[Samba] samba-ad restart fails occasionally

[mj]

Hi,

We are running sernet samba, and on one particular DC (debian 7.11, 
samba 4.5.6), when logrotate is ready rotating, "sernet-samba-ad 
restart" fails with:

> Shutting down SAMBA AD services : ...trying once more ... (warning).
> ...trying once more ... (warning).
.....
> ...trying once more ... (warning).
> Error: /usr/sbin/samba still running with PID=14755 from /var/run/samba/samba.pid ... failed!
> Starting SAMBA AD services : Warning: /usr/sbin/samba already running ! ... (warning).

So after all the script tries to start samba again, but since there 
appeared to be a process remaining, it fails to actually start, and 
this causes all kinds of failures:

- Replication fails with (WERR_CONNECTION_REFUSED)
- LDAP queries to that dc fail with query error: Transport endpoint is 
not connected

Rebooting the DC solves everything, but scheduling a reboot every 
morning is not an elegant solution. :-)

Here is the part of the sernet-samba-ad script where the restart occurs:

> 	PID=$(cat ${PIDFILE})
>
> 	if ! (readlink /proc/${PID}/exe | grep -q "^${BINARY}") ; then
> 		log_warning_msg "Warning: ${BINARY} not running with PID=${PID} from ${PIDFILE} ! "
> 		exit 0
> 	fi
>
> 	kill -15 ${PID}
> 	for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30; do
> 		sleep 3
> 		kill -0 ${PID} >/dev/null 2>&1 || break
> 		log_warning_msg "...trying once more "
> 		kill -15 ${PID}
> 	done
>
> 	kill -0 ${PID} >/dev/null 2>&1 || {
> 		log_success_msg ""
> 		rm -f ${PIDFILE}
> 		exit 0
> 	}

Hw dangerous would it be to replace the last "kill -15 ${PID}" with 
"kill -9 ${PID}"?

Does anyone have some nice suggestions how to solve this?

MJ