## TL;DR
Kubecost needs at least one full Prometheus scrape cycle plus its own aggregation pass before allocations appear, so a fresh install showing $0.00 everywhere is usually just impatience. Wait 10 to 15 minutes, confirm the Prometheus targets are up, and check that the cluster has the node pricing it needs.

## Use this when
- The Allocations page shows $0.00 for every namespace and workload.
- Kubecost was installed less than an hour ago.
- The bundled Prometheus shows targets up but allocations are empty.

## Not for this skill when
- Allocations show values that look wrong. That is a pricing or label problem.
- Only some workloads are zero. Check their labels and the aggregation window.

## Steps
1. Wait 10 to 15 minutes after install, then hard-refresh the Allocations page. Verify: the window selector covers a period after the install time.
2. Check the bundled Prometheus targets: port-forward to Prometheus and open the targets page. Verify: kubernetes-nodes, kube-state-metrics, and cadvisor targets are UP.
3. Confirm cost-model can reach Prometheus: check the cost-analyzer logs for scrape or query errors. Verify: no connection errors in the last 15 minutes.
4. Check node pricing is set: in Settings, confirm the provider integration or custom pricing has CPU and RAM rates. Verify: the Assets page shows nonzero node costs.
5. If still zero after 30 minutes, restart the cost-analyzer pod to force a fresh aggregation pass. Verify: allocations populate within the next cycle.

### External Prometheus variant
If Kubecost points at an existing Prometheus, confirm the required recording rules and metric retention are present. Verify: the Prometheus datasource test in Kubecost settings succeeds.

Compatibility: Kubecost (OpenCost-based) on any Kubernetes 1.24 and later.

## Why it happens
Allocations are derived from time-series metrics, not point-in-time state. Until Prometheus has scraped a full cycle and Kubecost has run its ETL over that window, there is literally no data to allocate, and the UI renders that as $0.00.

## Edge cases / pitfalls
- An aggregation window ending before the install time always shows zero. Extend the window forward.
- Short Prometheus retention starves longer allocation windows; align the two.
- Thanos or remote-write setups can delay metric availability by several minutes.
- Idle clusters with no running workloads show real zeros. Check that pods are actually scheduled.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_CBLXslGyCF-XyZ3oh_NOnQ