this gist is part of this series
- Connect all 3 NUC to power etc
- Disable secure boot in BIOS on all nodes
- Connect all 3 TB4 cables in a rinng topology:
using the numbers printed on the case of the intel13 nucs connect cables as follows (this is important):
- node 1 port 1 > node 2 port 2
- node 2 port 1 > node 3 port 2
- node 3 port 1 > node 1 port 2
- Prepare bootable usb key by downloading proxmox iso and tools such as 'rufus'
- Boot from USB
- choose GUI (why not!)
- accept terms
- choose the SSD NOT the NVME for the install
- set country, keyboard, etc
- choose a password (this will be for the user - root) and set email
- choose enp86s0 as the management interface (on proxmox 8 you will see the thundebolt netwrok too if cables are connected)
- set a name for each node - i used
pveX.mydomain.com
where X = the node number i wanted - set fixed IPv4 address for the management network
- 192.168.1.81/24 for node 1
- 192.168.1.82/24 for node 2
- 192.168.1.83/24 for node 3
- set external DNS server and gateway
- let the software partition disk as it sees fit
- review summary page
- removed the USB key and reboot when prompted
You should now have 3 nodes. Each should be accessible over the network with the IP addresses you specified. e.g 192.168.1.81:8006, 192.168.1.82:8006, 192.168.1.83:8006
The default username is root and you use the password you defined earlier. Consider making constrained users and doing security properly. I won't be doing that, i will be using root as this a PoC.
Note: i will not be using enterprise packages as this is a PoC for evaluation / homeLab purposes only
In node-name > updates > repositories
perform the following:
- add
No-Subscription
- add
Ceph Quincy No-Subscription
- disable
pve-enterprise
- disable
ceph-quincy enterprise
too
Open the shell and perform apt update && apt upgrade
to update all components and then reboot.
I think this is likely because proxmox as a base, is not heavy on the I/O and disk performance, another popular alternative Unraid, actually runs off a USB key. Thus, you have more "options" if you save your full NVME drive for after install, then it's easier to have that drive up for Ceph, or as partition it and wipe it as needed, and your actual containers and VMs can go on the NVME.