Public SourceMage Discussion List

[Seth Alan Woolley <seth AT positivism.org>]

On Wed, Aug 04, 2004 at 12:18:44PM +0200, Robert Helgesson wrote:
> On Wed, Aug 04, 2004 at 11:15:31 +0200, Rohan Nicholls wrote:
> > I did all that and it took a whole day, and the next day when I booted
> > to smgl it was happy, ran through loading the kernel, through runlevelS
> > and when it went to runlevel2 it had success getting networking going
> > and then froze (not in the crash sort of way, the keyboard responded but
> > it seemed as though it was waiting for something); Ctrl-c would not make
> > it skip to the next task, and the only thing that had any effect was
> > hitting Ctrl-Alt-Delete twice at which point it went into the shutdown
> > sequence.
> 
> I suspect it's a problem with the remote filesystem scripts, when
> installing init.d did you allow installation of smgl-default-remote-fs?
> Did you also allow it to be the default provider of the remote_fs
> facility? You can check this in /etc/sysconfig/facilities.
> 
> If this isn't the problem you can get a more verbose boot process by
> adding "set -x" just above the case statement at the end of
> /etc/init.d/rc, this should allow you to see where the bootup halts.

I've recently set init scripts to be default not enabled unless it is
marked ESSENTIAL=yes or the user says 'y' to it at the prompt.  Does
smgl-default-remote-fs need an essential on it?  If we need some stuff
installed by default while not having the essential 'logic' being
engaged, we might need another variable like DEFAULT_ENABLED=yes to
uncouple the two concepts of "if this fails, it is fatal" and "we really
do want this enabled, but it's no big deal if it fails."  Of course,
ESSENTIAL should always imply DEFAULT_ENABLED if we do go this route.

Is this needed?

Currently all init scripts are installed.  We just don't enable them all
right away without user intervention.  This leaves service starting and
stopping really up to the System Administrator except in cases where the
system would be unbootable otherwise (hence the ESSENTIAL mark on some
fundamental scripts).

On another issue...

It seems like this problem that is showing now is due to a circular
dependency in the init process acording to the wiki.  Is there some
reason for the circular dependency?  Usually legitimate circular
dependencies come up because an optional dependency has been turned on
in a second pass (something like post-invocation optional features that
require a reinvocation such as esgps/gimp-print/cups).  Anything else is
probably an illegitimate circular dependency and should be fixed, and
I'm not even sure why any init script would need a circular loop because
since we are writing the init scripts we can always write out the need
for reinvokation (unlike the printer spell example which requires a
recompile).  In other words, there's got to be a good solution to this
problem, even if we can't detect circular dependencies without rewriting
simpleinit to detect when 'a needs (b needs ...) c' and 'c needs a (or b
...)'.  Hopefully we can keep simpleinit simple.

> 
> > While I was wondering around the scripts I see that sourcemage has gone
> > for a simpler init solution than the sysV one, more leaning towards the
> > bsd one which I MUCH prefer.  I am about to visit the wiki and see if I
> > can figure out what is called when in the way of services and how.  So
> > just wanted to say nice choice for the init, as it actually looks
> > comprehensible to the likes of l'il ol' me.:-)
> 
> I'm glad you like the system. There is some information[1] about the
> init system in the wiki, most of it is in the design documentation.
> 
> [1] http://wiki.sourcemage.org/index.php?page=Init+System

Yeah, it's elegent; it's also exhibiting emergent behavior as well. ;)

I'm wondering if maybe parallel init might fix or be a contributor to
the problem.  I'm doubting it (on theoretical grounds), but it would be
interesting to see if the problem goes away in different init modes.

Seth

-- 
Seth Alan Woolley [seth at positivism.org], SPAM/UCE is unauthorized
Key id EF10E21A = 36AD 8A92 8499 8439 E6A8  3724 D437 AF5D EF10 E21A
http://smgl.positivism.org:11371/pks/lookup?op=get&search=0xEF10E21A
Security Team Leader Source Mage GNU/Linux http://www.sourcemage.org

Attachment: pgpWakpoeuDNN.pgp
Description: PGP signature