The "lra/test" module has been introduced to run integration tests with Arquillian containers. The main issue of this module is about a design choice made during the development phase: "lra-coordinator", "lra-participant and the entire "test/basic" module (which hosts the actual test cases) are deployed to the same Arquillian container (particularly, when the Wildfly container is selected). As specified in [1 - pag.175] and in the documentation of the Arquillian project, integration tests against REST APIs should "run as clients". This means that the actual test cases should run in a JVM while the web app(s) should run in the container (thus, in another JVM). The current design of the "lra/test" module does not respect this advice and, therefore, breaks the separation between server and client (and, as a consequence, does not entirely leverage the power of Arquillian).

To solve this issue, a review of the deployment process should be considered as the mechanism to (automatically) deploy the "lra-coordinator" and the required "lra-participant(s)" is based on the assumption that services and test cases are deployed to a unique container.

----------------------------------------

[1] Arquillian Testing Guide - John D. Ament