1. Four ways a Synology NAS fails
Synology units are everywhere in small and mid-sized businesses and in design studios, so plenty of them reach our bench. Our intake records sort the failures into four patterns:
- System partition corruption. The NAS answers a ping but the DSM web interface will not load, or it reports a system partition failure; over SSH the system volume mounts abnormally. Usually a power cut, a failed DSM upgrade or unstable memory.
- Degraded or crashed storage pool. Storage Manager reports a crashed pool or a corrupted volume, and one or more disks show as uninitialized or as having a failed system partition.
- Disks dropping out together. A backplane power fault or a SATA channel problem takes several disks offline at once, and the disks themselves may be fine.
- Human error. A shared folder or a whole volume deleted, a disk with data on it initialized, a RAID type migration that failed halfway.
The first thing to establish is whether you have a disk problem or a system problem. The handling is completely different.
2. Case: a design studio's DS920+ after a power cut
On August 18, 2026, a brand design company in Nancun, Panyu, was working late against a deadline when the office tower lost power. When it came back, the DS920+ would not start: four 4TB Western Digital Red disks in SHR, holding client source files, the asset library and the company's accounting photos, about 9TB in total. The power light came on, the status light sat at amber, the DSM page timed out, and a monitor attached to the unit reported "system partition failure, please reinstall DSM."
The office administrator nearly clicked reinstall. Here is the sentence that matters: the Synology "reinstall DSM" prompt normally rewrites only the system partition and leaves the data partition alone, but "initialize" and "rebuild storage pool" erase data. If you cannot tell which one you are looking at, click neither. Pull the power and send the unit in. They did the right thing and brought it to us.
What we did:
- Test the disks first, and leave the array alone. All four came out and went one at a time onto a read-only rig for SMART and surface scans. Disk 3 had 47 reallocated sectors and the count was still climbing; the other three were healthy. Diagnosis: the power cut damaged disk 3, SHR went degraded, and the system retried over and over in an abnormal state until the system partition collapsed.
- Image every disk. All four were mirrored read-only. Disk 3 was imaged in slow mode with bad-sector skipping, which took nine hours.
- Reassemble in Linux. SHR is mdadm software RAID (the RAID5 variant in this unit), on top of LVM2 and ext4. Working from the images, we rebuilt the md array, activated the LVM volume and mounted ext4. The directory tree came back complete, with the client source files and the asset library in place. ext4's journal did a lot of work for us after this particular power cut; the file-system damage was minor.
- Export and migrate. The 9TB went onto a new NAS the customer bought, a DS923+ with four 8TB disks, this time with snapshots enabled. Spot check: 50 PSD and AI source files opened at random, all fine; Lightroom rebuilt its catalog and every photo was there.
Billing: RMB 3,200 for reassembling and recovering the array, four-disk SHR with 9TB of valid data, plus RMB 400 for imaging the bad-sector disk, RMB 3,600 in total. The export ran to the customer's own new NAS, so we did not charge for media.
3. What you can try yourself, and when to stop
Reasonable to try on your own, with little risk to the data:
- The DSM page will not load but SSH gets in: the DSM service may simply have hung. A normal shutdown and start from the power button usually brings it back.
- One disk reports uninitialized while the pool is still degraded and usable: copy the critical data out while the array is online, then think about replacing the disk. A degraded pool is the last window you get, so do not spend that window on a rebuild.
- A shared folder deleted by mistake, with the recycle bin enabled: look in the #recycle directory.
- Snapshots were on, Btrfs volume plus Snapshot Replication: roll back from the snapshot. It is the best undo button Synology offers, and every Btrfs user should enable it, since the space it needs can be capped.
Stop and call someone:
- The pool has crashed and several disks are abnormal: do not press Repair Storage Pool or Initialize Disk.
- The disks are making noises and SMART reports a pile of bad sectors: image first, do everything else later. Poking at the original disk is secondary damage.
- Something went wrong after an interrupted RAID migration or expansion: the array metadata may be stuck mid-operation, and manual work will very likely scramble the disk order.
4. Why this kind of recovery works
Synology's storage layout is not a secret: Linux mdadm software RAID (SHR, RAID1, 5, 6 or 10) on top of LVM2, with ext4 or Btrfs as the file system. As long as you have a full-disk image, you can reassemble the array and mount the file system in a standard Linux environment, or with tools such as PC-3000 and UFS Explorer, whether or not the original unit will boot. That is also why NAS recovery success rates are generally decent; the layout is far friendlier than an encrypted SSD controller.
Two things are genuinely hard. One is extensive Btrfs metadata corruption, since ext4 is the more forgiving of the two. The other is a disk order or RAID parameter that someone scrambled and then wrote to. Which brings back the old rule: after an accident, writes are the enemy.
5. 2026 price reference for NAS recovery
| Case | Reference price | Notes |
|---|---|---|
| Accidental deletion or format, disks healthy | RMB 500-1,200 | Reassembly and recovery from images |
| Crashed pool or damaged system partition, disks healthy | RMB 1,500-3,200 | By disk count and data volume |
| Array recovery involving bad-sector disks | plus RMB 400-800 | Deep imaging of the bad sectors |
| Hardware-level repair, backplane, mainboard or power supply | RMB 300-800 plus parts | Repair the unit when the disks are good |
As with our other jobs, there is no recovery fee if the recovery fails, and diagnosis and assessment are free.
6. Four precautions for NAS owners
- Buy a UPS. A UPS costing RMB 500 or 600 would have prevented the whole incident. An abnormal power loss is the thing a NAS fears most, and a mid-range unit such as the Santak TG-BOX 850VA keeps it alive for 15 minutes, long enough for a clean shutdown.
- Turn on snapshots. A Btrfs volume plus Snapshot Replication is the antidote to ransomware and to a slip of the mouse. Configure it once and it runs for years.
- A NAS is not the only backup. The unit itself can be stolen, burned or encrypted. Critical data needs one more copy somewhere else, in the cloud or on a physically separate disk.
- Watch SMART. DSM Storage Manager shows disk health. When 05 (reallocated sectors), C5 (pending) or C6 (uncorrectable) is non-zero, buy a replacement before the disk drops out.
If the NAS has already failed: power it off, do not touch any initialize or repair button, then call 020-39029800. Tell us the model, the number of disks, the RAID type and what happened before the failure, and we can point you in the right direction on the phone.