StoreOnce Poor Dedupe on source appliance, good dedupe on replicated appliance

[JCI_Chris_B]

I have two new StoreOnce VSA systems setup.  I am using a CIFS share on one side to write backups to, which is replicated to the other appliance off-site.

One of the backup files is an uncompressed SVN dump, which should have very favorable characteristics for deuplication.

What I am seeing is that on the source appliance, I am getting very poor dedupe results when this file is added daily to the appliance.  However, the replicated appliance is getting good dedupe ratios on the same dataset, as would be expected based on the data type.

Is this expected behavior?  Does StoreOnce only do an inline dedupe?  Or is there an additional process that will eventually perform some post-process deduplication on the dataset on the source appliance?

I have used the StoreVirtual VSA on the very first release in the past and don't remember seeing this sort of thing, at least to the extreme that I am seeing it now.  See attached for screenshots of the source and destination CIFS share dedupe ratios.

[JCI_Chris_B]

This turns out to be an issue when files are copied via CIFS using "restartable mode" (e.g. "/Z" in xcopy/robocopy).

If this flag is used, the file copy takes much longer and it appears the system does not do any kind of deduplication on the incoming data.  By changing the copy commands to not use restartable mode, dedeuplication happens normally and the copy performance is also vastly improved.

To reclaim the storage if files have been copied in restartable mode, moving the file off the CIFS share, then moving it back onto the share (without using restartable mode) properly deduplicates the file and reclaims the previously ballooned space.