Stepped transition from single HDD to RAID5?

Hardware Hub Storage
1

Finances dictate that:

I need to start with a single NAS HDD (Toshiba N300 16tb?)

I want to end up with 4x HDD in RAID5(?). One disk at a time, over several months. Added benefit of hopefully spreading future failures out over time too.

Can I do this? Should I do this?

I guess LVM is almost mandatory?

I have 4 onboard SATA, and intend to use Linux RAID.

It’s for a ProxMox server that will hold video, steam cache / cafe server, home documents, ISO’s etc.

Opinions gratefully received! :slight_smile:

2

  1. Yes, but you’ll need to transfer the data back and forth if you’re going to migrate to RAID-5/Z

  2. No, most if not all will probably recommend ZFS

  3. See above

  4. Why bother with Proxmox? Seems like you’re adding just another layer of complexity for no obvious reason. Consider using Debian or whatever floats your boat. FreeBSD might also be an option and like likely less painful to maintain.

You might also want to look into Toshiba’s MG08/09 series over N300.

3

I approve of those drives. Great $/TB. Running 6 of them for 1.5 years now, all fine and running. People also like to get Seagate Exos series…basically the same, but another brand. HDDs are commodity and I don’t pick favorites.

You can and maybe you should. Proxmox runs on ZFS, so you maybe want ZFS. So consider another option: You start with one disk. If you get the 2nd disk, you attach the disk to the other and get a mirror and the full redundancy package. 3rd+4th disk just add another vdev → you get a RAID10. And easily expandable by adding two disks as a mirror until you run out of SATA ports.

With RAID5/Z, you need a 5th/6th drive to copy everything over once you do the array of 4.

4

  1. Is there a way I can keep data in place during expansion, or does everything get wiped if I add more disks?

  2. I’ll look at it again, but I understood that ZFS has a pretty big RAM overhead, considering I’m just running this at home with limited users. Is it not overkill for me?

  3. I am loading various containers and VM’s on it. (Jellyfin, Asterisk, game servers etc.)

Thanks for the tip! The price difference compared to the huge MTBF change seems well worth it, even with the higher power requirements.

Thanks!

5

I’d have preferred a low-power option, as I don’t need the speed, but there’s not much out there. SSD capacity sucks still, and the prices are horrendous, especially for NAS. Hopefully by the time these drives start failing, SSD’s will be a viable choice.

RAID 1 is a no brainer, but I wouldn’t want to end up with RAID 10, as it would be a huge waste of performance, and storage space that I can’t afford.

I’ll look at ZFS again though. Thanks!

6

RAID 10 is double the speed of a RAID5. 4x read speed and 2x write. And easy fast resilver. Only RAID0 is faster.

Well, storage is expensive. Cutting corners everywhere just because the budget is too low has all kinds of trouble around it. Like rearranging storage topology after creation and choice of subpar drives.
And it’s not the capacity that sucks. We have 30-60TB SSDs today. They’re just too expensive for most people so market will stick to 2-8TB models and small form factor like M.2

4TB SATA SSDs are well priced atm. Good drives can be had for 60$/TB. For everything else there is HDD with killer low prices. Save money for 4-6 drives and then put it to work.

7

ZFS does raid expansion these days, I’ve read about it but have never tried it myself, not sure how that’d work.

If I really wanted to, I’d try it in a VM.

I have a bunch of high capacity drives of various brands, ages and sizes and they’re mostly hosting media and backups.

They’re formatted individually with gpt - LVM - cryptsetup - btrfs.

… and then I’m using mergerfs + snapraid to have a singular mount point through which the space and files are exposed on the network via samba.

… and I’ve done various raid 1/10/5 with btrfs with smaller drives (3T / 8T), but at most 2 drive sizes at a time.

1 Like
8

If it were for business, then yep, i’d agree. But his is for my own home, where I have to weigh up costs a lot more carefully. I already prepare to pay the extra for enterprise level NAS drives with redundancy, but I have to do it in a way that’s compatible with family finances. I know there are bigger SSD’s out there, but for most of us mortals, a single 4Tb NAS SSD is already out of reach. And with 4 of them giving the same, or less, storage as a single NAS HDD, there’s only negatives for my use case scenario, apart from lower power usage.

9

I would suggest LVM. Does what I need. Could do what you need.

Nothing against ZFS. Always liked the idea, but never got past the friction. I configure and use a lot of different systems over time. LVM does what I need. ZFS is more trouble. (Queue the long discussion around ZFS integration with the Linux kernel.)

I tend to use LVM everywhere, by default, as it is easy. Easy to do fancy things (striped arrays, SSD cache, RAID), also.

If you are only setting up one box, either is fine.

1 Like
10

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.