[Mimedefang] Work directory names

[Joseph Brennan]

Mimedefang seems not to generate enough different names
for work directories for a very busy server.

This is on a Sun 4500 with eight processors, sendmail 8.12.8,
mimedefang 2.31.  I started sendmail using mimedefang (it had no
milter previously) at 09:53.  Errors began at 10:01.

Two samples below.  I have to disguise the sender addresses (as
xxx at yyy.com and zzzz at columbia.edu) but they were different from
each other.  It tried to use this particular name six times,
and others also.

# grep h2RF33fu026999: syslog
Mar 27 10:03:03 marionberry mimedefang[16960]: [ID 702342 mail.warning] 
h2RF33fu026999: Could not create directory 
/var/spool/MIMEDefang/mdefang-3E8312A7-0: File exists
Mar 27 10:03:03 marionberry sm-mta[26999]: [ID 801593 mail.notice] 
h2RF33fu026999: Please try again later
Mar 27 10:03:03 marionberry sm-mta[26999]: [ID 801593 mail.info] 
h2RF33fu026999: from=<xxx at yyy.com>, size=5232, class=0, nrcpts=0, 
proto=ESMTP, daemon=MTA-v4, relay=216.230.78.50.cypresscom.net 
[216.230.78.50]

# grep h2RF31fv026986 syslog
Mar 27 10:03:03 marionberry mimedefang[16960]: [ID 702342 mail.warning] 
h2RF31fv026986: Could not create directory 
/var/spool/MIMEDefang/mdefang-3E8312A7-0: File exists
Mar 27 10:03:03 marionberry sm-mta[26986]: [ID 801593 mail.notice] 
h2RF31fv026986: Please try again later
Mar 27 10:03:03 marionberry sm-mta[26986]: [ID 801593 mail.info] 
h2RF31fv026986: from=<zzzz at columbia.edu>, size=0, class=0, nrcpts=0, 
proto=ESMTP, daemon=MTA-v4, relay=dialup-ccts3-35.cc.columbia.edu 
[128.59.58.140]


No errors from start at 09:53 through 10:00.  Then:

# grep Mar.27.10:01.*Could.not.create.directory./var/spool/MIMEDefang 
syslog | wc -l
    241

# grep Mar.27.10:02.*Could.not.create.directory./var/spool/MIMEDefang 
syslog | wc -l
    316

# grep Mar.27.10:03.*Could.not.create.directory./var/spool/MIMEDefang 
syslog | wc -l
     25

I killed sendmail at 10:03:04.

Measuring by "grep Mar.27.09:54.*from= syslog | wc -l" and so on
for each minute, this is how much mail came in:

09:54	244
09:55	268
09:56	278
09:57	267
09:58	265
09:59	255
10:00	242
10:01	413
10:02	395

So there was a spike at 10:01 which is right where Mimedefang ran
into trouble with names.

The good news is the system load stayed around 4 to 5, which is
fine on an 8-processor box, and memory was more than 50% free.
I only halted it because of the name errors.

What to do now?

Joseph Brennan          Columbia University in the City of New York
postmaster at columbia.edu                 Academic Technologies Group