Register now or log in to join your professional community.
One of the NIC card is in the source ESXi and target ESXi is configured is enabled for vmotion service inorder to transfer the cpu,memory and kvm instructions during live vm transfer.This required data is transferred over dedicated line over Layer2.
And about the data to be transferred, it is only the memory (in execution) and cpu current instructions to be passed to the the target host to resume the VM on target ESXi inorder to retain it's state or current work after moved.
The copy of these data are not done in single short, it is done by small portion. For memory copy, memory bit map is used to transfer data part by part. Only when almost data is transferred and when nothing much is pending to transfer, vmware stun the VM on original ESXi and resume it on target ESXi by togetther copying the last changed bits on memory and cpu.
About the devices mapped to VM, the VMM component in the ESXi together with vmkernel, take care of the resource sharing. If any specific device mapped to the VM, if not available on the target ESXi, vmotion is not allowed and corresponding error is displayed to user.
Hi Noor,
When a vMotion task is initiated, the source host 'freezes' the VM and starts copying the VM RAM to the destination host. it happens very fast because nowadays transfer happens via teamed NICs backbones of 10s of Gbps, so moving few gigabytes of data from one to another extreme fast RAM module happens in a snap. Once done, the destination host handles the new VM and 'unfreeze' it back. All previous status of the VM will remain as is, as long as its storage is visible in both hosts.
The impact on the front end will most likely go unnoticed, applications usually 'wait' for their server if it did not response immediately, In worst case the end user's application may freeze for a few seconds and then response again normally, which happens all the time in any application.
Hope this will answer your question.
