Platform9 Managed Kubernetes Multi-master Architecture in Bare Metal Environment

A multi-master architecture comprises redundant master nodes for high availability in a Kubernetes cluster. More than one master nodes are present in a multi-master Kubernetes cluster. At a given time only one master node is active. When the active master node goes down, one of the other master nodes takes over and manages the cluster workload.

Multi-master Cluster in Platform9 Managed Kubernetes

Platform9 Managed Kubernetes supports highly available multi-master Kubernetes clusters in a bare metal environment with an operating system such as Ubuntu or CentOS.

When you want to design a multi-master Kubernetes cluster with Platform9 Managed Kubernetes, it is recommended to have a a minimum of three or five master nodes.

When you create a cluster with three or more masters, you can scale up or scale down the number of master nodes in the Kubernetes cluster, without a problem.

In case of multi-master clusters, one master node is active at a time. The active master node handles client requests. If the active master node goes down, one of the other master nodes becomes active.

More than half of the master nodes must be available at any given point in time for the etcd to work. The cluster aforementioned condition must be satisfied for etcd to work properly.

The Platform9 Managed Kubernetes API server runs on each master node of the multi-master Kubernetes cluster.

The etcd key-value store used by Kubernetes, has the information related to the cluster and etcd is available to all master nodes to harness the cluster-related information.

You can create multi-master Kubernetes clusters through Platform9 Clarity UI or through the Platform9 Managed Kubernetes API. Refer to the Platform9 Managed Kubernetes API documentation (accessible through Platform9 Clarity UI) for details on creating multi-master Kubernetes clusters using the API.

Architecture diagram

June 18, 2018