Datadog’s graphs use time aggregation to reduce the point count on a timeframe. This is done for performance reasons, because granularity higher than 350 data-points in a graph doesn’t provide additional insight.
Previously, we allowed the creation of monitors that use
max monitor aggregation with the
The only available query is
sum, which is the only mathematically accurate function with such behavior. This behavior applies to creating new monitors only, and does not affect editing existing monitors.
avg aggregation is applied on a count metric, it now performs AVG([5,1,2,1]) instead of AVG([5,1,2,Null,Null,Null,1]). This means that we are summing over all valid counts and dividing it by an arbitrary number of buckets with a non-null value.
If you zoom out - you might get an average of [6,2,1] which would produce a different result - if we think about number of buckets as how zoomed in/ out we arem then the zoom impacts the result.
As a workaround, change this monitor to alert on the
as_rate() form of this metric.