Provision a 3-node cluster

This is the canonical lab for getting a cluster off the ground. The goal is a working 3-node cluster ready for VM workloads, with traffic separation between management and cluster communication.

Environment

Steps

1. Post-install hardening on each host

On every node:

# Disable the enterprise repo if you do not have a subscription
sed -i 's/^/#/' /etc/apt/sources.list.d/pve-enterprise.list

# Add the no-subscription repo
echo "deb http://download.proxmox.com/debian/pve bookworm pve-no-subscription" \
  > /etc/apt/sources.list.d/pve-no-subscription.list

apt update && apt full-upgrade -y

2. Configure corosync network (optional but recommended)

Edit /etc/network/interfaces on each node to add the second NIC. Use a separate physical NIC for the corosync link to avoid head-of-line blocking with management traffic.

3. Initialise the cluster on pve-01

pvecm create mycluster --link0 10.0.1.11

Verify:

pvecm status
# Expected: Quorate: Yes, Nodes: 1, Active votes: 1

4. Join pve-02 and pve-03

On pve-01, get the join information:

pvecm expected 1    # Reset expected votes before joining

On pve-02 and pve-03:

pvecm add 10.0.0.11 --link0 10.0.1.12
pvecm add 10.0.0.11 --link0 10.0.1.13

5. Verify the cluster

From any node:

pvecm status
# Expected: Nodes: 3, Active votes: 3
pvecm nodes

In the GUI, the cluster view should show all three nodes green with identical versions.

Verification

• pvecm status reports Quorate: Yes and Nodes: 3
• The cluster dashboard in the GUI shows three green nodes
• /etc/pve/corosync.conf on every node lists all three nodes
• ha-manager status (after enabling HA on at least one VM) lists the manager

Cleanup

# On the nodes you want to remove:
pvecm delnode pve-0X

# On pve-01, after removing all others:
pvecm destroy

This is destructive — only run cleanup if you’re throwing away the lab.