ReFS / scale out / migrate to new array

[GabesVirtualWorld]

Hi
Customer has a 64TB physical RDM (Win2019 with ReFS) on which his long-term backup resides. Now we've bought a new storage array and we have to move his data to this new array. As we're moving from EMC to Pure, we have no migration tools at the array level. When migrating the data we also want to move from a physical RDM to VMDK (max 62TB). We're now looking at a migration scenario.

If we present a 62TB VMDK to the Win2019 VM with ReFS on it, we have to carefully copy from source to target, since ReFS will be needing time to dedupe and compress that data. The 64TB is 48TB filled, but decompressed this is 400TB of data which isn't going to fit if we just blindly copy everything. With smaller copy jobs we have to wait in between and keep the backup job paused to prevent any changes to the VBR files. Since this is time consuming for us, we were thinking for a different scenario.

Would it be possible to use scale-out for this? What if we create a scale-out repository for this VMDK and add it to the backup target, also add the current backup target to the scale-out and then put the scale-out with current backup into maintenance mode. Would VEEAM then copy the backup to the new volume? Would VEEAM know to "wait" for ReFS to finish with dedupe and compress before adding more files?

Other tips to migrate are also welcome.

[HannesK]

Hello,
If it really needs to be a VM (I always recommend physical repositories for performance and chicken / egg issues) I would attach the Pure volume directly with 64 or more TB to the VM. With iSCSI it's easy, with FC it should also work with NPIV.

Then I would copy all data with a block copy tool. dd is what I know that works fine.

Maintenance mode and evacuate would also inflate everything. Not an option as far as I see. Depending on on the retention, aging out the data with sealed mode would be an alternative.

Best regards,
Hannes

[GabesVirtualWorld]

Hmmm hadn't thought of dd. Good tip.
Just need to chew a bit on how I would get it into a VMDK because that would give me more freedom when moving to a new array in a few years,

Retention isn't an option since we'd like to have this done by October at last.

[HannesK]

in a few years there are hopefully be better solutions migrating block cloned backups (we are investigating that topic)