Hi all!

Please excuse the long post - I wanted to add some background to explain my position better.

A seemingly long time ago I ran qpsmtpd in front of my qmail servers (actually I have a couple still running old 0.10 code out there). For various configuration needs we moved from qmail to postfix, and hence at the time a postfix -> amavisd-new -> postfix environment. This seemed to give a good combination of features and performance and fit the bill quite nicely for a long period of time.

As many of you many or may not know - amavisd-new is a heavy process, we found that on a dual xeon server 10 child processes seemed to be the optimal configuration for the spam/virus scanning. All works well until a glut of email is delivered at one time which Postfix gladly accepts and queueus and slowly trickles through to amavisd-new. This and some other behavior has absolutely convinced me that it is time to move to a mail service that can function as the mx receiver and spam/virus filtering system at the same time (this way if one system is "busy" the remote system will resend to one of our many alternate mx systems).

This has led me back to qpsmtpd and this post - namely some integration questions based on our current configuration.

- Currently we use postfix to lookup/validate the recipient domain and in some cases the recipient user before accepting the message. This is a direct query to mysql - is this possible with qpsmtpd (without writing plugins)?

- With amavisd-new we are customizing the spam/virus thresholds per recipient domain and in some cases the user using a mysql lookup - is this possible with qpsmtpd?

- Can I quarantine messages marked as spam/virus directly within qpsmtpd?

We have a very high volume mail gateway (several million emails per day) , as I can tell I want to run the Apache::QPSMTP module for increased performance. I found a note about a multiplexing server...

"For very high concurrency, you can use it as a multiplexing server. Two large antispam companies' high-traffic spam traps have used this effectively since the second quarter of 2005, delivering concurrency as high as 10,000 on some occasions."

Is this documented somewhere that I can read more on?

Thanks in advance,
Max

Reply via email to