Failure symptoms, likely cause, and recoverability
The table below maps what a user actually observes to the underlying fault class, what to do first, and what usually determines whether the data can be recovered. It covers hard drives, SSDs, RAID arrays, and flash media.
Hard drive symptoms
| Symptom | Likely cause | Fault class | Correct first action | Effect on recoverability |
|---|---|---|---|---|
| Repeating click or beep on power-up | Head assembly failure or heads stuck on platters | Physical | Power down immediately, do not retry | Good if powered down early; each retry risks surface damage |
| Grinding or scraping noise | Head contact with the platter surface | Physical | Power down immediately | Reduced; damaged tracks may be permanently unreadable |
| No spin-up, no sound at all | Seized motor bearing, or electronics/PCB failure | Physical | Power down, do not attempt a board swap | Usually good — the media itself is typically undamaged |
| Detected by BIOS with wrong capacity or model name | Firmware / service area corruption | Physical (firmware) | Power down, do not run repair utilities | Usually good once the service area is repaired in a lab |
| Very slow reads, system freezes on certain files | Bad sectors and read retries | Mixed | Stop using the drive, image it before anything else | Time-critical; degradation is typically progressive |
| Drive asks to be formatted | File system or partition table corruption | Logical | Decline the format, stop writing to the drive | Usually good if nothing is written afterwards |
| Files deleted or volume reformatted | User or software action | Logical | Stop using the device immediately | Depends entirely on how much was written since |
| Burnt smell, visible scorching on the board | Electronics failure, often power related | Physical | Do not reapply power | Usually good; damage is normally confined to the board |
SSD and flash symptoms
| Symptom | Likely cause | Fault class | Correct first action | Effect on recoverability |
|---|---|---|---|---|
| Device not detected at all | Controller failure or power-supply fault on the board | Physical | Stop power cycling | Depends on controller family and encryption state |
| Detected with 0 MB capacity or a generic vendor name | Controller has entered a safe/panic mode | Physical (firmware) | Power down, do not initialise the disk | Often recoverable in a lab; initialising the disk is destructive |
| Read-only: files visible but nothing can be written | Controller has entered a protective read-only state | Firmware | Copy the data off immediately, to another device | Good while the state persists — it may not survive a reboot |
| USB drive reports “please insert disk” | Controller or NAND page failure | Physical | Stop retrying, do not format | Depends on controller; chip-level work may be required |
| Files deleted from a TRIM-enabled SSD | TRIM plus garbage collection | Logical | Power down immediately | Poor — deleted contents are commonly erased within seconds |
| Memory card unreadable after camera error | File system corruption or NAND wear | Mixed | Stop shooting to that card, do not let the device repair it | Usually good if the card is not written to again |
RAID, NAS and server symptoms
| Symptom | Likely cause | Fault class | Correct first action | Effect on recoverability |
|---|---|---|---|---|
| Array reported degraded, one member offline | Single member failure | Physical | Do not rebuild yet; label the drive order first | Good — but an immediate rebuild can destroy it |
| Array offline after a rebuild attempt | Second member dropped, or parity written over good data | Mixed | Power down the whole array | Variable; reconstruct from images, not from the controller |
| NAS boots but shares are empty | Volume or file system corruption on an intact array | Logical | Do not run the vendor repair or reset the configuration | Usually good if configuration is not overwritten |
| Controller replaced, array now shows as foreign or uninitialised | Metadata mismatch between controller and members | Logical | Do not initialise or import as new | Good — the member data is normally untouched |
| Accidentally reinitialised or reconfigured the array | Administrative action | Logical | Power down before anything is written | Depends on whether a rebuild or format followed |
Notes
- Fault class matters more than the device brand. Logical failures never require opening the device; physical failures always do.
- In almost every row above, the correct first action is to stop and remove power. Continued operation is the most common cause of avoidable, permanent loss.
- Recoverability commentary describes general engineering behaviour, not a prediction for any individual case. Only an evaluation of the specific device determines what is possible.