[kwlug-disc] SMART for SSDs

[B.S.]

----- Original Message -----
> From: Andrew Kohlsmith (mailing lists account) <aklists at mixdown.ca>
> To: KWLUG discussion <kwlug-disc at kwlug.org>
> Cc: 
> Sent: Thursday, July 30, 2015 12:10 AM
> Subject: Re: [kwlug-disc] SMART for SSDs
> 
>>  On Jul 29, 2015, at 11:59 PM, B.S. <bs27975 at yahoo.ca> wrote:
>>  It also notes that SMART will alert you LONG before the drive really dies - 
> years even, at a couple TB / year? And that when it dies ... it's really 
> dead. One last reboot and it's a brick. As Lori notes, the reallocated 
> sector count seems to be the real tell.
> 
> Is this SSD-specific? My experience with SMART and spinning platters is that if 
> SMART reports the drive is fine, be wary. I’ve never ever had a spinning platter 
> SMART warning flag a failing drive, only my RAID array saying that one of the 
> drives has failed because the kernel dropped it.
> 
> This is based on a good decade worth of smartd running on drives from all the 
> major vendors. Maybe they’ve finally got it working well with SSDs, but I don’t 
> trust SMART. Not one bit.

Others on this list deeply share your mistrust of (spinning platter) SMART, especially for false positives. I don't believe there is a problem with the intent, merely It Just Doesn't Work (tm). (As an effective tool / predictor / implementation.) And I don't dispute / I do trust their opinions.

However, the one takeaway that has seemed solid to me is ... if it doesn't give warning, you're largely OK. (At least as OK as with no smart at all.) But if it does give warning ... beware the high likelihood of it being a false positive. And better a false positive than no warning at all - you may unnecessarily or prematurely take corrective action, but better than the reverse. And one can run further cross-verification checks. (But I forget what they might be.)

If this isn't a reasonable take on it, I'd sure like to know.