Snowglobe
Fake but realistic telemetry generator to test your observability stack without real infrastructure
🔗 Visit Snowglobe
Description
Testing whether your monitoring dashboards, alerting rules, and tracing UI actually work well requires real traffic patterns — but standing up 20+ microservices just to generate test data is its own project. Snowglobe skips the real infrastructure entirely: it's a single small program that emits realistic OpenTelemetry logs, traces, and metrics as if a full distributed system were running, complete with the messy failures real systems produce.
It simulates up to 28 services (20 e-commerce-style plus 8 AI-focused) across 40 scenario flows, injecting the kind of problems that make observability tools earn their keep: lost messages, retry storms, rate limiting, saga compensation, timeout cascades, and general chaos scenarios. The AI-focused services emit spans following GenAI semantic conventions, so it's also useful for validating LLM-tracing tools specifically. It compiles to a single 6.4MB container with zero external dependencies, and you can tune complexity, throughput (2 to 350 traces per second), and error rates to match whatever load profile you're testing against. It's open source (Apache 2.0), actively maintained (69 commits, 17 stars, 2 forks at time of review), and completely free.
💬 Our review
The short version: Snowglobe solves a specific, unglamorous problem well — you need realistic telemetry to validate an observability setup, and standing up real services just to generate test data is wasteful busywork.
There's no well-known direct competitor doing exactly this; most teams either hand-roll synthetic traffic scripts or repurpose the official OpenTelemetry Demo app (a real, deployed e-commerce demo, not a lightweight generator), which is heavier to run and less focused on injecting realistic failure modes. Snowglobe's single 6.4MB binary with tunable throughput and built-in chaos scenarios (retry storms, saga compensation, timeout cascades) makes it meaningfully lighter-weight and more purpose-built for testing observability tooling specifically, including the newer GenAI/LLM tracing conventions most alternatives don't cover yet. Worth adding to your test suite if you're building or evaluating any observability/APM product; overkill if you just need a one-off smoke test.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Apache 2.0, gratuit, conteneur unique de 6.4 Mo sans dépendance externe.
Pros
Simule jusqu'à 28 microservices sans déployer de vraie infra
Injection réaliste de pannes (retry storms, saga compensation, timeout cascades)
Spans IA conformes aux conventions sémantiques GenAI
Conteneur unique 6.4 Mo, débit ajustable 2-350 traces/s, gratuit
Cons
Niche : utile seulement pour tester/valider des outils d'observabilité
Petite communauté (17 stars) malgré une maintenance active
Nécessite de comprendre OpenTelemetry pour en tirer profit