Cloud Provider Observability

Cloud Provider Observability provides a unified experience for you to manage multiple cloud environments, and streamline your monitoring and troubleshooting.

Overview

When you monitor and troubleshoot your cloud environment, you need different tools for each provider. This can raise costs and interfere with workflow and efficiency, not to mention a longer time to find and resolve issues. Cloud Provider Observability offers:

Straightforward setup
Instantaneous service discovery after setup
Preconfigured views for your data, including dedicated logs for each provider
Machine learning-powered predictions for proactive capacity planning and resource optimization
Seamless scaling

Note
Cloud Provider Observability doesn’t support AWS PrivateLink or Azure Private Link.

Cloud Provider Observability and cloud data sources

You can bring cloud telemetry into Grafana in two ways. The two ways differ in where your data is stored, how you query it, and how much you set up and manage yourself. Determine which one works best for your situation before you configure access, because the two ways have different setup processes.

One way is using Cloud Provider Observability, which is a managed app. After a guided setup, it discovers your resources, collects their metrics and logs, and stores them in your Grafana Cloud Mimir and Loki instances. If you use Cloud Provider Observability, you get the following features:

Preconfigured dashboards and alerts for popular services across AWS, Azure, and Google Cloud
Metrics and logs centralized in Grafana Cloud, so you can query and aggregate them uniformly across accounts, regions, and providers with PromQL and LogQL
Resource discovery after setup, with Azure autodiscovery adding all supported resources automatically and AWS and Google Cloud discovering resources within the scrape jobs you configure
A unified home page that shows resource health, alerts, and error logs across all three providers at a glance
Curated service health and availability views, and machine learning-powered predictions for capacity planning

The second way is using a cloud data source that connects Grafana directly to a provider’s monitoring service and queries it in place. Your metrics and logs stay in the provider, you query them with that provider’s own query format, and collection, dashboards, and alerts are mostly up to you. You connect a data source for each provider:

Because the data stays in the provider, you query it in place instead of storing a copy in Grafana Cloud. A data source connects through a single set of credentials, so centralizing or aggregating data across multiple accounts, regions, or environments depends on what the data source supports. For example, the Amazon CloudWatch data source connects to one AWS account and one region at a time.

The two also use different access and authentication. The Cloud Provider Observability setup has you create credentials that grant Grafana Cloud access to collect your data. For AWS, this includes a cross-account trust policy that lets Grafana Cloud assume a role in your account.

You configure a cloud data source with its own credentials on the data source settings page in Grafana Cloud. It doesn’t use the credentials you create for the Cloud Provider Observability integration. If you follow the Cloud Provider Observability setup steps while actually using a data source, or the reverse, the credentials you create won’t match what the tool expects.

Choose based on what you need:

Use Cloud Provider Observability when you want your cloud data stored and centralized in Grafana Cloud, uniform querying with PromQL and LogQL across accounts and providers, and preconfigured monitoring without building queries. Get started with AWS, Azure, or Google Cloud observability.
Use a cloud data source when you want to keep data in the provider and query it in place, work with provider-specific query formats, or access data the app doesn’t collect.

You can build custom dashboards with either approach, and you can use both in the same Grafana Cloud stack.