What is the difference between the OpAMP extension and the OpAMP supervisor in OpenTelemetry Collector deployments?

What makes the OpAMP extension different from the OpAMP supervisor?

The difference is mainly one of control plane depth. The OpAMP extension makes the collector visible to an OpAMP server and exposes its state, but it does not run the deployment. The OpAMP supervisor sits outside the collector process and can actively control lifecycle actions, which makes it the operational layer for managing collector instances rather than just observing them.

That distinction matters in real deployments because visibility and control answer different questions. An extension can tell you what a collector is doing; a supervisor can change what the collector does and keep it running across failures or updates. If you need only remote awareness, the extension is sufficient. If you need orchestration, the supervisor is the stronger component.

For a broader view of how OpenTelemetry pipelines are used in security and observability, see Observability, telemetry and SRE for security teams.

Where the operational boundary actually sits

The extension stays inside the collector’s own process model and is generally the lighter integration point. It is designed to publish read-only information such as description, health, and configuration so an OpAMP server can understand the collector without taking over the runtime. That makes it useful when the collector already has its own deployment tooling and you mainly want fleet visibility.

The supervisor is a separate process with authority over the collector lifecycle. It can push and receive configuration, start and stop the collector, restart it after failure, and coordinate updates. In practice, that means it is acting on behalf of the operator to manage availability and consistency, not merely reporting status. The separation is useful because it keeps the telemetry process and the orchestration process distinct.

This split is especially relevant when collectors are treated as managed estate, not just installed software. The extension fits environments where the collector is one component in a larger platform. The supervisor fits environments where the operator wants centralized rollout control, restart behavior, and coordinated configuration changes across many nodes.

For deployment patterns that emphasise pipeline handling and secure routing, the Security Data Pipeline Guide is the closest companion resource.

How to choose between them in practice

Choose the extension when the collector is already managed by your orchestration layer and OpAMP is there primarily for introspection or lightweight coordination. Choose the supervisor when you need OpAMP to be the control loop that keeps the collector aligned with desired state, especially when you care about restart behavior, rollout sequencing, or reacting to process failure without relying on another external manager.

The practical test is whether the tool needs authority over the process or only insight into it. If the answer is “insight,” the extension is enough. If the answer is “keep it running, reconfigure it, and update it,” you need the supervisor. That is why the two components are not competing implementations so much as different points on the same management spectrum.

In environments where the collector is part of a broader security or agentic observability stack, the operating model matters more than the feature list. The extension reduces blind spots, while the supervisor reduces drift and recovery time. A mature deployment may use both, but it should do so with a clear split between observation and control.

Practitioner Guidance

What to verify: Confirm whether your deployment architecture already has a process manager or orchestrator that owns restart and update behavior. If it does, the extension may be enough for OpAMP visibility; if it does not, the supervisor becomes the operational control point.

Decision rule: If the collector must be self-healing or remotely reconfigured as part of the OpAMP workflow, use the supervisor. If the collector only needs to publish state to an OpAMP server, use the extension and keep lifecycle control elsewhere.

Common mistake: Treating the extension as if it can replace runtime orchestration. That usually leads to a monitoring-only design that looks manageable in development but cannot enforce desired state, recover from failure, or coordinate updates in production.

Practitioner takeaway: The extension gives you telemetry about the collector, while the supervisor gives you operational authority over it. Pick the component that matches the level of control your deployment actually needs.

Related resources from NHI Mgmt Group

• OpenTelemetry Collector
• What do teams get wrong about batching and persistent queues in OpenTelemetry Collector deployments?
• Why do OpenTelemetry collector fleets become risky as they scale?
• How should teams govern custom OpenTelemetry collector builds in production?