I am replacing my parity drive with a larger disk.

 

I want to make the old parity drive the cache drive.

 

Will making the old parity drive the new cache drive mean that it has to be zeroed, and if so will the rest of the array be unavailable while it's being zeroed?

 

If so, I should do a "zero-only" preclear to avoid downtime?

 

Cheers,

 

Neil.

I am replacing my parity drive with a larger disk.

 

I want to make the old parity drive the cache drive.

 

Will making the old parity drive the new cache drive mean that it has to be zeroed, and if so will the rest of the array be unavailable while it's being zeroed?

Only drives that will participate in parity calculation have to be zeroed before being added to an existing protected array. The cache drive is not protected, and just needs to be formatted. I'd pull and examine a smart report just to be safe, but no need to preclear.

The drive does not have to be zeroed, however I would suggest a smart LONG test or as an easy test, doing the preclear (if it has not been done before). 

If you've already used Joe L's preclear on the drive, your good.

 

If you are familiar with the smartctl and can run the smart -t long then check it later, that would suffice.

The only reason to do the preclear is to exercise every sector and see the smart log comparison output.

 

I assumed (perhaps incorrectly) that since the drive was in service as a parity drive, that a parity check would have been run before it was removed and replaced.

 

To the OP, MAKE SURE YOU DO A PARITY CHECK THAT COMES BACK CLEAN BEFORE REMOVING THE DRIVE.  ;D

That way at least it's passed a full read scan recently, and a smart report should be sufficient. 

I assumed (perhaps incorrectly) that since the drive was in service as a parity drive, that a parity check would have been run before it was removed and replaced.

 

To the OP, MAKE SURE YOU DO A PARITY CHECK THAT COMES BACK CLEAN BEFORE REMOVING THE DRIVE.  ;D

That way at least it's passed a full read scan recently, and a smart report should be sufficient. 

 

Thanks all.

 

Yes, the parity drive has been read to and written to bajillions of times, so all I'm going to do is get a SMART report.

 

And, I have done several (3) full parity builds and parity checks as I have been adding drives in the last week or two -- the last one being about 6 hours ago. I'm not going to do any more! 

 

Thanks for the info. I shall just re-assign the drive to cache and expect a few minute wait while it formats!

 

Cheers,

 

Neil.

If you've done a parity build, then every sector has been written to, You're good.

Do a smart -t long test when the machine is idle.

I would do the long test before the drive is assigned to any slot. Otherwise a spin down will disrupt the test.

At the very least you'll have a mark in the SMART logs of when you put the drive in service as a cache drive.

Hmmm... how long (roughly) would a long SMART test take?

Depends on the size of the drive. It can be pretty long. sometimes 5 hours.

 

 

It will tell you when you submit the test, wait xxx minutes.

Thanks.

Thanks.

Make sure you disable any spin-down while performing the long test.  A spin-down will abort the test.

 

Joe L.

Thanks.

Make sure you disable any spin-down while performing the long test.  A spin-down will abort the test.

 

Joe L.

 

Good to know - but I'm surprised that it would be spun down while it's still actively in use.  I gather this means the spindown timer is maintained by UnRAID without "knowledge" of other disk use by Linux.    Does this mean add-ons that only use the cache drive (i.e. no writes to the array) could encounter the same issue?

 

Thanks.

Make sure you disable any spin-down while performing the long test.  A spin-down will abort the test.

 

Joe L.

 

Good to know - but I'm surprised that it would be spun down while it's still actively in use.  I gather this means the spindown timer is maintained by UnRAID without "knowledge" of other disk use by Linux.    Does this mean add-ons that only use the cache drive (i.e. no writes to the array) could encounter the same issue?

I'm pretty sure unRAID uses the built-in spin-down timer on the cache drive, but not sure if the SMART test triggers it.  Best to disable the spin-down during a long test.

 

Joe L.

The SMART test will not trigger activity that the kernel will detect. It's all done in firmware.

This is why I suggested doing the SMART long test before adding the drive to the array.