Shell Programming and Scripting

Customers specifying solutions rather than giving requirements is never good!

If you are getting sar data once an hour rather than every 20 minutes then you can do the maths to see hour many lines of data you want...

Otherwise I will hand over to cfajohnson's far better scripting abilities than mine.

The update to the script knowing that a line of data is added every hour is:

Code:

sar -f sa15 | grep -v Average | grep -v "^$" | grep -vi restart | grep -v System | grep -v AIX | tail -48 > sa15.$$
mailx -s "Last 48 hours sar data at `date`" user@domain.com < sa15.$$
rm sa15.$$

Please read the grep man page; look at the -e option.

I hope that this is the current thread (of many) for this topic.

To my mind the key point is in this statement from O/P. I could be wrong.

This ties up with my assumption that this is an incorrectly set up "sar" data collection and that there are multiple occurances of the day of the month appended into the sar(nn) data file. There is a strong hint in the quote above that you need to process more than one day file.
Depending on what time of day the report is run it would take two (or three) of these sar data files to get 48 hours data. This is further complicated at month boundaries. It is much too complicated to process in retrospect and requires some forward planning. If my guess is correct the sa(nn) files will grow as more data is appended.

A logical approach would be to prepare one day's summary data at the end of each day in a file named after the date in the sar heading. It is then a case of mailing the most recent two files concatonated.