Abhi wrote:
> Hi,
>
> does anyone has a script (possibly perl or shell) to monitor CPU,
> memory etc on run time ? I have written some scripts but they run on a
> pre-defined time i.e from cron or autosys. I want to put something at
> kernel level which alerts us whenever CPU starts going high.
>
> idea on how to write would also be helpful, I can then write it in
> perl or shell.
>
> cheers...



for a start search the IBM website for "RMC", which is the monitoring
built into aix, also look at "errnotify", a mechanism for automating
error log reporting. These are basic building blocks for writing your
own monitoring solution.

If you are more interested in reporting (which in contrast to monitoring
does not include means of having alerts sent under predifined conditions
and the like):

IBM also has a solution for performance reporting which to my knowledge
is neither supported nor publicly available but is very helpful (built
on nmon, rrdtool etc.). Ask your IBM reps for "smooth start".

and there is ganglia, opensource originally for reporting of hpc
clusters, but with IBM POWER extensions written by Michael Perzl ->
www.perzl.org
Takes some reading to make yourself aquainted, but is worth the effort.

And of course there are plenty of commercial solutions.

good luck
J. Gann

On Sep 12, 12:34*am, Joachim Gann wrote:
> Abhi wrote:
> > Hi,
>
> > does anyone has a script (possibly perl or shell) to monitor CPU,
> > memory etc on run time ? I have written some scripts but they run on a
> > pre-defined time i.e from cron or autosys. I want to put something at
> > kernel level which alerts us whenever CPU starts going high.
>
> > idea on how to write would also be helpful, I can then write it in
> > perl or shell.
>
> > cheers...
>
> for a start search the IBM website for "RMC", which is the monitoring
> built into aix, also look at "errnotify", a mechanism for automating
> error log reporting. These are basic building blocks for writing your
> own monitoring solution.
>
> If you are more interested in reporting (which in contrast to monitoring
> does not include means of having alerts sent under predifined conditions
> and the like):
>
> IBM also has a solution for performance reporting which to my knowledge
> is neither supported nor publicly available but is very helpful (built
> on nmon, rrdtool etc.). Ask your IBM reps for "smooth start".
>
> and there is ganglia, opensource originally for reporting of hpc
> clusters, but with IBM POWER extensions written by Michael Perzl ->www.perzl.org
> Takes some reading to make yourself aquainted, but is worth the effort.
>
> And of course there are plenty of commercial solutions.
>
> good luck
> J. Gann

Thanks for your reply J. I already have installed Ganglia into our
setup but the Ganglia is more of a visual monitoring. You cannot setup
notification in ganglia. Are you aware of any version of Ganglia which
has notification ? or any trigger which can execute an external
script ?

I will check-out the other options you mentioned. Thanks again for
your reply.

cheers...