SmartData and its values according to BackBlaze.

[kling]

Interresting article regarding drive failures and smart monitoring.
https://www.backblaze.com/blog/hard-drive-smart-stats/

I'm not smarter after reading the article, are you?

 

[cyberjock]

I'm not any smarter, but I could have written that article just based off of my own personal experience with SMART. :P

If you don't know about that stuff it's not a bad read. I would have written it a little better and differently, but those metrics are vague and obviously subjective. The ones they list ARE monitored by smartd and they WILL trigger emails if you have all of the necessary stuff setup for that to happen. I have gotten them (I get them as a text message because IMO they require immediate review and action).

I will elaborate a little on the phrase "The dirty industry secret? SMART stats are inconsistent from hard drive to hard drive." I think that what he is saying is that some SMART parameters that involve "rates" are not consisten. Seagate and Hitachi use error rates as a floating point integer value that reflects realtime. The first few digits represent a decimal and the last few the power of 10. With the other companies they assume that anything below a given value is "zero" and list it as such. Of course we know that things that are measured in "rates" are almost never zero, and hard drives are definitely never error free as the hard drive manufacturers freely mention the error rates in their statistics for the drives. Seagate and Hitachi may have some raw value like 13745479856, which may translate to some ultra low error rate, but on WD that would be a drive that just isn't functioning because of what the number means. In conclusion, their sentence makes it sound like SMART isn't "an established standard that can be trusted", but that's not the case at all. SMART is very good, should definitely be trusted, but you also have to know how to interpret the numbers for the hardware you own.

SSDs add even more trouble because SMART is somewhat useless for them. SMART wasn't designed for solid state drives quite like it was for platter based drives. It was designed to monitor key aspects of hard drives that have moving parts.

 

[solarisguy]

It would be nice to know what values of which SMART metrics guarantee manufacturer's warranty replacement...

 

[cyberjock]

It would be nice to know what values of which SMART metrics guarantee manufacturer's warranty replacement...

A failed SMART test is grounds for RMA.

 

[titan_rw]

A failed SMART test is grounds for RMA.

I had to argue (insist) one time when I had a WD Red that had failed a long test. This was in our non freenas nas that we use at work. I scheduled automated long tests because none had ever been performed by the previous admin.

A few smart tests later I get an email that one of the 4 drives failed a long. Call up WD to apply for an RMA, and he says something like, ohh but you can fix the failed smart test by dong a "low level format". Nevermind the fact it's been impossible to low level format a drives since MFM drives back in like 1980. I assume he meant a full surface drive zero, which is certainly not a 'low level format'. And possibly yes, a zero pass may have forced the problematic portion of the drives to start working again, but on a drive that's under warranty that is reporting it's failing a diagnostic test, why even bother trying.

I was insistent that the drive needs replacing as it couldn't be trusted anymore even if the zero pass did work. We'd never continue using it for reliability concerns. He eventually did go ahead and issue an rma, and I got it replaced. (I even printed off and shipped the smartctl -a output, highlighting the problem attributes / failed long tests)

I know Cyberjock doesn't like Seagate, but in my dealings with rma, they've been the easiest to deal with. It's only been WD I've gotten hassled about getting rma's.

That being said, when my seagates fail, I'm replacing them with WD's as I do think they are the better nas drive.