mailscanner dying

Mon Jan 7 13:43:24 GMT 2002

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello Scott,

What version of MailScanner are you running?

I had this same problem and worked with Julian to get it fixed.  The
version I am running was installed via RPM and is mailscanner-3.01-2

This is on a RedHat 7.2 system with all updates installed from RedHat.

JBB

Monday, January 07, 2002, 1:21:41 AM, you wrote:

SF> I am getting a similar result. The mailscanner perl task just vanishes.

SF> I set SpamAssassin to no.

SF> It was also fairly random, if I started mailscanner enough times the email
SF> finally got through.

SF> It was occuring on plaintext emails (that aren't scanned) and ones with
SF> attachments (scanned with innoculate). The innoculate part works fine.

SF> I never had mailscanner fail under mailscanner 2.53-1, even over a few
SF> month sof continuos running.

SF> under Debug=1, I ran the /usr/local/MailScanner/bin/check_mailscanner.linux
SF> directly, and saw a few segfaults in line 50, which is the close if I
SF> think, which doesn't make a lot of sense to me. One other time it said
SF> something about "$process $config "  which appears in the script file just
SF> above line 50.

SF> I am running :
SF> This is perl, v5.6.0 built for i386-linux

SF> I have redhat 7.2 - I am going to try any redhat ways of upgrading perl
SF> first, then try CPAN if there isn't a redhat upgrade.

SF> ps. (I haven't tried the adjusted sweep.pl for innoculate fixes yet (from
SF> Nick), as I haven't got enough other things in 3.0 stable enough)

SF> pps. I have had to set mailscanner to process individual mails, 1 at a
SF> time, and check the mailscanner process often. I am keen to see if anyone
SF> else is having this problem.

SF> regards
SF> Scott Farrell

SF> http://www.icconsulting.com.au
SF> ic Consulting - the people that make eBusiness happen.
SF> We offer e-business consulting and perform services. We deliver high impact
SF> consulting, and fast turn around projects for our clients.
SF> Ask us about Web Content Management,  Web Self Service, or working closer
SF> with your customers or suppliers.

SF> 0412 927 156,   02 9411 3622  mailto:sfarrell at icconsulting.com.au

SF>                     "Michael H.
SF>                     Warfield"                 To:     MAILSCANNER at JISCMAIL.AC.UK
SF>                     <mhw at WITTSEND.COM>        cc:
SF>                     Sent by:                  Subject:     Re: Can't get spam checking working?
SF>                     MailScanner
SF>                     mailing list
SF>                     <MAILSCANNER at JISCM
SF>                     AIL.AC.UK>

SF>                     07/01/02 04:37 AM
SF>                     Please respond to
SF>                     MailScanner
SF>                     mailing list

SF> On Sun, Jan 06, 2002 at 04:20:31PM +0000, Julian Field wrote:
>> At 15:53 06/01/2002, you wrote:
>> >> At this point is the mailscanner process still running? What's the
SF> last
>> >few
>> >> mailscanner lines in your maillog when this happens?
>> >
>> >        For me at least, the MailScanner process is gone.
>> >
>> >        Last MailScanner messages I see in maillog is "Scanning x
SF> messages,
>> >xxx bytes".  I restart the MailScanner by hand and get the same message
>> >(generally same x messages and xxx bytes) and if often continues.  Last
>> >night, I had to restart it 6 times in the same spot (i.e. I looked after
>> >starting it and it was gone again) but then it worked and was running
>> >this morning when I got up.  So it seems to be "non-deterministic" in
>> >some way.  Sendmail processes continue to run even after the mailscanner
>> >process dumps.

>> This has got to be bugs in Perl as it is non-deterministic. What error do
>> you get when it dies? Just a "Segmentation fault" or something more
SF> useful?

SF>         None of the above.  No "Segmentation fault" and no messages to
SF> standard out or standard error.  The only message I see is "Starting
SF> virus scanner..." and the process later exits.  Now, one thing I haven't
SF> done (but will the next time it does this) is to run it in debugging mode
SF> and keep it in the forground.

>> Unfortunately, quite a lot happens after that maillog entry before
SF> anything
>> more happens that will log to maillog, so it's not a very good indicator
SF> of
>> where in MailScanner it died. I'll look at reimplementing some of the
>> timeout code, as that's the only code I have ever written that has
>> repeatedly caused Perl to segmentation fault.

>> Does it still segmentation fault with all the spam checking switched off?
>> That would help narrow it down a bit.

SF>         Right now, the only spam checking it's running is one of the RBLs.
SF> SpamAssassin is disabled.

>> Can you also try starting it in debug mode and see if it still dies at
SF> the
>> same place. And does it log anything more before it dies? The best way of
>> starting it is with the check_mailscanner script. (Don't touch the
SF> sendmail
>> processes while doing this, they will happily carry on working)

SF>         I'll try that.

>> >> You may well need to upgrade your version of Perl, a few people are
>> >seeing
>> >> Perl core-dump due to internal bugs in Perl itself.
>> >
>> >        Because I installed SpamAssassin and used CPAN, I ended up with
>> >Perl upgraded to 5.6.1 anyways.  I've currently got SpamAssassin
SF> disabled
>> >in the MailScanner script because of the errors (tagging all messages)
>> >being reported and discussed in the other thread on this list.

>> Are you sure that you are using the upgraded Perl 5.6.1? You might have
>> ended up with 2 copies of Perl installed on your system. Do a
>> "/usr/bin/perl -v" just to check...

SF>         Very sure.  I remarked to some friends to watch out if they were
SF> running CPAN on a RedHat system and it tried to upgrade perl because the
SF> default would go into /usr/local/bin.  I had caught that when I saw it
SF> and directed the perl build to use /usr for the prefix, but here is the
SF> double check, anyways...

SF> [mhw at alcove mhw]$ /usr/bin/perl -v

SF> This is perl, v5.6.1 built for i586-linux

>> I've seen SpamAssassin tag *a few* messages that it shouldn't have, but
SF> the
>> SA is_spam() routine definitely always says they are spam, so my code is
>> only doing exactly what SpamAssassin says it should be.

SF>         Yeah, I've got SpamAssasin disabled in the MailScanner config
SF> right now and am only using it from procmail.  There are definitely some
SF> problems there, expecially vis-a-vis exit codes.  The code itself claims
SF> there is a -e switch to enable a non-zero exit code on spam, but the
SF> -e switch isn't even recognized.  I haven't even looked into the API
SF> calls yet or followed up over on the SpamAssassin side of the house.
SF> The plain text reporting is working incredibly well, though, and I
SF> like Vipul's Razor.  Getting it to play nicey nicey with MailScanner
SF> would be a nice benefit...

SF>         Oh...  BTW...  In one of your earlier messages I noticed a remark
SF> that you had seen times when SpamAssassin didn't exit.  I've noticed that
SF> there are times when it takes a LONG time to exit.  Seems to center
SF> around the RBL checks and seems to correspond to sluggishness in the
SF> DNS (which would make some sense).  I have yet to see the spamassassin
SF> command not eventually exit.  I did have problems with MailScanner
SF> initially, which I tracked down to the default of including the JANET
SF> UK RBL which seems to cause the MailScanner to hang on my system.
SF> Don't know if that is still the default in the lastest bundles, but
SF> it caused me some serious headaches till I realized what was misconfigured.

>> --
>> Julian Field                Teaching Systems Manager
>> jkf at ecs.soton.ac.uk         Dept. of Electronics & Computer Science
>> Tel. 023 8059 2817          University of Southampton
>>                             Southampton SO17 1BJ

SF>         Mike
SF> --
SF>  Michael H. Warfield    |  (770) 985-6132   |  mhw at WittsEnd.com
SF>   /\/\|=mhw=|\/\/       |  (678) 463-0932   |  http://www.wittsend.com/mhw/
SF>   NIC whois:  MHW9      |  An optimist believes we live in the best of all
SF>  PGP Key: 0xDF1DD471    |  possible worlds.  A pessimist is sure of it!

- --
Best regards,
 Jonathan                            mailto:jbayer at bayerfamily.net
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (MingW32)
Comment: For info see http://www.gnupg.org

iEYEARECAAYFAjw5pf0ACgkQLWek1tt+K51O3ACcDTtX3U/UWjkLidwYE0Y6+I2a
zGsAnR5unNLASnAFfLBnj3a78wHmeWEE
=7UDY
-----END PGP SIGNATURE-----