Running federated workloads

Federated workloads are workloads that are processed on multiple Kubernetes clusters at the same time. This is relatively easy to do for loosely-coupled and embarrassingly-distributed applications. However, if most of the processing can be done in parallel, often there is a join point at the end, or at least a central persistent store that needs to be queried and updated. It gets more complicated if multiple pods of the same service need to cooperate across clusters, or if a collection of services (each one of them may be federated) must work together and be synchronized to accomplish something.

Kubernetes federation supports federated services that provide a great foundation for such federated workloads.

Some key points for federated services are service discovery, cross cluster load-balancing, and availability zone fault tolerance.

Creating a federated service

A federated service creates a corresponding service in the federation's member clusters.

For example, to...