vdi-export failure causes orphaned/RO VDI

A vdi-export command failed because of lack of space on the target export area. This left an orphan snapshot VDI on a local SR that I could no longer delete with via XenCenter or via xe. I could not forget or destroy it as it claimed the VDI was in use by some other operation. Checking for any active tasks showed none, and even getting rid of the parent VM, rebooting all servers in the pool – nothing helped. I kept getting a "VDI is in use by some other operation" error. This should not happen, of course. An attempt to get rid of the VHD entry also failed. It told me the metadata DB is RO so I cannot do an lvm destroy on it. The workaround was to modify the LVM configuration file /etc/lvm/lvm.conf temporarily to allow that LVM entry to be deleted by setting "metadata_read_only = 0" after which I could destroy the LVM. It still took a subsequent reboot of the XenServer for the VDI entry to finally disappear and the space be released.

Failed operations should not result in such a state of breaking a VDI like this and appropriate cleanup steps should be taken to leave the SR in a "cleaner" state.