When an Oracle audit reaches a VMware estate, the conversation quickly turns to movement. Where could this virtual machine go? Where has it been? VMware keeps a record, and Oracle has learned to ask for it. vMotion and migration history become the evidence in an argument that a workload should be licensed not where it ran but everywhere it could have run. The argument is powerful as theatre and weak as contract, and knowing the difference is the whole defense.
How does Oracle use vMotion history in an audit?
Oracle uses vMotion history in an audit to argue that because a virtual machine moved, or could move, across the hosts in a cluster, every one of those hosts is licensable capacity. The migration logs are presented as proof that the Oracle workload had the run of the cluster, and the finding then prices every processor on every host the virtual machine could reach. A database that lived comfortably on two processors becomes a finding spanning an entire cluster, justified by a log showing the machine was once relocated for maintenance or balanced for performance.
The move from history to liability is the sleight of hand. A record of where a workload ran is turned into a claim about where it must be licensed, and the bridge between the two is Oracle's partitioning policy, not anything in the migration data itself. The logs are real. The leap from logs to cluster wide licensing is an interpretation, and interpretations can be contested.
Does vMotion history prove cluster wide usage?
vMotion history does not prove cluster wide usage; it proves where a virtual machine actually ran, which is a far narrower fact than the claim built on top of it. Migration logs record discrete events: this machine moved from this host to that host at this time. They do not establish that the contract requires licensing every host in reach, and they do not turn theoretical mobility into actual usage. The cluster wide conclusion comes entirely from the partitioning policy's treatment of soft partitioning, and that policy is a non contractual document.
This is the pivot of the defense. Contract language beats policy, and most signed agreements define the licensable unit by where the program is installed and running, not by where a hypervisor could in principle place it. Read against the contract, vMotion history is interesting context rather than the decisive evidence Oracle presents it as. The migration log answers a question the contract may never ask. An independent line by line review of Oracle findings typically cuts a claim 60 to 80 percent, and vMotion based virtualization claims reduce strongly because the gap between the policy and the signed terms is so wide.
An audit obtains six months of vMotion logs showing an Oracle virtual machine migrated across four of eight hosts in a cluster during routine maintenance windows. The opening finding licenses all eight hosts, citing the machine's mobility. The buyer responds with the contract, which ties licensing to installed and running processors, and with evidence that the workload ran on its home host for the overwhelming majority of the period, moving only briefly for maintenance. The defensible footprint settles on the genuine running capacity rather than the eight host cluster, removing the largest line in the finding.
How do you defend against a vMotion based claim?
You defend against a vMotion based claim by holding the finding to the contract's definition of the licensable unit, presenting your own evidence of where the workload genuinely ran, and documenting the controls that limited its reach. Start with the agreement and establish how it defines when a licence is required, which is almost always tied to where the program is installed and running rather than to potential mobility. Then build the evidential record: where the virtual machine actually executed, for how long, and within what boundaries. Your own data, presented first, frames the argument before Oracle's interpretation can.
Host affinity rules and other configuration controls strengthen the position by showing the workload was deliberately confined, narrowing both the technical reach and the credibility of the cluster wide claim. Where such controls exist, document them with dates and configuration evidence. Where they do not, the contract argument still stands on its own. The discipline is consistent throughout: accept the migration logs as a record of the past, refuse the leap to cluster wide licensing, and insist that the contract, not the partitioning policy, decides what is owed.
The vMotion argument is one move inside the wider virtualization claim. See disputing cluster wide virtualization claims and how scripts overcount in virtualized estates for the surrounding defense, and the full method in the Oracle virtualization licensing guide.