Windows Node Support

Kubecost can run on clusters with mixed Linux and Windows nodes. The Kubecost pods must run on a Linux node.

Deployment

When using a Helm install, this can be done simply with:

helm install kubecost \
--repo https://kubecost.github.io/cost-analyzer/ cost-analyzer \
--namespace kubecost --create-namespace \
-f https://raw.githubusercontent.com/kubecost/cost-analyzer-helm-chart/develop/cost-analyzer/values-windows-node-affinity.yaml

Detail

The cluster must have at least one Linux node for the Kubecost pods to run on:

Use a nodeSelector for all Kubecost deployments:

```
spec:
  nodeSelector:
    kubernetes.io/os: linux
  containers:
```

For DaemonSets, set the affinity to only allow scheduling on Windows nodes:

```
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - key: kubernetes.io/os
            operator: In
            values:
            - linux
```

See the list of all deployments and DaemonSets in this values-windows-node-affinity.yaml file:

kubecostMetrics:
  exporter:
    nodeSelector:
      kubernetes.io/os: linux

nodeSelector:
  kubernetes.io/os: linux

networkCosts:
  nodeSelector:
    kubernetes.io/os: linux

prometheus:
  server:
    nodeSelector:
      kubernetes.io/os: linux
  nodeExporter:
    enabled: true
    affinity:
      nodeAffinity:
        requiredDuringSchedulingIgnoredDuringExecution:
          nodeSelectorTerms:
            - matchExpressions:
              - key: kubernetes.io/os
                operator: In
                values:
                - linux
grafana:
  nodeSelector:
    kubernetes.io/os: linux

Metrics

• Collecting data about Windows nodes is supported by Kubecost

• Accurate node and pod data exists by default, since they come from the Kubernetes API.

• Kubecost requires cAdvisor for pod utilization data to determine costs at the container level.

• Currently, for pods on Windows nodes: pods will be billed based on request size.