rudder-agent init script with PID check, loop to force stop

rudder-agent_init_scriptv2.sh

#!/bin/bash

#====================================================================
# Start/stop script for CFEngine Community
# (http://www.cfengine.com).
#
# chkconfig: 2345 85 15
# description: CFEngine Community
#
### BEGIN INIT INFO
# Provides:          rudder-agent
# Required-Start:    $remote_fs $syslog $all
# Required-Stop:     $remote_fs $syslog
# Should-Start:      $network $time
# Should-Stop:       $network $time
# Default-Start:     2 3 4 5
# Default-Stop:      0 1 6
# Short-Description: CFEngine Community
# Description:       CFEngine Community init script provided by Normation
### END INIT INFO
#
# Copyright (C) 2010 Normation
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# GPL License: http://www.gnu.org/licenses/gpl.txt
#
#====================================================================

#====================================================================
# Changelog
#====================================================================
# Version 0.1 (12/2010):
# - First version
# Author: Jonathan Clarke <[email protected]>
# Version 0.2 (07/2013):
# - Add support for checking the "disable-agent" flag file
# Author: Jonathan Clarke <[email protected]>
#====================================================================

# Source environment variables from /etc/profile
. /etc/profile

# Array indexes
CFEXECD=0
CFSERVERD=1
CFMONITORD=2
CFENGINE_COMMUNITY_NAME[$CFEXECD]="cf-execd"
CFENGINE_COMMUNITY_NAME[$CFSERVERD]="cf-serverd"
CFENGINE_COMMUNITY_NAME[$CFMONITORD]="cf-monitord"

#====================================================================
# Default parameters (if /etc/default/{script_name} is not present)
#====================================================================
# Script specific
PROG_NAME=`basename $0 | sed 's/^[KS][0-9][0-9]//'` # For nice messages
OS=`uname -s`   # To adapt message printing
MYUID=`id -u`     # For UNIX compatibility => modify this command
MYGID=`id -g`     # For UNIX compatibility => modify this command
PS_COMMAND="ps -efww"  # This ensures full width for ps output but doesn't work on Solaris - use "ps -ef"
SYSLOG_FACILITY="local6"

# CFEngine Community directory and files
CFENGINE_COMMUNITY_PATH="/usr/local/cfengine"
CFENGINE_COMMUNITY_VAR_PATH="/var/cfengine"
CFENGINE_COMMUNITY_RUN[$CFEXECD]="1"
CFENGINE_COMMUNITY_RUN[$CFSERVERD]="1"
CFENGINE_COMMUNITY_RUN[$CFMONITORD]="1"
CFENGINE_COMMUNITY_BIN[$CFEXECD]="$CFENGINE_COMMUNITY_VAR_PATH/bin/cf-execd"
CFENGINE_COMMUNITY_BIN[$CFSERVERD]="$CFENGINE_COMMUNITY_VAR_PATH/bin/cf-serverd"
CFENGINE_COMMUNITY_BIN[$CFMONITORD]="$CFENGINE_COMMUNITY_VAR_PATH/bin/cf-monitord"
CFENGINE_COMMUNITY_PARAMS[$CFEXECD]=""
CFENGINE_COMMUNITY_PARAMS[$CFSERVERD]=""
CFENGINE_COMMUNITY_PARAMS[$CFMONITORD]=""
CFENGINE_COMMUNITY_PID_FILE[$CFEXECD]="$CFENGINE_COMMUNITY_VAR_PATH/cf-execd.pid"
CFENGINE_COMMUNITY_PID_FILE[$CFSERVERD]="$CFENGINE_COMMUNITY_VAR_PATH/cf-serverd.pid"
CFENGINE_COMMUNITY_PID_FILE[$CFMONITORD]="$CFENGINE_COMMUNITY_VAR_PATH/cf-monitord.pid"

RUDDER_DISABLE_AGENT_FLAG_FILE=/opt/rudder/etc/disable-agent

# Other
TIMEOUT="60"    # Max time to stop process

# Return functions' value
RETVAL=""

#====================================================================
# Message function
#====================================================================
message() {
  # $1: syslog level
  # $2: message

  if [ $OS = "Linux" ]
  then
    logger -p "${SYSLOG_FACILITY}.$1" -s -t $PROG_NAME -i "$2"
  else
    # Try without option -s
    logger -p "${SYSLOG_FACILITY}.$1" -t $PROG_NAME -i "$2"
    echo "$PROG_NAME: $2"
  fi
}

#====================================================================
# Load specific parameters
#====================================================================
if [ -f /etc/default/$PROG_NAME ]
then
  . /etc/default/$PROG_NAME
  message "info" "[INFO] Using /etc/default/$PROG_NAME for configuration"
else
  message "info" "[INFO] Using built-in configuration - this may cause some problems"
fi

#====================================================================
# Initial checks
#====================================================================

# Which daemons are we running?
DAEMONS=""
for daemon in $CFSERVERD $CFMONITORD $CFEXECD; do
  if [ "z${CFENGINE_COMMUNITY_RUN[$daemon]}" == "z1" ]; then DAEMONS="${DAEMONS} $daemon"; fi
done

# Make sure the pidfile directory exists with correct permissions
for daemon in ${DAEMONS}; do
  piddir=`dirname "${CFENGINE_COMMUNITY_PID_FILE[$daemon]}"`
  if [ ! -d "$piddir" ]; then
    mkdir -p "$piddir"
  fi
done

# Rights to execute binaries
for daemon in ${DAEMONS}
do
  if [ ! -x ${CFENGINE_COMMUNITY_BIN[$daemon]} ]
  then
    message "alert" "[ALERT] can't execute ${CFENGINE_COMMUNITY_NAME[$daemon]}"
    exit 1
  fi
done

# Is there a work directory?
if [ "$CFENGINE_COMMUNITY_VAR_PATH" -a ! -w "$CFENGINE_COMMUNITY_VAR_PATH" ]
then
  message "alert" "[ALERT] can't write to work directory $CFENGINE_COMMUNITY_VAR_PATH"
  exit 1
fi
message "info" "[INFO] Using $CFENGINE_COMMUNITY_VAR_PATH for CFEngine workdir"

# Are you root?
if [ $MYUID -ne 0 ]
then
  message "alert" "[ALERT] only root can launch CFEngine Community"
  exit 1
fi

#====================================================================
# Functions
#====================================================================
start_daemons() {
  RET=0

  # Check for disable-agent flag file
  if [ -f ${RUDDER_DISABLE_AGENT_FLAG_FILE} ]; then
    message "alert" "[ALERT] The ${RUDDER_DISABLE_AGENT_FLAG_FILE} flag file is present. The agent cannot be started."
    exit 3
  fi

  # It's time to start the daemons
  for daemon in ${DAEMONS}; do

    # Presence of PID file
    RUNNING_PID=`ps -efww | grep ${CFENGINE_COMMUNITY_BIN[$daemon]} | grep -v grep | tr -s " " | cut -f2 -d " "`
    [[ -f ${CFENGINE_COMMUNITY_PID_FILE[$daemon]} ]]  && PID=`cat ${CFENGINE_COMMUNITY_PID_FILE[$daemon]}` || PID=42
    if [[ $RUNNING_PID -eq $PID ]]
    then
      message "info" "[INFO] CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} is already running !"
      continue
    else
      # Start message
      message "info" "[INFO] Launching CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]}..."

      ${CFENGINE_COMMUNITY_BIN[$daemon]} ${CFENGINE_COMMUNITY_PARAMS[$daemon]}


      # Waiting loop
      i=1
      start_failed=0
      while [ ! -r ${CFENGINE_COMMUNITY_PID_FILE[$daemon]} ]
      do
        sleep 2
        if [ $i -eq $TIMEOUT -o $i -eq `expr $TIMEOUT / 2` ]
        then
          message "alert" "[ALERT] no PID file for ${CFENGINE_COMMUNITY_NAME[$daemon]} after $i times Trying again..."
          ${CFENGINE_COMMUNITY_BIN[$daemon]} ${CFENGINE_COMMUNITY_PARAMS[$daemon]}
        fi

        if [ $i -eq `expr $TIMEOUT + 10` ]
        then
          message "alert" "[ALERT] no PID file for ${CFENGINE_COMMUNITY_NAME[$daemon]} after $i times"
          RET=1
          start_failed=1
          break
        fi
        i=`expr $i + 1`
      done

      if [ ${start_failed} -eq 1 ]; then continue; fi

      # Is daemon launched?
      PID=`cat ${CFENGINE_COMMUNITY_PID_FILE[$daemon]}`
      if [ ! -e /proc/$PID ]
      then
        message "alert" "[ALERT] daemon ${CFENGINE_COMMUNITY_NAME[$daemon]} not running"
        RET=1
        continue
      else
        message "info" "[OK] CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} started after $i times"
      fi
    fi
  done

  if [ $RET -ne 0 ]; then exit $RET; fi

}

stop_daemons() {
  RET=0

  while [[ `$PS_COMMAND | grep ${CFENGINE_COMMUNITY_VAR_PATH}/bin | grep -v grep | wc -l` -gt 0 ]]
  do
    for daemon in ${DAEMONS}; do
      while [[ `$PS_COMMAND | grep ${CFENGINE_COMMUNITY_BIN[$daemon]} | grep -v grep | wc -l` -gt 0 ]]
      do
        # Stop message
        message "info" "[INFO] Halting CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]}..."

        # Presence of PID file
        # If the pid file is not readable or is empty, kill all process by name
        if [ ! -r ${CFENGINE_COMMUNITY_PID_FILE[$daemon]} -o ! -s ${CFENGINE_COMMUNITY_PID_FILE[$daemon]} ]
        then
          message "info" "[INFO] can't read PID file, not stopping ${CFENGINE_COMMUNITY_NAME[$daemon]}"
          RET=1
          sleep 2
        else
          PID=`cat ${CFENGINE_COMMUNITY_PID_FILE[$daemon]}`
          kill -INT $PID

          # Waiting loop
          # i[$daemon]=1
          i[$daemon]=`expr ${i[$daemon]} + 1`
          while [ -e /proc/$PID ]
          do
            if [ ${i[$daemon]} -eq $TIMEOUT ]
            then
              # Timeout
              message "alert" "[ALERT] ${CFENGINE_COMMUNITY_NAME[$daemon]} still running (PID $PID), try: $0 forcestop"
              exit 1
            fi
            i[$daemon]=`expr ${i[$daemon]} + 1`
            sleep 1
          done

          message "info" "[OK] CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} stopped after ${i[$daemon]} times"
        fi
      done
      sleep 1
    done
  done

  if [ $RET -ne 0 ]; then return $RET; fi

}

forcestop_daemons() {
  RET=0

  for daemon in ${DAEMONS}; do
    # Check if any process is running
    if [ `$PS_COMMAND | grep ${CFENGINE_COMMUNITY_BIN[$daemon]} | grep -v grep | wc -l` -eq 0 ]
    then
      message "info" "[INFO] Found no CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} processes running"

      # Cleanup PID file
      rm -f ${CFENGINE_COMMUNITY_PID_FILE[$daemon]}

      continue
    fi

    # Stop message
    message "info" "[INFO] Killing CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} with force..."

    # Presence of PID file
    # If the pid file is not readable or is empty, kill all process by name
    if [ ! -r ${CFENGINE_COMMUNITY_PID_FILE[$daemon]} -o ! -s ${CFENGINE_COMMUNITY_PID_FILE[$daemon]} ]
    then
      # Try a killall
      if [ -x /bin/pidof ]
      then
        /bin/kill -KILL `/bin/pidof ${CFENGINE_COMMUNITY_BIN[$daemon]}`
      else
        /usr/bin/killall -KILL ${CFENGINE_COMMUNITY_BIN[$daemon]}
      fi

      if [ $? -eq 0 ]
      then
        message "info" "[OK] all CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} processes killed with force"
      else
        message "alert" "[ALERT] Unable to kill CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} with force"
      fi
    else
      PID=`cat ${CFENGINE_COMMUNITY_PID_FILE[$daemon]}`
      kill -KILL $PID

      if [ $? -eq 0 ]
      then
        message "info" "[OK] CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} killed with force (PID $PID)"

        # Cleanup PID file
        rm -f ${CFENGINE_COMMUNITY_PID_FILE[$daemon]}
      else
        message "alert" "[ALERT] Unable to kill CFEngine Community ${CFENGINE_COMMUNITY_NAME[$daemon]} with force (PID $PID)"
        RET=1
        continue
      fi
    fi
  done

  if [ $RET -ne 0 ]; then exit $RET; fi

}

# Return 0 if all required daemons are running, 1 if any is stopped, 2 if we can't say, 3 if the disable-agent flag is present
daemons_status() {
  # Check for disable-agent flag file
  if [ -f ${RUDDER_DISABLE_AGENT_FLAG_FILE} ]; then
    return 3
  fi

  for daemon in ${DAEMONS}; do
    if [ ! -r ${CFENGINE_COMMUNITY_PID_FILE[$daemon]} ]
    then

      # Check if any process are running
      if [ `$PS_COMMAND | grep ${CFENGINE_COMMUNITY_BIN[$daemon]} | grep -v grep | wc -l` -eq 0 ]
      then
        return 1
      else
        return 2
      fi
    else
      PID=`cat ${CFENGINE_COMMUNITY_PID_FILE[$daemon]}`
    fi

    if [ ! -e /proc/$PID ]
    then
      return 1
    fi
  done

  return 0
}

display_status() {
  # Get status
  daemons_status

  status=$?

  if [ $status -eq 0 ]
  then
    message "info" "[INFO] All configured CFEngine Community processes are running"
  fi

  if [ $status -eq 1 ]
  then
    message "info" "[INFO] At least one configured CFEngine Community process is not running"
  fi

  if [ $status -eq 2 ]
  then
    message "info" "[INFO] Unable to determine CFEngine Community processes status"
  fi

  if [ $status -eq 3 ]
  then
    message "info" "[INFO] The ${RUDDER_DISABLE_AGENT_FLAG_FILE} flag is present, CFEngine Community is halted"
  fi

  # Exit with status code
  exit ${status}

}

#====================================================================
# Action switch
#====================================================================
case $1 in
  start)
    start_daemons
    ;;
  stop)
    stop_daemons || forcestop_daemons
    ;;
  forcestop)
    forcestop_daemons
    ;;
  restart)
    stop_daemons
    start_daemons
    ;;
  status)
    display_status
    ;;
  *)
    echo "Usage: $0 {start|stop|forcestop|restart|status}"
    exit 1
    ;;
esac

#====================================================================
# Exit
#====================================================================
exit 0

Hi @frbayart can you give us a diff ?
Or tell me the exact version you are using so that I can do it.