I've been using monit for a little while now to monitor a few
machines, and on one particular machine which has a low priority
process which mainly runs when the CPU would have been otherwise
idle (I.e. it is "nice -n 19"), I get high user CPU usage
notification, since monit considers user usage to be "cpu_user +
cpu_nice", but in reality it is a low priority processes which would
just get pushed out the way if anything really important want to get
the CPU. So I'd like to be able to specify a check rule like "if cpu
usage (user) > 70% then alert" but to exclude the portion which
came from nice. I have a couple of suggestions for possible ways to
implement it:
- Add a new resource name which only includes cpu_user time and
not cpu_nice time - backwards compatible
- Add a config option which excludes cpu_nice from the CPU user
stat - default to off for backwards compatibility
Of these two, I suspect the second might be simpler. Would this
feature be useful to anyone else out there?
Regards
Phil
|