Why it matters
Stopped VMs look free, but their disks are not. A project that nobody has started for a month is often a forgotten test or migration environment.How nOps finds it
nOps reads Compute Engine disk charges (persistent disk and Hyperdisk capacity, IOPS and throughput, not snapshots, images or local SSD) for each project over the last 30 complete days. A project is flagged when it was billed for disks on all 30 days, and in the last day, and had no Compute Engine vCPU, memory, GPU or local SSD usage, and no GKE Autopilot usage, in that time. The saving is the disk cost over the last 7 days, scaled to a month, and must be at least $5 a month. The cost of the snapshots you keep is not deducted, and static IP charges are left to their own recommendation. Projects whose spend on the service is at least 90% paid by credits or promotions are left out.Before you start
Confirm with the project owner that the VMs will not be started again, or that snapshots are enough to restore them.How to implement
- In the console, open Compute Engine → Disks for the project and list the disks with their In use by column.
- Confirm with the owner that the stopped VMs and unattached disks are not needed.
- Snapshot each disk you might need later:
gcloud compute snapshots create SNAPSHOT_NAME --source-disk=DISK_NAME --source-disk-zone=ZONE. - Delete the stopped VMs you no longer need, then delete the disks:
gcloud compute disks delete DISK_NAME --zone=ZONE. - Remove them from Terraform so they are not recreated.
Risks and considerations
- Billing has no link between disks and VMs, so check each disk before deleting it.
- Deleting a disk is permanent. Restoring from a snapshot creates a new disk and takes time.
- Snapshot storage is usually much cheaper than the disk, but it is not free.
FAQs
Does nOps flag a single stopped VM in a busy project?
Does nOps flag a single stopped VM in a busy project?
No. It flags a project only when no VM ran in it at all for 30 days, which makes a mistake much less likely.