[WildFly] Deployment cannot be rolled-back


we use Camunda since a while, but with the upgrade to version 7.15 and WildFly 22 we run into the following problem.

What we have is a EJB process application as described in the documentation (@Singleton, @Startup, etc.).
Additionally we have another Singleton (also with @Startup), which starts our components. In some cases an exception could be thrown, because for example the configuration for the application is checked. Like in the process application we have a PostConstruct and a PreDestroy handler.

If an exception is thrown in our components, the WildFly stucks in the deployment process and does not come to an end - at least until the deployment timeout is reached. The pre-destroy handlers are not executed. The WildFly then is not really usable anymore and cannot be stopped in a normal way. The process has to be killed.
In the past (Camunda 7.14 and WildFly 20) it worked fine. The pre-destroy was called and the deployment was rolled-back.
When I remove the process application it also works fine.

Is this a known problem or can you give us any advise to get rid of this problem? The kill of the WildFly process after every failed deployment (which can happen often during development time) and a restart is not an option for us.

As a hint: I used the provided WildFly with integrated Camunda process engine

I created an example application which shows a small (but important) part of our application:

I would be very thankful for any support!

Hey @timonz,

thanks for bringing this up and for creating https://jira.camunda.com/browse/CAM-13982 for this.
We will investigate the issue and let you know what we find out.