On Thu, Dec 07, 2006 at 04:10:06PM +0100, Bas van Schaik wrote:
Just like stated in the TODO.Debian file: the package is missing an
initscript to automatically export devices at boot time.
Some comments:
#!/bin/sh
# Start or stop one or more vblade-processes, including a logger process
#
# Bas van Schaik [EMAIL PROTECTED]
# Inspired by the init.d script from the Postfix package
No set -e?
PATH=/bin:/usr/bin:/sbin:/usr/sbin
DAEMON=/usr/sbin/vblade
NAME=vbladed
CONFIGFILE=/etc/vbladed.conf
It's more like a table of blade devices, much like fstab, crypttab,
raidtab, etc. Maybe calling it vbladetab would make more sense,
consistency-wise?
PIDSFILE=/var/run/vbladed.pids
test -f /etc/default/vbladed . /etc/default/vbladed # include
/etc/default/vbladed if it exists
Is that really needed? By default vblade can ship with an empty vblade
config, and then is already no-action. Having yet another config file to
change if you want to start using it seems a bit pointless here.
test -f $CONFIGFILE || exit 0 # exit if configfile doesn't exist
test -x $DAEMON || exit 0 # exit if daemon is not installed
if [ -f /etc/default/vbladed ]; then
. /etc/default/vbladed
This is the second time you include /etc/default/vbladed, anyway.
if [ ! $ENABLE_VBLADED = Y ] [ ! $ENABLE_VBLADED = y ]; then
# vbladed disabled in /etc/default/vbladed
exit 0
fi
fi
# Expects a file which will return a normalized version of the configfile
# (i.e. without comments and with double spaces trimmed out)
function normalizeConfig(){
configfile=$1
if [ ! -r $configfile ]; then
return
fi
cat $configfile | while read ln; do
ln=`echo $ln | grep -v ^# | tr -s | cut -d # -f 1`
if [ ! $ln = ]; then
echo $ln
fi
done
}
case $1 in
start)
echo Starting vblade daemon(s):
# parse /etc/vbladed.conf
counter=0
normalizeConfig /etc/vbladed.conf | while read ln; do
eth=`echo $ln | cut -d -f 1`
shelf=`echo $ln | cut -d -f 2`
slot=`echo $ln | cut -d -f 3`
bdevice=`echo $ln | cut -d -f 4`
Why not 'read eth shelf slot device', which does basicly the right thing
already?
if [ $eth = ] || [ $shelf = ] || [ $slot = ] ||
[ $bdevice = ]; then
echo malformed configuration line: \$ln\
Maybe you want to bail out with an error at this time?
else
echo -n $bdevice ($eth, shelf $shelf, slot $slot):
# Create temporary directory with named pipe
npipedir=`mktemp -d`
npipe=$npipedir/temp_pipe
mkfifo $npipe
# launchDaemon $shelf $slot $eth $bdevice | logger -t
vbladed[$pid] ## werkt
(
$DAEMON $shelf $slot $eth $bdevice 21
$npipe
This will not do what you expect, it'll *not* redirect stderr to $npipe,
but instead leaves it at what stdout used to point to (you're copying fd
1 to postion 2 before you've pointed fd 1 to $npipe).
pid=$!
logger -t vbladed[$pid] -t vbladed[$pid]
$npipe
Duplicate -t?
echo $pid $PIDSFILE
You do not actually clear the $PIDSFILE at the beginning of start, nor
do you guard for starting twice. Perhaps it'd be best to have one
pidfile for each vblade instance, by systematically name the .pid-files
like /var/run/vblade.$eth.$shelf.$slot.pid ? Then you can here detect
that such pid file already exists and say not starting because already
started.
# It's safe to remove the temporary directory
now
rm -r $npipedir
) echo OK || echo FAILED
The exit status of the (...) is determined by the last command, here 'rm
-r'. This is not really representing whether the vblade daemon was
succesfully launched. Also, because this happens in a loop for each
vblade instance, it might be more insightful for the admin to name the
bdevice that's being exported each time that one blade is started, or
similar (see Debian Policy 9.4, If the script starts one or more
daemons (...), http://www.debian.org/doc/debian-policy/ch-opersys.html#s9.4)
fi
done
;;
stop)
# Read file with PIDs
echo -n Stopping vblade daemon(s):
if [ ! -r $PIDSFILE ]; then
echo not running?
else
pids=`cat $PIDSFILE`
for pid in $pids; do
echo -n $pid
kill $pid /dev/null
done
echo
