PromQL Cheat Sheet

sum()

Sums values for time series across a dimension.

sum(rate(http_requests_total[5m]))

avg()

Calculates the average.

avg(rate(http_requests_total[5m]))

min()

Finds the minimum value.

min(rate(http_requests_total[5m]))

max()

Finds the maximum value.

max(rate(http_requests_total[5m]))

count()

Counts the number of series.

count(rate(http_requests_total[5m]))

stddev()

Standard deviation of values.

stddev(rate(http_requests_total[5m]))

stdvar()

Variance of values.

stdvar(rate(http_requests_total[5m]))

topk(k, expr)

Top k time series by value.

topk(5, rate(http_requests_total[5m]))

bottomk(k, expr)

Bottom k time series by value.

bottomk(5, rate(http_requests_total[5m]))

count_values()

Counts occurrences of values and groups them.

count_values("version", up)

rate()

Calculates per-second average rate of increase for counters over a time window.

rate(http_requests_total[5m])

irate()

Instantaneous rate of increase (based on the latest two samples).

irate(http_requests_total[5m])

increase()

Total increase of a counter in the specified time range.

increase(http_requests_total[5m])

delta()

Calculates difference between first and last value in range.

delta(cpu_usage_total[5m])

deriv()

Computes the per-second derivative.

deriv(cpu_usage_total[5m])

predict_linear()

Predicts the value of a gauge in the future based on linear regression.

predict_linear(cpu_usage_total[5m], 3600)

abs()

Returns the absolute value of a series.

abs(cpu_usage)

clamp_max()

Limits the maximum value.

clamp_max(cpu_usage, 80)

clamp_min()

Limits the minimum value.

clamp_min(cpu_usage, 20)

ceil()

Rounds values up to the nearest integer.

ceil(cpu_usage)

floor()

Rounds values down to the nearest integer.

floor(cpu_usage)

round()

Rounds values to the nearest integer or step.

round(cpu_usage, 0.1)

sort()

Sorts series by value (ascending).

sort(rate(http_requests_total[5m]))

sort_desc()

Sorts series by value (descending).

sort_desc(rate(http_requests_total[5m]))

time()

Returns the current time as a Unix timestamp (seconds).

time()

timestamp()

Returns the timestamp of the sample.

timestamp(cpu_usage)

offset

Shifts the time series in the past by a duration.

rate(http_requests_total[5m] offset 1h)

ignoring

Ignores certain labels in vector matching.

up == ignoring(instance) up

on

Keeps only specified labels for vector matching.

up == on(instance) up

group_left

Allows many-to-one matching.

up * on(instance) group_left(job) up

group_right

Allows one-to-many matching.

up * on(instance) group_right(job) up

label_replace()

Replaces or creates a label.

label_replace(up, "job", "new", "job", "old")

label_join()

Joins label values into a single string.

label_join(up, "instance_label", ",", "instance", "job")

vector()

Converts a scalar into a time series.

vector(5)

scalar()

Converts a time series into a scalar.

scalar(up)

bool

Compares values as a boolean (0 or 1) instead of filtering.

up == bool 1

Instant Vector

Single value per time series at a point in time.

up

Range Vector

Set of values over time for a series.

http_requests_total[5m]

+, -, *, /

Basic arithmetic.

cpu_usage + 5

%

Modulo operator.

cpu_usage % 2

^

Exponentiation.

cpu_usage ^ 2

==

Equal to.

up == 1

!=

Not equal to.

up != 1

>

Greater than.

up > 0

<

Less than.

up < 1

>=

Greater than or equal.

up >= 1

⇐

Less than or equal.

up ⇐ 1