Skip to main content
StashRaid

Storage Parameters

Usable Capacity

12 TB

Fault Tolerance

4

Array Resilience

Architecture

RAID 5 • Parity

Fault Tolerance

1 Concurrent Failures Allowed

Current Configuration

RAID 5 (Parity)

12 TB Usable

Total RAW: 16 TB

Disk 01DATA
Disk 02DATA
Disk 03DATA
Disk 04PARITY

Total Drives

4 x 4 TB

Efficiency

75% Utilized

Usable: 12 TBWasted to redundancy: 4 TB

Pros & Cons

  • PRO: Good balance of capacity/speed/safety

  • PRO: Cost-effective redundancy

  • CON: Slower writes due to parity calculation

    This setup prioritizes redundancy but has associated performance or capacity costs.

RAID Rebuild Time Guide

RAID rebuild time is one of the most critical — and most overlooked — factors in storage planning. A single drive failure triggers a rebuild window where your entire array is degraded, and the larger your drives, the longer and riskier that window becomes. This guide explains how to calculate rebuild times, why 20 TB+ drives change the math, and how to choose the right RAID level for your risk tolerance.

Rebuild Time Formulas by RAID Level

The basic formula for rebuild time is straightforward:

Rebuild Time (hours) = Drive Capacity (GB) / (Rebuild Speed (MB/s) × 3.6)

However, the amount of data that must be read varies by RAID level. RAID 5 reads every byte on every remaining drive to reconstruct parity. RAID 6 reads the same amount but computes two parity values per stripe — adding ~15–30% CPU overhead. ZFS RAIDZ2 rebuilds similarly but benefits from variable stripe widths.

RAID 5 / RAIDZ1

Rebuilds require reading N−1 drives in full. Single parity means one undetected read error during the rebuild kills the array. Historically recommended for 3–6 drives but increasingly risky with drives above 8 TB.

RAID 6 / RAIDZ2

Rebuilds require reading N−2 drives in full. Dual parity provides a safety net — one read error during rebuild does not cause failure. The CPU overhead for double parity computation adds ~20% to rebuild time. Recommended for 6+ drive arrays or any array with drives >10 TB.

RAID 10

Rebuilds require reading only 1 drive per mirror set — the surviving mirror member is copied to the replacement. This is the fastest rebuild of any RAID level since only one drive's worth of data needs to be transferred, not the entire array.

RAIDZ3

Rebuilds require reading N−3 drives in full. Triple parity allows two simultaneous read errors without data loss. Primarily used in enterprise ZFS deployments with drives >20 TB. Rebuilds are 25–40% slower than RAIDZ2 due to additional parity computation.

Rebuild Time Comparison Table

The table below shows estimated rebuild times for common drive sizes at three different rebuild speeds. Real-world speeds vary based on controller, drive model, and competing I/O.

Drive SizeRAID LevelAt 100 MB/sAt 150 MB/sAt 200 MB/sURE Risk
4 TBRAID 5~11 hrs~7 hrs~6 hrs~27%
4 TBRAID 6 / Z2~14 hrs~9 hrs~7 hrsProtected
8 TBRAID 5~22 hrs~15 hrs~11 hrs~47%
8 TBRAID 6 / Z2~27 hrs~18 hrs~14 hrsProtected
12 TBRAID 5~33 hrs~22 hrs~17 hrs~62%
12 TBRAID 6 / Z2~40 hrs~27 hrs~20 hrsProtected
20 TBRAID 5~56 hrs~37 hrs~28 hrs~79%+
20 TBRAID 6 / Z2~67 hrs~45 hrs~33 hrsProtected

Note: URE risk percentages assume consumer drives with a 1-in-1014 bit error rate. Enterprise drives (1-in-1015 or better) significantly reduce this risk. The RAID 5 percentage represents the probability of at least one URE occurring during a full-array read. RAID 6/Z2 can survive one URE during rebuild; the risk column shows single-URE exposure.

The URE Problem — Why Rebuilds Fail

Every hard drive has a specification called the URE (Unrecoverable Read Error) rate. This is the statistical probability that the drive's internal ECC cannot correct a bad sector during a read. For consumer drives, this is typically:

Consumer drive URE rate: 1 in 1014 bits read (roughly 1 error per 12.5 TB of data read).

Enterprise drive URE rate: 1 in 1015 bits read (roughly 1 error per 125 TB of data read).

During a RAID 5 rebuild, every remaining drive in the array must be read completely. If the array has five 12 TB drives and one fails, the remaining four must read 48 TB total. At 1 in 1014, the probability of hitting at least one URE is approximately:

P(URE) = 1 − (1 − 1/1014)48 × 1012 ≈ 62%

With a 20 TB array rebuild, the probability exceeds 79%. That means four out of five RAID 5 rebuilds on 20 TB consumer drives will fail due to a URE. This is the "second failure" problem — the rebuild itself becomes the most dangerous operation your array will ever perform.

Why 20 TB+ Drives Make RAID 5 Mathematically Dangerous

Do not use RAID 5 (or RAIDZ1) with drives 10 TB or larger.The combination of multi-day rebuild windows and high URE probability means your array is more likely to die during rebuild than to survive it. RAID 6 (or RAIDZ2) is the minimum recommendation for drives above 10 TB. For 20 TB+ drives, consider RAIDZ3 or triple-mirror configurations.

The Timing Problem

A 20 TB drive at a realistic 150 MB/s rebuild speed takes roughly 37 hours to rebuild in RAID 5. During those 37 hours, the array has no redundancy. A second drive failure — even from vibration, thermal stress, or a manufacturing defect — means total data loss. The probability of a second failure during a multi-day rebuild is non-trivial in any array with 4+ drives.

The Capacity Problem

Drive capacities have grown faster than read speeds. A 20 TB drive is roughly 50× larger than a 500 GB drive from a decade ago, but sequential read speeds have only increased ~3× (from ~80 MB/s to ~250 MB/s). The result: rebuild times scale with capacity, not with performance. A 20 TB array takes 50× longer to rebuild than a 500 GB array, not 3×.

Factors That Affect Rebuild Speed

Rebuild speed is not a fixed number. The following factors determine how fast your array can recover:

Drive Type

7200 RPM enterprise drives push 200–280 MB/s. 5400 RPM "NAS" drives manage 150–200 MB/s. SMR (shingled magnetic recording) drives drop to 50–80 MB/s under sustained write load and are unsuitable for RAID rebuilds. Always use CMR drives in RAID arrays.

Controller / Software RAID

Hardware RAID controllers with dedicated XOR engines rebuild faster than software RAID under CPU load. ZFS rebuilds benefit from the ARC cache — recently read data may already be cached. Linux mdadm rebuild speed is tunable via /proc/sys/dev/raid/speed_limit_min.

Competing I/O Load

A rebuild is a background operation. If the array is serving production traffic, rebuild speed drops proportionally. Many systems throttle rebuild I/O to prevent user-visible latency spikes. Scheduling a rebuild during low-activity hours can cut rebuild time by 30–50%.

Real-World Failure Scenarios

Scenario 1: RAID 5 with 6× 12 TB Drives (Consumer)

One drive fails. The remaining five drives (60 TB total) must be read to reconstruct parity. At 150 MB/s, the rebuild takes ~46 hours. With consumer URE rates, the probability of a URE during the full-array read is ~73%. If a URE is hit, the rebuild fails and the array is irrecoverably degraded. Verdict: RAID 5 is not viable at this size.

Scenario 2: RAID 6 with 6× 12 TB Drives (Enterprise)

One drive fails. Enterprise drives (1-in-1015 URE rate) are used. The rebuild reads 60 TB of data. The probability of a URE is ~13%, and RAID 6 can survive one URE during rebuild. Dual parity means a second URE would need to occur for data loss. The rebuild itself takes ~55 hours due to double-parity overhead. Verdict: Acceptable with monitoring.

Scenario 3: RAID 10 with 8× 20 TB Drives

One drive in a mirror pair fails. The surviving mirror drive is copied to a replacement. Only 20 TB needs to be read — not the full array. At 200 MB/s, the rebuild takes ~28 hours. No parity computation is needed, so rebuild speed is limited only by the drive's write speed.Verdict: Fastest rebuild, highest capacity overhead (50%).

Frequently Asked Questions

How long does a RAID 5 rebuild take with 20 TB drives?

At a typical rebuild speed of 100–200 MB/s, a 20 TB RAID 5 rebuild takes 28–56 hours. During that window, the array operates in degraded mode with no parity protection — a second failure means total data loss. This is why RAID 6 or RAIDZ2 is strongly recommended for drives 10 TB and larger.

What is the URE risk during a RAID rebuild?

A consumer hard drive has a URE (Unrecoverable Read Error) rate of 1 in 1014 bits, or roughly 1 error per 12.5 TB read. For a 12 TB RAID 5 rebuild, the array must read ~12 TB to reconstruct the failed drive — giving roughly a 62% chance of hitting a URE. For 20 TB drives, the probability of a failed rebuild exceeds 79% with consumer drives. Enterprise drives (1 in 1015) reduce this drastically to ~13%.

Why is RAID 5 dangerous with large drives?

RAID 5 has only a single parity drive. If any drive fails, the entire remaining set must be read without error to reconstruct the missing data. With consumer URE rates, drives over ~8 TB have a >50% chance of hitting a read error during rebuild. RAID 6 or RAIDZ2 (dual parity) is strongly recommended for drives 10 TB and larger, as it can survive one URE during the rebuild window.

What factors affect RAID rebuild speed?

The primary factors are: drive read/write speed (150–280 MB/s for modern HDDs), RAID controller or software overhead, rebuild priority/ionice settings, competing I/O load, interface bandwidth (SATA vs SAS vs NVMe), and the RAID level itself (RAID 6 requires computing two parity values per stripe, slowing rebuilds by 15–30% versus RAID 5). For detailed performance data, see our Performance Testing Methodology page.

Try the RAID Calculator

Estimate rebuild times for any RAID level, drive size, and configuration. Compare RAID 5 vs RAID 6 vs RAIDZ2 side by side with your actual drives and rebuild speeds.

Try the ZFS Calculator

Model your ZFS pool with specific vdev layouts, parity levels, and drive configurations. See estimated rebuild times and capacity efficiency for RAIDZ1, RAIDZ2, and RAIDZ3.