EasyUC
English
Aurora

Choose appearance

Software, updates and licence files are available in the customer portal. Go to downloads

Adding and removing nodes

Requirements, the procedure in the Manager, what happens technically and why a cluster should have three nodes.

Applies to: EasyUC 4.8.7 Last updated: October 2026

Requirements

  • A prepared EasyUC VM with the IP address you enter in the Manager.
  • The first node can reach this VM via SSH on TCP 15222.
  • Recommended: all nodes in the same network segment, with no firewall filtering between them. The ports are listed under Between cluster nodes.

Adding a node in the Manager

  1. Open the node overview in the Manager.

  2. Choose Add node.

  3. Enter the IP address of the prepared VM and, optionally, a name.

  4. Confirm. The Manager sets up the node, which then appears in the overview with its CPU and RAM usage.

What happens technically

The Manager sets up the new node on its own:

Adding a node The administrator enters IP address and name in the Manager. The Manager connects to the new VM via SSH on TCP 15222, sets the hostname and installs K3s from local files. The VM joins the cluster with the cluster token via TCP 6443 and becomes a server node. The node then appears in the overview. Administrator browser EasyUC Manager first node New VM prepared Cluster K3s · etcd Add node: IP, name SSH · TCP 15222 Set hostname K3s from local files Join with cluster token TCP 6443 Server node: control plane, etcd, workloads Node in the overview
Adding a node
  • The Manager connects to the new VM via SSH on TCP 15222.
  • It sets the hostname and installs K3s from the appliance's local files.
  • The node joins the cluster with the cluster token, via the Kubernetes API on TCP 6443.

Every node is a server

Every added node joins as a full K3s server. It is part of the control plane, an etcd member and also runs workloads. Pure worker nodes are not offered.

This means every node counts towards the etcd majority, see Why three nodes.

Removing a node

When you remove a node in the Manager, the following happens:

  1. Cordon

    No new workloads are scheduled on the node.

  2. Drain

    Workloads move to other nodes.

  3. Uninstall

    The node is uninstalled.

Important

Every service runs as a single instance. During the drain, the components on the node are stopped and restarted on another node, so the affected services are briefly interrupted. Remove nodes in a maintenance window.

Note

The last node cannot be removed. The IP address can only be changed on single-node installations.

Why three nodes

The control plane stays operational as long as a majority of the server nodes is available. This is a general principle of etcd and therefore of Kubernetes.

etcd majority and tolerated failures
Server nodesMajorityTolerated failuresMeaning
1 1 0 No redundancy
2 2 0 No gain in fault tolerance. If one node fails, the majority is lost.
3 2 1 Survives the loss of one node. Our recommendation.
5 3 2 Survives the loss of two nodes.

For redundancy we recommend three nodes, and in general an odd number. This protects the control plane. How the services themselves react to a failure is described under Redundancy.

Component placement

  • Several web and API components must not run twice on the same node (anti-affinity per host).
  • teams-sbc runs on the node where acd-media runs.
  • Both use the host network, so RTP firewall rules and the 1:1 NAT for Teams must point to that node.