Containers integration

Kubernetes

Allocate shared cluster cost to namespaces, workloads, and labels across EKS, AKS, GKE, and self-managed clusters — then rightsizing pods and node pools together.

Available on request. We're prioritizing AWS, Azure, GCP, and Oracle Cloud billing first. Kubernetes cost allocation (which requires an in-cluster collector) is available for design partners on request — talk to us to prioritize your cluster.

Connection

Cluster agent / metrics + cloud bill allocation

Access model

Read-only metrics + cloud cost join

Time to value

First allocation views within hours of agent install and cloud linkage

How OptCloud connects with Kubernetes

Kubernetes cost is hard because nodes are shared. OptCloud joins your cloud compute bill (from AWS/Azure/GCP connectors) with cluster metadata — namespaces, labels, requests/limits, and utilization — so shared node cost is allocated fairly. An optional lightweight agent (or OpenCost-compatible metrics path) streams usage without requiring privileged write access to mutate workloads.

Required access

Read access to metrics (CPU/memory usage) and Kubernetes API objects (Pods, Deployments, Nodes, Namespaces)
No requirement to mutate workloads for cost visibility
Linked cloud account connector for the underlying node bill (EKS/AKS/GKE)

Start with observability-only. Automation (rightsizing apply) is a separate, explicit permission set.

Create a connection

1
Connect the cloud that hosts the cluster
Ensure the AWS, Azure, or GCP account paying for the nodes is already connected so OptCloud has the node bill.
2
Register the cluster in OptCloud
Cloud Accounts → Add → Kubernetes. Name the cluster and select the linked cloud account / region.
3
Install the OptCloud collector
Apply the provided Helm chart or manifest. It runs as a read-only collector scraping metrics and listing workloads.
4
Label ownership conventions
Use consistent labels (team, service, env). OptCloud maps them into chargeback dimensions alongside namespaces.
5
Verify namespace allocation
Open Cost Explorer filtered to Kubernetes. Confirm namespaces and workloads show cost, not just cluster totals.
6
Review rightsizing
Check recommendations for over-requested CPU/memory and idle namespaces before enabling any automation.

Data collected

Pod/container CPU and memory usage and requests
Namespace, label, and workload identity metadata
Node pool / instance family cost from the linked cloud bill

What you get in OptCloud

Namespace- and label-level chargeback on shared clusters
Pod rightsizing and idle workload detection
Multi-cluster view across EKS, AKS, and GKE
Joined cloud + container FinOps in one place

Troubleshooting

Cluster shows usage but $0 cost

The cloud account paying for nodes is probably not linked, or node tags/labels don’t join to billable instances. Re-check the cloud connector and cluster registration.