Kubernetes Cost Allocation with OpenCost / Kubecost Prompt

You are a FinOps-minded platform engineer who has made multiple teams' Kubernetes spend legible and then cut it 30% without a single incident.

I will share:
- Cloud provider and how nodes are purchased (on-demand, spot, reserved, savings plans)
- Current cluster size and rough monthly bill
- Labeling/namespace conventions (or the lack of them)
- Who needs the numbers (eng leads, finance) and the cadence

Deliver an end-to-end cost-allocation plan:

1. **Allocation model first** — define how a node's cost splits into pods: by resource *requests* vs *usage*, plus shared/idle and system overhead. Explain why requests-based allocation drives better behavior (teams feel the cost of over-requesting) and where usage-based is fairer.

2. **Install & wire data** — OpenCost (or Kubecost) with accurate pricing: pull real cloud billing via the cloud cost integration rather than list prices, account for spot discounts, and reconcile against the actual invoice. Note the gap between estimated and billed cost and how to close it.

3. **Dimensions** — allocate by namespace, `team`/`cost-center`/`app` labels, controller, and pod. Provide the label taxonomy and an admission policy (Kyverno/Gatekeeper) that *requires* a cost-center label so nothing lands unattributed.

4. **Idle & waste** — surface idle cost (provisioned-but-unused), over-requested workloads (request vs P95 usage), orphaned PVCs and LoadBalancers, and abandoned namespaces. Rank by dollars, not percentages.

5. **Showback → chargeback** — a monthly report per team with trend, top movers, and one recommended action each. Decide showback vs chargeback and the org maturity needed for chargeback.

6. **Close the loop** — connect findings to VPA/right-sizing, spot adoption, and bin-packing, with an owner and a target for each.

Output: (a) the allocation methodology written so a finance partner understands it, (b) install/config notes incl. cloud-billing integration, (c) the required-label policy, (d) a sample monthly team report layout, (e) a prioritized waste-reduction backlog. Be blunt about precision limits.