Secondary clusters use a minimal Kubecost deployment to send their metrics to a central storage-bucket (aka durable storage) that is accessed by the primary cluster to provide a single-pane-of-glass view into all aggregated cluster costs globally. This aggregated cluster view is exclusive to Kubecost Enterprise.
Disable unnecessary containers/pods on secondary clusters. Note, that setting agentOnly
will disable the Kubecost UI on secondary clusters.
federatedETL:
federatedCluster: true
agentOnly: true
Grafana is not needed on secondary clusters.
global:
grafana:
enabled: false
proxy: false
Kubecost and its accompanying Prometheus collect a reduced set of metrics that allow for lower resource/storage usage than a standard Prometheus deployment.
prometheus:
server:
retention: 2d
Node-exporter is disabled by default. You should keep it this way if cluster/node right-sizing recommendations are not required.
{% hint style="info" %} Node-exporter must remain disabled if there is an existing DaemonSet. More info here. {% endhint %}
You can find complete installation guides and sample files on our repo.
Additional considerations for properly tuning resource consumption is here.