I have two brand NEW Hitachi 2TB drives (0F12117) and a few other drives that were laying around.  I was going to add the Hitachi drives and one of the 6 month old (Samsung 1TB) drives as a Cache drive.

 

One of the brand New Hitachi drives failed the pre-clear and one was ok.  The Samsung drive, which has about 40 hours on it also failed.

 

Does this really point to hard drive issues or is it more likely a cabling, sata port or ram issue.  I have had several older sata drives fail the pre-clear and I ended up selling them on Ebay for cheap because I didn't want to put them in the array and I had no other use for them.

 

I've previously run the memtest for 48 hours and had no errors, although that was about 3 months ago.

 

Just wondering if hard drive quality has really gone to hell or unRAID is really picky

I've heard 1 in 5 fail.

Just wondering if hard drive quality has really gone to hell or unRAID is really picky

Both probably... or rather, it is not unRAID that is picky... but us, as users of a data storage system, we are very picky.

(we want the data  read from the array to be the exact same as what we wrote)

AND

the drive capacity has grown to the point where I would be absolutely shocked if any drive is thoroughly tested before leaving the factory.  The electronics is tested, and some basic reading and writing has to occur to put the initial low-level sectors on the disk, but that is not the same as a full surface test similar to what we are doing in a "preclear"    Being "first-to-market" with a larger disk will mean the first few generations of disks are not as reliable as any older and more mature technology.  (translation... initially, do not expect 3TB and 4TB drives to be more reliable.)  Errors are far more likely to be detected soon after a drive is first put into service simply because bit density on the platters is so much higher and mechanical tolerances way smaller.

 

Basically, the manufacturer simply cannot read and write the disk much faster than we can, and there is no way they'll spend several hours burning in and testing every single drive.  It is FAR easier (and cost them far less) to just let us use the disks and deal with the cost of the RMA if they fail. 

 

Joe L.

I know drive errors are not that uncommon and bad sectors occur all the time.  Is it that unRAID can't compensate for these bad sectors? I always assumed it was the firmware on the drive that marked these sectors as bad and unusable protecting you against writing data to the these sectors.

 

Am I wrong in this assumption? If I am not, then why do bad sectors matter if they are marked as such?

 

I'm just trying to determine if it's absolutely imperative that I use 100% perfect drives.  Manufacturers will not RMA a drive unless there are enough bad sectors and I'd hate to have to keep playing the HD jackpot game, buying new drives until I find a pristine one.

Who say's unRAID can't compensate?

The disk does reallocate and unRAID does compensate to an extent. But if a drive cannot pass a pre-clear it will die once in your array after your precious data has been stored on it. A drive that fails the pre-clear can be RMA'd. If a drive looks marginal after a single pre-clear then a couple of more pre-clear passes can ready it for RMA.

 

A drive with a single reallocated sector is no big deal. But if the number of reallocated sectors keeps growing the disk will fail. Disks which fail in this manner can cause random unexpected problems in your server.

I know drive errors are not that uncommon and bad sectors occur all the time.  Is it that unRAID can't compensate for these bad sectors? I always assumed it was the firmware on the drive that marked these sectors as bad and unusable protecting you against writing data to the these sectors.

 

Am I wrong in this assumption? If I am not, then why do bad sectors matter if they are marked as such?

They are marked bad when they cannot be read.  Typically, they are not read until AFTER you've written them with your data.  By then, if it is your only copy of data, it is too late.

 

I'm just trying to determine if it's absolutely imperative that I use 100% perfect drives.  Manufacturers will not RMA a drive unless there are enough bad sectors and I'd hate to have to keep playing the HD jackpot game, buying new drives until I find a pristine one.

No, it is best if they are good from the beginning, and if you have a parity drive installed, and if a read error occurs, unRAID will re-construct the un-readable sector by reading all the other disks in the array AND it will write the un-readable sector to the drive with the faliure allowing its SMART firmware to re-allocate the sector it could not read.  It is self healing in that respect.  (to the extent that the drive has enough spare sectors)  If you see the re-allocated sector count continuing upward, it is a time to RMA the drive.

 

So... unRAID does "compensate" for the occasional bad sectors.

I'll have to attempt to re-clear those drives and take a closer look at the report to see what the problem is.  One of the drives was thrown back into a window system and has been working fine.  If a pre-clear failed drive will RMA, I'll have to see about that, it's still under warranty.

One of the brand New Hitachi drives failed the pre-clear

Would you please elaborate on what you mean by that statement?  Do you mean that the drive would not complete a pre-clear cycle?  Or do you mean that the drive did complete the preclear operation and you did not like the results that were displayed (i.e. read error rate, reallocated sectors, etc.)?

 

I'll have to attempt to re-clear those drives and take a closer look at the report to see what the problem is.

This leads me to believe that it was the latter.

 

One of the drives was thrown back into a window system and has been working fine.

Preclear is just a tool that you can use to evaluate the health/integrity of a HDD.  And the important part is that this evaluation is independent of whatever OS you will be using the HDD in.  Just because you installed a drive that showed "isues" in preclear into a windows system does not mean that the drive all of the sudden fixed itself.  The drive only appears to be "working fine" because you aren't seeing the level of detail in windows that you got from the preclear results. I think the phrase "Ignorance is bliss" would be appropriate to describe this situation.

I mean that when I come back after the 20 hours or so it says the drive FAILED the pre-clear with a few errors listed.  When I re test I'll save the logs.

 

When I say the drive is working find under windows I mean just that.  Just as a test I put it into a mirror under a 3ware 9600 card and filled the drive.  Ran a integrity test and it passed.  Ran a test using Hard Disk Sentinel and it passed as well with zero errors. 

 

As far as being "ignorant"; I suppose if I am failing to acknowledge that the pre-clear script is the end-all be-all of disk testing apps is being ignorant, then I guess I am.

Hard Disk Sentinel is just reading the SMART values. Unless you did a destructive surface scan, which should take as long as a pre-clear if done correctly.

Just a wild guess but have you tried to preclear the drives on a motherboard based on Intel or AMD chipset (not nVidia, VIA, ALi, SiS etc.)