OpenTelemetry Prometheus Exporter

opentelemetry-exporter-prometheus is the official OpenTelemetry Python exporter that bridges the OpenTelemetry metrics SDK to Prometheus. It converts OTel metric instruments (counters, up-down counters, gauges, and histograms) into Prometheus-format samples and exposes them for scraping via the standard prometheus_client library.

It plugs into the OpenTelemetry metrics pipeline as a MetricReader, so instrumentation written against the vendor-neutral OpenTelemetry API can be surfaced to a Prometheus server without changing application code. This lets teams standardize on OpenTelemetry instrumentation while keeping Prometheus as their metrics backend.

What You Get

• A PrometheusMetricReader that collects OpenTelemetry metrics for Prometheus scraping
• Mapping of OTel counters, up-down counters, gauges, and histograms to Prometheus metric types
• Integration with the prometheus_client library’s HTTP exposition
• An opentelemetry_metrics_exporter entry point for auto-configuration
• Correct handling of OTel attributes as Prometheus labels

Common Use Cases

• Exposing OpenTelemetry-instrumented application metrics to an existing Prometheus deployment
• Standardizing on vendor-neutral OTel instrumentation while keeping Prometheus as the metrics backend
• Scraping counters, gauges, and histograms from a Python service via a /metrics endpoint

Under The Hood

Architecture - The package lives in the opentelemetry-python monorepo at exporter/opentelemetry-exporter-prometheus and is deliberately small. src/opentelemetry/exporter/prometheus/__init__.py defines the PrometheusMetricReader (and an internal _AutoPrometheusMetricReader used by the auto-configuration entry point), which subclasses the OTel SDK MetricReader and, on collection, translates the SDK’s in-memory metric data into a Prometheus registry. _mapping.py holds the logic that maps OTel instrument types and attribute sets onto Prometheus metric names, types, and labels, delegating exposition to the prometheus_client Collector interface.

Tech Stack - Pure Python (3.10+), built with hatchling. It depends on opentelemetry-api and opentelemetry-sdk for the metrics pipeline and on prometheus_client (>=0.5.0) for Prometheus-format serialization and HTTP exposition.

Code Quality - The exporter ships a focused test suite — test_prometheus_exporter, test_mapping, and test_entrypoints — verifying instrument translation, label mapping, and auto-configuration wiring. As part of the CNCF OpenTelemetry project it follows that repo’s rigorous CI, linting, and review standards; it is marked Beta, with a documented limitation around multiprocessing environments.

API Design - Usage is idiomatic OpenTelemetry: construct a PrometheusMetricReader, pass it to a MeterProvider, and start the prometheus_client HTTP server — no exporter-specific glue is needed. The opentelemetry_metrics_exporter entry point enables fully zero-code setup under OTel auto-instrumentation. Familiarity with the OpenTelemetry metrics model is the main prerequisite, which sets a moderate learning curve.