Automated nmon collection

I’m working with a customer and I get this question

“My system is running slow, do you know why?”
“How do you know it’s running slow?”, I ask.
“I don’t know it just seems like it’s taking longer to do <whatever>”, they say.
“Do you have any trending that we can look at?”
“No – it was running fine. Why would I need that if it everything was fine?”, they ask.

I shake my head and die a little inside…
As a result of a few conversations like this I decided that rather than leaving it up to the customer to collect performance statistics on a regular basis, I need to do it for them. This gives me a few bits of information to use when they finally start coming to me with the inevitable performance problems and/or questions:
  • We have a baseline to work from!!!
  • We can determine what changed, if anything
  • It may only be in your head – we need to prove it’s not! (Some crazy people work on computers)
  • If there really is a problem, where do we even start looking?
    I whipped up this simple script to make it a lot easier to collect that baseline. [Nmon]( is now part of the [AIX]( operating system and there is really no reason why you shouldn’t be using it to collect data. It’s a very good “whole system health” type monitoring tool that you can get down and dirty with if needed.
  1. Change the variables to valid values for you!
  2. REPORT_RCPT – Who is going to receive this analysis?
  3. COMPANY – If you are collecting this for multiple companies (I am) put the company name here so the reports make sense when you get them.
  4. REPORTS_TO_KEEP – How many are we going to keep on the local system (not in your email!)
    Schedule the script in cron.
  5. A lot of people like to schedule this for midnight, however that’s when a lot of people schedule maintenance or backups. This splits those periods of high activity into multiple reports. Consider scheduling it for a regular “slow” period, like when people leave work ( 17:00 to 19:00 ) or usually about 06:00 before they come in to hit the system and nightly processes are finished.
    Collect the reports it sends to you – DON’T DELETE THEM! Copy them to your system if you need to get them out of email. Remember this is data collection not data collect and toss!!!
  6. [Analyze]( periodically so you have a quantitative value and idea what your system is doing – how often depends on your environment.
  7. When you have a performance problem later – reference earlier reports to determine what changed.


export DATE=`date +%m%d%Y`
export CURRENT_DIR=`pwd`

export REPORT_RCPT=""
export COMPANY="MyCompany"

# if the directory doesn't exist - create it

if [[ ! -d "/tmp/nmon" ]] then
 mkdir /tmp/nmon
cd /tmp/nmon

# Now lets get the one from yesterday and email to where is needs to be
export NMON_FILES=`ls -ctl | awk '{print $9}' | grep nmon | grep -v gz`
for i in $NMON_FILES; do
 export NEW_FILENAME=`echo $COMPANY ${i} | awk '{print $1 "_" $2}'`
 sort $i > $NEW_FILENAME
 tar cvf - $NEW_FILENAME | gzip -c > $NEW_FILENAME.tar.gz

 # we have the file we need now lets email it to whomever need it
 uuencode $NEW_FILENAME.tar.gz $NEW_FILENAME.tar.gz | mail -s "$COMPANY nmon report for $DATE" $REPORT_RCPT

 # Cleanup just a bit
 rm -f $i

find /tmp/nmon -type f -mtime +$REPORTS_TO_KEEP -exec rm -f {} \;

# Start NMON for the next day!
nmon -f -s 60 -c 1440

# Just incase you run it interactivly - return to where you started