Hardware compatibility
Use these topics to check server models, components, and known incompatibilities before purchasing or reusing hardware for a Virtuozzo Infrastructure cluster.
Compatible hardware
You can run Virtuozzo Infrastructure on any server that meets the system requirements, described in Requirements reference. Virtuozzo does not maintain its own list of certified server models or components. Because Virtuozzo Infrastructure is based on Red Hat Enterprise Linux (RHEL), you can refer to the Red Hat Ecosystem Catalog for compatible hardware. Use the RHEL version that corresponds to your product version.
When purchasing new equipment, hardware certified for the corresponding RHEL version is the safest choice. Hardware that is not listed in the catalog can still be used with Virtuozzo Infrastructure if it meets the system requirements and is not listed as incompatible below.
| Product version | Upstream RHEL version | Supported NVIDIA driver for vGPU |
|---|---|---|
| 8.0 (upcoming) | RHEL 10 | R595 branch, NVIDIA GRID v21 |
| 7.4 | RHEL 9 | R580 branch, NVIDIA GRID v19 |
| 7.3 Update 1 | RHEL 9 | R580 branch, NVIDIA GRID v19 |
| 7.3 | RHEL 9 | R560 branch, NVIDIA GRID v17 |
Recommendations
- Although a cluster can contain nodes with different hardware configurations, using similarly configured nodes provides more predictable performance and better resource balance.
- Use UEFI instead of BIOS where supported, particularly with NVMe drives.
- Test the cluster infrastructure thoroughly before putting it into production. Pay particular attention to potential failure points, such as SSDs and network adapter bonds.
Blade servers and external storage
Blade servers can be used when they meet the general server requirements, but they often provide only up to two internal drives. Using local Virtuozzo Storage as the main storage for compute volumes requires at least three SSDs or HDDs per server and a network connection of at least 10 Gbit/s between storage servers. With two SSDs mirrored for the system disk, no internal drives may remain for local storage.
You can use blade servers with supported external storage, for example, iSCSI, FC, or NFS, as the main storage for compute volumes. Ensure sufficient network performance for storage traffic, with at least 10 Gbit/s between blades participating in the storage cluster.
Even with external storage, Virtuozzo Storage remains mandatory for VM templates and some system components with the default storage policy
Consider the following use cases:
- Blade servers with only two internal drives and local Virtuozzo Storage as the main storage for compute volumes are not recommended and cannot be used.
- Blade servers with storage blades or at least three internal drives can be used for local storage, but the configuration is not recommended because storage options are limited.
- Blade servers with external storage can be used as regular compute servers. Store compute volumes on the external storage. A single system disk without RAID1 is possible but not recommended, because losing that disk loses the entire node.
Incompatible hardware
Some hardware may meet the minimum system requirements but is known to fail or cause data integrity issues with Virtuozzo Infrastructure. The following recommendations describe hardware to avoid.
What to avoid
- Desktop and consumer hardware: Do not use desktop-grade SSDs, network adapters, or switches. Such hardware is not designed for sustained enterprise workloads and may provide insufficient performance or reliability. Consumer-grade SSDs may also have low write endurance and inconsistent sustained performance.
- SSDs without power-loss protection: Disks used for metadata, cache, and database roles must have power-loss protection. Drives that ignore disk flushes can corrupt the file system if power is lost.
- Non-ECC memory: ECC RAM is required. Non-ECC memory may cause silent data corruption.
- QLC or desktop-grade SSDs for the system disk: Use enterprise SSDs with sufficient write endurance and consistent sustained performance.
- RAID volumes built from storage disks: Virtuozzo Infrastructure System must access each storage disk as a separate device. Present storage disks individually, for example, in JBOD or pass-through mode. Do not mirror cache devices in RAID1.
- SMR (shingled magnetic recording) HDDs: Use SMR HDDs only for the storage role and only on nodes with an SSD cache disk.
- PMem and NVRAM devices: These devices are not supported for cluster storage or cache.
Known hardware limitations
The following table lists specific hardware models, drivers, and configurations with known compatibility limitations.
For upgrades to version 8.0, the eligibility checks compare installed devices and loaded kernel drivers with RHEL 10 compatibility data. The table also includes hardware and drivers that may be reported as removed or unmaintained during the upgrade.
| Type | Vendor / model | Driver / PCI ID | Reason |
|---|---|---|---|
| Disk (SSD) | OCZ Vertex 3 | — |
Ignores disk flushes and reports writes as committed before the data reaches stable storage, which may cause file system corruption on power loss. |
| Intel 520 / X25-E / X25-M G2 | — |
Ignore disk flushes and are therefore unsuitable for workloads that require reliable data commits, including databases. |
|
| NIC | Intel EXPI9301CTBLK / Realtek 8129 | — |
Desktop adapters are not designed for sustained load and may not support full-duplex links. |
| Broadcom BCM57840 NetXtreme II 10/20 GbE / HPE FlexFabric 10Gb 2-port 536FLB | BNX2X
|
Limits the MTU to 3616, which may degrade cluster performance. The driver is reported as unmaintained during the upgrade to version 8.0. Use the BNXT driver instead. |
|
| NIC / RDMA adapter | Mellanox ConnectX-3 / ConnectX-3 Pro | mlx4_core; PCI IDs 15B3:1003–15B3:1010 |
Removed from RHEL 10. These adapters are reported as removed during the upgrade to version 8.0. |
| NIC / RDMA driver | Netronome NFP | nfp
|
Removed from RHEL 10. Replace the affected hardware or remove the driver dependency. |
| Cisco VIC usNIC | usnic_verbs
|
Removed from RHEL 10. Replace the affected hardware or remove the driver dependency. | |
| VMware Paravirtual RDMA | vmw_pvrdma
|
Removed from RHEL 10. Replace the affected hardware or remove the driver dependency. | |
| Software RDMA / iWARP | rdma_rxe, siw |
Removed from RHEL 10. Remove the driver dependency. |
|
| Network adapter | QLogic ISP3XXX | qla3xxx
|
Removed from RHEL 10. Nodes with these adapters are not eligible for upgrade to version 8.0 until the hardware is replaced or removed. |
| Storage adapter | QLogic ISP4010 / ISP4022 / ISP4032 iSCSI HBAs | qla4xxx
|
Removed from RHEL 10. Nodes with these adapters are not eligible for upgrade to version 8.0 until the hardware is replaced or removed. |
| Storage / platform driver | Legacy AAC RAID | aacraid
|
Available but unmaintained in RHEL 10. Consider replacing the affected hardware or moving workloads to supported nodes. |
| QLogic FCoE | bnx2fc
|
Available but unmaintained in RHEL 10. Consider replacing the affected hardware or moving workloads to supported nodes. |
|
| HISILICON SAS | hisi_sas_main
|
Available but unmaintained in RHEL 10. Consider replacing the affected hardware or moving workloads to supported nodes. |
|
| APM X-Gene AHCI | ahci_xgene
|
Available but unmaintained in RHEL 10. Consider replacing the affected hardware or moving workloads to supported nodes. |
|
| Seattle AHCI | ahci_seattle
|
Available but unmaintained in RHEL 10. Consider replacing the affected hardware or moving workloads to supported nodes. |
|
| Network configuration | Linux network teaming | team
|
Available but unmaintained in RHEL 10. Consider migrating network teaming configurations to supported bonding. |
| FireWire controller | FireWire devices | firewire_core
|
Removed from RHEL 10. Remove FireWire dependencies before upgrading to version 8.0. |
See also