This section provides the best practices to follow when configuring Proxmox with FlashArray as storage using the NVMe-TCP protocol.
Reference Architecture
A production-grade NVMe-TCP deployment consists of:
| Component: Best Practice | Reason |
|---|---|
| Proxmox Cluster: 3+ nodes for high availability | Minimum 3 nodes required for Proxmox HA quorum; allows maintenance without downtime |
| Dedicated Storage Network: Isolated network infrastructure for storage traffic | Prevents storage I/O contention with other traffic; ensures predictable performance; simplifies troubleshooting |
| NVMe-TCP Storage Array: Enterprise storage with multiple controllers | Multiple controllers eliminate single point of failure; enable active-active multipath for performance and redundancy |
| Multipath Configuration: Multiple redundant paths for failover and load balancing | Automatic failover on path failure; load distribution across all paths; no downtime during maintenance |
Deployment Topology
Deployment and Ongoing Operations Summary
Initial Deployment
- Configure the network infrastructure with redundancy
- Configure end-to-end MTU of 9000
- Load persistent NVMe kernel modules
- Generate host NQNs and register them with FlashArray
- Connect all paths (2 NICs × 4 portals = 8 paths)
- Set the IO policy to "queue-depth"
- Configure persistent connection
- Enable the nvmf-autoconnect service
- Create and share the LVM volume group
- Confirm that all nodes can access the storage
Ongoing Operations
- Monitor the path health daily
- Monitor the NVMe SMART data weekly
- Review performance metrics weekly
- Test the failover procedures quarterly
- Update the firmware during maintenance windows
- Verify backups regularly
- Document any configuration changes
- Keep runbooks updated
Read the following sections for more information on network, high-availability, redundancy, performance, security, monitoring and maintenance best practices.