System related information during performance testing

Resolving The Problem

--- Script Begin ---

#! /bin/ksh
echo Date: `date`
echo Date: `date +%H:%M:%S` > /tmp/$$

case $(uname) in
SunOS) ps -eo pid,ppid,time,etime,args >> /tmp/$$
;;
AIX) ps -eo pid,ppid,time,etime,thcount,scount,args >> /tmp/$$
;;
Linux) ps -eo pid,ppid,time,etime,thcount,scount,args >> /tmp/$$
;;
HP-UX) export UNIX95=1; ps -eo pid,ppid,time,etime,args >> /tmp/$$
;;
*) exit
;;

esac

FILE="/users/dcm/ps-`uname -n`-`date +%d`"
#echo $FILE

echo ================================ >> $FILE
cat /tmp/$$ | awk '/^Date:/ {D=$2} {print D, " ", $0}' >> $FILE

rm /tmp/$$

--- Script End ---

Some pointers for this script:
--------------------------------------------

This will generate reports specific to OS. For different reports just change the appropriate command (like "iostat" instead of "ps -eo..."). This script saves the reports under $FILE folder.

This script will run periodically and dump data into a file under a specified directory. This needs to be changed before deploying & running the script. It can be changed in the "FILE" variable. Just change the first 2 directory entries, don't change the actual file name starting with "ps-".

The script will need to be placed in the cron to run every 5-6 minutes. Don't need to run more frequently. Ensure that you have sufficient space wherever you decide to dump the output of this script.
Depending on the # of processes running, the output can be large.
Please beware that workloads that start up for short periods of time and stop repeatedly may not be captured by this script even though they may use quite a bit of CPU.