Why is my storage pool array synchronization taking so long?

[TMbuddy]

Applicable Versions: TOS 5/TOS 6

Applicable Models: All TNAS Series

1. What is Array Synchronization?

Array synchronization refers to the process where the system automatically redistributes data across all disks when operations such as disk replacement, expansion, or repair are performed on a storage pool array.


2. What Are the Impacts of Array Synchronization?

• Array synchronization consumes disk I/O resources, which may cause slowdowns or lag when reading/writing files or running applications during the process.

• Synchronization is a critical process for ensuring data security. While it may temporarily affect performance, normal operation will resume once completed. It is recommended to schedule synchronization during low-activity periods to minimize impact.

3. When Does an Array Require Synchronization?

Redundant arrays require synchronization during these operations to ensure data redundancy and integrity: array creation, repair, disk addition, online migration, online expansion, and rebuilding.

Arrays with Redundancy: TRAID+, TRAID (when disk count ≥ 2), RAID 1, RAID 5, RAID 6, RAID 10.

Arrays without Redundancy: TRAID (when disk count = 1), Single, JBOD, RAID 0.


4. How Long Does Array Synchronization Take?

Synchronization time depends on the following factors:

• Disk capacity: Larger capacities take longer (e.g., 10 TB disks may require hours to days).

• Disk performance: SSDs synchronize faster than HDDs.

• Workload: Frequent read/write operations during synchronization may slow the process.

• Synchronization speed: Limited by the disk's native read/write performance.

5. How to Accelerate Synchronization in TOS?

① Adjust Synchronization Speed

Path: Storage Management > Storage Pool > Settings > Array Sync Speed

• Options: Fast Sync (100-400 MB/s), Slow Sync (10-50 MB/s), or custom min/max speeds.

• Default sync speed: 50-200 MB/s.

• Actual speed is still constrained by disk performance.

② Enable RAID Fast Repair

Path: Storage Management > Storage Pool > Settings > RAID Fast Repair

• Skips unused space to speed up synchronization.

• Must be enabled before synchronization begins.

6. Example: RAID 1 Sync Time Estimation

Scenario: Two 1 TB disks in RAID 1, default sync speed (50-200 MB/s).

① Capacity Calculation

• Manufacturer’s nominal capacity: 1 TB = 1000 GB (decimal).

• OS/RAID calculation: 1 TB = 1024 GB (binary).

Effective sync capacity: 1 TB ≈ 931 GB (binary) ≈ 953,344 MB.

② Theoretical Time

• 953,344 MB ÷ 200 MB/s ≈ 4,767 s ≈ 79 min.

③ Real-World Time

• Inevitable delays: Speed fluctuations during sustained writes, RAID parity calculations (10%-20% overhead).

• Additional delays: User file operations, background tasks.

Actual sync time will typically exceed theoretical estimates.

[Tsagoth]

Who uses a 1TB drive, c'mon. I just bought an F6-424 and stuck six 28TB drives in it. It started formatting a TRAID volume before I cancelled it, and redid the storage pool as a RAID 6. The TRAID synch said it would take 4448 minutes, about 75 hours. It hasn't said how long the RAID 6 will take but presumably it will take more time not less. Why is it so slow ? On my old DROBO devices with six 8TB drives it would format the pool in about 10 minutes. These drives are super fast, why on earth does it take 75+ hours to simply format some chains of sectors available ? It should be driving all six of those disks at 150MB/sec minimum, this shouldn't take that long.

[CursaYang]

Hello, the time for array synchronization depends on the total capacity of the hard disks and their read/write speed. Therefore, this time is normal for 6 hard disks of 28T each.

[emmascott11]

You’re not just “formatting sectors” — RAID 6 has to build dual parity across all 6 × 28TB drives. That’s a huge amount of data to calculate and verify.
168TB raw capacity means the system is reading/writing tens of terabytes during sync. Even at ~150MB/s real-world sustained speed (which usually drops under parity load), 75+ hours is actually normal for full initialization. stickman hook
Older systems like some DROBO setups often used faster/thin initialization, which looks quick but delays the heavy work.
With disks that big, capacity — not drive speed — is the main reason it takes so long.

[MikeZhang]

Thank you very much for providing a detailed and professional explanation. This will help more people understand the relevant content more easily and clearly.