https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6075


Elsa Andrés <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #26 from Elsa Andrés <[email protected]>  2009-03-03 05:38:39 
PST ---
(In reply to comment #24)

> The log looks perfect, you got DNS responses to all of your 25 DNS queries,
> all of them within 0.8 seconds of sending a query.

Alright then, sirs. Closing this as "invalid" and sorry for the noise. Never
seen this before :-)

> This is a consequence of two things:

> - in your /etc/resolv.conf you are telling your resolver to try appending
>   one or more fields to the domain name if the attempt with original
>   name is unsuccessful. Get rid of your existing 'domain' and 'search'
>   directives in that file, and supply just a 'search' with no arguments.
>   Or at least do not specify something as vague as 'net' for your domain!
>   (according to your tcpdump in #12), but use your exact domain name in a
>   'search' option, and do not use a 'domain' option.

Indeed, that seems to be the culprit. After commented out that sentence
("search net") it returned the expected response (nxdomain).

> - the 'org.net' domain you have queried (appended .net to your ...njabl.org)
>   is registered by Cyberfusion / OKDIRECT.COM.  Such domains attempt to
>   fish for traffic from innocent users which make a typo when enetering an
>   URL, or sites like yours, which append '.net' to unsuccessful queries.
>   Gratuitous advertising. Don't let them have your mistyped queries!

> Btw, I have another hypothesis why your original queries to ISP may be 
> failing.
> You have 2 Gbps links from your host - but do you have gigabit connectivity
> all the way to your service provider? If your uplink is thin, you should have
> some traffic shaping with reasonable buffer size on your device sitting just
> before a bottleneck link - usually a firewall or a router. If this is some
> dumb device, your queries may have been victims of tail drop in a switch.
> Note that SpamAssassin made 29 queries in rapid succession, all within 22
> milliseconds. A solution in such scenario would be to turn on traffic
> shaper on firewall/router or on your host where SpamAssassin is running.
> But never mind, the solution with a caching name server within your
> own network is much better.

It makes sense. 

I had in mind to perform a "silly" test I was thinking on these past days: to
connect the host with a dial-up modem and see what happens, just to eliminate 2
things, a) dsl router and b) dsl line themselves. This server is running over
static ip address so I am binded to it. By using a dial-up modem (just for
testing purposes) I would get interesting results, in a way or another.

I don't mind to have a "named" service on this machine if just SA (or any other
service) need it (indeed, bind9 is a good partner), just wanted to be sure
where the problem was.

Thanks to all of you for helping me to debug this.


-- 
Configure bugmail: 
https://issues.apache.org/SpamAssassin/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

Reply via email to