Victor Duchovni:
> On Fri, Sep 11, 2009 at 02:53:17PM -0400, Victor Duchovni wrote:
>
> > On Fri, Sep 11, 2009 at 11:49:03AM -0700, Quanah Gibson-Mount wrote:
> >
> > >> Don't forget that sendmail(1) is outside the persistent master-managed
> > >> Postfix service and can en-queue mail (to disk) for later delivery, even
> > >> when Postfix is *not* running (provided it does run eventually...).
> > >
> > > Yeah, cron is calling the system sendmail, which is what is generating
> > > the
> > > mail in /var/spool/postfix/maildrop. For some reason, this particular
> > > system has nothing under /var/spool/postfix/public. I created a pickup
> > > FIFO and set the ownership/group accordingly, and now that error no
> > > longer
> > > shows.
> >
> > The solution is not to manually create FIFOs. Rather, you should either
> > run a second Postfix service, or replace the system sendmail(1) with
> > a symlink to the custom-built binary.
>
> I should also mention, that you could simply change the queue_directory
> setting in the otherwise disused default /etc/postfix/main.cf, and then
> the system "postdrop" will queue mail into your zimbra installation.
Better symlink the system sendmail to the "running" Postfix.
I would not recommend mixing sendmail/postdrop programs from a
newer Postfix release with the pickup daemon from an older Postfix
release. The other way around is rarely a problem as newer software
has to understand the queue file produced by older software.
Wietse
