rolling-count

Input Stack:

n: Int

expr: TimeSeriesExpr

⇨

Output Stack:

TimeSeriesExpr

Count the number of non-zero occurrences within a rolling window of datapoints. This operation is particularly valuable for alerting to reduce noise and ensure conditions persist over multiple datapoints before triggering alerts.

Parameters¶

expr: The time series expression to count occurrences for
n: Window size (number of datapoints including current value)

Counting Rules¶

Non-zero values: Only values != 0 are counted as occurrences
Missing data: NaN values are treated as zero (not counted)
Negative values: Negative non-zero values are counted as occurrences
Window-based: Uses fixed number of datapoints, not time duration

Alerting Pattern¶

A common alerting pattern using rolling count:

# Check if average CPU usage > 80%
name,cpuUser,:eq,:avg,80,:gt,

# Only alert if condition is true for 3+ of the last 5 datapoints
5,:rolling-count,3,:gt

Data Processing Example¶

Input	3,:rolling-count
0	0
1	1
-1	2
NaN	2
0	1
1	1
1	2
1	3
1	3
0	2

Examples¶

Counting positive occurrences in a 5-point window:

Before	After

:random, 0.4,:gt	:random, 0.4,:gt, 5,:rolling-count

:rolling-mean - Rolling average of values
:rolling-max - Rolling maximum values
:rolling-min - Rolling minimum values
:gt - Generate boolean conditions for counting