PromQL builder

A set of libraries for writing and composing PromQL queries as-code in Go, Python or Typescript.

Examples

Go

package main

import (
    "fmt"

    "github.com/grafana/promql-builder/go/promql"
)

// time() - demo_batch_last_success_timestamp_seconds > 3600
func batchJobsWithNoSuccessInLastHour() *promql.BinaryExprBuilder {
    return promql.Gt(
        promql.Sub(
            promql.Time(),
            promql.Vector("demo_batch_last_success_timestamp_seconds"),
        ),
        promql.N(3600),
    )
}

// method_code:http_errors:rate5m{code="500"} / ignoring(code) method:http_requests:rate5m
func errorRatioPerHTTPMethod() *promql.BinaryExprBuilder {
    return promql.Div(
        promql.Vector("method_code:http_errors:rate5m").Label("code", "500"),
        promql.Vector("method:http_requests:rate5m"),
    ).Ignoring([]string{"code"})
}

// sum by(device) (node_filesystem_free_bytes)
func freeDiskSpacePerDevice() *promql.AggregationExprBuilder {
    return promql.Sum(
        promql.Vector("free_disk_space_per_device"),
    ).By([]string{"device"})
}

// 90th percentile request latency over last 5 minutes per path and method
// histogram_quantile(0.9, sum by(le, path, method) (
//	rate(demo_api_request_duration_seconds_bucket[5m])
// ))
func requestLatency90thPercentilePerPathAndMethod() *promql.FuncCallExprBuilder {
    return promql.HistogramQuantile(0.9,
        promql.Sum(
            promql.Rate(
                promql.Vector("demo_api_request_duration_seconds_bucket").Range("5m"),
            ),
        ).By([]string{"le", "path", "method"}),
    )
}

func main() {
    fmt.Println(batchJobsWithNoSuccessInLastHour())
    fmt.Println(errorRatioPerHTTPMethod())
    fmt.Println(freeDiskSpacePerDevice())
    fmt.Println(requestLatency90thPercentilePerPathAndMethod())
}

Python

from promql_builder.builders.promql import (
    div,
    gt,
    histogram_quantile,
    n,
    rate,
    sub,
    sum,
    time,
    vector,
)

# time() - demo_batch_last_success_timestamp_seconds > 3600
def batch_jobs_with_no_success_in_last_hour():
    return gt(
        sub(
            time(),
            vector("demo_batch_last_success_timestamp_seconds"),
        ),
        n(3600),
    )


# method_code:http_errors:rate5m{code="500"} / ignoring(code) method:http_requests:rate5m
def error_ratio_per_http_method():
    return div(
        vector("method_code:http_errors:rate5m").label("code", "500"),
        vector("method:http_requests:rate5m"),
    ).ignoring(['code'])


# sum by(device) (node_filesystem_free_bytes)
def free_disk_space_per_device():
    return sum(
        vector("free_disk_space_per_device"),
    ).by(["device"])


# 90th percentile request latency over last 5 minutes per path and method
# histogram_quantile(0.9, sum by(le, path, method) (
#	rate(demo_api_request_duration_seconds_bucket[5m])
# ))
def request_latency_90th_percentile_per_path_and_method():
    return histogram_quantile(0.9,
        sum(
            rate(vector("demo_api_request_duration_seconds_bucket").range("5m")),
        ).by(["le", "path", "method"]),
    )


if __name__ == '__main__':
    print(str(batch_jobs_with_no_success_in_last_hour()))
    print(str(error_ratio_per_http_method()))
    print(str(free_disk_space_per_device()))
    print(str(request_latency_90th_percentile_per_path_and_method()))

Typescript

import * as promql from "@grafana/promql-builder";

// time() - demo_batch_last_success_timestamp_seconds > 3600
function batchJobsWithNoSuccessInLastHour() {
    return promql.gt(
        promql.sub(
            promql.time(),
            promql.vector('demo_batch_last_success_timestamp_seconds'),
        ),
        promql.n(3600)
    );
}

// method_code:http_errors:rate5m{code="500"} / ignoring(code) method:http_requests:rate5m
function errorRatioPerHTTPMethod() {
    return promql.div(
        promql.vector("method_code:http_errors:rate5m").label("code", "500"),
        promql.vector("method:http_requests:rate5m"),
    ).ignoring(["code"]);
}

// Free disk space per device
// sum by(device) (node_filesystem_free_bytes{machine="foo"})
function freeDiskSpacePerDevice() {
    return promql.sum(promql.vector("free_disk_space_per_device").label("machine", "foo")).by(["device"]);
}

// 90th percentile request latency over last 5 minutes per path and method
// histogram_quantile(0.9, sum by(le, path, method) (
//	rate(demo_api_request_duration_seconds_bucket[5m])
// ))
function requestLatency90thPercentilePerPathAndMethod() {
    return promql.histogramQuantile(
        0.9,
        promql.sum(
            promql.rate(
                promql.vector("demo_api_request_duration_seconds_bucket").range("5m"),
            ),
        ).by(["le", "path", "method"]),
    );
}

const expressions = [
    freeDiskSpacePerDevice(),
    batchJobsWithNoSuccessInLastHour(),
    errorRatioPerHTTPMethod(),
    requestLatency90thPercentilePerPathAndMethod(),
];

expressions.forEach(expression => console.log(expression.toString()))

Maturity

This project should be considered as "public preview". While it is used by Grafana Labs, it is still under active development.

Additional information can be found in Release life cycle for Grafana Labs.

Note

Bugs and issues are handled solely by Engineering teams. On-call support or SLAs are not available.

License

Apache 2.0 License