My DNS not always available on the web

[Barry Margolin]

In article <7o6gtu$85t$1 at nnrp1.deja.com>,
Dan Winchester  <andy.stowell at virgin.net> wrote:
>Hi,
>
>Everyday I seem to get two or three emails from people saying that when
>they try and access my site (http://www.freelancers.net) they get a DNS
>error. Both DNS servers are based here in my office (as is the
>webserver) so I know they are always up as I never have to restart them.
>
>Ok, so what I generally do is try and track down what DNS they would be
>using based on their email address and query it for my domain. More
>often than not my query seems to time out with a 'no DNS for host
>freelancers.net' type message. But then if I run the same query again it
>resolves immediately.
>
>So, is the problem here that my DNS is just too slow? If so how can I
>speed it up? Here are the name servers:

I think the problem may be that there are many levels of indirection
involved.  Here's a typical scenario when the user's server doesn't have
anything related to your domain or ISP in its cache (I'll assume that it
does have common TLD and country domain info cached, since that's typical).

It queries a root/gtld server for www.freelancers.net.  The server returns
a referral:

FREELANCERS.net.	172800	NS	NS1.FUNKTION.CO.UK.
FREELANCERS.net.	172800	NS	NS2.FUNKTION.CO.UK.

Now it needs to look up ns1.funktion.co.uk or ns2.funktion.co.uk.  Most
likely it will have the .CO.UK NS records in its cache (per my assumption
above), so it doesn't need to start at the root servers again.  It will
pick one of the .CO.UK servers and query it.  This will return:

funktion.co.uk.	14400	NS	ns1.bb-online.net.
funktion.co.uk.	14400	NS	ns2.bb-online.net.

Now it has to look up ns1.bb-online.net or ns2.bb-online.net.  It will
query one of the root/gtld servers.  This time it gets an answer rather
than a referral (I'm not sure why -- these don't seem to be necessary glue
records):

ns1.bb-online.net.	172800	A	194.168.163.59

Now it queries this server and gets:

ns1.funktion.co.uk.	86400	A	212.19.75.168

and then it finally gets to send the original query to this server, which
responds with the answer.

All these queries take time, and if it has to go through all of them the
querier may time out, especially if there's some network congestion that
causes packets to be lost, requiring one or more to be retried (from my
desktop in Massachusetts most of these queries took 100-200 ms, so the
total time should be around .5-.7 seconds if there are no errors).

-- 
Barry Margolin, barmar at bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.