Slide 2 of 8

How data source queries work

Data source architecture

Data source architecture: Grafana queries cloud provider APIs in real-time

Important: This section covers Grafana data source plugins, which are a core Grafana capability available in all Grafana editions (OSS, Enterprise, and Cloud). Data sources are not part of the Cloud Provider Observability product. They’re included in this journey because they offer an alternative path to cloud visibility that may suit certain use cases.

How it works

StepDescription
1Configure cloud provider credentials in Grafana.
2Data stays in CloudWatch / Azure Monitor / Cloud Monitoring.
3Grafana queries cloud API in real-time.
4No data duplication or additional storage.

When to choose this approach

Choose a data source when…Choose Cloud Provider Observability when…
You want to avoid extra ingestion and storage costYou want prebuilt dashboards and alerts
The data you need is within your cloud’s retention windowYou need to retain data longer than your cloud keeps it
You’re comfortable using each cloud’s own query languageYou want one query language across clouds (PromQL/LogQL)
You want to keep setup simpleYou want to avoid per-query API cost and rate limits

Script

Everything we covered in the metrics and logs sections, collecting metrics and logs and storing them in Grafana Cloud, is the Cloud Provider Observability product.

What we’re covering now is different. Data source plugins are a core Grafana capability that exists in every edition: open source Grafana, Grafana Enterprise, and Grafana Cloud. They’re not part of Cloud Provider Observability, but we cover them here because they offer a legitimate alternative path.

So how does this work? You configure credentials for your cloud provider. When you open a dashboard, Grafana calls the CloudWatch API, the Azure Monitor API, or the Cloud Monitoring API in real-time.

The data never leaves your cloud provider. Grafana just fetches what you need for the current view.

The appeal is simplicity. No collection infrastructure, no storage costs, no data duplication.

A few things to know: query performance depends on the cloud API, you use the cloud’s own query language such as CloudWatch’s syntax or Azure’s KQL, and alerting makes an API call for every evaluation, which can add up.

This approach fits when you want to avoid extra ingestion and storage cost, when the data you need is within your cloud provider’s retention window, or when you’d rather query in place than centralize. You still build dashboards, set alerts, and correlate cloud data with your other Grafana data.

/
Previous Next