Strange upgrade issue - VM corruption

[Dr_Virt]

Had a single host test environment running ESXi 5.5. Decided to upgrade to 6 last night. Shut down all non-critical VMs and paused the critical ones (Domain Controller, File Server, Logging Server). Installed vCenter and Update Manager 6. Registered ESXi host with vCenter. Scanned and remediated without issue. Verified all was in good working order and began to restart the VMs.

First, tried to resume the Domain Controller.  Received a failure notice - "The specified device is not a valid physical disk." I verified the VMDK file was correct. I removed from inventory and then added it back in. I removed the disk from the VM and added it back in. All failed.  I eventually deleted the VM while saving the flat file off to the side. I created a new VM with the same configuration, updated the MAC address, and placed the backup flat file in the new VM folder structure. Fired up the VM and it worked perfectly.

I had to repeat this process for each paused VM.

The shutdown VMs fired up without issue.

Any experience anything similar where a paused VM has is VMDK header file corrupted during upgrade when suspended?

[CoolRam]

will try this scenario and revert back. in the cluster with multiple host never seen such issue.

[a_p_]

Any experience anything similar where a paused VM has is VMDK header file corrupted during upgrade when suspended?

I can't provide specific technical details or reasons for this, but you should always cleanly shut down VMs if you are going to upgrade or patch a host. With pausing/suspending a VM all runtime information is stored in a .vmss file, which is used when you resume the VM. If issues like this happen with suspended VMs, it's often only helps to "pull the VM's power plug", i.e. delete the VM's .vmss file.

André

[Dr_Virt]

Little more information upon further digging. Somehow, during the upgrade, the datastore for these VMs seems to be have been messed up. I had to evacuate the datastore and reformat it.

Continued testing...