Guest Operating System Notes¶

This page covers OS-specific considerations, recommendations, and behaviors when running virtual machines on HVM clusters.

The HVM controls referenced on this page are selected from Advanced Options on the Configure step during provisioning. See VM Advanced Options for field availability and reconfiguration behavior.

Windows Guests¶

Cores Per Socket¶

Windows 10 and Windows 11 display only one socket in Task Manager regardless of the actual CPU topology. If your Windows VM has multiple vCPUs and you want the guest to see them as cores within a single socket (rather than multiple single-core sockets), set Cores Per Socket equal to the total vCPU count during provisioning.

For example, a VM with 8 vCPUs and Cores Per Socket set to 8 will present as 1 socket with 8 cores. If Cores Per Socket is left at the default (1), Windows sees 8 sockets with 1 core each, which may confuse certain applications and licensing models.

VirtIO Drivers¶

Windows does not include VirtIO drivers natively. To achieve optimal I/O performance, VirtIO drivers must be installed during the initial Windows setup:

1. When provisioning a Windows VM from an ISO, enable Attach VirtIO Drivers in Advanced Options

2. This mounts the VirtIO driver ISO as a secondary CD-ROM

3. During the Windows “Where do you want to install Windows?” step, click Load Driver

4. Browse to the mounted VirtIO ISO and select the appropriate driver for your Windows version (e.g., viostor for the storage controller)

5. Complete the Windows installation

6. After Windows is installed and you are logged in, navigate to the VirtIO ISO drive in Windows Explorer

7. Right-click virtio-win-guest-tools and select Install

8. Step through the installer — this installs all remaining VirtIO drivers (network, balloon, serial) and the QEMU Guest Agent

Hardware Compatibility¶

• Modern (Q35) — Use for Windows 8/Server 2012 and later. Supports PCIe, VirtIO, and modern features.

• Legacy (i440fx) — Use for Windows XP, Windows Server 2003, or other guests that cannot load VirtIO drivers. This provides emulated IDE storage and legacy PCI bus.

TPM and Secure Boot¶

Windows 11 requires both TPM 2.0 and Secure Boot. Enable both TPM and Secure Boot (which also requires UEFI) in Advanced Options when provisioning Windows 11 VMs.

Windows 10 does not require TPM or Secure Boot but supports both for enhanced security features like BitLocker.

Guest Customization (Sysprep)¶

When cloning a Windows VM or provisioning from a template, HPE Morpheus Software runs Sysprep via the QEMU Guest Agent to generalize the image (new SID, hostname, network identity). This requires:

• The VirtIO drivers and QEMU Guest Agent to be installed in the template (virtioSupported must be true)

• The template to have been prepared with sysprep /generalize (or have the Sysprep utility available)

If the QEMU Guest Agent is not available, guest customization will be skipped and the VM may have identity conflicts.

Hostname Length¶

Windows hostnames are automatically truncated to 15 characters (the NetBIOS limit). If your naming policy generates longer names, be aware that the Windows hostname will differ from the HPE Morpheus Software display name.

Agent Communication¶

HPE Morpheus Software communicates with Windows VMs using:

1. QEMU Guest Agent (preferred) — works without network connectivity, used for Sysprep, command execution, and file operations

2. WinRM (fallback) — requires network connectivity and WinRM to be enabled in the guest

Ensure WinRM is configured with winrm quickconfig and set to automatic startup in your Windows templates.

Diagnosing Inter-Host Network Throughput¶

Do not use a nominal 10 Gb NIC speed as a throughput guarantee for a Windows VM. End-to-end results depend on the guest driver and CPU, VM sizing, host load, Virtual Switch and bond configuration, MTU, and every physical link in the path.

Before changing Windows settings:

1. Confirm the VirtIO network driver is installed and current for the supported guest image.

2. Test in both directions with a dedicated throughput tool and multiple streams; record latency, retransmits, guest CPU, and host CPU during the test.

3. Compare same-host and different-host VM results to isolate the virtual/guest path from the physical uplinks.

4. Verify negotiated physical link speed, bond state, switch counters, VLANs, and end-to-end MTU. Resolve drops, errors, or MTU mismatch first.

5. Record the Windows version, VirtIO driver version, HVM release, VM vCPU/memory, test command, and topology so results are repeatable.

As a bounded diagnostic, an administrator can inspect RSS and TCP auto-tuning from an elevated PowerShell session:

Get-NetAdapterRss
Get-NetAdapterAdvancedProperty -Name "<adapter>"
netsh int tcp show global

If RSS is disabled and the installed VirtIO driver supports it, test enabling RSS on the specific adapter with Enable-NetAdapterRss -Name "<adapter>" and repeat the same benchmark. Revert with Disable-NetAdapterRss -Name "<adapter>" if throughput, CPU use, or stability regresses.

Do not disable checksum or large-send offload, or add MaxUserPort and TcpTimedWaitDelay registry values, as a general HVM optimization. Those changes affect guest-wide networking and should be used only for a measured workload under Microsoft and HPE Support direction, with the original adapter and registry values recorded for rollback.

Linux Guests¶

Cloud-Init¶

Linux VMs use cloud-init by default for guest customization. HPE Morpheus Software generates cloud-init configuration (user data, meta data, and network data) and delivers it via an ISO mounted to the VM at boot.

Cloud-init handles:

• Hostname and network configuration

• SSH key injection

• User account creation

• Post-provisioning scripts

QEMU Guest Agent¶

For enhanced management capabilities (graceful shutdown, filesystem freeze for snapshots, live IP reporting), install the QEMU Guest Agent in your Linux templates:

• Ubuntu/Debian: apt install qemu-guest-agent

• RHEL/CentOS: yum install qemu-guest-agent

• Enable and start: systemctl enable --now qemu-guest-agent

The QEMU Guest Agent enables HPE Morpheus Software to communicate with the VM even when network-based access (SSH) is unavailable.

Clock and Time Synchronization¶

HVM 9.1 configures the libvirt domain XML so that guest time stays synchronized with the Host:

• Linux guests use kvmclock as the time source with RTC set to catchup mode. This ensures the guest clock recovers after a pause (live migration, snapshot, host suspend) without requiring manual NTP correction.

• Windows guests use localtime for the clock offset rather than UTC, matching the Windows default expectation. The hypervisor clock is set to catchup mode as well.

These settings are applied automatically to new VMs and to existing VMs on their next restart. No user configuration is required. Guest-level NTP or Windows Time Service remains recommended for long-running workloads, but the hypervisor-level catchup reduces drift during VM lifecycle events.

VirtIO Drivers¶

Modern Linux kernels (2.6.25+) include VirtIO drivers natively. No additional driver installation is needed. The Hardware Compatibility setting should be left at Modern (Q35) for all supported Linux distributions.

For very old Linux distributions (kernel < 2.6.25) that lack VirtIO support, select Legacy (i440fx) hardware compatibility.

Cores Per Socket¶

Linux correctly identifies and utilizes multi-socket topologies, so Cores Per Socket configuration is less critical than for Windows. However, for NUMA-aware workloads, aligning Cores Per Socket with the host’s physical cores-per-NUMA-node can improve memory locality. See VM Compute Configuration for vNUMA configuration.

Template Preparation¶

Before converting a Linux VM to a template:

1. Clear cloud-init state: sudo cloud-init clean --logs --machine-id

2. Remove SSH host keys: sudo rm -f /etc/ssh/ssh_host_*

3. Clear machine ID: sudo truncate -s 0 /etc/machine-id

4. Remove persistent network rules if present

5. Power off the VM

General Considerations¶

Power Operations & Shutdown Behavior¶

HVM/KVM VMs support separate graceful and forced stop operations. From an Instance or VM detail page, select Stop Server or Stop, then choose the shutdown type:

• Graceful Shutdown — The default. Sends an ACPI power-button signal and waits up to five minutes for the guest to report that it is shut off.

• Force Stop (immediate power-off) — Immediately powers off the VM without notifying the guest operating system.

If a graceful shutdown reaches the five-minute timeout, HPE Morpheus Software leaves the VM running and reports the timeout. It does not automatically change to a forced stop. Resolve the guest shutdown issue or submit a separate Force Stop (immediate power-off) request.

The same shutdown-type selector is used for one or multiple selected HVM/KVM Instances or VMs. Graceful shutdown remains selected by default. Instance stops also offer Mute Monitoring for planned downtime. The selected shutdown type is recorded in activity and audit history; forced stops are identified as force operations.

How ACPI shutdown works with different guest configurations:

When the QEMU Guest Agent matters for power-related operations:

While the guest agent does not affect the graceful or forced Stop Server operation itself, it is used for:

• Filesystem freeze/thaw — Before snapshots, the guest agent freezes guest filesystems for consistency

• Graceful reboot — The agent can execute a clean reboot command inside the guest

• Guest OS information — Reporting IP addresses, OS version, and hostname back to HPE Morpheus Software