Hi, i am new to Unraid but not to computing (30+ years experience).

The software has created interest in me because it's quite different to professional approaches, yet very useful in the home domain it seems.

 

I have some hard disks (20TB) that i want to use for backups. Actually, twice the number to allow for two backup sets.

1) I am wondering if i can format the disks as stand-alone disks with e.g. ZFS or XFS and put them into Unraid to share their contents via SMB? No fear of accidentally formatting them, anything to take care of?

That way, i don't need to connect a USB-SATA dock to the system to-be-backed-up, i can just backup via network (10Gbe) and, once completed, take out the disk and store it.

2) I would also like to be able to use Mergerfs so i can have two disks appear as one - i read some Unraid users are doing already?

 

Thanks a lot for any advice.

Unclear what you have in mind. Will you have an Unraid array and these "backup" disks will be additional disks not assigned (Unassigned Devices)?

 

 

unassigned means not part of an array but still be able to be shared? 

Then this sounds promising and that's what i meant.

15 hours ago, unraidinterested said:

not part of an array but still be able to be shared? 

yes, but you could also put each disk into a seperate "pool" and share the content via smb.

 

The difference is (the term "array" is currently very misleading, because there are two types of arrays which have not much in common. Let us call the "UNRAID array" "UA" and a ZFS RAID array "ZR" in the following text)

 

The UA consists of a "bunch of disks" with optional one or two parity disks. The advantage of this type of array is, that shares can split over disk boundaries (they are automatically collected together again an presented to the client as one piece), the disks can be any size, disks can be added later on, every single disk contains a normal filesystem and can be read by any linux system without any tricks. The optional parity drives sum up all blocks of the data disks and allow a failed disk to be replaced without any loss of data ( dont ask me why there can be 2 parity drives, its a thing for the overcautious types that assume "more than one" drive to fail at once). Drawback is a rather large overhead so with fast lan and fast disks you notice a degregation of speed. There is not read caching, a ssd cache can be added, but only as a write cache for NEW data. (yeah thats bad, I ask for a read cache some years now already, no answer from limetech until now).  The only limitations are "1) the parity disk(s) need to be the largest one(s) of all data disks" and "2) maximum of 30 (+2) drives can be used (a limit that you hardly can hit with common PC hardware)"

 

The ZR is a standalone Array based on ZFS and RAID technology. You need disks of same size/type. Depending on the amount of drives its either RAID0, 1, or 5. It contains a bidirectional cache (called "ARC") in RAM which is managed by ZFS itself. Its said to be much faster. I dont know, I have bought UNRAID because I did not want RAID anymore. I was fed up to get replacements for failed drives after some years. So dont ask me how good this works.

 

What you never should do or even consider is to put an ZR into an UA (with parity)! This will make you box creep around like an old dog in the hot sun...

This is because the UA parity has to follow each write operation and the ZR uses idle times to "optimize" its internal disks. So there are permanent writes which slow down everything until the IO Bottleneck kicks in and the machine almost comes to a full stop. You do not want this!!!

 

So either trade speed for flexibility or use fast access and be aware that there are problems to come in a few years...

 

From your reading I guess you want it as a backup solution. Then I would recommend to use the UA because you can expand it really easily later on or swap out data disks if they become unstable or something. And if everything fails, you can pull out the datadisks and use it in different machines without loss of any data (but you have to be aware that folders might have become splitted, so it may take more than one disk to copy to get all the content of a single folder (files itselfs are never splitted).

And the "mergerfs" you are looking for is automatically builtin. You can select "when to split", but the merge is always there. You can also select if disks are filled up one by one first or split evenly.

 

 

 

Edited November 24, 20241 yr by MAM59

Hello @MAM59, many thanks for your answer, it has helped me tremendously! 😀👍

As a matter of fact, i have now organized a trial on a VM with simulated hard disks, just to see the software in action and it's better for understanding. I also stumbled on Space Invader One's set of formidable videos on Youtube and watched maybe 10 or so explaining some things i figured would be the essential topics, like intro, moving files, Apps, some use cases, parity, mover, caching - in the parity regard, he also explained the RMW, that everything has to be reread, then calculated and written again. 

That Unraid is special was already somewhat known to me but the more i dig into the topic, it blows me away how special it actually is and how large the community is.

 

I operate a self-built Truenas Scale and it was a 2 year project to even build it (leave alone the day-to-day operations which i have yet to fully master, like ZFS backups). I think that Unraid with its UA has a particular use case, therefore i am wondering what's their way forward as in 7.0 they seem to support all the ZFS part. I would never use Unraid for speed currently, i think they need to concentrate on the UA for the future as well (in my mind, that's the flexibility, e.g. being able to just throw in leftover disks that makes home users use it).

 

I ordered some parts to build my setup, i.e. some disks, a used mainboard/CPU/RAM for rather cheap but still luseful. I hope to be able to create a machine that can host some backup storage for sudden needs (i.e. you notice how complicated working with big-sized NAS it can become if you cannot shift data around, one NAS culminates in the need for another and so forth). I also hope i can put it to sleep when not in use so the power consumption should not be an issue.

 

For 1), i think my format of XFS was an accidental good decision  to use the 20TB backup disks in a hotswap cage, then add them to a non-parity protected UA. They should mount and i can create a share and put a "how-to-fill-the-drive" policy, which should get 2) covered. 

Edited November 26, 20241 yr by unraidinterested