We were in a process of moving 32TB of cifs data. The customer had a setup of DFS-R with all data on vhdx drives on the Netapps with 8.3.1p2.
Hyper-v was communication smb3 to connect to the datastores. DFS-R had many issues with the replication and also backup was suffering.
The redesign was a direct CIFS share from a Netapp filer with snapmirror for DR and snapvault for the backup.
In the initial phase we copied 2x 5TB to the new cifs shares. After 4 days of robocopy we had our first baseline. When we started our incremental update it did run for 4 days again. When we checked the huge logs it looked like not a single file was skipped, even from users that left the company.
A test showed us that the size from some source files differed with the size from the destination files.(pictures/movies) We tried changing all robocopy and netapp parameters… nothing helped. The problem was caused by ODX. This parameter is designed to improve network copy.
We couldn’t disable the parameter on our SVM, it would impact our Hyper-V performance. So we created a separate SVM and disabled ODX.
After weeks of chewing and testing… I would love to share this to prevent other people wasting their time.
“>vserver cifs options modify -vserver [name] -copy-offload-enabled false”