OpenTelemetry Glossary

This glossary defines key terms and concepts used throughout the OpenTelemetry course.

A-C

Attribute

A key-value pair containing metadata that annotates a Span, Metric, or Log Record. Attributes provide context like http.method, db.statement, or customer.id.

Baggage

A mechanism for propagating custom key-value pairs across process boundaries along with the trace context. Unlike attributes, baggage is available to all downstream services but is not automatically recorded as telemetry unless explicitly configured.

Cardinality

The number of unique values in a set. In metrics, High Cardinality refers to attributes with many unique values (e.g., User IDs), which can cause excessive memory usage and storage costs in time-series databases.

Collector

A vendor-agnostic proxy that can receive, process, and export telemetry data. It consists of Receivers, Processors, and Exporters.

Consistent Hashing

A technique used in distributed systems to distribute keys (like Trace IDs) across a changing set of nodes (Collectors) with minimal rebalancing. Essential for ensuring Trace Affinity in Tail Sampling.

Context

The state that is propagated across service boundaries. It typically includes the Trace ID, Span ID, and Baggage.

D-L

Exporter

A component in the SDK or Collector responsible for sending telemetry data to a backend (e.g., Jaeger, Prometheus, OTLP endpoint).

Extension

A component in the Collector that provides capabilities not directly related to processing telemetry, such as health checks (health_check), performance profiling (pprof), or authentication (oidc).

Head-Based Sampling

A sampling strategy where the decision to sample a trace is made at the very beginning (the “head”) of the request. This is efficient but may miss interesting traces that error out later.

Instrumentation

The code that gathers telemetry. It can be Manual (using the API directly) or Automatic (using agents or libraries to attach to frameworks).

Log Record

A recording of an event. In OpenTelemetry, logs are structured and can be correlated with Traces and Spans.

M-O

Metric

A numerical measurement recorded over time. Common types include Counters, Gauges, and Histograms.

OTLP (OpenTelemetry Protocol)

The native protocol used by OpenTelemetry for encoding and transmitting telemetry data. It is efficient, gRPC-based (with HTTP/JSON options), and vendor-neutral.

P-R

ParentBased

A sampling strategy that respects the sampling decision of the parent span. If the parent was sampled, the child is sampled. If the parent was dropped, the child is dropped.

Processor

A component in the Collector or SDK that modifies telemetry data before it is exported (e.g., batching, sampling, filtering, attributes modification).

Propagation

The mechanism of moving Context (Trace ID, Span ID, Baggage) from one process to another, typically via HTTP headers (e.g., W3C Trace Context).

Receiver

A component in the Collector responsible for ingesting data. Common receivers include OTLP (for OTel data), Jaeger (for legacy traces), and Host Metrics (for system stats).

Remote Sampling

A mechanism where the SDK polls a central authority (like the Collector or Jaeger) to dynamically fetch sampling configuration, allowing changes without redeployments.

Resource

A set of attributes that describe the entity producing telemetry, such as service.name, host.name, or deployment.environment.

S-Z

Sampler

A component that decides whether a trace should be recorded and exported.

Service (Collector)

In the context of the Collector configuration, a Service defines the enabled pipelines (traces, metrics, logs) and the extensions that should be active.

Span

The basic unit of work in a trace. A span represents an operation (e.g., an HTTP request, a database query) and contains a start time, duration, attributes, logs, and a reference to a parent span.

Tail-Based Sampling

A sampling strategy where the decision to sample is made after the entire trace has been collected (at the “tail”). This allows for intelligent decisions (e.g., “Keep 100% of errors”) but requires buffering all spans in memory.

Trace

A representation of a request’s execution path through a distributed system. It is a tree of Spans.

TraceIDRatio

A probabilistic sampling strategy that samples a fixed percentage of traces based on the hash of their Trace ID.

Tracer

The API interface used to create Spans.

W3C Trace Context

The standard HTTP header format (traceparent, tracestate) used by OpenTelemetry for context propagation.