Spring Boot Websocket STOMP Example

How to create a Spring/Stomp server with a Java client

Posted on 08 January 2017

In our current project we want to add a service that uses websockets to push messages to our mobile applications. While the documentation on Spring Websockets + STOMP is excellent when it comes to implementing a service that is consumed by a simple web application, the example on how to use the STOMP client doesn’t really align very well with the short getting started guide. Since I hit a few snags in the implementation I’m creating this example so you don’t have to.

This blog post is accompanied with an example repository.

Introduction

For the spike I’m working on now we want to have our mobile client communicate with our back end via websockets. On top of that we want to use the STOMP messaging protocol / broker since we are likely going to need a pub/sub mechanism anyway. The main difference from the Getting Started is that we don’t need the SockJS compatibility layer since we control both the server and the client.

The service

Relative to the Getting Started there are a few key changes. First of all I prefer to use Lombok in my projects. You’ll notice that the POJO’s have generated getters and setters and the @Slf4j annotation creates a Logger for us.

Secondly I have removed the SockJS layer from the service. This is a really minor change in the endpoint configuration. Instead of this:

    @Override
    public void registerStompEndpoints(StompEndpointRegistry registry) {
        registry.addEndpoint("/hello").withSockJS();
    }

You set it up like this:

    @Override
    public void registerStompEndpoints(StompEndpointRegistry registry) {
        registry.addEndpoint("/hello");
    }

I also added a TimeSender component that, unsurprisingly, broadcasts time updates every 5 seconds. It does this broadcast on the same '/topic/greeting' topic as the GreetingController responds on to keep the client simple. It’s a Spring @Scheduled task so don’t forget to enable task scheduling with @EnableScheduling in your configuration!

The rest of the service is more or less the same as the one in the Getting Started: Application is the main entry point, WebSocketConfig contains the configuration and GreetingController handles the receiving of the client’s name and greeting them.

If you build the project using maven (mvn clean install) you can run the service standalone using java -jar target/spring-boot-websocket-1.0.jar. You can also run it from your IDE by simply running Application.

The JavaScript client

The only real change made on the JS client side was removing the SockJS compatibility layer. I removed the script tag loading the sockjs.js source and changed the bit where you create the client from this:

    var socket = new SockJS('/gs-guide-websocket');
    stompClient = Stomp.over(socket);

To this:

    stompClient = Stomp.client('ws://localhost:8080/hello');

The Java client

The Java STOMP client was a bit less straightforward. It had me stumped (stomped?) for a while. Although I could connect to the service and send messages it would not allow me to receive any responses. Not just that; I wasn’t getting any errors either. Why? Well what the documentation unfortunately does not make very clear is that the StompSessionHandlerAdapter you tend to extend has an empty implementation of handleException(). So whenever something goes wrong it just swallows the exception. That only took me like 2 hours to figure out :D. So the first order of business was to add 'proper' exception handling to MySessionHandler:

    @Override
    public void handleException( ... ) {
        exception.printStackTrace();
    }

In my opinion this should be the default behavior. Note: you can’t simply re-throw the exception here; it will disappear.

So now that we are actually seeing exceptions I found out I was doing something wrong:

    org.springframework.messaging.converter.MessageConversionException:
			No suitable converter, payloadType=class ... .Greeting,
			handlerType=class ... .MySessionHandler
	at ... .DefaultStompSession.invokeHandler(DefaultStompSession.java:443)

I went through different permutations of removing the StringMessageConverter, returning the Type of String inside session handler .getPayloadType(), etc. The problem was both a lack of understanding on my part and (again) unclear documentation. First of all; the STOMP 'frames' you receive will have a mime type. It’s not really easy to spot (I found out by dumping all the headers to std out) but the service implementation, because we are returning Greeting objects from the controller, is using Jackson to serialize these objects to JSON. This will also result in those frames getting the application/json mimetype. The StringMessageConverter will only work on text/plain messages.

The second part that was confusing was the documentation. It claims that if Jackson is on the classpath it will use a Jackson message convertor. I don’t know how it tries to figure this out but it didn’t happen automatically, this is why I needed to add it myself:

    stompClient.setMessageConverter(new MappingJackson2MessageConverter());

That got my basic Java client working: the ServiceClient configures the connection and adds a handler, MySessionHandler implements a handler that sends a message to the WS service and then subscribes to the 'topic/greeting' topic. It now will receive the same messages as the JS client does.

Conclusion

Although I do think the Spring documentation has some gaps here it was a fun journey of discovery! I hope this example is useful and save you some time. If you have any questions or feedback feel free to raise an issue in this repository!