DBA Scripts coding considerations for best compatibility to any platform

Did you find yourself trying to convert your DBA scripts to work on other OS or UNIX flavors ?

This is a typical issue DBAs cope with, the solutions can be:

1. Each system has its own best scripting code (PowerShell for windows, bash for Linux), let's keep some versions of the same code.
2. Write the code in a common language for all systems

• Shell: In the past I used the old Posix shell (sh) which exited in all Unix systems. These days I use bash as is very common. 
• Perl: Oracle installation already has its own Perl, no need to install it.
• Java: Oracle installation already has its own java, no need to install it.
• Python: Python should be installed, but the libraries that can be used are huge, and you can write shorter code.

3. Write in sql/plsql as much as you can, the external code will be minimal (cmd / sh).
4. Work remotely from your convenient operating system, using your scripts written in the language you master.
5. Use a few macros injected to your terminal.
6. Why use scripts, I use Toad / Cloud Control - it is much easier.
7. Leave me alone, I like typing the same commands over and over.

I use most of the above options :) but try to have one version of script.
I am writing in Bash, but have started moving to Python (for new and complicated needs).
Recently I had a need to work on Solaris. Because I believe in using one code for all, I started to modify my code with a lot of IFs.
Above that, I needed to adjust external utilities like "sed" and "awk" since they do not work the same (Solaris Vs Linux).
Luckily I found that from Solaris 11 there are built-in gnu commands which are compatible with Linux, that are located in a different directory and start with the letter "g": "sed" is "gsed", "awk" is "gawk" and so on.
I decided that the gnu utility commands will be a variable in all of my scripts. All scripts will call a master environment file that decides which version to use.

cat platform.env
platform=`uname -s`
case "$platform"
in
   "SunOS")  os=Solaris
             ORATAB=/var/opt/oracle/oratab
             AWK=gawk
             GREP=ggrep
             EGREP=gegrep
             SED=gsed
             HOST_NAME=$(hostname)
             DF=gdf
             FIND=gfind
             USER=${LOGNAME}           
             CAT=gcat
             CHOWN=gchown
             CHMOD=gchmod
             CP=gcp
             ECHO=gecho
             MKDIR=gmkdir
             CUT=gcut
             SUDO=/usr/local/bin/sudo
             ;;
   "Linux")  os=Linux
             ORATAB=/etc/oratab
             AWK=awk
             GREP=grep
             EGREP=egrep
             SED=sed
             HOST_NAME=$(hostname -f)
             DF=df
             FIND=find
             CAT=cat
             CHOWN=chown
             CHMOD=chmod
             CP=cp
             ECHO=echo
             MKDIR=mkdir
             CUT=cut
             SUDO=sudo
             ;;
   #"HP-UX")  os=hpunix;;
   #  "AIX")  os=aix;;
         *)  echo "Sorry, $platform is not currently supported." | tee -a $LOGFILE
             exit 1;;
esac

In my last blog I discussed about identifying $ORACLE_SID and $ORACLE_HOME Basics commands for DBA part 1 - finding instance.
Now we will see a generic script version which uses the variables above:

#!/bin/bash
export SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
source ${SCRIPT_DIR}/platform.env
if [ -f ${ORATAB} ]
then
    for ORA_SID in `${CAT} ${ORATAB} | ${GREP} -v ^'#' | ${GREP} -v ^$ | ${AWK} -F ":" ' { print $1 }' | uniq`
    do
        if [[ ! "${ORA_SID^^}" == *"ASM"* ]] && [[ ! "${ORA_SID^^}" == *"MGMTDB"* ]]
        then
                echo working on ${ORA_SID}
                export ORACLE_SID=${ORA_SID}
                export ORACLE_HOME=`${CAT} $ORATAB | ${GREP} ^${ORACLE_SID}: | ${GREP} -v "^#" | ${GREP} -v "^$" | ${CUT} -d":" -f 2`
                Put your code here, using the variables above
        fi
    done
fi

Please comment and tell us about your way of using DBA scripts
Yossi Nixon