Switching from independent disks to raid in place

Hey everyone,

over the years I have accumulated a couple HDDs in my main system that I have tried to use equally to keep my data sorted, but as you can probably guess it has become a mess at this point and I am getting concerned that any one day could be the last of me seeing any particular chunk of data.

Running freenas I found zfs to be a nice filesystem to work with and I think it could help me keep my data safe and ordered.
But now for the problem: I would need at least 4 drives(which is just the mininum for the raid level), to run raidz2 but buying the drives that could hold around 8TB of data in this configuration is out of my budget, so I am looking for a way that I can more or less incrementally get my data to safety while not breaking the bank.

Thanks in advance for any tips you might have.

A few notes:
My PSU should be able to handle a fair amount of disks with 1kW
I do already have a flashed raid controller in IT mode in the system, so adding a couple drives should not be a problem.

1 Like

Thanks for the question. Please can you clarify:

  1. How much data you are looking to store? You mention 8TiB, is that the amount of data or the amount of disk space you want to have.

  2. I’m assuming you have backups, at least one full copy not on your main drives. If not, before you even think about moving data, buy one big drive and copy everything to it, put it in a box or give it to a relative to keep safe.

  3. What do you want the pool to be used for? if it is just archive of old files then a mirror is fine, so you only need 2 big disks. If you want more performance, ie, for game library, then you will want to stripe drives (raid 10 or mirrored Vdevs).

  4. RaidZ2 is not what you need. Technically you can use 4 disks but that means you only get 2 disks of data and 2 disks redundancy, but the performance of 1 disk. you will get more performance from striping 2 disks and mirroring (mirrored vdevs).

  5. You talk about adding drives. Remember for RAIDZ the drives will all need to be the same size, or will default to the smallest size. Adding “lots of disks” is sometimes worse than buying one big disk, at least on your main machine, and using the old ones for your backups (easy to split a backup)

Take some time to think about why you want to build a resilient pool, how you want to access the data (fast or slow) and what tiered storage model you really need. Feel free to ask more questions!

There are around 5 active threads on this topic at the moment on L1, so have a read through some of the other ideas, not just mine. Good luck

Good luck.

1 Like

Curious on your title if you mean in place, as in to add raid, and not format existing disks. If so, this is not possible with any method I know, and if anyone tells you it is, make sure you have a backup. And if you have a backup, there is no reason to do it in place.

As stated, running raid z2 on 4 drives is not a great option, and raid 10 is a better setup. Same usable space, and generally a better setup with 4 drives.

Look at shucking drives possibly. I have recently built an array out of 3 10tb drives at $165 a piece for a raidZ (1 disk parity). I’m aware that it’s not safe, but I have critical data backed up on another array, so data loss is more of an annoyance than a concern.

Definitely overkill, so you are fine.

I’m assuming that you have a ton of media that you are storing. If not, please let us know so we have an understanding of what performance is needed. Generally speaking, more drives in a raidZ (Raid Z2 and Raid Z3 included) are faster with more drives, but that starts at 5 drives on a Z2, and 7 drives on a Z3 (compared to Raid 10). IF you are storing media, and not actively “working with data” on the drives, then generally it won’t matter.

I’ll also offer another option, that I WILL NOT RECOMMEND, but is something I have done in the past. You can build a degraded array using sparse files, intentionally removing parity until you add a drive in. I have done this before when doing a raidZ2 array, but did not have the budget for the last disk. I built it with a sparse file, and when I had the money for the last disk, I added it and resilvered. NEVER DO THIS WITHOUT A BACKUP! I can not stress that enough. Your data is NOT safe while you do that. In my case, it was as safe as a RaidZ (1 disk parity) but that’s not safe. The only reason to do that is budget, and only ever considered if you have a good backup in place. I knew I wanted a raidZ2, but couldn’t afford more than a RaidZ, and you can’t convert it after the fact without doing something like this.

TLDR on most of the above. Never ever do what I did, unless you have a backup. Hopefully I stressed this enough.

1 Like

First of, I think I owe saying that I am a university student so that’s why I have a very tight budget, I am not trying to use anyone here as price aggregator.

Thanks for the reply, sure I can clarify some things.

  1. I did mention 8TB and it’s roughly the size I would need to store most of what I have right now. “Most” instead of “all” because I don’t want perfect to be the enemy of better.

  2. No, none of my workstation data is backed up that’s why I want more data redundancy with a raid and just blowing away 350-500 bucks on something I store cold isn’t really an attractive prospect to be honest.

  3. The pool would hold everything from coding projects over my music library to games, that’s why I ended up with raidz2 since it offers a sweetspot of performance, redundancy and usable size.(according to the zfs guides I have read)

  4. I know that raidz2 with 4 disks is borderline comical, but I just picked the lowest possible drive number. Ideally it would be more of course.

  5. What I meant with “adding drives” is that, should it be necessary, I can manage a couple more physical drives without having to do some janky USB hub stuff.

I am sorry, I know “in places” is the wrong term but I didn’t find a good other way to express that I would rather not buy a whole new set of drives at once.

I would really like to hear more about that because the price seems great.

I have heard of this and it’s pretty tempting to do in all honesty because it would give the incremental upgrade element that would soften the blow to the budget. That the data is at risk is the obviously prohibitive factor.

I think I should add that right now I have 7 HDDs in my system and using that number is why I came to raidz2 instead of anything else.
I thought about buying 7 super cheap (and small) drives but all my data wouldn’t fit on there and 14 disks is too much for my case.

Byte My Bits has a great series on youtube. Instead of typing forever, I am going to direct you there as he goes through the up’s and downs, and how to work around the 3.3v “issue” that older power supply’s have as the sata spec has changed. It’s pretty simple. I can recommend grabbing 2 of the 8TB drives and adding them in raid1 (mirrored). Your data will be safe on a mirror, and you can easily add another mirrored vdev to the pool to add space later with complete protection if you ever need more space.

Thanks for all the clarifying points and responding clearly. Nice to have a chance to close out the thread.

Fully agree with @kdb424. Watch Jason’s videos as he gets the point across well. Also good networking content on there.

Depending on your location I would generally say walking into a supermarket (best buy, Walmart style places) and looking for external USB hard drives in massive sizes is always the cheapest option. The drive itself will be totally random. I’ve had reds and greens and blues (WD language) as well as the dreaded white label ones.

I won’t bend on this one. RAID is not a backup, ever. RAID has only one job, to maintain uptime in critical production loads. If you want more storage add more JBOD partitions. Always budget for cold backup. I started my data hoarding career as a student (back then it was CD rips to MP3, showing my age). I lost a drive array with work on it too. I cried.

I’ve seen and been involved in a bunch of threads this week (must be something in the water) on L1 with people wanting to build massive disk arrays with expansion in mind. I really don’t understand the drive (pun intended).

Buying six 2TiB disks to use raidz2 to get 8TiB will not be materially cheaper or more resilient than mirroring a pair of 8TiB disks and adding a cold backup. If I want more space in a year, I could nuke the mirror and make a raid 10, I still have my backup.

Trust me, running 6 drives in a non enterprise chassis will get warm and create cabling jungles. If any disk dies, your best recovery approach is to copy the data to a new array, as swapping the dead drive is not going to be plug and play. The more disks you have, the more vibration, increasing failure scenarios.

For your use case less is more. I would strongly suggest spending your limited funds on a single big backup disk and copy data to it once. Then if you are feeling brave, raid / mdadm your existing JBOD disks for capacity purposes and maintain regular backup rotations. You can add a second disk to the rotation later if you can afford it.

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