I have a Time Machine sparsebundle backup on an NTFS volume on a USB drive. When I wanted to restore from it I connected it and the volume was mounted as read-only, but I was unable to mount the sparsebundle and migration assistant didn't offer it as a source to restore from.
After copying the sparsebundle to a read-write HFS+ volume (on a USB drive) I was offered the option to restore from it, leading to my question.
What are the requirements to be able to restore from a sparsebundle? Is it that the containing volume is writable, or something else?
Background
I'm doing my backups over SMB, so I want a file system that's journalled on Linux (i.e. I can't use HFS+) and from which I can restore when using it as a USB drive (so far not NTFS). I'm aware of paid drivers for NTFS on macOS, but I'm initially looking for a free solution for Mojave.
The ntfs-3g project seemed completely dead judging by its website, and the install/compile instructions seemed lacking and outdated (packages from 2015 as opposed to the latest release in 2017).
Best Answer
I have experienced problems using sparsebundles on APFS volumes, whereby writing ~50gb lead to a backup that had incorrect checksums (determined by running tmutil verifychecksums and seeing a failure).
I then switch to using sparsebundles on HFS+ volumes, and no more problems.
Worth mentioning that you are probably using sparsebundles that are themselves HFS+ formatted, if you are using them for Time Machine.
If I had to take a guess, I would say it probably has more to do with the filesystem that the sparsebundle lives on (not the filesystem internal to the sparsebundle)... and less to do with the read/write vs read-only issue. You can test this simply... try to mount your sparsebundle while it lives on a read-only HFS+ volume... does that work? (can you mount?)
Yes, I realize that the filesystem where the sparsebundle lives should not matter, but I have myself seen that sometimes it does. One reason is quite simply that filesystems themselves have bugs. APFS, NTFS, and HFS are no exception to this.
As some further detail, I noticed that verifying checksums for a sparsebundle on an APFS volume took significantly longer than performing the same operation on a similar sparsebundle on an HFS+ volume.