Uptrace — Open Source APM with OpenTelemetry

Introduction

Uptrace is an open-source APM (Application Performance Monitoring) platform built on OpenTelemetry. It collects distributed traces, metrics, and logs from your applications and infrastructure, stores them in ClickHouse for high-performance querying, and presents them through an intuitive web dashboard.

What Uptrace Does

• Collects distributed traces, metrics, and logs via the OpenTelemetry protocol (OTLP)
• Stores telemetry data in ClickHouse for fast aggregation and long-term retention
• Provides a web dashboard for exploring traces, analyzing latency, and debugging errors
• Supports alerting rules based on metrics thresholds and error rate conditions
• Groups similar spans and errors automatically to reduce alert noise

Architecture Overview

Uptrace runs as a single Go binary that accepts OTLP data over gRPC and HTTP. Incoming traces are processed, grouped by service and operation, and written to ClickHouse tables optimized for time-series queries. Metrics are stored in ClickHouse using materialized views for efficient aggregation. The web frontend (built with Vue.js) queries the backend API to render trace waterfalls, service maps, and metric charts. ClickHouse handles both the hot path (recent data) and cold storage (historical data) in a single deployment.

Self-Hosting & Configuration

• Deploy with Docker Compose including Uptrace, ClickHouse, and PostgreSQL (for metadata)
• Configure via a YAML file specifying ClickHouse connection, retention policies, and projects
• Set up projects and DSN tokens to separate telemetry by application or team
• Configure alerting channels (email, Slack, PagerDuty) for threshold-based notifications
• Scale by tuning ClickHouse resources — a single node handles millions of spans per day

Key Features

• Native OpenTelemetry support — no proprietary agents or SDKs required
• ClickHouse backend delivers sub-second query performance across billions of spans
• Automatic span grouping surfaces the most impactful performance issues
• Multi-project support with separate DSN tokens for access control
• Built-in metrics dashboards for system and application-level KPIs

Comparison with Similar Tools

• Jaeger — trace visualization only; Uptrace adds metrics, logs, alerting, and span grouping
• SigNoz — similar OpenTelemetry APM; Uptrace uses ClickHouse directly for simpler operations
• Grafana Tempo — trace backend that requires Grafana for visualization; Uptrace is an all-in-one package
• Datadog — commercial APM; Uptrace is fully open-source and self-hosted
• New Relic — commercial full-stack observability; Uptrace covers core APM needs at zero license cost

FAQ

Q: Does Uptrace replace Prometheus? A: Uptrace can ingest and store OpenTelemetry metrics, but it does not replace Prometheus for pull-based monitoring. Many teams use both — Prometheus for infrastructure and Uptrace for application traces.

Q: How much storage does Uptrace need? A: It depends on trace volume. ClickHouse compresses data efficiently. A typical application sending 1 million spans per day uses roughly 1-2 GB of storage per month.

Q: Can I send data from applications already instrumented with OpenTelemetry? A: Yes. Point your OTLP exporter at Uptrace and add the DSN header. No code changes beyond exporter configuration are needed.

Q: Is there a hosted version? A: Uptrace offers a cloud-hosted option for teams that prefer not to self-host. The open-source and hosted versions share the same codebase.

Sources

• https://github.com/uptrace/uptrace
• https://uptrace.dev/get/open-source-apm.html