Hands-On Infrastructure Monitoring with Prometheus

histogram_quantile()

This function takes a float, which defines the required quantile (0 ≤ φ ≤ 1), and an instant vector of the gauge type as arguments. Each time series must have a le label (which means less than or equal to) to represent the upper bound of a bucket. This function also expects one of the selected time series to have a bucket named such as +Inf, which works as the catch-all, the last bucket of the cumulative histogram. Since histograms that are generated by Prometheus client libraries use counters for each bucket, rate() needs to be applied to convert them into gauges for this function to do its work. The time that's range selected for the range vector will then correspond to the window for the quantile calculation. Although ...