This is a weird article. It points out that yes, the server OS itself needs a partition sure - but if the author is using Alpine because as they say it’s so small, why bother with unreliable and potentially catastrophic USB boot? To save a few megabytes of disk space? Crazy.
It's deffinately not the perfect file server. From using usb drives, 15 year old atom CPU, 2gb of ram and raid 5 for the drives... Just some of the reasons.
Granted, yes, the author should have chosen a more specific and less clickbaity title. But titles need to be short and pithy when dealing with the goldfish attention spans many people have.
So a better, more accurate title could be,
>“The perfect low-cost/no-cost file server utilizing old/upcycled and possibly obsolete hardware”
Except that doesn’t exactly roll off the tongue with aplomb.
Wasn't RAID 5 considered less than ideal because in the event of one disk failing, you're likely to get a failure during resilver due to the amount of strain put on the remaining disks during the process?
Plus, when rebuilding a failed RAID-5 disk, the system is engaging in mass calculations to rebuild that data from the “fingerprints” stored on the other drives.
Conversely, with RAID-10, we have mirrored drives (the 1 part of the RAID) which are then striped across the pairs that exist (the 0 part of the RAID). As such, IIRC rebuilds are a lot faster because it is a direct copy instead of a CPU-intensive rebuild of the data.
Plus, you have (technically) a n-drive failure capability, where n = the number of mirrored pairs of drives. Provided that you don’t have more than one drive per pair fail, of course. So for a RAID-10 of four drives, where you have a stripe across two mirrored pairs of drives, you could have up to two drives fail without experiencing data loss, so long as each failed drive was from a different mirrored pair.
Plus, RAID-10 is much faster than RAID-5, and much, much faster than RAID-6.
Yes, this is why RAID5 isn't advised for (most) modern arrays anymore.
The time it takes to rebuild a pool full of 10TB+ disks is going to be extremely long, and the chance of a second disk failing during that rebuild is much higher than smaller capacities like 2-3TB that would have been more common 10-15 years ago.
For thia reason, its often recommended to use RAID6 for new arrays and treat the second failable disk as strictly a spare for during a rebuild.