KubeCrash: Cooperating Clusters, Linkerd’s new take on multi-cluster

Linkerd has long had support for allowing communication across cluster boundaries that’s secure, independent of network topology, and completely transparent to the application. This support was designed for a world in which individual clusters were generally isolated from one another, and crossing the cluster boundary was a bit unusual — the kind of situation where you had a primary cluster and a failover, or a separate cluster in another AZ that you only wanted to use for special situations.

The world is changing, though. Now, it’s also common to see clusters used more like namespaces, with a single application spanning multiple clusters by design. Linkerd 2.14 brings powerful new tools to the table to handle this new world, with new features that support direct pod-to-pod communication across clusters, while preserving all the power of Linkerd’s workload-based identity, policy management, and sophisticated routing capabilities configured with the Gateway API. In this talk, we’ll break down how it all works and how it can help you in your multi-cluster world.