If a virtual machine has more than 255 cores per socket, Fast Suspend/Resume (FSR) or migration tasks might fail
If a VM has more than 255 cores per socket, tasks to migrate, fast resume, or hot-plug the VM fail. In the vmware.log you see a message similar to:
yyyy-mm-ddTHH:MM:54.488Z In(05) vmx - [msg.checkpoint.inConsistentCoresPerSocket] The suspended image contains a coresPerSocket value (0) that does not match with VM's actual coresPerSocket value (256)
yyyy-mm-ddTHH:MM:54.891Z In(05) vmx - PStrIntern expansion: nBkts=256
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - MigrateSetStateFinished: type=2 new state=MIGRATE_FROM_VMX_FINISHED
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - MigrateSetState: Transitioning from state MIGRATE_FROM_VMX_CHECKPT (11) to MIGRATE_FROM_VMX_FINISHED (12).
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - Migrate: Caching migration error message list:
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - [msg.checkpoint.migration.failedReceive] Failed to receive migration.
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - [msg.checkpoint.inConsistentCoresPerSocket] The suspended image contains a coresPerSocket value (0) that does not match with VM's actual coresPerSocket value (256).
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - Msg_Post: Error
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - [msg.checkpoint.inConsistentCoresPerSocket] The suspended image contains a coresPerSocket value (0) that does not match with VM's actual coresPerSocket value (256).
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - [msg.checkpoint.migration.failedReceive] Failed to receive migration.
yyyy-mm-ddTHH:MM:54.893Z In(05) vmx - ----------------------------------------
yyyy-mm-ddTHH:MM:54.894Z In(05) vmx - Module 'CheckpointLate' power on failed.
yyyy-mm-ddTHH:MM:54.894Z No(00) vmx - PowerOnTiming: ALL took 116596139 us
yyyy-mm-ddTHH:MM:54.894Z In(05) vmx - VMX_PowerOn: ModuleTable_PowerOn = 0
yyyy-mm-ddTHH:MM:54.894Z In(05) vmx - TOOLS refreshing value for isoImageExists (refreshCount=1, lastCount=0).
yyyy-mm-ddTHH:MM:54.894Z In(05) vmx - ToolsISO: Refreshing imageName for 'rhel7-64' (refreshCount=1, lastCount=1).
ESXi 8.x
When a virtual machine is configured with cores per socket (CPS) ≥ 256, this causes failures in operations that rely on checkpointing, such as vCPU hot-add/hot-plug (FSR), Suspend/Resume, and vMotion.
This issue exists in vSphere 8.0 and later, starting with virtual hardware version 20 (HWv20), where support for 256 CPS was introduced.
WorkAround : Reduce coresPerSocket to 128
The issue is permanently resolved in ESXi 8.0 Update 3h