Skip to main content

04 – Grafana Tempo Deep-Dive

Learning Objectives​

  • Explain Tempo’s architecture and index-free design.
  • Compare Tempo to Jaeger/Zipkin/X-Ray/vendor backends.
  • Understand push ingestion and query flow.

Architecture Overview​

  • Distributor → Ingester → Compactor → Storage → Query Frontend/Querier.
  • Object storage (S3/GCS/Azure/MinIO) holds blocks of trace data; compaction improves query speed.
  • Index-free: avoids costly secondary indexes; TraceQL + block metadata power queries.

Push vs Pull​

  • Prometheus pulls metrics; Tempo receives pushes from collectors/SDKs using OTLP/Jaeger.

Tempo vs Others​

  • Jaeger backend relies on indices in ES/Cassandra; Tempo optimizes for cheap object storage.
  • Zipkin: simple, but not built for multi-tenant petabyte scale.
  • X-Ray/vendor APM: managed, but can be costlier or lock you in. Tempo keeps data portable.

Minimal Tempo Config (filesystem for dev)​

Create tempo.yaml:

server:
  http_listen_port: 3200

distributor:
  receivers:
    otlp:
      protocols:
        grpc:
        http:

storage:
  trace:
    backend: local
    local:
      path: /tmp/tempo/traces

compactor:
  compaction:
    block_retention: 168h  # 7 days

Hands-on Lab​

  1. Start Tempo with the config above (via Docker in Section 5).
  2. Verify Tempo UI/API on http://localhost:3200 (Explore via Grafana in Section 5).

Deliverables​

  • Running Tempo accepting OTLP.

Quiz (Self-check)​

  • Name core Tempo components and their roles.
  • Why is index-free storage a cost advantage?

Resources​

  • Grafana Tempo Architecture Docs
  • Tempo Config Reference

Visuals​

Tempo Architecture Diagram​