Why SMART is passing while it indicates failure for RAW read error rate?

Question

I have an external USB disk (WD Elements Caviar Green 2 To, WD20EADS-00W4B0) that started to fail (slow reads). My system logs didn't show anything unusual, this definitely indicates some supposedly serious problem.

To investigate, I've installed smartmontools, and checked SMART status. SMART passed but indicate the following issue:

...

SMART overall-health self-assessment test result: PASSED

...

1 Raw_Read_Error_Rate     0x002f   001   001   051    Pre-fail  Always   FAILING_NOW 64414

...

I've already started to clone my data (not before because this data is not highly critical), and will replace it in a few days. But I don't understand why SMART indicates passing, and not a WARNING instead since I believe the drive will soon die. Can you explain this to me ?

As a side question, would it be safe to reuse the drive after low-level formatting ?

Solution

Keep in mind, SMART is not a Pass/Fail kind of thing, but a set of stats that together can give you a general idea of a drives health. as such not all stats are created equal. Some are critical (high reallocated sector count) where as other are not so important.

Since this is the case, anyone who writes smart analysis software is making a decision as to which stats are important and which aren't, based on the stat itself and its value.

in this case, yes, I agree a Raw read error rate indicating a pre-fail condition is definitely cause for concern, and I would not trust that disk for much. I think the smartmontools folks should have flagged it as important, but they obviously choose not to.

No, this is a permant error (almost all SMART errors are permanent), so formatting should have no impact on the issue. per this article, this error pertains to either the RW heads, or the disk platter surface, both of which would require replacement: http://www.ariolic.com/activesmart/smart-attributes/raw-read-error-rate.html

Comments

[ybart]

My understanding is that the disk itself reports the verified SMART status (it is reported as self assessment). However, I assume it will completely fail in a few days or weeks, so I won't reuse it at all. I will read more documentation and comment.

[EKW]

For the record, SMART can be a pass/fail, depending on the manufacturer and how they choose to configure it. Unfortunately, many manufacturers have excessively 'generous' failure thresholds, thus SMART often doesn't 'fail' until the drive is dead. Some SMART utilities ignore the set thresholds entirely, substituting their own limits during analysis.

[Frank Thomas]

@EKW, it is true that software can and does make pass/fail assumptions based on SMART statistics, and sometimes that software operates at a low level and may disallow user choices because of them, but that is still software making its own choices on how to react to the data. SMART info is just data, so it doesn't make any choices on its own. Obviously, the SMART data attempts to reflect the health of the drive, but like all data attempting to model reality, it doesn't match perfectly, so, yes, when the drive mechanically seizes, you could call that fail, but its not failing due to SMART.