Weird RAID configurations for redundancy

Jan 12, 2026
3
0
1
Let's say I have 10 drives of 6 TB each and want to make sure the storage is safe. Can stripe/raid0 5 pairs of them to make 5 pairs of 12TB and then mirror/raid1 them so that all 5 pairs have the same data on them? i would be happy with 12 TB storage and still have data even if 4(worst case) - 8(best case) drives fail.

Yes, this is overkill. Yes, I'm paranoid. Yes, I have geological redundancy as well. I'm just curios about this in Proxmox.
 
Yes, this kind of weird raid configuration is possible doing via zfs (mirror), mdadm (raid01) or combination of "hw-raidctrl with mdadm (raid01)" while is nothing special to proxmox because possible in each linux ... but if it's really recommended is another question to have 5 times same data.
:)
 
You can do this with ZFS, albeit manually (not from webUI). You could also create 2 5-way mirror with 5 disks each, then create a RAID0 with those two vdev. Something like choosing stripping "vertically" or "horizontally". No idea on how it would perform speed wise.
I let to you to decide if that level of redundancy makes sense or not i.e. I would prefer to have another computer and use zfssend/zfsreceive to cover the case where the PSU or mobo kills all disks at once ;)
 
  • Like
Reactions: UdoB
You can absolutely do this with ZFS, I wouldn't use the RAIDZ to host VMs (os + applications) though:

This is especially true if they happen to be hdds since rotating discs ain't great for vm performance to begin with. For bulk data the story is different of course and depends on the specific workload (databases are a different thing than a fileserver etc)I would go with a striped mirror (zfs lingo for RAID10) since with that you will have a good compromise between capacity, redundancy and performance/IOPs.