hi ya peter
okay... I see it now...
I just checked /etc/rc.d/init.d/autofs ( yup redhat box )
...
stop)
#
# change -TERM to -USR2 here ?
#
--->> kill -TERM $(/sbin/pidof /usr/sbin/automount)
rm -f /var/lock/subsys/automount
;;
reload|restart)
....
while read pid tt stat time command; do
echo "$command" >>$TMP2
if ! grep -q "^$command" $TMP2; then
--->> kill -USR2 $pid
echo "Stop $command"
fi
done
)
....
#
# I'd add this to it
#
force_kill)
echo "AutoFS Going down...and might NOT come back up..."
sleep 10 ;
--->> kill -TERM $(/sbin/pidof /usr/sbin/automount)
rm -f /var/lock/subsys/automount
;;
guess we can change those options around ??
so as it is... autofs stop will kill autofs...
( which is what I usually do )
and "autofs reload" will try to reload only the changes
to auto.master/auto.misc
guess one could experiment with changing -TERM to -USR2
to stop it from shutdown completely and not being able to restart ?
thanx
alvin
> > > > yes... if you could...please use some kind of detection schem
> > > > where autofs will refuse to shutdown and just flag that /home
> > > > is in use.... instead of going down anyway...and not be able
> > > > to restart to remount the other dirs that the users now wants
> > > > but can't get since autofs won't come back up without killing
> > > > all the old processes...
> > > >
> > >
> > > Sure; currently that's what autofs uses if you send it SIGUSR1; would
> > > it make sense if I made it behave that way for SIGTERM as well?
> >
> > Am not 100% sure...but doesn't SIGTERM kill the process ??
>
> SIGTERM is the "normal" kill (not kill -9, which is SIGKILL).
> Software can, and does, trap SIGTERM for a clean shutdown. In the
> case of autofs, it umounts what it cans and then puts autofs into
> catatonic mode.
>
> > All I'm wondering is autofs flags that it cannot umount a used mount point
> > and merrily continues it's way out....and cannot get autofs restarted...
> >
> > - so am, wondering, can we give it an option to force it to continue
> > and shutdown like it does now, knowing that it ( autofs) won't come back up
> > .. the other option is to NOT shutdown if it finds "used mountpoints"
>
> You already have this option: SIGTERM (and SIGQUIT) does the former,
> and SIGUSR2 (not SIGUSR1 as I said before) does the latter.
>
> Perhaps it would make more sense to have SIGTERM do the latter (I
> presume SIGTERM is what the rc scripts send) and only have SIGQUIT do
> the former?
>
> > - I have this problem with /usr/local and don't really want to
> > shutdown the machine... and similarly if /var/spool/mail was
> > unmounted and autofs shutdowns when no one was reading mail...
> > and we're stuck with whatever used mountpoints we have...
> > no /var/spool/mail... which is not critical .. just annoying..
> >
> > - if I play with /home....than all users are kicked off anyway
> > and the machine rebooted to test the changes...
>
> -hpa
>
