Cluster installation options
Clusters can be installed in different flavors regarding the number and role of control plane nodes. This decision has impacts on the availability of the cluster and on how hardware resources are shared between cluster operation and workload.
The core differences between the available options are the number of control plane nodes and the role of the control plane nodes. The number of control plane nodes can be either one or three. The role can be either dedicated control plane nodes or nodes that are both control and compute nodes.
You can install Red Hat® OpenShift® cluster for the following topologies:
- A single-node OpenShift Container Platform cluster (SNO): A node that is both a control and compute.
- A three-node OpenShift Container Platform cluster : A compact cluster that has three control plane nodes that are also compute nodes.
- Highly available OpenShift Container Platform cluster (HA): Three control plane nodes with any number of compute nodes.
Single-node OpenShift (SNO)
SNO always has one single control node. You cannot change the setup in a later stage to add more control nodes and build the cluster as a high availability (HA) cluster. In a minimum installation, SNO can be installed as one single node that contains control and compute.
SNO supports adding compute nodes to the SNO cluster, but doesn’t allow adding control plane nodes. The impact of having only one control node is that updates of the control and its services cause planned outages of the cluster management. The workloads are not impacted by the update. Furthermore, in case of a disaster, the single control node must be backed up so that the cluster can get back into an operational state. If the etcd database or the root volume gets lost, the whole cluster cannot be managed anymore. If those risks are managed and downtime for the management is acceptable, SNO might be a good option for various workloads. Because IBM Z® and IBM® LinuxONE already have a very high availability, this can be an option to optimize your CPU footprint for Red Hat OpenShift on the platform. Some CPU costs are lower, since the services do not need to synchronize as for other setups.
Three-node cluster
Red Hat OpenShift clusters can be set up with three nodes as a compact cluster. Those three nodes have both roles, control and compute. Therefore, it provides the availability for the workload and management. Furthermore, updates do not impact cluster management availability. It can be migrated to a full HA cluster by adding extra nodes and separating the roles of control and compute nodes. A three-node cluster is the preferred option in comparison with SNO, when your application requires HA and an option to migrate to a full cluster at a later time. In comparison with a full cluster, the costs for this environment are slightly lower because less communication between nodes is required for combined control and compute nodes.
High availability (HA) full cluster
A high availability (HA) cluster can be set up with five nodes and more. This is the standard setup to target the highest availability and scalability. This cluster setup is built for continuous uptime even during hardware failures or maintenance.