The purposes of this repository is to refine the the cluster-cpu-capacity Report and only pull out the worker nodes for purposes of subscription managment.
Guidance from writing-custom-queries in the metering-operator project was used to create this report.
The queries that are used for that report have been duplicated and modified to only pull the worker node data from Prometheus.
The idea is that this would run in a cluster and the endpoint can be queried monthly, quarterly, etc... and used for billing purposes.
The cluster-cpu-capacity report averages the number of CPUs between the last scheduled time and current scheduled time.
These CRD's assume the Metering operator is installed on an OpenShift cluster. For help with that, see the operator-setup folder.
The sample Report is R-test-cluster-worker-cpu-capacity.yml - this report is currently configured to run hourly, during minute 36. It can be tailored to run on whatever schedule makes sense.
In general:
- Report CRDs are prefixed with R.
- ReportDataSource CRDs are prefixed with RDS.
- ReportQuery CRDs are prefixed with RQ.
# git clone https://github.com/jaywryan/worker-node-metering
# cd worker-node-metering
# kubectl create -n $METERING_NAMESPACE -f ./reportsThe reports are exposed via the API endpoint of the metering operator.
METERING_ROUTE_HOSTNAME="$(oc -n $METERING_NAMESPACE get routes metering -o jsonpath='{.spec.host}')"
TOKEN="$(oc -n $METERING_NAMESPACE serviceaccounts get-token reporting-operator)"
REPORTNAME=test-cluster-worker-cpu-capacity
curl -H "Authorization: Bearer $TOKEN" -k "https://$METERING_ROUTE_HOSTNAME/api/v1/reports/get?name=$REPORTNAME&namespace=$METERING_NAMESPACE&format=csv"
period_start - The begining timestamp for the particular reporting period
period_end - The ending timestamp for the reporting period
total_cluster_capacity_cpu_core_seconds - The total number vCPU seconds (# of cores per second of the reporting period) for the cluster
avg_cluster_capacity_cpu_cores - The average of the raw vCPU data (collected every minute) during the reporting period
avg_node_count - The average number of worker nodes available during the reporting period.