Live Partition Mobility
Live Partition Mobility moves a running IBM i, AIX or Linux partition from one IBM Power server to another without shutting it down, so an IBM i shop can empty a server for maintenance without an outage window. It is part of PowerVM Enterprise Edition, and IBM requires every I/O device on the moving partition to be virtual and served by a Virtual I/O Server.
What does Live Partition Mobility do?
It picks up a whole partition and sets it down on a different Power server while the partition keeps running. Think of moving a fish from one tank to another without ever taking it out of the water. The jobs keep going. The users stay signed on.
IBM's own words for it are short: partition mobility is a component of the PowerVM Enterprise Edition hardware feature, and it moves AIX, IBM i and Linux logical partitions from one system to another. A logical partition, or LPAR, is one slice of a Power server running its own operating system. The move carries the processor state, the memory, the attached virtual devices and the connected users. You start it from the Hardware Management Console, from IBM PowerVC, or from PowerVM NovaLink.
What is the difference between an active move and an inactive move?
The two kinds of move IBM documents
| Question | Active move | Inactive move |
|---|---|---|
| IBM's other name for it | Live Partition Mobility | Cold partition mobility |
| State of the partition | Running, including the operating system and applications | Powered off |
| Planned downtime | None. The partition is never shut down. | The partition is already stopped. |
IBM Power 11 documentation for the Power E1180, machine type 9080-HEU, Live Partition Mobility topic.
What does an IBM i partition need before it can move?
More than most shops expect. IBM publishes a separate requirements list for IBM i, and it is stricter than the list for AIX or Linux. The short version is that the partition may not own any hardware of its own.
- PowerVM Enterprise Edition licensed on both the source and the destination server.
- All I/O for the partition virtual and hosted by a Virtual I/O Server, using virtual SCSI, virtual Fibre Channel with N_Port ID Virtualization, or virtual Ethernet.
- No physical I/O adapters on the partition, and no Logical Host Ethernet Adapter.
- External storage only, zoned and masked so a Virtual I/O Server on each server can reach it.
- Both servers on the same network subnet.
- The RestrictedIO Partition property turned on.
- IBM i 7.1.0 Group PTF SF99707 level 4 or later, VIOS 2.2.1.4 or later, and Hardware Management Console 7.7.5 or later.
IBM also rules out a few states rather than a few parts. The partition cannot be the alternative error logging partition, cannot be sitting at Dedicated Service Tools, cannot have HSL OptiConnect or Virtual OptiConnect turned on, and cannot host or be hosted by another IBM i partition. A virtual tape or optical device has to be unconfigured first, and a Fibre Channel tape drive that is varied on will stop the move. The version numbers above are the floor IBM published, not the level you want in 2026. Check IBM's Fix Level Recommendation Tool for the pairing you actually own.
Can a partition with an accelerator card move?
Not as an active move, and this is the part worth knowing before you buy. IBM's rule is that all I/O on the moving partition must be virtual and supported by a Virtual I/O Server. A Spyre Accelerator card is a physical PCIe card plugged into a slot and handed to one partition. That is physical I/O. A Virtual I/O Server can hand out virtual disk and virtual network, but it does not hand out a slice of an accelerator card.
So the AI inference partition is tied to the server that holds its cards. The IBM i partition next to it can still move freely, because its disk and network come from a Virtual I/O Server. Plan the two differently: the IBM i workload is portable, the inference workload is not.
IBM has not published Live Partition Mobility guidance that names the Spyre Accelerator. The reasoning above comes from IBM's general rule that every I/O device on a moving partition must be virtual. Confirm the behavior for your firmware level, your accelerator feature code and your release with IBM or your reseller before you build a plan around it.
What has to be in place first?
Four things to set up before the first move
- 1. License PowerVM Enterprise Edition on both serversPartition mobility is a component of PowerVM Enterprise Edition. Both the source and the destination server need it. A lower PowerVM edition does not move partitions.
- 2. Make every device on the partition virtualDisk and network come from a Virtual I/O Server as virtual SCSI, virtual Fibre Channel or virtual Ethernet. Anything physical and dedicated to the partition has to go first. For IBM i, IBM allows no physical adapters at all.
- 3. Put the storage and the network where both servers can see themThe storage has to be external and zoned so a Virtual I/O Server on each server can reach it. The network has to be the same subnet from both sides. IBM also wants matching logical memory block sizes and enough free memory and processor capacity on the destination.
- 4. Name a mover and keep the heartbeat aliveAt least one Virtual I/O Server on each server needs the Mover Service Partition attribute turned on. That is the partition that carries the move. IBM also checks that Resource Monitoring and Control connections exist between the moving partition, the service processors and the mover partitions.
Drawn from IBM's support guidance on performing Live Partition Mobility and its separate requirements list for IBM i partitions.
Related Terms
AI Workload Placement
The decision about which system, operating environment, and partition executes an AI model or agent, separately from the IBM i application supplying its data.
AI Inference Partition
A separate supported operating-system partition used to execute an AI model while an IBM i partition may remain the production application and data source.
PowerVM
IBM's hypervisor for Power servers, built into system firmware. It is what lets a Linux partition running an AI model sit on the same machine as the production IBM i partition, each with its own processors, memory and cards.
Virtual I/O Server
VIOS is a dedicated IBM Power partition that shares one set of physical disk, Fibre Channel and Ethernet adapters out to every other partition on the server. For an IBM i shop that means a Linux partition running an AI model can get its storage and its network connection from VIOS instead of buying its own physical cards.
Hardware Management Console
The HMC is the console an IBM i shop uses to create the partition an AI model runs in, then hand that partition processors, memory and the PCIe slot holding an accelerator card while production keeps running. IBM Power 11 servers need HMC release V11R1M1110 or later.
Dynamic Logical Partitioning
Dynamic logical partitioning, or DLPAR, moves processors, memory and card slots in and out of a running IBM Power partition without restarting it. IBM documents it for the Spyre Accelerator, so an AI card can join or leave a Linux inference partition while the IBM i production partition keeps working.
Spyre Accelerator
IBM's PCIe AI inference card. On IBM Power 11 it carries 32 cores, 128 GB of memory, and more than 300 TOPS inside a 75 watt envelope.