SMART Self-Tests Explained: Short, Long, and Conveyance — When to Run Each One
Short, long, or conveyance — which SMART self-test should you actually run? The answer depends on the question you are asking the drive. A short test answers “do the basics work?” A long test answers “can every part of the surface be read?” A conveyance test answers a narrower question still: “did moving this drive damage it?”
Most guides stop at the commands. The commands are the easy part. Choosing the test, and reading the result line it leaves behind, is where the value sits.
The three tests, side by side
| Test | What it checks | Typical time | Best for |
|---|---|---|---|
| Short | Electrical and mechanical checks, plus a read/verify of a portion of the disk | About 2 minutes or less in one source’s table; under ten minutes in another’s, varying by manufacturer | A routine quick check; a first pass on a drive you have just connected |
| Long (extended) | The short test’s checks extended to reading the entire disk surface | No fixed limit — duration scales with disk size; several hours on large drives | Suspected unreadable areas; a thorough check before trusting a drive with data |
| Conveyance | Damage incurred during transport | A few minutes | A drive that has just been shipped or moved — ATA drives only |
Two details in that table do real work. The short test reads only a portion of the disk, so its pass is a sample, not a survey. And the conveyance test is exclusive to ATA drives — asking an NVMe drive for one is asking for a test that does not exist on it.
Match the test to the situation
- Routine check on a healthy drive: short. It is brief by design and can reliably confirm a faulty disk when the fault sits in the area it covers.
- A used drive before you trust it, or a suspected bad area: long. Only the long test reads the entire surface, so only it can find the unreadable sector hiding outside the short test’s sample.
- A drive that just arrived by courier or moved house: conveyance, if it is an ATA drive. That is the one job this test exists for.
- A drive already reporting errors, clicking, or struggling to be read: none of them. See the limits below — this is the exception that matters most.
Run one: the procedure (smartctl)
The documented smartctl flow, on Linux, for a drive that is stable and backed up:
- Ask the drive how long its tests take. Run
sudo smartctl -c /dev/sdX(replace sdX with your drive). The output lists a recommended polling time for each test — one recorded example shows 2 minutes for short, 157 minutes for extended, and 5 minutes for conveyance on the same drive. - Start the test you chose. Run
sudo smartctl -t short /dev/sdX,-t long, or-t conveyance. The test runs inside the drive, in the background, so other work can continue. - Leave it alone while it runs. Tests can run while the drive is mounted, but expect some drop in performance. Do not stack a long test on top of other disk-intensive work — NAS software documentation warns against scheduling self-tests at the same time as a scrub or resilver, for the same reason.
- Read the log. Run
sudo smartctl -l selftest /dev/sdXto see the results of the tests the drive has run. - Abort only if you must.
sudo smartctl -X /dev/sdXcancels a running test. An aborted test leaves an aborted result in the log — that is a record of your interruption, not a drive verdict.
What success looks like: the log’s newest entry names the test you ran and reads “Completed without error.” Anything else is information of a different kind — read it with the next section.
Reading the log: what the result line actually says
The self-test log is a history, newest first. Three readings cover most of what home users will meet:
- “Completed without error” after a long test is the strongest result a self-test gives: the whole surface was read. It still is not a promise about next month.
- “Completed without error” after a short test clears the sampled portion and the electrical and mechanical checks — and says nothing about the rest of the surface.
- A read failure means the drive found an area it could not read during the test. A failed self-test is a clear indication the drive should be replaced — copy data off first if it is not already backed up, and do not re-run tests hoping for a kinder result.
The limits: what no self-test can prove
Self-tests are read-side diagnostics run by the drive on itself. Their limits follow from that. A short test cannot clear areas it never read. A long test adds read load for hours, which is why timing and scheduling matter. And a passing test today does not retire the ordinary rule of this site: monitoring matters because drives change.
The hard limit bears repeating plainly. Do not run self-tests on drives that are already reporting errors unless specifically instructed by support: the tests add load to a drive that is already struggling, and can stress it further. On a drive in that state, the correct sequence is data off first, diagnostics later — or never.
— Dex Carter, Tech Writer