Skip to contentSkip to conversation

checking if a job still uses the CPU

DISCUSSION martin f krafft 1 participant 2 messages Nov 17, 2004

lore
Share & tools
lore
#0 lore raw

I want to check a job periodically whether it still uses the CPU. If not, I want to kill it. The latest solution I found is at

http://madduck.net/~madduck/scratch/worker.sh
http://madduck.net/~madduck/scratch/job.sh

worker.sh starts job.sh and polls /proc/$PID/stat:16 (the 16th
field: cutime) regularly. As soon as it stopped changing, the job is killed.

This works fine as long as job.sh does not start jobs itself which run longer than the period used to check in the worker (7 seconds in my example). If job.sh's children take longer than 7 seconds to complete, job.sh might get killed because the cutime value of a process in /proc is only updated whenever a child returns.

Is there another way to achieve what I am trying to do? How can I kill jobs that are idling and have not used the CPU in $DELAY time?

Thanks,

Signature · 10 lines
martin;              (greetings from the heart of the sun.)
  \____ echo mailto: !#^."<*>"|tr "<*> mailto:" net@madduck
 
invalid/expired pgp subkeys? use subkeys.pgp.net as keyserver!
spamtraps: [email protected]
 
"we are like shop windows in which we are continually arranging,
 concealing or illuminating the supposed qualities other ascribe to us
 -- in order to deceive ourselves."
                                                 - friedrich nietzsche
#1 lore raw

I should clarify: all jobs I intend to start this way are CPU-intensive, so that it should be easy to determine whether a job is idling.

Signature · 7 lines
martin;              (greetings from the heart of the sun.)
  \____ echo mailto: !#^."<*>"|tr "<*> mailto:" net@madduck
 
invalid/expired pgp subkeys? use subkeys.pgp.net as keyserver!
spamtraps: [email protected]
 
micro$oft could shit in a box, and most people would buy it.