Hi Dan,

Here's a stack trace for a spamass-milter that's stuck.  Note that it's not the "main" spamass-milter that is getting stuck;  It's a child of the main spamass-milter process.


mailserver#     pstack 29702/26935
29702:  /etc/mail/sbin/spamass-milter -p /var/run/spamass.sock -f -r 10 -u spa
 ff143720 _ZN9__gnu_cxx18__exchange_and_addEPVii (42ca0, ffffffff, 1, 3e, 0, 0) + 20
 ff12862c _ZNSs9_M_mutateEjjj (42b80, 0, 0, 42c98, 0, 0) + 1c0
 ff128df4 _ZNSs12_M_leak_hardEv (42b80, 0, 0, 0, 0, 0) + 58
 ff128eb0 _ZNSsixEj (42b80, 0, 0, 0, 0, 0) + 18
 000151d0 _ZN12SpamAssassin10local_userEv (42b80, 1, 95f8c, 14048, fef6a2b4, fef729f8) + 10
 00015c74 _ZN12SpamAssassin7ConnectEv (42b50, 1a9b8, cc, 1, 40, 0) + 5ac
 00016a78 Z11mlfi_headerP8smfi_strPcS1 (33ec0, 42da0, 42da9, fe6fbe38, 8, 42da9) + 38
 ff3743d8 st_header (0, 16a40, ffffffff, ff38a6d0, ff38a000, fe6fbed0) + 5c
 ff373a5c mi_engine (33ec0, 7, ff38a000, 42da0, 7, 16664) + 2d0
 ff375dd4 mi_handle_session (33ec0, 33ec0, 14264, 0, ff38a000, ffffffff) + 54
 ff375418 mi_thread_handle_wrapper (33ec0, fe6fc000, 0, 0, ff375414, 0) + 4
 fef408e8 _lwp_start (0, 0, 0, 0, 0, 0)

Hi Dan,
 
Thanks for a reply.  Sure, I can run those commands the next time spamass-milter  gets stuck, but it's been running really well for me lately.  
 
One thing... "prstat -L" doesn't tell me a "thread number" as far as I can tell.  But I can run those commands by using the pid.

Hi Anonymous,

No, we never found a solution for this problem.  Usually, it'll occurs in a cluster of 4 times in a week, then not at all for a month.

Hi Folks,

I am running spamass-milter 0.3.1 with sendmail 8.14.1 and 3.2.4.
I am doing this on a Solaris 10 computer.

For the most part, spamass-milter works like a champ.  I'm using these flags:
/etc/mail/sbin/spamass-milter -p /var/run/spamass.sock -f -r 10 -u spamass-milter-default

So it automatically rejects e-mails with scores greater than 10.

My problem is that every few days, spamass-milter gets stuck.  When it is stuck, it is always a child spamass-milter (so the main
spamass-milter continues to work).  From "truss -p <pid>, it apperas that spamass-milter is doing nothing.  From the log file, I see the following errors:

Milter (spamassassi): timeout before data read, where=body
Milter (spamassassin): to error state

And although spamass-milter is "stuck", the e-mail gets through to our user without delay, though clearly it has not been processed by spamassassin because it contains no X-Spam-Status headers.

We sometimes get the above Milter error messages and spamass-milter DOESN'T get stuck.

Does anyone know how I can figure out what's going on and fix this problem?  Once spamass-milter hits 100% CPU usage, my e-mail server gets really slow and my users get grumpy.