J and I and Me: Approaches for Über / Fat JAR Deployment in Java EE

Approaches for Über / Fat JAR Deployment in Java EE

I believe the deployment and monitoring model for Java EE Application Server is outdated. If you want to learn more about my opinion refer to my article at JAXenter, the slides or my Interview at InfoQ. Instead I advocate a model to deploy an application including all the needed infrastructure as a large JAR file. This model is supported e.g. by Vert.x, Dropwizard, the Play Framework and of course Spring Boot. However, what about Java EE applications? While in theory there is no reason why a Java EE should not be deployed as a huge JAR file - but there are not that many solutions to actually do that. So here are some approaches that might be worth looking at:

Spring Boot actually does support some Java EE APIs. A Blog Post describes this in more detail. It shows JAX-RS with Jersey, transactions with JTA and the Java EE annotation @Transactional, and also JSR 330 annotations for Dependency Injection. However, the example still uses some Spring classes and some technologies like EJB are not supported. But Spring Boot has a quite advanced model for metrics and monitoring, see Spring Boot Actuator.
TomEE supports this model to some extend. See this older blog post. Also David Belvins of the TomEE team has quick demo and another demo.
There is Backset that supports CDI, JSF and JPA - with EJB, JTA and JAX-RS on the roadmap.
Finally this repro shows how to use Undertow with JAX-RS in a main method.

Thanks for the discussion over at Twitter and all the help I got there. If you know any other options - please leave a comment. Thanks!

¶ 10:36

Comments:

Here is a link Matthias Wessendorf provided: https://rmannibucau.wordpress.com/2015/01/20/did-you-ever-realize-openejb-applicationcomposers-is-the-ee-spring-boot/

# posted by

Eberhard Wolff : 26 Januar, 2015 12:55

If you want to take this idea one step further: Instead of bundling a "fat" JAR build a Docker image.

Put Glassfish (or any other JEE AppServer) + your deployed app into the container and voila you have a self contained JEE app.

# posted by

Anonym : 26 Januar, 2015 15:08

Thanks for the comment! However, I think creating a Docker image is something different: You still need to deal with the monitoring and the deployment model of the App Server - which IMHO are considerable drawbacks. You could create a Docker image with a fat JAR, too - but that would solve different issues i.e. providing a verified operating system with the app and more lightweight virtualization.

# posted by

Eberhard Wolff : 26 Januar, 2015 15:16

Great article

# posted by

Sandro Agboka : 27 Januar, 2015 00:47

Good read! I fully agree! Looks like app server vendors are starting to recognize the need for lightweight deployment models too. See Wildfly Swarm (https://github.com/wildfly-swarm/wildfly-swarm), Payara Micro (http://www.payara.fish/introducing_payara_micro). I guess the microservices architectures drives the mindset change.

Regards,
Andreas

# posted by

Unknown : 30 April, 2015 00:32

Kommentar veröffentlichen

<< Home

Über mich