This is documentation for the next version of Mimir. For the latest stable release, go to the latest version.

Documentation Grafana Mimir Visualize Native histograms

Visualize native histograms

Prometheus native histograms are a data type in the Prometheus ecosystem that allow you to produce, store, and query a high-resolution histogram of observations. To learn more about the native histograms data type and how to start sending native histograms to Grafana Mimir, refer to Send native histograms to Mimir.

Prometheus Query Language

Use the Prometheus Query Language (PromQL) to query native histogram metrics. The data type of your query results depends on the query and the underlying data.

For more information about PromQL, refer to Querying Prometheus.

The following examples show common ways to derive floating point data type from native histograms data for visualizations and also how to convert existing queries using classic histograms into queries using native histograms.

Note that the native histogram queries do not include the _bucket, _sum and _count suffixes of classic histograms.

Query your histogram’s count or sum

To query the total count of observations within a histogram, use the following queries:

PromQL Copy

# Native histograms:
histogram_count(sum(request_duration_seconds))

# Classic histograms:
sum(request_duration_seconds_count)

To query the total sum of observed values, use the following query:

PromQL Copy

# Native histograms:
histogram_sum(sum(request_duration_seconds))

# Classic histograms:
sum(request_duration_seconds_sum)

Find rate of observations

To query the rate of all observations calculated over 5 minute time window, use the following query:

PromQL Copy

# Native histograms:
histogram_count(sum(rate(request_duration_seconds[5m])))

# Classic histograms:
sum(rate(request_duration_seconds_count[5m]))

To query the rate of observations between two values such as 0 and 2 seconds, use the following query:

PromQL Copy

# Native histograms:
histogram_fraction(0, 2, sum(rate(request_duration_seconds[5m])))
*
histogram_count(sum(rate(request_duration_seconds[5m])))

# Classic histograms:
sum(rate(request_duration_seconds_bucket{le="2.5"}[5m]))

There is a native histogram function that estimates the fraction of the total number of observations that fall within a certain interval, such as [0, 2]. For more information, refer to histogram fraction.

Classic histograms have no such function. Therefore, if the lower and upper bounds of the interval do not line up with the bucket boundaries of a classic histogram, you have to estimate the fraction manually.

Quantiles

To query an upper bound of observation values that 95% of observations fall under, use the following query:

PromQL Copy

# Native histograms:
histogram_quantile(0.95, sum(rate(request_duration_seconds[5m])))

# Classic histograms:
histogram_quantile(0.95, sum by (le) (rate(request_duration_seconds_bucket[5m])))

Create Grafana dashboards

The panel types Histogram and Heatmap are compatible with the native histogram data type. Use these panel types to visualize a query, such as:

PromQL Copy

sum(rate(request_duration_seconds[$__rate_interval]))

For all other panel types, for example Timeseries, use one of the histogram functions in the Prometheus Query Language to visualize a derived float time series.

Grafana classic explore

In Explore, use one of the histogram functions in Prometheus Query Language to visualize a derived float time series.