Connect to the shell of the ESXi host system and browse to location of the datastore (/vmfs/volumes/{datastore name}) to which you want to identity orphoned virtual machine hard disks and invoke the following command. This will return all virtual machine hard disks to which the modified date is seven days in the past (-mtime +7) and return the filename pattern match (find -name “*flat.vmdk*) with the modified date timestamp (-exec stat -c “%n %y” {} \;).
find -iname "*flat.vmdk" -mtime +7 -exec stat -c "%n %y" {} \;
This approach assumes that if the virtual machine hard disk has not been touched in the specified timespan the virtual machine hard disk is orphoned. This may not be accurate, as the virtual machine may be in a powered off state for a period that exceeds the timespan and the virtual machine may be valid. The output should be similar to the below:
./vmfs/volumes/424b26ec-4294ea41/ubuntu-01/ubuntu-01-flat.vmdk 2017-01-29 10:52:03.000000000 ./vmfs/volumes/424b26ec-4294ea41/ubuntu-02/ubuntu-02-flat.vmdk 2017-01-29 10:17:46.000000000 ./vmfs/volumes/424b26ec-4294ea41/ubuntu03/ubuntu03-flat.vmdk 2017-01-29 11:04:18.000000000 ./vmfs/volumes/424b26ec-4294ea41/ubuntu04/ubuntu04-flat.vmdk 2017-01-29 10:17:49.000000000 ./vmfs/volumes/424b26ec-4294ea41/ubuntu05/ubuntu05-flat.vmdk 2017-01-29 10:52:03.000000000
An alternative method, is using RVTools which can identify orphoned virtual machines and hard disks (categorised as ‘zombie’) from the inventory scan of a vCenter Server System or ESXi Host System. On completion of the inventory scan, the vHealth tab will display any virtual machines and hard disks identified and may be exported for reference.
Deans Blog 