need opinions about startup script

Bob Langford said the following on 16/09/2004 02:38:

 > Here's the issue: it seems to me that this startup script should
 > call apachectl to do the real work, instead of reimplementing it.

Why? apachectl is just a wrapper. You could say that it's the other way
around (apachectl reimplementing the apache start/stop script).

 > But I've searched for and found many examples of apache startup
 > files, and NONE of them do that -- they all ignore apachectl.

The only daemon that does that (which I know off) is sendmail (at least
on a Debian system), but, ofcourse, sendmail is not Apache.

 > I've learned over the years that this situation will eventually
 > lead to some hard-to-find problem. If the system starts automatically

Can you give an example? I've administered a few Apache boxes on
different platforms and have never encountered a problem with this.

 > one way, and manually another way, it's harder to be SURE that
 > they are the same.

I see what you mean and theoretically you're right. Still, I've not had
problems with it.

 > So, why don't people use apachectl in their system startup scripts?

There are differences in start/stop scripts and wrapper scripts, where
the latter usually has more options. The start/stop scripts are usually
run by /etc/init.d/rc (or something like that), which is controlled by
init, whereas a wrapper script is more a "user" (instead of "daemon") thing.

For Apache the difference might be very little, but there are a few
start/stop scripts which have to run when /usr (and thereby /usr/bin and
/usr/sbin) are not yet mounted (or already u(n)mounted in case of going
down).

You might want to symlink the start/stop script to apachectl as Davide
pointed out (not that he really suggested it), but that may raise other
problems and I would recommend against it.

--
Regards
Harrie<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: need opinions about startup script 

In article <414b2be8$0$21106$e4fe514c@news.xs4all.nl>,
dinges_danges_donges.RemoveThis@hotmail.com says...
 > Bob Langford said the following on 16/09/2004 02:38:
 >
  > > Here's the issue: it seems to me that this startup script should
  > > call apachectl to do the real work, instead of reimplementing it.
 >
 > Why? apachectl is just a wrapper. You could say that it's the other way
 > around (apachectl reimplementing the apache start/stop script).
 >
  > > But I've searched for and found many examples of apache startup
  > > files, and NONE of them do that -- they all ignore apachectl.
 >
 > The only daemon that does that (which I know off) is sendmail (at least
 > on a Debian system), but, ofcourse, sendmail is not Apache.
 >
  > > I've learned over the years that this situation will eventually
  > > lead to some hard-to-find problem. If the system starts automatically
 >
 > Can you give an example? I've administered a few Apache boxes on
 > different platforms and have never encountered a problem with this.
 >
  > > one way, and manually another way, it's harder to be SURE that
  > > they are the same.
 >
 > I see what you mean and theoretically you're right. Still, I've not had
 > problems with it.
 >
  > > So, why don't people use apachectl in their system startup scripts?
 >
 > There are differences in start/stop scripts and wrapper scripts, where
 > the latter usually has more options. The start/stop scripts are usually
 > run by /etc/init.d/rc (or something like that), which is controlled by
 > init, whereas a wrapper script is more a "user" (instead of "daemon") thing.
 >
 > For Apache the difference might be very little, but there are a few
 > start/stop scripts which have to run when /usr (and thereby /usr/bin and
 > /usr/sbin) are not yet mounted (or already u(n)mounted in case of going
 > down).
 >
 > You might want to symlink the start/stop script to apachectl as Davide
 > pointed out (not that he really suggested it), but that may raise other
 > problems and I would recommend against it.
 >
 >
I've learned a bit more about it, and at least on Redhat systems,
the /etc/init.d/* scripts have a specific set of exit statuses that
they are supposed to return, and apachectl does not return the correct
ones, so doing a symlink to apachectl won't work just right (although it
will be very close; maybe close enough for most people). So a wrapper
is called for.

And that's what I'm planning to do. That way, whatever options I need
to define for Apache (-DSSL, etc) can be put in apachectl, and the init
script doesn't have to be kept in sync.

Thanks to those who responded. The Net comes through again!
--
Bob Langford, Silicon Masters Consulting, Inc.
langford.RemoveThis@silicon-masters.com <a style='text-decoration: underline;' href="http://www.silicon-masters.com/" target="_blank">http://www.silicon-masters.com/</a>
phone: 804-674-1253 fax: 804-745-7803<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: need opinions about startup script