OTLP metrics export to Prometheus

Introduction

Prometheus was designed and optimized for pull-based monitoring, where it discovers targets and scrapes metrics endpoints at regular intervals. This model is central to its architecture, supporting features like service discovery and consistent target-based collection.

With the growing adoption of OpenTelemetry, newer versions of Prometheus have introduced support for receiving push-based metrics via OTLP. In this setup, OpenTelemetry SDKs export metrics using OTLP over HTTP, and Prometheus acts as an OTLP receiver instead of scraping metrics. This approach can be used in simpler setups, experiments, or local development environments. However, for production deployments using OpenTelemetry, it is strongly recommended to use an OpenTelemetry Collector as an intermediary.

This guide explains how to configure direct OTLP metric export from OpenTelemetry SDKs to a Prometheus OTLP endpoint. It covers required environment variables, exporter configuration, and key considerations such as service identification, export intervals, and operational trade-offs.

Prerequisite

Before you begin, make sure the following requirements are met:

• Set up Prometheus. Follow the example prometheus.yml configuration in this Prometheus guide.
• Enable the OTLP receiver

Once you have Prometheus set up, you can move on to configure your application to send metrics directly to an OTLP ingestion endpoint.

Use environment variables

You can configure OpenTelemetry SDKs and instrumentation libraries with standard environment variables. Set the environment variables before starting your application. The following OpenTelemetry variables are needed to send OpenTelemetry metrics to a Prometheus server on localhost:

export OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
export OTEL_EXPORTER_OTLP_METRICS_ENDPOINT=http://localhost:9090/api/v1/otlp

Turn off traces and logs when using Prometheus if you only need metrics:

export OTEL_TRACES_EXPORTER=none
export OTEL_LOGS_EXPORTER=none

The default push interval for OpenTelemetry metrics is 60 seconds. This can be adjusted depending on monitoring requirements. For example, a 15-second interval provides more responsive metrics and faster alerting at the cost of higher network and processing overhead.

export OTEL_METRIC_EXPORT_INTERVAL=15000

If your instrumentation library does not provide service.name and service.instance.id out-of-the-box, it is highly recommended to set them. Without these attributes, it becomes difficult to reliably identify services or distinguish between instances, making debugging and aggregation significantly harder. The example below assumes that the uuidgen command is available on your system.

export OTEL_SERVICE_NAME="my-example-service"
export OTEL_RESOURCE_ATTRIBUTES="service.instance.id=$(uuidgen)"

Configure telemetry

Update your OpenTelemetry configuration to use the same exporter and reader from the OTLP setup in your language SDK documentation. If the environment variables are set up and loaded correctly, the OpenTelemetry SDK reads them automatically.