I was experiencing the same exact problem. It was unclear why the process handling SMTP connections was dying and nothing was getting logged. I saw a correlation to large amounts of spam traffic and assumed it was some exploit or buffer overflow, it ended up being much simpler.

In your esmtp configuration file there is a path to set the logger:

TCPDOPTS="-stderrlogger=/usr/sbin/courierlogger"

I had an old path in there, and when traffic hit 35 processes, it would attempt to log a warning:

write(2, "WARN: 35 active connections.\n", 29) = -1 EPIPE (Broken pipe)

to the non existent (pipe?) and kill off the parent process.

This is a bit of a subtle misconfiguration, and only debugable via strace when you are actually hitting the ceiling of your MAXDAEMONS=40. (Sam, don't tear my head off, but would it make any sense to refuse to start if that pipe is non-existent or not die if it can't be written to?)

Hope this solves your problem.

Keywords:

esmtp smtp dying failing silent