systemd for 11.10 ?

[Steve Langasek]

On Mon, May 09, 2011 at 10:58:24AM -0500, Patrick Goetz wrote:

> My point about learning a new system was based on the observation
> that there still seem to be a number of kinks in upstart; in
> particular, we've had timing problems with autofs and statd, both of
> which intermittently refused to start until we tweaked the upstart
> scripts.

Thanks to Clint Byrum's persistence, I believe the issues with statd have
been fixed in natty; if they haven't already been backported to 10.04, we
should certainly do so.

I'm not aware of the status of autofs.

> Also, the fact that packages like apache still don't have an upstart
> script indicates the technology isn't fully mature.

I don't agree.  It mostly indicates that the transition is incomplete,
because it takes time and thought to convert an init script to an upstart
job (and we are not yet able to share this workload with Debian).  This is
why sysvinit scripts are still supported in upstart for compatibility.

> This is very funny, but there are some points raised in Lennart's
> comparison:

>   http://0pointer.de/blog/projects/why.html

> that merit careful reflection; e.g. shell-free bootup,
> mount/automount handling, and disabling of services without editing
> files, to name a few.

Shell-free bootup is not an end unto itself; the end goal is to boot the
system efficiently.  As such, I think there's an open question about whether
shell-free bootup isn't a false optimization.  Yes, it potentially makes the
system boot a little bit faster (fewer subprocesses to execute == less
overhead), but at the cost of making it *much* more difficult for the admin
to adjust the boot-time behavior when needed.

Including mount handling in this list is deceptive.  upstart does not handle
mounting, because we have a separate process, mountall, which does so[1].  I
believe the current thinking is that mountall should eventually be merged
into upstart itself; nevertheless, saying that systemd is better than
upstart because it handles mounting directly is a bit like saying that
sendmail is better than postfix because queue management is built in instead
of having to run a separate executable.  There are so many other factors
that contribute to whether one architecture or the other is appropriate that
such a one-liner comparison is meaningless.

> No one, however, can tell me that the accepted method for disabling
> services in upstart isn't a kludge!  <:)

This is an admitted weakness of upstart currently, but hardly an
insurmountable one.  I can certainly understand how this might be a deciding
factor for a sysadmin about which one they prefer to *administer* - but it's
not the sort of thing that should determine what a distro will *ship*,
because addressing this doesn't really call for changes to the design, just
for providing appropriate tools for admin use.  (The current version of
upstart introduces support for job override files that can be updated
without need to modify the .conf files on disk, if that's the actual concern
here.)

-- 
Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
Ubuntu Developer                                    http://www.debian.org/
slangasek at ubuntu.com                                     vorlon at debian.org

[1] Lennart at least acknowledges this in a footnote.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 828 bytes
Desc: Digital signature
URL: <https://lists.ubuntu.com/archives/ubuntu-devel/attachments/20110510/f3b41510/attachment.pgp>