When deploying API Connect on an OpenShift cluster, what are the minimum number of nodes required for high availability for the platform?

For a high-availability (HA) platform deployment, both the underlying OpenShift cluster and the IBM API Connect application must be resilient. OpenShift requires a minimum of three master (control plane) nodes to establish a stable etcd quorum, which is essential for control plane HA and preventing "split-brain" scenarios. Additionally, a minimum of three worker (compute) nodes is the standard requirement for a production API Connect deployment. This allows for proper pod distribution across failure domains, supporting rolling updates and ensuring the service remains available even if one worker node fails.

A. 2 Master Nodes and 2 Worker Nodes: Two master nodes cannot form a stable etcd quorum, making the control plane susceptible to failure and not highly available.

B. 1 Master Node and 2 Worker Nodes: A single master node represents a single point of failure for the entire cluster's control plane, which is a non-HA configuration.

C. 1 Master Node and 3 Worker Nodes: A single master node is a single point of failure, making this an invalid configuration for a high-availability platform.

1. IBM API Connect v10.0.5 Knowledge Center: In the section on planning for deployment

it specifies the node requirements. For production environments

it states: "For a production deployment

a minimum of 3 worker nodes is recommended." This addresses the worker node requirement for the application's HA.

Source: IBM API Connect v10.0.5 Documentation

"Hardware and software requirements for OpenShift". (Navigating under Installing -> Planning your deployment -> Deployment on OpenShift and Cloud Pak for Integration).

2. Red Hat OpenShift Container Platform 4.10 Documentation: The official documentation for OpenShift

the underlying platform

details the requirements for a production-level (HA) cluster. It specifies the need for three control plane nodes.

Source: Red Hat OpenShift Container Platform 4.10 Documentation

Section 2.3

"Node requirements". The documentation states

"For production environments... a cluster requires three control plane nodes." It also recommends

"For production or other multi-node clusters

use at least three compute nodes."