I’ve read a lot on pre-clearing disks prior to adding them to an array. I understand pre-clear does the following:

 

1. Clears the disks so that unRAID does not need to do the same

2. Stress tests disks to rule out bad disks

 

Let’s park the idea of stress testing the disk. The disks I’m concerned about have been in service for a while and I know they are good.

 

Coming back to clearing the disk. I believe this is only needed when adding disks to an existing array with a valid parity. Emphasis on "existing array with a valid parity".

 

If I were to start a NEW array without a parity disk, unRAID just needs to format the data disks to one of the supported file systems and bring the array of online. I can then stop the array and add a parity drive which would trigger and parity build.

 

Using the above process I can completely skip the clearing/pre-clearing process for NEW arrays. Am I correct?

Correct, but how do you know the disks are good? Have you done any diagnostic tests or looked at their SMART?

 

Just because you have used them in some other system and aren't aware of any problem doesn't necessarily mean they are good enough for unRAID. All bits of all disks are used when rebuilding a failed disk.

Yes.

 

As to testing, I would at least view the SMART information on the disks, just to be sure you don't have a silent problem waiting. Unraid needs every sector to work perfectly, not just the sectors that have been used in the past for data. A disk that worked perfectly in another OS can still cause issues when it comes to rebuilding a failed disk in unraid.

And if it was used in Windows, be aware that Windows is notoriously bad at telling you the drive is NFG until it actually drops dead.

Correct, but how do you know the disks are good? Have you done any diagnostic tests or looked at their SMART?

 

Just because you have used them in some other system and aren't aware of any problem doesn't necessarily mean they are good enough for unRAID. All bits of all disks are used when rebuilding a failed disk.

 

I have examined their SMART attributes and find no issues. These disks have had data written to and read from them frequently. They were daily drivers. So I believe it would be safe to assume they are good drives. Secondly, even if I were to run 1 or 2 pass of pre-clear, that would not necessarily mean they would not fail on the third. I think pre-clearing for stress testing makes sense with new disks to rule out any DOAs (infant mortality).

Yes.

 

As to testing, I would at least view the SMART information on the disks, just to be sure you don't have a silent problem waiting. Unraid needs every sector to work perfectly, not just the sectors that have been used in the past for data. A disk that worked perfectly in another OS can still cause issues when it comes to rebuilding a failed disk in unraid.

 

Reallocated sector count and Current pending sector are zero. Do I need to look at any other attributes as well?

 

And if it was used in Windows, be aware that Windows is notoriously bad at telling you the drive is NFG until it actually drops dead.

 

The disks were in use on a Mac and were almost full.

I pre-clear each of the old disks which I reused for the server (and certainly each of the new disks too).

 

Back a long time ago when "Norton Disk Doctor" was a thing, I noticed that my failed sectors tend to be towards the very end of the disk. So "almost full" is no guarantee that sector n-1 has been touched and thus there's still a relatively higher chance of it being part of the DOA population (just not dead enough). Pre-clear (or any other proper stress test software) would ensure the entire disk is tested.

 

You are 100% right that 1st pass does not guarantee 2nd pass and so on. However, the marginal risk after each pass becomes much smaller. I just run it once to ensure the entire disk is tested.

 

However, if you don't want to run a pre-clear, you do not have to as it is not a requirement.

Back a long time ago when "Norton Disk Doctor" was a thing, I noticed that my failed sectors tend to be towards the very end of the disk. So "almost full" is no guarantee that sector n-1 has been touched and thus there's still a relatively higher chance of it being part of the DOA population (just not dead enough). Pre-clear (or any other proper stress test software) would ensure the entire disk is tested.

 

That makes perfect sense. I went ahead and pre-cleared the disks once and they came out good! Thanks!

Deja-Vu. It seems like this topic keeps coming up every other day.

 

See the other existing discussions for benefits.