Resolution of "very urgent problem with MailScanner & Postfix..."
ree at THUNDERSTAR.NET
Mon Nov 10 19:47:01 GMT 2003
Thanks, this turned out to be a missing config in spam.assassin.prefs.conf
with the path to the dcc executable.
On Fri, 7 Nov 2003, Ugo Bellavance wrote:
> > -----Message d'origine-----
> > De : Ron E. [mailto:ree at THUNDERSTAR.NET]
> > Envoyé : Friday, November 07, 2003 6:35 PM
> > À : MAILSCANNER at JISCMAIL.AC.UK
> > Objet : Resolution of "very urgent problem with MailScanner &
> > Postfix..."
> > I thought I had better let everyone know what I cam up with after all
> > this.
> > Quick recap: Running MailScanner & Postfix for months,
> > suddenly earlier
> > this week a backlog started accumulating and could only be halted by
> > temporarily stopping the incoming postfix, letting
> > MailScanner catch up,
> > then starting it back up again. Found out that this was fixed in the
> > latest stable build of MailScanner but didn't upgrade yet. (Note that
> > along with the backlog was the problem where duplicate truncated
> > messages were being delivered, discussed earlier on this
> > list. - "still
> > being delivered" error, etc)
> > Anyway, no problems until late in the week when another backlog began
> > accumulating. Didn't mess around at this point, but upgraded.
> > Unfortunately the backlog continued to grow until again I stopped the
> > incoming postfix to allow mailscanner to catch up. Did things
> > like turned
> > off RBLs do help this along. Started incoming postfix again
> > at which point
> > another backlog immediately started accumulating. Inspection
> > of the logs
> > showed mailscanner taking quite awhile on each message. Various
> > performance settings tweaked, message batch reduced to 10
> > from 100, and
> > changed "Max Normal Queue Size" to 500 (from 5000).
> > Note that throughout all this the other oddity was that a service
> > MailScanner restart would stop everything fine but then
> > MailScanner would
> > not start - another service MailScanner start would be needed
> > to get it
> > going again. I finally noticed what I should have noticed originally -
> > there were too many mailscanner processes running (I had 5
> > max children
> > set). I believe this may have been why I was still having a
> > problem with
> > the "still being delivered" error. I killed those of course and that
> > handled the starting problem.
> > Next I started running debug mode, again something I should have done
> > earlier. Found that razor, pyzor and dcc were not working.
> > Razor seemed to
> > be causing the biggest delay, followed by pyzor. Dcc was just
> > erroring.
> > Found that despite doing the usual server discovery for razor
> > & pyzor, the
> > servers were not updating for some reason. Managed to get ahold of the
> > correct server IPs and then all of a sudden pyzor & razor
> > started working.
> > Hopefully all this will be of use to someone else who may run into a
> > similar situation.
> > The last unresolved thing is dcc - if anyone has any ideas on
> > fixing this,
> > I'd be very interested. The errors I get are:
> > no executable dccproc found
> > DCCifd is not available: no r/w dccifd socket found
> It means that you dont run dccifd, but it usually turns automatically to dccproc.
> Do you see something like dccproc available just after?
> I have a procedure to use dccifd instead of dccproc if you want.
> >From what I found, dccifd is some kind of daemon, so the 8 MB of code is loaded in memory and stays there.
> If you use dccproc, the code is loaded each time.
> > I checked, though, and there is an executable dccproc in
> > /usr/local/bin/dccproc
> > Perhaps spamassassin just doesn't know where to find dccproc?
> > I installed the latest version but no change.
> > In any case, thanks to all who assisted, particularly Julian.
> > Regards,
> > Ron
More information about the MailScanner