've read all the messages on the list archives about dnsserver issues,
and I've tried to isolate the dns problem with no results. I don't have
congestion in my links, and dns queries, running "by-hand" the
dnsservers respond very quickly (1-2 seg).

Of course, i had to increase the number of dnsservers to 20, to avoid
filling up the queue and the squid process to abort.

Now, with 20 processes configured, I'm getting 11 of them used.

One strange thing i've noticed is that the dns.median_svc_time in the
last 60 minute average of counters is about 1 second, wich seems very
razonable for me... how are these values related?

What do you think?

Is there any way i could debug these situation?

The cache is running in a redhat 6.0 system, with 2.2SATBLE4 compiled in
it, 64 MB RAM, 800 MB swap directory and 50.000 acceses a day. The cache
is configured with one ICP parent.

Thanks a lot for your help.

Yours,
 

-- 
---------------< LINUX: The choice of a GNU generation >----------------
David Guerrero                                      E-mail: david@boe.es
Unidad Informatica                                    Telf: 91 384 16 43
B.O.E. (Boletin Oficial del Estado)         Ministerio de la Presidencia