Any other utility can used to check CPU time like "top"?

[yyghp]

I am going to write a script to monitor the oracle processes, when one of them continues to consume CPU 100% for about 1000 minutes, then send me an email.
Although "top" can be used to check the current "Raw CPU percentage" ("%CPU") and "Number of system and CPU seconds process has consumed" (TIME) of each process, but I don't know how to use "top" output for scripting.
Is there any other utility like "top" to get the "%CPU" and "TIME" of the process?
Thanks!

[Jeff Schussele]

Hi,

The best way to do this would be to leverage MeasureWare to collect data & PerfView to analyze it.
These are purchased products, but are designed to get to the process level much better than anything else.

My 2 cents,
Jeff

[James R. Ferguson]

Hi:

If you would like to capture the output of 'top' into a file use the '-f file' switch and argument. For instance, to capture one snapshot, do:

# top -d 1 -f /tmp/top.output

Otherwise, I suggest using 'ps'. If you leverage the UNIX95 option, the outpuf of the time can be displayed in the format "[dd-]hh:mm:ss" instead of "mmmm:ss". This can make things easier.

Regards!

...JRF...

[rmueller58]

I've wrote a couple of scripts I put in cron for the same purpose trying to isolate which app, that front ends for an Informix DB that is sucking CPU and memory..

#checkproc -- This script view only user related processes
ps -ef |grep -v informix|grep -v root |grep -v lp|awk '{print $4, $1, $2, $7, $8, $9, $10, $11, $12,$13, $14, $15}' |sort -r| head -10

#chkps.sh
# this file appends information in a log to show High Water Marks through the day. Note Column #3
export line1=`checkproc |head -2 |tail -1`
export line2=`checkproc |head -3 |tail -1`
export line3=`checkproc |head -4 |tail -1`
export line4=`checkproc |head -5 |tail -1`
export line5=`checkproc |head -6 |tail -1`
export d=`date +%D`
export t=`date +%T`
echo $d $t $line1 >> /restore/`date +%Y%m%d`chkproc.log
echo $d $t $line2 >> /restore/`date +%Y%m%d`chkproc.log
echo $d $t $line3 >> /restore/`date +%Y%m%d`chkproc.log
echo $d $t $line4 >> /restore/`date +%Y%m%d`chkproc.log
echo $d $t $line5 >> /restore/`date +%Y%m%d`chkproc.log


it is a way to log it at very least, I use it for troubleshooting a vendor app that periodically gives me headaches.. I've isolated several apps that the vendor was forced to re-write because of this.

Generally the DB is not the problem unless you don't have it tuned correctly.. More likely it is an application with inefficient queries..

[yyghp]

Thanks guys!

Jeff: I don't think I can get it approved to buy an extra product for this purpose, but thanks for your advice!

James: I really don't want to use temporary file for scripting unless I don't have any other choice, that's why I raise the question here. And how "ps" can show the "%CPU" ? ( I need to grep the oracle process keeps consuming 100% CPU ), like this:

CPU TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
4 ? 4243 oracle 239 20 4562M 12460K run 1029:15 100.57 100.40 oracleRMS_PROD

Rex: I read through your script, but where to get "%CPU" ?

Thanks!

[Greg Vaidman]

if you use the following command line, you can get the top output (mostly) unformatted so that you can parse it inline in a script. This bypasses the temporary file.

TERM=ansi top -d 1 | col -b |
awk '/^CPU[[:space:]]*TTY/,EOF{print}' | tail +2

[rmueller58]

Column #3 is CPU%

[yyghp]

Hi rex, your script is based on command "ps", but "ps" doesn't get "CPU%", right? So, how can the script shows "CPU%"? Thanks!

[Greg Vaidman]

ps can get %cpu with the following command line:

UNIX95= ps -e -o ruser,pid,pcpu,args

the 3rd column will be %cpu

see the man page for ps if you need additional columns or want it formatted differently.

[yyghp]

but how I can know that process keeps running more than 1000 mintues?
thanks!

[Greg Vaidman]

well, you can add cpu to the list of options in the ps command.

but in any case, either with ps or top, the %cpu reported is just for the last scheduling interval. So if you see 100% and 1000 minutes in any one report, it only means that it used 100% over the past couple of seconds. you need to collect data over time and parse it to really say that it used >x% over any period of time.

[yyghp]

I think I'd better use "top" for this purpose, thanks everyone!