Skip to main content

Persistent Volumes

Persistent volumes provide storage that exists independently of your instance lifecycle. Unlike the ephemeral storage that comes with an instance, data stored on a persistent volume is preserved when instances are terminated and can be reattached to new instances.

Key Concepts​

  • Independent lifecycle — Volumes are not tied to a specific instance. You can create, attach, detach, and delete volumes separately from your instances.
  • Reusable storage — Attach the same volume to different instances over time. This is useful for datasets, model weights, or any data you want to persist between sessions.
  • Works with VMs and Docker — Volumes can be attached to both virtual machine and Docker container instances.

Use Cases​

  • Machine learning — Store large model weights or training datasets on a volume so they persist between training runs.
  • Development environments — Keep your development environment and project files on a volume that survives instance restarts.
  • Databases — Run databases with data stored on persistent volumes to ensure data durability.

Multi-Datacenter Access​

Volumes can be accessible from multiple datacenters when the underlying storage cluster is shared. The volume API endpoints (/public/volumes/list, /public/volumes/create, and /public/team-volumes/*) return a datacenters field — an array of all datacenter names where the volume is available. The existing datacenter_name field remains for backward compatibility.

danger

Volumes are deleted when your account balance is depleted. Ensure you maintain a positive balance to avoid data loss.

Step-by-Step Guide​

Step 1: Choose a Supported Data Center​

Persistent volumes are currently available only with certain data center partners. When launching a rental, look for GPU cards labeled "Persistent storage support" (e.g., RTX PRO 6000).

If the selected GPU doesn't show this label, you won't be able to create a persistent volume with that instance.

Step 2: Create a Volume During Instance Setup​

In the instance creation flow:

  1. Choose a GPU that supports persistent storage.
  2. When prompted, select "Create a Volume" and choose a size.

The volume will be automatically attached to your instance and appear under the Volumes tab.

Step 3: Mount the Volume in Your VM​

Once your instance is running, connect to it and list available block devices:

lsblk

Your volume will appear as a block device — usually vdb. You can identify it by the size you chose during setup.

To prepare and mount it:

# Format the disk (do this only once, on first use)
sudo mkfs.ext4 /dev/vdb

# Create a mount point
sudo mkdir -p /mnt/my-volume

# Mount the volume
sudo mount /dev/vdb /mnt/my-volume

# Give your user ownership
sudo chown -R riftuser:riftuser /mnt/my-volume

Replace my-volume with the name you chose for your volume.

caution

Only format the disk (mkfs.ext4) on first use. Formatting an existing volume will erase all data on it.

Step 4: Verify Persistence​

To confirm your data persists across instances:

  1. Write a test file:

    echo "volume test $(date)" > /mnt/my-volume/test.txt

  2. Stop your instance.

  3. Launch a new instance and select "Use Existing Volume" to attach the same volume.

  4. Mount the volume again:

    sudo mount /dev/vdb /mnt/my-volume

  5. Check that the file is still there:

    cat /mnt/my-volume/test.txt

If the file contents appear, your data is persisting correctly.

API Reference​

Volumes can also be managed through the CLI and REST API. See the API documentation for the full list of volume operations.