Source: rsync
Version: 3.1.1-3
Tags: patch
User: la...@debian.org
Usertags: initscript-stop-action
Hi,
rsync's initscript does not check whether the rsync process has actually
terminated before returning successfully.
This can result in a non-deterministic race condition where resources that are
required by the corresponding "start" action (eg. TCP ports, exclusive
filesystem locks, etc.) are not yet available to the new instance because the
older process is still releasing them in the background.
This then results in the new process failing to start which is often hidden
from the sysadmin unless a monitoring system is in use.
For the same/parallel reasons, the stop and restart actions should return a
non-zero exit code if the daemon failed to stop. Note that adding "sleep" calls
merely masks the problem and is not a valid fix.
Patch attached.
Regards,
--
,''`.
: :' : Chris Lamb
`. `'` la...@debian.org / chris-lamb.co.uk
`-
diff --git a/debian/init.d b/debian/init.d
index 3bf5167..3cb0447 100644
--- a/debian/init.d
+++ b/debian/init.d
@@ -111,8 +111,13 @@ case "$1" in
;;
stop)
log_daemon_msg "Stopping rsync daemon" "rsync"
- start-stop-daemon --stop --quiet --oknodo --pidfile $RSYNC_PID_FILE
- log_end_msg $?
+ start-stop-daemon --stop --quiet --oknodo --retry 30 --pidfile
$RSYNC_PID_FILE
+ RETVAL="$?"
+ log_end_msg $RETVAL
+ if [ $RETVAL != 0 ]
+ then
+ exit 1
+ fi
rm -f $RSYNC_PID_FILE
;;
@@ -126,8 +131,7 @@ case "$1" in
if $RSYNC_ENABLE; then
log_daemon_msg "Restarting rsync daemon" "rsync"
if [ -s $RSYNC_PID_FILE ] && kill -0 $(cat $RSYNC_PID_FILE)
>/dev/null 2>&1; then
- start-stop-daemon --stop --quiet --oknodo --pidfile
$RSYNC_PID_FILE || true
- sleep 1
+ start-stop-daemon --stop --quiet --oknodo --retry 30 --pidfile
$RSYNC_PID_FILE
else
log_warning_msg "rsync daemon not running, attempting to start."
rm -f $RSYNC_PID_FILE
