Je vous mets une copie de mon post sur unitedbsd (en anglais)

I tried a dedicated server at OVH and tried to install several BSDs...
It is entirely possible to install OmniOSce (r151046 LTS and the latest r151052) on this dedicated server "KS-STOR | Intel Xeon-D 1521" at OVH.

The process is simple, whereas you can not use any installer version (.iso, dd-usb) nor OVH template with cloud.qcow2 image. You must use the cloud.raw image.

Boot into rescue mode (debian-10), restart your server then log in via SSH

ssh root@SERVER-IP

List the storage devices connected to the server :

lsblk

Erase the disks (not mandatory.. your choice)

nohup dd if=/dev/zero of=/dev/sda bs=1M &
nohup dd if=/dev/zero of=/dev/sdb bs=1M &
nohup dd if=/dev/zero of=/dev/sdc bs=1M &
nohup dd if=/dev/zero of=/dev/sdd bs=1M &
nohup dd if=/dev/zero of=/dev/nvme0n1 bs=1M &

If you don't want to erase the NVMe entirely, you really should erase the beginning and the end to be sure that there isn't an old ZFS pool label (I had a previous zpool from my FreeBSD installation on it which made the server crash after 5min the first time - FAULTED rpool) :

zpool labelclear -f /dev/nvme0n1
dd if=/dev/zero of=/dev/nvme0n1 bs=1M count=100
dd if=/dev/zero of=/dev/nvme0n1 bs=1M seek=$(( $(blockdev --getsz /dev/nvme0n1) * 512 / 1024 / 1024 - 100 )) count=100 

Monitor when the erase operations are complete:

ps -ef | grep 'dd if=' | grep -v grep

Install parted, zstd et fwupd

apt update
apt install parted zstd fwupd -y

Mettre à jour la liste des firmwares disponibles (si si, c'est important).

fwupdmgr refresh

fwupdmgr get-updates

fwupdmgr update

Create GPT tables on the disks (again, not mandatory, just to clean...)

parted /dev/sda mklabel gpt
parted /dev/sdb mklabel gpt
parted /dev/sdc mklabel gpt
parted /dev/sdd mklabel gpt
parted /dev/nvme0n1 mklabel gpt

Download the OmniOSce cloud.raw image and uncompress it

wget https://downloads.omnios.org/media/stable/omnios-r151052.cloud.raw.zst
zstd -d omnios-r151052.cloud.raw.zst

Burn the image to one of the HDDs:

dd if=omnios-r151052.cloud.raw of=/dev/nvme0n1 bs=1M status=progress
sync

Import the pool:

zpool import
zpool import -f rpool

Mount it:

mkdir -p /mnt/root
mount -t zfs rpool/ROOT/omnios-r151052 /mnt/root

Add your public SSH key:

mkdir /mnt/root/root/.ssh
echo "YOUR-PUBLIC-SSH-KEY-HERE" > /mnt/root/root/.ssh/authorized_keys 

Export the pool

zpool export rpool

In your OVH interface, set the boot to disk this time (instead of rescue).
Restart the server and use the IPMI/KVM console (via your web browser, for example... it works very well with Firefox contrary to what is said) to watch the server boot from the raw image.
It will stop displaying after : "Loading /boot/defaults/loader.conf"...
But goes on...

Wait for 1-2 minutes more, erase your local known_hosts file then connect by SSH:

ssh root@IP-SERVER

Tadam !

You should quick deactivate some services :

svcadm disable svc:/network/routing-setup:default
svcadm disable svc:/network/routing/route:default
svcadm disable svc:/network/routing/ndp:default
svcadm disable svc:/system/cloud-init:initlocal
svcadm disable svc:/system/cloud-init:init
svcadm disable svc:/system/cloud-init:modules
svcadm disable svc:/system/cloud-init:final

Also, create an ipadm backup script (I lost my static IP configuration and had to go back to the rescue mode.. don't know why. I had previously configured it from dhcp to static, but it did not remain after reboot):

vi /etc/rc3.d/S99network-setup

#!/bin/bash

# Erase the IP configurations of the network cards :
ipadm delete-ip ixgbe1 2>/dev/null
ipadm delete-ip ixgbe0 2>/dev/null

# Wait for them to be ready again
sleep 5

# Configure the first interface :
ipadm create-ip ixgbe0
ipadm create-addr -T static -a local=SERVER-IP/CIDR ixgbe0/v4

# Wait again
sleep 5

# Add the default route :
route -p add default GATEWAY-IP

exit 0

Give the permission to execute :

chmod +x /etc/rc3.d/S99network-setup