Unraid Parity Rebuild + Parity Check Estimator
This estimator works out how long an Unraid data-disk rebuild or parity build takes, how long the array runs with reduced protection once you add the wait for a replacement, what protection is left while it runs, and what to check before you start. For a routine parity check it shows the same estimate as our parity check calculator.
When a disk drops out of an Unraid array, parity is what lets its contents be rebuilt. Until the rebuild finishes, single parity has no redundancy left and dual parity has one disk to spare. The length of that window, and what you check before starting, are the parts you control.
Do not treat this as an ordinary rebuild. Investigate first.
A rebuild or parity build reads every disk hard for hours. With another disk already showing problems, that is when it is most likely to fail too.
- Read the warning disk's SMART report and the system log; check its cable, power and controller port.
- If critical data is not backed up and the array is readable, consider copying the most important files first, knowing that heavy reading also stresses a weak disk.
- If more than one disk is in trouble, get recovery advice before starting. Our guide to NAS data recovery costs in Australia covers your options.
Estimated rebuild time
…
Protection while it runs
Degraded-window risk, from independent disk failures
Before you start
- A disabled disk is not always a dead one. A failed/disabled disk (red X) is not always broken (bad connection, power issue, glitch); a disabled disk judged healthy can be rebuilt onto itself instead of replaced. (Unraid source, applies to Current docs (no version stated), checked 2026-10-06) Check its SMART report, cable and power before you buy a replacement.
- Check the emulated disk. While a disk is disabled, Unraid emulates it from parity plus the other working drives, and its data stays accessible. (Unraid source, applies to Current docs (no version stated), checked 2026-10-06) Before rebuilding, confirm it mounts and its files look right: a rebuild reproduces what parity holds, it does not repair a damaged filesystem.
- Check every other disk. Look at SMART on each, especially pending and reallocated sectors, and reseat cables.
- Backups. A backup reduces what you lose if this goes wrong; it does not make the rebuild safer. If your critical data is not backed up, consider copying it first.
- Replacement size. Each parity disk must be at least as large as the largest data disk (dual parity: both, but they may differ in size from each other). (Unraid source, applies to Current docs (no version stated), checked 2026-10-06) If your replacement is larger than the parity disk: A parity swap procedure exists for replacing a data disk with one larger than the current parity disk. (Unraid source, applies to Current docs (no version stated), checked 2026-10-06) That is a multi-stage job, not one rebuild, and this tool does not time it.
Routine parity check at the same speed
How we calculate
- Operation time = disk size ÷ average progress speed, in decimal units (1 TB = 1,000,000 MB): 8 TB at 160 MB/s is
8,000,000 ÷ 160 ÷ 3,600= 13.9 hours. A data rebuild is timed on the size of the disk being rebuilt; a parity build on the parity disk's size. If your replacement is larger than the disk it replaces, enter the larger size: the result then errs on the long side. - Average progress speed is set by the slowest disk, the controller and other activity, and by how fast the target disk can take a long sustained write. Without your own figure we use an assumed range (table below), scaled for activity. An SMR target gets a deliberately wide range, not a fixed penalty, because sustained SMR writes vary by model.
- Degraded window = wait before the rebuild starts + rebuild time (slow end). The wait is often longer than the rebuild.
- Degraded-window risk is the same calculation as our Drive Failure Risk Estimator: each other disk fails inside the window with probability
1 − (1 − AFR)hours ÷ 8,760, independently. With single parity one more failure loses data; with dual parity it takes two. While a single parity disk is being built the array has no redundancy, so any data disk failing counts. Real failures cluster (same batch, same heat, the rebuild load), so the figure is a simplified planning estimate, not a forecast, and clustering can make the real risk higher. - Dual parity adds no time here: the rebuild still writes one disk at the pace of the slowest disk it reads.
- Read errors are not priced as a whole-disk failure. NTK's reading, not an Unraid statement: an unreadable sector with no redundancy left affects the data in that place, typically part of one file, not the whole rebuild.
- A health warning on another disk replaces the risk band with a stop: no figure describes a rebuild onto an array that is already failing.
| Slowest disks | Rebuild or parity build | Parity check (reads) |
|---|---|---|
| 7200 rpm class hard drives | 120 to 160 | 120 to 160 |
| 5400 rpm class or older hard drives | 90 to 120 | 90 to 120 |
| An SMR hard drive in the array | 40 to 110 | 90 to 120 |
| SATA SSDs | 150 to 300 | 400 to 550 |
| Activity | Speed left |
|---|---|
| Idle: nothing else running | 100% to 100% |
| Light: some streaming or reads | 85% to 95% |
| Moderate: active use, some writes | 65% to 85% |
| Heavy: big copies, VMs or containers busy | 40% to 70% |
Worked example
Four disks (three data, one parity), single parity, rebuilding an 8 TB data disk, 7200 rpm class drives, nothing else running, a spare on the shelf (about 12 hours to swap), drives at 0.87% AFR.
- Rebuild: 8 TB ÷ 160 MB/s = 14 h to 8 TB ÷ 120 MB/s = 19 h.
- Degraded window: 12 + 18.5 hours = about 31 h.
- Chance one of the other three disks fails in that window: 0.0091%, band Low. With dual parity (five disks), two would have to fail: under 0.0001%.
- The same rebuild onto a 16 TB SMR disk with no observed speed: 40 h to 4.6 days. The width is the point: enter the speed you actually see once it starts.
Rebuild, parity build and parity check compared
| Operation | Reads | Writes | Protection while it runs |
|---|---|---|---|
| Data disk rebuild | Parity and every other data disk | The replacement data disk | Single parity: none left. Dual parity: one more disk can fail. |
| Parity build (replace or upgrade) | Every data disk | The parity disk | Only parity disk: none until it finishes. One of two: the other still protects. |
| Adding a second parity disk | Every data disk | The new parity disk | The first parity disk keeps protecting the array. |
| Parity check | Every data and parity disk | Parity, only where a correcting check finds a mismatch | Full protection; it checks it. |
A parity check is timed in the Parity Check Calculator, which shares this page's speed assumptions and formula.
Unraid facts this page uses
Each fact below is taken from Unraid's own documentation or release notes, re-found at the source on the date shown. Latest stable Unraid release when checked: 7.3.3.
| Fact | Applies to | Source |
|---|---|---|
| What Unraid does by default | ||
| Latest stable Unraid OS release is 7.3.3 (docs release notes: 7.3.3 dated 2026-10-08). | Unraid OS stable, as of 2026-10-12 (7.4.0-rc.1 dated 2026-10-08 is a pre-release) | Unraid source, checked 2026-10-12 |
| After an unsafe shutdown Unraid automatically starts a parity check using the mode set under Settings > Scheduler > Parity Check (non-correcting by default); it is not automatically a correcting check. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| What the Unraid documentation advises | ||
| During a disk rebuild Maintenance Mode can optionally be selected to speed up the rebuild; the array is not accessible while it runs. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| For array devices with sustained writes, particularly parity, the docs prefer CMR over drive-managed SMR; SMR can have significantly lower sustained write performance during parity syncs and rebuilds. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| How the array behaves | ||
| A failed/disabled disk (red X) is not always broken (bad connection, power issue, glitch); a disabled disk judged healthy can be rebuilt onto itself instead of replaced. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| While a disk is disabled, Unraid emulates it from parity plus the other working drives, and its data stays accessible. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| Each parity disk must be at least as large as the largest data disk (dual parity: both, but they may differ in size from each other). | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| A parity swap procedure exists for replacing a data disk with one larger than the current parity disk. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| Adding a parity disk works like adding a data disk: assign it and start the array; parity builds while files stay accessible, possibly slower. Either parity disk must be at least the largest data disk. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| Unraid does not stripe across drives: each file is stored on a single disk. | Current docs (no version stated) | Unraid source, checked 2026-10-06 |
| Unraid's parity-protected array holds up to 30 devices, 28 data and 2 parity, on an Unleashed or Lifetime licence; a Starter licence allows up to 6 devices including parity. | Current licence tiers (licensing FAQ, current docs) | Unraid source, checked 2026-10-14 |
Spares, warranty and recovery
The wait for a replacement counts as degraded time, and it is often longer than the rebuild. A spare on the shelf removes most of it; a warranty claim usually adds to it, because the replacement comes back after the failed drive goes in. Enter the wait you would really have above. A manufacturer warranty is the maker's own promise. Your first step for a faulty product is usually the retailer that sold it. The ACCC explains your rights under the Australian Consumer Law at accc.gov.au.
If a rebuild fails, restore from backup. Without one, professional recovery is possible in many cases; providers include Payam Data Recovery and Ontrack, and quotes vary by failure type. See NAS data recovery costs in Australia.
Frequently Asked Questions
How long does an Unraid disk rebuild take?
About the size of the disk being rebuilt divided by the speed the rebuild progresses at. At 120 to 160 MB/s an 8 TB disk takes about 14 h to 19 h. The slowest disk, the controller and other activity set the pace, so the best figure is the average speed of a past parity check or rebuild on your own server. Enter it as your observed speed.
Is a disabled disk a failed drive?
A failed/disabled disk (red X) is not always broken (bad connection, power issue, glitch); a disabled disk judged healthy can be rebuilt onto itself instead of replaced. Check the disk's SMART report, its cable and power, and the system log before you buy a replacement.
What should I check before starting a rebuild?
While a disk is disabled, Unraid emulates it from parity plus the other working drives, and its data stays accessible. Also check SMART on every other disk, reseat cables, and make sure no second disk is showing errors.
Should I rebuild in Maintenance Mode?
During a disk rebuild Maintenance Mode can optionally be selected to speed up the rebuild; the array is not accessible while it runs. This tool does not apply a fixed speed-up for it: it treats Maintenance Mode as an array with nothing else running.
What happens if another disk fails during the rebuild?
With single parity, one more disk failing before the rebuild finishes is more than parity can rebuild, and data can be lost. With dual parity, the array can lose one more disk and still rebuild. Unraid does not stripe across drives: each file is stored on a single disk. A backup is what covers a loss beyond parity.
Will one unreadable sector ruin the rebuild?
An unreadable sector on another disk is not the same as a second disk failing. NTK's reading: with no redundancy left it affects the data in that place, typically part of one file, not the whole rebuild. With dual parity there is still parity to rebuild that block. Regular parity checks find weak sectors while redundancy is still intact.
Can I rebuild onto an SMR drive?
For array devices with sustained writes, particularly parity, the docs prefer CMR over drive-managed SMR; SMR can have significantly lower sustained write performance during parity syncs and rebuilds. SMR drives can slow down badly under long sustained writes, which a rebuild is, by an amount that varies by model and how full the drive cache is. This tool uses a deliberately wide range for an SMR target instead of a fixed penalty, and a speed you have observed on your own server beats it.
Should I back up before a rebuild?
A backup reduces what you lose if the rebuild goes wrong; it does not make the rebuild itself safer. If your critical data is not backed up and the array is readable, consider copying the most important files first. If more than one disk is showing problems, get recovery advice before you start, because heavy reading can push a failing disk over.
Method last reviewed: 2026-10-14. Unraid facts last checked at source: 2026-10-14.