Earlier this week I was working on an ESXi VM host that had a huge (4TB) amount of local storage. Ideally, I would have liked to provision a single datastore on the server, but as I found out, this isn’t possible; ESXi supports datastore sizes up to 2TB, but no higher. The odd thing about that limit is that ESXi can see all 4TB of space, and will let you provision a datastore that supposedly has 4TB of capacity, but in reality only has about half that. Since my volume had approximately 3.7TB of space, I was winding up with a datastore that had 1.6TB of capacity.
I decided instead to try and create two datastores on the same virtual disk. This, again, met with failure, because you can only create one datastore on a single LUN (in this case, a virtual disk). So I ventured back into the RAID controller and split the virtual disk in half (and reinstalled ESXi) and was then able to provision two datastores of equal size.
The odd thing I found about this was that when creating my datastores on each of my 1.8TB virtual disks, ESXi reported the entire capacity available. Whatever happened to the 200-400GB of capacity lost when I first attempted to do this is beyond me. I’m also not entirely sure why there’s a 2TB limit on datastores in the first place; I can understand the 2TB limit on individual VMDK files, but why the datastores are capped like that is beyond me, although I’m sure there’s a computer scientist out there with an explanation for me.
Leave a Reply