Custom Metrics Example

This GoFr example demonstrates the use of custom metrics through a simple HTTP server that creates and populates metrics. GoFr by default pushes metrics to port 2121 on /metrics endpoint.

To run the example use the command below:

console

go run main.go

Exporting via OTLP (push)

The same application can push these metrics to any OTLP collector or backend without touching the Go code — it's purely a configs/.env change. GoFr builds its meter with both a Prometheus pull reader and (when METRICS_EXPORTER=otlp) an OTLP push reader on the same instruments, so /metrics stays scrapeable and metrics are pushed — no double-counting.

Config	Meaning
`METRICS_EXPORTER`	`otlp` to enable push (unset/`prometheus` = pull only)
`METRICS_URL`	collector/backend endpoint (`host:port` for gRPC, URL for HTTP)
`METRICS_PROTOCOL`	`grpc` (default) or `http`
`METRICS_EXPORT_INTERVAL`	push interval, seconds (default `30`)
`METRICS_TEMPORALITY`	`cumulative` (default), `delta`, or `lowmemory`
`METRICS_HEADERS`	`key=value,key=value` request headers (auth, routing)
`METRICS_AUTH_KEY`	shorthand: sets the `Authorization` header
`METRICS_INSECURE`	`false` (default) keeps TLS; set `true` only for a local plaintext collector
`METRICS_PORT`	pull endpoint port; `0` disables it (push-only)

METRICS_PORT controls pull and METRICS_EXPORTER controls push, fully independently. Push-only serverless mode = METRICS_EXPORTER=otlp + METRICS_PORT=0.

Run locally

Start a collector (prints received metrics, re-exposes them for Prometheus):

bash

cd docker && docker compose up

Add the OTLP settings to configs/.env — a local collector listens in plaintext, so this is one of the few cases where METRICS_INSECURE=true is appropriate:

METRICS_EXPORTER=otlp
METRICS_URL=localhost:4317
METRICS_INSECURE=true

Then run the app and generate some metrics:

bash

go run .
curl -XPOST localhost:9011/transaction

Watch them arrive in the collector's logs, or browse Prometheus at localhost:9090. If nothing shows up, confirm METRICS_EXPORTER=otlp is set — without it the app only serves the pull endpoint.

Backend recipes (env only — no code changes)

Managed backends terminate TLS, so keep the secure default (METRICS_INSECURE unset/false).

Datadog — HTTP intake, delta temporality required, static API key:

METRICS_EXPORTER=otlp
METRICS_PROTOCOL=http
METRICS_URL=https://api.datadoghq.com/api/intake/otlp/v1/metrics
METRICS_TEMPORALITY=delta
METRICS_HEADERS=dd-api-key=<YOUR_DD_API_KEY>

Grafana Cloud — OTLP gateway, Basic auth:

METRICS_EXPORTER=otlp
METRICS_PROTOCOL=http
METRICS_URL=https://otlp-gateway-<region>.grafana.net/otlp
METRICS_AUTH_KEY=Basic <base64(instanceID:token)>

New Relic — OTLP endpoint, license-key header:

METRICS_EXPORTER=otlp
METRICS_URL=otlp.nr-data.net:4317
METRICS_HEADERS=api-key=<YOUR_NR_LICENSE_KEY>

Google Managed Prometheus (keyless) — see examples/using-gcp-metrics.