Add a node

Adding a node is the install’s join flow, on a running cluster.

1. Create a join code. Console: Admin › Metal › Add node.
2. Install the machine from the same ISO as a joiner, choosing Join a cluster. An older ISO upgrades itself first, which is untested on real hardware.
3. Approve it, usually as the same as the first node.
4. Revoke the join code.

What to expect

• Role. Below the planned control-plane count the node is hybrid. Beyond it, hypervisor: it runs instances, not the control plane. The count of control-plane nodes cannot be left even beyond the planned size.
• The cluster must be ready. A join waits until the cluster is READY, no other node is joining or being removed, no upgrade runs and every node has applied the current membership. Joins run one at a time.
• The network. Storage and overlay must reach the new node at their MTU. If they do not, the join fails and readiness names the path.
• Disks. A disk approved as ceph-osd becomes an OSD on Ceph’s next reconcile pass, with no further step. A compute disk becomes a pool on that node.
• Existing nodes are not rebooted. Their control plane and a few monitoring services restart to take the new peer, one node at a time, in join order.

Afterwards run stratumctl admin doctor: it flags a node whose OVN version has drifted from the cluster’s before it breaks floating IP traffic. Then check readiness.

If a join fails, the node is taken back out of every cluster it reached. Fix the cause and stratumctl admin nodes joins retry <id>.