> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:clamav-users-
> [EMAIL PROTECTED] On Behalf Of Phillip Salzman
> Sent: Wednesday, May 04, 2005 11:03 AM
> To: ClamAV users ML
> Subject: RE: [Clamav-users] clamav dying & accept() errors in syslog
> 
> It could be, but the -STABLE we're running is from over a year after
the
> "fix."
> 
> Thanks,
> 
> Phillip Salzman
> 
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] On Behalf Of Trog
> Sent: Wednesday, May 04, 2005 10:57 AM
> To: ClamAV users ML
> Subject: Re: [Clamav-users] clamav dying & accept() errors in syslog
> 
> On Wed, 2005-05-04 at 16:48 +0100, Nigel Horne wrote:
> > On Wednesday 04 May 2005 16:40, Phillip Salzman wrote:
> > >
> > > --
> > > May  4 13:39:03 lora clamav-milter: ClamAv: accept() returned
> invalid
> > > socket (Result too large), try again
> > > May  4 13:39:28 lora last message repeated 9 times
> > > --
> > >
> > > I'm currently using clamav 0.84e on FreeBSD 4.9-stable.  They are
on
> > > Dell hardware.
> > >
> 
> I believe that's a FreeBSB bug:
> 
> http://www.rhyolite.com/pipermail/dcc/2003/001409.html
> 
> -trog 

Just as an update incase anyone else is having issues, I've upgraded the
machines that were failing to FreeBSD 5.4-STABLE (via installing 5.4-rc4
and rebuilt the source tree from CVS (yesterday's source tree).)

The errors are still in place, and the newer machine that I brought up
which was originally working, is also having the problem now.

=  May  6 13:29:23 jor-l clamav-milter: ClamAv: accept() returned
invalid
=  socket (Result too large), try again

The other machine that was originally fine is also getting the above
message, in addition to a sig11.  So far the sig11 has not appeared on
the jor-l machine (but it was before the 5.4-S upgrade.)

=  May  6 13:44:07 zod kernel: pid 93239 (clamav-milter), uid 106:
exited on 
=  signal 11 

These are pretty heavily loaded machines.  They each handle about 2M
msgs/day, but at the same time they're pretty beefy with lots of
resources (which aren't appearing to be used very much.)

Would someone have any suggestions on tuning clamav for such a load?  I
was using it with clamd and doing clamav-milter --external, but am not
now.  Any suggestions would be very much appreciated.

In order to survive (and not every customer calling) I had to set a 10s
timeout for the milter to talk to the socket.  This sadly means we're
missing viruses, but it allows the SMTP server to keep responding...

Thanks for anything that anyone can provide,


Phillip Salzman
_______________________________________________
http://lurker.clamav.net/list/clamav-users.html

Reply via email to