Skip to content

Deletion and retention

Deleting one resource is final once its operation succeeds. Deleting a whole project, or closing the organization, has a recovery window in which you can change your mind.

Deleting a resource

Each delete dialog says what goes with the resource and what stays.

You delete What goes with it What stays
An instance Its system disk, its addresses, and volumes the location is set to delete with it (the dialog names them) Other attached volumes, public IPs (still allocated), security groups
A volume Its data Nothing; delete its snapshots first
A snapshot The copy Volumes already made from it
A load balancer Its listeners, pools and health checks The instances behind it; its public IP, detached but still allocated; its certificates
A network Its range and its routing attachment Other networks, private routing
An instance group Nothing, by default Its instances, unless you choose to delete them too
A schedule Nothing The snapshots it made, no longer pruned
A launch template Its versions and its user data Instances launched from it
A certificate The certificate and its key in the key manager Nothing

What blocks a delete

Some deletes are refused until something else is gone. The refusal names what is in the way.

Resource Refused while Code
Instance It is locked, or a change to it is running —
Volume It is attached, or has snapshots volume_in_use, volume_has_snapshots
Snapshot A volume is being made from it snapshot_restore_pending
Image It is protected in the location image_protected
Network Instances, load balancers or others have addresses on it resource_in_use
Network (detach routing) A public IP or a route uses it public_ips_in_use, route_uses_network
Security group An instance or load balancer uses it; or it is default security_group_in_use, default_security_group
Server group It has instances server_group_in_use
Internet gateway A public IP in that location is attached public_ips_in_use
Load balancer pool A listener uses it, or it has members or a health check lb_child_in_use
Certificate A listener serves it, as default or by host name certificate_in_use
Launch template A launch is creating its instances, or an instance group uses it launch_template_in_use
Custom role Someone holds it role_in_use

Nothing blocks deleting an SSH key, a public IP, a schedule, an instance group or a whole load balancer.

Delete a project

Project admins, owners and admins delete a project from its Settings → Danger zone, or from the organization's Settings → Projects. Choose Delete only if empty, or Delete everything in it: instances, load balancers, public IPs, routers, networks and security groups go within minutes, while volumes, snapshots and images are kept for the recovery window. Step by step: Projects.

When What happens
At once The project is suspended: nothing new can be made in it
During the recovery window (7 days unless agreed otherwise) Cancel deletion, then Resume project, brings it back. After Delete everything, only volumes, snapshots and images come back
Delete now, or the window ends Everything left is deleted; this cannot be cancelled once started
Afterwards Activity and usage records are kept
What happens when a project is deleted or the organization closed Two timelines. Deleting a project, which project admins, owners and admins can do: at once the project is suspended and nothing new can be made in it. During the recovery window, 7 days unless agreed otherwise, Cancel deletion and then Resume project bring it back. When the window ends, or at once with Delete now, everything left is deleted, and this cannot be cancelled once started. Afterwards, activity and usage records are kept. With Delete everything in it, instances, load balancers, public IPs, routers, networks and security groups go within minutes, while volumes, snapshots and images are kept for the window and are the only things that come back after Cancel deletion. With Delete only if empty, the deletion is refused while anything is left. Closing the organization, owners only: at once every instance stops, pending invitations are revoked and nothing can be changed, though owners can still read everything. During the recovery window, 30 days unless agreed otherwise, everything is kept, Cancel closure brings it back and an owner can request a data export of its records. When the window ends, every project is deleted and every member removed; this cannot be undone. Afterwards the audit log and usage records are kept. At once Recovery window Window ends or Delete now Afterwards Project Window: 7 days Project admins, owners and admins Suspended Nothing new can be made in it Cancel deletion then Resume project brings it back Deleted for good; this cannot be cancelled once started Kept activity and usage records With Delete everything in it Within minutes instances, load balancers, public IPs, routers, networks, security groups Kept for the window volumes, snapshots and images: only these come back after Cancel deletion With Delete only if empty Refused while anything is left Organization Window: 30 days Owners only Read only Every instance stops; pending invitations are revoked. Owners can still read everything Cancel closure brings it back: everything is kept until the window ends. An owner can request a data export of its records Closed Every project deleted, every member removed. This cannot be undone Kept the audit log and usage records
Deleting a project and closing the organization: what happens at once, during the recovery window, when it ends and afterwards. Each window is as shown unless agreed otherwise.

The project's schedules pause when its deletion is requested; its launch templates, schedules, instance groups and certificates go with it. Stacks and clusters made with a location's own tools must be deleted with those tools; the page names anything that stops the deletion, and the deletion carries on once it is gone.

Clean up a project safely

Before you delete a project, or to stop paying for what you no longer use:

  1. Copy any data you need off its volumes, and save images you want to keep elsewhere.
  2. Delete instance groups (with their instances) and schedules, so nothing creates new resources.
  3. Delete load balancers, then instances.
  4. Release public IPs: they stay allocated, and counted, after their instance is gone.
  5. Delete snapshots, then volumes, then images you no longer need.
  6. Delete the Internet gateway if nothing needs the Internet, then networks you made.
  7. Check Usage the next day: anything still counting is still there.

Close the organization

Owners only, from Settings → Danger zone → Close organization. Every instance stops at once; everything is kept for the recovery window (30 days unless agreed otherwise), when Cancel closure brings it back. When the window ends, every project is deleted and every member removed. While it is closing, an owner can download a copy of its records. See Settings.

People leaving

  • Removing someone from the organization ends their access at once, deletes their personal SSH keys and service accounts, and signs them out of every location within 15 minutes. What they made stays.
  • Deleting a service account stops its API keys at once, ends its project access and deletes its cloud credentials in every location.