unix performance monitoring script

[avik1983]

hey,
i am in bit prob..can u help me to write a script in unix (can use perl also) to monitor
1.disk usage
2.cpu usage
3.memory usage

in every 10 seconds(for example)..
it should display the three usage result in screen in every 10 seconds.
it will be better if there is an option for user to set the monitoring time also..
pls help guys...

[ramudukamudu]

Hi, I am also facing similar problem. I need to monitor the memory usage, idle %CPU percentage while my application is running. right now I am using "top" utility. But it is difficult to extact required information from "top". "top" is generating the O/P continuously. I want one shot answer as "vmstat","iostat" utilities give.
Could you suggest the solution??
regards,
ramudu.

[arne]

Hi, I am also facing similar problem. I need to monitor the memory usage, idle %CPU percentage while my application is running. right now I am using "top" utility. But it is difficult to extact required information from "top". "top" is generating the O/P continuously. I want one shot answer as "vmstat","iostat" utilities give.
Could you suggest the solution??
regards,
ramudu.

For Linux, consider using the information provided under /proc:

/proc/diskstat
/proc/cpustat
/proc/meminfo

This is where many monitoring tools under Linux get their info from.
For specific processes look under /proc/<PID>/

Can't say anything about UNIX, though ...

Hope that helps.

[drhowarddrfine]

In Unix it's du and df for disk usage.
I found this.

[ramudukamudu]

Thanks flocks,
That's really helpful.

[cybervegan]

You could always make a crontab job to dump the values, but you'd probably get less of a performance hit by writing a short perl or shell script to do it, and then "daemonize" it.

I use Python the most, so that's what I'll use to prototype:

import time
cpuidle = 0 # adjust to the word index of the stat you want, counting from 0
# when you do "cat /proc/cpustat"
diskfree = 0 # adjust as per cpuidle
memfree = 9 # adjust as per cpuidle (this one is probably right)
logfile = open("/var/log/perflog","w")
while True:
logfile.write("CPU: %s DISK: %s MEM: %s\n" % (
file("/proc/cpustat").read().split(" ")[cpuidle],
file("/proc/diskstat").read().split(" ")[diskfree],
file("/proc/meminfo").read().split(" ")[memfree]))
logfile.flush()
time.sleep(10)

You could probably convert this to perl quite easily...

reagards,
-cybervegan

[nano2]

You could always make a crontab job to dump the values, but you'd probably get less of a performance hit by writing a short perl or shell script to do it, and then "daemonize" it.

I use Python the most, so that's what I'll use to prototype:

import time
cpuidle = 0 # adjust to the word index of the stat you want, counting from 0
# when you do "cat /proc/cpustat"
diskfree = 0 # adjust as per cpuidle
memfree = 9 # adjust as per cpuidle (this one is probably right)
logfile = open("/var/log/perflog","w")
while True:
logfile.write("CPU: %s DISK: %s MEM: %s\n" % (
file("/proc/cpustat").read().split(" ")[cpuidle],
file("/proc/diskstat").read().split(" ")[diskfree],
file("/proc/meminfo").read().split(" ")[memfree]))
logfile.flush()
time.sleep(10)

You could probably convert this to perl quite easily...

reagards,
-cybervegan

Will this work for hp and solaris ???

[ghostdog74]

hey,
i am in bit prob..can u help me to write a script in unix (can use perl also) to monitor
1.disk usage
2.cpu usage
3.memory usage

in every 10 seconds(for example)..
it should display the three usage result in screen in every 10 seconds.
it will be better if there is an option for user to set the monitoring time also..
pls help guys...

have you tried to code anything yet? read up anything on shell scripting before this?