Basic Prerequisites

This article describes the generic node level prerequisites for creating any type of Kubernetes cluster using PMK.

Supported Clouds

Platform9 supports the following Public Clouds:

• AWS - Create native PMK clusters using PMK provider for AWS and/or import EKS clusters
• Azure - Create native PMK clusters using PMK provider for Azure and/or import AKS clusters
• Google - Import GKE clusters

Supported Operating Systems

Regardless of whether you are creating a Kubernetes cluster on-premises or in a public cloud using PMK, PMK supports the following operating systems for the nodes of a Kubernetes cluster.

• CentOS 7+ (64-bit)
• RHEL 7+ (64-bit)
• Ubuntu LTS 18.04 (64-bit)
• Ubuntu LTS 20.04 (64-bit)

System Requirements

These requirements ensure that your Kubernetes cluster nodes are provided with enough resources to run core Kubernetes services, monitoring and other required services that PMK deploys, and your containerized applications that will run on top of Kubernetes.

Test / Proof of Concept (PoC) Clusters

To create a simple test Kubernetes cluster using PMK, each node that is to be part of the cluster should have:

CPU Minimum: 4 vCPUs

RAM Minimum: 16 GB of RAM

Storage

• Minimum Size: 30 GB
• Free Space: 20 GB

Production Clusters

To run a production grade Kubernetes cluster, each node that is to be part of the cluster should have:

CPU 1 CPU Core per 4 GB of RAM.

RAM Minimum of 16 GB of RAM

Storage

• Minimum Size: 60 GB
• At least 40 GB of free disk space. 20 GB of disk space will be used by the Operating System and Kubernetes components.
• Make sure that the primary / OS volume has at least 8 GB of unallocated storage space. This would be the default path used by etcd on your master nodes, unless you explicitly specify a different path for etcd storage.
• Make sure that the /var/lib directory exists on a file system with at least 30 GB of free disk space. This path is used by your containers running on Kubernetes for their local disk / storage.

Networking

• At least one physical (or VLAN backed) NIC with an IP address.

This type of configuration should be used to create a production cluster that you will use either for your CI/CD or to run production applications.

Production Cluster Best Practices

• A production Kubernetes cluster should be created with at least 3 master nodes.
• Best practice is to not run your workloads on the masters in a production cluster, but to have separate worker nodes in the cluster with sufficient capacity to run your workloads