0.6.3 experiment(s)

Janne Karhunen janne.karhunen at gmail.com
Wed Oct 21 14:53:57 BST 2009


Hi,

We (maemo.org team) have been an old user of the 0.3 series and
have managed to run upstart quite nicely; it just works. Really nice.

I just started experimenting with new 0.6 stable series to see how
things will work out this time. Looks like the old scripts pass the
config parser with just a few modifications and things are looking
good (we're not using any of the sysv-compat stuff; startup scripts
are mostly native to upstart).

However, with 0.6.3 things crash and burn very early on bootup to
very simple script exec I could not instantly explain. Initial startup
sequence is roughly as follows:
kernel --> preinit --> upstart(init) --> sysconfig-early -->

Sysconfig-early in this case looks simply;

start on startup
console output
pre-start script
  echo "echo something"
end script
exec /etc/init.d/sysconfig-early

Construct is more than simple enough and works like a charm on
0.3 series. Running this with 0.6.3 yields following output on debug
(with startup event still being startup);

..
event_new: Pending startup event
Handling startup event
event_pending_handle_jobs: New instance sysconfig-early
sysconfig-early goal changed from stop to start
sysconfig-early state changed from waiting to starting
event_new: starting
event_new: Pending starting event
Handling starting event
event_finished: Finished starting event
sysconfig-early state changed from starting to pre-start
Kernel panic - not syncing: Attempted to kill init!

Now, any wild guesses what I'm looking at? As it never even
makes it to the pre-start script I would assume something
goes badly haywire while setting up the job. Some environment
changes perhaps?


-- 
// Janne



More information about the upstart-devel mailing list