[OPLINTECH] If you heard a load noise about 7:20 this morning.

[Chad Neeper]

This was a set of offlist messages that I'm responding to on-list because
they're relevant. I've removed the identifying info from the person
responding to the OP since I didn't ask for permission. But then again, I
didn't ask Bob's permission either. Tough. Deal with it.  ;-)


Just to clarify/reinforce this a bit, don't use anything other than
Spinrite Level 1 or 2. DEFINITELY NOT higher than 2 though. Level 3 and
above will write to the drive, which is detrimental to SSDs. Essentially
the reads Spinrite does will trigger the SSD's internal error correcting by
forcing it to read each part of the SSD. If possible, any bad spots
detected by the ssd will be relocated or re-written.

What I've come to find as the downside to SSDs vs traditional spinning
drives though is the way they tend to fail. So far of the handful of SSDs
I've had fail, every one of them was working one minute and then not the
next. Not a single one has faded out slowly with bad sectors. At least with
traditional drives sometimes you get a drive that up and dies on you, but
more often than not its the platters that wear out and Spinrite can come to
the rescue. Doesn't seem to be the case with SSDs. All told, though, I
still hugely prefer SSDs over traditional drives.

About memory tests:  I've learned a few things about RAM and come to
respect it even more. For starters I use memtest86+ to test RAM and let it
run for as long as possible, preferably days, but at least several passes.
memtest86+ uses various patterns to test the bits in various ways. This is
necessary because it's possible for a single pattern to actually miss bad
bits. For instance, a simple example would be writing all 11111111's out
could miss the fact that a bit is stuck on 1 and won't flip to 0. So, the
obvious thing to do is then write out all 00000000's? That could catch out
a bit that's stuck on 1. But what about a bit that's stuck to it's
neighbor? memtest does a really good job at testing the RAM, but even a
current computer with a current normal amount of RAM can actually take
hours to complete a full pass of all the various pattern iterations
necessary to adequately test the RAM. It's also quite possible that a bit
will be ok on the first couple of passes but then fail on a later pass.

When we think of RAM failures, we tend to think of the more obvious
failures...like when Windows BSODs or a kernel panics because the RAM has a
significant enough patch of bad bits in the right spots that an important
binary wigs out. But it can actually be a lot more insidious than that.
Sometimes, the bad bits are just "insignificant" enough that the OS doesn't
obviously fail, but bad bits are still creeping into your environment. It
could be a long time before something really bad happens that makes you
think there might be a hardware problem. In the meantime, that
uber-important spreadsheet which you neglected to back up is slowly getting
clobbered. Those are the worst. It's also why ECC RAM is so important on
servers. The extra bits help to detect any stuck bits and halt the system
before the bad data creeps in.



>From Wikipedia on Spinrite:
Spinrite can be run and can be effective on SSDs, but running in a
higher-level mode than 1 or 2 is detrimental, as it wears the SSD by
writing to it unnecessarily. In episode 194 of the podcast Security Now!
<http://en.wikipedia.org/wiki/Security_Now!> Gibson said that he could "see
absolutely no possible benefit to running SpinRite on a solid-state drive"
and later "SpinRite is all about mechanics and magnetics, neither of which
exist, by design, in an SSD".[10]
<http://en.wikipedia.org/wiki/SpinRite#cite_note-10> In episode 338 Gibson
clarified "it is actually detrimental because [solid-state drives] don't
like to be written", but also pointing out that a read-only run could be
beneficial: "SpinRite's Level 1 is a read-only scan, and doing that on an
SSD makes a lot of sense. Do a read-only scan of an SSD, it'll show the
SSD's controller that it's got a problem reading a sector, and then it'll
map that out or rewrite it in order to strengthen that sector, if possible.
So that ends up being a value for SpinRite on solid-state drives." [11]
<http://en.wikipedia.org/wiki/SpinRite#cite_note-11>Also, Gibson posted on
his website that "SpinRite is seeing many successes [...] with non-spinning
solid-state (thumb) drives!".


Hope someone finds this useful.
Chad


______________________________
*Chad Neeper*
Senior Systems Engineer

*Level 9 Networks*
740-548-8070 (voice)
866-214-6607 (fax)

*Full LAN/WAN consulting services -- Specialized in libraries and schools*


On Thu, Jul 31, 2014 at 12:02 PM, Bob Neeper <neeperro at oplin.org> wrote:

>  Good suggestion about SpinRite, in read only mode.
> Not sure if it would work though if the hard drive was truly not found by
> the BIOS.
> Had some mechanical drives that I don't think it could see.
>
> Anyway I did another data backup, ran the Intel SSD Toolbox which showed
> the C: part as good.
> Half the SSD is Truecrypte'd and Toolbox doesn't like it.
> Updated Toobox and ran it again.
>
> Rebooted and all is okay.
>
> I will pretend it was a fluke due to having to force the laptop down last
> night for some reason.
>
> But I think I'll try to find my original HD tonight and image it.
>
> Bob
>
> On 7/31/2014 9:19 AM, someone wrote:
>
>  Oh no!!!! Did you try running SpinRite on level 2 or 3? It supposedly
> can repair errors even on SSDs… When watching the memory test, I invariably
> have the same thought: “This is silly.”
>
>
>
> Drink coffee, and cross your fingers!
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.oplin.org/pipermail/oplintech/attachments/20140731/8fc41180/attachment.html>