cpu - Plugin to monitor CPU usage.
All Linux systems
The following is default configuration
[cpu] env.HZ 100 env.scaleto100 no
"scaleto100" may be "yes" or "no". With "yes" all CPU-related values are scaled for a limit of 100. With "no" (the default) all CPU-related values are scaled for a limit of n * 100 ("n" being the number of CPUs).
See "BUGS" for an explanation of the "HZ" setting.
You can also set warning and critical levels for each of the data series the plugin reports. The following environment variables are used as default for all fields:
But each field can be controlled separately:
env.system_warning env.system_critical env.user_warning env.user_critical env.nice_warning env.nice_critical env.idle_warning env.idle_critical
For some kernels there is also the following settings:
env.iowait_warning env.iowait_critical env.irq_warning env.irq_critical env.softirq_warning env.softirq_critical env.steal_warning env.steal_critical env.guest_warning env.guest_critical
The plugin shows cpu usage in percent. In case of more than one core it displays 100% for each core.
If a core is 100% busy there will be no "iowait" showing, that only shows if the CPU has nothing else to do while it waits on IO. Therefore a 100% busy core can hide a lot of iowait. Please refer to the IO latency and other disk related graphs for further information about IO performance.
#%# family=auto #%# capabilities=autoconf
Some combinations of hardware and Linux (probably only 2.4 kernels) use 1000 units/second in /proc/stat corresponding to the systems HZ. (see /usr/src/linux/include/asm/param.h). But Almost all systems use 100 units/second and this is our default. Even if Documentation/proc.txt in the kernel source says otherwise. - Finding and fix by firstname.lastname@example.org
Otherwise none known