Cloud metrics on Grafana Labs

Types of cloud metrics

Fri, 02 Oct 2026 08:47:49 -0500

Types of metrics you can collect

Each cloud provider exposes its metrics through its own monitoring service, and Grafana can collect them. The metric names differ between providers, but the categories are the same. The services below are examples, not a complete list.

Category	AWS	Azure	GCP
Compute	EC2, Lambda	VMs, Functions	GCE, Cloud Functions
Containers	ECS	AKS, ACI	GKE, Cloud Run
Databases	RDS, DynamoDB	Azure SQL, Cosmos DB	Cloud SQL, Spanner
Storage	S3, EBS	Blob, Disks	Cloud Storage
Networking	ELB, CloudFront	Load Balancer, CDN	Cloud Load Balancing

Questions you can answer

With cloud metrics in Grafana, you can answer questions like:

What’s the CPU utilization trend across all my compute instances?
Which services are approaching capacity limits?
How do costs correlate with traffic patterns?
What’s the p99 latency on my load balancers?

AWS metrics scrape jobs

Fri, 02 Oct 2026 08:47:49 -0500

AWS metrics: scrape jobs

Complexity: Simple | Infrastructure: None | Latency: 5 min default

Create an IAM role that Grafana Cloud assumes; you don’t hand over access keys. Choose which regions and CloudWatch namespaces to scrape, such as EC2, RDS, or Lambda. Each listed service includes a default set of metrics. A custom namespace is not in that list: you add the namespace and enter its metrics. Grafana Cloud queries the CloudWatch API on a schedule, pulls the metrics in, and converts them to Prometheus format so you can query them with PromQL.

Considerations

Configured entirely in the Grafana Cloud UI, with no AWS infrastructure to deploy
60+ AWS services supported
Metrics in Prometheus format for PromQL
Default scrape interval of 5 minutes (configurable)
CloudWatch API cost and rate limits at scale
IAM setup for cross-account access
Resources must be tagged; scrape jobs export only tagged resources
A scrape job has a resource limit per service across regions. Split the job by region, or ask Grafana support to raise the limit
Metrics for a deleted resource can keep appearing for about three hours. Remove its tags before you delete it

When to use

Quick start and evaluation
Proof-of-concept setups
Small environments with moderate metric volume

For ongoing production monitoring, use metric streams instead.

Set it up

Open this path in your Grafana Cloud stack for a fully interactive experience, or read through it to understand the process first. You’ll need access to your AWS account with permissions to create IAM roles.

Learning path

Monitor AWS resources

Monitor AWS resources with Grafana Cloud Provider Observability by collecting Amazon CloudWatch metrics through pre-built dashboards.

14 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

AWS metrics: metric streams

Fri, 02 Oct 2026 08:47:49 -0500

AWS metrics: metric streams

Complexity: Moderate | Infrastructure: Stream + Firehose + IAM | Latency: Near real time

You create a CloudWatch Metric Stream that selects which metrics to export. The stream delivers to an Amazon Data Firehose stream, which batches the metrics and forwards them to Grafana Cloud’s ingestion endpoint. You deploy the Metric Stream, the Firehose delivery stream, and the IAM roles in your AWS account, and you can template all of it with Terraform or CloudFormation.

You also create a resource metadata scrape job. It periodically fetches your AWS resource tags and metadata, which Grafana Cloud combines with the streamed metrics to enrich them with context like resource ARNs and tags.

Considerations

Near real-time delivery; the reference configuration buffers for 60 seconds before delivery, and the first metrics can take 2-3 minutes to appear
If the stream is active and metrics still don’t appear, check that the access policy token has metric:write and that the Firehose ingest URL is correct
Scalability
No metric scrape schedule; the resource metadata scrape job still runs on an interval
AWS infrastructure to deploy: the Metric Stream, an Amazon Data Firehose delivery stream, and IAM roles
A resource metadata scrape job to enrich metrics with resource ARNs and tags
Upfront setup, which you can template with Terraform or CloudFormation

Note
Both Terraform and CloudFormation need an access policy token with metric:write, which requires the Admin role to create.

When to use

Production monitoring where near-real-time delivery matters
Environments that scale across multiple AWS accounts
Setups where you want no metric scrape schedule

Documentation

View the AWS metric streams documentation.

Azure metrics: serverless

Fri, 02 Oct 2026 08:47:49 -0500

Azure metrics: serverless

Complexity: Simple | Infrastructure: None | Latency: Minutes

This is the serverless collection method: Grafana Cloud collects Azure metrics through the Azure Monitor API, with no collector or agent to deploy. Collecting with Grafana Alloy is a separate, agent-based method.

Create a service principal in Microsoft Entra ID with read access to Azure Monitor and provide those credentials to Grafana Cloud. Grafana Cloud queries the Azure Monitor API, discovers your resources automatically, and pulls metrics in, converting them to Prometheus format so you can query them with PromQL.

Considerations

Managed by Grafana Cloud, with no Azure infrastructure to deploy
Simple service principal authentication
Automatic resource discovery
Metrics in Prometheus format for PromQL
Collected every two minutes
Occasional Batch API throttling (529 errors) can cause short gaps
One credential can cover several subscriptions. The service principal needs the Monitoring Reader role on each subscription you collect
Client secrets expire. A deleted or disabled service principal, or the wrong tenant ID, shows up as an authentication failure
Newly accessible subscriptions can take up to an hour to appear in discovery

When to use

Getting started with Azure metrics
Minimal infrastructure preference

Set it up

Open this path in your Grafana Cloud stack for a fully interactive experience, or read through it to understand the process first. You’ll need access to your Azure account with permissions to create service principals and assign roles. The Collect Azure metrics with the serverless solution documentation covers the same steps in reference form.

Learning path

Monitor Azure resources with Grafana Cloud

Monitor Azure resources using Grafana Cloud Provider Observability and visualize Azure Monitor metrics through pre-built dashboards without deploying a collector.

20 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

Azure metrics: Grafana Alloy

Fri, 02 Oct 2026 08:47:49 -0500

Azure metrics: Grafana Alloy

Complexity: Flexible | Infrastructure: Linux, Windows, or MacOS VM | Latency: Configurable

The serverless method filters by resource tag and can add static labels, but it can’t rewrite labels the way Alloy’s relabel rules can. When you need that, use Grafana Alloy, Grafana’s open source telemetry collector. Deploy Alloy once, as a Linux VM or a single-replica Kubernetes workload (for example on AKS), not a DaemonSet. It collects metrics from across all your subscriptions. Alloy can authenticate with an Azure managed identity, so there are no service principal credentials to rotate. Its processing pipeline lets you filter and relabel metrics before they’re stored.

Considerations

Multiple subscriptions from one Alloy deployment
Managed identity support, so no keys to rotate
Processing pipeline for filtering and relabeling
Alloy infrastructure to deploy, maintain, and update
Learning curve

When to use

Want managed identity
Need relabel rules before metrics are stored

Documentation

View the Azure Alloy metrics documentation.

GCP metrics: Grafana Alloy

Fri, 02 Oct 2026 08:47:49 -0500

GCP metrics: Grafana Alloy

Note
The documented way to collect GCP metrics is Grafana Alloy.

Complexity: Moderate | Infrastructure: Linux, Windows, or MacOS host | Latency: Configurable

Alloy collects from Cloud Monitoring using its embedded Stackdriver exporter, then sends the metrics to Grafana Cloud. Its processing pipeline lets you filter and relabel metrics before they’re stored. Provide credentials through Google Application Default Credentials, including a service account key or the identity of the VM or pod Alloy runs on.

Considerations

Multiple projects from one Alloy deployment
Application Default Credentials, including the identity of the VM or pod
Processing pipeline for filtering and relabeling
Metrics in Prometheus format for PromQL
Alloy infrastructure to deploy, maintain, and update
Run Alloy as a single instance, not a DaemonSet

Documentation

View the GCP metrics documentation.

Choose your metrics collection method

Fri, 02 Oct 2026 08:47:49 -0500

Collection method by cloud provider

Cloud	Quick start	Production
AWS	Scrape jobs	Metric streams
Azure	Serverless	Serverless / Alloy
GCP	None	Alloy

Which method by scenario

Scenario	Method
AWS production	Metric streams
Azure production	Serverless
Quick evaluation (any cloud)	Scrape jobs (AWS) / Serverless (Azure)
Multi-cloud environment	Grafana Alloy
Want managed or workload identity	Grafana Alloy

Set up your chosen method

AWS metrics

Azure metrics

GCP metrics

GCP Alloy (docs)

Next steps: Set up your chosen method, then continue to the logs section to add log collection. Metrics tell you something’s wrong. Logs tell you why.

Cloud metrics on Grafana Labs

Types of cloud metrics

Types of metrics you can collect

Questions you can answer

AWS metrics scrape jobs

AWS metrics: scrape jobs

Considerations

When to use

Set it up

Monitor AWS resources

Open in Grafana Cloud

AWS metrics: metric streams

AWS metrics: metric streams

Considerations

When to use

Documentation

Azure metrics: serverless

Azure metrics: serverless

Considerations

When to use

Set it up

Monitor Azure resources with Grafana Cloud

Open in Grafana Cloud

Azure metrics: Grafana Alloy

Azure metrics: Grafana Alloy

Considerations

When to use

Documentation

GCP metrics: Grafana Alloy

GCP metrics: Grafana Alloy

Considerations

Documentation

Choose your metrics collection method

Collection method by cloud provider

Which method by scenario

Set up your chosen method

AWS metrics

Azure metrics

GCP metrics

Metrics resources

Resources

Documentation

Blog posts

Community