3 or 5 HDDs for RAIDZ1?

[.n3]

Hey,

I want to setup a homeserver with proxmox and homeassitant, nextCloud, frigate nvr, nodered, ... I want to use 7-8TB for my surveillance and 6TB for private data. The surveillance can be lost and the private data will be backup. The backup will be through proxmox backup and the backup will be stored on an different system.

But what is better for a RAIDZ1? 3x 6/12TB WD Red Plus, or 5x 4TB WD Red Plus? I think 5x 4TB will be better because "cheaper" and more read speed. Are there also other (dis)advantages?

I don't want to use RAIDZ2, because of the cost and it brings me only a better availabiltity/uptime but not more data security. Therefor I have my backup.

best regards

 

[ness1602]

Yes,usually for read more disks in raidz the better,so i would go with 5x4tb.

 

[.n3]

Ok, but which disadvantage do I get?

 

[ness1602]

No,this is raidz1, there aren't any disadvantages. There are if there was an option to user raid10, but since you are only using raidz1 as an option, this is the preferred,atleast for me. Because,if the drive dies, smaller drive gets faster resilvered.

 

[.n3]

OK thank, then I will for the RAIDZ! with 5 disks My prefered case has also 5 hot swap drives so it will fit perfectly.

Just for my curisosity. Is there any statistic how many drives dies during a resilvered? For a risk analysis it is important to know the likelihood.

 

[LnxBil]

No,this is raidz1, there aren't any disadvantages.

You will have space waste depending on the volblocksize and ashift value. Be aware of that.

 

[.n3]

Thank you for the hint. I read the information but do not understand the spacewast exactly. If I use the default config, everythink should be fine, right?

# zpool create -f -o ashift=12 <pool> raidz1 <device1> <device2> <device3> <device4> <device5>

 

[ness1602]

Right,i use this config in 100% cases.

 

[LnxBil]

Thank you for the hint. I read the information but do not understand the spacewast exactly. If I use the default config, everythink should be fine, right?

No, you will always have to potential for space waste with raidz*, only mirroring will not suffer from this. The degree varies of course and is defined by the ashift value, the number of disks in a vdev, the number of vdevs and the choosen volblocksize. This topic is often disregarded and people will come back to ask where their space went. Just google "zfs space waste" and you will see what I mean (PVE forum and reddit if full of it). I personally also ran in to this problem many years ago.

 

[Dunuin]

When using the default ashift=12 then you need to switch the volblocksize from default 8K to 16K for 3 disk raidz1 and to 32K for a 5 disk raidz1.

More disks also consume more electricity. And Raidz2 will increase data integrety a bit. Once a single disk fails your whole pool is without bit rot protection until the pool got resilvered. So while resilvering, the smallest error will cause unrecoverable data corruption. With a raidz2 your pool would be still bit rot protected while it is resilvering a single failed disk.