To build a highly available (HA) Proxmox cluster, your storage architecture is absolutely critical. HA in Proxmox depends on shared storage so that VMs can move between nodes without losing access to their disks.

1. Use Shared Storage

All nodes must be able to access the same VM disk images.

Best Options:

a. Ceph (Highly Recommended)

• Integrated into Proxmox VE
• Scalable, distributed, fault-tolerant
• Provides block storage (RBD), object storage, and file storage
• Supports replication and self-healing
• Excellent for VM disk storage

Pros:

• No single point of failure
• Fully redundant
• Great performance with SSDs and fast networks (10/25/40/100 Gbps)

Requirements:

• Minimum 3 nodes (for quorum and redundancy)
• Prefer SSDs or NVMe for journals / DB / OSDs
• Fast and redundant networking (dedicated Ceph network recommended)

b. ZFS over iSCSI or NFS

If you already have a ZFS NAS (like TrueNAS), you can:

• Export ZFS volumes via iSCSI (preferred) or NFS
• Connect Proxmox nodes to the shared storage

Pros:

• ZFS data integrity
• Snapshot/replication support
• Lower complexity than Ceph

Cons:

• Less resilient than Ceph (single NAS is a SPOF unless HA NAS setup is used)
• Requires reliable external storage system

c. Enterprise SAN (iSCSI/FC)

• Proxmox supports connecting to SANs via iSCSI or Fibre Channel.
• VMs use LVM over iSCSI or raw block devices.

Pros:

• High performance, enterprise features
• Existing infrastructure use

Cons:

• Expensive
• Still a SPOF unless SAN itself is HA

Not Recommended for HA

• DRBD: Older method, more complex, less used now that Ceph is widely adopted.

Storage Redundancy & Performance

Example HA Cluster with Ceph

3+ Node Proxmox Cluster with Ceph

• Each node:
  • Dual CPUs
  • 128–256 GB RAM
  • 2× 10GbE (1 for cluster, 1 for Ceph)
  • 3+ SSDs or NVMe (used as Ceph OSDs)
  • Proxmox with built-in Ceph
• Setup:
  • Use Proxmox GUI to deploy Ceph
  • Enable HA for critical VMs
  • Use Proxmox replication for fast recovery of non-HA VMs

Recommended: