Prerequisites

Before deploying Service Mesh Manager on your cluster, complete the following tasks.

Create a cluster

You need a Kubernetes cluster to run Service Mesh Manager (and optionally, Streaming Data Manager). If you don’t already have a Kubernetes cluster to work with, create one with one of the methods described in Create a test cluster.

CAUTION:

Supported providers and Kubernetes versions

The cluster must run a Kubernetes version that Service Mesh Manager supports: Kubernetes 1.21, 1.22, 1.23, 1.24.

Service Mesh Manager is tested and known to work on the following Kubernetes providers:

  • Amazon Elastic Kubernetes Service (Amazon EKS)
  • Google Kubernetes Engine (GKE)
  • Azure Kubernetes Service (AKS)
  • On-premises installation of stock Kubernetes with load balancer support (and optionally PVCs for persistence)

Resource requirements

Make sure that your Kubernetes cluster has sufficient resources. The default installation (with Service Mesh Manager and demo application) requires the following amount of resources on the cluster:

Only Service Mesh Manager Service Mesh Manager and Streaming Data Manager
CPU - 12 vCPU in total
- 4 vCPU available for allocation per worker node (If you are testing on a cluster at a cloud provider, use nodes that have at least 4 CPUs, for example, c5.xlarge on AWS.)		 - 24 vCPU in total
- 4 vCPU available for allocation per worker node (If you are testing on a cluster at a cloud provider, use nodes that have at least 4 CPUs, for example, c5.xlarge on AWS.)
Memory - 16 GB in total
- 2 GB available for allocation per worker node		 - 36 GB in total
- 2 GB available for allocation per worker node
Storage 12 GB of ephemeral storage on the Kubernetes worker nodes (for Traces and Metrics) 12 GB of ephemeral storage on the Kubernetes worker nodes (for Traces and Metrics)

These minimum requirements need to be available for allocation within your cluster, in addition to the requirements of any other loads running in your cluster (for example, DaemonSets and Kubernetes node-agents). If Kubernetes cannot allocate sufficient resources to Service Mesh Manager, some pods will remain in Pending state, and Service Mesh Manager will not function properly.

Enabling additional features, such as High Availability increases this value.

The default installation, when enough headroom is available in the cluster, should be able to support at least 150 running Pods with the same amount of Services. For setting up Service Mesh Manager for bigger workloads, see scaling Service Mesh Manager.

Install the Service Mesh Manager tool

Install the Service Mesh Manager command-line tool. You can use the Service Mesh Manager CLI tool to install Service Mesh Manager and other components on your cluster.

Note: The Service Mesh Manager CLI supports macOS and Linux (x86_64). On Windows, install the Windows Subsystem for Linux (WSL) and use the Linux binary.

  1. Install the Service Mesh Manager CLI for your environment.

  2. Set Kubernetes configuration and context.

    The Service Mesh Manager command-line tool uses your current Kubernetes context, from the file named in the KUBECONFIG environment variable (~/.kube/config by default). Check if this is the cluster you plan to deploy the product by running the following command:

    kubectl config get-contexts

    If there are multiple contexts in the Kubeconfig file, specify the one you want to use with the use-context parameter, for example:

    kubectl config use-context <context-to-use>

Deploy Service Mesh Manager

After you have completed the previous steps, you can install Service Mesh Manager on a single cluster, or you can form a multi-cluster mesh right away.

Note: The default version of Service Mesh Manager is built with the standard SSL libraries. To use a FIPS-compliant version of Istio, see Install FIPS images.

Select the installation method you want to use:

You can install Service Mesh Manager on a single cluster first, and attach additional clusters later to form a multi-cluster mesh.

Accessing the Service Mesh Manager binaries

Create a test cluster