Persistent disks left unattached, 7+ days after a detach, that bill for their full size

Detached disks bill for every provisioned gigabyte

Disk pricing charges Persistent Disk by provisioned space: “If you provision a 200 GB disk, you are billed for that entire disk space, regardless of how you use it, until you relinquish it.” Google’s example puts a 200 GB SSD Persistent Disk at $34.00 a month in US regions.

Attachment does not change that. The disk modification guide says “you incur costs for a disk until you delete it, even if the disk isn’t attached to any instances.” Disks survive VM deletion whenever autoDelete is false, which is how most orphaned disks are born.

Listing disks nothing is using

A disk’s users field lists the instances it is attached to. The -key:* filter form matches disks where that field is empty:

gcloud compute disks list \
  --filter="-users:*" \
  --format="table(name,zone,sizeGb,type,lastDetachTimestamp)"

Conditions for an orphaned disk

ZopNight treats a disk as unattached when its status is READY and no instance uses it. For a disk that was attached and later detached, it then requires at least 7 days since the detach, using the lastDetachTimestamp Compute Engine records, so a disk moved between VMs during maintenance is not flagged. A disk that was never attached has no detach time, so it is flagged as soon as it is seen unattached, whatever its age.

Disks it leaves alone

Kubernetes-managed volumes are skipped: names starting pvc-, gke- or kubernetes-dynamic-pvc-, and disks carrying the goog-gke-volume label from the GKE Persistent Disk CSI driver. Their lifecycle belongs to the cluster, and a released volume is often reclaimed by a StatefulSet. If ZopNight has no price for the disk, it stays silent instead of reporting a zero.

Pricing a disk nobody reads

saving = full monthly cost of the disk
cost after fix = 0

Deleting an unattached disk removes its whole standing charge, so no fraction is applied.

Deleting a disk safely

1. Confirm no instance or instance template expects the disk; check its labels and description for an owner.
2. Take a snapshot if the contents might matter: gcloud compute snapshots create SNAPSHOT_NAME --source-disk=DISK_NAME --source-disk-zone=ZONE.
3. Delete the disk: gcloud compute disks delete DISK_NAME --zone=ZONE.
4. Set autoDelete on data disks of future VMs you do not intend to keep, so they go with the instance.