Parity drive questions

[Julius]

I've read about what is actually on the parity drive(s), but there's some things I don't really get about its sync operations;

 

I have an array with now ~5 TB of data used on it.

Two parity drives of 6 TB each.

I have 10 TB HGST DC drives that I wanted to expand parity size with, so I added the 10TB's to unRAID and precleared them.

Then stopped the array, replaced the 1st Parity drive with a 10TB one.

Much to my surprise it was parity-syncing way more than the 5TB that is actually content. In fact, it even apparently synced way more than 6TB of the 2nd Parity disk.

I would expect a speed up after the 5TB, since all other space is just nothingness. Why is it reading and writing from and to nothing, when just a quick format with a TOC would suffice?

[itimpi]

It is actually making sure that all parity content beyond the largest data drive size is all zeroes.      This is required to ensure parity remains valid even if you later add a new data drive larger than your current largest data drive (but not larger than your smallest parity drive).

[bonienl]

26 minutes ago, Julius said:

when just a quick format with a TOC would suffice?

There is no file system on the parity disk and for it to remain in sync over its complete size, it needs to be zeroed.

[Julius]

Aha, so, beyond what is actual content it is purposely writing zeros onto the 10TB surface, even if that's never going to be used?

Sorry for the silly dinosaur questions here, I come from a time when writing to disk surface would be considered wasteful, and only done when absolutely required, so I was assuming a bit flip would say "rest of the surface not needed" and that would be it. Just so you know where I'm coming from; It already didn't feel good to have to 'preclear' them, it seems needlessly strenuous for the disks.

Must be me, but I'd be designing this a different way.

[itimpi]

6 hours ago, Julius said:

Aha, so, beyond what is actual content it is purposely writing zeros onto the 10TB surface, even if that's never going to be used?

Sorry for the silly dinosaur questions here, I come from a time when writing to disk surface would be considered wasteful, and only done when absolutely required, so I was assuming a bit flip would say "rest of the surface not needed" and that would be it. Just so you know where I'm coming from; It already didn't feel good to have to 'preclear' them, it seems needlessly strenuous for the disks.

Must be me, but I'd be designing this a different way.

You never have to pre-clear disks - that is purely voluntary!   Some people like to deliberately do it to stress test new drives but that is their decision.

 

writing zeroes to the portion of the parity disk Beyond the largest current data drive simplifies greatly the process for later adding larger data disks as it means Unraid does not need to do anything to the parity drive at that point.   I agree that potentially a different approach could perhaps be used but that is the current decision that Limetech have made.   I think they assume that you soon will be getting data drives of the size of the parity disk(s).

[Julius]

2 hours ago, itimpi said:

I agree that potentially a different approach could perhaps be used but that is the current decision that Limetech have made.   I think they assume that you soon will be getting data drives of the size of the parity disk(s).

And that is very true, I guess, since prices have been dropping a lot lately for drives under 12 TB. Anyway, thanks for explaning.