AWS Docker Containers running Wildfly / Java EE - domain cluster or standalone? [closed]

Question

I am running a large EE application on wildfly in containers. Each instance is currently standalone, but I dont think that this is optimal.

I imagine that I should be running a domain cluster, and have each container register itself and deregister itself as part of its life-cycle. This at least would allow me to share configuration settings and monitor the health of the cluster and of individual members from some single point.

What are the best practices in this regard?

For those eager to talk about wildfly-swarm, I have not yet started using it, but your comments will be much appreciated and are invited.

Answer 1

It's a complex question. But in general, for implementing such kind of automation you need to build right container images and inject dynamic variables later (IP/hostname and credentials). The main rule that simplifies creating self-registering instances is to put only one Worker per container. Then you can use a predefined hostname in the host-slave.xml as well as specify user and password via env variables for being able to connect each new Worker to the Controller.

The clustering in domain mode can be a tricky task, please take a look at the article that provides hints and tips on running WildFly Managed Domain in Containers: Automatic Micro Clustering and Scaling - high level overview of the potential issues while migrating WildFly from VMs to containers.