Cloud Resources & Metrics
What KubeSense discovers and measures in each cloud — 1550 metrics across 116 resource types, from AWS, GCP, Azure, MongoDB Atlas, Confluent Cloud and Kong.
Overview
Alongside your Kubernetes estate, KubeSense discovers cloud resources and collects their metrics through each provider's own monitoring API, using read-only credentials. No agent is installed in your cloud accounts for metric collection.
These pages are the reference for what that covers: every resource type discovered, every metric collected, and the provider-side query detail behind each one — so you can confirm coverage before switching an integration on, look up the provider metric behind a KubeSense chart, or reconcile a KubeSense dashboard against your own cloud console.
| Provider | Metrics | Resource types | Read from |
|---|---|---|---|
| AWS | 591 | 44 | Amazon CloudWatch, plus RDS Performance Insights for database internals |
| GCP | 191 | 27 | Google Cloud Monitoring |
| Azure | 364 | 30 | Azure Monitor |
| MongoDB Atlas | 187 | 3 | Atlas Administration API measurements endpoints |
| Confluent Cloud | 184 | 7 | Confluent Cloud Metrics API |
| Kong | 33 | 5 | Prometheus plugin on Kong Gateway, Konnect analytics for Konnect control planes |
| Total | 1550 | 116 |
Metric tiers
Every resource type is collected at one of two tiers, set per integration when you configure it.
- Core — the default. A focused set covering health, saturation, throughput and errors: enough to run dashboards and alerts without a large increase in metric volume.
- Expanded — a superset of Core, adding deeper per-service detail. Choose it for the services you actively operate; it increases the number of stored series and, on some providers, the number of monitoring API calls billed to your account.
Tiers are set per resource type rather than per integration, so you can run Expanded on the handful of services that matter and leave the rest on Core.
How to read the tables
Each provider page opens with the resource catalogue — every toggle on that integration's Resource Types selector, the entity kinds it produces, and whether metrics are available for it. A type marked Inventory only is discovered and shown in the Infrastructure explorer, but has no metric series.
Then one table per resource type lists the metrics, alongside the provider-side detail needed to find the same series in your own console:
- AWS — the CloudWatch namespace, the dimension the metric is scoped by, and the statistic requested
- GCP — the full Cloud Monitoring metric type and the alignment applied
- Azure — the Azure Monitor aggregation requested
- MongoDB Atlas, Confluent Cloud, Kong — the provider's own metric or measurement name
Metrics your account does not publish: The tables describe what KubeSense will ask each provider for. A metric that depends on an optional feature — Enhanced Monitoring on RDS, Container Insights on ECS and EKS, Atlas Search on MongoDB Atlas, the Prometheus plugin on Kong — returns no data until that feature is enabled, and KubeSense simply records nothing for it. Nothing needs to be turned off in KubeSense to account for services or features you do not use.
Provider references
- AWS — 30 resource types, CloudWatch and RDS Performance Insights
- GCP — 24 resource types, Cloud Monitoring
- Azure — 31 resource types, Azure Monitor
- MongoDB Atlas — clusters, nodes and databases
- Confluent Cloud — clusters, topics, consumer groups, connectors, ksqlDB, Flink
- Kong — gateways, services, routes, upstreams, consumers
To connect an account in the first place, see Resource & Metric Collection for AWS, GCP or Azure.