Mounting shared storage

Mount persistent storage volumes to your GPU instances across different providers.

Overview

Persistent storage volumes provide shared, network-attached storage that can be attached to your GPU instances. These volumes persist independently of your instances, so data is not lost when an instance terminates.

Volumes attach to CPU nodes as well as GPU instances. The volume step in the CPU deploy wizard is the same one the GPU wizard uses, and the provider rules below apply unchanged. The exception is a Spheron AI CPU node in Noida or Chennai, which takes no volume, so the wizard hides the volume step.

Key features:

Volumes persist independently of instances; data is not lost when an instance terminates
A single volume can be attached to multiple instances simultaneously (provider-dependent)
Up to 10 volumes can be attached to a single instance on Verda and on Spheron AI (Canada 1, Montreal 1, Norway 1, Oslo 1, US); Spheron AI in Noida and Chennai, Spheron MS and Spheron Con take one volume per instance
Provider-specific mounting protocols (NFS or block device)
Volumes can be attached and detached from running instances without data loss on Verda, Spheron AI, and Spheron ES; Spheron MS binds a volume at instance launch only, and so do Spheron AI in Noida and Chennai and Spheron Con, which mounts it at /workspace in the container
Resizing is grow-only and available on Verda, Spheron ES, and Spheron Con only
Billed for the full provisioned size for as long as the volume exists, attached or not

Prerequisites

Before mounting a volume, ensure you have:

Created a persistent storage volume via the Spheron dashboard (Create Volume page)
Attached the volume to your GPU instance
Retrieved the volume's connection details (virtual IP, mount path, protocol)
SSH access to your instance

Provider-specific mounting guides

Select your provider below for detailed mounting instructions:

Verda

Protocol: Network File System (NFS) Status: Available

Mount shared filesystems on Verda instances using datacenter-region NFS endpoints and pseudopaths.

View Verda mounting guide

Key features:

NFS protocol with high-performance nconnect option
Datacenter-specific NFS endpoints (e.g., nfs.fin-01.datacrunch.io)
Pseudopath-based filesystem identification
Up to 10 volumes can be attached to a single instance
Supports concurrent multi-instance access within the same datacenter
Cluster-wide mounting via pdsh

Spheron AI

Protocol: Block device (ext4, xfs, or ntfs) or shared filesystem, depending on region Status: Available

Spheron AI offers two kinds of volume, depending on the region. In Canada 1, Montreal 1, Norway 1, Oslo 1, and the US, volumes are raw block devices that you format and mount with your preferred filesystem. In Noida and Chennai, volumes are shared filesystems that mount automatically at /home/jl_fs.

View Spheron AI mounting guide

Key features:

Block volumes: up to 50 TB each and up to 10 per instance, with attach and detach while the instance runs. Each block volume attaches to one instance at a time
Block volumes: direct-attached storage with local disk throughput, formatted as ext4, xfs, or ntfs
Block volumes: UUID-based fstab configuration with nofail for safe boot persistence
Shared volumes: 50 GB to 2,048 GB, one per instance, chosen at deploy time, and shareable by several instances (see Shared volumes in Noida and Chennai)
Per-region size limits and pricing in the create-volume wizard
Available across Canada 1, Montreal 1, Norway 1, Oslo 1, US, Noida, and Chennai

Spheron ES

Protocol: virtiofs (shared filesystem) Status: Available

Mount shared filesystems on Spheron ES (Extra Supply) instances using virtio-attached devices identified by a mount tag.

View Spheron ES mounting guide

Key features:

virtiofs protocol with directly attached shared filesystems
Mount tag based identification (copyable from the volume card)
Up to 256 TiB per volume (sizeInGb: 262144 on the API)
Same volume can be mounted on multiple instances simultaneously, read-write or read-only
Volumes can be resized in place without recreation or data migration (shrinking is not supported)
Attach at deploy time without a restart, or hot-attach with a brief VM restart (data preserved)
Multiple volumes can be attached to a single instance
Available across EU North 1 (Finland), EU West 1 (France), ME West 1 (Middle East), and US Central 1 (US)

Spheron MS

Protocol: Block device (ext4) Status: Available

Attach a block volume to a Spheron MS instance at deployment time, then format and mount it as a raw block device.

View Spheron MS mounting guide

Key features:

Direct-attached block storage with local disk throughput
Up to 16 TB per volume, one volume per instance
Wired in before the machine boots, so the disk is present on first connection
UUID-based fstab configuration with nofail for safe boot persistence
Available across US, Canada, Norway, Poland, Israel, India, Japan, Singapore, and Australia regions

Spheron Con

Protocol: Network volume, mounted for you Status: Available

Storage for containers. Pick the volume when you deploy a container and it is mounted at /workspace.

View Spheron Con volume guide

Key features:

Mounted at /workspace with nothing to format or mount
10 GB to 4,096 GB per volume, one volume per container
Several containers in the same region can share one volume
Grow and rename at any time; shrinking is not supported
For containers only, not virtual machines

Spheron AM and Massed Compute

Status: Not available

Spheron AM and Massed Compute do not support volumes. Keep anything that must survive termination off the instance, or deploy on a provider that supports volumes.

General concepts

Volume lifecycle

Create Volume → Attach to Instance → Mount → Use → Unmount → Detach → Delete
     ↓              ↓                  ↓       ↓         ↓         ↓        ↓
  Available     Available          Attached  Attached  Attached  Available Deleted

Common operations

Creating volumes: Create persistent storage volumes from the Create Volume page in the Spheron dashboard. Specify the provider, name, and size in GB.

Attaching volumes: Attach a volume to a running instance from the volume details page in the dashboard. The volume and instance must use the same provider. On Spheron MS, Spheron Con, and Spheron AI in Noida and Chennai, select the volume in the deploy wizard instead, because those volumes cannot be attached to a running instance. On Spheron ES, attaching to a running instance restarts it.

Mounting: Follow the provider-specific mounting instructions in the guides linked above.

Detaching volumes: Unmount the volume from the instance first, then detach it from the dashboard. Volume data is preserved after detachment.

What's next

Volume API Reference: Create and manage volumes programmatically
SSH Connection Setup: Secure instance access
Security Best Practices: Secure your storage and data
Getting Started: Complete deployment guide

For questions, use the chat widget on app.spheron.ai.