Cloud Pak for Business Automation is licensed by Virtual Processor Cores (VPC) of the pods that carry a license annotation; the IBM License Service (part of the foundational services) measures the CPU limits / requests of those pods per hour and keeps the peak per month. Workflow counts the workflow server / authoring pods (and their PFS, Workplace parts as documented per version); databases outside the cluster, OpenShift itself and the operator do not count. Ratios between capabilities (e.g. Workflow vs Content) are defined in the license terms - the License Service already applies them and reports "CP4BA VPCs".
# current and historical usage (cluster-admin or license-service role)
oc get route -n ibm-licensing | grep ibm-licensing-service-instance
TOKEN=$(oc get secret ibm-licensing-token -n ibm-licensing -o jsonpath='{.data.token}' | base64 -d)
curl -sk "https://<licensing-route>/products?token=$TOKEN" # products and peak VPC per month
curl -sk "https://<licensing-route>/snapshot?token=$TOKEN" -o snapshot.zip # audit snapshot for IBM (keep 2 years)
# per pod: the annotation that makes a pod licensed
oc get pod icp4adeploy-bawins1-baw-server-0 -o jsonpath='{.metadata.annotations}' | tr ',' '\n' | grep productMetricPractical points: pod limits drive the count on most releases (set them consciously, question on sizing applies); non-production environments count unless licensed as such (production / non-production is a term of the entitlement, not a technical flag); the same entitlement covers ODM, ADS, Content, RPA and BAI, so moving a workload between capabilities does not change the pool; Workflow Process Service and the SaaS tenant are user-based instead. Keep the monthly License Service snapshots - they are the audit evidence.
References