Health monitoring on Grafana Labs

How Fleet Management works

Tue, 18 Aug 2026 05:43:38 -0700

How it works

Collectors register with Fleet Management, receive remote configurations, and send health telemetry back to Grafana Cloud.

The flow

Register: The collector registers itself with Fleet Management, declaring its identity, its attributes, and that it’s ready to receive remote configuration.
Discover: Each collector appears in the Fleet Management inventory, where you can see its health status, version, operating system, and the attributes you’ve assigned, such as env, region, or team.
Configure: You create configuration pipelines and assign them to collectors with attribute matching, for example, every collector where env=production.
Distribute: Collectors poll the Fleet Management API for matching configurations.
Monitor: Collector health telemetry flows back to Fleet Management dashboards.

Key concepts

Concept	What it means
Collector	A telemetry collector registered with Fleet Management
Configuration pipeline	A piece of config that can be assigned remotely
Attribute	Key-value pair used to match pipelines to collectors
remotecfg block	Local config that registers a collector with Fleet Management

Local + remote configs

Grafana Alloy runs local and remote configurations in parallel, isolated from each other. This means:

Remote config errors don’t break local config.
You can mix and match based on your needs.
Gradual migration from local to remote is safe.

The OpenTelemetry Collector works differently. Its Supervisor merges local and remote configuration and doesn’t automatically roll back to the last working remote config, so validate remote changes carefully when using it.

Access tokens and policies

Tue, 18 Aug 2026 05:43:38 -0700

Access tokens

Access tokens allow collectors to authenticate with Fleet Management. You’ll create one and add it to your collector configuration.

How it works

An access policy defines what a token can do: what it can read, what it can write, and which services it can access. An access token is a credential generated from that policy.

Think of it like a keycard. The policy says “this card opens the Fleet Management door,” and the token is the actual card the collector carries.

To connect a collector, you create an access policy with the right scopes, generate a token from it, and add that token to the local configuration of the collector or OpAMP Supervisor. When the collector starts, it uses the token to authenticate with Fleet Management and register itself in your inventory.

Required scopes

Scope	Purpose
`set:alloy-data-write`	Lets Alloy collectors register, report health, and receive remote configurations
`set:otel-data-write`	Lets OpenTelemetry collectors register, report health, and receive remote configurations

Where the token goes

remotecfg {
    url = "https://fleet-management-<CLUSTER>.grafana.net/"
    basic_auth {
        username = "<INSTANCE_ID>"
        password = "<YOUR_ACCESS_TOKEN>"  // ← token goes here
    }
}

In the path, you’ll create an access token and use it to register your Alloy collector.

Attributes

Tue, 18 Aug 2026 05:43:38 -0700

What are attributes?

An attribute is a key-value pair that describes a collector, like a label or tag. Fleet Management uses attributes to match configuration pipelines to collectors.

Why use attributes

When you have 50 collectors, you don’t want to manage them one by one. You want to say “apply this configuration to all production collectors” or “update all collectors in the us-east region.” Attributes make that possible. You match configuration pipelines to collectors based on their attributes.

The better you categorize your fleet with attributes, the more powerful remote configuration becomes:

Configure at scale by applying one pipeline to every collector that shares an attribute.
Organize your fleet by attributes like env, region, or team.

Three types of attributes

Type	Description	Set by
Reserved	Preset attributes such as `collector.os`, `collector.version`, `collector.ID`	Fleet Management (auto)
Local	Set in the collector or Supervisor local config file	You
Remote	Set in the Fleet Management user interface in Grafana Cloud	You

OpenTelemetry collectors report additional reserved keys, such as service.name, os.type, host.*, and cloud.*.

Health status indicators

Tue, 18 Aug 2026 05:43:38 -0700

What Fleet Management checks

Fleet Management continuously monitors your collectors and checks three things. Based on these checks, each collector receives a health status.

Check	Question it answers
Check-in	Has the collector contacted Fleet Management in the last 30 minutes?
Alerts	Are any user-created alerts firing for this collector?
Config errors	Did the collector load its remote configuration without errors?

Fleet Management doesn’t create alerts. It surfaces active alerts you’ve created in your stack.

Health status indicators

Status	Color	Meaning	Action
Healthy	Green	All checks passing, collector is operational.	None needed
Warning	Yellow	Non-critical alert, or the collector hasn’t checked in.	Investigate soon
Error	Red	Critical alert, or a remote configuration error.	Investigate immediately
Unknown	Gray	Collector hasn’t reported any telemetry yet (no heartbeat).	Confirm it started

A quick glance tells you what needs attention. Green is fine, yellow needs a look soon, and red needs investigating now. Gray means a collector hasn’t reported in yet, so check that it started and can reach Grafana Cloud.

Self-monitoring

Tue, 18 Aug 2026 05:43:38 -0700

Self-monitoring pipelines

Self-monitoring pipelines are configuration that Fleet Management automatically assigns to each registered collector, instructing it to collect and report its own health data, such as CPU usage, memory consumption, component health, and internal logs.

Why self-monitoring matters

Without self-monitoring, you’d have a blind spot with no visibility into whether your collectors themselves are healthy.

Scenario	Without self-monitoring	With self-monitoring
Collector runs out of memory	You don’t know until metrics stop arriving.	You see memory usage climbing.
A component fails	Silent failure, missing data.	Component dashboard shows it’s unhealthy.
Collector is overloaded	Data arrives late or not at all.	Throughput metrics show the issue.

You don’t have to configure this. It’s automatic. Self-monitoring is observability for your observability.

Monitor collector health

Tue, 18 Aug 2026 05:43:38 -0700

Monitor collector health

Learning path

Monitor collector health

Welcome to the Grafana learning path that teaches you how to use Grafana Fleet Management to monitor and troubleshoot issues with your collectors.

11 min

Beginner

Docs & blog posts

Open in Grafana Cloud

Complete this learning path directly in your Grafana Cloud stack with an interactive learning experience.

Open in Grafana Cloud

View on website