Datastore out of space - disk too big - bit of a mess

[Bitstreams]

I'm supporting a free ESXI installation on an HP server running a Windows Server VM. A while back we added disks to the hardware array and expanded it to 9TB. We then expanded the main data disk (4TB) on the server (under vSphere) to 9TB but this failed. We then learnt that there was a 4TB (ish) limit under the version we're running. So the vmdk file thinks it's 9TB and the server thinks it's 4TB and wont expand. We've been running this way for a few months.

We had a power fail last week which threw out the CBT tracking on the backups and last night I carried out a CBT reset. The backup started again and overnight the datastore ran out of space. I have cleared a checkpoint and some logs and now the server is up again, but obviously I'm concerned that I will have a problem during the backup tonight.

So what to do? There isn't enough space on the host to create a new datastore, but I can create a datastore with 10TB on a local NAS. So one option is to move the vmdk file to the new datastore on the NAS, which will stop the current datastore running out of space, but I'll still have a vmdk file that is 9TB but only offers 4TB.

Or I could create a new datastore on the NAS and another 4TB disk on it, attach it to the VM, move all the data under Windows, delete the wrong-sized vmdk file and create a new one and then move all the data back. 

The host is Esxi 6.0 

I have asked the client if they would like us to provide a new server but they're not keen. Any ideas?

[a_p_]

If I understand this correctly, you now have a Windows VM with a 9TB thick provisioned virtual disk, and 5 TB as unused from within the guest OS?

In this case - and with he NAS available - you may consider to migrate the VM to the NAS, and then migrate it back to the host's VMFS datastore as "Thin provisioned". This way the 5 TB unused disk space would still show up in the VM, but would not consume any disk space on the datastore.

Moving/copying the guest data to a new virtual as you mentioned disk is also an option, but you have to take care of the configured file system permissions, shares. However, this approach may (depending on the Windows version) allow you to configure a GPT partition table on the new virtual disk, which will let you create a larger partition, and increase its size it in the future if required.

André

[Bitstreams]

"If I understand this correctly, you now have a Windows VM with a 9TB thick provisioned virtual disk, and 5 TB as unused from within the guest OS?"

Not quite, the guest OS sees it as a 4TB disk (Disk Manager under Windows shows it as having just 12mb of space to grow). 

So the disk under Windows is 4TB, the disk under VSphere is 4TB, but the VMDK file is 9TB

[ChrisFD2]

Do as @a_p_ suggested and Storage vMotion it to the NAS as thin provisioned then do the same in reverse. It will then only consume data that the guest OS has actually written to its disk. It's the only supported way to change the provisioning type.

[a_p_]

>>> We then learnt that there was a 4TB (ish) limit under the version we're running.
Out of curiosity. Which "version" are you referring to? Windows, storage, VMFS, ...?

André