Grafana Cloud

Host-hours pricing

Database Observability billing has three components: a per-host hourly fee, metrics usage, and logs usage. This page explains how each component is calculated.

Billing components

ComponentRateWhat it measures
Host hoursSee Grafana Cloud pricingEach connected database instance
MetricsStandard Grafana Cloud Metrics ratePrometheus metrics from database exporters
LogsStandard Grafana Cloud Logs rateQuery samples, schema details, and other log-based telemetry

Host hours

What counts as a host

A host is a single database instance. The following each count as one host:

  • One self-managed MySQL or PostgreSQL server
  • One AWS RDS instance
  • Each Aurora instance within a cluster (writer and reader instances are billed separately)
  • One Google CloudSQL instance
  • One Azure Database instance

Multiple databases or schemas on the same instance count as one host.

When you are billed

You are billed for each hour that Grafana Alloy maintains a connection to your database instance. Billing is based on connectivity, not query activity.

ScenarioBilled?Reason
Database running, Alloy connected, queries executingYesAlloy is connected
Database running, Alloy connected, no query activityYesAlloy is still connected
Database stopped (for example, RDS instance paused)NoAlloy cannot connect, so no telemetry is sent
Aurora read replica active for 3 hours3 hours onlyBilled for actual connection time
Alloy shut down or disconnectedNoNo connection means no telemetry

Note

Billing uses hourly metering, not monthly high-water mark. You pay only for the hours each instance is connected. Ephemeral replicas and instances that scale in and out are billed only for the time they exist.

Note

Alloy v1.14 or later is required for Database Observability. This version includes finer-grained controls for tuning telemetry collection. To update, refer to the Alloy release notes.

Controlling host hour costs

To stop billing for an instance:

  • Stop monitoring: Remove the instance’s configuration from your Alloy config and restart Alloy.
  • Pause the database: If you stop an RDS instance or shut down a self-managed server, Alloy loses connectivity and billing stops automatically.
  • Deactivate Database Observability: To stop all Database Observability billing, deactivate Database Observability from the Manage App tab in the Database Observability configuration page in your Grafana Cloud console.

Metrics

Database Observability collects Prometheus metrics through prometheus.exporter.mysql and prometheus.exporter.postgres components. Metrics are billed at the standard Grafana Cloud Metrics rate.

An active series is a unique combination of metric name and label values that received data in the last 30 minutes.

Factors that increase metrics volume

  • More unique query patterns (each normalized query creates distinct series)
  • More databases or schemas per instance
  • Heavy use of table partitioning (each partition is collected as a separate table)

Logs

Database Observability sends query samples, schema details, explain plans, and other diagnostic data as logs to Grafana Cloud Loki. Logs are billed at the standard Grafana Cloud Logs rate.

Factors that increase log volume

  • Shorter collection intervals (configured via collect_interval in the collector block)
  • More unique query patterns
  • Larger schemas with more tables
  • Heavy table partitioning

Controlling metrics and logs costs

You can reduce metrics and logs costs by tuning how Alloy collects data:

  • Increase collection intervals: Set a longer collect_interval on query_samples, query_details, or explain_plans to collect less frequently. For example, increasing query_samples from the default 10s to 60s reduces log volume significantly (although it decreases visibility on slow queries).
  • Disable collectors you don’t need: Turn off explain_plans or schema_details entirely if you don’t use the Explain Plans or Schema tabs. This eliminates the logs those collectors generate.
  • Exclude system schemas: Filter out schemas like rdsadmin, cloudsqladmin, and azure_sys to avoid collecting data from system schemas.
  • Exclude service accounts and monitoring users: Prevent queries from replication users, cloud admin users (for example, rdsadmin) or the Alloy monitoring user from generating telemetry.

For environment-specific configuration examples, including high-volume and low-resource presets, refer to Tune Alloy collection.

Warning

Heavy use of table partitioning significantly increases telemetry volume. If your tables use extensive partitioning (for example, 100 tables with 100 partitions each), contact Grafana Cloud support for custom sizing guidance.

Estimating costs

Your total monthly cost is the sum of host hours, metrics, and logs. The host hour fee is predictable based on the number of instances you monitor. Metrics and logs costs vary based on your workload characteristics.

For help estimating your total cost, refer to Grafana Cloud pricing or contact Grafana Cloud support.

Retention

Telemetry typeFree tierPro tier
Metrics14 days13 months
Logs (including query samples)14 days30 days

Extended retention is available on the Enterprise tier.

Grace period

Some customers may be eligible for a grace period on the per-host fee. Contact your Grafana account team for details.

How Database Observability charges appear on your bill

If you have a Grafana Cloud flexible spend commitment, Database Observability charges burn down from your existing commitment. No separate commitment or contract amendment is required.

If you are on a pay-as-you-go plan, Database Observability charges appear as usage on your Grafana Cloud invoice.