Empty messages, again (on 4.25-9)

[Jacques Caruso]

Hi everybody,

I have upgraded to 4.25-9 since Thursday, and today the dreaded
empty-mail problem has resurfaced. The problems always appear in the
same order :

* MS stops processing messages without logging anything. The messages
  start to clog up the postfix.in/deferred queue
* We notice the problem (usually when the queue is very huge, around
  ~1400 messages today) and restart MS
* Processing of messages restarts, the system load goes through the
  roof (from ~3 to ~15)
* Postfix begins to emit 'Skipped, still being delivered' messages, as
  MS accesses messages in the (wrong) postfix.in/incoming queue :

[root at sceuzi][~]# cat lsof.out | grep postfix.in/incom
MailScann 15893     postfix   18uW  REG        8,2       2362    1802747 /var/spool/postfix.in/incoming/4/4ABB91B81FB
MailScann 15900     postfix   25uW  REG        8,2       1672    1802536 /var/spool/postfix.in/incoming/7/701A91B8128
MailScann 15900     postfix   30uW  REG        8,2       2204    1804071 /var/spool/postfix.in/incoming/7/707E41B8727
MailScann 15900     postfix   31uW  REG        8,2       2306    1804100 /var/spool/postfix.in/incoming/7/7F8C41B8744
MailScann 15900     postfix   32uW  REG        8,2      25385    1804118 /var/spool/postfix.in/incoming/7/7959C1B8756
MailScann 15900     postfix   33uW  REG        8,2       3954    1802533 /var/spool/postfix.in/incoming/7/713D81B8125
MailScann 15900     postfix   34uW  REG        8,2       2161    1802605 /var/spool/postfix.in/incoming/7/76CE31B816D
MailScann 15900     postfix   35uW  REG        8,2       2636    1803233 /var/spool/postfix.in/incoming/7/7CCAC1B83E1
MailScann 15910     postfix   20uW  REG        8,2       1998    1802462 /var/spool/postfix.in/incoming/B/BDF961B80DE
MailScann 15924     postfix   17uW  REG        8,2       1771    1802541 /var/spool/postfix.in/incoming/7/761651B812D
MailScann 15945     postfix   27uW  REG        8,2       1614    1803853 /var/spool/postfix.in/incoming/7/73F9C1B864D
cleanup   17875     postfix   16u   REG        8,2      81920    1802800 /var/spool/postfix.in/incoming/5/57FDC1B8230
cleanup   18615     postfix   16u   REG        8,2          0    1802797 /var/spool/postfix.in/incoming/4/45A501B822D
MailScann 21344     postfix   18u   REG        8,2       2362    1802747 /var/spool/postfix.in/incoming/4/4ABB91B81FB
MailScann 21348     postfix   20u   REG        8,2       1998    1802462 /var/spool/postfix.in/incoming/B/BDF961B80DE

I noticed that a new version appeared on 29/11. Still, the changelog
lists only two more fixes, apparently unrelated to my problem :

        - Fixed bug where sender of bulk precedence mail would be sent some warnings
          if their mail was identified as spam.
        - Fixed duplicate logging of New Batch messages. Abandoned support of
          syslog-ng until I can test it properly on my own systems.

Like last time, I put the result of 'lsof' and 'ps -afx' on a web
server :

http://aragorn.monaco.net/tmp/ms/2/lsof.out.txt
http://aragorn.monaco.net/tmp/ms/2/psafx.out.txt

Could someone give me hints about :

* preventing MS from stopping to process mail ? This happens every few
  days for apparently no reason. The processes just sit idle.
* preventing MS from accessing mail in the postfix.in/incoming queue ?
  I've thought about permissions but I don't see how to let MS read in
  deferred but not in incoming (after all, they need to have the same
  permissions for Postfix to operate).

Any help (like a hint to a probable cause for the problems) will be
greatly appreciated.

Cheers,
-- 
[ Jacques Caruso <jacques at monaco.net>                  Développeur PHP ]
[ Monaco Internet                           http://monaco-internet.mc/ ]
[ Tél : (+377) 93 10 00 43                        Clé PGP : 0x41F5C63D ]
[ * Tired of choosing the lesser of two evils? Vote Cthulhu in 2004! * ]