> ## Documentation Index > Fetch the complete documentation index at: https://docs.derekdinh.com/llms.txt > Use this file to discover all available pages before exploring further. # Identifying and Responding to Storage Drive Failure > Spot early warning signs of a failing drive, run SMART diagnostics with CrystalDiskInfo or smartctl, back up data, and know when to replace the drive. Storage drive failure is one of the few hardware problems where delayed action can mean permanent, unrecoverable data loss. Unlike a GPU that simply stops producing video, a dying drive can corrupt or destroy years of files in the time it takes to finish a single boot cycle. If you suspect your drive is failing, treat this guide as urgent — back up data before you attempt any repair. Drives rarely fail without warning. Learn to recognise the early indicators so you can act before catastrophic failure occurs. **Hard Disk Drives (HDDs):** * **Clicking, grinding, or repetitive ticking sounds** — the read/write head is failing or the platters are damaged. This is a critical emergency; shut down immediately. * **Unusually slow read/write speeds** — a file copy that used to take 10 seconds now takes several minutes. * **Files or folders that disappear or become corrupted** spontaneously. * **CRC (Cyclic Redundancy Check) errors** during file copies or backups. **Solid State Drives (SSDs) and NVMe Drives:** * **System freezes of several seconds** during normal operation, particularly when accessing files. * **Drives that intermittently disappear** from the operating system and reappear on reboot. * **Files that fail to open** or return read errors. * **Sudden dramatic performance degradation** (sustained write speeds dropping to single-digit MB/s on a drive that was previously fast). If you hear clicking or grinding sounds from a spinning hard drive, power down the machine immediately. Continued operation can turn a recoverable failure into complete physical destruction of the platters and your data. Self-Monitoring, Analysis, and Reporting Technology (SMART) is built into virtually every drive manufactured after 2005. It tracks dozens of internal health indicators and can reveal impending failure before symptoms become obvious. **Windows — CrystalDiskInfo** 1. Download and install [CrystalDiskInfo](https://crystalmark.info/en/software/crystaldiskinfo/) (free, portable version available — no installation needed). 2. Launch the application. Each connected drive appears as a tab. 3. Check the overall health status badge: **Good** (healthy), **Caution** (monitor closely), or **Bad** (replace immediately). 4. Pay particular attention to these SMART attributes: | Attribute ID | Name | Concern Threshold | | ------------ | -------------------------- | -------------------- | | 05 | Reallocated Sectors Count | Any value above 0 | | C5 | Current Pending Sectors | Any value above 0 | | C6 | Uncorrectable Sector Count | Any value above 0 | | BB (SSD) | Uncorrectable Error Count | Any value above 0 | | E9 (SSD) | Media Wearout Indicator | Below 10 is critical | **Linux and macOS — smartctl (smartmontools)** ```bash theme={null} # Install on Debian/Ubuntu sudo apt install smartmontools # Install on macOS via Homebrew brew install smartmontools # List all drives sudo smartctl --scan # Run a short self-test on /dev/sda (takes ~2 minutes) sudo smartctl -t short /dev/sda # Display full SMART data including all attributes sudo smartctl -a /dev/sda # For NVMe drives, use the NVMe device path sudo smartctl -a /dev/nvme0 ``` ```bash theme={null} # Read the self-test results log after the test completes sudo smartctl -l selftest /dev/sda ``` A **passed** SMART status does not guarantee a healthy drive. SMART catches approximately 30–40% of drive failures before they occur. Slow performance, CRC errors, or unusual sounds should be taken seriously even when SMART reports "Good." Before running any repair tools, create a complete backup of your data. Repair utilities like `chkdsk` and `fsck` write to the drive and can, in rare cases, trigger a drive that is on the edge of failure to fail completely. **Quick backup options:** * **External USB drive** — copy critical files manually using your file manager, or use Windows Backup/macOS Time Machine for a full system image. * **Cloud storage** — upload critical documents, photos, and projects to OneDrive, Google Drive, or Dropbox. Note that upload speeds may limit how much you can protect quickly. * **Disk imaging (recommended)** — create a sector-level clone of the drive before it deteriorates further: ```bash theme={null} # Linux: create a raw disk image with ddrescue (more resilient than dd on bad sectors) sudo apt install gddrescue # Clone source drive /dev/sda to image file on an external drive mounted at /mnt/backup sudo ddrescue -d -r3 /dev/sda /mnt/backup/drive_image.img /mnt/backup/drive_image.log ``` ```powershell theme={null} # Windows: create a system image via the Control Panel # Control Panel → Backup and Restore (Windows 7) → Create a system image ``` Do not back up to another partition on the same physical drive. A drive failure affects the entire disk — all partitions will be lost simultaneously. After securing your backup, use `chkdsk` to scan the volume for file system errors and attempt to recover readable data from bad sectors. Open **Command Prompt** as Administrator (right-click → Run as administrator): ```cmd theme={null} :: Scan the C: drive and repair file system errors :: /r locates bad sectors and recovers readable data :: /f fixes errors on the disk chkdsk C: /r /f ``` Because the system volume (C:) is in use when Windows is running, `chkdsk` will prompt you to schedule the scan for the next restart. Type **Y** and restart. ```cmd theme={null} :: To check a secondary data drive (e.g., D:) that is not in use: chkdsk D: /r /f :: View the chkdsk results log after it runs (search the Event Viewer): :: Event Viewer → Windows Logs → Application → Source: Wininit ``` On a large or heavily fragmented HDD, `chkdsk /r` can take several hours to complete. Do not interrupt it once started. If it freezes for more than 30 minutes on a single percentage point, the drive is likely beyond software repair. On Linux, `fsck` performs the same role as `chkdsk` — scanning and repairing the file system. It must be run on an **unmounted** file system; running it on a mounted partition can cause further corruption. ```bash theme={null} # Identify your drive and partition layout lsblk -f # For a partition that can be unmounted (e.g., /dev/sdb1, a secondary drive): sudo umount /dev/sdb1 sudo fsck -y /dev/sdb1 # For the root partition, boot from a live USB (Ubuntu, Fedora, etc.) first, then: sudo fsck -y /dev/sda1 # For ext4 file systems, you can force a full check: sudo fsck.ext4 -f -y /dev/sda1 # View detailed output: sudo fsck -v -y /dev/sda1 ``` ```bash theme={null} # If fsck reports: "UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY" # Boot into recovery mode or a live environment and run: sudo fsck -y /dev/sda1 # The -y flag automatically answers "yes" to all repair prompts ``` On macOS, the equivalent utility is **Disk Utility** (Applications → Utilities → Disk Utility → First Aid) or the command-line `diskutil`: ```bash theme={null} # Run First Aid on a volume sudo diskutil repairVolume /dev/disk2s1 ``` Even if `chkdsk` or `fsck` completes successfully, a drive that has developed bad sectors is on a downward trajectory. Use the following criteria to make the replacement decision. **Replace immediately if:** * CrystalDiskInfo or `smartctl` reports a **Bad** or **Caution** status with reallocated sectors, pending sectors, or uncorrectable errors above 0. * The drive produces clicking, grinding, or repetitive seek sounds. * `chkdsk` or `fsck` found and had to repair errors. * The drive has disappeared from the system more than once without explanation. * The drive is more than **5 years old** (HDDs) or is showing an SSD wear indicator below 10%. **Can continue monitoring if:** * SMART status is "Good" with all critical attributes at zero. * Performance is within expected range for the drive's age. * `chkdsk` or `fsck` found **no errors**. * The slowness was explained by a one-time cause (e.g., disk full, index rebuild). - **NVMe SSD** — best performance, ideal for OS and applications. Look for drives with a DRAM cache for sustained write performance (Samsung 990 Pro, WD Black SN850X). - **SATA SSD** — good performance, lower cost per GB, compatible with older systems lacking M.2 slots (Samsung 870 EVO, Crucial MX500). - **HDD** — highest capacity per dollar, suitable for bulk storage and backups where speed is not critical (Seagate IronWolf for NAS, WD Blue for desktop storage). Never use a drive that has been given a **Bad** SMART rating as a backup destination or secondary storage. A second failure in a short period is highly probable. Some storage failures require physical intervention or specialised data recovery tools beyond the scope of standard troubleshooting. * **Clicking or grinding hard drive** with unrecovered, irreplaceable data — stop all writes and refer to a professional data recovery service (DriveSavers, Ontrack). Do not run `chkdsk` or any write-based tool on a clicking drive. * **Drive not detected** in Disk Management (Windows) or `lsblk` (Linux) after trying different SATA/NVMe ports and cables. * **Multiple drives failing** in a short period — this can indicate a failing PSU delivering unstable voltage rather than individual drive problems. * **RAID array degraded** or multiple disks offline simultaneously. When escalating, provide your SMART report export from CrystalDiskInfo (File → Save Smart Information), the `smartctl -a` output, the drive's make/model/age, and a description of the symptoms and timeline.