On Thu, 2008-05-01 at 15:57 -0400, Jean-Paul Natola wrote: > > -----Original Message----- > From: Ross Boylan [mailto:[EMAIL PROTECTED] > Sent: Thursday, May 01, 2008 3:33 PM > To: Jean-Paul Natola > Cc: [EMAIL PROTECTED]; SpamAssassin > Subject: RE: Connection timed out > > On Thu, 2008-05-01 at 13:54 -0400, Jean-Paul Natola wrote: > > OPTIONS="--create-prefs --max-children 5 --helper-home-dir \ > > --username=mail --socketpath=/var/run/spamd/socket" > > > > >I'm running on a Pentium 4 with hyperthreading, which appears as 2 CPU's > > >to the OSs. There's really only 1 CPU. I wonder if that could have > > >something to do with the trouble. > > ------------------------------------------------------------------------ > > > > How much ram do you have, are you exceeding your physical mem? > > > > > > > 2G physical + 2G swap. I don't think I'm exceeding it. In response to Randy's query, there aren't out of memory messages in the log. > > > I'm not entirely sure about Debian , I use BSD but me with my measly 512 mb > of ram, I have higher max children- > These are my settings > > --min-children=2 --max-children=8 --min-spare=2 --max-spare=3 > --max-conn-per-child=100 -r > > Maybe its timing out on large messages? Are you limiting message size to > scan? I only scan messages under 50k.
The fact that it only happens at night and that each day's incidents are usually within a few minutes of each other suggests that some system maintenance task is getting in the way, maybe just by slowing things down. I run a backup over the same period; at various times it can be disk or CPU intensive (compression). But the fact that the times of the errors vary by up to 4 hours from day to day (e.g., 1:30 v 4:30am) doesn't seem to fit that theory, since I'd expect the housekeeping jobs to run at more consistent times. It looks as if there were only 1 or 2 messages being processed, so it's hard to see how the current max-children setting would cause the problem. Ross
