Read dedicated server drive health without a destructive test
Drive health reports help investigate hardware warnings, but a healthy status cannot guarantee future reliability. This procedure is for a dedicated-server administrator with root access and smartmontools already installed. VPS guests usually see virtual devices and may not expose the physical drive's health. Ask hosting support for host evidence rather than treating missing SMART support as a failed drive.
Identify the hardware carefully
date -Is
lsblk -o NAME,TYPE,SIZE,MODEL,SERIAL,MOUNTPOINTS
sudo smartctl --scan-open
The scan discovers supported devices and access methods. It opens devices for identification, so it can wake a sleeping drive; it is not a write benchmark. Save serial numbers privately to distinguish drives of the same model. Do not post complete serial inventories publicly.
Hardware RAID controllers may hide member drives or require a controller-specific device option. Do not guess controller indexes or use commands that change RAID configuration. If the scan does not clearly identify the intended drive, stop and ask support which device path and access method are correct.
Read one device report
sudo smartctl -a /dev/nvme0
The device above is an example, not a universal path. Use the device and any required type option identified for your machine. Reading the report does not start an extended self-test, erase media or repair errors. Avoid adding test, security or configuration options from unrelated examples.
Read the model, protocol and report sections before interpreting values. ATA attributes are often vendor-specific. NVMe exposes different health fields, including critical warnings, temperature and data-integrity counters. Raw numbers from unlike models are not directly comparable. The installed smartctl manual explains the output and its exit-status bitmask; a nonzero status has more detail than a simple generic command failure.
Correlate warnings with the incident
Review existing kernel logs for I/O errors, resets or timeouts around the same UTC interval. Check filesystem free space and application errors separately. A full filesystem is not evidence of failing NAND, and a temperature reading without workload history is not a complete cooling diagnosis.
Compare reports over time using the same device identity. New or increasing media errors deserve attention even when a summary still looks reassuring. Do not clear logs or reset counters to make a dashboard green. Preserve the report before any replacement or provider investigation.
Escalate without stressing the drive
If there are critical warnings, read failures or recurring resets, reduce nonessential work and verify your independent backups. Contact support with the service ID, affected device identity, timestamps and relevant redacted report sections. Agree a replacement and data-protection plan before changing hardware.
Long self-tests and benchmarks can be useful in a planned diagnostic window, but this article does not prescribe them on a busy production drive. A failing drive may degrade further under added load. There is no rollback for a read-only health report; the useful outcome is preserved evidence and a supported next action, not an improvised repair command.
Related instructions
Find disk-space use. Verify a downloaded backup.
Reference: Official documentation. Check the documentation installed on your own operating-system release when an option differs.