In Certain Fibre Channel Environments, VMotion Does Not Succeed and the Virtual Machine Remains Running on the Source Machine (2265)
This might occur with QLogic cards in one of the following situations:
- A particular VMware ESX Server machine is in a crashed state
- A disk array with a bad port is powered on but does not allow any Fibre Channel logins.
To verify that you are having the most common version of this problem, on the VMotion destination VMware ESX Server machine, check the VMX log file. The log file shows:
vmx| MigrateStateUpdate: Transitioning from state 9 to 10.vmx| Unable to initialize swap file /vmfs/volumes/43e11658-vmx| Module Migrate power on failed.
This indicates that a fabric event (host adds, switch adds, rezoning, and so on) has occurred during VMotion. At the same time, there is a bad port on a disk array or another machine on the fabric that is powered on but not fully operational. Due to issues with the QLogic driver on the healthy running machine, I/O to its storage is delayed during its rediscovery response to the fabric event and VMotion fails. This problem might also occur if any one machine or Fibre Channel device somewhere in your Fibre Channel fabric has a partially non-responsive failure mode.
VMotion typically fails because the destination virtual machine cannot power on. In the most common case, the swap file of the destination virtual machine cannot be opened. This results only in a VMotion failure. The virtual machine continues to execute on its source server. There is no loss of data or availability of the virtual machine.