Bonus Tip – Unrecoverable Errors in Raid 5

One often encounters the claim that RAID 5 is supposedly dead because declared unrecoverable read error rate (URE) of high capacity hard drives 1e-15 is too high a RAID5 array rebuild is supposedly likely to fail before completion.

It turns out, this worry was based mostly on naive misunderstanding of the failure modes of RAID controllers. URE usually do not cause complete rebuild failure, only a failed stripe.

In reality, you can RAID 5 just fine with howsoever big enterprise disks, if 1) you use just a few of them in the RAID set, say 3 disks; 2) you can tolerate occasional very rare failed stripe (which can cause few files being corrupted). This can be the case for multi-day backups, for example.

In case you want to be on the safe side regarding failed stripes using RAID5 + high capacity disks, you can create multiple smaller arrays on the RAID5 disk set. Say, with 3x18TB disks, total available space is 36TB, but if you partition it into 6x6TB arrays , then any failure during one array rebuild can affect only 1/6 of the data.