On Monday 23 July 2007, Gene Heskett wrote:
> On Monday 23 July 2007, David Baron wrote:
> >>I mean the obvious stuff like "viagra" and such. Usually the spam is
> >> caught but sporadically it does get through.
> >>
> >>What is happening.

> >
> >Simply, there are no X-Spam headers on these (and none or some of the
> > "ham" as well). In other words, messages are being delivered before sa is
> > running!
> >
> >Originally, I was starting everything by their standard packaged
> > /etc/init.d scripts. All is fine except sa would start before the
> > internet connection was working, sa_update would fail, leaving ALL spam
> > getting through! (To me, this behavior is a bug--no sa_update today?
> > Leave yesterday's rules intact.)
> >
> >So I included in my 99z_end-all-catch-all script (needed because of other
> >things) which tries until ntpdate has worked implying a working internet
> >connection and then ran sa_update and started spamassassin from there. The
> >problem is now a catch 22: Fetchmail and exim are running beforehand and a
> >few messages can get delivered before sa comes up.
> >
> >I could move all of this to my script but there must be a better, more
> > correct and standard way to accomplish this. Any ideas?

> Humm, with my lashup here that Joanne helped me setup, S78spamassassin
> starts a few copies of spamd, and fetchmail is started much later in
> S99local. Its fetchmail that calls procmail, and its procmail that calls
> the spamd's, so there is no time that SA can be bypassed.
> I thought everyone was doing it. Somebodies better idea isn't?

Problem is that the S78 will start spamassassin but that start does not
necessarily get a valid rule-set. For that, the internet connection must be
up at the time. So I moved the spamassassin start to a 99 level script. But
fetchmail can be up before. So I guess the fetchmail start needs to be moved
to after the effective spamassassin start.

Problem is that with every upgrade, those rc#.d files may be restored if I am
not careful :-)