I earn commissions when you sign up or buy through the affiliate links on this page, at no extra cost to you.

Homelab · Guide · By Mohammed Almuhanna · Updated June 2026

RAID Rebuild Risk and UREs

When a disk in a single-parity array dies, the rebuild has to read every block on every surviving disk to reconstruct the one you lost. Bigger disks and more of them mean more data the rebuild has to read without a single hiccup, over more hours, with no redundancy left to fall back on. I run SHR-1 on my own Synology, five Seagate IronWolf Pro 12TB on single-drive redundancy, and I sleep fine. The only reason I sleep fine is that I keep a real offsite backup. This page is exactly why that backup is not optional.

What a URE is

Consumer drives are rated for roughly one unrecoverable read error per 10^14 bits read. For most of a drive's life that number is meaningless to you. During a full rebuild it stops being meaningless, because you are reading so much data that the odds of hitting one bad sector go from "never think about it" to "this can actually happen to me". On a traditional single-parity array (hardware or mdadm RAID 5), one URE in the wrong place during a rebuild fails the whole rebuild and takes the array with it. ZFS RAIDZ1 handles it better: it cannot reconstruct the blocks under that bad sector, so it loses those specific files and tells you which, but the pool stays online. Either way you have lost data you cannot get back without a backup, and a second drive failure mid-rebuild ends the array on both.

Why drive size made this worse

A 2 TB rebuild barely reads anything. A 16 TB rebuild reads eight times as much, runs for far longer, and the entire time the array has zero redundancy left. More data to read and a longer window with no safety net both push the odds the wrong way. That is the real reason single parity (RAID 5, RAIDZ1) gets waved off for large modern drives, and it is a fair warning, not marketing.

The fix

If the array is large and you do not have a backup you trust, move to double parity, RAID 6 or RAIDZ2, so a second failure or a URE during the rebuild does not end you. I do not run double parity myself, and I am not nervous about it, because single parity plus a real offsite copy is a setup I stand behind. Without that copy I would not run single parity on big drives. The calculator flags it when your config lands in the risky single-parity-on-large-drives zone. See also ZFS RAIDZ explained.

Common questions

What is an unrecoverable read error (URE)?

A URE is a sector the drive just cannot read back correctly, even after it retries. Consumer drives are rated for roughly one URE per 10^14 bits read. During a full rebuild you read so much data that the odds of hitting one go from negligible to something that can actually happen, and a URE mid-rebuild costs you the whole array on RAID 5, or just the files it sits under on RAIDZ1. A second drive failure mid-rebuild ends either one.

Why is single parity riskier on bigger drives?

Rebuilding a failed drive reads every block on every surviving drive. A 16 TB rebuild reads eight times as much data as a 2 TB rebuild, over many more hours, all while the array has zero redundancy left. More data to read and a longer window with no safety net both raise the chance that a URE or a second failure costs you data during the rebuild.

Should I use RAID 5 or RAIDZ1 with large drives?

On large modern drives without a backup, no. Move to double parity, RAID 6 or RAIDZ2, so a URE or a second failure during the rebuild is survivable. Single parity is fine on smaller drives, or on a backed-up home array. That is exactly what I run, SHR-1 on my Synology with a real offsite copy behind it, and the backup is the part that makes single parity reasonable.

Does double parity remove the need for backups?

No. Double parity makes the rebuild window survivable, but RAID of any kind is not a backup. You still need a real offsite copy for the things RAID cannot save you from, like a deleted file, ransomware, or the whole machine dying.