Scalability issues with maRBL (the policy server for Postfix)?

[Brian Wong <bw...@gmail.com>]

I have been using maRBL v1.1 for two weeks now and the results of
selective greylisting based on RBL and p0f is impressive. Although I
am happy with the results, from my experience I do not believe that it
scales well.

Anywhere between 1 to 3 days of running marbl my Postfix logs complain:

Dec  6 11:39:19 <hostname> postfix/smtpd[20015]: warning: timeout on
127.0.0.1:2552 while reading input attribute name
Dec  6 11:39:19 <hostname> postfix/smtpd[20015]: warning: problem
talking to server 127.0.0.1:2552: Connection timed out

Then eventually Postfix starts defferring:

Dec  6 11:41:00 <hostname> postfix/smtpd[20015]: NOQUEUE: reject: RCPT
from <snip>: 451 4.3.5 Server configuration problem; from=<snipped>
to=<snipped> proto=ESMTP helo=<snipped>

The connections are held too long and...

Dec  6 11:40:08 <hostname> postfix/master[23860]: warning: service
"smtp" (25) has reached its process limit "100": new clients may
experience noticeable delays
Dec  6 11:40:08 <hostname> postfix/master[23860]: warning: to avoid
this condition, increase the process count in master.cf or reduce the
service time per client

The warnings about the process limit is just a side effect of maRBL
not responding either at all or not fast enough. Disabling the maRBL
policy server in the Postfix configuration is the only way to resume
mail flow. Resuming use of it immediately results in the same problem.

Memory usage of maRBL steadily increased from around 4MB to 40MB over
the course of 5 days. Anyone else experiencing any issues with maRBL?
I have another policy server (Policyd) that is called on every email
yet does not exhibit any of these issues. Thanks in advance.

Perl 5.8.5
Net::Server 0.94
Net::RBLClient 0.4