4.65.3-1 Segmentation fault at end of batch

Lukas Garberg lukas at spritelink.net
Mon Feb 18 23:06:06 GMT 2008


Hi list,

I run a spam filter solution with three identical (cloned) servers running
Linux  2.6.21.5 (Slackware 12.0) with MailScanner 4.65.3-1, postfix 2.4.6,
SpamAssassin 3.2.4 & perl 5.8.8.
Hardware is Pentium 4 2.6 GHz with 1 GB RAM. A 128 MB ramdisk is mounted on
MailScanners incoming directory.

I should maybe also note that all the three machines use the same SA
bayes-database stored in SQL on a fourth machine.

Now to the problem:
_Two_ of the machines have a strange problem with MailScanner
segfault:ing at
the end of almost every batch; this is what a ps aux | grep MailS looks
like:

root at xxx:/opt# ps aux | grep MailS
postfix   3462  0.0  1.9  23936 19896 ?        Ss   00:22   0:00
MailScanner: starting child
postfix   4467  3.8  0.0      0     0 ?        Z    00:35   0:02
[MailScanner] <defunct>
postfix   4475  4.2  0.0      0     0 ?        Z    00:35   0:02
[MailScanner] <defunct>
postfix   4484  4.4  0.0      0     0 ?        Z    00:35   0:02
[MailScanner] <defunct>
postfix   4490  4.9  0.0      0     0 ?        Z    00:35   0:02
[MailScanner] <defunct>
postfix   4493  5.6  0.0      0     0 ?        Z    00:35   0:02
[MailScanner] <defunct>
postfix   4503  6.5  0.0      0     0 ?        Z    00:35   0:02
[MailScanner] <defunct>
postfix   4508  7.6  0.0      0     0 ?        Z    00:36   0:02
[MailScanner] <defunct>
postfix   4512  9.2  0.0      0     0 ?        Z    00:36   0:02
[MailScanner] <defunct>
postfix   4517 12.1  0.0      0     0 ?        Z    00:36   0:02
[MailScanner] <defunct>
postfix   4527 16.4  0.0      0     0 ?        Z    00:36   0:02
[MailScanner] <defunct>
postfix   4533 23.8  4.8  55764 49944 ?        S    00:36   0:01
MailScanner: waiting for messages
postfix   4539 60.6  4.7  54828 48940 ?        S    00:36   0:01
MailScanner: starting child
root      4543  0.0  0.0   2004   636 pts/0    R+   00:36   0:00 grep MailS

Note that one of the three machines runs perfectly!

If I run MailScanner with Debug & Debug SpamAssassin set to yes this is
the end
of the output of /opt/MailScanner/bin/check_MailScanner:

[4843] dbg: check: is spam? score=13.138 required=5
[4843] dbg: check:
tests=BAYES_99,RCVD_IN_PBL,RCVD_IN_SORBS_DUL,RCVD_IN_XBL,RDNS_DYNAMIC,
STOX_REPLY_TYPE,TRACKER_ID,TVD_SPACE_RATIO
[4843] dbg: check:
subtests=__ANY_OUTLOOK_MUA,__CT,__CTE,__CTYPE_CHARSET_QUOTED,
__CT_TEXT_PLAIN,__DOS_HAS_ANY_URI,__DOS_RCVD_MON,__DOS_RCVD_TUE,__FH_HAS_XMSMAIL,
__FH_HAS_XPRIORITY,__HAS_ANY_URI,__HAS_MIMEOLE,__HAS_MSGID,__HAS_MSMAIL_PRI,
__HAS_RCVD,__HAS_SUBJECT,__HAS_X_MAILER,__HDR_ORDER_FTSDMCXXXX,__HELO_NO_DOMAIN,
__LAST_UNTRUSTED_RELAY_NO_AUTH,__MIMEOLE_MS,__MIME_VERSION,__MISSING_REF,
__MSGID_DOLLARS_MAYBE,__MSGID_DOLLARS_OK,__MSGID_OK_HEX,__MSGID_RANDY,
__NONEMPTY_BODY,__NO_INR_YES_REF,__OE_MSGID_2,__OE_MUA,__RATWARE_0_TZ_DATE,
__RCVD_IN_SORBS,__RCVD_IN_ZEN,__RDNS_DYNAMIC_HCC,__RDNS_DYNAMIC_IPADDR,
__RDNS_INDICATOR_TYPE,__SANE_MSGID,__TOCC_EXISTS,__XM_MSOE6,__XM_MS_IN_GENERAL,
__XM_OUTLOOK_EXPRESS
[4843] dbg: learn: auto-learn? ham=0.1, spam=12, body-points=10.716,
head-points=10.716, learned-points=4
[4843] dbg: learn: auto-learn? no: inside auto-learn thresholds, not
considered ham or spam
/opt/MailScanner/bin/check_mailscanner: line 131:  4822 Segmentation
fault      $process $config


I've seen a few other posts to the list with similar problems, and most
answers
have been bayes-related. I, at least in my case, don't find a bayes
error very
probable since the MailScanner successfully scans & learns all messages
but the
last one (Max Unscanned Messages Per Scan = 10) and everything runs
smoothly on
one of the threes machines, with a common bayes DB.

Any ideas on how to go further into solving this problem?

Thank you in advance,
Lukas Garberg


More information about the MailScanner mailing list