Quick answer: A non-zero Reallocated Sector Count (SMART attribute 5) means sectors have already been remapped to spare pool — that number only goes up, never down, and once the spare pool is exhausted the drive starts losing real data. Back up immediately, then replace the drive.
The drive firmware keeps a small pool of spare sectors. When a read fails, it grabs a spare, remaps the bad block, and logs the event in attribute 5. This is why the drive still "works" — it's silently papering over failures. The problem: mechanical wear, head crashes, and heat damage don't happen once. If you've got 8 reallocated sectors this month, you'll likely have 80 next month and 800 the month after. I've seen drives limp along for a year with 4 reallocated sectors, and I've seen drives go from 12 to dead in a week. You don't get to pick which one you have.
Common trigger I see constantly: aging 2.5" laptop drives (Seagate ST2000LM007, WD WD10SPZX) running 24/7 in an external enclosure or NAS with poor airflow. Heat plus mechanical wear equals rapid sector remapping. Same story with old desktop 7200RPM units (WD Blue WD10EZEX, Seagate Barracuda ST2000DM008) that have been running since 2016.
How to read the S.M.A.R.T. data
Don't trust the Windows "SMART status: OK" in Device Manager — it's a dumb pass/fail based on threshold, not trend. Get raw values with a real tool. CrystalDiskInfo on Windows, smartctl on Linux/macOS. Look at the RAW value, not the normalized one.
sudo smartctl -a /dev/sda | grep -iE "reallocated|pending|uncorrect|spin"Key attributes to watch, in order of how bad they are:
| ID | Attribute | What it means |
|---|---|---|
| 5 | Reallocated Sector Count | Sectors already remapped. Bad. |
| 197 | Current Pending Sector | Unstable sectors waiting to be remapped. Worse. |
| 198 | Offline Uncorrectable | Read failures during background scans. Very bad. |
| 10 | Spin Retry Count | Motor struggling to spin up. Mechanical death incoming. |
If attribute 197 (pending) or 198 (uncorrectable) is non-zero, stop reading and go back up your data. Those mean the drive is failing to read sectors it already has on disk.
Fix steps (in order — do not skip 1)
- Back up everything you care about right now. Not tomorrow, not after you "check one more thing." Copy to another drive, the cloud, whatever. If a file won't copy, use
ddrescueorrobocopy /R:2 /W:2so it doesn't hang on bad blocks forever. - Get the raw SMART values. On Windows, install CrystalDiskInfo (portable version works fine). On Linux,
smartctl -a. Write down attribute 5, 197, 198 raw numbers. Take a screenshot. - Check the health trend. Run the check again in 24 hours. If attribute 5 went up at all — even by one — the drive is actively failing. If 197 or 198 ticked up, it's dying fast.
- Run a long SMART self-test. This isn't a fix, it's a diagnostic. It sweeps the whole platter surface and forces the firmware to remap or flag weak sectors. Takes 2–6 hours on a 2TB drive. Check the result with
smartctl -l selftest /dev/sda. - Do NOT run chkdsk /r or fsck on the dying drive as your first move. I know everyone says to. On a drive with reallocated sectors, a full surface scan can push the firmware past its spare pool and kill access to your data. Back up first, then run it if you still want to.
- Check the SATA cable and power. Seriously. A bad SATA cable causes CRC errors that can be misread as sector issues. Look at attribute 199 (UDMA CRC Error Count). If 199 is climbing but 5 is stable, it's your cable, not the drive. Swap the cable, reseat both ends, retest.
- Replace the drive if attribute 5 is above ~20 and climbing. That's my threshold, and it's conservative. Some folks say 100. I don't. A drive that's remapped 50 sectors is on borrowed time.
sudo smartctl -t long /dev/sda
# wait, then:
sudo smartctl -l selftest /dev/sdaIf the drive is already unmountable or throwing I/O errors
Don't fight it. Image the raw device with ddrescue before you try anything clever:
sudo ddrescue -d -r3 /dev/sda /mnt/backup/drive.img /mnt/backup/drive.mapThat retries bad blocks 3 times, skips past them, and keeps a map file so you can resume. Once you've got the image, mount it read-only and pull files. Trying to repair a failing drive in place is how people lose everything.
If ddrescue can't read the drive at all — powering up but not spinning, or clicking — stop. That's a head crash or seized bearing. No software fixes that. Send it to a data recovery lab (DriveSavers, Ontrack) if the data matters. Expect $700–$2500.
Prevention that actually helps
- Run a quarterly SMART check. Set a calendar reminder. CrystalDiskInfo can run at startup and alert on threshold breach. Do it.
- Keep drives cool. Below 40°C at idle. Above 50°C accelerates wear. If your NAS drives sit at 52°C, fix the airflow.
- Don't use desktop drives in NAS enclosures. Buy NAS-rated drives (WD Red Plus, Seagate IronWolf) with higher vibration tolerance and better TLER behavior.
- Replace drives at 5–6 years regardless of SMART. Mechanical drives have a bathtub failure curve. The right side of the curve is a cliff.
- RAID is not backup. Reallocated sectors don't respect RAID. You still need a separate backup.
The number that matters isn't how many sectors are bad. It's how fast that number is growing. A drive with 40 reallocated sectors from three years ago and no movement is a different animal than one with 4 sectors added this week.