Hi,

I just assembled and installed unRAID on my home lab server

-- Hardware wise I have 5 drives :

• 1Tb NVMe that I plan to use as cache and mover will move dayly / weekly

• 500Gb SSD that I plan to use as apps/vm/docker --> plex, Home Assistant, adguardHome, one flavor of homer-heimdall-homepage-homarr, and sonarr

• 3x 18Tb HDD that I wanted to configure

  • 2x HDD to data sensitive to be protected

  • 1x HDD to store media, mainly my transcoded music and films, also the sonarr downloads that will be read and streamed by the plex srvr

-- The use (the way I imagine) :

• the cache will handle the sonarr downloads and then the downloads should be moved to the media HDD

• similar to an automated daily backup of the family phone's pics and vids using SMBSync or something similar to specific shares (it's a shame this app is no longer maintained)... these should be moved to the protected drives

• Home Assistant will run all the time from apps drive

• the protected data won't be use intensive. and today, I have like 1.2Tb of data to protect so I really believe that 18Tb should get me several years of peace of mind *and I know that "big enough is what she said" also what every person expanding their array/pool once said. . ..

• I might install a VM or two to play around... probably to run android studio; fusion360....

now, the questions :

1) how do I set up docker and the VM manager to use the apps-SSD for apps data / yet to use the cache to write ? (I guess this will be clearer down the road)

2) about the "protected array" in my vast understanding (joking) is that for only 2 drives, calculate parity makes no sense and that just mirroring (which if I got right might be achieved by a pool instead of the array) gives the same protection, using less resources... so I dig a bit on the wise internet and I found a couple of posts, like this one basically saying that with only 2 drives on the array, unraid will set-up a RAID1 like array, not calculating parity... the posts I found (google found) are "old" so, could someone confirm this and eventually telle me if would it be better to set up an array OR a mirrored pool ?

3) I let the file systems all in "auto" which I believe will use XFS, and this is fine by me, but just for the protected array/pool I wonder if ZFS would be a better pick ?

Thank you all in advance for your time, patience and wisdom sharing replies.

First, some clarifications about your notions of protected storage as implemented in Unraid.

In Unraid, the array consists of those disks that can be protected by parity. You can have 0, 1, or 2 parity disks in the array. 0 (no failure protected), 1 (1 failure protected), or 2 (2 simultaneous failures protected). These are listed separately from "pools". The way Unraid implements parity, each data disk is an independent filesystem that can be read all by itself on an appropriate OS, such as Unraid or any Linux. And parity (and parity2) is a separate disk that only contains parity bits. https://docs.unraid.net/unraid-os/getting-started/what-is-unraid/#arrays-and-parity

An array with a single parity and single data disk is a mirror simply because of the way the parity calculation works. You can add additional data drives to the array and they would also be protected by parity, but it would no longer be a mirror.

In addition to the array, you can have one or more pools, each can have one or more disks depending on the filesystem chosen for the pool. These pools can be used for caching user share writes that get moved to the array, or just for fast storage.

Each User Share has settings that control how it uses storage; which drives/pools to use, whether to move, etc. User shares are how Unraid allows folders to span drives and pools.

Typically, the docker/VM related shares appdata, domains, system will go to a fast pool and not be moved. Other user shares can be cached and moved to the array, or just written directly to the array.

Parity, mirrors, various forms of raid available with Unraid array and pools, should not be considered a substitute for backup. Plenty of more common ways to lose data besides a bad disk, including user error. If you accidentally delete something, for example, Unraid will consider it deleted and restoring from backup would be the best way to recover.

The array can be whatever filesystem, but XFS is the usual choice since each drive in the array is independent. A multidisk pool must use btrfs or ZFS.

Don't cache initial data load or any continuous user share write larger than cache. It is impossible to move from fast cache to slow array as fast as you can write to cache, and trying to move while still writing will slow things down even more. Standard way to cache is to schedule mover for off hours. Default is daily in the middle of the night.

Another point to note is that you should not rely only on parity to protect your data. For important files you also want another copy as a backup, ideally not kept on your Unraid system.

yes, I'm trying to implement a 3-2-1 including a NAS in the family-in-law's house.

14 hours ago, trurl said:

An array with a single parity and single data disk is a mirror

does this means that, in this case, the parity disk can be unplugged-plugged somewhere else and read ?

2 hours ago, jorgemarmo said:

does this means that, in this case, the parity disk can be unplugged-plugged somewhere else and read ?

In the special case of a the main array having 1 data drive and parity1 then the answer is yes.