[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Sendmail milter SPF (was: run levels (priority) and chkconfig --add)



On Wed, 2004-10-20 at 23:47, Alexander Dalloz wrote:
> Am Do, den 21.10.2004 schrieb Paul Howarth um 0:34:
> 
> > > Oct 17 09:37:47 kjc386 sendmail[2296]: i9HDblkJ002296: Milter 
> > > (spf-milter): error connecting to filter: Connection refused by 
> > > /var/spf-milter/spf-milter.sock
> > > Oct 17 09:37:47 kjc386 sendmail[2296]: i9HDblkJ002296: Milter 
> > > (spf-milter): to error state
> > > Oct 17 09:37:47 kjc386 sendmail[2296]: i9HDblkJ002296: Milter: 
> > > initialization failed, temp failing commands
> > > 
> > > And the email is then bounced (temporarily) by sendmail.
> > > 
> > > > Does the socket get created in /var/spf-milter?
> > > 
> > > Not sure, I haven't looked whilest it was dead.  It certainly exists 
> > > after I restart it.
> > 
> > It shouldn't be there when the milter is dead, but it should be there
> > when it's just started, before sendmail is started.
> 
> But if the socket is missing when Sendmail comes up there will be an
> error warning in the maillog produced by Sendmail.

Hmm, true.

> > Could be an "unsafe socket" problem; make sure that none of the
> > directories / /var /var/spf-milter are world-writable.
> 
> Then too Sendmail would complain in the maillog.

True again.

It's strange that no error message is produced when the milter dies.
Usually sendmail is quite forthcoming about things it doesn't like.

> It could be a problem that an old socket file resides and the SPF milter
> server startup does not fully create a new socket file. You could
> prevent this situation by writing in the start() function of the SPF
> milter init script a line
> 
> rm -f /var/spf-milter/spf-milter.sock
> 
> right before the daemon line with which the milter is started.

The perl milter (http://spf.pobox.com/sendmail-milter-spf-1.41.pl)
unlinks any pre-existing socket as part of its start-up procedure. Not
sure about the C version, but I've not seen that problem with that
version.

> (Wouldn't be a more common location for the socket - and other milter
> files - /var/lib/spf-milter/ or /var/spool/spf-milter/?)

Or /var/run/spf-milter. It's there basically because that's where Mark
(who wrote the perl milter) put it and it wasn't trivial to move it
elsewhere.

The C spfmilter package has the socket in /var/run/spfmilter

Paul.
-- 
Paul Howarth <paul city-fan org>


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]