ha/service-mgmt/sm/scripts/sm

#! /bin/sh
#
# Copyright (c) 2014 Wind River Systems, Inc.
#
# SPDX-License-Identifier: Apache-2.0
#
# chkconfig: - 84 84
# processname: sm
# description: Service Management Engine
#
### BEGIN INIT INFO
# Description: sm
#
# Short-Description: Service Management Engine.
# Provides: sm
# Required-Start: $network
# Should-Start: $syslog
# Should-Stop: $null
# Required-Stop: $network
# Default-Start: 3 5
# Default-Stop: 0 6
### END INIT INFO

. /etc/init.d/functions

RETVAL=0

SM_NAME="sm"
SM="/usr/bin/${SM_NAME}"
SM_PIDFILE="/var/run/${SM_NAME}.pid"

if [ ! -e "${SM}" ] 
then
    logger "${SM} is missing"
    exit 5
fi

PATH=/sbin:/usr/sbin:/bin:/usr/bin:/usr/local/bin
WORKER_RESERVED_FILE=/etc/platform/worker_reserved.conf
SM_PLATFORM_CORES_FILE=/var/run/sm/.platform_cores

export PATH

case "$1" in
    start)
        if [ -f $WORKER_RESERVED_FILE ]; then
            . $WORKER_RESERVED_FILE
            mkdir /var/run/sm -p
            echo $PLATFORM_CPU_LIST > $SM_PLATFORM_CORES_FILE
        fi
        sm_args=""
        if [ "`/usr/bin/facter is_virtual`" = "true" ]
        then
            sm_args="--interval-extension 10 --timeout-extension 30"
        fi

        echo -n "Starting ${SM_NAME}: "
        c=0
        p=$(pidof ${SM})
        # pidof /usr/bin/sm return 2 pids. When SM main process is killed
        # subprocess id is returned until the subprocess goes away.
        # calling start-stop-daemon --start too early will fail
        #
        # Add a loop below to wait up to 10 seconds for sub process to finish.
        # Sub process terminates in around 5 seconds by itself, try killing it.
        # A slight longer wait time to see main process goes away is not a concern,
        # as if the SM is actually running  (possibly started by pmon or systemd),
        # sm is already functioning.
        c=0
        p=$(pidof ${SM})
        while [[ $c -lt 10 && ${p} ]]
        do
           logger "SM waiting ${p}"
           if [[ $(echo ${p} | grep "^[0-9]*$") ]]; then
              # only subprocess running, try killing it
              kill -9 ${p}
           fi
           sleep 1
           c=$(( ${c} + 1 ))
           p=$(pidof ${SM})
        done

        running=0
        if [[ "${c}" == "10" ]]; then
           if [[ $(echo ${p} | grep "^[0-9]*$") ]]; then
              kill -9 ${p}
           elif [[ $(echo ${p} | grep "^[0-9]* [0-9]*$") ]]; then
              running=1
              logger "pidof ${SM} still  running."
              RETVAL=0
           fi
        fi

        if [[ "${running}" == "0" ]]; then
           start-stop-daemon --start -b -x ${SM} -- ${sm_args}
           RETVAL=$?
        fi

        if [ ${RETVAL} -eq 0 ]
        then
           echo "OK"
        else
           echo "FAIL ${RETVAL}"
           RETVAL=1
        fi
        ;;

    stop)
        echo -n "Stopping ${SM_NAME}: "
        if [ -n "`pidof ${SM}`" ]
        then
           killproc ${SM}
        fi

        SHUTDOWN_TIMEOUT=20
        count=0
        while [ ${count} -lt ${SHUTDOWN_TIMEOUT} ]
        do
           pidof ${SM} &> /dev/null
           rc=$?
           if [ ${rc} -eq 1 ]
           then
              echo "OK"
              break
           fi
           count=`expr ${count} + 1`
           sleep 1
        done

        pidof ${SM} &> /dev/null
        rc=$?
        if [ ${rc} -eq 0 ]
        then
           echo "FAIL"
           RETVAL=7
        fi

        rm -f ${SM_PIDFILE}
        ;;

    status)
        pid=`cat ${SM_PIDFILE} 2>/dev/null`
        if [ -n "${pid}" ]
        then
           if ps -p ${pid} &>/dev/null
           then
              echo "${SM_NAME} is running"
              RETVAL=0
           else
              echo "${SM_NAME} is not running but has pid file"
              RETVAL=1
           fi
        else
           echo "${SM_NAME} is not running"
           RETVAL=3
        fi
        ;;

    restart)
        pid=`cat ${SM_PIDFILE} 2>/dev/null`
        if [ -n "${pid}" ]
        then
           kill -USR2 ${pid}
        fi
        $0 stop
        sleep 1
        $0 start
        ;;

    reload)
        pid=`cat ${SM_PIDFILE} 2>/dev/null`
        if [ -n "${pid}" ]
        then
           echo "${SM_NAME} reload"
           kill -HUP ${pid}
        fi
        ;;

    force-reload)
        echo "${SM_NAME} force-reload"
        $0 restart
        ;;

    *)
        echo "usage: $0 { start | stop | status | restart | reload | force-reload }"
        ;;
esac

exit ${RETVAL}