What a Raff volume is, how it persists, how it attaches, and what currently varies between volumes
Updated May 8, 2026A Raff volume is a single block device that you attach to a VM. Every volume on the platform today shares the same characteristics — Raff offers one volume class at the moment, so the choice you actually make at create time is size and billing term, not a performance tier.This page describes the model so you know what behaviour to expect, where volumes sit relative to your other storage options, and which properties are fixed vs which the platform may expand later.
Single tier — NVMe SSD block storage (HDD tier coming)
Block driver
virtio-blk (devices appear as /dev/vda, /dev/vdb, …)
Size range
10 - 1000 GiB
Resize
One-way grow only — no shrinking
Attach model
Single-attach — one VM at a time
Region scope
Volumes are bound to the region of the VM they attach to
Project scope
Volumes inherit the project of their attached VM
Lifecycle
Independent of any VM — survive VM delete
Filesystem
You choose at create time (Linux: ext4 / xfs; Windows: NTFS via Disk Management)
There are no performance tiers, IOPS classes, or “general purpose vs optimized” variants today. Future tiers may be added; this page will list them when they exist.
A volume is independent of the VM it attaches to. Conceptually, a volume sits next to the VM, not inside it.
VM event
Volume result
VM stop / start
Stays attached, contents intact
VM reboot
Stays attached, contents intact
VM resize (compute)
Stays attached, contents intact
VM reinstall (Change OS)
Stays attached — survives the OS-disk wipe
VM factory reset
Stays attached — preserved
VM destroy
Detached, not deleted — keeps existing in your account, billing for storage, until you delete the volume separately
This is the practical reason to put data you can’t lose on a volume, not on the VM’s base disk. The base disk is destroyed with the VM; volumes outlive it.See Storage model for the full survival matrix across all storage objects.
A volume can be attached to one VM at a time. This is a hard constraint — Raff doesn’t support multi-attach (where two VMs see the same volume simultaneously).
A volume’s region and project come from the VM it’s attached to. Three rules follow:
A volume can only be reattached within its original region. You can’t take a us-east volume and attach it to a VM in another region.
Volumes follow their VM’s project. Attach to a VM in project prod and the volume sits in prod.
Standalone (unattached) volumes sit in the dashboard’s currently-selected region until you attach them. Pick the region carefully when creating an unattached volume — same one-way constraint applies.
To “move” a volume across regions: snapshot it, restore the snapshot in the new region, then delete the original. (Cross-region snapshot restore depends on platform support — see Create a snapshot.)