Re: spamass-milter dies under heavy load

[Dan Nelson]

In the last episode (Sep 20), Matt Cramer said:
> I am running two HP-UX 11.0 servers with ISC BIND 8.12.3.  They
> connect to a dual processor linux 2.4.4 machine running spamassassin,
> spamass-milter 0.1.2 (compiled with ISC sendmail 8.12.3 milter libs).

There's no such thing as ISC sendmail, and bind jumped from 8.3.3 to 9 :)
If your HP-UX boxes are your mail servers and the Linux boxes are your
spamassassin servers, you want to put spamass-milter on the HP-UX box,
and edit the SPAMC define in config.h to "spamc -d linuxhost". 
Theoretically, you could run `` SPAMC="spamc -d linuxhost" ./configure ''
but for some reason that doesn't work.  And I just figured out why :)
Next release you will be able to do that.
 
> spamass-milter is locking up (not accepting connections on the tcp
> socket).  I can not get it to dump core by sending it a kill -11! 
> Has anyone else experienced this problem?  Under light load it runs
> like a champ, but when put in our production environment it just
> chokes.  Our two mail servers process about about 8000 messages /
> day, each.  Sometimes it will run for not more than 15 minutes before
> needing to be shutdown and restarted.

-11 is SIGSEGV, and applications are allowed to ignore (or trap) that. 
Try -9 aka -KILL.  Also make sure your resource limits allow you to
coredump.  Some OSes have a "gcore" command which will generate a
corefile from a running processes without killing it.

Some other options are attaching to the processs with a debugger, or
truss/strace/ktrace to show the syscalls the process is calling.
 
-- 
        Dan Nelson
        address@hidden