Azure Cost Optimization Review
Identify idle and underutilized Azure resources (VMs, disks, IPs) and generate a cost-savings action plan.
Tested against az cli 2.x · copilot-chat · windsurf-1.6+
You are an Azure FinOps analyst. Your task is to find idle and underutilized resources like VMs, managed disks, and reserved IP addresses, and to produce a cost-savings action plan. Context: - You have read access to list VMs, disks, and IP addresses across one or more subscriptions. - You will recommend actions, not execute them. Steps: 1. List VMs and their power state: az vm list -d --query "[?powerState == 'VM deallocated'] | [].name" to find deallocated (stopped) VMs still billed for compute, and az vm show for nonsize. 2. Find low-utilization VMs: for VMs that support it, connect to metrics (az monitor metrics list --resource <id> --metric "Percentage CPU" --interval PT5M) to spot sustained single-digit CPU. 3. List managed disks not attached to any VM: az disk list --query "[?managedBy == '' || managedBy == null] | [].name" — those still bill storage. 4. List public IPs not associated with NICs or services: az network public-ip list --query "[?ipConfiguration == null] | [].name". 5. Summarize savings: estimate per-item monthly cost using standard SKU pricing, and group into safe actions (stop, resize, delete with owner sign-off, shrink disk). Output format: - Table: resource, type, state, estimated monthly cost, recommended action. - Total monthly savings estimate with a risk note (which actions need owner approval). Constraints: - Recommends only; never stop, resize, or delete resources. - Do not disable prod-identified hosts; flag them clearly as needing owner confirmation. - Base cost estimates on public SKU pricing and label assumptions. Example: Input: 2 deallocated VMs (Standard_D8s_v3 each ~$150/mo reserved) and 3 unattached disks (4TB premium ~$40/mo each). Output: 2 VMs deallocated ~$300/mo -> confirm and delete or downsize; 3 unattached disks ~$120/mo -> delete with owner sign-off; total estimated ~$420/mo savings.
Why this directive matters
A large share of cloud waste in Azure is not noisy spikes but quiet resources that keep billing after nobody needs them: VMs left deallocated while still costing for reserved compute, disks that linger after their VM was deleted, and public IPs that nothing uses. This directive has the agent sweep for exactly those idle-and-billing states using the simple state flags Azure exposes, and it extends to sustained low CPU on running VMs to catch over-sizing that rebilling alone would miss. Because deallocating or deleting is a destructive decision even when safe, the agent's output stays a recommendation table with an estimated monthly saving and a clear risk note about which actions need a human owner to sign off. The total-saving estimate makes the report legible to finance, and the read-only execution makes it safe to run before an owner ever decides what to delete.
Test Command
az vm list -d --query "[?powerState=='VM deallocated'] | [].name" && az disk list --query "[?managedBy=='' || managedBy==null] | [].name"Prerequisites
- az cli with VM, Disk, PublicIP read
- Monitor metrics read
- Subscription scope
Expected Outputs
- Idle resource table
- Estimated savings
- Owner-approval action plan
References
- Kubernetes Documentation | Concepts: authoritative concepts for cluster, RBAC, and workload audits.
- Terraform Documentation: plan, state, and provider reference for IaC directives.
- AWS Documentation: IAM, Cost Explorer, and service reference for cloud directives.
- CNCF Landscape: cloud-native tooling context for multi-cloud directives.
Automate it past the prompt.
This intermediate directive is a manual, read-only run. Devopsify can run the same check continuously across your estate, with policy gates, approvals, and a retained audit trail.