User Survey

[Scott James Remnant]

On Sun, 2006-09-03 at 18:48 +0200, Clément Stenac wrote:

> > Define "diverging" ... we're changing init! that's a pretty big,
> > whopping divergence for a start.  It's obviously going to have knock-on
> > effects on everything else.
> 
> Actually, it can be seen as "providing another init, and, in this goal,
> making sure that init can be changed".
> 
Right, that's what the patches so far have been trying to do.  It wasn't
possible to actually replace init with something else (not really
surprising, given Debian Policy specifies sysvinit) so we had to tweak a
few things.

> I don't think we're diverging if we're having a separate upstart
> package, a separate upstart-initscripts package. We are if we maintain a
> separate sysvinit package.
> 
During the migration, we expect to have some scripts in initscripts and
some in an upstart jobs package; this is probably a sequence of changes
Debian don't need to take wholesale.

> Btw, something I haven't really understood. How will dependencies for
> services be handled. If I understand correctly, all the "base"
> initscripts will be upstart jobs and have the correct "dependencies"
> through events.
> 
That's correct.  Upstart isn't a dependency-based init system, unlike
initng, it works purely on events.  The principal difference is that
it's backwards.  A dependency-based init system starts network because
apache depends on it, an event-based init system starts apache because
network has been started.

(Note that the LSB stuff is neither, it simply uses that information to
perform ordering of the boot sequence)

> But how will upstart handle things like "apache needs network to be set
> up"
> 
"start on network" ?

> "tomcat needs postgresql to be up"
> 
"start on postgresql" ?

> Will some upstart initscripts read LSB tags in apache and tomcat's
> standard init scripts or will we have to change these ?
> 
No.  If apache and tomcat wanted to use upstart events, they would have
to have their init scripts replaced with upstart job files.  This is
desirable anyway in the long-run as it means we can provide service
management, etc. for them.

In the short term, the init scripts are run by /etc/init.d/rc as usual
which has no concept of dependencies, and just runs things in order.

The rcS job (which exec's /etc/init.d/rcS) would be run when we have
writable filesystems, so would "finish up" the configuration.

Scott
-- 
Have you ever, ever felt like this?
Had strange things happen?  Are you going round the twist?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/upstart-devel/attachments/20060903/a9bb82f2/attachment.pgp