Thread 106343857

Right now, I have my proxmox OS installed on /dev/nvme0 using the ext4 format. I am using LVM thin to create the volume group called "pve", which spans /dev/nvme0 and /dev/sda. In the future when I buy more pairs of drives, I want to create a new volume group called "pvf" spanning some new drives /dev/nvme1 and /dev/sdb. I want to configure my server so it automatically mirrors (like in RAID 1) the data lv from the pve vg to a new lv in the pvf vg so that there is data redundancy in the event of a failure. I also want to be able to expand pve in the future by adding /dev/sdc to pve and expand the mirrored volume group by adding /dev/sdd to pvf in the future while retaining the mirroring characteristics. I also want to avoid having to wipe data when extending the volume group with new drives or configuring the data mirroring. How can I accomplish this?

Sounds like you want zfs. Use zfs striped mirrors its like raid10 and you can just add disks to it.

bump for yoyu op, but i think you'd have better luck breaking your question into parts or summarizing it better or posting it in /hsg/

>>106344411
definitely needs running through chatgpt

Since you’re already using Proxmox:

If you want the simplest redundancy/expansion model go ZFS with mirrored vdevs. It’s the most future-proof and aligns exactly with your “add pairs of disks” plan.

If you want to stay with LVM use a single VG with lvcreate --type raid1 and forget about “separate mirroring between VGs”.

If you insist on keeping pve and pvf separate you need DRBD (or similar block replication like LINSTOR).

>>106343857 (OP)
Just reinstall it by using a Debian install. Do mdadm raid 1 first, then set up Proxmox on top of the debian install.

I tried cleaner formatting in a different board before and got relentlessly mocked for "reddit spacing". Never again XD

>>106344407
I don't know a lot about ZFS (or sysadmin in general), but when extending a zpool with a new drive, does it concatenate the drive by default like LVM? Or does it stripe data thus requiring rebalancing?

>>106343857 (OP)
the proxmox documentation really sucks ass and dick and balls

>>106343857 (OP)
this might not be what you're looking for, but you should also question how much you want inside proxmox versus provided to it.

all my lxc (20 or so) live on one partition that's about 250GB, and one special lxc is given ownership of the computer's other drives, and passes them to the other lxc using sshFS over the private network. this is nice because it means the lxc installations are small and simple.

>>106343857 (OP)
I just installed debian and then the proxmox packages onto an ext4 drive because I'm not autistic enough for zfs.

why even use this shit?
just install a small system like arch and use kvm + qemu or whatever