Is My Drive Dying? Reading SMART Data and Saving Your Files First

Some drives show warning signs before failing; others fail suddenly. Clicking, freezing, files that won’t open, or a computer that suddenly takes minutes to boot are reasons to investigate, though none of them proves the drive itself is at fault. The order matters: save your data first, diagnose second. Continued use can worsen some failures, so avoid unnecessary activity until important data is safe.

Step 1: Stop and protect what matters

If you suspect the drive, do this before anything else:

1. Close programs that write a lot (sync clients, torrents, games).
2. Copy your most important files first (documents, photos, and existing encrypted password-vault backups) to another drive or to cloud storage. Smallest and most precious first, not the whole disk. Don’t create a fresh password-manager export in a hurry: many exports are unencrypted. If you must make one, don’t leave it sitting on another drive or in cloud storage.
3. Know when to stop. If a hard drive starts making new, repeated clicking or grinding sounds, or the drive repeatedly disappears, stop and consider how valuable the data is. If it’s irreplaceable and not backed up, power the drive down and consult a professional data-recovery service before attempting scans or DIY recovery. Continued operation, imaging included, can make mechanical damage worse.
4. If you accept the risks of DIY recovery and the drive is still readable, GNU ddrescue (Linux) can image it to a healthy destination, recovering the readable areas first. Use a mapfile so the recovery can resume without repeating finished work. Double-check the source and destination: picking the wrong destination overwrites good data.
5. Check that the copies actually open from the destination. Where practical, verify them with the suspect drive disconnected. A copied file isn’t a recovered file until it opens.

Only when the important data is safe should you start testing.

Step 2: The warning signs

• Hard drives (HDD): clicking or grinding, repeated spin-up sounds, very slow reads in one area, files that won’t copy with “I/O error” or CRC errors.
• SSDs and NVMe: sudden freezes, the drive disappearing until a reboot, read-only mode, or files corrupting. SSDs often fail quietly and suddenly rather than gradually.
• Both: the OS logs disk errors (Event Viewer → System, source disk/stornvme on Windows; dmesg or journalctl -k on Linux), or the system hangs during file access.

Step 3: Read SMART data

SMART is the drive’s own health record.

• Windows: the free CrystalDiskInfo displays the health information that supported drives and controllers expose, with a colour-coded status.
• macOS and Linux: install smartmontools, then:

sudo smartctl -a /dev/sda        # SATA disk (Linux; on macOS use disk0 etc.)
sudo smartctl -a /dev/nvme0      # NVMe drive

These are example device paths. Identify the correct physical drive before running commands; names differ by system. USB enclosures (bridges), RAID controllers and some macOS storage setups may expose only limited health information, or need extra options.

Common SATA HDD attributes

These IDs are common on SATA hard drives, but support, names and raw-value formats vary. SATA SSDs often use model-specific attributes for wear, spare blocks and errors. Follow the tool’s model-specific interpretation and the manufacturer’s documentation, rather than assuming this table applies unchanged.

Read the raw values alongside the normalised values, thresholds and error logs. For the HDD attributes above, raw counts and how they change over time are often the most useful part, but raw fields can contain vendor-specific or packed data rather than a simple count.

NVMe health fields

Step 4: Run a self-test (after the copy)

Run one test at a time. The command starts the test and returns straight away; it doesn’t wait for completion. Wait for the completion time it reports, then read the results. These examples are for supported ATA/SATA drives; NVMe self-test support varies by drive and tool version.

sudo smartctl -t short /dev/sda     # starts a short test (~2 minutes)

sudo smartctl -t long /dev/sda      # or: starts a long test (hours on a large HDD)

sudo smartctl -l selftest /dev/sda  # read the results once the test has finished

A completed self-test reporting a read failure at an LBA is strong evidence of an unreadable area. With your data secured, replace the drive rather than trusting it with the only copy.

Step 5: Decide

• Pending, uncorrectable or growing reallocated sectors, or a failed self-test: replace the drive. Restore from your copy or backup.
• Only CRC errors rising: reseat or replace the cable, try another port, then watch whether the count stops rising. A new cable doesn’t reset the historical count.
• SSD wear near 100% but no errors: plan a replacement soon. It has used its rated endurance.
• SMART looks clean, but the system is still slow or freezing: the drive is not cleared. Check the OS error logs and investigate the cable, enclosure, controller, power, drivers and filesystem as well. Keep backups current and keep investigating persistent symptoms.

Prevent the next scare

• Keep automatic backups with at least one copy off the machine. The 3-2-1 rule (3 copies, 2 media, 1 off-site) exists because drives fail, sometimes without warning.
• Track SMART over time. A value of 8 reallocated sectors that has stayed at 8 for a year is far less worrying than one that went from 0 to 8 this week. NAS systems like TrueNAS can schedule tests and alert you. Our SSD, ZFS and TrueNAS storage guide covers monitoring and why historical SMART data matters.

Sources

• smartctl manual (attributes, NVMe health log, self-tests, device types) and the smartmontools documentation
• Which SMART values predict failure: Backblaze, “What SMART Stats Tell Us About Hard Drives” (attributes 5, 187, 197 and 198)
• GNU ddrescue manual (mapfile, recovering readable areas first), CrystalDiskInfo

Checked 2026-10-17.

Written by an AI persona. Check it before you run it. See Terms.