Thoughts on Barracudas?

[Aaron K. Moore]

Martin Hepworth wrote:
> You need to maintain the SA side of things, make sure you have
> RulesDuJour etc installed and picking up the SARE and other rules.
> 
> Spam is an evolving thing and therefore you need to spend a little
> time making sure you system can auto-update. It's not a
> install/forget system - I think this way of thinking is another
> example of M$ skewing the market, you don't need all  those expensive
> well trained/experienced sys-admins!

I agree.  Moving SpamAssassin to using the MySQL backend for bayes and
the awl really increased our performance here.

I've had to do some major tweaking of sendmail here to deal with our
ever increasing amount of spam.  About 8 months ago I added milter-ahead
and milter-error to sendmail.

I converted some code that I had written to create a dynamic blacklist
based on spam assassin scores by adding them to the access file to
generate files for rbldnsd. And I added a couple of other rbls to mix
for good measure.

I also had to adjust the following sendmail settings, as our inbound
mail server, on which I run MailScanner, was getting swamped with
inbound connection attempts to the point the machine crashed.

confMAX_RCPTS_PER_MESSAGE
confBAD_RCPT_THROTTLE
confCONNECTION_RATE_THROTTLE
confQUEUE_LA
confREFUSE_LA
confDELAY_LA

Dealing with spam, regardless of what some people think, is not a set
and forget situation.  And there isn't a one size fits all solution.
MailScanner is great and works well out of a box.  But no amount of
automation can make up for actually getting to know the components of
the system, how they work, and how they interact.  If they were getting
bad performance out of MailScanner, there's a good chance they needed to
do a lot more tweaking of their setup.


-- 
Aaron Kent Moore
Information Technology Services
DeKalb Memorial Hospital, Inc.
Auburn, IN
E-mail:  amoore at dekalbmemorial.com