When exactly are drives available for warranty service?

[dtemp]

I'm wondering what SMART parameters the different hard drive manufacturers use to determine if your drive is bad enough to qualify for warranty service. In other words, if you RMAed a suspect drive, and it didn't meet any of the criteria, they would just send the drive back and tell you it's your fault.

Would they accept an RMA if any of the critical SMART attributes (5, 196, 197, 198) were non-zero? Instead, would the raw values on those critical attributes need to be above the "THRESH" value (which seems to be 36-140 for my drives for #5 and always 0 for #196-198)? Would there need to be errors on a long SMART test?

Are the answers to the above different for each manufacturer?

 

[Rand]

Well I would assume that the manufacturers with own analyzing software will take that software's verdict as reference;
I know Seagate does (or did).

Parameters will o/very likely vary by tool; I am not sure whether those just query smart or go to chip directly and get more info from it...

From what i know of Toshiba (who dont have their own software) it depends on whether you have warranty at all - if yes the are told to be fairly relaxed ; if not then they wont exchange at all. Havent tried it yet;)

 

[dtemp]

I guess my question might be too exacting when what I'm looking for is anecdotal evidence on what people have successfully RMAed. I don't know if reallocated sectors needs to be in the dozens or hundreds or thousands for them to care, for example.

I don't have a failing drive at the moment, although I know how drives behave as they near the end of their life, and sometimes you like to submit a drive giving you errors before it is actually kicked out of the pool.

 

[Rand]

As i said, from my experience -
if a vendors software says drive is ok then no repair.
If it says failed - replacement. Thats what Seagate does (or did).

 

[dtemp]

Ok. So you are relying on vendor tools for whether it is due an RMA or not. I know the SeaTools and other programs exist; I'm trying to take a shortcut to avoid using them. I don't have a desktop I can stick a drive in and reboot, I would have to restart my ESXi server into some boot disk to check a drive, which would involve shutting down all my services, etc. I'm also not sure if those tools modify the disk, and makes them need resilvering if they aren't RMAable and you choose to keep them in the pool.

It would be nice to know what SMART figures make a drive more-than-likely RMAable without needing to test them with manufacturer software. I know I've RMAed totally hosed Seagate disks successfully without testing them first, although those disks weren't even close to borderline. Thanks for your notes though.

 

[Rand]

Ah.
From my experience no modification is done on the disks. I was able to read ok disks to linux raids without issue (havnt checked zfs though).
And sorry, not really sure which paramters make it more likely for an actual issue to be present beyond those you mentioned.
I dont think any non zero value will do - massive increase or custom higher threshold are likely needed, depending on specific disk (consumer, enterprise grade)
Manufactureres don't want to replace after all;) They just have to sometimes.

 

[solarisguy]

@dtemp, if what you had asked for was possible, then almost everybody would be running RAID-Z1 and RAID-Z3 might not even exist ;-)

 

[dtemp]

@dtemp, if what you had asked for was possible, then almost everybody would be running RAID-Z1 and RAID-Z3 might not even exist ;-)

I don't care about the drive's actual propensity to fail; I can already gauge that from the SMART parameters. All I care about is quantifying when the drive can be RMAed or not ;)