perfdata

Perfdata, short for performance data, is a structured set of metrics emitted by software components to quantify operational characteristics such as response time, throughput, or resource usage. In monitoring practice, perfdata is designed to be machine-readable and consumable by graphing, alerting, and analysis tools. The term is most closely associated with Nagios and related ecosystems, where plugins emit perfdata alongside human-readable output.

Typical perfdata uses a pipe to separate a human message from the numeric data. After the pipe, one or more fields appear in the form label=value[unit];[warn];[crit];[min];[max]. The value may include a unit suffix (such as %, ms, MB). Each field encodes a single metric; fields are separated by spaces. Both the thresholds and the range limits help the monitoring system generate alerts and scale graphs.

Example: "DISK OK - free space: 42.3% left | free_space=42.3%;20;10;0;100 used_space=2.1GB;1;2;0;5"

Not all monitoring stacks use Nagios-style perfdata. Some exporters implement bare JSON or other formats. Prometheus uses a dedicated exposition format rather than a pipe-delimited perfdata string. Perfdata remains popular due to its compactness and compatibility with legacy plugins.

Best practices include keeping labels concise and unique, avoiding spaces, normalizing units, including sensible min and max to enable scaling, and avoiding leakage of sensitive data. Use graphs to complement thresholds and document the metric definitions to aid interpretation.