Re: spamass-milter goes to stop state?

[Dan Nelson]

In the last episode (Jul 05), Tony Shadwick said:
> I've been wrestling with this for months and I'm just now getting
> around to dealing with it.
> 
> I have a pair of mail exchangers that experience a pretty heavy load.
> There doesn't appear to be one particular cause that trips this, but
> at random times, sometimes days or even weeks apart, the
> spamass-milter process will go to "Stop" state (at least that is what
> `top` says) on FreeBSD 5.3.  Doing `kill -9 (pid)` doesn't do it. 
> There's no way to kill it other than rebooting the system.
> 
> Any ideas what might trigger this?  System specs are different on
> both machines, but both are running the most current 5.3 kernel
> sources available, however this has happened on every kernel we've
> built. :\ We use generic sources, enabling SMP on one, and the other
> not.

I have seen this, but only in a testsuite of threaded program that
fork/exec'ed heavily.  If you send the process a signal (something that
won't cause an exit, like SIGCHLD, or SIGINFO), it should free up and
continue on its way.  Try putting a 5.4 kernel on if you haven't tried
already, and see if the problem clears up.  You don't have to update
any userland binaries or libraries, just the kernel.  Or, switch to
good old libc_r threads (via a libmap.conf entry), since spamass-milter
doesn't need real kernel threads.
 
-- 
        Dan Nelson
        address@hidden