Histogram metrics are a powerful way to capture the distribution of values for a given measurement, such as response times, request sizes, or durations. Compared to a counter that measures the overall total in a single value, a histogram counts the number of occurrences at different thresholds (called buckets). It is a multi-value counter, aggregating data into buckets, allowing you to analyze percentiles, averages, and the overall shape of your data.
For a deeper dive into histograms and their advantages, see the following series of the OpenTelemetry blog posts:
No special configuration is needed to ingest histograms from OpenTelemetry (OTLP ingest API) or Prometheus sources (Dynatrace Operator). Histogram metrics are sent automatically when your environment receives OpenTelemetry or Kubernetes data.
The OpenTelemetry Exponential Histogram is not supported as a histogram: the histogram's min|max|sum|count are ingested but the buckets aren't.
If any of below happens, the OpenTelemetry ingest API returns the 400 or 200 with partial success responses.
sum aren't ingested. This happens when negative values are recorded.NaN or Infinite are invalid.Dynatrace Query Language (DQL) allows you to analyze histogram metrics using the percentile aggregation. It calculates the requested percentile across all the buckets for each time slot.
For example:
timeseries percentile(http_request_duration_seconds_bucket, 99)
The query calculates the 99th percentile of values based on the ingested histogram buckets.
Histograms represent data as a series of buckets, each containing the count of values that fall within a specific range. For example, http_request_duration_seconds_bucket contains counts of request durations grouped by predefined ranges (le dimension).
The percentile function uses the bucket data to estimate the requested percentile (for example, the 99th percentile). It interpolates the value at which 99% of the observed data points fall below, based on the cumulative counts in the histogram buckets.
This query is particularly useful for analyzing latency or performance metrics, as percentiles provide insights into the worst-case scenarios experienced by users.
You can visualize histogram metrics in Dynatrace Notebooks or Dashboards using DQL queries. However, be aware of the following limitations:
timeseries command). Visualizing the distribution of occurrences across buckets for a given period of time (typically using a bar chart visualization) is NOT supported at this time.For more information, see Edit visualizations for Notebooks and Dashboards.
The timeseries percentile function, which is necessary to query histograms, is only available to DPS customers with the Metrics powered by Grail rate card. For more information, see Metrics powered by Grail overview (DPS).