Handling ‘state’ in Java WebSocket applications

By and large, there are two kinds of states in a WebSocket application

  • User/client specific: related to a connected user/Session e.g. user ID, list of subscriptions, last message received etc.
  • Global: state which is relevant across your application and something which all connected users/Sessions might be able to use

User specific state

This can be handled using getUserProperties method on the Session object – this exposes a Map which you can use to store anything (Object type) using a String type key

Global state

There are multiple options here as well. Please note that these are scoped to a specific Endpoint

  • getUserProperties in EndpointConfig – it exposes the same Map interface as the one in Session. Since the WebSocket runtime creates a single instance of an EndpointConfig object per Endpoint , it can be used a global state store

  • Another option is to encapsulate some of the common/global logic in a custom Configurator implementation which can be accessed & used within the endpoint logic

Further reading


Posted in Java, Java EE | Tagged , , | 2 Comments

Using CDI with Java EE Concurrency Utilities

This blog post explores usage of CDI along with Java EE Concurrency Utilities – specifically using CDI beans as managed tasks. Here is the sample application on Github

Lets begin with a quick overview

Java EE Concurrency Utilities provides APIs and constructs to manage concurrency within Java EE applications. Many of the Java EE components have specific concurrency semantics e.g. EJBs, JAX-RS resources, WebSocket endpoints etc. Writing components with custom concurrency properties was traditionally difficult since starting unmanaged threads in a Java EE container was forbidden i.e. one was not able to leverage Java SE concurrency libraries. With Concurrency Utilities, Java EE applications have access to Managed versions of the Java SE counterparts, namely,

  • ManagedExecutorSevice
  • ManagedScheduledExecutorSevice
  • ManagedThreadFactory

.. and a bunch of other APIs as well, but the above ones are the Java EE equivalent of the Java SE concurrency APIs

Tasks as CDI beans

Both ManagedExecutorSevice and ManagedScheduledExecutorService can accept tasks to execute (in a container managed thread pool) in the form of Runnable and Callable instances. The good thing is that these tasks can be CDI beans as well. Points worth noting are

  • these CDI beans can be injected into other components as well as inject other beans
  • the scope of the CDI beans which can be used as tasks as restricted to @ApplicationScoped and @Dependent (for details read section of the specification)

Here is a summary of ..

.. what’s going on in the application. For more details, refer to the the README and explore the code

  • Tasks are POSTED via a REST interface and the client gets back a HTTP 202 (Accepted) in response along with a task id
  • the BackgroundTask (CDI bean) is executed in a background thread by the ManagedExecutorService  – it is injected (@Inject) and a different instance is created on every invocation since the CDI bean is marked @Dependent and the JAX-RS resource is created on each request by the client
  • the status is store in a @Singleton EJB (TaskStore) – this is injected in the BackgroundTask CDI bean
  • status of each task (in progress, failed, completed) can be tracked via a REST interface by querying against a task ID
  • one can also get the status of all tasks

Further reading


Posted in Java, Java EE | Tagged , , , , , , | 4 Comments

Trending Meetup groups with Redis and Java EE

This is an application which displays the currently trending Meetup groups based on their (live) RSVPs feed. It’s built using Java EE 7 (uses WebSocket client & server APIs, Singleton EJB timers and CDI events to wire things up) and Redis

To run

  • get the code
  • start Redis
  • change Redis connection details here and here
  • mvn clean install
  • deploy the WAR file in any Java EE 7 (or above) compliant container
  • access the application in your browser e.g. http://localhost:8080/meetup-trending/

You should see something like this – Score represents the frequency of the group occurence in the RSVPs (popularity)




Posted in Java, Java EE | Tagged , , , , , , , | Leave a comment

Github: WebSocket applications…

I recently pushed a couple of WebSocket samples to Github. Both are based around the notion of a chat service – a canonical WebSocket example! These were originally meant to serve as additional material to the Java WebSocket API Handbook in order for readers to understand API usage in practical and also get hands-on  by tinkering with the code


Posted in Java EE | Tagged , , , | Leave a comment

Quick tip: managing Stateful EJBs in WebSocket endpoints

@Stateful EJBs can be injected in WebSocket endpoints (supported by the WebSocket specification). There is an one-to-one association between the WebSocket client & endpoint (which is by default) as well as the injected Stateful EJB instance, which makes it an good candidate for storing client specific state. It offers advanced semantics as compared to simple java.util.Map interface exposed by getUserProperties method in javax.websocket.Session)

But, what happens to the EJB when ….

… the WebSocket session terminates (either from the client or server side) ?? The Stateful instance can still linger in memory before its destroyed

  • Their removal can be tuned by using the @StatefulTimeout annotation (but why wait ??), or
  • you can choose to passivate them by using passivationCapable=true with  @Stateful (but why passivate the EJB when the WebSocket connection itself is closed ?)

The Tip

Implement a @Remove annotated method in the Stateful EJB and call it from the @OnClose callback method ni your WebSocket server endpoint implementation. This will ensure that the  EJB is removed from the memory immediately rather than depending upon other factors

Further reading

Posted in Java, Java EE | Tagged , , , , , , | Leave a comment

WebSocket endpoint as Singleton EJB

By default …

… a WebSocket implementation creates a new (server) endpoint instance per client. In case you need a single instance, you can implement this using a custom ServerEndpointConfig.Configurator (by overriding the getEndpointInstance method)

The catch: you might have to sacrifice some of the (Java EE) platform related services like dependency injection and interceptors

More details here

Alternate solution ?

A similar behavior can be achieved by decorating the WebSocket endpoint with @Singleton

This approach has the caveat of not being a standard feature outlined by the spec (although injection of EJBs as well as interceptors support is clearly mentioned in the Java WebSocket spec Sec 7.1)

Concurrency semantics ?

In case of a @Singleton, all the clients will interact with the one-and-only server endpoint instance. Here is a quick summary of how the EJB as well as WebSocket threading semantics are applied

  • The Singleton bean default approach WRITE lock ensures single threaded access across all connected clients
  • If thread-safety is not a concern (e.g. in case where you do not deal with client specific data/state in your logic) and in case the single-threaded access model proves to be a bottleneck, override the default behavior by switching to a READ lock which allows concurrent threads to access the methods (unless of course a WRITE lock is not already in effect)

Note: The above mentioned semantics are with respect to ALL the WebSocket clients. From the point of view of a single client, the default strategy of one thread at a time, per endpoint instance per client continues to apply (more details here)

Further reading..


Posted in Java, Java EE | Tagged , , , , | 3 Comments

New release: Java WebSocket API handbook

Happy to announce the initial release of the Java WebSocket API handbook. This is still a work in progress but moving at a strady pace. You can read this on Gitbook or grab your PDF/ePub/Mobi version from Leanpub



Here is a quickie about the book

The book is what it says it is – a handbook, a quick reference, a fast track guide. It covers the nitty gritty of the Java WebSocket API which is a standard (specification) for building WebSocket applications. It is targeted (primarily) towards Java/Java EE developers and can be used in various capacities

  • As a getting started with the Java WebSocket API guide
  • You are already well versed with this API and its constructs, but need a quick peek/reference to a specific API, it’s usage, nuances etc.
  • Maybe you’re just curious about what Java has to offer in terms of WebSocket support – feel free to check it out

Current status

The majority of the big ticket chapters are complete





Posted in Java, Java EE | Tagged , , , , , | Leave a comment

Microservices messaging on Oracle Cloud using Apache Kafka

Here is a blog I posted on the Oracle Cloud Developer Solutions portal. This is the first of a two-part series which shows asynchronous messaging b/w microservices with the help of a simple example (application)



Technical components

Oracle Cloud

Open source

  • Apache Kafka: scalable pub-sub message hub
  • Jersey: Used to implement REST and SSE services. Uses Grizzly as a (pluggable) runtime/container



Posted in Cloud, Java, Oracle Application Container Cloud, Oracle Cloud, Oracle PaaS | Tagged , , , , , , , , | Leave a comment

(eBook) ‘REST assured with JAX-RS’ now available on Gitbook




You can now read REST assured with JAX-RS online on Gitbook, or grab the PDF from Leanpub.. Happy reading!

Posted in Java, Java EE | Tagged , , , , | Leave a comment

Using Asynchronous timeouts in the Java WebSocket API

Sending messages in an asynchronous manner avoid blocking the sending thread. This is a great where your solution needs to scale in order to support a large number of clients.

But there is a limit on how long can we wait for the asynchronous process to complete

The Java WebSocket API gives you a few options in this regard

Async Timeout support

  • first and foremost, there is a notion of a timeout and this can be configured using the setSendTimeout method in the RemoteEndpoint.Async interface
  • secondly, the failure result manifests itself using the Future object or java.websocket.SendResult

How do timeouts manifest ?

It depends on which strategy you’re using in order to send your messages

  • Callback based
  • Future (java.util.concurrent) based

In case you are using the java.websocket.SendHandler i.e. the callback handler route, the timeout exception details will be available via SendResult.getException()

If you chose to use the Future to track the completion, calling it get method will result in a java.util.concurrent.ExecutionException

Further reading


Posted in Java, Java EE | Tagged , , , , , | Leave a comment