Multi-Node Cluster
A multi-node cluster consists of three or more physical servers assembled together into a single Kubernetes system. A three node cluster is capable of losing an entire physical node with little or no visible impact to end users—administrators can demonstrate this ability by literally pulling the power on a server while the system is in use.
Three or more nodes are required. A two-node cluster cannot provide redundancy because of quorum requirements inherent to the Kubernetes platform.
When multiple nodes are present in the cluster, client connections can be received on any node; see the 'Cluster Load Balancing' section (below) for a description of different techniques that can be used to route incoming traffic to different servers.
A three-node cluster utilizing the same hardware as described previously (16 cores, 32GB RAM) is sufficient for serving 2,500 concurrently active users. By adding additional nodes to the cluster, virtually unlimited scale is achievable.
Note: Simply increasing the nodes in the cluster will not produce a resilient or highly-scalable system. To achieve these goals, special configuration of Instant Connect is required to assure system components are sufficiently distributed across nodes. Consult the product guide for details.