Mailinglist Archive: opensuse-security (320 mails)

< Previous Next >
Re: [suse-security] What's keeps changing my inet.d sequence
  • From: John Andersen <jsa@xxxxxxxxxxxxxx>
  • Date: Thu, 22 May 2003 00:08:26 -0800
  • Message-id: <200305220008.26033.jsa@xxxxxxxxxxxxxx>
On Wednesday 21 May 2003 22:15, Arjen de Korte wrote:
> > If webmin won't start before the network is up, then accepting
> > the default of starting it before will yield a 100% failure rate.
> From the startup script of webmin-1.070-23:
> # Provides: webmin
> # Required-Start: $network $syslog
> # Required-Stop: $network
> # Default-Start: 2 3 5
> # Default-Stop: 0 1 6
> # Description: Start or stop the Webmin server
> There. Webmin already depends on $network, so it will not be started before
> the network startup script is run.

Those comments are ONLY used to set the order of the links, they do not
affect anything at the time that the services in the runlevel (including
webmin) are actually started.

Therefore, if you see the webmin script linked into a runlevel BEFORE
network, then It will still fail - regardless of what the comments say.

Thats when you have to find out WHY insserv linked it into the runlevel
in the wrong order, - and failing to get insserv to behave, you have
to change the order yourself.

As you know, the alphabetical order of the links
in the /etc/init.d/rc?.d directory sets the order they are run. If those
are wrong the Optimum way you make them right is setting the Required-Start
list to include network and running insserv.

insserv then attempts to change the order - if it condescends
to work at all.

In the Webmin example the Original Poster said his order
started webmin prior to network, so clearly insserv failed to
set the order in his case.

I have a similar case.

In my case, insserv insists on starting Shorewall AFTER vmware
in spite of my having vmware as a pre-requsite to shorewall.

That would be bad enough, I would just do the arangement the
old fassion way (by hand) EXCEPT THAT something in the
suse environment keeps undoing it. THATS the bit I was
questioning. How to suppress that behavior.

But the METHOD of changing the names (hence the order) of the
runlevel scripts is all we are discussing, the fact that you MUST
change the script name to change the order is a given.

You see, this differes a great deal from
your assertion that changing the order of the scrips (by virtue of
the name change) is, in your opinion, unjustified. Changing the
order is necessary, normal, and expected.

John Andersen

< Previous Next >