fasthttp prometheus-middleware

Prometheus middleware for fasthttp

Exports metrics for request duration request_duration_seconds with http status code as code and http request method + endpoint/route as path f.e code="200",path="GET_/health", code="201",path="POST_/foo"

Example

using fasthttp/router

package main

import (
"log"

fasthttpprom "github.com/carousell/fasthttp-prometheus-middleware"
"github.com/fasthttp/router"
"github.com/valyala/fasthttp"
)

func main() {

	r := router.New()
	p := fasthttpprom.NewPrometheus("")
	p.Use(r)

	r.GET("/health", func(ctx *fasthttp.RequestCtx) {
		ctx.SetStatusCode(200)
		ctx.SetBody([]byte(`{"status": "pass"}`))
		log.Println(string(ctx.Request.URI().Path()))
	})

	log.Println("main is listening on ", "8080")
	log.Fatal(fasthttp.ListenAndServe(":"+"8080", p.Handler))

}

Example metrics for above code in /metrics endpoint

request_duration_seconds_bucket{code="200",path="GET_/health",le="0.01"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.02"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.04"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.06"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.08"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.1"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.15"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.25"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.4"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.6"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="0.8"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="1"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="1.5"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="2"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="3"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="5"} 25063
request_duration_seconds_bucket{code="200",path="GET_/health",le="+Inf"} 25063
request_duration_seconds_sum{code="200",path="GET_/health"} 0.14781658099999923
request_duration_seconds_count{code="200",path="GET_/health"} 25063

Claude Code skills

This repo ships two project skills (in .claude/skills/) for common maintenance work. In Claude Code, invoke them by name or with a slash command.

verify-metrics

Boots the example server, fires representative requests, scrapes /metrics, and reports the emitted request_duration_seconds series (labels + bucket boundaries). Since the library has no unit tests, this is the primary way to confirm a change is correct.

Use it after any change to prometheus.go — bucket edits, dependency upgrades, or labeling logic. Ask Claude:

/verify-metrics

It confirms, among other things, that /values/123 collapses to the label path="GET_/values/{id}" (cardinality control) rather than exploding per-id.

tune-buckets

Safely changes the histogram bucket boundaries for request_duration_seconds (the Buckets slice in prometheus.go). It restates the resulting slice, keeps values strictly ascending, verifies the output via verify-metrics, and flags the downstream blast radius (dashboards / alerts / SLOs) since these boundaries are consumed by everyone who upgrades. Ask Claude:

/tune-buckets add a 0.05 bucket and remove 5