Re: [vps-mail] filtering non local account with spamassassin and clamav

["Bennett Lanford" <ben@xxxxxxxxxxxxx>]

ADNET Ghislain <gadnet@xxxxxxxx> said:

> > I haven't seen the ClamAV milter in ports (but it might be there now). It 
> > used to require the next version of the gcc compiler. I ran ClamAV as a 
> > milter for awhile, but after running for awhile, it started reporting 
> > out-of-memory errors. (I think it might have a memory leak somewhere ...)
> 
> Very interesting. Do a restart of sendmail solve the memory problem or
> do you had to reboot the vps ?

The memory leak isn't in sendmail. Actually, MIMEDefang was reporting it. (I 
had daemons running for ClamAV, SpamAssassin (spamd), and MIMEDefang.) Things 
worked fine for nearly a month before the problem showed up. 

I'm wondering if it might have been a problem with a particular virus 
signature. (My ClamAV virus signatures are updated periodically via cron.) 
Maybe one of the new signatures caused ClamAV to loop or recurse and not free 
up memory? ...) For the moment, I've disabled MIMEDefang, and am back to using 
procmail to call SpamAssassin, with procmail rules to match viruses ... (... 
But I will return later! If nothing else, I will use MIMEDefang without having 
it call ClamAV. MIMEDefang also "defangs" incoming mail by making attached 
viruses non-executable, if you tell it to ...)

> > You can also run both of the above with the MIMEDefang milter. I ran that
> > setup for several weeks ... then started getting out-of-memory errors. (I
> > *think* it was called by MIMEDefang's calls to ClamAV ...)
> 
> this seems to be something to watch in the milter thing, memory leak
> then, thanks for the info.

I think the sendmail side of milters is rock solid. The uncertainty is in the 
milters (other programs that register themselves with sendmail) themselves: 
some are better than others. 

-- 
Bennett Lanford
ben@xxxxxxxxxxxxx


======================================================================
This is <vps-mail@xxxxxxxxxxxx>       <http://www.perlcode.org/lists/>
======================================================================