Thanks, Duane for your kind comment. Yes I have tried different
methods. I used to use cgpsa. cgpav was only for antiviral stuff and
it failed miserably when I tried to have it doing the spamassassin
calls also. It seems that no matter how I call spamd there is an issue.
I'm using scanspam.sh as an execute call in CGPro. In fact I think
I've actually identified where in the shell script the issue may be
occurring. Here's an excerpt from the key part:
/var/CommuniGate/spam/spamprep "$myCgate/$QueuePath" "$ReturnPath"
"$Username" |
/usr/bin/spamc -d 127.0.0.1 -t 100 -u "$Username" >> "$myCgate/
Submitted/$NewFile"
mv /var/CommuniGate/Submitted/$NewFile /var/CommuniGate/Submitted/
$FinalFile
CG Pro gives execute scripts 2 minutes to finish or it kills them.
Possibly it calls the script again, or so I thought, which might be
the source for the multiple .tmp extensions and long processing times.
If the first line above is calling the spamd, then CG Pro kills the
script before the mv command adds the .sub extension. That would
account for the orphaned .tmp files that have the spamd processing
finished, but never got submitted.
Still that does not tell me that there is a problem with CG Pro and
shell scripts. It could likely just as much mean that spamd is slowed
because of a memory leak in that code.
Notice also that the -t 100, which means that spamd should finish
processing within 100 seconds (or so I understand) SHOULD mean that
CGPro shouldn't ever reach the 120 second (2 minute) limit that would
cause it to kill the sh process.
When I watch the submitted folder, MOST of the processing however
occurs very quickly and there are usually not more that 3 or 4 .tmp
files being processed. Until the spam load increases. That's where my
suspicion of a memory leak in spamd comes from.
In order to further test this, I'm considering altering the script
above to actually call another script basically containing the two
lines above, thereby preventing CGPro from killing the script
prematurely. That's a next step. If the delay in processing is still
present then, I would think that I've moved suspicion away from CG Pro/
spamd interaction as a cause for this.
Ron Smith
[EMAIL PROTECTED]
"Having an email problem is painful, but character-building."
On Jul 25, 2008, at 9:46 AM, Duane Hill wrote:
For a test, ever thought about changing to a different script?
Even through we have two Postfix border servers doing filtering, I
still have something on our internal (antiquated) CommuniGate server.
I am using SA 3.2.5 with cgpsa and there hasn't been one bit of an
issue.
http://www.tffenterprises.com/cgpsa/
-d
