I make Backups from Snapshots of zfs Datasets. Due to the nature of zfs i expect the Data of the Snapshot is correct even in case of a disk failure.
The storage location of the restic repository is unconfident. But if there would be an error i expect restic to make a report.
So i assume a restored backup is either guaranteed to be correct or fails or at least restic will show an error message in the restore process.
How is this guarantee influenced by the cache? What happens if it is stored on unreliable media outside the zfs pool and there occur an unreported failure on the device?
Are there still checksums involved so restic will report this failure?
Simply said:
Is it safe to have the restic cache on unreliable media?
Ok, i have done a bit investigation myself in the meantime.
i flipped some bits in a cachefile to simulate an error. Surprisingly nothings happens. Than i noticed that restic silently overwrites the damaged cachefile with correct data from repository. At second thought this behaviour looks acceptable but some warning might be suitable. At least it could be a hint for hardware malfuction.
Not printing a warning is intentional. Restic does not enforce that the file in the cache are actually fully written to disk. So if your system or restic crashes at the wrong point in time, the cache (and only the cache! The data in the repository is not affected by these issues) might contain incomplete files. So most warnings about broken cache files would turn out to be false positives.