Some questions about "predictive failure" hard drives on Smart Array P400

[ivc]

Hello.
I know, there are a lot of topics with the same sympthoms, but my question is not "how to replace the hard drive?". 
I've got a Proliant DL380 G5 server with Smart Arrray P400 controller.
There are 8 SAS drives (HP EG0146FARTR) in RAID 1+0.
During 2 years I has 3 drives with "Predictive to fail" status in SMART. HP's support adviced me to replace them. They send me new drives (as a warranty).

Now I have 3 drives with "Predictive to fail" status again. I don't wanna change them for few reasons: mail in Russia works really bad, warranty period will over soon, and it seems like problem is not in the drives.

Now I'm asking for help to understand the reason of "predictive to fail" status of the drives.

So, during all the 3 years I has 3 drives failed in Bay 2 (!) and 3 other drives in Bay 7, Bay 6 and Bay 3.
OS (Windows Server 2008 Standard x64) is running and working good.

I think there are two possible reasons: SMART data is not true or something wrong with P400 controller.

After drive's and controller's firmware update status was not changed. ACU version is last.

But I found something in ADU report.

Physical Drive (146 GB SAS) 1I:1:6

Last Failure Reason: Hot Removed (0x14) Hot Plug Count 3 (0x00000003)

It was never hot removed! I'm always change it with power off. The same with drive in Bay 3.

SMART Predictive Failure Errors Are Recorded In The Factory MP Data For This Drive

SMART Predictive Failure Errors Are Recorded In The Since Power MP Data For This Drive

What is Factory MP Data?

There are no read errors since firmware update.

Read Errors Hard (Since Factory): 0x0005f9aa

Read Errors Hard (Since Reset): 0x00000000

The same with drive in Bay 3.

Perfomance test is OK for all drives.

Physical Drive (146 GB SAS) 2I:1:2

Last Failure Reason: Aborted Command (0x0e)

What it means?

Read Errors Hard (Since Factory): 0x00000000

Predictive Failure Errors: 0x000034e5

If there are no read errors, so what's predictive failure errors?

Last question: Can I clear SMART data?

Thanks!

[Torsten.]

In general:
A disk is always a bit bigger the specified, so it has room to relocate bad blocks. It also has always a couple of bad blocks since it was produced, they will never be used. It is allowed to have a couple of r/w errors, but if a threshold is reached, the controller will predictive fail the disks, this means, there are too many errors already on this drive. Smart data is written by the disk itself, this means the disk itself detected the errors and noted them down.
IMHO read errors are bit more serious, because if the disk cannot write to a block, it will write to another. If data cannot be read, you have a problem ...