mirrors/prometheus
1
0
Fork 0
mirror of https://github.com/prometheus/prometheus.git synced 2025-08-06 14:17:12 +02:00
Code Issues Projects Releases Wiki Activity

promql: add tests to demonstrate extrapolation below zero

Browse Source 
This shows how float counters cannot go below zero when extrapolationg
for rate/increase, and how histograms do not have that protection yet,
leading to an overestimation of the rate/increase.

This also demonstrates edge cases where the count extrapolation does
not need to be limited, but an individual bucket still goes below
zero.

Signed-off-by: beorn7 <beorn@grafana.com>
This commit is contained in:
beorn7 2025-07-03 14:05:19 +02:00
parent 419d436a44
commit c0a13223e7
1 changed files with 35 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
promql/promqltest/testdata/native_histograms.test vendored
View File

@ -1373,3 +1373,38 @@ eval instant at 1m histogram_fraction(-Inf, +Inf, histogram_nan)
expect info msg: PromQL info: input to histogram_fraction has NaN observations, which are excluded from all fractions for metric name "histogram_nan"
{case="100% NaNs"} 0.0
{case="20% NaNs"} 0.8
clear
# Tests to demonstrate how an extrapolation below zero is prevented for a float counter, but not for native histograms.
# I.e. the float counter that behaves the same as the histogram count might yield a different result after `increase`.
load 1m
metric{type="histogram"} {{schema:0 count:15 sum:25 buckets:[5 10]}} {{schema:0 count:2490 sum:75 buckets:[15 2475]}}x55
metric{type="counter"} 15 2490x55
# End of range coincides with sample. Zero point of count is reached within the range.
eval instant at 55m increase(metric[90m])
{type="histogram"} {{count:2497.5 sum:50.45454545454545 counter_reset_hint:gauge buckets:[10.09090909090909 2487.409090909091]}}
{type="counter"} 2490
# End of range does not coincide with sample. Zero point of count is reached within the range.
eval instant at 54m30s increase(metric[90m])
{type="histogram"} {{count:2520.8333333333335 sum:50.92592592592593 counter_reset_hint:gauge buckets:[10.185185185185187 2510.6481481481483]}}
{type="counter"} 2512.9166666666665
# End of range coincides with sample. Zero point of count is reached outside of (i.e. before) the range.
# This means no change of extrapolation is required for the histogram count (and neither for the float counter),
# however, the 2nd bucket's extrapolation will reach zero within the range. The overestimation is visible
# easily here because the last sample in the range coincides with the boundary, where the 2nd bucket has
# a value of 2475 but has increased by 2476.2045454545455 according to the returned result.
eval instant at 55m increase(metric[55m15s])
{type="histogram"} {{count:2486.25 sum:50.227272727272734 counter_reset_hint:gauge buckets:[10.045454545454547 2476.2045454545455]}}
{type="counter"} 2486.25
# End of range does not coincide with sample. Zero point of count is reached outside of (i.e. before) the range.
# This means no change of extrapolation is required for the histogram count (and neither for the float counter),
# however, the 2nd bucket's extrapolation will reach zero within the range.
eval instant at 54m30s increase(metric[54m45s])
{type="histogram"} {{count:2509.375 sum:50.69444444444444 counter_reset_hint:gauge buckets:[10.13888888888889 2499.236111111111]}}
{type="counter"} 2509.375