Find idle and orphaned Azure resources that still cost money

Distinguish an orphan from an idle resource

An orphaned resource has lost the relationship that normally makes it useful, such as a managed disk with no attached VM. An idle resource is still connected but has little observed activity. Neither condition alone proves that the resource can be removed: a disk may be retained for recovery, and a quiet VM may run a monthly job. Build a candidate list, then establish purpose and ownership.

Inventory the full subscription scope

Use Azure Resource Graph or the portal to list resources across the subscriptions you are responsible for. Capture the full resource ID, type, region, resource group and owner tags. A resource group name is a useful clue but is not evidence that every resource in it belongs to the same application. Missing permissions can also make an attached resource appear to have no visible parent.

1. List managed disks and inspect attachment and disk-state information.
2. Review snapshots alongside their source disks and retention requirements.
3. Inspect public IP associations, network interfaces and load-balancer dependencies.
4. Record resources whose purpose or owner is unknown for follow-up.

Confirm which candidates are actually billed

Match each candidate to its recent Cost Management records. A network interface can be worth cleaning up without representing a direct recurring charge of its own. Disks, public IP addresses and other associated services have different pricing rules. Prioritize observed cost over resource count, and do not turn missing billing records into an assumed zero.

Check a representative activity window

For idle workloads, review CPU, memory where collected, disk activity, network traffic and application schedules. Include backup windows, month-end processing and disaster-recovery requirements. A stopped VM can still incur compute charges if it remains allocated. A deallocated VM stops normal VM compute billing, but disks and other retained resources can continue to cost money.

Plan retirement with the workload owner

Confirm retention requirements, backups, dependencies and any resource locks before making a change. Record the resource ID and the reason for the decision. Where the application supports it, use a reversible shutdown or detach-and-observe stage before permanent deletion. A snapshot is another billable resource and should have its own retention plan rather than becoming a permanent replacement for the item being removed.

Measure the result instead of assuming savings

Illustrative example: three unused disks each incurred ₹600 in the previous complete month. Their historical cost totals ₹1,800 per month; that is a planning estimate, not a guaranteed immediate invoice reduction. Deleting them midway through a month leaves earlier charges, and any replacement snapshots add costs. Check the following complete period. Cloudledger’s Orphaned Resources view can help prioritize candidates, while billing drilldowns provide evidence for the review.