Monitor Oracle Processes

#!/bin/ksh

############################################################################
#
# Color Codes
#
############################################################################

ESC_SEQ="\033["

NORM=$ESC_SEQ"0m"

RED=$ESC_SEQ"0;31m"
LRED=$ESC_SEQ"1;31m"
LGREEN=$ESC_SEQ"1;32m"

############################################################################
#
# Create errcheck function
#
############################################################################

function errcheck {

if [[ $1 = 0 ]]
then return
fi

case $1 in

2)   echo "ORACLE_MONITOR:ERROR:2:ORACLE_PROCESS_OFFLINE"
     echo "---------------------------------------------"
     echo "The following processes have been detected as OFFLINE..."
     echo -e ${LRED}""
     crscheck |grep -i off |egrep -v "${SDS_SI_IGNORE}"
     echo -e ${NORM}""
     echo -e ${LGREEN}"Action:${NORM} Contact a DBA to investigate the cause of this failure."
     echo "---------"
     echo ""
     echo ""
     SDS_ERRORS=`expr $SDS_ERRORS + $SDS_ERR`;;
4)   echo "ORACLE_MONITOR:ERROR:4:ORACLE_PROCESS_NOT_RUNNING"
     echo "------------------------------------------------------"
     echo "The following errors have been detected..."
     echo -e ${LRED}""
     echo -e "Oracle Instance: $f, has no $n process running."
     echo -e ${NORM}""
     echo -e ${LGREEN}"Action:${NORM} Contact a DBA to investigate the cause of this failure."
     echo "---------"
     echo ""
     echo ""
     if [[ $SDS_ERRORS -eq 4 ]] || [[ $SDS_ERRORS -eq 6 ]]
     then :
     else SDS_ERRORS=`expr $SDS_ERRORS + $SDS_ERR`
     fi;;
8)   echo "ORACLE_MONITOR:ERROR:8:ASM_PROCESS_NOT_RUNNING"
     echo "------------------------------------------------------"
     echo "The following errors have been detected..."
     echo -e ${LRED}""
     echo -e "$SDS_ASM has no $n process running."
     echo -e ${NORM}""
     echo -e ${LGREEN}"Action:${NORM} Contact a DBA to investigate the cause of this failure."
     echo "---------"
     echo ""
     echo ""
     if [[ $SDS_ERRORS -eq 8 ]] || [[ $SDS_ERRORS -eq 10 ]] || [[ $SDS_ERRORS -eq 14 ]]
     then :
     else SDS_ERRORS=`expr $SDS_ERRORS + $SDS_ERR`
     fi;;
16)  echo "ORACLE_MONITOR:ERROR:16:GRID_LISTENER_NOT_RUNNING"
     echo "------------------------------------------------------"
     echo "The following errors have been detected..."
     echo -e ${LRED}""
     echo -e "There is no listener process running."
     echo -e ${NORM}""
     echo -e ${LGREEN}"Action:${NORM} Contact a DBA to investigate the cause of this failure."
     echo "---------"
     echo ""
     echo ""
     SDS_ERRORS=`expr $SDS_ERRORS + $SDS_ERR`;;
32)  echo "Not set yet";;
64)  echo "Not set yet";;
128) echo "Not set yet";;
*)   echo "ORACLE_MONITOR:ERROR:999:SCRIPT_FAILURE_UNKNOWN_ERROR_CODE"
     echo "----------------------------------------------------------"
     echo -e ${LGREEN}"Action:${NORM} Contact a DBA to investigate the cause of this failure."
     exit 999;;

esac

}

############################################################################
#
# Create crscheck function
#
############################################################################

function crscheck {

RSC_KEY=$1
AWK=/bin/awk    # if not available use /usr/bin/awk

# Table header:echo ""
$AWK \
  'BEGIN {printf "%-45s %-10s %-18s\n", "HA Resource", "Target", "State";
          printf "%-45s %-10s %-18s\n", "-----------", "------", "-----";}'

# Table body:
$GRID_HOME/bin/crsctl status resource | $AWK \
 'BEGIN { FS="="; state = 0; }
  $1~/NAME/ && $2~/'$RSC_KEY'/ {appname = $2; state=1};
  state == 0 {next;}
  $1~/TARGET/ && state == 1 {apptarget = substr($2,index($2,"ZZZZZ"),7); state=2;}
  $1~/STATE/ && state == 2 {appstate = $2; state=3;}
  state == 3 {printf "%-45s %-10s %-18s\n", appname, apptarget, appstate; state=0;}'
}

############################################################################
#
# SCRIPT START
#
############################################################################

############################################################################
#
#       Script :        oracle_monitor.ksh
#
#       Date :          12th Jan 2011
#
#       Author :        Mark Ramsay
#
#       Copyright:      www.markramsay.com 2011
#
#       Description :   Checks to make sure various processes are running
#
#       Parameters :    None
#
# History       Date      Name      Reason
#               ----      ----      ------
#               dd/mm/yy  xxx       yyyyyyy
#
############################################################################

#
# Configure Variables
#

export SDS_COMPONENTS="pmon smon dbw lgwr"
export SDS_ASM_COMPONENTS="pmon smon dbw lgwr"
export SDS_GRID_LISTENER=lsnr
export SDS_RAC_IGNORE=""
export SDS_SI_IGNORE="ora.diskmon|ora.ons"
export SDS_ERRORS=0
export SDS_ERR=0

#
# Testing Variables Only - Unhash for Use.
#

#export SDS_SI_IGNORE="ora.dskmon|ora.ons"
#export SDS_COMPONENTS="pon son dbw lgr"
#export SDS_ASM_COMPONENTS="pon smon dbw lgwr"
#export SDS_GRID_LISTENER=lnr

#
# Check health using crs
#

crscheck |grep -i off |egrep -v "${SDS_SI_IGNORE}" >/dev/null

if [[ $? -eq 0 ]]
then SDS_ERR=2
errcheck $SDS_ERR
fi

#
# Get a list of DB instances
#

SDS_LIST=`egrep -v '(^M.....)00' /etc/oratab |egrep '(^M.....)'|awk -F\: '{print $1}'` #Line will need to be updated for the relevant database.

#
# Open loop for instances
#

for f in $SDS_LIST
do

  #
  # Open loop for components
  #

  for n in $SDS_COMPONENTS
  do

  #
  # Check DB process is running for the given instance
  #

  ps -ef |grep $f |grep $n |grep -v grep >/dev/null

  if [[ $? -eq 1 ]]
  then SDS_ERR=4
  errcheck $SDS_ERR
  fi

  #
  # End loop for components
  #

  done

#
# End loop for instances
#

done

#
# Check ASM processes
#

SDS_ASM=`grep ^\+ASM /etc/oratab | awk -F\: '{print $1}'`

#
# Open loop for ASM components
#

for n in $SDS_ASM_COMPONENTS
do

#
# Check to see if ASM component is running
#

ps -ef |grep $SDS_ASM |grep -v grep |grep $n >/dev/null

if [[ $? -eq 1 ]]
then SDS_ERR=8
errcheck $SDS_ERR
fi

#
# End loop for ASM components
#

done

#
# Check listener
#

ps -ef |grep $SDS_GRID_LISTENER |grep -v grep >/dev/null

if [[ $? -eq 1 ]]
then SDS_ERR=16
errcheck $SDS_ERR
fi

#
# Exit script with error code
#

exit $SDS_ERRORS