Repeated spamd dying due to SIGCHLD signal 11

[a....@ukgrid.net]

Hi,

   I have a problem on one server that I see several times an hour  
this problem logged by spamd

Wed May  5 10:04:43 2010 [88823] info: spamd: handled cleanup of child  
pid [90622] due to SIGCHLD: DIED, signal 11 (000b)

And in the main messages file a corresponding error regarding the perl process

May  5 10:04:43 <kern.info> nu kernel: pid 90622 (perl), uid 26:  
exited on signal 11

This is a problem on a relatively new server:

FreeBSD 8.0
SpamAssassin version 3.3.1
   running on Perl version 5.10.1

On an older FreeBSD version I am having no problems, both should be  
configured the same re mail and spamassassin. Today I have also copied  
over the .spamassassin data directory from the working server to the  
dodgy server in case there was any file corruption but the problem  
persists.
A similar problem I found searching the errors seems to be:

https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6127

However this isnt resolved currently, and also I dont have a GB sized  
auto whitelist file...

Any ideas on this? Whats my next step? Some more debugging? Any advise  
gratefully received,

thanks Andy.

PS Below I include a full exert from the spamd log for a process that  
dies in this way:

Wed May  5 10:34:03 2010 [96093] info: spamd: server successfully  
spawned child process, pid 98044
Wed May  5 10:34:03 2010 [96093] info: prefork: adjust: 0 idle  
children less than 1 minimum idle children. Increasing spamd children:  
98044 started.
Wed May  5 10:34:03 2010 [98044] info: spamd: connection from  
localhost [127.0.0.1] at port 48998
Wed May  5 10:34:12 2010 [98044] warn: bayes: cannot open bayes  
databases /var/spool/mqueue/.spamassassin/bayes_* R/W: lock failed:  
File exists
Wed May  5 10:34:12 2010 [98044] info: spamd: processing message  
<rt...@ukgrid.net> for mailnull:26
Wed May  5 10:34:19 2010 [98044] warn: bayes: cannot open bayes  
databases /var/spool/mqueue/.spamassassin/bayes_* R/W: lock failed:  
File exists
Wed May  5 10:34:29 2010 [98044] warn: bayes: cannot open bayes  
databases /var/spool/mqueue/.spamassassin/bayes_* R/W: lock failed:  
File exists
Wed May  5 10:34:37 2010 [98044] info: spamd: clean message (1.3/4.8)  
for mailnull:26 in 33.9 seconds, 2794 bytes.
Wed May  5 10:34:37 2010 [98044] info: spamd: result: . 1 -  
BAYES_00,J_CHICKENPOX_27,MISSING_HEADERS,NO_RELAYS,REPLYTO_WITHOUT_TO_CC  
scantime=33.9,size=2794,user=mailnull,uid=26,required_score=4.8,rhost=localhost,raddr=127.0.0.1,rport=48998,mid=<rt...@ukgrid.net>,bayes=0.000000,autolearn=no
Wed May  5 10:34:38 2010 [98044] info: spamd: connection from  
localhost [127.0.0.1] at port 25754
Wed May  5 10:34:47 2010 [96093] info: spamd: handled cleanup of child  
pid [98044] due to SIGCHLD: DIED, signal 11 (000b)

Re: Repeated spamd dying due to SIGCHLD signal 11 SOLVED

[a....@ukgrid.net]

Hi all,

   ok I did some furhter work and when testing (sa-learn --sync) the  
bayes db integrity its not good. So basically the problem was caused  
by a corrupt bayes DB.

Thanks Andy.

Quoting a.smith@ukgrid.net:

Re: Repeated spamd dying due to SIGCHLD signal 11

[a....@ukgrid.net]

Hi,

   Ok, update 2:

I proved that my system had the perl but (via a test script) as  
previously mentioned. Ive therefore downgraded perl to perl 5.8.9. I  
still have the same problem with SpamAssassin perl processes crashing.
I´ve take some more logs from spamd with debugging enabled, this is  
the last thing you see before it goes wrong:

Thu May  6 01:10:01 2010 [88183] dbg: locker: safe_lock: breaking  
stale /var/spool/mqueue/.spamassassin/bayes.lock: age=1273104000  
now=1273104601
Thu May  6 01:10:02 2010 [88183] dbg: locker: safe_lock: trying to get  
lock on /var/spool/mqueue/.spamassassin/bayes with 6 retries
Thu May  6 01:10:02 2010 [88183] dbg: locker: safe_lock: link to  
/var/spool/mqueue/.spamassassin/bayes.lock: link ok
Thu May  6 01:10:02 2010 [88183] dbg: bayes: tie-ing to DB file R/W  
/var/spool/mqueue/.spamassassin/bayes_toks
Thu May  6 01:10:02 2010 [88183] dbg: bayes: tie-ing to DB file R/W  
/var/spool/mqueue/.spamassassin/bayes_seen
Thu May  6 01:10:02 2010 [88183] dbg: bayes: found bayes db version 3
Thu May  6 01:10:02 2010 [88169] dbg: prefork: child 88183: just exited
Thu May  6 01:10:02 2010 [88169] dbg: prefork: child 88183: entering state 4
Thu May  6 01:10:02 2010 [88169] info: spamd: handled cleanup of child  
pid [88183] due to SIGCHLD: DIED, signal 11 (000b)

Anyone any ideas?

Re: Repeated spamd dying due to SIGCHLD signal 11

[a....@ukgrid.net]

Hi,

   An update.... it looks like the problem may well be this Perl bug  
that affects Perl 5.10.1 in the FreeBSD ports tree...

http://rt.perl.org/rt3//Public/Bug/Display.html?id=69973

thanks Andy.