Currently there is no easy way of fixing this, we have an issue for implementing this tough.
Probably the simplest way of re-sync the DB, is to shutdown the VM manually in the hypervisor (virsh -c qemu:///system destroy one-<vm_id>) and after that, just resume the VM again so OpenNebula will resume the VM in the expected host.
Note that this will only work if you have shared storage, if not you will need to copy the information of the system datastore.