End-to-End Kubernetes with Rancher, RKE2, K3s, Fleet, Longhorn, and NeuVector

Regional vs. Hub & Spoke: A Decision Guide

At this stage, we have a working cluster deployed using RKE2. Our goal is to deploy Rancher Manager and use it to manage this cluster and the workloads running on it. To understand the different strategies for deploying Rancher Manager, let's consider the following scenario: Let's say that for high availability and disaster recovery reasons, we have multiple clusters deployed in different regions. We want to use Rancher Manager to manage all these clusters.

A few questions arise at this level: Should we deploy Rancher Manager in each region, or should we deploy a single Manager to manage all the clusters? What are the pros and cons of each strategy?

There are 2 main strategies for deploying the Manager in this scenario: the regional and the hub & spoke strategy.

• The regional strategy: Deploy Rancher Manager in each region and use it to manage the clusters in that region. This option is more complex and requires more resources.

• The hub & spoke strategy: Deploy Rancher Manager in a single region and use it to manage all the clusters in all regions. This option is simpler and requires fewer resources.

In both cases, you can run the Manager on a Kubernetes cluster or a standalone server. If the high availability of the Rancher Manager is a requirement, deploying it on a cluster is the way to go. It is also worth mentioning that your workloads should be deployed on a separate cluster from the Rancher Manager.

ℹ️ Deploying Rancher Manager on the same cluster as the managed cluster seems to be the simplest option, but it is not recommended.

In the following sections, we will discuss the pros and cons of our two strategies and provide decision criteria that may help you choose the best approach for various scenarios.

The Hub & Spoke Strategy

We can qualify this strategy as "centralized" because a single Rancher Manager instance is used to manage multiple clusters in different regions. It has the following advantages:

• Centralized Management: Simplifies visibility and operational control by providing a single control plane for managing all regions and environments.

• Cluster Independence: Kubernetes clusters operate independently, even when the Rancher hub is temporarily unavailable.

• Efficient Monitoring: Centralized reporting and metrics make monitoring and troubleshooting across all downstream clusters easier.

• Resource Optimization: Consolidates management tasks into a single Rancher instance, thus reducing the effort required to manage multiple Rancher installations.

• Global View: Provides a unified interface for managing Kubernetes clusters across geographies, which benefits multi-region deployments.

However, this strategy has some disadvantages:

• Network Delays: Communication between the Rancher hub and downstream clusters depends on network reliability, which may affect performance.

• Risk of Downtime: If the Rancher hub fails, global provisioning of new services pauses until recovery. Individual clusters remain operational, but their central management is unavailable.

• Scaling Concerns: The hub requires more resources as the number of downstream clusters increases, so capacity planning must be done carefully.

ℹ️ A downstream cluster is a Kubernetes cluster attached and managed by a Rancher server.