Independent Australian Storage & Infrastructure Authority
RAID Rebuild Time + Risk Estimator
This estimator works out how long a RAID rebuild takes for your drive size and NAS, how long the array runs degraded once you add the wait for a replacement, and the chance another needed drive fails before it finishes.
When a drive fails, the array keeps running without its full redundancy until the replacement is in and rebuilt. The length of that window is what you can actually change: a spare on hand, a quieter array and dual parity each shrink a different part of the risk.
Rebuild time = capacity written to the replacement drive ÷ effective rebuild rate. Units are decimal: 1 TB = 1,000,000 MB, so 8 TB at 100 MB/s is 8,000,000 ÷ 100 ÷ 3,600 = 22.2 hours.
Effective rebuild rate means progress on the new drive, in MB/s. The speed profiles are assumptions for typical home and small-office NAS units, shown as ranges because CPU, drives, RAID type, filesystem and workload all change the real rate. The range on the result runs from the fast end to the slow end of the profile.
Used-space repair applies only when you choose Synology DSM Fast Repair with a RAID type Synology lists for it (RAID 1, 5, 6, 10, SHR-1, SHR-2). Synology documents that Fast Repair skips unused space, is on by default, and switches to Regular Repair when the pool is more than 80% full. It also documents that Fast Repair needs a storage pool with volumes created on it, and that the repair of block-level LUNs (supported on DSM 6.1 and earlier, still working after an update) cannot be accelerated (Synology Knowledge Center, checked 2026-10-14). If you are not sure Fast Repair applies to your pool, or it has been turned off, use the full rebuild time. Every other case is a full rebuild.
Degraded window = wait for the replacement + rebuild time (slow end).
Rebuild-window risk is the same calculation as the Drive Failure Risk Estimator: with each remaining drive failing in the window with probability p = 1 − (1 − AFR)hours ÷ 8,760, the chance that enough of the drives that matter fail to lose data. Drives are assumed to fail independently, so the figure is a simplified planning estimate, not a forecast. Real failures can cluster (same batch, the rebuild's own load), which can make the real risk higher.
Data read = the other drives in a parity array (or the one mirror partner) read in full, or their used space under Fast Repair. It is reported for read errors and never divided by the rebuild rate.
Read errors at spec = bits read ÷ the vendor's rating. A rating of 1 per 1014 bits is a worst-case specification, so the result is what the drives could meet if they performed exactly at that limit, not a measured chance of failure.
Worked example
A 4-drive RAID 5 of 8 TB drives on a typical home NAS (50 to 100 MB/s), full rebuild, spare drive on the shelf (12 hours to swap), drives at 0.87% AFR.
Rebuild time: 8 TB ÷ 100 MB/s = 22 h to 8 TB ÷ 50 MB/s = 44 h.
Degraded window at the slow end: 12 + 44.4 hours = 2.4 days.
Chance one of the other three drives fails in that window: 0.017%, band Low.
Data read from the three survivors: 24 TB. At a rating of 1 per 1014 bits that is about 1.9 unreadable sectors if the drives performed exactly at their worst-case rating; at 1 per 1015, about 0.19. Real drives usually do much better than their rating.
The same array on Synology with Fast Repair, half full, at 100 MB/s: about 11 h, because only used space is rebuilt.
Rebuild time by drive size
Full rebuild times from the same formula as the estimator, at four effective rebuild rates (MB/s columns), plus a Synology Fast Repair estimate for a pool half full at 100 MB/s. Use the estimator above for your own RAID type, wait and speed.
Drive size
30 MB/s
50 MB/s
100 MB/s
150 MB/s
Fast Repair*
2 TB
19 h
11 h
5.6 h
3.7 h
2.8 h
3 TB
28 h
17 h
8.3 h
5.6 h
4.2 h
4 TB
37 h
22 h
11 h
7.4 h
5.6 h
6 TB
2.3 days
33 h
17 h
11 h
8.3 h
8 TB
3.1 days
44 h
22 h
15 h
11 h
10 TB
3.9 days
2.3 days
28 h
19 h
14 h
12 TB
4.6 days
2.8 days
33 h
22 h
17 h
14 TB
5.4 days
3.2 days
39 h
26 h
19 h
16 TB
6.2 days
3.7 days
44 h
30 h
22 h
18 TB
6.9 days
4.2 days
2.1 days
33 h
25 h
20 TB
7.7 days
4.6 days
2.3 days
37 h
28 h
22 TB
8.5 days
5.1 days
2.5 days
41 h
31 h
24 TB
9.3 days
5.6 days
2.8 days
44 h
33 h
26 TB
10 days
6 days
3 days
2 days
36 h
28 TB
10.8 days
6.5 days
3.2 days
2.2 days
39 h
30 TB
11.6 days
6.9 days
3.5 days
2.3 days
42 h
32 TB
12.3 days
7.4 days
3.7 days
2.5 days
44 h
* Synology DSM Fast Repair only, on supported RAID types, on a storage pool with volumes created on it (not block-level LUNs from DSM 6.1 and earlier), with Fast Repair left on, and only while the pool is 80% full or less. Other platforms rebuild every block. Times are the rebuild alone: add the wait for a replacement drive.
What to do during a rebuild
Check your backup is current before the rebuild finishes. If anything else goes wrong, it becomes your only copy.
Reduce load. Pause media transcoding, backup jobs and large copies. Every competing read or write slows the rebuild and lengthens the window.
Keep the power stable. A UPS stops a power cut from interrupting the rebuild.
Watch the other drives. Look at the health status and SMART readings your NAS reports for the remaining drives. Rising reallocated or pending sector counts are a warning sign. A clean reading is not a guarantee: many failures come without warning.
Order a replacement straight away if you had no spare, because the wait counts as degraded time.
Single or dual parity, spares and rebuild time
There is no drive size at which RAID 5 stops being reasonable for everyone. Longer rebuilds, more drives, same-age drives, a long wait for a replacement and a missing or untested backup all make dual parity (RAID 6, SHR-2, RAIDZ2) worth its extra drive. A hot spare removes the replacement wait but adds no redundancy; a cold spare removes most of the same wait without using a bay; dual parity protects you during the window itself. Change the inputs above to see which part of your window each one removes.
If the rebuild fails
Restore from your backup. If there is no usable backup, professional RAID recovery is possible in many cases. Australian providers include Payam Data Recovery, Ontrack and Secure Data Recovery, and quotes vary by failure type and drive count. Our guide to NAS data recovery costs in Australia covers what drives the price.
Frequently Asked Questions
How long does a RAID rebuild take?
Roughly the replacement drive's capacity divided by the rate the rebuild progresses. An 8 TB drive at 50 to 100 MB/s takes about 22 to 44 hours, a 24 TB drive about 67 to 133 hours. The rate depends on the NAS hardware, the drives, the RAID type and how busy the array is, so the best figure is the progress rate your own NAS reports during a rebuild or a scrub. Enter it as a custom speed.
Why does a rebuild read so much more data than it writes?
To rebuild one drive in a parity array, the NAS reads the matching blocks from every remaining drive and calculates the missing data. A 4-drive RAID 5 of 16 TB drives reads up to 48 TB to write 16 TB. Those reads happen in parallel, so the rebuild time follows the 16 TB being written at the rebuild rate, not the 48 TB being read. The read volume matters for read errors, which this tool reports separately.
Does Synology Fast Repair make rebuilds faster?
Yes, when it applies. Synology documents that Fast Repair skips the unused space in a storage pool, that it is enabled by default, that it covers RAID 1, 5, 6, 10, F1, SHR-1 and SHR-2 pools, and that a pool more than 80% full is repaired with Regular Repair instead. It also says Fast Repair works only on storage pools that have volumes created on them, and cannot speed up the repair of block-level LUNs (a DSM 6.1 and earlier feature that still works after an update). Because it is on by default, it can also have been turned off. This tool applies the saving only when you choose Synology DSM Fast Repair and a supported RAID type; if you are not sure Fast Repair applies to your pool, read the full rebuild time instead. Other platforms are estimated as a full rebuild.
Can I use my NAS during a rebuild?
Usually yes, but everything you do competes with the rebuild for the same drives. Streaming, large copies and backup jobs slow the rebuild down and keep the array degraded for longer. If the data matters, pause heavy jobs until the rebuild finishes. Many NAS systems also let you set how much priority a rebuild gets over normal use in their storage settings.
What is a URE and will one stop my rebuild?
A URE (unrecoverable read error) is a sector a drive cannot read. Vendors rate drives as having fewer than 1 such error per 10 to the power of 14 or 15 bits read. That is a worst-case specification, not a measured rate, and most drives do much better. During a single-parity rebuild there is no redundancy left to fix an unreadable block, and what happens next depends on the NAS software: some mark the affected data as damaged and carry on, others stop the rebuild. With dual parity the second parity can usually rebuild that block.
Should I use RAID 5 or RAID 6 with large drives?
It depends on how long you would run degraded and what a loss would cost you, not on one drive size. Larger drives, more drives, same-age drives, a long wait for a replacement and no tested backup all favour dual parity. The cost is one more drive of capacity. Run your own array through this estimator and the Drive Failure Risk Estimator to see the rebuild-window risk for each.
Is a hot spare worth it?
A hot spare starts the rebuild as soon as a drive fails, which removes the time spent noticing the alert and getting a replacement. It does not add redundancy, and it uses a bay and power the whole time. A cold spare on the shelf removes most of the same wait, and dual parity protects you during the rebuild itself. Which is worth more depends on how long your replacement wait would be and how long the rebuild runs.
What happens if the rebuild fails?
It depends on the RAID type. With single parity and a drive already down, a second drive failure usually loses the array. With dual parity, one more failure leaves the data intact but with no redundancy until the rebuild completes. In Unraid, each data disk has its own filesystem, so a failure beyond parity loses the files on the failed data disks, not the whole array. Restore from backup first; if there is none, professional recovery is possible in many cases and quotes vary by failure type.