SMART said PASSED with 2,184 uncorrectable errors
The hard drive in my NAS says PASSED. Then you read the attributes:
| Attribute | Raw |
|---|---|
5 Reallocated_Sector_Ct |
26 |
187 Reported_Uncorrect |
2184 |
197 Current_Pending_Sector |
0 |
198 Offline_Uncorrectable |
0 |
9 Power_On_Hours |
11198 |
187 is Reported Uncorrectable Errors: 2,184 of them. Its normalised value had bottomed out at 001. Backblaze tracks it as one of five SMART stats that point to a failing drive, and the overall status still says PASSED.
187 is a lifetime counter
It never goes down, so it can’t tell you whether the drive is failing now or failed once, years ago. 197 is the present-tense one, sectors pending right now, and it was 0. So was 198.
So I tested it, three ways, on the box itself so the network wasn’t the bottleneck:
- write 8 GB of random data through the filesystem, drop the cache, read it back, compare SHA-256
badblocksover every sector of the raw device- a SMART long self-test, the firmware’s own full surface scan
All three passed, and none of the counters moved between before and after. Seven weeks later, after copying about 246 GB from a Windows drive onto the NAS, they still hadn’t.
The rule
So:
- it holds bulk and second copies only: downloads, media, scratch. never the only copy of anything.
- the SSD is the disk to trust
smartctl -t longevery few months, and retire it the moment 197 goes above 0 or 5 starts climbing