A photo studio in Nansha wrote to us about its Synology DS920+: four bays, two 4TB Western Digital Red drives in SHR, and only 40GB of space left. Lightroom had stopped writing to the catalog. They wanted both 4TB drives replaced with 8TB drives, and they asked whether we could handle it and whether the data would survive the swap.
The short answer: SHR expands by hot-swapping one disk at a time, the data stays put, and the job runs for the better part of two days. Any single step done carelessly can still wreck it. Here is the full record.
1. Before you start: three rules
- Back up first. During a rebuild the disks run under constant load, and an old drive can fail at any point. We rsynced the 1.2TB of finished edits and the client originals to a portable drive before anything else; that step alone took nine hours. Rebuilding an array with no backup means betting every file on the health of two old disks.
- Test the new disks first. Both WD Red Plus 8TB units (WD80EFZZ) went into a drive dock for a full slow scan, a complete read pass in Victoria. Only when they came back with zero bad sectors did we install them. A new drive that is already sick turns a half-finished rebuild into a much bigger problem.
- Protect the power. The studio had no UPS, so we borrowed a Santak TG-BOX 850VA and put the NAS and the router on it. That call saved the job: the next afternoon the breaker really did trip.
2. The process: one disk at a time
SHR expands by replacing disks one after another, and each new disk has to finish repairing before the next one comes out.
| Step | Action | Elapsed |
|---|---|---|
| 1 | Hot-remove bay 1 (4TB), insert the new 8TB, then Storage Manager → HDD/SSD → Action → Replace Drive | 2 minutes |
| 2 | The array drops to degraded and starts repairing on its own, rebuilding onto the new disk and growing the partition | 22 hours |
| 3 | Once the status is back to healthy, repeat steps 1 and 2 for disk 2 | 24 hours |
| 4 | With both disks swapped, Storage Manager → Storage Pool → Action → Expand File System, if DSM has not done it automatically | 2 hours |
Total: about 48 hours. The NAS stayed online and usable the whole time, but throughput dropped because the rebuild owned the I/O. Retouching exports ran at roughly half speed for two days, which is expected.
3. The interruption: power lost at 61%
Disk 2 was 61% through its rebuild when the studio air conditioner tripped the breaker. The UPS carried the NAS for 12 minutes and it shut down cleanly, because we had set DSM to enter safe mode on power loss. When the power came back, DSM resumed the rebuild from 61%. No data lost, no errors reported.
That is the resume capability of an SHR or RAID rebuild, and it depends entirely on a clean shutdown. A hard cut with no UPS leaves a real chance of inconsistent array metadata, which means checking and repairing the array by hand with mdadm. That is a far worse position to be in.
4. Why the result is 7.2TB and not 14TB
The customer stared at the total after the expansion: 7.2TB. Here is the explanation. Two equal disks in SHR are a RAID1 mirror: one disk holds the data, the other holds the copy. Two 8TB disks therefore give about 7.2TB usable, since 8TB nominal is 7.28TiB and the system partition comes out of that.
To actually double the capacity you either add a third disk, because SHR only starts offering capacity maximization at three disks, or move to SHR-2 or RAID5 and expand disk by disk. Redundancy is not wasted space; it is a discount taken up front against recovery cost. Cleanroom platter recovery on a 4TB drive runs RMB 3,000-8,000, and the cost of mirroring sits well below that.
5. Reading the SMART data: is a retired disk still usable?
We pulled SMART from both 4TB Reds before retiring them.
- Disk 1: 26,000 power-on hours, roughly three years without a break, zero reallocated sectors and zero pending sectors. Healthy, so it became the cold backup disk.
- Disk 2: 12 reallocated sectors and 3 pending. Physical bad sectors, which makes the disk scrap.
Without this planned expansion, disk 2 might not have lasted the winter. If it had dropped out, SHR would have gone degraded with disk 1 as the only remaining copy, which is how a data disaster starts.
That is why we tell NAS owners to look at SMART twice a year. If 05 (reallocated sectors), C5 (pending) or C6 (uncorrectable) is anything other than zero, it is time to plan a replacement.
6. Appendix: measured rebuild times on a DS920+
| Old disks → new disks | Rebuild per disk | Total for two disks |
|---|---|---|
| 4TB → 8TB | 22-24 hours | about 48 hours |
| 4TB → 16TB | 40-46 hours | about 4 days |
| 8TB → 16TB | 38-44 hours | about 3.5 days |
Rebuild time scales with disk capacity and runs inversely to NAS performance. Keep the power steady while it runs, do not reboot the NAS, and do not pile on heavy read and write jobs at the same time.
ACCPC handles the full range of Synology and QNAP NAS work: expansion and drive swaps with a data verification report, array repair, recovery of accidentally deleted data, ransomware decryption assessment and system migration. We work on site in Panyu and Nansha and take shipped units from other districts. The expansion service fee is RMB 600 for up to four bays, including 48 hours of monitoring; data backup is quoted separately.