Volumes
Create, attach, expand, detach, and remove persistent block devices.
Volumes provide persistent block devices for droplet root disks and application data. Each volume belongs to one project and region and uses a storage policy.
Create a volume
Open Storage → Volumes → Create volume, choose the region, policy, and size, and review quota and pool capacity. Sizes use integer GiB. Zero, negative, fractional, and unsupported sizes are rejected rather than rounded.
A new empty volume contains no partition table or filesystem. After attachment, format and mount it using guest operating system tools. A volume restored from a snapshot contains the captured disk data and should not be formatted as empty.
Attach a volume
Select Attach and choose a droplet in the same project and region. The hypervisor and storage system must support the requested attachment. Infra assigns an attachment ID and exposes a stable device identifier to the guest.
Attachment identity and writer ownership describes device serials, stale attachment references, and migration handover. A writable volume supports one active droplet attachment. A second writable attachment is rejected. Sharing application data between guests requires a service designed for concurrent access; a block device alone does not provide a shared filesystem.
Use stable device identifiers inside the guest. Kernel names such as /dev/vdb
can change when attachment order changes. Confirm the device size and existing
filesystem before writing to it.
Expand a volume
Open Resize, enter a larger integer GiB value, and review quota and pool capacity. Expansion changes the block device after the operation succeeds. It cannot shrink the device or undo a completed expansion.
The guest might require a device rescan, partition expansion, and filesystem expansion. The correct steps depend on its operating system and filesystem. Infra does not automatically run arbitrary filesystem tools inside the guest.
If storage expansion fails before the new size is committed, the existing size remains. If the device has expanded but guest filesystem growth fails, retain the larger volume and repair the guest filesystem separately.
Detach a volume
Stop application writes, unmount the filesystem, and select Detach. A root volume requires a stopped droplet. A data volume can be detached from a running guest only when its device supports hot removal and guest usage has stopped.
A forced detach from an unreachable host requires proof that its previous writer cannot continue. Storage leases prevent simultaneous writers, but an operator must not bypass fencing to resolve a network partition. Unclean removal can require filesystem recovery.
Change a storage policy
Select a compatible policy in the same region. Check its encryption domain and inherited I/O limits before applying the transition. Infra previews replica movement, additional capacity, and feature differences. The existing policy remains active until the transition reaches its required state. Unsupported transitions are rejected with the affected volume ID.
Delete and retain
An attached volume cannot be deleted. Detach it and check deletion protection, snapshots, and backup requirements first. Deleting a volume does not expire independent backup copies. Snapshot dependencies can retain underlying blocks until their own retention ends.
See Snapshots for short-term recovery points and Backup policies for scheduled copies and retention.
Guest file deletion and physical pool usage are not equivalent. Use discard and space reclamation to distinguish filesystem free space, logical quota, and blocks retained by recovery points.
Online expansion
Supported storage policies allow block-device expansion while an eligible guest continues running. The guest filesystem must independently support online growth for the full application procedure to remain online. Review discard and allocation when logical and physical usage differ.
Object buckets serve object data through a separate S3 path. Cross-region replication maintains explicit recovery targets without changing this volume's regional attachment boundary.
Storage prices use provisioned logical size for volumes and separate retained-block accounting for snapshots. Replicated volume rates include the published replica count.