problem with insserv and bootchart, bootlogd...
pere at hungry.com
Fri Dec 12 16:33:16 UTC 2008
> Following the last message on debian-devel-announce, I've tested
> insserv on a Debian Lenny system, to be able to start processes
> concurrently, as some bootcharts done with bootchart package shown
> some unused time during boot of my laptop (Thinkpad X31).
Hi. Great to hear another test report.
> About bootchart, it's already signaled in bug 491391, but I didn't
> find any valuable clue to correct this. The stop-bootlogd should
> also be at the end.
I do not remember the details of these scripts, but guess the
dependencies should be adjusted. Not sure the bootlogd stuff is
working at all. It has not been maintained for a few years.
> There is a conflict between bootchart and stop-bootlogd dependencies
> on rmlogin, and rmnologin depending of pseudo-task $all. I wonder if
> this latter is a good choice, does rmnologin should depend on gdm
> instead of being the last "service" started ? does gdm works before
> rmnologin hasn't started ?
It might be a good idea to adjust the dependencies of rmnologin. I
guess it is a matter of policy if it should only depend on $remote_fs
to allow early login, or if it should depend on some later scripts or
$all to block logins until the boot is complete.
> About the parallel start with CONCURRENCY variable, I wonder which
> are the differences between 'shell' solution and 'startpar' ? the
> former seems much simpler than the latter...
The startpar option serialize the boot messages, while the shell
solution do not. In addition, the startpar program is capable of
ordering scripts using the dependency information to allow for more
parallelization. This is not yet used in Debian, but I hope we can
enable it in the future.
More information about the initscripts-ng-devel