The struggle with Hazelcast queue persistence
In this blog I will outline why we used Hazelcast for queueing messages in-memory distributed over a cluster and how we achieved higher resilience by persisting the queue’s content. I will explain the pitfalls and difficulties that we encountered and how I constantly switched between praising and condemning Hazelcast. The problem to solve I’m currently working in a project for a large customer data backend. The prod system consists of a load balanced cluster of five VMs each running two Tomcat instances hosting our application.