Reallocated, Pending and Uncorrectable Sectors: What Fails a Drive

A used hard drive fails our inspection when SMART attribute 5 (reallocated sectors), 197 (current pending) or 198 (offline uncorrectable) shows any nonzero raw value, or when a SAS drive reports medium errors. A single grown defect on a SAS drive is a warning, not an automatic fail. Here is what each number means and why the thresholds are strict.

Three attributes that matter

SATA drives report dozens of SMART attributes, and most of them are noise for buying decisions. Three describe the physical health of the platter surface, and they are the ones to read first.

ID	Name	What it counts	Our threshold
5	Reallocated Sector Count	Sectors the drive gave up on and remapped to spare area	Raw must be 0
197	Current Pending Sector Count	Sectors that failed a read and are waiting for a write to decide their fate	Raw must be 0
198	Offline Uncorrectable Sector Count	Sectors that failed during an offline scan or self-test and could not be corrected	Raw must be 0

Reallocated sectors (5)

Every hard drive ships with a pool of spare sectors. When the firmware decides a sector is unreliable, it copies the data (if it can still read it) to a spare and remaps the address. The operating system never sees this happen.

A small reallocated count does not make a drive unusable. It does mean the surface has started to degrade in at least one spot, and large-fleet observations consistently show drives with reallocated sectors fail at a higher rate than drives with none. That correlation is why we fail a drive on any reallocation. The drive may run for years, but we cannot certify that it will.

Pending sectors (197)

A pending sector is a sector the drive tried to read and could not. It has not been remapped yet because the firmware does not know whether the problem is the media or a transient write issue. The next write to that address resolves it: if the write succeeds and verifies, the pending count drops and nothing is remapped. If it fails, the sector is reallocated and attribute 5 increments.

Pending sectors are the most dangerous of the three for a buyer, because the data at that address is currently unreadable. On a drive in service, a pending sector is a sector your filesystem cannot get back without redundancy. A used drive advertised with "only one pending sector" is a drive with an unresolved read failure.

Offline uncorrectable (198)

Attribute 198 counts sectors that failed during the drive's own offline scanning or a self-test. It overlaps with 197 in practice; the distinction is that 198 was found by the drive scanning itself, not by a host read. A nonzero 198 after a completed extended self-test means the test found a bad spot and could not fix it.

Raw versus normalized values

Each SMART attribute has a normalized value (usually 100 or 200 when healthy, counting down), a worst value, a threshold, and a raw value. Tools like smartctl print all four. The normalized value is what the drive uses to decide whether to report an overall SMART failure, and manufacturers set the thresholds generously. A drive can have hundreds of reallocated sectors and still show a normalized value above threshold, which means the overall health status reads PASSED.

Read the raw value. For attributes 5, 197 and 198 the raw value is a plain count on nearly every drive. Ignore the overall PASSED or FAILED verdict for buying decisions; it only fires when a drive is close to dead. Our post on how to read a SMART report covers the output line by line.

One caveat: some Seagate attributes (1, 7 and 195) report enormous raw numbers that are not error counts. They pack several counters into one field. Attributes 5, 197 and 198 are not affected by this; a raw value of 8 means eight sectors.

SAS drives: grown defects and medium errors

SAS and other SCSI drives do not use the SATA attribute table. The equivalent data lives in log pages, which smartctl reads with the same command.

• Primary defect list (P-list). Defects mapped at the factory before the drive shipped. Every drive has some. They are irrelevant to health.
• Grown defect list (G-list). Defects the drive found and remapped in service. This is the SAS equivalent of SATA attribute 5. smartctl prints it as "Elements in grown defect list."
• Error counter log. Separate read, write and verify counters with columns for errors corrected with and without delay, errors corrected by rereads or rewrites, and uncorrected errors. The "total uncorrected errors" column is the one that matters. Large values in the corrected columns are normal, especially for reads; the drive's ECC works constantly.
• Medium errors. Uncorrected read errors attributed to the platter surface. These are the SAS equivalent of a pending or uncorrectable sector.

Our rule for SAS is slightly different from SATA. Zero grown defects and zero uncorrected errors is a pass. A single grown defect with zero uncorrected errors and a passed extended self-test is marked WARN on the test report, disclosed in the listing, and the drive is sold with that note. Two or more grown defects, or any uncorrected medium error, is a fail. The reasoning is that one remapped sector on a drive with 36 billion of them is within what the manufacturer's own reliability model allows, while a growing list indicates an active problem.

Why the self-test matters

SMART attributes only record problems the drive has already encountered. A drive that has sat idle for a year can have a developing bad spot that no read has touched yet. The SMART extended self-test (sometimes called the long test) reads every sector on the drive and reports the first unreadable address if it finds one. On an 18 TB drive this takes well over a day, which is why many sellers skip it.

A used drive with a clean attribute table and no recent self-test has not been proven. A drive with a clean attribute table after a completed extended self-test has had its whole surface read within the last few days. That is the difference between "no known problems" and "checked."

FAQ

Is one reallocated sector really a problem?

For a drive you already own in a redundant array, not necessarily; watch it. For a drive you are paying for, it is a reason to pay less or buy a different unit. We do not sell SATA drives with any reallocated sectors.

Why does the SMART overall status say PASSED when there are bad sectors?

Because the manufacturer's threshold for the normalized value is set low enough that only severe degradation trips it. The overall status is a last-resort alarm, not a health grade.

Can pending sectors go away on their own?

Yes. A pending sector is cleared when the address is written. If the write succeeds the count drops with no reallocation. If it fails, the sector is remapped and attribute 5 goes up instead. Either way the pending count returns to zero.

What we do at Atomic Racks

Every used drive we sell gets a SMART extended self-test, a fio benchmark and NIST SP 800-88 sanitization with read-back verification. Attributes 5, 197 and 198 must be zero on SATA, and SAS drives follow the grown-defect rule above. Hours are read from the drive and never reset, and units are priced in power-on-hour tiers. The QR tamper seal opens each unit's public test report. 90-day warranty on standard items, up to 2 years on certified tiers, 30-day returns, free shipping to the lower 48. See how we test.