Documentation Grafana Cloud Observe and act Instrument and send data Grafana Alloy Introduction Telemetry flow

How telemetry flows through Grafana Alloy

Alloy moves telemetry through connected components, from sources to backends. Your configuration defines those components and the connections between them. Follow those connections to see where data enters the pipeline and where it ends up.

Define every connection explicitly

Alloy doesn’t transform, route, or process telemetry unless your configuration tells it to. You define every connection between components, and telemetry moves only along the paths you create.

A changed component alters the telemetry it handles. A missing connection means data doesn’t reach the next stage.

Components connect through exports, receiver references, and attributes such as forward_to. Build data pipelines explains how these connections form pipelines.

Follow the pipeline stages

Connected components form pipelines. Most telemetry pipelines use some combination of four functions: discovery, ingestion, transformation, and output. Discovery and transformation are optional, and you can chain multiple components in the same function or branch to multiple outputs.

flowchart LR

  Discovery[Discovery]
  Ingestion[Ingestion]
  Transformation[Transformation]
  Output[Output]

  Discovery -.->|targets| Ingestion
  Ingestion -->|telemetry| Transformation
  Transformation -->|telemetry| Output

  %% Grafana styling
  classDef grafana fill:#ffffff,stroke:#F05A28,stroke-width:2px,rx:8,ry:8,color:#1f2937,font-weight:600;

  class Discovery,Ingestion,Transformation,Output grafana

In pull-based pipelines, discovery components pass scrape targets to ingestion components such as prometheus.scrape. OpenTelemetry pipelines start at otelcol.receiver.* and skip discovery.

Ingestion collects or receives telemetry and converts it to an internal format. Transformation components modify, filter, route, or sample that data. You can also connect ingestion directly to output and skip transformation.

Output components send telemetry to backends.

When a component supports multiple signal types, connect each type separately through the pipeline.

Build data pipelines has multi-stage examples and pipeline patterns. Choose a component helps you pick components by signal type.

See how telemetry moves in a configuration

When you read a configuration, follow the data path from source to destination:

1. Start at ingestion components and note what signal type each one handles.
2. If the pipeline uses discovery, follow targets from discovery components into ingestion.
3. Follow each component’s output to the next component in the chain.
4. Note any transformation components and what they change.
5. Identify where each path ends at an output component.

Connection order determines execution order, not the textual order of components in the file. Pipelines can branch to multiple outputs or share components across paths. Pipeline patterns covers fan-out and chain processing.

The Alloy UI visualizes these connections. Use Debug to inspect component pipelines in a running instance.

Next steps

• Start with Get started for configuration syntax and component basics.
• Use Build data pipelines to connect components and apply pipeline patterns.
• Use Choose a component to pick components for metrics, logs, traces, and profiles.
• Follow Collect and forward data for end-to-end collection examples.
• Read How Alloy works for architecture and capabilities.
• Look up behavior in the Component reference.