Skip to main content
Base rates for reasoning about HDD risk, and the recording standard that depends on them. The purpose of this page is to replace adjectives with numbers.

​
The rule

No drive record in any repo calls a disk “dying,” “worn,” “end-of-life,” or “on borrowed time.” Those are predictions written as observations. Record the measured values, cite a base rate from this page, and let the reader do the arithmetic. This is not stylistic. Age-based intuition about disks is mostly wrong. A drive at 6–7 years is not end-of-life stock to be replaced individually; the largest public dataset does not support that read.

​
Source data

Backblaze Drive Stats Q1 2026 — 341,263 drives, 30,203,180 drive-days in the quarter, 529,968,464 lifetime. AFR (annualized failure rate) is the share of drives expected to fail over a year of continuous service.

​
Failure vs. age

The finding that matters most, from Are Hard Drives Getting Better? Let’s Revisit the Bathtub Curve: The age at which the failure rate spikes moved later by roughly two years between those cohorts. Backblaze’s own conclusion is that the classic bathtub curve no longer describes their population. In their words: “the neat story of early failures, calm middle age, and gentle decline no longer fits the world our drives inhabit.” Consequence. A 6–7.5 year old drive is not self-evidently near failure. It sits inside the band where a modern cohort still shows a low per-drive AFR. That band runs at or before the age where the 2021 cohort’s rate peaked. Age alone is not a defect, and it is not recorded as one.

​
What actually predicts failure

Backblaze and Google independently identified a small set of SMART attributes that correlate with impending failure. Backblaze uses five, chosen because they are consistent across manufacturers: Google’s analysis found attributes 5, 187, 197, and 198 consistently correlated. It also reported a finding about a drive’s first scan error, meaning SMART 187 going non-zero. After that error, the drive is 39× more likely to fail within 60 days than a drive with none. Two limits, because they change how strongly these may be cited:
  • Both analyses are univariate, one attribute at a time. They establish correlation, not a calibrated per-drive probability.
  • Not every model implements every attribute. A missing attribute is not a reading of zero. Where one is absent, say so and substitute the evidence that does exist, normally a completed extended self-test.

​
Recording standard

  1. Record measurements, not prognosis. Power-on hours, the five predictive attributes, self-test outcomes with their LifeTime hour, temperature.
  2. Read self-test results from smartctl -l selftest, never the -c summary. -c reports “completed without error” even when no test has run; confirm the LifeTime hour advanced.
  3. State absence explicitly. “Attribute 197 not implemented on this model.” Never silently imply zero.
  4. Cite a base rate rather than an adjective. “65,281 h, all five predictive attributes zero or unimplemented, extended self-test completed without error” beats “high-hours refurb nearing end of life.”
  5. Redundancy is the control, not drive selection. A non-zero population AFR means any drive can fail at any age. That is what raidz and off-box copies absorb. Replacing healthy disks on an age heuristic is not a protection strategy; see data protection classes.

​
Applying it

Two cases from this lab, both measured rather than estimated. A drive kept in service. A 6 TB SATA unit at 65,281 power-on hours (7.5 years) reads zero on every predictive attribute it implements. It also completed a full extended surface scan without error, hours before entering service. Attribute 197 is not implemented on that model, so the completed scan stands in as that evidence. Nothing measured on it indicates a defect, and its age sits below the 2021-cohort failure peak. It was put into a raidz1 vdev. A drive removed from service. A different 6 TB unit reached 56 Current Pending Sectors with 0 reallocated and 0 offline-uncorrectable. Attribute 197 is one of the five predictors. That makes this an evidence-backed reason to pull a disk from a pool that had no redundancy to spare. Note what that does and does not establish. A non-zero 197 raises failure probability materially, but it does not mean the drive is dead. And 0 reallocated plus 0 offline-uncorrectable means the platters had not lost data. Retest standalone before any disposal decision. Correlated age is a vdev property, not a drive property. Members bought together and run together fail less independently than the base rate assumes. That matters during a rebuild, when the surviving disks do their heaviest reading. That is an argument for rebuild-window planning and verified restores, not for calling any individual disk suspect.