Can someone explain why DPM thinks it has to transfer 6.4TB of “changed” data for a Hyper-v host recovery point? From what I can tell, only a tiny fraction of that data is needed/used to create the recovery point. For more information, read on.
I have two protected member servers, comparably configured, at two remote sites. The servers are running Windows Server 2012 R2 Standard (as Hyper-V Host) and are hosting Windows Server 2012 R2 Standard (guest w/dedup enabled). The only real difference between these sites are the size of the .vhdx files:
Site A: has a protected member that has a 11TB DATA.vhdx file
Site B: has a protected member that has a 16TB DATA.vhdx file
These systems are being protected with DPM 2016 UR4 (v5.0.342.0). The guest is protected by an off-site DPM server connected via a 100 mbps WAN link. While the host is protected by an on-site DPM server utilizing a 1GB connection.
DPM is supposed to query the NTFS Change Journal and transfer what has changed from the last sync job. This appears to be true for the off-site DPM server protecting the guest over the WAN connection, with both sites transferring a small number of GB every night and finishing the recovery point within an hour.
However, when using the on-site DPM server to protect the volume that contains the .vhdx, it doesn’t appear to work correctly (read as, efficiently). As an example, Site A is doing data transfers of 650GB to 1.5TB every night and the situation is even worst for Site B. It is transferring 6.2TB to 6.4TB every night, taking 17hours to complete a single backup.
Move over, it is clear that not all of the data being transferred is stored in the Protection Group’s Recovery point. I know this because, I have 31 recovery points with only 1.8TB (in total) of data being used by the on-site DPM server.
Can anyone explain this behavior?
UPDATE #1:
Instead of protecting the volume that contains the .vhdx, I am now protecting via the Hyper-v method. There is no change in the behavior. My next move is to disable Dedup and Defrag process and monitor for a few days.
Site A Host transfers:
Site B Host transfers: