Skip to main content

Connect Kubernetes

Summary​

Kubernetes cost visibility in Costimizer uses an open-source metrics collector deployed via Helm. It collects pod-level and node-level usage data and sends it to Costimizer using Prometheus remote write.

This works with any Kubernetes cluster, self-managed or cloud-hosted (EKS, AKS, GKE).

Required role: Organization Manager or Manage Cloud Credentials
Estimated time: 20 minutes

Prerequisites​

  • Access to the Costimizer UI with the required permissions
  • A running Kubernetes cluster
  • kubectl access to the target cluster
  • Helm v3+ installed on your machine

Create the Kubernetes data source in Costimizer​

  1. Log in to Costimizer.
  2. Go to System → Accounts → Connect Cloud Account.
  3. Select Kubernetes.
  4. Fill in:
    • Name — a meaningful label (e.g., prod-eks-cluster)
    • User — a username for the collector to use
    • Password — a secure password for authentication
    • Flavor-based cost model (optional): if enabled, Costimizer uses actual cloud provider pricing for node costs. If disabled, a default cost model is applied and can be adjusted later in the Nodes tab.
  5. Click Connect.

Kubernetes connection form in Costimizer

Kubernetes connection — select Kubernetes and fill fields

Verify account​

After connecting, review the Kubernetes account details page to verify your Kubernetes ID, Cloud Account ID, User, and Cost Model. Costs will initially display as $0 until metric generation begins.

Kubernetes account details — verify IDs and cost model

Kubernetes account — cost model settings

Setting up the cost model manually if the selected default model mentioned above:

Kubernetes cost model manual setup

Deploy the Helm collector​

Step 1: Add Helm repo​

Run this command on your local machine (pointed at the target cluster):

helm repo add costimizer https://costimizerinc.github.io/helm-charts

Helm repo add command

Step 2: Install collector​

helm install kube-cost-metrics-collector costimizer/kube-cost-metrics-collector \
  --set prometheus.server.dataSourceId=<CLOUD_ACCOUNT_ID> \
  --set prometheus.server.username=<USERNAME> \
  --set prometheus.server.password=<PASSWORD> \
  --set prometheus.server.remote_write[0].url=https://stg.costimizer.ai/... \
  --set prometheus.server.remote_write[0].name=costimizer \
  --set prometheus.server.remote_write[0].basic_auth.username=<USERNAME> \
  --set prometheus.server.remote_write[0].basic_auth.password=<PASSWORD> \
  --set 'prometheus.server.remote_write[0].headers.Cloud-Account-Id=<CLOUD_ACCOUNT_ID>' \
  --namespace costimizer --create-namespace

Where to find your values​

After creating the Kubernetes data source in Costimizer, click Kubernetes Integration (top-right of the account page). The modal shows the exact Helm install command pre-filled with your CLOUD_ACCOUNT_ID, USERNAME, and PASSWORD.

Validate cost data​

After deployment, metric generation is not immediate. Once the collector is running, navigate to:

  • Resources — to see Kubernetes workloads and node costs
  • Pools — to view workloads organised by team or project
  • Recommendations — to see Kubernetes-specific optimisation opportunities

Validate Kubernetes cost data in Resources and Recommendations

Key takeaways​

  • Kubernetes costs = cluster infrastructure + pod requests + storage + networking
  • Connect the parent cloud account first for complete cost picture
  • Right-sizing pod requests often yields 15–25% savings