On Thu, Jun 20, 2002 at 12:40:57PM -0500, Michael D. Schleif wrote:
>
> We have a third party application that we are running under debian. It
> is a complex application that uses many executables, some of which are
> not always running, are called by the main executable periodically; but,
> the periodicity is not constant nor predictable.
>
> One of these un-predictable processes is causing us problems. When it
> comes to life, always a new pid, it tries to grab *all* of the cpu and
> it may live for many minutes.
>
> We need to quantify our problem to the software developers; but, how can
> we do this?
>
> top is running and we can empirically see the problem, when it occurs
> and when we happen to be looking.
>
> How can we automate this profiling of this process?
How about replacing the executable with a script (and renaming the
original to ${original_name}.real). The script could contain something
along the lines of:
#!/bin/sh
( echo $0 is about to start
for a in "$@"
do
echo -- Parameter: \""$a"\"
done
echo Get Pizza.
) | mail -s "System Gale Warning" yourname-here@localhost
sleep 600
exec $0.real "$@"
At least this way you will get a "heads-up", so you are prepared before
it suddenly hits you.
Depending on your application, this may provide enough information
(=command line parameters) for you to reproduce the problem at will.
PS: Don't forget to rename things back afterwards :-)
HTH
--
Karl E. Jørgensen
karl@jorgensen.com
www.karl.jorgensen.com
Please read http://www.pantsfullofunix.net before reporting bugs in my code.
It won't help you, but it will make me feel better.
Attachment:
pgpgeioxhn7y7.pgp
Description: PGP signature