Documentation feedback: /docs/sources/mimir/visualize/native-histograms/_index.md

[ccronemberger]

For the histogram_count description I think it would be helpful to mention that the count is cumulative.

[dimitarvdimitrov]

@krajorama @tacole02 can you guys take a look at this one?

[tacole02]

I can add a note to https://grafana.com/docs/mimir/latest/visualize/native-histograms/#query-your-histograms-count-or-sum stating that the count is cumulative. What do you think, @krajorama ?

[krajorama]

Hi,

I don't quite understand the comment.

In what sense would you say the count is cumulative ?

If we look at the count compared to individual buckets in a single native histogram, it is true that the count is a sum of all individual bucket counts - as stated in the documentation - but I wouldn't say "cumulative" because that could be confused with something else.

If we look at it from a time series perspective, the count may or may not be cumulative. Currently programs will emit native histograms that count observations over time - so cumulative. However you can write a recording rule to record the rate of such native histograms, in which case the resulting native histogram will store rate values , which are not cumulative.

The function histogram_count works the same on cumulative and non-cumulative (i.e. gauge) native histograms, it just returns the count value.

I feel this is beyond the scope of our documentation, but what we could do is to reference the new specification document that @beorn7 wrote recently: https://prometheus.io/docs/specs/native_histograms/. The document does mention the above flavors: https://prometheus.io/docs/specs/native_histograms/#flavors.

[tacole02]

I think this is a really good idea, @krajorama . Adding a link to the spec could provide more useful context overall. Let me know if this is how you'd like to proceed, and I can make the change.