Linkerd: Up and Running

Chapter 12. Multicluster Communication with Linkerd

Every Kubernetes cluster represents a single security and operational failure domain. As you look at scaling out your platform to accommodate more teams, more customers, and more use cases, you will inevitably run into the question of how you want to distribute your apps. Do you want to use large regional clusters with all your production apps in one place? Do you want to use purpose-built clusters for each app or each team? Most teams end up somewhere in the middle, with some shared clusters and some purpose-built for certain apps or categories of apps.

Linkerd aims to make the technical implementation problems around running multiple clusters easier to solve.

Types of Multicluster Setups

Linkerd supports two styles of multicluster configurations: gateway-based multicluster and Pod-to-Pod multicluster. Gateway-based multicluster setups are easier to deploy; Pod-to-Pod setups offer more advanced functionality. You can choose which is best for a given situation, and you can even use both in the same cluster at the same time, if desired.

Gateway-Based Multicluster

Linkerd’s gateway-based multicluster setup routes communications between clusters through a special workload that Linkerd calls a gateway, which is reachable via a LoadBalancer Service. This means that gateway-based multicluster connections don’t require any particularly demanding network configuration: all that’s required for gateway-based multicluster communications ...