URE - Unrecoverable read error

[gustomucho]

Hello,

I am in the processing of migrating all my data on a brand new Unraid server.

Having read around a bit, I find myself a bit worried about the dreaded UREs.

 

The spec sheet of my disks (WD reds) states that the reported URE rate is < 1x10^14.

Since I am using 10TB and 12TB disks with 1 disk parity, using aggressive figures (and not conservative ones), the way I understand it is that in the process of rebuilding a disk from parity, the odds of encountering a URE are fairly high.

 

What would happen in the event of a URE during a disk rebuild? Would the rebuild process completely stop, or would some of the data simply become corrupted? If the effect is data corruption, how much of the data would become corrupted? Would it only be 1 byte, effectively making one single file corrupted?

 

I have no idea if I should even be worried about this, which is why I am asking for help from you guys

Does ECC RAM play any role in this?

Would the addition of a second parity drive decrease the possibility of it happening? I would rather not have to do this for monetary considerations.

 

I know I've asked alot of questions already, but I would appreciate it if someone with knowledge or experience about this could enlighten me some.

 

Thanks

Edited February 23, 2021 by gustomucho

[JorgeB]

6 hours ago, gustomucho said:

What would happen in the event of a URE during a disk rebuild?

The rebuild would continue, with single parity there would be some data corruption, disk sectors are usually 512B, so that would be the minimum corruption size, or possibly it would affect a block, which is 4KB.

 

6 hours ago, gustomucho said:

Does ECC RAM play any role in this?

No, but it's also good to have if you care about data integrity.

 

6 hours ago, gustomucho said:

Would the addition of a second parity drive decrease the possibility of it happening?

Yes.

[itimpi]

It is worth remembering that the quoted URE value is for I/O operations, and each sector is 512 bytes.   The likelihood needs to be based on the sectors on the drive rather than its capacity.

[gustomucho]

That's very good to know.

The risk of URE I had calculated is much lower if the drive size has to be divided by 512 bytes.

 

Thanks for the answers. I'll hold on a little while longer before adding a second parity.

Edited February 24, 2021 by gustomucho