Four drives is a tricky size, and not generally one I would use with any kind of RAID other than a dual mirror, and even then I’d only use a dual mirror for performance critical applications where data integrity is less important (and honestly, for anything performance critical, I wouldn’t use a hard drive anyway)
Dual mirrors are not very space efficient though. If you want more capacity you can go with RAID5/RAIDz1 but that has generally not been recommended for quite some time now.
The 2007 article “Why RAID5 Stops working in 2009” was quite controversial at the time, but over time just about everything in it has been accepted as true by those in the know.
The primary reason is this:
As drives have become larger and larger, they have a frighteningly large number of bits. Each of those 28TB drives of yours have 2.24*10^14 bits.
If you ever have to resilver that pool, and it is close to being full, the resilver process will be reading three times that (the three remaining good drives) so, 6.72*10^14 bits.
Depending on your drive you will likely have a claimed URE (unreadable error rate) of either 10^14 or 10^15.
If you only have one redundant drive, if one drive fails or needs to be replaced, you are reading from the rest during your rebuild without any redundancy at all. In other words if one bit is either unreadable or flips, you have at least one bad file.
If your drives are rated at a URE of 10^14 this will happen 6.72 times during the average resilver for you. If your drives are rated at 10^15, it will happen an average of 0.672 times.
Ideally for data integrity, if the data matters to you, you want at least two redundant drives, such that you still have redundancy during a resilver, and replace any failed drive immediately as soon as it fails before another one does, so you won’t suffer from URE’s or flipped bits during that resilver.
It seems like you have already bought the drives, but honestly, if you had reached out before doing so, I would probably have recommended getting more smaller drives rather than four 28TB drives. A six drive RAID6/RAIDz2 pool would be perfect for this application and to get the same amount of available space as a dual mirror with the drives you did buy could have been done with six 14TB drives. Getting the same available space as a RAID5/RAIDz pool with the drives you did buy could have been achieved with six 21TB drives. This would have been better for data integrity long term.
That said, RAID should not be mistaken for a backup. It can help you with uptime (not having to restore from backup) but in and of itself it is not backup. If the data matters to you, you should always have a backup (preferably offsite) in addition to raid.
With something like Jellyfin - however - you might consider any data on the pool to be replaceable, in which case maybe it doesn’t matter.
If it is a mixed pool with some replaceable data (media files) and some data that is irreplaceable (family photos, personal files, etc.) I’d highly recommend keeping a regular offsite backup of the irreplaceable data.
RAID is great, but it only protects you against drive failure. It doesn’t protect you against lightning strikes, fires or floods that take out the whole system, pool corruption (which is granted pretty unlikely with ZFS, but has known to happen), errant keystrokes or the likes of malware/viruses/ransomware.
For these you need a true copy somewhere else you can restore from. All RAID should ever really be is for convenience and timeliness, to reduce the risk of you having to actually restore from backup. It should never be your only protective measure for critical data.
