CPU Load Check Reference

All configuration fields, sampling parameters, and alert thresholds for the CPU Load check in octoja.

Written By Stefan Steuer

Last updated 28 days ago

Configuration options and output metrics for the CPU Load check.

The CPU Load check samples processor usage on the device's local CPU over a short window and compares the result against percentage thresholds. The agent collects several CPU readings per check run and reports both the average and the peak.

Configuration fields

FieldTypeRequiredDefaultDescription
Sampling duration (seconds)integerNo (default applies)5Total time spent collecting CPU samples on each run. Clamped to 1–60
Sampling interval (ms)integerNo (default applies)500Time between individual CPU readings. Clamped to 100–5000
Evaluate againstchoice (Peak / Average)No (default applies)PeakWhether the threshold check compares the peak sample or the average of all samples
Warning (%)integerNo (default applies)80Warning is raised when the selected value (peak or average) exceeds this percentage. Clamped to 0–100
Critical (%)integerNo (default applies)95Critical alert is raised when the selected value exceeds this percentage. Clamped to 0–100

Alert behaviour

ConditionSeverity
Selected metric ≥ Critical thresholdCritical
Selected metric ≥ Warning threshold (and below Critical)Warning
Selected metric below the Warning thresholdOK
No CPU samples could be collected during the runWarning, with the message "Unable to collect CPU samples"
An unexpected error occurs during the runFailure (shown as Error in the dashboard)

Metrics collected

  • Average CPU usage over the sampling window (%)
  • Peak CPU usage in any single sample (%)
  • Number of samples collected
  • The threshold metric used for the alert (Peak or Average)

Platform support

PlatformSupported
WindowsYes
LinuxYes
macOSYes

Notes

  • The check runs every 5 minutes on the agent. The check interval is not user-configurable.
  • If you set the Warning threshold higher than the Critical threshold by mistake, the agent automatically swaps them — the lower value is treated as Warning, the higher as Critical.
  • Sampling happens inside a single check run; the sampling window (default 5 seconds at 500 ms intervals = ~10 samples) is independent of the check interval (every 5 minutes).
  • Use the Peak evaluation metric for workloads where any spike should fire an alert. Use Average for workloads that tolerate brief peaks (batch jobs, video encoding, compilation).

See also: Configure a CPU Load Check