ZFS on Proxmox: Performance and Efficiency Best Practices

Whether you're running a small lab or managing a production cluster, ZFS gives Proxmox administrators enterprise-grade storage capabilities: snapshots, compression, replication, and rock-solid data protection.

In this session, Allan Jude and JT Pennington will walk through practical ways to get the most out of ZFS in Proxmox. From tuning for performance to designing efficient replication, you'll come away with tested practices you can put to use right away.

What You’ll Learn:

•  How to tune ZFS for performance in Proxmox environments
• Compression and caching strategies for different workloads
• Replication workflows that balance efficiency with resilience
• Common pitfalls and how to avoid them

Top Questions from the Session—Answered!

🗨️ How do you implement high availability (HA) for storage in Proxmox? What's the best way to ensure real-time data replication, and how can you minimize disruption during storage failover for VMs running databases?

    For true real-time failover, we recommend shared drives with multipath SAS connections accessible by both active and standby storage servers. When the active node fails, a floating IP moves to the standby, keeping data instantly available with zero replication delay. For disaster recovery, ZFS replication to a second site offers protection with a slight delay—around 15 minutes—while databases may still benefit from app-level replication for seamless recovery.

🗨️ Are larger hard drives becoming scarce due to AI data center buildouts? What should you do when you can't source new enterprise spinning hard drives?

    Yes, we have seen enterprise HDD lead times stretching to eight weeks or more as demand from AI data centers grows. If you are looking for alternatives, we recommend choosing recertified drives over refurbished ones: they're unused units that have had additional QA before being released, offering solid reliability and cost savings. Avoid "refurbished" drives unless you can verify usage hours and source quality. And always keep 10–15% of your pool free to buy time for replacements when supply is tight.

🗨️ Do you recommend CPU pinning for NVME systems?

    CPU pinning can help improve performance, but it's about balance rather than strict one-to-one pinning. NVMe devices use multiple queues, and the operating system distributes them across CPU cores. You can reserve some cores for other tasks or tune interrupt moderation to reduce CPU overhead. The key is to balance the load across CPUs to ensure consistently fast I/O without unnecessary interruptions.

Date: October 23, 2025

Time: 11:00 AM EDT 

Duration: 50 minutes.