How to recognise a bad drive in Linux

The commands in the first four sections cover SATA drives, both spinning disks and SSDs. You compare read speed against the typical range for that drive type, then inspect the drive's SMART attributes. NVMe drives report health differently and need their own commands, covered in the last section. Use the results to decide whether a disk needs assessing.

Identify your drives

Use the lsblk command to list the drives in your system:

lsblk

Depending on your disk layout, this shows the available drives, for example sda, sdb, sdc and sdd.

Test disk speed

Benchmark disk speed with hdparm. Install it first:

# Ubuntu / Debian
apt-get install -y hdparm

# AlmaLinux / Rocky Linux
yum install -y hdparm

hdparm is Linux-only, it relies on Linux-specific ATA/SCSI ioctls and isn't packaged for FreeBSD. On FreeBSD, use camcontrol or diskinfo instead for the equivalent checks.

Then run hdparm against the drive you want to test, for example:

hdparm -t /dev/sda

For more reliable results:

• Run the test on an inactive system, ideally after a reboot, so a minimal amount of system resources are in use.
• Run the command two or three times and take the average of the results.

These are generic, industry-typical ranges for the drive class, not Worldstream-specific benchmarks. They apply to SATA drives only, see NVMe drives for why hdparm figures mean nothing on NVMe.

Actual speeds vary with your specific configuration. Contact support to check whether the speed you're seeing is normal for your setup, and include the test output when you do.

Check disk status (SMART)

Use smartmontools to read a drive's S.M.A.R.T. (Self-Monitoring, Analysis, and Reporting Technology) data. Install it first:

# Ubuntu / Debian
apt-get install -y smartmontools

# AlmaLinux / Rocky Linux
yum install -y smartmontools

# FreeBSD
pkg install -y smartmontools

The examples below use /dev/sdb; replace this with the drive you're checking.

smartctl -i /dev/sdb

smartctl -s on /dev/sdb

smartctl -t short /dev/sdb

smartctl -l selftest /dev/sdb

smartctl -A /dev/sdb

Any value above 0 on the three counters warrants investigation, even a single one: it means the drive returned or nearly returned data it could not correct. If any of these values is outside the expected range, open a support ticket with the smartctl output so the disk can be assessed.

Check for ATA errors

SMART attributes don't cover ATA-level errors. Check the kernel log for those instead:

dmesg | grep -i ata

NVMe drives

NVMe drives need different commands throughout. They are not ATA devices, so lsblk lists them as nvme0n1 rather than sda, and the SMART attributes above (Reallocated_Sector_Ct, Current_Pending_Sector, Reported_Uncorrect) do not exist on them.

hdparm -t also understates NVMe badly. It measures buffered sequential reads one request at a time, which is nowhere near what an NVMe drive delivers in parallel, so a healthy drive will look like it fails. Treat any hdparm figure on NVMe as a comparison against that same drive's earlier results, never as a pass or fail. For a real throughput measurement, use fio with several jobs and a queue depth above 1.

Read NVMe health from the controller device:

smartctl -a /dev/nvme0

Kernel-level NVMe errors show up under a different tag than ATA ones:

dmesg | grep -i nvme

As with SATA drives, open a support ticket with the smartctl output if anything looks wrong.