Hello, I currently have a home server mainly for media, in which I have an SSD for the system and 2 6TB hard drives set up in raid 1 using mdadm, its the most I can fit in the case. I have been getting interested in ZFS and wanting to expand my storage since it’s getting pretty full. I have 2 12TB external hard drives. My question is can I create a pool (I think that’s what they are called), using all 4 of these drives in a raidz configuration, or is this a bad idea?
(6TB+6TB) + 12TB + 12TB, should give me 24TB, and should work even if one of the 6TB or 12TB fails if I understand this correctly.
How would one go about doing this? Would you mdadm the 2 6TB ones into a raid 0 and then create a pool over that?
I am also just dipping my toes now into Nixos so having a resource that would cover that might be useful since the home server is currently running Debian. This server will be left at my parents house and would like it to have minimal onsite support needed. Parents just need to be able to turn screen on and use the browser.
Thank you
You must log in or register to comment.
Is there any particular reason you’re interested in using ZFS?
How do you intend to move over the data on the 2x6 array if you create a new pool with all the drives?
mdadm RAID1 is easy to handle, fairly safe from write holes and easy to upgrade.
If it were me I’d upgrade to a 2x12 array (mdadm RAID1 or ZFS mirror, whichever you want), stored internally. And use the old 6 TB drives as cold storage external backups with Borg Backup. Not necessarily for the media files but you must have some important data that your don’t want to lose (passwords, 2FA codes, emails, phone photos etc.)
I wouldn’t trust USB-connected arrays much. Most USB enclosures and adapters aren’t designed for 24/7 connectivity, and arrays (especially ZFS) are sensitive to the slightest error. Mixing USB drives with any ZFS pool is a recipe for headache IMO.
I could accept using the 2x6 as a RAID1 or mirror by themselves but that’s it. Don’t mix them with the internal drives.
Not that there’s much you could do with that drive setup, since the sizes are mismatched. You could try unraid or snapraid+mergerfs which can do parity with mismatched drives but it’s meh.
Oh and never use RAID0 as the bottom layer of anything, when a drive breaks you lose everything.
If there’s an offline backup, they could create a degraded RAIDz1 with the 2 12T disks, copy the data from the 6Ts over, create the 12T linear volume out of the 6Ts, add it to the degraded RAIDz1 and wait for it to resilver. If no hardware fails and they don’t punch in a wrong keystroke, it should work.
Most USB enclosures and adapters aren’t designed for 24/7 connectivity
This is true. I’m using 8 external USB drives in two RAIDz1s and I had to ensure their controllers don’t overheat. For example I’ve had 4 WD Elements standing vertically, stacked next to each other. The inner two’s controllers would overheat during initial data transfer and disconnect. Spacing then apart resolved this for my ambient environment. In the other pool, I had a new WD Elements overheat on its own, without taking ambient heat. I resolved that by adhering a small heatsink to the SATA-USB controller in the enclosure. I also drilled a hole in the enclosure immediately above the heatsink for better ventilation. I later applied this mod to the of the drives of the same model.
Mixing USB drives with any ZFS pool is a recipe for headache IMO.
Crucially however between the issues above and accidental cable unplugging, ZFS hasn’t lost any data or caused any undue headache. If anything, getting back to a working state has been easier in some occasions as it would automatically detect a missing drive is back, resilver if needed and go its merry way. The headache I’ve observed most of the time has been of the sort - a message that zpool is not healthy, a drive has shown errors and/or missing, resolve drive issues if any, reconnect drive, no affected applications, no downtime. The much less often observed issue, probably twice over the last 5 years has been of the sort - applications are down, zpool isn’t reading/writing or is missing, more than one drives is disconnected due to a cable snafu, shutdown, reconnect drives, boot, ZFS detects the drives and it proceeds as if nothing happened. All in, the occasions on which I had to manipulate ZFS over that last 5 years is around 5, most during the initial data transfer load. The previous LVM + mdraid setup I had required more work to get back in shape after a drive was kicked out for one reason or another. So yes USB can definitely present issues that you wouldn’t see in an internal application, especially if some of your USB enclosure controllers are shit, but in my anecdotal experience, ZFS is very capable in handling those gracefully and with less manual intervention than the standard Linux solutions. If anything, ZFS has been less sensitive to hardware problems.
I feel like what you’re saying here, in effect, is “USB connected drives in a RAID are a bad idea, but if you’re going to do it, ZFS is the way to go.”
Hahaha. Good one!
Well not quite. More like “USB connected drives in RAID could be less reliable than internal and software can deal with it. ZFS makes that easier than LVM+mdraid.” The downside of LVM+mdraid in my experience is that it needs more commands typed in to repair an array if something’s gone wrong. It probably doesn’t break much more than ZFS would under the same hardware conditions and it probably can recover from the same conditions ZFS could. USB drives can present more failure modes than internal but one of the points of RAID is to mitigate hardware failures. So I’m considering USB drives as just shittier drivers whose shittiness the software should be able to hide. So far that has been borne out in practice in my anecdata. I’ve used both LVMRAID (LVM+built-in mdraid) and ZFS with questionable USB drives and both have handled them without data loss and rare downtime, less than once a year. ZFS requires less attention. With all of that said ZFS does of course provide data integrity checking and correction which is a significant plus over LVM+mdraid. It’s already saved me from data corruption due to RAM I had no idea had a problem. RAM that passed Memtest86+'s first pass. Little did I know that it fails on subsequent passes… Yes the first and subsequent passes are different. So I’d use ZFS with USB or internal disks whenever I have the choice to. 😂
