I dont know much about zfs so I’m not 100% sure. Looks like I will need to search youtube and this forum for some ideas
You might want to put the fast drives into their own device, because even one slow drive will slow down the entire thing since it will have to wait on it.
I have a btrfs array of six HDDs where two of them are 7,200 RPM drives (they were a good deal) and it is the same speed as if they were all 5,400. But it does work.
I’ve done mixed speed arrays using Linux MD and btrfs but never ZFS, however I don’t see any reason it wouldn’t work.
How do you plan to back the data up?
I have no idea. I was under the impression zfs handles the backup. I’m assuming there are multiple types of zfs configurations?
zfs can only be as good as the redundancy you setup in it.
if you setup a bunch a single drive vdevs… dont be surprised when you lose the pool when one drive fails.
(and dont blame zfs)
ZFS acts more like raid but has special sauce that protects against things like bit rot. You typically wouldnt rely on raid alone to be your backup. I use ZFS as my backup but my nas is on another machine entirely and I rsync to it. If you dont have a backup solution and ZFS fails in some way… you’re just as boned as you would be with any other FS.
1 Like
Whoa, there.
Remember that RAID is not a backup. With ZFS (or btrfs or LVM) snapshots it can cover some backup scenarios. But what will you do if the power supply blows out with a huge spike and sets all of your drives on fire? Or thieves steal all of your stuff?
That’s why I’ve got an 8 TB external USB drive I use to actually back up my important files and whatever else will fit.
I have two external drives, actually, and swap them out in a bank safe deposit box.
Other than that you can use online backup services which are very slow, generally, but will keep your stuff safe.
Or some people like tape for backup. It can store a lot more than a HDD and each tape is easy to move to safe storage (like a bank box), but I always thought that I can buy a lot of external HDDs for the cost of one tape drive.
1 Like
Well… yes and no.
LTO-7 Tapes are 15/6 TB at ~80€ each, LTO-8 are 30/12 TB at ~170€ each (that is, compressed/uncompressed capacity).
But the issue with tapes aren’t the tapes, it’s the drives, cause those can easily cost you midrange 4-figures a pop.
And AFAIK they are also neither upwards nor downwards compatible. So once you get a drive you have one type of tape to use and that’s it. So when the drive dies you either get a new one of the same type, if they are still available, or you’re boned. So you’d also need to migrate them every now and then.
Let’s use this hypothetical scenario… I have one 4tb hard drive setup in zfs. How many hard drives would I need to correctly backup that drive? Then later down the road I run out of space and throw in another 4tb HD into the mix… What is the ratio of drives that I would need for storage and backup?
I think you misunderstand how ZFS works. You can set up a single drive… but theres basically no advantage to doing so.
In ZFS you have whats called a vdev. A vdev is responsible for its own redundancy. You cant add a disk to a vdev. A zpool is made up of vdevs. You can add a vdev to a zpool but you cannot remove them. Since all the redundancy happens at the vdev level, if you have a single drive vdev and you lose that drive, you lose the pool. Also if you set up a vdev, in say raidz1 for instance with your 3x1tb drives, your pool is now only ever going to be as fault tolerant as raidz1 regardless of if you add a raidz2 vdev to the pool.
You would not really ever want to add single drives since that would defeat the purpose of the redundancy making your pool reliant on a single disk vdev. lose that disk, lose the pool.
I think you might want to look into a synology or a similar mdraid setup where its more forgiving.
As far as backing it up is concerned, how much storage you need is based on the amount of data you have, nothing else.
2 Likes
Agreed, I’ve had 7200 and 10K drives in a scale out file server and they just end up slowing down to stay cool no matter what.
Unless you really need the speed for a specific purpose or are interested in future proofing yourself that much more, the slower drives will probably perform the same without also investing in major cooling and airflow options on top of all of it.
Thank you for that. I completely missed the point of zfs. I was under the impression you just keep adding drives to the pool until you max out. This makes sense now. I will have to buy more than one disk to make this worth it.
1 Like
This is the major disadvantage to ZFS. You do want to plan out your upgrades and buy a bunch of disks all at once to match what you have in redundancy. If you have a 6 disk raidz2 vdev, you really want to match that with another 6 disks in a raidz2.
If you’re just looking for a nas setup to dump random size disks into, I recommend unraid or synology or xpenology if you dont want to buy into the synology ecosystem.
I want to do exactly that. I want to dump random size disks into a nas. I will look into the three technologies that you mentioned
unraid is interesting because as its name implies, its not a raid. It still has parity, but it dumps all the parity on to one disk. The upside to unraids quirky setup is if you lose the array, you dont lose everything, just the disks that died.
synology does something a little different in that it breaks larger disks up into smaller chunks so that you can utilize the storage on them. They call it synology hybrid array.
xpenology is just the synology operating system made to run on non synology hardware. Theres some caveats to this too though.
Ultimately backups are what are most important. RAID in any form isnt a backup, its about uptime and availability. In the case of ZFS its also about data integrity. I use unraid for my nas but backup to a box running ZFS.
unraid sounds exactly like what I would like to do. I remember I watched some videos about the unraid subject and I thought that was zfs. I was totally wrong LOL. So I think I know what technology that I would like to use; now it’s time to figure out what other hardware to get(mobo, cpu and ram) or if I should repurpose my every day windows machine(remove windows and install unraid).
Do you want to virtualize? Do you care about docker?
Virtualization like virtualbox? I use it, but for whatever reason it just seems a little slow if I use use a gui like gnome in my Ubuntu guest OS. I probably haven’t configured it properly. Virtualbox is running on my every day machine. Command prompt runs rather fast. The only experience that I have with docker is running Home Assistant.
The specs of my current pc are:
Asus X99-A
Intel Core i7 5820K @3.30GHz 6cores
DDR4 Quad channel 16 GBytes
AMD Radeon R9 200 Series
I was going to say something senable 
but I waited
I’d go for capacity over rpm.
RPM impacts random IOPs performance more than sequential, and your access pattern is mostly sequential.
For a NAS for media playing, 5400 rpm drives are also quieter.
IN short: buy 5400 rpm, and favour capacity over rpm. You don’t need rpm in this application. And these days if you’re dealing with an application where RPM is important, you’re likely better off biting the bullet for SSD cache (and again, use cheap/large 5400 rpm on the backend) or pure SSD.