[TriLUG] network storage drive spin up policy: how often for longest life? And can the SMART parms be hacked (reset)?

[David Burton via TriLUG]

Your kitty school taught you well, Mauricio.

People often make that mistake with RAID arrays. They buy *identical* new
drives, with the *same* date of manufacture, made in the *same* factory,
and they subject them to nearly identical usage patterns, and then they're
surprised at how "unlucky" they are when two of the drives fail on the same
day, and they lose the data from their "redundant" RAID array.

The principle doesn't apply merely to computer storage, but to many other
things, as well. If you build something with redundant parts, for
reliability, then you should try to ensure that the parts are *different*.

For example, if you use redundant bolts, to make something stronger than it
needs to be, for increased reliability, you really should get your bolts
from different sources, in case one batch is substandard. Unfortunately,
hardly anyone does that.

If you take apart a coffee pot, these days, you'll typically find *two*
thermal fuses, connected in series. My understanding is that the reason for
that is that coffee makers were burning down people's houses, when the
thermostats stuck, and the thermal fuses failed to work. Here's an article
about it:
https://www.electrical-forensics.com/Coffeemakers/CoffeeMakers.html

To avoid killing customers and their families, and to avoid lawsuits,
coffee maker manufacturers started putting *two* thermal fuses, in series,
in their coffee pots.

However, for that to really improve the protection offered, the two fuses
*need* to be *different*: different batches, certainly, and preferably
different brands or specs.

But they usually aren't. If you take apart your coffee maker and you'll
probably find two *exactly* *identical* thermal fuses, with a high
likelihood that they were made on the same assembly line, on the same day
-- which means that two are probably no better than one. (About half of all
engineers graduated in the bottom halves of their classes.)

I had a Mr. Coffee fail, and out of curiosity I disassembled it, to find
out why. I checked the two thermal fuses, expecting to find that one of
them was "open," and I was surprised to instead find that they *both* were
open.

Think about that. As soon as *either* of them opened, all power was cut to
the appliance, yet they had *both* opened. That means those two thermal
fuses must have been well and truly *identical*.

Fortunately, they were functional, and identical. If they'd been bad and
identical, then a stuck thermostat might have burned down my house.

If you're an engineer, and you want to ensure that two parts are not
identical, so that they cannot possibly be identically defective, then to
foolproof the manufacture of your product you need to spec the two parts
differently. So the Mr. Coffee designer should have spec'd two slightly
different temperature ratings for the two thermal fuses, and the bridge
designer should design with slightly different bolt lengths for the
redundant bolts, just to make sure that they cannot come from the same
(possibly embrittled or otherwise defective) batch.

That's why my doubly-mirrored RAID 1 server has two *different* brands of
drives in it (and the two drives of the same brand have substantially
different dates of manufacture).

Dave


On Wed, Dec 2, 2020 at 1:27 PM Mauricio Tavares via TriLUG <
trilug at trilug.org> wrote:

>       ... in kitty school told us to never get all drives from the same
> batch (multiple vendors
> or multiple orders might address that)
>