Milter-greylist

I am not sure how best to proceed with the following:

I have 3 Servers running Sendmail/MailScanner/Milter-Greylist after 4-6
hours the greylist daemon bails causing socket errors in sendmail. One of
the servers is so badly affected that it kills the ability to log in even
from the console.

At first I was using the Multi-MX sync feature but found that failed
within 4 hours or so with messages indicating that 1024 connections where
queued. I reconfigured my Load Balancers to have simple persistence based
on sender IP and set each server up to run without MX sync and we got
greater time between failures. Looking at an active server that had just
become unstable I noticed that it had 250ish greylist processes and
Sendmail could not connect to any of them.

My thoughts are that its a process/thread limit rather than a memory issue
since each machine has 1024 Meg.

Any suggestions?

Thanks for the help.


Garry Davies