Kubernetes

Kubernetes Inventory gives you cost and efficiency analysis for your Kubernetes clusters — broken down by namespace, workload, node, pod, and persistent storage — with the option to automate workload right-sizing using the AKO optimizer.

When to use it

• You want to understand what Kubernetes workloads are costing and why.
• You need to identify over-provisioned namespaces or underutilized nodes.
• You want to right-size workload resource requests using automated policy.

Open Kubernetes

Go to Cloud Infrastructure > Kubernetes from product navigation.

Review cluster efficiency

1. Filter by provider, account, and cluster.
2. Select a date range.
3. Review cluster-level cost and resource summaries.
4. Compare Utilized, Requested, and Provisioned CPU and memory:
   • A large gap between Provisioned and Utilized indicates over-provisioning.
   • A gap between Requested and Utilized indicates workloads requesting more than they use.
5. Check node cost by pricing model: On-Demand, Spot, and Reserved.
6. Drill into namespaces, workloads, nodes, pods, and persistent volumes.
7. Click the utilization or cost trend to open a time-based investigation modal.

Inventory views

View	What it shows
Clusters	All connected clusters with cost and efficiency summary
Nodes	Node-level cost, instance type, and utilization
Pods	Pod-level resource usage and cost allocation
Persistent Volumes	Storage volume cost and utilization

AKO — Kubernetes Optimizer

AKO (the Kubernetes Optimizer) is available if enabled for your tenant. It provides policy-driven workload right-sizing with human approval and automation controls.

Use AKO

1. Open the AKO view from the Kubernetes area.
2. Filter by cluster, environment, namespace, provider, or review status.
3. Use the workload state views:
   • All workloads — complete inventory
   • Managed by policy — workloads under an active AKO policy
   • Awaiting review — changes ready for human approval
   • Blocked — workloads that cannot be optimized due to policy conflicts or missing data
   • Low confidence — insufficient history to make a reliable recommendation
4. Select a workload and review the proposed CPU and memory request change.
5. Approve the change when ready.

AKO policy settings

For each policy you can configure: target namespace, environment, confidence threshold, automation mode (recommend-only or auto-apply with approval), and review cadence.

Best practices

• Investigate clusters where Provisioned is significantly higher than Utilized — this is typically the biggest Kubernetes cost opportunity.
• Run AKO in recommend-only mode for at least two weeks before enabling auto-apply.
• Separate production and non-production policies — non-prod can tolerate more aggressive right-sizing.
• Review blocked and low-confidence workloads periodically as more data accumulates.

Troubleshooting

AKO requires connected clusters. If the AKO workload table is empty, verify that the Kubernetes integration is connected and clusters are being discovered. See Missing or delayed data.