summary metric

A metric type that calculates percentiles inside the application process and exports the finished answers.

In more detail

A summary is exact for the one process that computed it, but percentiles cannot be averaged or added, so summaries from several instances cannot be combined into a service-wide figure. Prometheus recommends histograms in most cases; a summary suits a single process that needs an exact figure.

Example

Averaging two instances' 99th percentiles gave 0.6 s while the true 99th percentile of both was 0.9 s.

Related terms

Taught in

References

Certification courses