We have a customer who is trying to use MTV to migrate from a cluster running CNV 4.16.6 to 4.17.4.

Source Cluster:
OCP 4.16.32 
CNV 4.16.6(virtualization operator version 4.16.6) and the Destination 

Cluster:
OCP 4.17.4 
CNV 4.17.4

During the migration attempt, they are seeing behavior similar to  https://issues.redhat.com/browse/CNV-45016 and https://access.redhat.com/solutions/7081671  Its unclear whether that should be in 4.17.4.While attempting to migrate we get:
       error:
          phase: CopyDisks
          reasons:
          - 'Unable to process data: Unable to transfer source data to scratch space:
            unable to write to file: write /scratch/tmpimage: no space left on device'

We see the source device is sized 53687091200 and when we create the DataVolume on the import, we are requesting 53687091200.

During the migration, it looks like we are using the dell-powerstore storage class, at least for our scratch space.  This is filesystem based, so it seems we aren't creating that with enough overhead just like the KCS.  

The customer stated "What I'm finding interesting is that the data migration process for these VMs seems to create a scratch space with 5.5% overhead, but if I watch the amount of used space during the process, I believe it dies before it uses that 5.5%. For example, I watched a migration of a 50G disk where it had ~53G of scratch space, but the migration quit as soon as it reached 50G."