Host hours billing

Several Grafana Cloud products bill on host hours rather than on the volume of telemetry they produce. This page explains the model those products share. What counts as a host differs by product, so read this page for the mechanics and the product’s own page for its definition.

Products billed on host hours:

Application Observability
Kubernetes Monitoring, which bills container hours as well
Database Observability
The knowledge graph, which extends the host hours billing of the product that unlocked it

How host hours are counted

A host hour is one host being active for one hour. Your monthly host hours are the number of active hosts multiplied by the hours each one is active:

host hours = number of active hosts x hours each host is active

A host that runs continuously through a 30 day month accumulates 720 host hours. A host that runs for half the month accumulates roughly 360.

Two properties follow from this:

You’re billed for time, not peak count. Metering accumulates while hosts are active rather than sampling your highest host count and charging for it all month. Short-lived and autoscaled hosts cost only the time they exist.
Turning a host off stops the meter. When a host stops sending the telemetry that identifies it, it stops accruing host hours.

What counts as a host

Each product defines a host in terms of the thing it monitors. The definitions aren’t interchangeable.

Product	A host is	Also billed
Application Observability	A physical or virtual operating system instance	Telemetry
Kubernetes Monitoring	A Kubernetes node	Container hours
Database Observability	A single database instance	Metrics and logs

Some things that look like hosts aren’t billed as one. A container isn’t a host under Application Observability, because it isn’t a full operating system instance. A Pod isn’t a host under Kubernetes Monitoring, because Pods run on nodes and the node is the host. Under Database Observability, multiple databases or schemas on one instance count as a single host.

How activity is determined

A host is active when it’s currently sending the telemetry that identifies it to the product. Each product looks for a different signal:

Application Observability identifies hosts from OpenTelemetry resource attributes on your telemetry. If none of the supported attributes are present, the service isn’t counted for host hours billing. For the attributes and how to set them, refer to Application Observability host hours pricing.
Kubernetes Monitoring counts hosts and containers reported by kube-state-metrics. Without that exporter, nodes and containers can’t be detected.
Database Observability meters each hour that Grafana Alloy holds a connection to the instance. Billing follows connectivity rather than query activity.

Because host identification depends on how you’ve configured collection, a misconfigured collector can both under-report and over-report your bill. Refer to each product’s setup documentation before assuming a host count is wrong.

Telemetry billed alongside host hours

Host hours products still generate metrics, logs, traces, and profiles, and that telemetry is billed at standard Grafana Cloud rates in addition to host hours.

Some plans include telemetry credits that offset part of this. Credits vary by product and by when your plan started, so refer to the product’s page for what applies to you.