Arch Linux

KaibutsuX

Member

Registered: 2015-04-22

Posts: 22

Raid drive size frustrations

I have a 7 device RAID6 array built on 10TB drives (9.1TiB or 10.00073TB) (10000725425664 bytes)

I was running out of physical room and decided to take advantage of a 20TB drive which I planned on partitioning in 2 for 9.1TB each.

However, apparently, the drive is actually 20000588955648 bytes, which works is 20.00059TB which works out to two partitions of 10.000295TB. So if I make the first partition the same size as my existing 7 drive partitions, my leftover space isn't large enough to be added to the raid.

I guess this is why you're supposed to build in a larger buffer than using the entire drive.

What are my options here? Can I resize the existing partitions to be smaller without data loss? Thinking through it I can't even figure out how that would be done, because as soon as 1 drive is removed from the array to be shrunk, it couldn't be added back until the last drive was also shrunk.

Am I just screwed?

Trilby

Inspector Parrot

Registered: 2011-11-29

Posts: 30,459

Website

Re: Raid drive size frustrations

Not a direct answer, but wouldn't partitioning a single physical device to play the role of two in a raid 6 be a bad idea in the first place?  It'd make the whole thing really a raid 5 as it couldn't recover from the failure of two devices if one of the two was this new 20TB device.

---

"UNIX is simple and coherent" - Dennis Ritchie; "GNU's Not Unix" - Richard Stallman

frostschutz

Member

Registered: 2013-11-15

Posts: 1,647

Re: Raid drive size frustrations

You can shrink whatever is on the RAID and set it to use a slightly smaller size. That should allow you to add two partitions of your 20TB even if they're slightly smaller than your original drives.

However, Trilby is correct. Doing this would harm your redundancy. Your RAID would no longer survive any double failure if one drive fulfills two roles in the same array. (It would only work for partial failures, bad sectors, but that's gambling).

Also it would be terrible for performance since for every request, the 20TB drive would have to deal with additional seeks across the entire device. Well, this could be optimized via the device mapper so these seeks are smaller, or you could pick a raid layout where this drive stores parity exclusively so it is not involved for reads (as long as all drives are intact), but... is there a point...

You could split your array... so you end up with two arrays, one 4x10TB another 5x10TB where your 20TB fulfills one single role in either array. Redundancy-wise that would work but it doesn't gain you anything. RAID-6 wastes a lot of space on redundancy and with two separate arrays, you'd end up paying that cost twice. You'd still end up reducing to RAID-5 just without the performance issues.

Do you have any data that doesn't need redundancy? You could add one 10TB partition to your RAID-6 and use the remaining 10TB for something else entirely. You could still make that additional partition a RAID (single drive RAID-1) so you could grow it if you were to add more 20+TB drives in the future.

Last edited by frostschutz (2024-05-08 06:58:36)

KaibutsuX

Member

Registered: 2015-04-22

Posts: 22

Re: Raid drive size frustrations

Trilby, thats a good point, I hadn't really considered that, I was just going for the more efficient use of physical space.

I have done the partition to 10TB and use the rest for something else elsewhere since there are a few 14TB drives in the array, but man I was really looking forward to 2 more 10TB partitions.

I guess I could just eat it, use the 20 for something else as an entire drive and maybe purchase a single 10 to add to the raid.

Thanks for the responses, I learned something.