Running JAXRS services in the Cloud in 5 minutes… or less!

Here is a really simple post about how to push REST services in the Cloud. Nothing really technical nor advanced, just some notes and sample using amazing tools CXF + Heroku…

Last time I was speaking about putting some SOAP Web services in the Cloud with Heroku, this time it is the same with REST services… The approach is exactly the same but it uses the JAXRS implementation provided by Apache CXF.

The REST service illustrates how to annotate the Java interface to returns JSON-based responses like:

Once implemented and configured (it uses Spring with the famous WEB-INF/beans.xml file), pushing it to Heroku is as simple as last time, nothing new here. Heroku needs a Procfile to start, and the Maven-based project is configured to generate what the Procfile needs: A shell file which launches a Jetty instance running Apache CXF and all the REST stuff.

The code is located at You can run it locally to test before pushing to heroku:

  • mvn install
  • sh target/bin/webapp

Petals DSB is not Petals ESB

Almost true… In fact Petals DSB uses and extends Petals ESB in several ways. When I started to think about extending the Enterprise Service Bus, it was just to avoid all the JBI stuff at the management level i.e. use a real, simple and efficient API. So I added many management stuff exposed as Web services : Bind external services, expose internal services (oh yes bind + expose = proxy), get endpoints, activate things, etc…
One other goal was to avoid to use closed protocols for inter node communication: The ESB uses at least three ports to create inter node communications for JMX, NIO and SOAP. So why not, just using an open protocol like SOAP and just one port? This is what I did, I changed some implementations to use the same port and the same protocol for all services.

All this stuff has been developed focusing on extensibility and easier development. This is mainly because the ESB can be hard to extend for newbies, there are so many things inside… Today the DSB is not only a Distributed Service Bus, it is also a framework so that developers can easily extend the DSB without the need to know how it works inside. Some examples? Want to expose a kernel service : Add the JAXWS @WebService annotation. Want to subscribe to Web service notifications : Annotate you Java method with @Notify. Want to be notified about new endpoints : Add the @RegistryListener annotation. That’s all, you do not have to search for the Web service server to expose your service, nor read all the WSN documentation to receive notifications, … Simple, efficient.

There are other things you can do, mostly all is detailed in the DSB page.

Create a management application in 3 hours with Play

Story, code, compare

Yet another ‘nightly project’ (thanks to current house build project and the lack of sleep it brings). This time I needed to be able to manage the so-famous Service Bus from some Web enabled tooling. I already developed such tool in a research project but the fact is that the licence of some libraries are not compatible with the petalslink open source approach. The second thing is that it is GWT based (which bores me, has almost 100 libraries dependencies and takes 10 minutes to compile). So this application is a good candidate to compare development productivity between GWT and the Play Framework. Here is a summary of the application creation:

  • I bootstraped the application yesterday during lunch between 1PM and 2PM. I was already able to invoke most of the interesting service bus actions with the help of service bus SOAP API.
  • I added some pages and actions last night, let say that since it was between 11PM and 1AM I was not very productive…
  • I fixed some bugs this morning

As a result, I think I worked around 3 hours on this application. I think I spent one hour to resolve a dependency conflict between Play and a CXF dependency but as a result I have a good result which is almost equivalent in functionality to the GWT based application. I still miss some operations but I do not need them for now… I feel ashamed to say how long it tooks me to create the GWT version…


Let’s talk about deployment… I did not have time to play with heroku to push this application in the cloud. Tis will be a future step but since I use Play and git, I am able to push the code to github and then pull it on an OVH server I rent. I am able to provide this instance for some project partners if they want to manage the service bus. The time it took? 5 minutes (wget play, git clone and play start…).


The Play enabled application is available on github at chamerling/dsbmanager-webapp.

Some Play Framework, Service Bus, WS Notification and Web Sockets…

The Context

In the previous post I was introducing some tests I did with Play Framework and Web sockets. To summarize, it was just ‘about’ receiving messages on the Play! application and pushing them to the browser. This time, let’s go one step forward: Let’s add some infrastructure stuff to do something more real…

In the current case, I want to introduce a service bus which allows to create a real integration between service consumers and producers (I need to write another article in response to this nice videocast about integration from Zenexity guys ‘EAI & ESB n’apportent rien si les applications ne sont pas intégrables et interopérables’, but I really need more time to explain my thoughs…).
Using a service bus in the current case (and not for this case in fact…) must bring some added value. Here I choose to show that a service bus, even if it is not so lightweight at all, can provide some real cool features that you can have out of the box. Now, let’s add some event stuff to switch to an event-based world where we can have tons of event producers and let’s say thousands of event consumers. Since I can not setup such hude amount of actors, let’s say that we have one event producer and two event consumers:

  • The event producer wants to publish some stuff somewhere. To illustrate, let’s say that we have a weather sensor connected to our platform.
  • The platform provides a list of topics which producers can use to publish data. One is the weather topic which will be used by the producer above.
  • The event consumers want to be notified on new weather data i.e. as soon as the weather sensor publish new data. To keep it simple, they need to subscribe to the weather topic provided by the platform.

To recap, in the event context, the event producer only knows the service he has to push weather data to, the event consumers just have to subscribe to a topic they are interested in. All the knowledge stuff about producers, consumers, topics and all the mapping is delgated at the service bus level. Yes, true it is exactly like in some topic-based messaging stuff because at the end of the day, it is topic-based messaging stuff…

The Stuff

Now we can speak about the stuff we are going to use in the software point of view for notificaiton actors…

  • The event producer will not be a sensor but a Play! application. The application sends Web service notifications message to the service bus on a given topic.
  • The service bus is (of course) the Petals Distributed Service Bus with some Web service notification modules inside.
  • The event consumers are 1/A Play! application exposing a Web service to receive notification it subscribes to and 2/ A local java application displaying OS X notifications using Growl (let’s use JavaGrowl I published some days ago…). Note that the Play! application which have subscribed to notifications pushes them to their clients (browsers) using the funky Websocket stuff.

Let’s look at what really happens in a short video:

  • I use the Play! powered application play-soap-wsnclient to subscribe to notification on behalf of the play-soap-websocket Web application.
  • In Eclipse, I start a Web service notification powered Web service which subscribes to the same notification topic. Its listener is configured to use JavaGrowl to display incoming notifications.
  • I use play-soap-wsnclient to send notification messages to the notification service hosted on the service bus.
  • Once received, the service bus forwards the notification to all the subscribers using internal routing and WSN stuff.
  • The play-soap-websocket Web application receives a notification and push it to the client browser using Websocket.
  • At the same time, the Java application also receives a notification and display it using Growl.

One (or more) step(s) further…

And what if we have something which is not a Web service which subscribes to notifications? With the help of a service bus like Petals ESB/DSB, we just have to add a component which knows how to speak with the subscriber. For example, let’s say that SOAP is bad and that REST is the best thing ever. Can we have REST services receiving notifications? Yes, we can! Let’s just add the REST connector to the service bus. Another protocol/transport/format? Develop and add the new one. This is where the service bus can also help you. Hopefully, there are also other things which are possible with a service bus, we will see it later in other posts if I have time (as usual): Let’s think about business processes, service orchestration, transformation…
Next time we will have a look on what we can do if we use the distributed feature of the service bus, for example, receiving some notifications on one node and be able to notify subscribers which are bound to other nodes…

Source Code

Invoke business services from service bus kernel

Let’s introduce what it means…

While Petals ESB does not provide any solution to invoke services directly from the kernel/component Java code, the DSB now provides a solution for that… By using this DSB feature, we can really use the power of service oriented architecture directly in our code. No more direct calls to services, you just need to call the client and say which service you want to invoke. The service bus will resolve the endpoint and route the message to the right place by using the right way/path. By using this way, you do not have to care if you need to speak with a SOAP service, a JMS queue or a REST service: Just call the service, the DSB will do the rest.

Now some code…

So, let’s have a look on the client code and first, let’s imagine that you have a web service which is bound to the service bus. Once bound, a DSB endpoint is available inside the DSB, and all messages which are sent to this endpoint will be forwarded to the final web service. One basic approach is to create a web service hosted by the service bus which ‘consumes’ the DSB service. As a result, a SOAP message sent to the DSB web service will be forwarded to the DSB endpoint which will forward it again to the final web service. Allright, but this is just for external SOAP clients and it is impossible to invoke anything firectly from the service bus kernel.
Now, let’s call the DSB endpoint from Java code… To put ourselves in a **real** case, let say that you are DSB kernel developer and that you want to say hello to the world (oh s*** really?) by using the *HelloWorldService* which is bound to the bus. Let’s code this client:

So, just get a *org.petalslink.dsb.service.client.Client* from the *org.petalslink.dsb.service.client.ClientFactory* factory, create your *org.petalslink.dsb.service.client.Message* and send it by using the *sendReceive* method. That’s all… Let’s try to clarify what it means with a simple schema:

Hacker inside!

It looks so simple… I just deeply hacked the petals service bus kernel to provide such a thing. Why? Because the service bus is also deeply based on the JBI specification: Creating a client means that the client need to be able to receive responses and acts itself as a service on the service bus point of view. It is also built using all the component context stuff. No component context means no way to send/receive messages (component contexts are just created when JBI components are installed).
So it is really important to release the client when you do not use it anymore: There are tons of listeners and threads in the different layers which are just used to receive messages. If you do not release the client, there will be unused resources for a long time.