• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Unrecoverable error....

S

SomeGuy133

2[H]4U
Joined
Apr 12, 2015
Messages
3,447
So can someone elaborate on what it is and how it works. You run a chance (making this up) 1 unrecoverable error for every 6 TB written so you may run across that when you do a rebuild with 6 TB drives. (this is an example not saying numbers are accurate. Wanting to understand how it happens (the concept)).

Is this error on a single bit or the whole drive?

Is it only a single bit or is it a sector?

Is the rest of the drive sound except that one spot?

If it is only a single bit/sector do hardware RAIDs or software RAIDs know to grab that single bad error from a different parity? Or are they dumb and assume the drive is SOL and attempt the next drive which runs an equal chance of being SOL from a single error?

It seems kinda stupid that they can't just grab from a different parity the single file/bit that is needed. If that is the case it leave a huge room for failed builds for such a small avoidable error.

If you have 6TB drives and its URE is 12TB that is a 50/50 chance of failing so with 2 drives you still run a dangerous chance of being SOL if it is dumb like the above situation BUT if it can simply grab the single bat spot from a different drive the chances of 2 drives being SOL goes from a 50/50 to like 1 in 4.0e12 or is it even lower then that? See the huge difference? I want to know how this works so i can plan accordingly in my build.

Again as always thanks for the help guys!!!!
 
You run a chance (making this up) 1 unrecoverable error for every 6 TB written so you may run across that when you do a rebuild with 6 TB drives
Click to expand...

I have found in practice of raid scrubs of 70+ TB each week for many years that the 1 unrecoverable bit per 12TB of data is off by several orders of magnitude for tested working drives. For me I consider a working drive a a drive that passes a 4 pass badblocks without a single bad block reported (takes around 50 hours for a 4TB drive). What I see (and have seen for over a decade) is working drives will have a much lower error rate ( somewhere on the order of 1 in 500 TB written) while drives that are about to die will have a much higher URE rate ( 1 in several hundred MB at the worst ).


Is this error on a single bit or the whole drive?
Click to expand...

Its a single corrupt sector per X amount of bits transferred.

Is the rest of the drive sound except that one spot?
Click to expand...
If the drive is not about to die I expect to not see any more UREs for a full drive read (well actually many given my observed corrupt sector rate). If it is on the way out you may see 1000s of corrupt sectors in a full pass.

If it is only a single bit/sector do hardware RAIDs or software RAIDs know to grab that single bad error from a different parity?
Click to expand...

Yes. If the drive reports an error the raid will know how to fix it provided it has enough parity. If the drive does not report the error (silent error - possibly caused by a bug in the drives firmware) the raid should be able to recover with no dataloss if it has more than 1 parity for example raid6.
 
Last edited:
coudnt the drives built in error recovery be able to recover this data? Especially with the new 4K sectors where they use a much longer error recovery string of like 32 bits/sector I think instead of 8 bits for 512bytes..
 
AlienTech said:
coudnt the drives built in error recovery be able to recover this data? Especially with the new 4K sectors where they use a much longer error recovery string of like 32 bits/sector I think instead of 8 bits for 512bytes..
Click to expand...

A URE will only happen when the ECC of the sector is not enough to recover the error. Meaning more bits of the sector are unreadable than the ECC protects against. I just edited the part about 1 bit in my first reply to reflect that.
 
Last edited:
One of the things I remember from the PR was that using a larger crc would reduce this URE.. But since they are using 4K sectors and are using half the number of crc bits, eg 8 bits/512 byte sectors would be 64 bits instead of 32.. For their claim to be valid it would mean that there are many areas on the drive with more than 8 bit errors but they are not consequent so using 32bits would give more usable sectors. But also a failure here would also mean the medium has some major problems.

What I mean is, a failure of 8 bits far less critical than a failure of 32bits. As 8 bit failures are more common than 32bit. So more critical.. You have to pay more attention to 32bit failures.. Any URE on a 4K sector drive should be immediately investigated..
 
You must log in or register to reply here.
Back
Top