Random problems with filesystem corruption.

[Rashkae]

On 2019-12-15 3:14 p.m., Ralf Mardorf via ubuntu-users wrote:

> However, in my experiences with spinning disks, SMART data was more or
> less useless. I can't comment how useful it is for SSDs, since none of
> my SSDs failed.
> 
> 

Smart data can actually predict over 70% of drive failures in advance.
If you search around for a Blackblaze report on "What Smart tells us.",
they lay it out all well.  I don't have exact statistics of my own to
share, but it has certainly been my experience over the years and most
drive failures can indeed be spoted in SMART ahead of any other failures.

The key here is to look at the actual values.  The Smart "Failed/Passed"
metric used by manufacturers is completely useless.  By the time a Drive
actually admits it has failed, you'll be lucky to get data off of it.


So far, I've only experienced 1 SSD Failure, but it as well showed up as
errors in the SMART data before any complete failure, so that's a good
start... (but SSD's have been comparatively very durable so far, so
that's also good news.)


Job 1, of course, is always to make sure you have a good back-up
strategy, since drives can and *will* fail, and the only question is how
long they can go before they do.  But periodically monitoring SMART will
save you a ton of time and headache by allowing you to replace most
drives when they are only start to suffer.