About this article: AI-assisted writing, reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
Physical hosts in a hosting platform need firmware updates, hypervisor patches and hardware repairs. To do that work without switching customers' servers off, the provider moves running virtual machines to another host first. VMware calls this vMotion; Microsoft calls it live migration.
This article explains how the move works, what a hosted Windows server and its users may notice, which servers cannot be moved this way, and what to agree with the provider.
Why it matters
A provider that can move servers while they run can maintain its hosts without a customer outage. A customer whose server cannot be moved, or whose application reacts badly to the move, needs to know that before the maintenance night, not after.
How it works
VMware vMotion
Broadcom states that migration with vSphere vMotion allows virtual machine processes to continue working throughout a migration. It describes three stages:
- vCenter Server verifies that the VM is in a stable state on its current host.
- The VM's state, including memory, registers and network connections, is copied to the target host.
- The VM resumes its activities on the new host.
If errors occur during migration, Broadcom states that the VM reverts to its original state and location. When only the host changes, the virtual disk stays where it is, on storage shared between the two hosts. vMotion can also move the disk to another datastore at the same time, including where the hosts do not share storage.
Broadcom describes migrations with vMotion as transparent to the running VM. The VM keeps its identity, including the MAC addresses of its network cards.
Hyper-V live migration
Microsoft describes live migration as a Hyper-V feature that lets an administrator transparently move running Virtual Machines from one Hyper-V host to another without perceived downtime. Microsoft names draining a host of VMs before upgrading or decommissioning it as one of its uses, and pairs it with failover clustering for high availability.
"Without perceived downtime" is not "nothing happens"
Both vendors' wording describes what users can expect to experience, not a guarantee that no application can tell. Broadcom's own documentation shows the edge cases:
- Latency-sensitive applications. Broadcom states that for applications such as voice over IP and high-frequency trading, vMotion and DRS are usually deactivated. From vSphere 8.0, a VM at hardware version 20 or later can be configured to notify an application inside the guest when a vMotion starts and finishes. The migration waits for the application to acknowledge, which Broadcom notes may lengthen the overall migration. The mechanism is off by default.
- Graphics-accelerated VMs. For VMs with NVIDIA vGPU, Broadcom documents a per-VM "stun time" limit: the period during which the VM is paused at the switch-over can be estimated, and a limit can stop a migration whose estimate is too long.
The writer's reading: for an ordinary Windows server running business applications, a brief pause at switch-over is the expected effect. Whether a given application notices it depends on that application's sensitivity to short delays.
Servers that cannot be moved live
Broadcom lists conditions that block vMotion, including:
- A virtual device backed by something only the source host has, such as its physical CD drive, or by a device on an administrator's client computer. These must be disconnected first.
- A destination host whose processors cannot offer the same instructions. Broadcom states that processors must come from the same vendor class, AMD or Intel; clock speed, cache size and core count can differ.
- Hardware-specific needs, such as a 3D renderer set to hardware on a destination without a suitable GPU.
A server that cannot move live has to be shut down, migrated cold or left on the host while that host's maintenance waits.
What to choose and why
Agree these points with a hosting provider:
- Is host maintenance done with live migration? If yes, the customer can expect a brief pause rather than a restart.
- Which of my servers cannot be moved live, and why? Broadcom's list includes devices tied to one host and special hardware.
- What happens to those servers during maintenance: a planned shutdown, a cold migration or a delayed maintenance window?
- Are maintenance windows announced, and can sensitive servers be moved at a time the business chooses?
- Are anti-affinity rules kept during maintenance? Moving servers off one host can place a redundant pair together.
- Is my application sensitive to short pauses? Ask the application vendor, and if so, discuss whether vMotion notifications or a scheduled move would help.
A hypothetical example
An invented business runs a hosted accounting server and a phone system server with the same provider. The provider migrates both live during Tuesday evening maintenance. Accounting users notice nothing. The phone server's vendor reports that calls can drop during short pauses, so the business and provider agree to move that server only outside business hours, after notice.
Risks and limits
- Vendor wording describes user experience. "Transparent" and "without perceived downtime" are not a promise that no application will notice. Test sensitive applications with a planned move on isolated test servers before relying on it.
- Host updates that need VM changes are separate. A host patch may later require a VM hardware or VMware Tools update that does restart the server; ask how those are scheduled (writer guidance).
- Migration needs capacity. The remaining hosts must have room for the moved servers.
- No commands are given here. The sources describe configuration in detail.
Sources
Opened 14 November 2026.
- Broadcom TechDocs: vSphere Virtual Machines Migration, updated 28 October 2026.
- Broadcom TechDocs: Migrating Virtual Machines with vSphere vMotion, updated 28 October 2026.
- Broadcom TechDocs: Virtual Machine Conditions and Limitations for vSphere vMotion.
- Broadcom TechDocs: vSphere vMotion Notifications for Latency Sensitive Applications.
- Broadcom TechDocs: How to Set a Stun Time Limit for Your vGPU Virtual Machines.
- Broadcom TechDocs: CPU Compatibility and vSphere Enhanced vMotion Compatibility.
- Microsoft Learn: Live Migration Overview; displayed Last updated: 2020-09-17.
Related service: Cloud hosting.
Related future topics
- Are your redundant servers on the same physical host?
- When the provider's dashboard and your server disagree.
- Planning maintenance windows for hosted business applications.
Learning notes
- vMotion and Hyper-V live migration move running servers between hosts for maintenance.
- vMotion copies memory, registers and network connections, then resumes the VM; on error, it reverts.
- "Without perceived downtime" describes user experience; latency-sensitive applications can notice.
- Host-tied devices and CPU incompatibility block live migration.
- Agree with the provider which servers cannot move live and how they are handled.
This article was originally published on .