# Kubernetes All Units

**Purpose:** The container inventory. One row per cluster with its namespace, node, pod, and container counts and its total CPU, memory, and storage, plus a unit selector for changing the object level.

![Kubernetes All Units grid showing cluster name, namespaces, nodes, pods, containers and CPU, memory and storage totals in an empty state](https://visualoneintelligence.com/wp-content/uploads/2026/08/voi-docs-k8s-all-units.png)

## When to use it

- Confirm which clusters are under collection
- Compare cluster scale by pod and container density rather than by node count
- Find clusters with high CPU utilization against their allocated total
- Reconcile the container estate against what the platform teams believe they run

## Key areas and metrics

| Column | What it counts |
| --- | --- |
| Cluster Name | The cluster |
| Namespaces | Logical partitions within it |
| Nodes | Machines providing capacity |
| PODs | Scheduled pods |
| Containers | Running containers across those pods |
| CPU Total, Memory Total, Storage Total | Aggregate allocated resource |
| % CPU | Utilization against the CPU total |

A **Select Unit** dropdown at the top changes the object level the grid reports at, and helper text notes that double-clicking opens detailed cluster reports. The grid carries the standard search, saved column settings, view controls, and row grouping.

## Read the ratios, not the counts

Raw counts say little about a cluster on their own, because clusters vary enormously in node size. The useful readings are ratios between the columns.

- **Containers per pod.** Close to one is the common case. Substantially higher indicates sidecar-heavy workloads, which changes how resource requests should be read.
- **Pods per node.** High density is efficient and leaves less room to absorb a node failure.
- **Namespaces against pods.** Many namespaces holding few pods each usually reflects per-team or per-environment partitioning rather than workload scale.
- **% CPU against CPU Total.** Low utilization on a large allocation is the container equivalent of an over-provisioned VM.

## Tips

> An empty grid here means no container platform collector is configured, not that no clusters exist. Every other screen in this section will be empty for the same reason.

## Related screens

[Kubernetes Budgeting Report](https://visualoneintelligence.com/docs/kubernetes-budgeting-report/) attaches cost to these objects. [Trend All Clusters](https://visualoneintelligence.com/docs/kubernetes-trend-all-clusters/) shows how the counts are moving. [All Units: Hosts](https://visualoneintelligence.com/docs/virtual-all-units-hosts/) covers the machines underneath a self-managed cluster.

---

Source: https://visualoneintelligence.com/docs/kubernetes-all-units/