ReFS fast-clone and synthetic fulls and free disk space

[FireExit]

Hi,

With a local repository of 30TB, it has <1TB free. Windows Server 2019 with ReFS. The properties of all the files and folders show 35TB of total data, which makes sense as there are some synthetic fulls already on there so block cloning seems to be working.

One job is trying to make a new synthetic full from a forward-incremental chain, the full will be around 3TB. It uses up the free disk space and fails. My understanding is that with fast-clone, the synthetic full should be made of references to existing blocks in the original full VBK and the chain of diffs, is that right? If so, why does it take hundreds of GB of space and fail?

[FireExit]

Oops I don't have a support ID because I wouldn't bother support about this, so I guess it will be deleted; I can't delete it myself anymore.

[Gostev]

Hi,

You're correct: the creation of synthetic full on ReFS should not consume ANY physical disk space except a few MB for metadata, because 100% of its blocks are reused from existing backup files. That is despite the resulting synthetic full file size will show, say, 3TB (this is logical file size, not physical disk space it occupied in the process of creation).

But you're correct in that you should contact support to understand why it's not the case for you and why fast cloning was not leveraged for this particular synthetic full.

Thanks

[FireExit]

Thank you Gostev, I have raised support case 06349468 about it.