OpenTelemetry three pillars - metrics, logs, traces

OpenTelemetry (OTel) is the standard for observability. It unifies metrics, logs, and traces under one framework. This lesson covers the three pillars and the OTel components.

The three pillars

Each pillar answers a different question:

• Metrics for aggregate behaviour and alerts.
• Logs for events and debugging.
• Traces for latency analysis across services.

Why unify?

Historically, three separate systems:

• Prometheus for metrics.
• ELK / Loki for logs.
• Jaeger / Zipkin for traces.

Each has its own agent, format, and query language. Switching between them is manual.

OTel unifies them:

• One agent (the OTel Collector) ingests all three.
• One schema (OpenTelemetry Protocol, OTLP) for transport.
• One set of tools for visualisation (Grafana, Honeycomb, Lightstep).

OpenTelemetry components

• OTel SDK: library for applications to emit metrics/logs/traces.
• OTel Collector: receives OTLP, processes, exports to backends.
• OTel Protocol (OTLP): the wire format.
• Auto-instrumentation: agents that auto-instrument common libraries (Java, Python, Node, etc.).

Pillar-specific details

Metrics

• Counters, gauges, histograms.
• Used for: aggregate behaviour, alerting, capacity planning.
• Backend: Prometheus, Datadog, New Relic, etc.

Logs

• Structured events with attributes.
• Used for: debugging, event correlation, audit.
• Backend: Loki, Elasticsearch, Splunk, etc.

Traces

• Spans representing units of work.
• A trace is a tree of spans (e.g. one HTTP request → multiple service calls).
• Used for: latency analysis, dependency mapping.
• Backend: Jaeger, Tempo, Zipkin, Honeycomb.

OTel data model

A common data model across pillars:

• Resource: the entity (host, service, container).
• Attribute: a key-value (e.g. http.method=GET).
• Event: a log record.
• Span: a unit of work in a trace.
• Metric data point: a measurement.

All have a resource, attributes, and a timestamp. The schema is consistent across pillars.

OTel Collector

The Collector is the workhorse:

# /etc/otel-collector/config.yaml
receivers:
  otlp:
    protocols:
      grpc:
      http:

processors:
  batch:

exporters:
  prometheus:
    endpoint: 0.0.0.0:8889
  loki:
    endpoint: http://loki:3100/loki/api/v1/push
  jaeger:
    endpoint: jaeger:14250

service:
  pipelines:
    metrics:
      receivers: [otlp]
      processors: [batch]
      exporters: [prometheus]
    logs:
      receivers: [otlp]
      processors: [batch]
      exporters: [loki]
    traces:
      receivers: [otlp]
      processors: [batch]
      exporters: [jaeger]

One Collector, multiple pipelines, multiple backends.

Knowledge check