Virtualized TrueNAS storage ownership
TrueNAS HBA Passthrough: VM Storage Controller Guide
Virtualizing TrueNAS changes the most important hardware question from “which disk is fastest?” to “which layer owns the physical storage controller?” Current TrueNAS documentation warns against routine production use of virtual disks for critical pools and, when virtualization is used, recommends passing physical disks or the entire storage controller to the TrueNAS VM where possible. Whole-HBA passthrough creates the cleanest ownership boundary when the host supports IOMMU and the controller can be isolated.
Quick answer
What is the cleanest HBA passthrough design for TrueNAS?
Give the TrueNAS VM exclusive ownership of a dedicated HBA and every pool disk attached to it. Keep the hypervisor boot device and any host-required storage on separate motherboard SATA/NVMe or another controller. Verify VT-d/AMD-Vi, IOMMU grouping and reset behavior on the exact platform before migrating data. A marketplace “IT mode” HBA does not guarantee that the motherboard can isolate it.
Live Amazon hardware
Current HBA, expander, cable and controller listings for this decision
The Sprint 6A catalogue separates HBAs, SAS expanders, PCIe SATA cards, cable families and cooling hardware. Used and refurbished enterprise controllers remain eligible when the listing matches the exact model, but marketplace IT-mode wording is treated only as evidence to verify after delivery.
Buying decision
Whole-controller ownership is cleaner than mixing host and guest access
The hypervisor should not mount or depend on disks behind a controller that will be handed to TrueNAS. If the host needs one of those drives, redesign the cabling or add a separate controller. This prevents the most dangerous ambiguity in a storage VM: two software layers believing they own the same physical path.
Interactive decision tool
TrueNAS HBA Passthrough Planner
Use this planner to screen controller class, port count, expander need, PCIe budget, cabling and storage ownership before buying hardware. Results are planning guidance; they cannot inspect your motherboard lane map, firmware, IOMMU groups, backplane wiring or the exact card delivered by a marketplace seller.
Compatibility checklist
Four checks before adding more NAS drives
Keep direct disk visibility
TrueNAS and Unraid both benefit from controllers that expose individual disks rather than hiding them behind a hardware RAID virtual disk.
Count connectors, not just ports
Match SFF-8087, SFF-8643, direct SATA and backplane connectors, and confirm breakout direction before buying cables.
Budget PCIe bandwidth and slots
A physical x16 slot can be electrically narrower. Plan controller throughput, free slots and lane sharing around the final drive count.
Treat airflow as part of the controller
Enterprise HBAs and expanders were designed for server airflow. Desktop NAS cases often need deliberate fan placement over the controller heatsink.
TrueNAS documents virtualization as a special case
Current TrueNAS hardware guidance says OpenZFS works best with physical storage devices and warns that virtualizing TrueNAS with virtual disks is not a supported production design for critical data.
When a virtualized design is unavoidable or intentionally chosen, the documentation says to pass physical disks or the entire storage controller where possible. The goal is to preserve direct device semantics for ZFS.
Whole-HBA passthrough creates one clear storage owner
Passing the entire HBA means the guest receives the PCIe controller, not eight separately emulated disks. TrueNAS can then enumerate attached devices through the controller path and maintain stable serial-based identity.
This boundary is easier to reason about during a host reboot or recovery because the hypervisor does not also manage the pool disks.
IOMMU support must exist in CPU, firmware and motherboard topology
VT-d or AMD-Vi support is necessary for direct PCIe assignment, but enabling a BIOS toggle is only the first step. The HBA must appear in an IOMMU group that can be safely assigned without taking unrelated devices with it.
Consumer motherboard lane switches and shared multifunction devices can complicate isolation. Verify the actual group layout; no product page can promise that a specific HBA will be isolatable in every board.
Keep the hypervisor boot disk off the passed controller
The host must start before it can launch the TrueNAS VM, so its boot storage cannot depend on an HBA that will be detached from the host and assigned to the guest.
Use motherboard NVMe/SATA or another dedicated controller for the hypervisor. This also makes recovery easier if the storage VM is intentionally stopped.
Do not split one HBA between host and TrueNAS
A passed-through PCIe device is intended to be owned by one VM. Trying to reserve some downstream disks for the host while passing the same controller to TrueNAS defeats that ownership model.
If the host needs local data disks, attach them through a different controller. Physical separation is much easier to audit than a complicated hybrid design.
HBA mode still matters inside the virtualized architecture
The fact that a controller is passed through does not make hardware RAID desirable. TrueNAS still wants direct disk visibility behind the controller, so HBA/JBOD/IT-style behavior remains the target.
Verify the card after it is visible in the guest. Check serials, SMART information where applicable, link speeds and firmware before importing or creating the pool.
Controller reset behavior can affect VM lifecycle
Some PCIe devices do not reset cleanly when a VM is stopped and started without rebooting the host. TrueNAS documentation for PCI passthrough elsewhere warns that devices without defined reset mechanisms can behave inconsistently.
Test repeated VM shutdown/start cycles and host reboots. A storage VM should not require improvisation after every maintenance window.
Cable and backplane topology still applies inside the guest
Passthrough changes software ownership, not physical signal paths. A 9207 still needs SFF-8087 cabling, a 9300-class HBA still uses its SAS3 connector family, and expanders still share uplink bandwidth.
Document cable groups and drive serials before passing the controller. The guest should inherit a topology that is already physically understandable.
Host backups and TrueNAS pool backups are separate problems
Backing up the hypervisor configuration does not back up the TrueNAS data pool. Likewise, ZFS snapshots in the guest do not protect the hypervisor boot device or VM configuration.
Maintain independent backups for both layers. Recovery documentation should explain how to restore the host, reattach the HBA and then import or recover the TrueNAS pool.
Performance depends on more than passthrough overhead
Direct PCIe assignment removes much of the virtual storage stack, but throughput can still be constrained by HBA PCIe lanes, expander uplinks, network speed, CPU scheduling and the drives themselves.
Measure real scrub, replication and client workload performance. Avoid using one synthetic disk benchmark as proof that the full storage path is healthy.
Production risk should influence whether virtualization is worth it
Virtualization can consolidate hardware and provide flexible compute management, but the storage stack becomes dependent on the hypervisor, IOMMU setup and VM lifecycle.
For critical data, weigh that complexity against a separate bare-metal NAS. The right choice is an operational decision, not simply whether passthrough technically works.
Test a complete failure-and-recovery cycle before relying on it
Shut down the TrueNAS VM cleanly, reboot the host, confirm the HBA returns, verify every disk identity, and practice recovering from a disabled VM or replaced controller.
A passthrough design is ready only when its recovery sequence is documented and repeatable. Keep current backups outside the virtualized pool throughout testing.
Questions people ask
TrueNAS HBA Passthrough questions
Does TrueNAS support being virtualized?
It can run virtualized, but current TrueNAS hardware guidance warns against ordinary production use with virtual disks for critical data. If virtualizing, direct disk or whole-controller passthrough is preferred where possible.
Is whole-HBA passthrough better than passing individual disks?
It usually creates a cleaner storage-ownership boundary because the guest receives the physical controller and attached disks directly.
Do I need VT-d or AMD-Vi?
Yes for PCIe device assignment, and the motherboard firmware and IOMMU grouping must also allow the HBA to be isolated.
Can the hypervisor boot from a disk on the passed HBA?
That is a poor design because the host needs its boot storage before the guest can own the controller. Keep host boot storage elsewhere.
Can the host use one disk on the HBA while TrueNAS uses the others?
Not if the entire HBA is passed through. Use a separate controller for host-owned disks.
Does passthrough remove the need for HBA/IT mode?
No. TrueNAS still benefits from direct disk visibility behind the controller, so avoid hiding pool devices inside hardware RAID virtual disks.
Will any motherboard isolate an HBA?
No guarantee. IOMMU groups and lane topology vary by platform. Verify the exact motherboard and firmware.
Can I pass through a 9207-8i or 9300-8i?
Potentially, when the host supports and isolates the device. The controller model alone does not guarantee passthrough success.
What should I test before importing my pool?
Test HBA detection, serial visibility, SMART access, VM restart behavior, host reboot, sustained I/O and recovery from a stopped VM.
Is passthrough a substitute for backups?
No. It changes device ownership only. Maintain independent backups of important data and the hypervisor configuration.
Official references and methodology
Verify the exact controller, cable path and operating-system behavior
Virtualization guidance follows the current TrueNAS Hardware Guide. The planner deliberately cannot verify IOMMU groups, reset support or motherboard lane topology. Hardware listing claims are used only for model selection; platform passthrough capability must be confirmed on the actual host.
As an Amazon Associate, Cloudzat may earn from qualifying purchases. Marketplace model names, firmware claims, condition, connector layouts, cable direction and seller descriptions can be wrong or change. Verify the delivered board, firmware and cable path before connecting production disks.